1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006 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. Most of them are also available
20 in C++. @xref{C++ Extensions,,Extensions to the C++ Language}, for
21 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 * Decimal Float:: Decimal Floating Types.
37 * Hex Floats:: Hexadecimal floating-point constants.
38 * Zero Length:: Zero-length arrays.
39 * Variable Length:: Arrays whose length is computed at run time.
40 * Empty Structures:: Structures with no members.
41 * Variadic Macros:: Macros with a variable number of arguments.
42 * Escaped Newlines:: Slightly looser rules for escaped newlines.
43 * Subscripting:: Any array can be subscripted, even if not an lvalue.
44 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
45 * Initializers:: Non-constant initializers.
46 * Compound Literals:: Compound literals give structures, unions
48 * Designated Inits:: Labeling elements of initializers.
49 * Cast to Union:: Casting to union type from any member of the union.
50 * Case Ranges:: `case 1 ... 9' and such.
51 * Mixed Declarations:: Mixing declarations and code.
52 * Function Attributes:: Declaring that functions have no side effects,
53 or that they can never return.
54 * Attribute Syntax:: Formal syntax for attributes.
55 * Function Prototypes:: Prototype declarations and old-style definitions.
56 * C++ Comments:: C++ comments are recognized.
57 * Dollar Signs:: Dollar sign is allowed in identifiers.
58 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
62 * Label Attributes:: Specifying attributes of labels and statements.
63 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
64 * Alignment:: Inquiring about the alignment of a type or variable.
65 * Inline:: Defining inline functions (as fast as macros).
66 * Extended Asm:: Assembler instructions with C expressions as operands.
67 (With them you can define ``built-in'' functions.)
68 * Constraints:: Constraints for asm operands
69 * Asm Labels:: Specifying the assembler name to use for a C symbol.
70 * Explicit Reg Vars:: Defining variables residing in specified registers.
71 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
72 * Incomplete Enums:: @code{enum foo;}, with details to follow.
73 * Function Names:: Printable strings which are the name of the current
75 * Return Address:: Getting the return or frame address of a function.
76 * Vector Extensions:: Using vector instructions through built-in functions.
77 * Offsetof:: Special syntax for implementing @code{offsetof}.
78 * Atomic Builtins:: Built-in functions for atomic memory access.
79 * Object Size Checking:: Built-in functions for limited buffer overflow
81 * Other Builtins:: Other built-in functions.
82 * Target Builtins:: Built-in functions specific to particular targets.
83 * Target Format Checks:: Format checks specific to particular targets.
84 * Pragmas:: Pragmas accepted by GCC.
85 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
86 * Thread-Local:: Per-thread variables.
87 * Binary constants:: Binary constants using the @samp{0b} prefix.
88 @c APPLE LOCAL blocks 7205047 5811887
89 * Blocks:: Anonymous functions (closures).
93 @section Statements and Declarations in Expressions
94 @cindex statements inside expressions
95 @cindex declarations inside expressions
96 @cindex expressions containing statements
97 @cindex macros, statements in expressions
99 @c the above section title wrapped and causes an underfull hbox.. i
100 @c changed it from "within" to "in". --mew 4feb93
101 A compound statement enclosed in parentheses may appear as an expression
102 in GNU C@. This allows you to use loops, switches, and local variables
103 within an expression.
105 Recall that a compound statement is a sequence of statements surrounded
106 by braces; in this construct, parentheses go around the braces. For
110 (@{ int y = foo (); int z;
117 is a valid (though slightly more complex than necessary) expression
118 for the absolute value of @code{foo ()}.
120 The last thing in the compound statement should be an expression
121 followed by a semicolon; the value of this subexpression serves as the
122 value of the entire construct. (If you use some other kind of statement
123 last within the braces, the construct has type @code{void}, and thus
124 effectively no value.)
126 This feature is especially useful in making macro definitions ``safe'' (so
127 that they evaluate each operand exactly once). For example, the
128 ``maximum'' function is commonly defined as a macro in standard C as
132 #define max(a,b) ((a) > (b) ? (a) : (b))
136 @cindex side effects, macro argument
137 But this definition computes either @var{a} or @var{b} twice, with bad
138 results if the operand has side effects. In GNU C, if you know the
139 type of the operands (here taken as @code{int}), you can define
140 the macro safely as follows:
143 #define maxint(a,b) \
144 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
147 Embedded statements are not allowed in constant expressions, such as
148 the value of an enumeration constant, the width of a bit-field, or
149 the initial value of a static variable.
151 If you don't know the type of the operand, you can still do this, but you
152 must use @code{typeof} (@pxref{Typeof}).
154 In G++, the result value of a statement expression undergoes array and
155 function pointer decay, and is returned by value to the enclosing
156 expression. For instance, if @code{A} is a class, then
165 will construct a temporary @code{A} object to hold the result of the
166 statement expression, and that will be used to invoke @code{Foo}.
167 Therefore the @code{this} pointer observed by @code{Foo} will not be the
170 Any temporaries created within a statement within a statement expression
171 will be destroyed at the statement's end. This makes statement
172 expressions inside macros slightly different from function calls. In
173 the latter case temporaries introduced during argument evaluation will
174 be destroyed at the end of the statement that includes the function
175 call. In the statement expression case they will be destroyed during
176 the statement expression. For instance,
179 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
180 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
190 will have different places where temporaries are destroyed. For the
191 @code{macro} case, the temporary @code{X} will be destroyed just after
192 the initialization of @code{b}. In the @code{function} case that
193 temporary will be destroyed when the function returns.
195 These considerations mean that it is probably a bad idea to use
196 statement-expressions of this form in header files that are designed to
197 work with C++. (Note that some versions of the GNU C Library contained
198 header files using statement-expression that lead to precisely this
201 Jumping into a statement expression with @code{goto} or using a
202 @code{switch} statement outside the statement expression with a
203 @code{case} or @code{default} label inside the statement expression is
204 not permitted. Jumping into a statement expression with a computed
205 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
206 Jumping out of a statement expression is permitted, but if the
207 statement expression is part of a larger expression then it is
208 unspecified which other subexpressions of that expression have been
209 evaluated except where the language definition requires certain
210 subexpressions to be evaluated before or after the statement
211 expression. In any case, as with a function call the evaluation of a
212 statement expression is not interleaved with the evaluation of other
213 parts of the containing expression. For example,
216 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
220 will call @code{foo} and @code{bar1} and will not call @code{baz} but
221 may or may not call @code{bar2}. If @code{bar2} is called, it will be
222 called after @code{foo} and before @code{bar1}
225 @section Locally Declared Labels
227 @cindex macros, local labels
229 GCC allows you to declare @dfn{local labels} in any nested block
230 scope. A local label is just like an ordinary label, but you can
231 only reference it (with a @code{goto} statement, or by taking its
232 address) within the block in which it was declared.
234 A local label declaration looks like this:
237 __label__ @var{label};
244 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
247 Local label declarations must come at the beginning of the block,
248 before any ordinary declarations or statements.
250 The label declaration defines the label @emph{name}, but does not define
251 the label itself. You must do this in the usual way, with
252 @code{@var{label}:}, within the statements of the statement expression.
254 The local label feature is useful for complex macros. If a macro
255 contains nested loops, a @code{goto} can be useful for breaking out of
256 them. However, an ordinary label whose scope is the whole function
257 cannot be used: if the macro can be expanded several times in one
258 function, the label will be multiply defined in that function. A
259 local label avoids this problem. For example:
262 #define SEARCH(value, array, target) \
265 typeof (target) _SEARCH_target = (target); \
266 typeof (*(array)) *_SEARCH_array = (array); \
269 for (i = 0; i < max; i++) \
270 for (j = 0; j < max; j++) \
271 if (_SEARCH_array[i][j] == _SEARCH_target) \
272 @{ (value) = i; goto found; @} \
278 This could also be written using a statement-expression:
281 #define SEARCH(array, target) \
284 typeof (target) _SEARCH_target = (target); \
285 typeof (*(array)) *_SEARCH_array = (array); \
288 for (i = 0; i < max; i++) \
289 for (j = 0; j < max; j++) \
290 if (_SEARCH_array[i][j] == _SEARCH_target) \
291 @{ value = i; goto found; @} \
298 Local label declarations also make the labels they declare visible to
299 nested functions, if there are any. @xref{Nested Functions}, for details.
301 @node Labels as Values
302 @section Labels as Values
303 @cindex labels as values
304 @cindex computed gotos
305 @cindex goto with computed label
306 @cindex address of a label
308 You can get the address of a label defined in the current function
309 (or a containing function) with the unary operator @samp{&&}. The
310 value has type @code{void *}. This value is a constant and can be used
311 wherever a constant of that type is valid. For example:
319 To use these values, you need to be able to jump to one. This is done
320 with the computed goto statement@footnote{The analogous feature in
321 Fortran is called an assigned goto, but that name seems inappropriate in
322 C, where one can do more than simply store label addresses in label
323 variables.}, @code{goto *@var{exp};}. For example,
330 Any expression of type @code{void *} is allowed.
332 One way of using these constants is in initializing a static array that
333 will serve as a jump table:
336 static void *array[] = @{ &&foo, &&bar, &&hack @};
339 Then you can select a label with indexing, like this:
346 Note that this does not check whether the subscript is in bounds---array
347 indexing in C never does that.
349 Such an array of label values serves a purpose much like that of the
350 @code{switch} statement. The @code{switch} statement is cleaner, so
351 use that rather than an array unless the problem does not fit a
352 @code{switch} statement very well.
354 Another use of label values is in an interpreter for threaded code.
355 The labels within the interpreter function can be stored in the
356 threaded code for super-fast dispatching.
358 You may not use this mechanism to jump to code in a different function.
359 If you do that, totally unpredictable things will happen. The best way to
360 avoid this is to store the label address only in automatic variables and
361 never pass it as an argument.
363 An alternate way to write the above example is
366 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
368 goto *(&&foo + array[i]);
372 This is more friendly to code living in shared libraries, as it reduces
373 the number of dynamic relocations that are needed, and by consequence,
374 allows the data to be read-only.
376 @node Nested Functions
377 @section Nested Functions
378 @cindex nested functions
379 @cindex downward funargs
382 A @dfn{nested function} is a function defined inside another function.
383 @c APPLE LOCAL begin nested functions 4357979
384 Nested functions are not supported for GNU C++ and are disable by
385 default on FreeBSD. The @option{-fnested-functions} and
386 @option{-fno-nested-functions} options can be used to enable and
387 disable nested function suppport. The nested function's name is local
388 to the block where it is defined. For example, here we define a
389 nested function named @code{square}, and call it twice:
390 @c APPLE LOCAL end nested functions 4357979
394 foo (double a, double b)
396 double square (double z) @{ return z * z; @}
398 return square (a) + square (b);
403 The nested function can access all the variables of the containing
404 function that are visible at the point of its definition. This is
405 called @dfn{lexical scoping}. For example, here we show a nested
406 function which uses an inherited variable named @code{offset}:
410 bar (int *array, int offset, int size)
412 int access (int *array, int index)
413 @{ return array[index + offset]; @}
416 for (i = 0; i < size; i++)
417 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
422 Nested function definitions are permitted within functions in the places
423 where variable definitions are allowed; that is, in any block, mixed
424 with the other declarations and statements in the block.
426 It is possible to call the nested function from outside the scope of its
427 name by storing its address or passing the address to another function:
430 hack (int *array, int size)
432 void store (int index, int value)
433 @{ array[index] = value; @}
435 intermediate (store, size);
439 Here, the function @code{intermediate} receives the address of
440 @code{store} as an argument. If @code{intermediate} calls @code{store},
441 the arguments given to @code{store} are used to store into @code{array}.
442 But this technique works only so long as the containing function
443 (@code{hack}, in this example) does not exit.
445 If you try to call the nested function through its address after the
446 containing function has exited, all hell will break loose. If you try
447 to call it after a containing scope level has exited, and if it refers
448 to some of the variables that are no longer in scope, you may be lucky,
449 but it's not wise to take the risk. If, however, the nested function
450 does not refer to anything that has gone out of scope, you should be
453 GCC implements taking the address of a nested function using a technique
454 called @dfn{trampolines}. A paper describing them is available as
457 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
459 A nested function can jump to a label inherited from a containing
460 function, provided the label was explicitly declared in the containing
461 function (@pxref{Local Labels}). Such a jump returns instantly to the
462 containing function, exiting the nested function which did the
463 @code{goto} and any intermediate functions as well. Here is an example:
467 bar (int *array, int offset, int size)
470 int access (int *array, int index)
474 return array[index + offset];
478 for (i = 0; i < size; i++)
479 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
483 /* @r{Control comes here from @code{access}
484 if it detects an error.} */
491 A nested function always has no linkage. Declaring one with
492 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
493 before its definition, use @code{auto} (which is otherwise meaningless
494 for function declarations).
497 bar (int *array, int offset, int size)
500 auto int access (int *, int);
502 int access (int *array, int index)
506 return array[index + offset];
512 @node Constructing Calls
513 @section Constructing Function Calls
514 @cindex constructing calls
515 @cindex forwarding calls
517 Using the built-in functions described below, you can record
518 the arguments a function received, and call another function
519 with the same arguments, without knowing the number or types
522 You can also record the return value of that function call,
523 and later return that value, without knowing what data type
524 the function tried to return (as long as your caller expects
527 However, these built-in functions may interact badly with some
528 sophisticated features or other extensions of the language. It
529 is, therefore, not recommended to use them outside very simple
530 functions acting as mere forwarders for their arguments.
532 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
533 This built-in function returns a pointer to data
534 describing how to perform a call with the same arguments as were passed
535 to the current function.
537 The function saves the arg pointer register, structure value address,
538 and all registers that might be used to pass arguments to a function
539 into a block of memory allocated on the stack. Then it returns the
540 address of that block.
543 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
544 This built-in function invokes @var{function}
545 with a copy of the parameters described by @var{arguments}
548 The value of @var{arguments} should be the value returned by
549 @code{__builtin_apply_args}. The argument @var{size} specifies the size
550 of the stack argument data, in bytes.
552 This function returns a pointer to data describing
553 how to return whatever value was returned by @var{function}. The data
554 is saved in a block of memory allocated on the stack.
556 It is not always simple to compute the proper value for @var{size}. The
557 value is used by @code{__builtin_apply} to compute the amount of data
558 that should be pushed on the stack and copied from the incoming argument
562 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
563 This built-in function returns the value described by @var{result} from
564 the containing function. You should specify, for @var{result}, a value
565 returned by @code{__builtin_apply}.
569 @section Referring to a Type with @code{typeof}
572 @cindex macros, types of arguments
574 Another way to refer to the type of an expression is with @code{typeof}.
575 The syntax of using of this keyword looks like @code{sizeof}, but the
576 construct acts semantically like a type name defined with @code{typedef}.
578 There are two ways of writing the argument to @code{typeof}: with an
579 expression or with a type. Here is an example with an expression:
586 This assumes that @code{x} is an array of pointers to functions;
587 the type described is that of the values of the functions.
589 Here is an example with a typename as the argument:
596 Here the type described is that of pointers to @code{int}.
598 If you are writing a header file that must work when included in ISO C
599 programs, write @code{__typeof__} instead of @code{typeof}.
600 @xref{Alternate Keywords}.
602 A @code{typeof}-construct can be used anywhere a typedef name could be
603 used. For example, you can use it in a declaration, in a cast, or inside
604 of @code{sizeof} or @code{typeof}.
606 @code{typeof} is often useful in conjunction with the
607 statements-within-expressions feature. Here is how the two together can
608 be used to define a safe ``maximum'' macro that operates on any
609 arithmetic type and evaluates each of its arguments exactly once:
613 (@{ typeof (a) _a = (a); \
614 typeof (b) _b = (b); \
615 _a > _b ? _a : _b; @})
618 @cindex underscores in variables in macros
619 @cindex @samp{_} in variables in macros
620 @cindex local variables in macros
621 @cindex variables, local, in macros
622 @cindex macros, local variables in
624 The reason for using names that start with underscores for the local
625 variables is to avoid conflicts with variable names that occur within the
626 expressions that are substituted for @code{a} and @code{b}. Eventually we
627 hope to design a new form of declaration syntax that allows you to declare
628 variables whose scopes start only after their initializers; this will be a
629 more reliable way to prevent such conflicts.
632 Some more examples of the use of @code{typeof}:
636 This declares @code{y} with the type of what @code{x} points to.
643 This declares @code{y} as an array of such values.
650 This declares @code{y} as an array of pointers to characters:
653 typeof (typeof (char *)[4]) y;
657 It is equivalent to the following traditional C declaration:
663 To see the meaning of the declaration using @code{typeof}, and why it
664 might be a useful way to write, rewrite it with these macros:
667 #define pointer(T) typeof(T *)
668 #define array(T, N) typeof(T [N])
672 Now the declaration can be rewritten this way:
675 array (pointer (char), 4) y;
679 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
680 pointers to @code{char}.
683 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
684 a more limited extension which permitted one to write
687 typedef @var{T} = @var{expr};
691 with the effect of declaring @var{T} to have the type of the expression
692 @var{expr}. This extension does not work with GCC 3 (versions between
693 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
694 relies on it should be rewritten to use @code{typeof}:
697 typedef typeof(@var{expr}) @var{T};
701 This will work with all versions of GCC@.
704 @section Conditionals with Omitted Operands
705 @cindex conditional expressions, extensions
706 @cindex omitted middle-operands
707 @cindex middle-operands, omitted
708 @cindex extensions, @code{?:}
709 @cindex @code{?:} extensions
711 The middle operand in a conditional expression may be omitted. Then
712 if the first operand is nonzero, its value is the value of the conditional
715 Therefore, the expression
722 has the value of @code{x} if that is nonzero; otherwise, the value of
725 This example is perfectly equivalent to
731 @cindex side effect in ?:
732 @cindex ?: side effect
734 In this simple case, the ability to omit the middle operand is not
735 especially useful. When it becomes useful is when the first operand does,
736 or may (if it is a macro argument), contain a side effect. Then repeating
737 the operand in the middle would perform the side effect twice. Omitting
738 the middle operand uses the value already computed without the undesirable
739 effects of recomputing it.
742 @section Double-Word Integers
743 @cindex @code{long long} data types
744 @cindex double-word arithmetic
745 @cindex multiprecision arithmetic
746 @cindex @code{LL} integer suffix
747 @cindex @code{ULL} integer suffix
749 ISO C99 supports data types for integers that are at least 64 bits wide,
750 and as an extension GCC supports them in C89 mode and in C++.
751 Simply write @code{long long int} for a signed integer, or
752 @code{unsigned long long int} for an unsigned integer. To make an
753 integer constant of type @code{long long int}, add the suffix @samp{LL}
754 to the integer. To make an integer constant of type @code{unsigned long
755 long int}, add the suffix @samp{ULL} to the integer.
757 You can use these types in arithmetic like any other integer types.
758 Addition, subtraction, and bitwise boolean operations on these types
759 are open-coded on all types of machines. Multiplication is open-coded
760 if the machine supports fullword-to-doubleword a widening multiply
761 instruction. Division and shifts are open-coded only on machines that
762 provide special support. The operations that are not open-coded use
763 special library routines that come with GCC@.
765 There may be pitfalls when you use @code{long long} types for function
766 arguments, unless you declare function prototypes. If a function
767 expects type @code{int} for its argument, and you pass a value of type
768 @code{long long int}, confusion will result because the caller and the
769 subroutine will disagree about the number of bytes for the argument.
770 Likewise, if the function expects @code{long long int} and you pass
771 @code{int}. The best way to avoid such problems is to use prototypes.
774 @section Complex Numbers
775 @cindex complex numbers
776 @cindex @code{_Complex} keyword
777 @cindex @code{__complex__} keyword
779 ISO C99 supports complex floating data types, and as an extension GCC
780 supports them in C89 mode and in C++, and supports complex integer data
781 types which are not part of ISO C99. You can declare complex types
782 using the keyword @code{_Complex}. As an extension, the older GNU
783 keyword @code{__complex__} is also supported.
785 For example, @samp{_Complex double x;} declares @code{x} as a
786 variable whose real part and imaginary part are both of type
787 @code{double}. @samp{_Complex short int y;} declares @code{y} to
788 have real and imaginary parts of type @code{short int}; this is not
789 likely to be useful, but it shows that the set of complex types is
792 To write a constant with a complex data type, use the suffix @samp{i} or
793 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
794 has type @code{_Complex float} and @code{3i} has type
795 @code{_Complex int}. Such a constant always has a pure imaginary
796 value, but you can form any complex value you like by adding one to a
797 real constant. This is a GNU extension; if you have an ISO C99
798 conforming C library (such as GNU libc), and want to construct complex
799 constants of floating type, you should include @code{<complex.h>} and
800 use the macros @code{I} or @code{_Complex_I} instead.
802 @cindex @code{__real__} keyword
803 @cindex @code{__imag__} keyword
804 To extract the real part of a complex-valued expression @var{exp}, write
805 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
806 extract the imaginary part. This is a GNU extension; for values of
807 floating type, you should use the ISO C99 functions @code{crealf},
808 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
809 @code{cimagl}, declared in @code{<complex.h>} and also provided as
810 built-in functions by GCC@.
812 @cindex complex conjugation
813 The operator @samp{~} performs complex conjugation when used on a value
814 with a complex type. This is a GNU extension; for values of
815 floating type, you should use the ISO C99 functions @code{conjf},
816 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
817 provided as built-in functions by GCC@.
819 GCC can allocate complex automatic variables in a noncontiguous
820 fashion; it's even possible for the real part to be in a register while
821 the imaginary part is on the stack (or vice-versa). Only the DWARF2
822 debug info format can represent this, so use of DWARF2 is recommended.
823 If you are using the stabs debug info format, GCC describes a noncontiguous
824 complex variable as if it were two separate variables of noncomplex type.
825 If the variable's actual name is @code{foo}, the two fictitious
826 variables are named @code{foo$real} and @code{foo$imag}. You can
827 examine and set these two fictitious variables with your debugger.
830 @section Decimal Floating Types
831 @cindex decimal floating types
832 @cindex @code{_Decimal32} data type
833 @cindex @code{_Decimal64} data type
834 @cindex @code{_Decimal128} data type
835 @cindex @code{df} integer suffix
836 @cindex @code{dd} integer suffix
837 @cindex @code{dl} integer suffix
838 @cindex @code{DF} integer suffix
839 @cindex @code{DD} integer suffix
840 @cindex @code{DL} integer suffix
842 As an extension, the GNU C compiler supports decimal floating types as
843 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
844 floating types in GCC will evolve as the draft technical report changes.
845 Calling conventions for any target might also change. Not all targets
846 support decimal floating types.
848 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
849 @code{_Decimal128}. They use a radix of ten, unlike the floating types
850 @code{float}, @code{double}, and @code{long double} whose radix is not
851 specified by the C standard but is usually two.
853 Support for decimal floating types includes the arithmetic operators
854 add, subtract, multiply, divide; unary arithmetic operators;
855 relational operators; equality operators; and conversions to and from
856 integer and other floating types. Use a suffix @samp{df} or
857 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
858 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
861 GCC support of decimal float as specified by the draft technical report
866 Translation time data type (TTDT) is not supported.
869 Characteristics of decimal floating types are defined in header file
870 @file{decfloat.h} rather than @file{float.h}.
873 When the value of a decimal floating type cannot be represented in the
874 integer type to which it is being converted, the result is undefined
875 rather than the result value specified by the draft technical report.
878 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
879 are supported by the DWARF2 debug information format.
885 ISO C99 supports floating-point numbers written not only in the usual
886 decimal notation, such as @code{1.55e1}, but also numbers such as
887 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
888 supports this in C89 mode (except in some cases when strictly
889 conforming) and in C++. In that format the
890 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
891 mandatory. The exponent is a decimal number that indicates the power of
892 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
899 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
900 is the same as @code{1.55e1}.
902 Unlike for floating-point numbers in the decimal notation the exponent
903 is always required in the hexadecimal notation. Otherwise the compiler
904 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
905 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
906 extension for floating-point constants of type @code{float}.
909 @section Arrays of Length Zero
910 @cindex arrays of length zero
911 @cindex zero-length arrays
912 @cindex length-zero arrays
913 @cindex flexible array members
915 Zero-length arrays are allowed in GNU C@. They are very useful as the
916 last element of a structure which is really a header for a variable-length
925 struct line *thisline = (struct line *)
926 malloc (sizeof (struct line) + this_length);
927 thisline->length = this_length;
930 In ISO C90, you would have to give @code{contents} a length of 1, which
931 means either you waste space or complicate the argument to @code{malloc}.
933 In ISO C99, you would use a @dfn{flexible array member}, which is
934 slightly different in syntax and semantics:
938 Flexible array members are written as @code{contents[]} without
942 Flexible array members have incomplete type, and so the @code{sizeof}
943 operator may not be applied. As a quirk of the original implementation
944 of zero-length arrays, @code{sizeof} evaluates to zero.
947 Flexible array members may only appear as the last member of a
948 @code{struct} that is otherwise non-empty.
951 A structure containing a flexible array member, or a union containing
952 such a structure (possibly recursively), may not be a member of a
953 structure or an element of an array. (However, these uses are
954 permitted by GCC as extensions.)
957 GCC versions before 3.0 allowed zero-length arrays to be statically
958 initialized, as if they were flexible arrays. In addition to those
959 cases that were useful, it also allowed initializations in situations
960 that would corrupt later data. Non-empty initialization of zero-length
961 arrays is now treated like any case where there are more initializer
962 elements than the array holds, in that a suitable warning about "excess
963 elements in array" is given, and the excess elements (all of them, in
964 this case) are ignored.
966 Instead GCC allows static initialization of flexible array members.
967 This is equivalent to defining a new structure containing the original
968 structure followed by an array of sufficient size to contain the data.
969 I.e.@: in the following, @code{f1} is constructed as if it were declared
975 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
978 struct f1 f1; int data[3];
979 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
983 The convenience of this extension is that @code{f1} has the desired
984 type, eliminating the need to consistently refer to @code{f2.f1}.
986 This has symmetry with normal static arrays, in that an array of
987 unknown size is also written with @code{[]}.
989 Of course, this extension only makes sense if the extra data comes at
990 the end of a top-level object, as otherwise we would be overwriting
991 data at subsequent offsets. To avoid undue complication and confusion
992 with initialization of deeply nested arrays, we simply disallow any
993 non-empty initialization except when the structure is the top-level
997 struct foo @{ int x; int y[]; @};
998 struct bar @{ struct foo z; @};
1000 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
1001 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1002 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
1003 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
1006 @node Empty Structures
1007 @section Structures With No Members
1008 @cindex empty structures
1009 @cindex zero-size structures
1011 GCC permits a C structure to have no members:
1018 The structure will have size zero. In C++, empty structures are part
1019 of the language. G++ treats empty structures as if they had a single
1020 member of type @code{char}.
1022 @node Variable Length
1023 @section Arrays of Variable Length
1024 @cindex variable-length arrays
1025 @cindex arrays of variable length
1028 Variable-length automatic arrays are allowed in ISO C99, and as an
1029 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1030 implementation of variable-length arrays does not yet conform in detail
1031 to the ISO C99 standard.) These arrays are
1032 declared like any other automatic arrays, but with a length that is not
1033 a constant expression. The storage is allocated at the point of
1034 declaration and deallocated when the brace-level is exited. For
1039 concat_fopen (char *s1, char *s2, char *mode)
1041 char str[strlen (s1) + strlen (s2) + 1];
1044 return fopen (str, mode);
1048 @cindex scope of a variable length array
1049 @cindex variable-length array scope
1050 @cindex deallocating variable length arrays
1051 Jumping or breaking out of the scope of the array name deallocates the
1052 storage. Jumping into the scope is not allowed; you get an error
1055 @cindex @code{alloca} vs variable-length arrays
1056 You can use the function @code{alloca} to get an effect much like
1057 variable-length arrays. The function @code{alloca} is available in
1058 many other C implementations (but not in all). On the other hand,
1059 variable-length arrays are more elegant.
1061 There are other differences between these two methods. Space allocated
1062 with @code{alloca} exists until the containing @emph{function} returns.
1063 The space for a variable-length array is deallocated as soon as the array
1064 name's scope ends. (If you use both variable-length arrays and
1065 @code{alloca} in the same function, deallocation of a variable-length array
1066 will also deallocate anything more recently allocated with @code{alloca}.)
1068 You can also use variable-length arrays as arguments to functions:
1072 tester (int len, char data[len][len])
1078 The length of an array is computed once when the storage is allocated
1079 and is remembered for the scope of the array in case you access it with
1082 If you want to pass the array first and the length afterward, you can
1083 use a forward declaration in the parameter list---another GNU extension.
1087 tester (int len; char data[len][len], int len)
1093 @cindex parameter forward declaration
1094 The @samp{int len} before the semicolon is a @dfn{parameter forward
1095 declaration}, and it serves the purpose of making the name @code{len}
1096 known when the declaration of @code{data} is parsed.
1098 You can write any number of such parameter forward declarations in the
1099 parameter list. They can be separated by commas or semicolons, but the
1100 last one must end with a semicolon, which is followed by the ``real''
1101 parameter declarations. Each forward declaration must match a ``real''
1102 declaration in parameter name and data type. ISO C99 does not support
1103 parameter forward declarations.
1105 @node Variadic Macros
1106 @section Macros with a Variable Number of Arguments.
1107 @cindex variable number of arguments
1108 @cindex macro with variable arguments
1109 @cindex rest argument (in macro)
1110 @cindex variadic macros
1112 In the ISO C standard of 1999, a macro can be declared to accept a
1113 variable number of arguments much as a function can. The syntax for
1114 defining the macro is similar to that of a function. Here is an
1118 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1121 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1122 such a macro, it represents the zero or more tokens until the closing
1123 parenthesis that ends the invocation, including any commas. This set of
1124 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1125 wherever it appears. See the CPP manual for more information.
1127 GCC has long supported variadic macros, and used a different syntax that
1128 allowed you to give a name to the variable arguments just like any other
1129 argument. Here is an example:
1132 #define debug(format, args...) fprintf (stderr, format, args)
1135 This is in all ways equivalent to the ISO C example above, but arguably
1136 more readable and descriptive.
1138 GNU CPP has two further variadic macro extensions, and permits them to
1139 be used with either of the above forms of macro definition.
1141 In standard C, you are not allowed to leave the variable argument out
1142 entirely; but you are allowed to pass an empty argument. For example,
1143 this invocation is invalid in ISO C, because there is no comma after
1150 GNU CPP permits you to completely omit the variable arguments in this
1151 way. In the above examples, the compiler would complain, though since
1152 the expansion of the macro still has the extra comma after the format
1155 To help solve this problem, CPP behaves specially for variable arguments
1156 used with the token paste operator, @samp{##}. If instead you write
1159 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1162 and if the variable arguments are omitted or empty, the @samp{##}
1163 operator causes the preprocessor to remove the comma before it. If you
1164 do provide some variable arguments in your macro invocation, GNU CPP
1165 does not complain about the paste operation and instead places the
1166 variable arguments after the comma. Just like any other pasted macro
1167 argument, these arguments are not macro expanded.
1169 @node Escaped Newlines
1170 @section Slightly Looser Rules for Escaped Newlines
1171 @cindex escaped newlines
1172 @cindex newlines (escaped)
1174 Recently, the preprocessor has relaxed its treatment of escaped
1175 newlines. Previously, the newline had to immediately follow a
1176 backslash. The current implementation allows whitespace in the form
1177 of spaces, horizontal and vertical tabs, and form feeds between the
1178 backslash and the subsequent newline. The preprocessor issues a
1179 warning, but treats it as a valid escaped newline and combines the two
1180 lines to form a single logical line. This works within comments and
1181 tokens, as well as between tokens. Comments are @emph{not} treated as
1182 whitespace for the purposes of this relaxation, since they have not
1183 yet been replaced with spaces.
1186 @section Non-Lvalue Arrays May Have Subscripts
1187 @cindex subscripting
1188 @cindex arrays, non-lvalue
1190 @cindex subscripting and function values
1191 In ISO C99, arrays that are not lvalues still decay to pointers, and
1192 may be subscripted, although they may not be modified or used after
1193 the next sequence point and the unary @samp{&} operator may not be
1194 applied to them. As an extension, GCC allows such arrays to be
1195 subscripted in C89 mode, though otherwise they do not decay to
1196 pointers outside C99 mode. For example,
1197 this is valid in GNU C though not valid in C89:
1201 struct foo @{int a[4];@};
1207 return f().a[index];
1213 @section Arithmetic on @code{void}- and Function-Pointers
1214 @cindex void pointers, arithmetic
1215 @cindex void, size of pointer to
1216 @cindex function pointers, arithmetic
1217 @cindex function, size of pointer to
1219 In GNU C, addition and subtraction operations are supported on pointers to
1220 @code{void} and on pointers to functions. This is done by treating the
1221 size of a @code{void} or of a function as 1.
1223 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1224 and on function types, and returns 1.
1226 @opindex Wpointer-arith
1227 The option @option{-Wpointer-arith} requests a warning if these extensions
1231 @section Non-Constant Initializers
1232 @cindex initializers, non-constant
1233 @cindex non-constant initializers
1235 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1236 automatic variable are not required to be constant expressions in GNU C@.
1237 Here is an example of an initializer with run-time varying elements:
1240 foo (float f, float g)
1242 float beat_freqs[2] = @{ f-g, f+g @};
1247 @node Compound Literals
1248 @section Compound Literals
1249 @cindex constructor expressions
1250 @cindex initializations in expressions
1251 @cindex structures, constructor expression
1252 @cindex expressions, constructor
1253 @cindex compound literals
1254 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1256 ISO C99 supports compound literals. A compound literal looks like
1257 a cast containing an initializer. Its value is an object of the
1258 type specified in the cast, containing the elements specified in
1259 the initializer; it is an lvalue. As an extension, GCC supports
1260 compound literals in C89 mode and in C++.
1262 Usually, the specified type is a structure. Assume that
1263 @code{struct foo} and @code{structure} are declared as shown:
1266 struct foo @{int a; char b[2];@} structure;
1270 Here is an example of constructing a @code{struct foo} with a compound literal:
1273 structure = ((struct foo) @{x + y, 'a', 0@});
1277 This is equivalent to writing the following:
1281 struct foo temp = @{x + y, 'a', 0@};
1286 You can also construct an array. If all the elements of the compound literal
1287 are (made up of) simple constant expressions, suitable for use in
1288 initializers of objects of static storage duration, then the compound
1289 literal can be coerced to a pointer to its first element and used in
1290 such an initializer, as shown here:
1293 char **foo = (char *[]) @{ "x", "y", "z" @};
1296 Compound literals for scalar types and union types are is
1297 also allowed, but then the compound literal is equivalent
1300 As a GNU extension, GCC allows initialization of objects with static storage
1301 duration by compound literals (which is not possible in ISO C99, because
1302 the initializer is not a constant).
1303 It is handled as if the object was initialized only with the bracket
1304 enclosed list if the types of the compound literal and the object match.
1305 The initializer list of the compound literal must be constant.
1306 If the object being initialized has array type of unknown size, the size is
1307 determined by compound literal size.
1310 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1311 static int y[] = (int []) @{1, 2, 3@};
1312 static int z[] = (int [3]) @{1@};
1316 The above lines are equivalent to the following:
1318 static struct foo x = @{1, 'a', 'b'@};
1319 static int y[] = @{1, 2, 3@};
1320 static int z[] = @{1, 0, 0@};
1323 @node Designated Inits
1324 @section Designated Initializers
1325 @cindex initializers with labeled elements
1326 @cindex labeled elements in initializers
1327 @cindex case labels in initializers
1328 @cindex designated initializers
1330 Standard C89 requires the elements of an initializer to appear in a fixed
1331 order, the same as the order of the elements in the array or structure
1334 In ISO C99 you can give the elements in any order, specifying the array
1335 indices or structure field names they apply to, and GNU C allows this as
1336 an extension in C89 mode as well. This extension is not
1337 implemented in GNU C++.
1339 To specify an array index, write
1340 @samp{[@var{index}] =} before the element value. For example,
1343 int a[6] = @{ [4] = 29, [2] = 15 @};
1350 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1354 The index values must be constant expressions, even if the array being
1355 initialized is automatic.
1357 An alternative syntax for this which has been obsolete since GCC 2.5 but
1358 GCC still accepts is to write @samp{[@var{index}]} before the element
1359 value, with no @samp{=}.
1361 To initialize a range of elements to the same value, write
1362 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1363 extension. For example,
1366 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1370 If the value in it has side-effects, the side-effects will happen only once,
1371 not for each initialized field by the range initializer.
1374 Note that the length of the array is the highest value specified
1377 In a structure initializer, specify the name of a field to initialize
1378 with @samp{.@var{fieldname} =} before the element value. For example,
1379 given the following structure,
1382 struct point @{ int x, y; @};
1386 the following initialization
1389 struct point p = @{ .y = yvalue, .x = xvalue @};
1396 struct point p = @{ xvalue, yvalue @};
1399 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1400 @samp{@var{fieldname}:}, as shown here:
1403 struct point p = @{ y: yvalue, x: xvalue @};
1407 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1408 @dfn{designator}. You can also use a designator (or the obsolete colon
1409 syntax) when initializing a union, to specify which element of the union
1410 should be used. For example,
1413 union foo @{ int i; double d; @};
1415 union foo f = @{ .d = 4 @};
1419 will convert 4 to a @code{double} to store it in the union using
1420 the second element. By contrast, casting 4 to type @code{union foo}
1421 would store it into the union as the integer @code{i}, since it is
1422 an integer. (@xref{Cast to Union}.)
1424 You can combine this technique of naming elements with ordinary C
1425 initialization of successive elements. Each initializer element that
1426 does not have a designator applies to the next consecutive element of the
1427 array or structure. For example,
1430 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1437 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1440 Labeling the elements of an array initializer is especially useful
1441 when the indices are characters or belong to an @code{enum} type.
1446 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1447 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1450 @cindex designator lists
1451 You can also write a series of @samp{.@var{fieldname}} and
1452 @samp{[@var{index}]} designators before an @samp{=} to specify a
1453 nested subobject to initialize; the list is taken relative to the
1454 subobject corresponding to the closest surrounding brace pair. For
1455 example, with the @samp{struct point} declaration above:
1458 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1462 If the same field is initialized multiple times, it will have value from
1463 the last initialization. If any such overridden initialization has
1464 side-effect, it is unspecified whether the side-effect happens or not.
1465 Currently, GCC will discard them and issue a warning.
1468 @section Case Ranges
1470 @cindex ranges in case statements
1472 You can specify a range of consecutive values in a single @code{case} label,
1476 case @var{low} ... @var{high}:
1480 This has the same effect as the proper number of individual @code{case}
1481 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1483 This feature is especially useful for ranges of ASCII character codes:
1489 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1490 it may be parsed wrong when you use it with integer values. For example,
1505 @section Cast to a Union Type
1506 @cindex cast to a union
1507 @cindex union, casting to a
1509 A cast to union type is similar to other casts, except that the type
1510 specified is a union type. You can specify the type either with
1511 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1512 a constructor though, not a cast, and hence does not yield an lvalue like
1513 normal casts. (@xref{Compound Literals}.)
1515 The types that may be cast to the union type are those of the members
1516 of the union. Thus, given the following union and variables:
1519 union foo @{ int i; double d; @};
1525 both @code{x} and @code{y} can be cast to type @code{union foo}.
1527 Using the cast as the right-hand side of an assignment to a variable of
1528 union type is equivalent to storing in a member of the union:
1533 u = (union foo) x @equiv{} u.i = x
1534 u = (union foo) y @equiv{} u.d = y
1537 You can also use the union cast as a function argument:
1540 void hack (union foo);
1542 hack ((union foo) x);
1545 @node Mixed Declarations
1546 @section Mixed Declarations and Code
1547 @cindex mixed declarations and code
1548 @cindex declarations, mixed with code
1549 @cindex code, mixed with declarations
1551 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1552 within compound statements. As an extension, GCC also allows this in
1553 C89 mode. For example, you could do:
1562 Each identifier is visible from where it is declared until the end of
1563 the enclosing block.
1565 @node Function Attributes
1566 @section Declaring Attributes of Functions
1567 @cindex function attributes
1568 @cindex declaring attributes of functions
1569 @cindex functions that never return
1570 @cindex functions that return more than once
1571 @cindex functions that have no side effects
1572 @cindex functions in arbitrary sections
1573 @cindex functions that behave like malloc
1574 @cindex @code{volatile} applied to function
1575 @cindex @code{const} applied to function
1576 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1577 @cindex functions with non-null pointer arguments
1578 @cindex functions that are passed arguments in registers on the 386
1579 @cindex functions that pop the argument stack on the 386
1580 @cindex functions that do not pop the argument stack on the 386
1582 In GNU C, you declare certain things about functions called in your program
1583 which help the compiler optimize function calls and check your code more
1586 The keyword @code{__attribute__} allows you to specify special
1587 attributes when making a declaration. This keyword is followed by an
1588 attribute specification inside double parentheses. The following
1589 attributes are currently defined for functions on all targets:
1591 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1592 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1593 @code{format}, @code{format_arg}, @code{no_instrument_function},
1594 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1595 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1596 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1597 @code{gnu_inline} and @code{externally_visible}. Several other
1598 attributes are defined for functions on particular target systems. Other
1599 attributes, including @code{section} are supported for variables declarations
1600 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
1601 (@pxref{Variable Attributes}), for types (@pxref{Type Attributes}),
1602 and labels (@pxref{Label Attributes}).
1604 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
1605 You may also specify attributes with @samp{__} preceding and following
1606 each keyword. This allows you to use them in header files without
1607 being concerned about a possible macro of the same name. For example,
1608 you may use @code{__noreturn__} instead of @code{noreturn}.
1610 @xref{Attribute Syntax}, for details of the exact syntax for using
1614 @c Keep this table alphabetized by attribute name. Treat _ as space.
1616 @item alias ("@var{target}")
1617 @cindex @code{alias} attribute
1618 The @code{alias} attribute causes the declaration to be emitted as an
1619 alias for another symbol, which must be specified. For instance,
1622 void __f () @{ /* @r{Do something.} */; @}
1623 void f () __attribute__ ((weak, alias ("__f")));
1626 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1627 mangled name for the target must be used. It is an error if @samp{__f}
1628 is not defined in the same translation unit.
1630 Not all target machines support this attribute.
1632 @item aligned (@var{alignment})
1633 @cindex @code{aligned} attribute
1634 This attribute specifies a minimum alignment for the function,
1637 You cannot use this attribute to decrease the alignment of a function,
1638 only to increase it. However, when you explicitly specify a function
1639 alignment this will override the effect of the
1640 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1643 Note that the effectiveness of @code{aligned} attributes may be
1644 limited by inherent limitations in your linker. On many systems, the
1645 linker is only able to arrange for functions to be aligned up to a
1646 certain maximum alignment. (For some linkers, the maximum supported
1647 alignment may be very very small.) See your linker documentation for
1648 further information.
1650 The @code{aligned} attribute can also be used for variables and fields
1651 (@pxref{Variable Attributes}.)
1654 @cindex @code{always_inline} function attribute
1655 Generally, functions are not inlined unless optimization is specified.
1656 For functions declared inline, this attribute inlines the function even
1657 if no optimization level was specified.
1660 @cindex @code{gnu_inline} function attribute
1661 This attribute should be used with a function which is also declared
1662 with the @code{inline} keyword. It directs GCC to treat the function
1663 as if it were defined in gnu89 mode even when compiling in C99 or
1666 If the function is declared @code{extern}, then this definition of the
1667 function is used only for inlining. In no case is the function
1668 compiled as a standalone function, not even if you take its address
1669 explicitly. Such an address becomes an external reference, as if you
1670 had only declared the function, and had not defined it. This has
1671 almost the effect of a macro. The way to use this is to put a
1672 function definition in a header file with this attribute, and put
1673 another copy of the function, without @code{extern}, in a library
1674 file. The definition in the header file will cause most calls to the
1675 function to be inlined. If any uses of the function remain, they will
1676 refer to the single copy in the library. Note that the two
1677 definitions of the functions need not be precisely the same, although
1678 if they do not have the same effect your program may behave oddly.
1680 If the function is neither @code{extern} nor @code{static}, then the
1681 function is compiled as a standalone function, as well as being
1682 inlined where possible.
1684 This is how GCC traditionally handled functions declared
1685 @code{inline}. Since ISO C99 specifies a different semantics for
1686 @code{inline}, this function attribute is provided as a transition
1687 measure and as a useful feature in its own right. This attribute is
1688 available in GCC 4.1.3 and later. It is available if either of the
1689 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1690 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1691 Function is As Fast As a Macro}.
1693 Note that since the first version of GCC to support C99 inline semantics
1694 is 4.3, earlier versions of GCC which accept this attribute effectively
1695 assume that it is always present, whether or not it is given explicitly.
1696 In versions prior to 4.3, the only effect of explicitly including it is
1697 to disable warnings about using inline functions in C99 mode.
1699 @cindex @code{flatten} function attribute
1701 Generally, inlining into a function is limited. For a function marked with
1702 this attribute, every call inside this function will be inlined, if possible.
1703 Whether the function itself is considered for inlining depends on its size and
1704 the current inlining parameters. The @code{flatten} attribute only works
1705 reliably in unit-at-a-time mode.
1708 @cindex functions that do pop the argument stack on the 386
1710 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1711 assume that the calling function will pop off the stack space used to
1712 pass arguments. This is
1713 useful to override the effects of the @option{-mrtd} switch.
1716 @cindex @code{const} function attribute
1717 Many functions do not examine any values except their arguments, and
1718 have no effects except the return value. Basically this is just slightly
1719 more strict class than the @code{pure} attribute below, since function is not
1720 allowed to read global memory.
1722 @cindex pointer arguments
1723 Note that a function that has pointer arguments and examines the data
1724 pointed to must @emph{not} be declared @code{const}. Likewise, a
1725 function that calls a non-@code{const} function usually must not be
1726 @code{const}. It does not make sense for a @code{const} function to
1729 The attribute @code{const} is not implemented in GCC versions earlier
1730 than 2.5. An alternative way to declare that a function has no side
1731 effects, which works in the current version and in some older versions,
1735 typedef int intfn ();
1737 extern const intfn square;
1740 This approach does not work in GNU C++ from 2.6.0 on, since the language
1741 specifies that the @samp{const} must be attached to the return value.
1745 @cindex @code{constructor} function attribute
1746 @cindex @code{destructor} function attribute
1747 The @code{constructor} attribute causes the function to be called
1748 automatically before execution enters @code{main ()}. Similarly, the
1749 @code{destructor} attribute causes the function to be called
1750 automatically after @code{main ()} has completed or @code{exit ()} has
1751 been called. Functions with these attributes are useful for
1752 initializing data that will be used implicitly during the execution of
1756 @cindex @code{deprecated} attribute.
1757 The @code{deprecated} attribute results in a warning if the function
1758 is used anywhere in the source file. This is useful when identifying
1759 functions that are expected to be removed in a future version of a
1760 program. The warning also includes the location of the declaration
1761 of the deprecated function, to enable users to easily find further
1762 information about why the function is deprecated, or what they should
1763 do instead. Note that the warnings only occurs for uses:
1766 int old_fn () __attribute__ ((deprecated));
1768 int (*fn_ptr)() = old_fn;
1771 results in a warning on line 3 but not line 2.
1773 The @code{deprecated} attribute can also be used for variables and
1774 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1777 @cindex @code{__declspec(dllexport)}
1778 On Microsoft Windows targets and Symbian OS targets the
1779 @code{dllexport} attribute causes the compiler to provide a global
1780 pointer to a pointer in a DLL, so that it can be referenced with the
1781 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1782 name is formed by combining @code{_imp__} and the function or variable
1785 You can use @code{__declspec(dllexport)} as a synonym for
1786 @code{__attribute__ ((dllexport))} for compatibility with other
1789 On systems that support the @code{visibility} attribute, this
1790 attribute also implies ``default'' visibility, unless a
1791 @code{visibility} attribute is explicitly specified. You should avoid
1792 the use of @code{dllexport} with ``hidden'' or ``internal''
1793 visibility; in the future GCC may issue an error for those cases.
1795 Currently, the @code{dllexport} attribute is ignored for inlined
1796 functions, unless the @option{-fkeep-inline-functions} flag has been
1797 used. The attribute is also ignored for undefined symbols.
1799 When applied to C++ classes, the attribute marks defined non-inlined
1800 member functions and static data members as exports. Static consts
1801 initialized in-class are not marked unless they are also defined
1804 For Microsoft Windows targets there are alternative methods for
1805 including the symbol in the DLL's export table such as using a
1806 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1807 the @option{--export-all} linker flag.
1810 @cindex @code{__declspec(dllimport)}
1811 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1812 attribute causes the compiler to reference a function or variable via
1813 a global pointer to a pointer that is set up by the DLL exporting the
1814 symbol. The attribute implies @code{extern} storage. On Microsoft
1815 Windows targets, the pointer name is formed by combining @code{_imp__}
1816 and the function or variable name.
1818 You can use @code{__declspec(dllimport)} as a synonym for
1819 @code{__attribute__ ((dllimport))} for compatibility with other
1822 Currently, the attribute is ignored for inlined functions. If the
1823 attribute is applied to a symbol @emph{definition}, an error is reported.
1824 If a symbol previously declared @code{dllimport} is later defined, the
1825 attribute is ignored in subsequent references, and a warning is emitted.
1826 The attribute is also overridden by a subsequent declaration as
1829 When applied to C++ classes, the attribute marks non-inlined
1830 member functions and static data members as imports. However, the
1831 attribute is ignored for virtual methods to allow creation of vtables
1834 On the SH Symbian OS target the @code{dllimport} attribute also has
1835 another affect---it can cause the vtable and run-time type information
1836 for a class to be exported. This happens when the class has a
1837 dllimport'ed constructor or a non-inline, non-pure virtual function
1838 and, for either of those two conditions, the class also has a inline
1839 constructor or destructor and has a key function that is defined in
1840 the current translation unit.
1842 For Microsoft Windows based targets the use of the @code{dllimport}
1843 attribute on functions is not necessary, but provides a small
1844 performance benefit by eliminating a thunk in the DLL@. The use of the
1845 @code{dllimport} attribute on imported variables was required on older
1846 versions of the GNU linker, but can now be avoided by passing the
1847 @option{--enable-auto-import} switch to the GNU linker. As with
1848 functions, using the attribute for a variable eliminates a thunk in
1851 One drawback to using this attribute is that a pointer to a function
1852 or variable marked as @code{dllimport} cannot be used as a constant
1853 address. On Microsoft Windows targets, the attribute can be disabled
1854 for functions by setting the @option{-mnop-fun-dllimport} flag.
1857 @cindex eight bit data on the H8/300, H8/300H, and H8S
1858 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1859 variable should be placed into the eight bit data section.
1860 The compiler will generate more efficient code for certain operations
1861 on data in the eight bit data area. Note the eight bit data area is limited to
1864 You must use GAS and GLD from GNU binutils version 2.7 or later for
1865 this attribute to work correctly.
1867 @item exception_handler
1868 @cindex exception handler functions on the Blackfin processor
1869 Use this attribute on the Blackfin to indicate that the specified function
1870 is an exception handler. The compiler will generate function entry and
1871 exit sequences suitable for use in an exception handler when this
1872 attribute is present.
1875 @cindex functions which handle memory bank switching
1876 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1877 use a calling convention that takes care of switching memory banks when
1878 entering and leaving a function. This calling convention is also the
1879 default when using the @option{-mlong-calls} option.
1881 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1882 to call and return from a function.
1884 On 68HC11 the compiler will generate a sequence of instructions
1885 to invoke a board-specific routine to switch the memory bank and call the
1886 real function. The board-specific routine simulates a @code{call}.
1887 At the end of a function, it will jump to a board-specific routine
1888 instead of using @code{rts}. The board-specific return routine simulates
1892 @cindex functions that pop the argument stack on the 386
1893 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1894 pass the first argument (if of integral type) in the register ECX and
1895 the second argument (if of integral type) in the register EDX@. Subsequent
1896 and other typed arguments are passed on the stack. The called function will
1897 pop the arguments off the stack. If the number of arguments is variable all
1898 arguments are pushed on the stack.
1900 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1901 @cindex @code{format} function attribute
1903 The @code{format} attribute specifies that a function takes @code{printf},
1904 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1905 should be type-checked against a format string. For example, the
1910 my_printf (void *my_object, const char *my_format, ...)
1911 __attribute__ ((format (printf, 2, 3)));
1915 causes the compiler to check the arguments in calls to @code{my_printf}
1916 for consistency with the @code{printf} style format string argument
1919 The parameter @var{archetype} determines how the format string is
1920 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1921 or @code{strfmon}. (You can also use @code{__printf__},
1922 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1923 parameter @var{string-index} specifies which argument is the format
1924 string argument (starting from 1), while @var{first-to-check} is the
1925 number of the first argument to check against the format string. For
1926 functions where the arguments are not available to be checked (such as
1927 @code{vprintf}), specify the third parameter as zero. In this case the
1928 compiler only checks the format string for consistency. For
1929 @code{strftime} formats, the third parameter is required to be zero.
1930 Since non-static C++ methods have an implicit @code{this} argument, the
1931 arguments of such methods should be counted from two, not one, when
1932 giving values for @var{string-index} and @var{first-to-check}.
1934 In the example above, the format string (@code{my_format}) is the second
1935 argument of the function @code{my_print}, and the arguments to check
1936 start with the third argument, so the correct parameters for the format
1937 attribute are 2 and 3.
1939 @opindex ffreestanding
1940 @opindex fno-builtin
1941 The @code{format} attribute allows you to identify your own functions
1942 which take format strings as arguments, so that GCC can check the
1943 calls to these functions for errors. The compiler always (unless
1944 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1945 for the standard library functions @code{printf}, @code{fprintf},
1946 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1947 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1948 warnings are requested (using @option{-Wformat}), so there is no need to
1949 modify the header file @file{stdio.h}. In C99 mode, the functions
1950 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1951 @code{vsscanf} are also checked. Except in strictly conforming C
1952 standard modes, the X/Open function @code{strfmon} is also checked as
1953 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1954 @xref{C Dialect Options,,Options Controlling C Dialect}.
1956 The target may provide additional types of format checks.
1957 @xref{Target Format Checks,,Format Checks Specific to Particular
1960 @item format_arg (@var{string-index})
1961 @cindex @code{format_arg} function attribute
1962 @opindex Wformat-nonliteral
1963 The @code{format_arg} attribute specifies that a function takes a format
1964 string for a @code{printf}, @code{scanf}, @code{strftime} or
1965 @code{strfmon} style function and modifies it (for example, to translate
1966 it into another language), so the result can be passed to a
1967 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1968 function (with the remaining arguments to the format function the same
1969 as they would have been for the unmodified string). For example, the
1974 my_dgettext (char *my_domain, const char *my_format)
1975 __attribute__ ((format_arg (2)));
1979 causes the compiler to check the arguments in calls to a @code{printf},
1980 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1981 format string argument is a call to the @code{my_dgettext} function, for
1982 consistency with the format string argument @code{my_format}. If the
1983 @code{format_arg} attribute had not been specified, all the compiler
1984 could tell in such calls to format functions would be that the format
1985 string argument is not constant; this would generate a warning when
1986 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1987 without the attribute.
1989 The parameter @var{string-index} specifies which argument is the format
1990 string argument (starting from one). Since non-static C++ methods have
1991 an implicit @code{this} argument, the arguments of such methods should
1992 be counted from two.
1994 The @code{format-arg} attribute allows you to identify your own
1995 functions which modify format strings, so that GCC can check the
1996 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1997 type function whose operands are a call to one of your own function.
1998 The compiler always treats @code{gettext}, @code{dgettext}, and
1999 @code{dcgettext} in this manner except when strict ISO C support is
2000 requested by @option{-ansi} or an appropriate @option{-std} option, or
2001 @option{-ffreestanding} or @option{-fno-builtin}
2002 is used. @xref{C Dialect Options,,Options
2003 Controlling C Dialect}.
2005 @item function_vector
2006 @cindex calling functions through the function vector on the H8/300 processors
2007 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2008 function should be called through the function vector. Calling a
2009 function through the function vector will reduce code size, however;
2010 the function vector has a limited size (maximum 128 entries on the H8/300
2011 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2013 You must use GAS and GLD from GNU binutils version 2.7 or later for
2014 this attribute to work correctly.
2017 @cindex interrupt handler functions
2018 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
2019 ports to indicate that the specified function is an interrupt handler.
2020 The compiler will generate function entry and exit sequences suitable
2021 for use in an interrupt handler when this attribute is present.
2023 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
2024 SH processors can be specified via the @code{interrupt_handler} attribute.
2026 Note, on the AVR, interrupts will be enabled inside the function.
2028 Note, for the ARM, you can specify the kind of interrupt to be handled by
2029 adding an optional parameter to the interrupt attribute like this:
2032 void f () __attribute__ ((interrupt ("IRQ")));
2035 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2037 @item interrupt_handler
2038 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2039 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2040 indicate that the specified function is an interrupt handler. The compiler
2041 will generate function entry and exit sequences suitable for use in an
2042 interrupt handler when this attribute is present.
2045 @cindex User stack pointer in interrupts on the Blackfin
2046 When used together with @code{interrupt_handler}, @code{exception_handler}
2047 or @code{nmi_handler}, code will be generated to load the stack pointer
2048 from the USP register in the function prologue.
2050 @item long_call/short_call
2051 @cindex indirect calls on ARM
2052 This attribute specifies how a particular function is called on
2053 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2054 command line switch and @code{#pragma long_calls} settings. The
2055 @code{long_call} attribute indicates that the function might be far
2056 away from the call site and require a different (more expensive)
2057 calling sequence. The @code{short_call} attribute always places
2058 the offset to the function from the call site into the @samp{BL}
2059 instruction directly.
2061 @item longcall/shortcall
2062 @cindex functions called via pointer on the RS/6000 and PowerPC
2063 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2064 indicates that the function might be far away from the call site and
2065 require a different (more expensive) calling sequence. The
2066 @code{shortcall} attribute indicates that the function is always close
2067 enough for the shorter calling sequence to be used. These attributes
2068 override both the @option{-mlongcall} switch and, on the RS/6000 and
2069 PowerPC, the @code{#pragma longcall} setting.
2071 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2072 calls are necessary.
2075 @cindex indirect calls on MIPS
2076 This attribute specifies how a particular function is called on MIPS@.
2077 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2078 command line switch. This attribute causes the compiler to always call
2079 the function by first loading its address into a register, and then using
2080 the contents of that register.
2083 @cindex @code{malloc} attribute
2084 The @code{malloc} attribute is used to tell the compiler that a function
2085 may be treated as if any non-@code{NULL} pointer it returns cannot
2086 alias any other pointer valid when the function returns.
2087 This will often improve optimization.
2088 Standard functions with this property include @code{malloc} and
2089 @code{calloc}. @code{realloc}-like functions have this property as
2090 long as the old pointer is never referred to (including comparing it
2091 to the new pointer) after the function returns a non-@code{NULL}
2094 @item model (@var{model-name})
2095 @cindex function addressability on the M32R/D
2096 @cindex variable addressability on the IA-64
2098 On the M32R/D, use this attribute to set the addressability of an
2099 object, and of the code generated for a function. The identifier
2100 @var{model-name} is one of @code{small}, @code{medium}, or
2101 @code{large}, representing each of the code models.
2103 Small model objects live in the lower 16MB of memory (so that their
2104 addresses can be loaded with the @code{ld24} instruction), and are
2105 callable with the @code{bl} instruction.
2107 Medium model objects may live anywhere in the 32-bit address space (the
2108 compiler will generate @code{seth/add3} instructions to load their addresses),
2109 and are callable with the @code{bl} instruction.
2111 Large model objects may live anywhere in the 32-bit address space (the
2112 compiler will generate @code{seth/add3} instructions to load their addresses),
2113 and may not be reachable with the @code{bl} instruction (the compiler will
2114 generate the much slower @code{seth/add3/jl} instruction sequence).
2116 On IA-64, use this attribute to set the addressability of an object.
2117 At present, the only supported identifier for @var{model-name} is
2118 @code{small}, indicating addressability via ``small'' (22-bit)
2119 addresses (so that their addresses can be loaded with the @code{addl}
2120 instruction). Caveat: such addressing is by definition not position
2121 independent and hence this attribute must not be used for objects
2122 defined by shared libraries.
2125 @cindex function without a prologue/epilogue code
2126 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2127 specified function does not need prologue/epilogue sequences generated by
2128 the compiler. It is up to the programmer to provide these sequences.
2131 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2132 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2133 use the normal calling convention based on @code{jsr} and @code{rts}.
2134 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2138 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2139 Use this attribute together with @code{interrupt_handler},
2140 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2141 entry code should enable nested interrupts or exceptions.
2144 @cindex NMI handler functions on the Blackfin processor
2145 Use this attribute on the Blackfin to indicate that the specified function
2146 is an NMI handler. The compiler will generate function entry and
2147 exit sequences suitable for use in an NMI handler when this
2148 attribute is present.
2150 @item no_instrument_function
2151 @cindex @code{no_instrument_function} function attribute
2152 @opindex finstrument-functions
2153 If @option{-finstrument-functions} is given, profiling function calls will
2154 be generated at entry and exit of most user-compiled functions.
2155 Functions with this attribute will not be so instrumented.
2158 @cindex @code{noinline} function attribute
2159 This function attribute prevents a function from being considered for
2162 @item nonnull (@var{arg-index}, @dots{})
2163 @cindex @code{nonnull} function attribute
2164 The @code{nonnull} attribute specifies that some function parameters should
2165 be non-null pointers. For instance, the declaration:
2169 my_memcpy (void *dest, const void *src, size_t len)
2170 __attribute__((nonnull (1, 2)));
2174 causes the compiler to check that, in calls to @code{my_memcpy},
2175 arguments @var{dest} and @var{src} are non-null. If the compiler
2176 determines that a null pointer is passed in an argument slot marked
2177 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2178 is issued. The compiler may also choose to make optimizations based
2179 on the knowledge that certain function arguments will not be null.
2181 If no argument index list is given to the @code{nonnull} attribute,
2182 all pointer arguments are marked as non-null. To illustrate, the
2183 following declaration is equivalent to the previous example:
2187 my_memcpy (void *dest, const void *src, size_t len)
2188 __attribute__((nonnull));
2192 @cindex @code{noreturn} function attribute
2193 A few standard library functions, such as @code{abort} and @code{exit},
2194 cannot return. GCC knows this automatically. Some programs define
2195 their own functions that never return. You can declare them
2196 @code{noreturn} to tell the compiler this fact. For example,
2200 void fatal () __attribute__ ((noreturn));
2203 fatal (/* @r{@dots{}} */)
2205 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2211 The @code{noreturn} keyword tells the compiler to assume that
2212 @code{fatal} cannot return. It can then optimize without regard to what
2213 would happen if @code{fatal} ever did return. This makes slightly
2214 better code. More importantly, it helps avoid spurious warnings of
2215 uninitialized variables.
2217 The @code{noreturn} keyword does not affect the exceptional path when that
2218 applies: a @code{noreturn}-marked function may still return to the caller
2219 by throwing an exception or calling @code{longjmp}.
2221 Do not assume that registers saved by the calling function are
2222 restored before calling the @code{noreturn} function.
2224 It does not make sense for a @code{noreturn} function to have a return
2225 type other than @code{void}.
2227 The attribute @code{noreturn} is not implemented in GCC versions
2228 earlier than 2.5. An alternative way to declare that a function does
2229 not return, which works in the current version and in some older
2230 versions, is as follows:
2233 typedef void voidfn ();
2235 volatile voidfn fatal;
2238 This approach does not work in GNU C++.
2241 @cindex @code{nothrow} function attribute
2242 The @code{nothrow} attribute is used to inform the compiler that a
2243 function cannot throw an exception. For example, most functions in
2244 the standard C library can be guaranteed not to throw an exception
2245 with the notable exceptions of @code{qsort} and @code{bsearch} that
2246 take function pointer arguments. The @code{nothrow} attribute is not
2247 implemented in GCC versions earlier than 3.3.
2250 @cindex @code{pure} function attribute
2251 Many functions have no effects except the return value and their
2252 return value depends only on the parameters and/or global variables.
2253 Such a function can be subject
2254 to common subexpression elimination and loop optimization just as an
2255 arithmetic operator would be. These functions should be declared
2256 with the attribute @code{pure}. For example,
2259 int square (int) __attribute__ ((pure));
2263 says that the hypothetical function @code{square} is safe to call
2264 fewer times than the program says.
2266 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2267 Interesting non-pure functions are functions with infinite loops or those
2268 depending on volatile memory or other system resource, that may change between
2269 two consecutive calls (such as @code{feof} in a multithreading environment).
2271 The attribute @code{pure} is not implemented in GCC versions earlier
2274 @item regparm (@var{number})
2275 @cindex @code{regparm} attribute
2276 @cindex functions that are passed arguments in registers on the 386
2277 On the Intel 386, the @code{regparm} attribute causes the compiler to
2278 pass arguments number one to @var{number} if they are of integral type
2279 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2280 take a variable number of arguments will continue to be passed all of their
2281 arguments on the stack.
2283 Beware that on some ELF systems this attribute is unsuitable for
2284 global functions in shared libraries with lazy binding (which is the
2285 default). Lazy binding will send the first call via resolving code in
2286 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2287 per the standard calling conventions. Solaris 8 is affected by this.
2288 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2289 safe since the loaders there save all registers. (Lazy binding can be
2290 disabled with the linker or the loader if desired, to avoid the
2294 @cindex @code{sseregparm} attribute
2295 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2296 causes the compiler to pass up to 3 floating point arguments in
2297 SSE registers instead of on the stack. Functions that take a
2298 variable number of arguments will continue to pass all of their
2299 floating point arguments on the stack.
2301 @item force_align_arg_pointer
2302 @cindex @code{force_align_arg_pointer} attribute
2303 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2304 applied to individual function definitions, generating an alternate
2305 prologue and epilogue that realigns the runtime stack. This supports
2306 mixing legacy codes that run with a 4-byte aligned stack with modern
2307 codes that keep a 16-byte stack for SSE compatibility. The alternate
2308 prologue and epilogue are slower and bigger than the regular ones, and
2309 the alternate prologue requires a scratch register; this lowers the
2310 number of registers available if used in conjunction with the
2311 @code{regparm} attribute. The @code{force_align_arg_pointer}
2312 attribute is incompatible with nested functions; this is considered a
2316 @cindex @code{returns_twice} attribute
2317 The @code{returns_twice} attribute tells the compiler that a function may
2318 return more than one time. The compiler will ensure that all registers
2319 are dead before calling such a function and will emit a warning about
2320 the variables that may be clobbered after the second return from the
2321 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2322 The @code{longjmp}-like counterpart of such function, if any, might need
2323 to be marked with the @code{noreturn} attribute.
2326 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2327 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2328 all registers except the stack pointer should be saved in the prologue
2329 regardless of whether they are used or not.
2331 @item section ("@var{section-name}")
2332 @cindex @code{section} function attribute
2333 Normally, the compiler places the code it generates in the @code{text} section.
2334 Sometimes, however, you need additional sections, or you need certain
2335 particular functions to appear in special sections. The @code{section}
2336 attribute specifies that a function lives in a particular section.
2337 For example, the declaration:
2340 extern void foobar (void) __attribute__ ((section ("bar")));
2344 puts the function @code{foobar} in the @code{bar} section.
2346 Some file formats do not support arbitrary sections so the @code{section}
2347 attribute is not available on all platforms.
2348 If you need to map the entire contents of a module to a particular
2349 section, consider using the facilities of the linker instead.
2352 @cindex @code{sentinel} function attribute
2353 This function attribute ensures that a parameter in a function call is
2354 an explicit @code{NULL}. The attribute is only valid on variadic
2355 functions. By default, the sentinel is located at position zero, the
2356 last parameter of the function call. If an optional integer position
2357 argument P is supplied to the attribute, the sentinel must be located at
2358 position P counting backwards from the end of the argument list.
2361 __attribute__ ((sentinel))
2363 __attribute__ ((sentinel(0)))
2366 The attribute is automatically set with a position of 0 for the built-in
2367 functions @code{execl} and @code{execlp}. The built-in function
2368 @code{execle} has the attribute set with a position of 1.
2370 A valid @code{NULL} in this context is defined as zero with any pointer
2371 type. If your system defines the @code{NULL} macro with an integer type
2372 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2373 with a copy that redefines NULL appropriately.
2375 The warnings for missing or incorrect sentinels are enabled with
2379 See long_call/short_call.
2382 See longcall/shortcall.
2385 @cindex signal handler functions on the AVR processors
2386 Use this attribute on the AVR to indicate that the specified
2387 function is a signal handler. The compiler will generate function
2388 entry and exit sequences suitable for use in a signal handler when this
2389 attribute is present. Interrupts will be disabled inside the function.
2392 Use this attribute on the SH to indicate an @code{interrupt_handler}
2393 function should switch to an alternate stack. It expects a string
2394 argument that names a global variable holding the address of the
2399 void f () __attribute__ ((interrupt_handler,
2400 sp_switch ("alt_stack")));
2404 @cindex functions that pop the argument stack on the 386
2405 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2406 assume that the called function will pop off the stack space used to
2407 pass arguments, unless it takes a variable number of arguments.
2410 @cindex tiny data section on the H8/300H and H8S
2411 Use this attribute on the H8/300H and H8S to indicate that the specified
2412 variable should be placed into the tiny data section.
2413 The compiler will generate more efficient code for loads and stores
2414 on data in the tiny data section. Note the tiny data area is limited to
2415 slightly under 32kbytes of data.
2418 Use this attribute on the SH for an @code{interrupt_handler} to return using
2419 @code{trapa} instead of @code{rte}. This attribute expects an integer
2420 argument specifying the trap number to be used.
2423 @cindex @code{unused} attribute.
2424 This attribute, attached to a function, means that the function is meant
2425 to be possibly unused. GCC will not produce a warning for this
2429 @cindex @code{used} attribute.
2430 This attribute, attached to a function, means that code must be emitted
2431 for the function even if it appears that the function is not referenced.
2432 This is useful, for example, when the function is referenced only in
2435 @item visibility ("@var{visibility_type}")
2436 @cindex @code{visibility} attribute
2437 This attribute affects the linkage of the declaration to which it is attached.
2438 There are four supported @var{visibility_type} values: default,
2439 hidden, protected or internal visibility.
2442 void __attribute__ ((visibility ("protected")))
2443 f () @{ /* @r{Do something.} */; @}
2444 int i __attribute__ ((visibility ("hidden")));
2447 The possible values of @var{visibility_type} correspond to the
2448 visibility settings in the ELF gABI.
2451 @c keep this list of visibilities in alphabetical order.
2454 Default visibility is the normal case for the object file format.
2455 This value is available for the visibility attribute to override other
2456 options that may change the assumed visibility of entities.
2458 On ELF, default visibility means that the declaration is visible to other
2459 modules and, in shared libraries, means that the declared entity may be
2462 On Darwin, default visibility means that the declaration is visible to
2465 Default visibility corresponds to ``external linkage'' in the language.
2468 Hidden visibility indicates that the entity declared will have a new
2469 form of linkage, which we'll call ``hidden linkage''. Two
2470 declarations of an object with hidden linkage refer to the same object
2471 if they are in the same shared object.
2474 Internal visibility is like hidden visibility, but with additional
2475 processor specific semantics. Unless otherwise specified by the
2476 psABI, GCC defines internal visibility to mean that a function is
2477 @emph{never} called from another module. Compare this with hidden
2478 functions which, while they cannot be referenced directly by other
2479 modules, can be referenced indirectly via function pointers. By
2480 indicating that a function cannot be called from outside the module,
2481 GCC may for instance omit the load of a PIC register since it is known
2482 that the calling function loaded the correct value.
2485 Protected visibility is like default visibility except that it
2486 indicates that references within the defining module will bind to the
2487 definition in that module. That is, the declared entity cannot be
2488 overridden by another module.
2492 All visibilities are supported on many, but not all, ELF targets
2493 (supported when the assembler supports the @samp{.visibility}
2494 pseudo-op). Default visibility is supported everywhere. Hidden
2495 visibility is supported on Darwin targets.
2497 The visibility attribute should be applied only to declarations which
2498 would otherwise have external linkage. The attribute should be applied
2499 consistently, so that the same entity should not be declared with
2500 different settings of the attribute.
2502 In C++, the visibility attribute applies to types as well as functions
2503 and objects, because in C++ types have linkage. A class must not have
2504 greater visibility than its non-static data member types and bases,
2505 and class members default to the visibility of their class. Also, a
2506 declaration without explicit visibility is limited to the visibility
2509 In C++, you can mark member functions and static member variables of a
2510 class with the visibility attribute. This is useful if if you know a
2511 particular method or static member variable should only be used from
2512 one shared object; then you can mark it hidden while the rest of the
2513 class has default visibility. Care must be taken to avoid breaking
2514 the One Definition Rule; for example, it is usually not useful to mark
2515 an inline method as hidden without marking the whole class as hidden.
2517 A C++ namespace declaration can also have the visibility attribute.
2518 This attribute applies only to the particular namespace body, not to
2519 other definitions of the same namespace; it is equivalent to using
2520 @samp{#pragma GCC visibility} before and after the namespace
2521 definition (@pxref{Visibility Pragmas}).
2523 In C++, if a template argument has limited visibility, this
2524 restriction is implicitly propagated to the template instantiation.
2525 Otherwise, template instantiations and specializations default to the
2526 visibility of their template.
2528 If both the template and enclosing class have explicit visibility, the
2529 visibility from the template is used.
2531 @item warn_unused_result
2532 @cindex @code{warn_unused_result} attribute
2533 The @code{warn_unused_result} attribute causes a warning to be emitted
2534 if a caller of the function with this attribute does not use its
2535 return value. This is useful for functions where not checking
2536 the result is either a security problem or always a bug, such as
2540 int fn () __attribute__ ((warn_unused_result));
2543 if (fn () < 0) return -1;
2549 results in warning on line 5.
2552 @cindex @code{weak} attribute
2553 The @code{weak} attribute causes the declaration to be emitted as a weak
2554 symbol rather than a global. This is primarily useful in defining
2555 library functions which can be overridden in user code, though it can
2556 also be used with non-function declarations. Weak symbols are supported
2557 for ELF targets, and also for a.out targets when using the GNU assembler
2561 @itemx weakref ("@var{target}")
2562 @cindex @code{weakref} attribute
2563 The @code{weakref} attribute marks a declaration as a weak reference.
2564 Without arguments, it should be accompanied by an @code{alias} attribute
2565 naming the target symbol. Optionally, the @var{target} may be given as
2566 an argument to @code{weakref} itself. In either case, @code{weakref}
2567 implicitly marks the declaration as @code{weak}. Without a
2568 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2569 @code{weakref} is equivalent to @code{weak}.
2572 static int x() __attribute__ ((weakref ("y")));
2573 /* is equivalent to... */
2574 static int x() __attribute__ ((weak, weakref, alias ("y")));
2576 static int x() __attribute__ ((weakref));
2577 static int x() __attribute__ ((alias ("y")));
2580 A weak reference is an alias that does not by itself require a
2581 definition to be given for the target symbol. If the target symbol is
2582 only referenced through weak references, then the becomes a @code{weak}
2583 undefined symbol. If it is directly referenced, however, then such
2584 strong references prevail, and a definition will be required for the
2585 symbol, not necessarily in the same translation unit.
2587 The effect is equivalent to moving all references to the alias to a
2588 separate translation unit, renaming the alias to the aliased symbol,
2589 declaring it as weak, compiling the two separate translation units and
2590 performing a reloadable link on them.
2592 At present, a declaration to which @code{weakref} is attached can
2593 only be @code{static}.
2595 @item externally_visible
2596 @cindex @code{externally_visible} attribute.
2597 This attribute, attached to a global variable or function nullify
2598 effect of @option{-fwhole-program} command line option, so the object
2599 remain visible outside the current compilation unit
2603 You can specify multiple attributes in a declaration by separating them
2604 by commas within the double parentheses or by immediately following an
2605 attribute declaration with another attribute declaration.
2607 @cindex @code{#pragma}, reason for not using
2608 @cindex pragma, reason for not using
2609 Some people object to the @code{__attribute__} feature, suggesting that
2610 ISO C's @code{#pragma} should be used instead. At the time
2611 @code{__attribute__} was designed, there were two reasons for not doing
2616 It is impossible to generate @code{#pragma} commands from a macro.
2619 There is no telling what the same @code{#pragma} might mean in another
2623 These two reasons applied to almost any application that might have been
2624 proposed for @code{#pragma}. It was basically a mistake to use
2625 @code{#pragma} for @emph{anything}.
2627 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2628 to be generated from macros. In addition, a @code{#pragma GCC}
2629 namespace is now in use for GCC-specific pragmas. However, it has been
2630 found convenient to use @code{__attribute__} to achieve a natural
2631 attachment of attributes to their corresponding declarations, whereas
2632 @code{#pragma GCC} is of use for constructs that do not naturally form
2633 part of the grammar. @xref{Other Directives,,Miscellaneous
2634 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2636 @node Attribute Syntax
2637 @section Attribute Syntax
2638 @cindex attribute syntax
2640 This section describes the syntax with which @code{__attribute__} may be
2641 used, and the constructs to which attribute specifiers bind, for the C
2642 language. Some details may vary for C++. Because of infelicities in
2643 the grammar for attributes, some forms described here may not be
2644 successfully parsed in all cases.
2646 There are some problems with the semantics of attributes in C++. For
2647 example, there are no manglings for attributes, although they may affect
2648 code generation, so problems may arise when attributed types are used in
2649 conjunction with templates or overloading. Similarly, @code{typeid}
2650 does not distinguish between types with different attributes. Support
2651 for attributes in C++ may be restricted in future to attributes on
2652 declarations only, but not on nested declarators.
2654 @xref{Function Attributes}, for details of the semantics of attributes
2655 applying to functions. @xref{Variable Attributes}, for details of the
2656 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
2657 semantics of attributes applying to variables. @xref{Type
2658 Attributes}, for details of the semantics of attributes applying to
2659 structure, union and enumerated types. @xref{Label Attributes}, for
2660 details of the semantics of attributes applying to labels and
2663 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
2664 An @dfn{attribute specifier} is of the form
2665 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2666 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2667 each attribute is one of the following:
2671 Empty. Empty attributes are ignored.
2674 A word (which may be an identifier such as @code{unused}, or a reserved
2675 word such as @code{const}).
2678 A word, followed by, in parentheses, parameters for the attribute.
2679 These parameters take one of the following forms:
2683 An identifier. For example, @code{mode} attributes use this form.
2686 An identifier followed by a comma and a non-empty comma-separated list
2687 of expressions. For example, @code{format} attributes use this form.
2690 A possibly empty comma-separated list of expressions. For example,
2691 @code{format_arg} attributes use this form with the list being a single
2692 integer constant expression, and @code{alias} attributes use this form
2693 with the list being a single string constant.
2697 An @dfn{attribute specifier list} is a sequence of one or more attribute
2698 specifiers, not separated by any other tokens.
2700 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
2701 In GNU C, an attribute specifier list may appear after the colon
2702 following a label, other than a @code{case} or @code{default} label.
2703 GNU C++ does not permit such placement of attribute lists, as it is
2704 permissible for a declaration, which could begin with an attribute
2705 list, to be labelled in C++. Declarations cannot be labelled in C90
2706 or C99, so the ambiguity does not arise there.
2708 In GNU C an attribute specifier list may also appear after the keyword
2709 @code{while} in a while loop, after @code{do} and after @code{for}.
2711 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
2712 An attribute specifier list may appear as part of a @code{struct},
2713 @code{union} or @code{enum} specifier. It may go either immediately
2714 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2715 the closing brace. The former syntax is preferred.
2716 Where attribute specifiers follow the closing brace, they are considered
2717 to relate to the structure, union or enumerated type defined, not to any
2718 enclosing declaration the type specifier appears in, and the type
2719 defined is not complete until after the attribute specifiers.
2720 @c Otherwise, there would be the following problems: a shift/reduce
2721 @c conflict between attributes binding the struct/union/enum and
2722 @c binding to the list of specifiers/qualifiers; and "aligned"
2723 @c attributes could use sizeof for the structure, but the size could be
2724 @c changed later by "packed" attributes.
2726 Otherwise, an attribute specifier appears as part of a declaration,
2727 counting declarations of unnamed parameters and type names, and relates
2728 to that declaration (which may be nested in another declaration, for
2729 example in the case of a parameter declaration), or to a particular declarator
2730 within a declaration. Where an
2731 attribute specifier is applied to a parameter declared as a function or
2732 an array, it should apply to the function or array rather than the
2733 pointer to which the parameter is implicitly converted, but this is not
2734 yet correctly implemented.
2736 Any list of specifiers and qualifiers at the start of a declaration may
2737 contain attribute specifiers, whether or not such a list may in that
2738 context contain storage class specifiers. (Some attributes, however,
2739 are essentially in the nature of storage class specifiers, and only make
2740 sense where storage class specifiers may be used; for example,
2741 @code{section}.) There is one necessary limitation to this syntax: the
2742 first old-style parameter declaration in a function definition cannot
2743 begin with an attribute specifier, because such an attribute applies to
2744 the function instead by syntax described below (which, however, is not
2745 yet implemented in this case). In some other cases, attribute
2746 specifiers are permitted by this grammar but not yet supported by the
2747 compiler. All attribute specifiers in this place relate to the
2748 declaration as a whole. In the obsolescent usage where a type of
2749 @code{int} is implied by the absence of type specifiers, such a list of
2750 specifiers and qualifiers may be an attribute specifier list with no
2751 other specifiers or qualifiers.
2753 At present, the first parameter in a function prototype must have some
2754 type specifier which is not an attribute specifier; this resolves an
2755 ambiguity in the interpretation of @code{void f(int
2756 (__attribute__((foo)) x))}, but is subject to change. At present, if
2757 the parentheses of a function declarator contain only attributes then
2758 those attributes are ignored, rather than yielding an error or warning
2759 or implying a single parameter of type int, but this is subject to
2762 An attribute specifier list may appear immediately before a declarator
2763 (other than the first) in a comma-separated list of declarators in a
2764 declaration of more than one identifier using a single list of
2765 specifiers and qualifiers. Such attribute specifiers apply
2766 only to the identifier before whose declarator they appear. For
2770 __attribute__((noreturn)) void d0 (void),
2771 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2776 the @code{noreturn} attribute applies to all the functions
2777 declared; the @code{format} attribute only applies to @code{d1}.
2779 An attribute specifier list may appear immediately before the comma,
2780 @code{=} or semicolon terminating the declaration of an identifier other
2781 than a function definition. At present, such attribute specifiers apply
2782 to the declared object or function, but in future they may attach to the
2783 outermost adjacent declarator. In simple cases there is no difference,
2784 but, for example, in
2787 void (****f)(void) __attribute__((noreturn));
2791 at present the @code{noreturn} attribute applies to @code{f}, which
2792 causes a warning since @code{f} is not a function, but in future it may
2793 apply to the function @code{****f}. The precise semantics of what
2794 attributes in such cases will apply to are not yet specified. Where an
2795 assembler name for an object or function is specified (@pxref{Asm
2796 Labels}), at present the attribute must follow the @code{asm}
2797 specification; in future, attributes before the @code{asm} specification
2798 may apply to the adjacent declarator, and those after it to the declared
2801 An attribute specifier list may, in future, be permitted to appear after
2802 the declarator in a function definition (before any old-style parameter
2803 declarations or the function body).
2805 Attribute specifiers may be mixed with type qualifiers appearing inside
2806 the @code{[]} of a parameter array declarator, in the C99 construct by
2807 which such qualifiers are applied to the pointer to which the array is
2808 implicitly converted. Such attribute specifiers apply to the pointer,
2809 not to the array, but at present this is not implemented and they are
2812 An attribute specifier list may appear at the start of a nested
2813 declarator. At present, there are some limitations in this usage: the
2814 attributes correctly apply to the declarator, but for most individual
2815 attributes the semantics this implies are not implemented.
2816 When attribute specifiers follow the @code{*} of a pointer
2817 declarator, they may be mixed with any type qualifiers present.
2818 The following describes the formal semantics of this syntax. It will make the
2819 most sense if you are familiar with the formal specification of
2820 declarators in the ISO C standard.
2822 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2823 D1}, where @code{T} contains declaration specifiers that specify a type
2824 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2825 contains an identifier @var{ident}. The type specified for @var{ident}
2826 for derived declarators whose type does not include an attribute
2827 specifier is as in the ISO C standard.
2829 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2830 and the declaration @code{T D} specifies the type
2831 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2832 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2833 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2835 If @code{D1} has the form @code{*
2836 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2837 declaration @code{T D} specifies the type
2838 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2839 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2840 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2846 void (__attribute__((noreturn)) ****f) (void);
2850 specifies the type ``pointer to pointer to pointer to pointer to
2851 non-returning function returning @code{void}''. As another example,
2854 char *__attribute__((aligned(8))) *f;
2858 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2859 Note again that this does not work with most attributes; for example,
2860 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2861 is not yet supported.
2863 For compatibility with existing code written for compiler versions that
2864 did not implement attributes on nested declarators, some laxity is
2865 allowed in the placing of attributes. If an attribute that only applies
2866 to types is applied to a declaration, it will be treated as applying to
2867 the type of that declaration. If an attribute that only applies to
2868 declarations is applied to the type of a declaration, it will be treated
2869 as applying to that declaration; and, for compatibility with code
2870 placing the attributes immediately before the identifier declared, such
2871 an attribute applied to a function return type will be treated as
2872 applying to the function type, and such an attribute applied to an array
2873 element type will be treated as applying to the array type. If an
2874 attribute that only applies to function types is applied to a
2875 pointer-to-function type, it will be treated as applying to the pointer
2876 target type; if such an attribute is applied to a function return type
2877 that is not a pointer-to-function type, it will be treated as applying
2878 to the function type.
2880 @node Function Prototypes
2881 @section Prototypes and Old-Style Function Definitions
2882 @cindex function prototype declarations
2883 @cindex old-style function definitions
2884 @cindex promotion of formal parameters
2886 GNU C extends ISO C to allow a function prototype to override a later
2887 old-style non-prototype definition. Consider the following example:
2890 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2897 /* @r{Prototype function declaration.} */
2898 int isroot P((uid_t));
2900 /* @r{Old-style function definition.} */
2902 isroot (x) /* @r{??? lossage here ???} */
2909 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2910 not allow this example, because subword arguments in old-style
2911 non-prototype definitions are promoted. Therefore in this example the
2912 function definition's argument is really an @code{int}, which does not
2913 match the prototype argument type of @code{short}.
2915 This restriction of ISO C makes it hard to write code that is portable
2916 to traditional C compilers, because the programmer does not know
2917 whether the @code{uid_t} type is @code{short}, @code{int}, or
2918 @code{long}. Therefore, in cases like these GNU C allows a prototype
2919 to override a later old-style definition. More precisely, in GNU C, a
2920 function prototype argument type overrides the argument type specified
2921 by a later old-style definition if the former type is the same as the
2922 latter type before promotion. Thus in GNU C the above example is
2923 equivalent to the following:
2936 GNU C++ does not support old-style function definitions, so this
2937 extension is irrelevant.
2940 @section C++ Style Comments
2942 @cindex C++ comments
2943 @cindex comments, C++ style
2945 In GNU C, you may use C++ style comments, which start with @samp{//} and
2946 continue until the end of the line. Many other C implementations allow
2947 such comments, and they are included in the 1999 C standard. However,
2948 C++ style comments are not recognized if you specify an @option{-std}
2949 option specifying a version of ISO C before C99, or @option{-ansi}
2950 (equivalent to @option{-std=c89}).
2953 @section Dollar Signs in Identifier Names
2955 @cindex dollar signs in identifier names
2956 @cindex identifier names, dollar signs in
2958 In GNU C, you may normally use dollar signs in identifier names.
2959 This is because many traditional C implementations allow such identifiers.
2960 However, dollar signs in identifiers are not supported on a few target
2961 machines, typically because the target assembler does not allow them.
2963 @node Character Escapes
2964 @section The Character @key{ESC} in Constants
2966 You can use the sequence @samp{\e} in a string or character constant to
2967 stand for the ASCII character @key{ESC}.
2970 @section Inquiring on Alignment of Types or Variables
2972 @cindex type alignment
2973 @cindex variable alignment
2975 The keyword @code{__alignof__} allows you to inquire about how an object
2976 is aligned, or the minimum alignment usually required by a type. Its
2977 syntax is just like @code{sizeof}.
2979 For example, if the target machine requires a @code{double} value to be
2980 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2981 This is true on many RISC machines. On more traditional machine
2982 designs, @code{__alignof__ (double)} is 4 or even 2.
2984 Some machines never actually require alignment; they allow reference to any
2985 data type even at an odd address. For these machines, @code{__alignof__}
2986 reports the @emph{recommended} alignment of a type.
2988 If the operand of @code{__alignof__} is an lvalue rather than a type,
2989 its value is the required alignment for its type, taking into account
2990 any minimum alignment specified with GCC's @code{__attribute__}
2991 extension (@pxref{Variable Attributes}). For example, after this
2995 struct foo @{ int x; char y; @} foo1;
2999 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3000 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3002 It is an error to ask for the alignment of an incomplete type.
3004 @node Variable Attributes
3005 @section Specifying Attributes of Variables
3006 @cindex attribute of variables
3007 @cindex variable attributes
3009 The keyword @code{__attribute__} allows you to specify special
3010 attributes of variables or structure fields. This keyword is followed
3011 by an attribute specification inside double parentheses. Some
3012 attributes are currently defined generically for variables.
3013 Other attributes are defined for variables on particular target
3014 systems. Other attributes are available for functions
3015 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
3016 (@pxref{Function Attributes}), types (@pxref{Type Attributes}) and
3017 labels (@pxref{Label Attributes}). Other front ends might define
3018 more attributes (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3020 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
3021 You may also specify attributes with @samp{__} preceding and following
3022 each keyword. This allows you to use them in header files without
3023 being concerned about a possible macro of the same name. For example,
3024 you may use @code{__aligned__} instead of @code{aligned}.
3026 @xref{Attribute Syntax}, for details of the exact syntax for using
3030 @cindex @code{aligned} attribute
3031 @item aligned (@var{alignment})
3032 This attribute specifies a minimum alignment for the variable or
3033 structure field, measured in bytes. For example, the declaration:
3036 int x __attribute__ ((aligned (16))) = 0;
3040 causes the compiler to allocate the global variable @code{x} on a
3041 16-byte boundary. On a 68040, this could be used in conjunction with
3042 an @code{asm} expression to access the @code{move16} instruction which
3043 requires 16-byte aligned operands.
3045 You can also specify the alignment of structure fields. For example, to
3046 create a double-word aligned @code{int} pair, you could write:
3049 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3053 This is an alternative to creating a union with a @code{double} member
3054 that forces the union to be double-word aligned.
3056 As in the preceding examples, you can explicitly specify the alignment
3057 (in bytes) that you wish the compiler to use for a given variable or
3058 structure field. Alternatively, you can leave out the alignment factor
3059 and just ask the compiler to align a variable or field to the maximum
3060 useful alignment for the target machine you are compiling for. For
3061 example, you could write:
3064 short array[3] __attribute__ ((aligned));
3067 Whenever you leave out the alignment factor in an @code{aligned} attribute
3068 specification, the compiler automatically sets the alignment for the declared
3069 variable or field to the largest alignment which is ever used for any data
3070 type on the target machine you are compiling for. Doing this can often make
3071 copy operations more efficient, because the compiler can use whatever
3072 instructions copy the biggest chunks of memory when performing copies to
3073 or from the variables or fields that you have aligned this way.
3075 The @code{aligned} attribute can only increase the alignment; but you
3076 can decrease it by specifying @code{packed} as well. See below.
3078 Note that the effectiveness of @code{aligned} attributes may be limited
3079 by inherent limitations in your linker. On many systems, the linker is
3080 only able to arrange for variables to be aligned up to a certain maximum
3081 alignment. (For some linkers, the maximum supported alignment may
3082 be very very small.) If your linker is only able to align variables
3083 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3084 in an @code{__attribute__} will still only provide you with 8 byte
3085 alignment. See your linker documentation for further information.
3087 The @code{aligned} attribute can also be used for functions
3088 (@pxref{Function Attributes}.)
3090 @item cleanup (@var{cleanup_function})
3091 @cindex @code{cleanup} attribute
3092 The @code{cleanup} attribute runs a function when the variable goes
3093 out of scope. This attribute can only be applied to auto function
3094 scope variables; it may not be applied to parameters or variables
3095 with static storage duration. The function must take one parameter,
3096 a pointer to a type compatible with the variable. The return value
3097 of the function (if any) is ignored.
3099 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3100 will be run during the stack unwinding that happens during the
3101 processing of the exception. Note that the @code{cleanup} attribute
3102 does not allow the exception to be caught, only to perform an action.
3103 It is undefined what happens if @var{cleanup_function} does not
3108 @cindex @code{common} attribute
3109 @cindex @code{nocommon} attribute
3112 The @code{common} attribute requests GCC to place a variable in
3113 ``common'' storage. The @code{nocommon} attribute requests the
3114 opposite---to allocate space for it directly.
3116 These attributes override the default chosen by the
3117 @option{-fno-common} and @option{-fcommon} flags respectively.
3120 @cindex @code{deprecated} attribute
3121 The @code{deprecated} attribute results in a warning if the variable
3122 is used anywhere in the source file. This is useful when identifying
3123 variables that are expected to be removed in a future version of a
3124 program. The warning also includes the location of the declaration
3125 of the deprecated variable, to enable users to easily find further
3126 information about why the variable is deprecated, or what they should
3127 do instead. Note that the warning only occurs for uses:
3130 extern int old_var __attribute__ ((deprecated));
3132 int new_fn () @{ return old_var; @}
3135 results in a warning on line 3 but not line 2.
3137 The @code{deprecated} attribute can also be used for functions and
3138 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3140 @item mode (@var{mode})
3141 @cindex @code{mode} attribute
3142 This attribute specifies the data type for the declaration---whichever
3143 type corresponds to the mode @var{mode}. This in effect lets you
3144 request an integer or floating point type according to its width.
3146 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3147 indicate the mode corresponding to a one-byte integer, @samp{word} or
3148 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3149 or @samp{__pointer__} for the mode used to represent pointers.
3152 @cindex @code{packed} attribute
3153 The @code{packed} attribute specifies that a variable or structure field
3154 should have the smallest possible alignment---one byte for a variable,
3155 and one bit for a field, unless you specify a larger value with the
3156 @code{aligned} attribute.
3158 Here is a structure in which the field @code{x} is packed, so that it
3159 immediately follows @code{a}:
3165 int x[2] __attribute__ ((packed));
3169 @item section ("@var{section-name}")
3170 @cindex @code{section} variable attribute
3171 Normally, the compiler places the objects it generates in sections like
3172 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3173 or you need certain particular variables to appear in special sections,
3174 for example to map to special hardware. The @code{section}
3175 attribute specifies that a variable (or function) lives in a particular
3176 section. For example, this small program uses several specific section names:
3179 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3180 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3181 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3182 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3186 /* @r{Initialize stack pointer} */
3187 init_sp (stack + sizeof (stack));
3189 /* @r{Initialize initialized data} */
3190 memcpy (&init_data, &data, &edata - &data);
3192 /* @r{Turn on the serial ports} */
3199 Use the @code{section} attribute with an @emph{initialized} definition
3200 of a @emph{global} variable, as shown in the example. GCC issues
3201 a warning and otherwise ignores the @code{section} attribute in
3202 uninitialized variable declarations.
3204 You may only use the @code{section} attribute with a fully initialized
3205 global definition because of the way linkers work. The linker requires
3206 each object be defined once, with the exception that uninitialized
3207 variables tentatively go in the @code{common} (or @code{bss}) section
3208 and can be multiply ``defined''. You can force a variable to be
3209 initialized with the @option{-fno-common} flag or the @code{nocommon}
3212 Some file formats do not support arbitrary sections so the @code{section}
3213 attribute is not available on all platforms.
3214 If you need to map the entire contents of a module to a particular
3215 section, consider using the facilities of the linker instead.
3218 @cindex @code{shared} variable attribute
3219 On Microsoft Windows, in addition to putting variable definitions in a named
3220 section, the section can also be shared among all running copies of an
3221 executable or DLL@. For example, this small program defines shared data
3222 by putting it in a named section @code{shared} and marking the section
3226 int foo __attribute__((section ("shared"), shared)) = 0;
3231 /* @r{Read and write foo. All running
3232 copies see the same value.} */
3238 You may only use the @code{shared} attribute along with @code{section}
3239 attribute with a fully initialized global definition because of the way
3240 linkers work. See @code{section} attribute for more information.
3242 The @code{shared} attribute is only available on Microsoft Windows@.
3244 @item tls_model ("@var{tls_model}")
3245 @cindex @code{tls_model} attribute
3246 The @code{tls_model} attribute sets thread-local storage model
3247 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3248 overriding @option{-ftls-model=} command line switch on a per-variable
3250 The @var{tls_model} argument should be one of @code{global-dynamic},
3251 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3253 Not all targets support this attribute.
3256 This attribute, attached to a variable, means that the variable is meant
3257 to be possibly unused. GCC will not produce a warning for this
3261 This attribute, attached to a variable, means that the variable must be
3262 emitted even if it appears that the variable is not referenced.
3264 @item vector_size (@var{bytes})
3265 This attribute specifies the vector size for the variable, measured in
3266 bytes. For example, the declaration:
3269 int foo __attribute__ ((vector_size (16)));
3273 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3274 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3275 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3277 This attribute is only applicable to integral and float scalars,
3278 although arrays, pointers, and function return values are allowed in
3279 conjunction with this construct.
3281 Aggregates with this attribute are invalid, even if they are of the same
3282 size as a corresponding scalar. For example, the declaration:
3285 struct S @{ int a; @};
3286 struct S __attribute__ ((vector_size (16))) foo;
3290 is invalid even if the size of the structure is the same as the size of
3294 The @code{selectany} attribute causes an initialized global variable to
3295 have link-once semantics. When multiple definitions of the variable are
3296 encountered by the linker, the first is selected and the remainder are
3297 discarded. Following usage by the Microsoft compiler, the linker is told
3298 @emph{not} to warn about size or content differences of the multiple
3301 Although the primary usage of this attribute is for POD types, the
3302 attribute can also be applied to global C++ objects that are initialized
3303 by a constructor. In this case, the static initialization and destruction
3304 code for the object is emitted in each translation defining the object,
3305 but the calls to the constructor and destructor are protected by a
3306 link-once guard variable.
3308 The @code{selectany} attribute is only available on Microsoft Windows
3309 targets. You can use @code{__declspec (selectany)} as a synonym for
3310 @code{__attribute__ ((selectany))} for compatibility with other
3314 The @code{weak} attribute is described in @xref{Function Attributes}.
3317 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3320 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3324 @subsection M32R/D Variable Attributes
3326 One attribute is currently defined for the M32R/D@.
3329 @item model (@var{model-name})
3330 @cindex variable addressability on the M32R/D
3331 Use this attribute on the M32R/D to set the addressability of an object.
3332 The identifier @var{model-name} is one of @code{small}, @code{medium},
3333 or @code{large}, representing each of the code models.
3335 Small model objects live in the lower 16MB of memory (so that their
3336 addresses can be loaded with the @code{ld24} instruction).
3338 Medium and large model objects may live anywhere in the 32-bit address space
3339 (the compiler will generate @code{seth/add3} instructions to load their
3343 @anchor{i386 Variable Attributes}
3344 @subsection i386 Variable Attributes
3346 Two attributes are currently defined for i386 configurations:
3347 @code{ms_struct} and @code{gcc_struct}
3352 @cindex @code{ms_struct} attribute
3353 @cindex @code{gcc_struct} attribute
3355 If @code{packed} is used on a structure, or if bit-fields are used
3356 it may be that the Microsoft ABI packs them differently
3357 than GCC would normally pack them. Particularly when moving packed
3358 data between functions compiled with GCC and the native Microsoft compiler
3359 (either via function call or as data in a file), it may be necessary to access
3362 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3363 compilers to match the native Microsoft compiler.
3365 The Microsoft structure layout algorithm is fairly simple with the exception
3366 of the bitfield packing:
3368 The padding and alignment of members of structures and whether a bit field
3369 can straddle a storage-unit boundary
3372 @item Structure members are stored sequentially in the order in which they are
3373 declared: the first member has the lowest memory address and the last member
3376 @item Every data object has an alignment-requirement. The alignment-requirement
3377 for all data except structures, unions, and arrays is either the size of the
3378 object or the current packing size (specified with either the aligned attribute
3379 or the pack pragma), whichever is less. For structures, unions, and arrays,
3380 the alignment-requirement is the largest alignment-requirement of its members.
3381 Every object is allocated an offset so that:
3383 offset % alignment-requirement == 0
3385 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3386 unit if the integral types are the same size and if the next bit field fits
3387 into the current allocation unit without crossing the boundary imposed by the
3388 common alignment requirements of the bit fields.
3391 Handling of zero-length bitfields:
3393 MSVC interprets zero-length bitfields in the following ways:
3396 @item If a zero-length bitfield is inserted between two bitfields that would
3397 normally be coalesced, the bitfields will not be coalesced.
3404 unsigned long bf_1 : 12;
3406 unsigned long bf_2 : 12;
3410 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3411 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3413 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3414 alignment of the zero-length bitfield is greater than the member that follows it,
3415 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3435 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3436 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3437 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3440 Taking this into account, it is important to note the following:
3443 @item If a zero-length bitfield follows a normal bitfield, the type of the
3444 zero-length bitfield may affect the alignment of the structure as whole. For
3445 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3446 normal bitfield, and is of type short.
3448 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3449 still affect the alignment of the structure:
3459 Here, @code{t4} will take up 4 bytes.
3462 @item Zero-length bitfields following non-bitfield members are ignored:
3473 Here, @code{t5} will take up 2 bytes.
3477 @subsection PowerPC Variable Attributes
3479 Three attributes currently are defined for PowerPC configurations:
3480 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3482 For full documentation of the struct attributes please see the
3483 documentation in the @xref{i386 Variable Attributes}, section.
3485 For documentation of @code{altivec} attribute please see the
3486 documentation in the @xref{PowerPC Type Attributes}, section.
3488 @subsection Xstormy16 Variable Attributes
3490 One attribute is currently defined for xstormy16 configurations:
3495 @cindex @code{below100} attribute
3497 If a variable has the @code{below100} attribute (@code{BELOW100} is
3498 allowed also), GCC will place the variable in the first 0x100 bytes of
3499 memory and use special opcodes to access it. Such variables will be
3500 placed in either the @code{.bss_below100} section or the
3501 @code{.data_below100} section.
3505 @node Type Attributes
3506 @section Specifying Attributes of Types
3507 @cindex attribute of types
3508 @cindex type attributes
3510 The keyword @code{__attribute__} allows you to specify special
3511 attributes of @code{struct} and @code{union} types when you define
3512 such types. This keyword is followed by an attribute specification
3513 inside double parentheses. Seven attributes are currently defined for
3514 types: @code{aligned}, @code{packed}, @code{transparent_union},
3515 @code{unused}, @code{deprecated}, @code{visibility}, and
3516 @code{may_alias}. Other attributes are defined for functions
3517 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
3518 (@pxref{Function Attributes}), variables (@pxref{Variable
3519 Attributes}), and labels (@pxref{Label Attributes}).
3521 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
3522 You may also specify any one of these attributes with @samp{__}
3523 preceding and following its keyword. This allows you to use these
3524 attributes in header files without being concerned about a possible
3525 macro of the same name. For example, you may use @code{__aligned__}
3526 instead of @code{aligned}.
3528 You may specify type attributes either in a @code{typedef} declaration
3529 or in an enum, struct or union type declaration or definition.
3531 For an enum, struct or union type, you may specify attributes either
3532 between the enum, struct or union tag and the name of the type, or
3533 just past the closing curly brace of the @emph{definition}. The
3534 former syntax is preferred.
3536 @xref{Attribute Syntax}, for details of the exact syntax for using
3540 @cindex @code{aligned} attribute
3541 @item aligned (@var{alignment})
3542 This attribute specifies a minimum alignment (in bytes) for variables
3543 of the specified type. For example, the declarations:
3546 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3547 typedef int more_aligned_int __attribute__ ((aligned (8)));
3551 force the compiler to insure (as far as it can) that each variable whose
3552 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3553 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3554 variables of type @code{struct S} aligned to 8-byte boundaries allows
3555 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3556 store) instructions when copying one variable of type @code{struct S} to
3557 another, thus improving run-time efficiency.
3559 Note that the alignment of any given @code{struct} or @code{union} type
3560 is required by the ISO C standard to be at least a perfect multiple of
3561 the lowest common multiple of the alignments of all of the members of
3562 the @code{struct} or @code{union} in question. This means that you @emph{can}
3563 effectively adjust the alignment of a @code{struct} or @code{union}
3564 type by attaching an @code{aligned} attribute to any one of the members
3565 of such a type, but the notation illustrated in the example above is a
3566 more obvious, intuitive, and readable way to request the compiler to
3567 adjust the alignment of an entire @code{struct} or @code{union} type.
3569 As in the preceding example, you can explicitly specify the alignment
3570 (in bytes) that you wish the compiler to use for a given @code{struct}
3571 or @code{union} type. Alternatively, you can leave out the alignment factor
3572 and just ask the compiler to align a type to the maximum
3573 useful alignment for the target machine you are compiling for. For
3574 example, you could write:
3577 struct S @{ short f[3]; @} __attribute__ ((aligned));
3580 Whenever you leave out the alignment factor in an @code{aligned}
3581 attribute specification, the compiler automatically sets the alignment
3582 for the type to the largest alignment which is ever used for any data
3583 type on the target machine you are compiling for. Doing this can often
3584 make copy operations more efficient, because the compiler can use
3585 whatever instructions copy the biggest chunks of memory when performing
3586 copies to or from the variables which have types that you have aligned
3589 In the example above, if the size of each @code{short} is 2 bytes, then
3590 the size of the entire @code{struct S} type is 6 bytes. The smallest
3591 power of two which is greater than or equal to that is 8, so the
3592 compiler sets the alignment for the entire @code{struct S} type to 8
3595 Note that although you can ask the compiler to select a time-efficient
3596 alignment for a given type and then declare only individual stand-alone
3597 objects of that type, the compiler's ability to select a time-efficient
3598 alignment is primarily useful only when you plan to create arrays of
3599 variables having the relevant (efficiently aligned) type. If you
3600 declare or use arrays of variables of an efficiently-aligned type, then
3601 it is likely that your program will also be doing pointer arithmetic (or
3602 subscripting, which amounts to the same thing) on pointers to the
3603 relevant type, and the code that the compiler generates for these
3604 pointer arithmetic operations will often be more efficient for
3605 efficiently-aligned types than for other types.
3607 The @code{aligned} attribute can only increase the alignment; but you
3608 can decrease it by specifying @code{packed} as well. See below.
3610 Note that the effectiveness of @code{aligned} attributes may be limited
3611 by inherent limitations in your linker. On many systems, the linker is
3612 only able to arrange for variables to be aligned up to a certain maximum
3613 alignment. (For some linkers, the maximum supported alignment may
3614 be very very small.) If your linker is only able to align variables
3615 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3616 in an @code{__attribute__} will still only provide you with 8 byte
3617 alignment. See your linker documentation for further information.
3620 This attribute, attached to @code{struct} or @code{union} type
3621 definition, specifies that each member (other than zero-width bitfields)
3622 of the structure or union is placed to minimize the memory required. When
3623 attached to an @code{enum} definition, it indicates that the smallest
3624 integral type should be used.
3626 @opindex fshort-enums
3627 Specifying this attribute for @code{struct} and @code{union} types is
3628 equivalent to specifying the @code{packed} attribute on each of the
3629 structure or union members. Specifying the @option{-fshort-enums}
3630 flag on the line is equivalent to specifying the @code{packed}
3631 attribute on all @code{enum} definitions.
3633 In the following example @code{struct my_packed_struct}'s members are
3634 packed closely together, but the internal layout of its @code{s} member
3635 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3639 struct my_unpacked_struct
3645 struct __attribute__ ((__packed__)) my_packed_struct
3649 struct my_unpacked_struct s;
3653 You may only specify this attribute on the definition of a @code{enum},
3654 @code{struct} or @code{union}, not on a @code{typedef} which does not
3655 also define the enumerated type, structure or union.
3657 @item transparent_union
3658 This attribute, attached to a @code{union} type definition, indicates
3659 that any function parameter having that union type causes calls to that
3660 function to be treated in a special way.
3662 First, the argument corresponding to a transparent union type can be of
3663 any type in the union; no cast is required. Also, if the union contains
3664 a pointer type, the corresponding argument can be a null pointer
3665 constant or a void pointer expression; and if the union contains a void
3666 pointer type, the corresponding argument can be any pointer expression.
3667 If the union member type is a pointer, qualifiers like @code{const} on
3668 the referenced type must be respected, just as with normal pointer
3671 Second, the argument is passed to the function using the calling
3672 conventions of the first member of the transparent union, not the calling
3673 conventions of the union itself. All members of the union must have the
3674 same machine representation; this is necessary for this argument passing
3677 Transparent unions are designed for library functions that have multiple
3678 interfaces for compatibility reasons. For example, suppose the
3679 @code{wait} function must accept either a value of type @code{int *} to
3680 comply with Posix, or a value of type @code{union wait *} to comply with
3681 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3682 @code{wait} would accept both kinds of arguments, but it would also
3683 accept any other pointer type and this would make argument type checking
3684 less useful. Instead, @code{<sys/wait.h>} might define the interface
3692 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3694 pid_t wait (wait_status_ptr_t);
3697 This interface allows either @code{int *} or @code{union wait *}
3698 arguments to be passed, using the @code{int *} calling convention.
3699 The program can call @code{wait} with arguments of either type:
3702 int w1 () @{ int w; return wait (&w); @}
3703 int w2 () @{ union wait w; return wait (&w); @}
3706 With this interface, @code{wait}'s implementation might look like this:
3709 pid_t wait (wait_status_ptr_t p)
3711 return waitpid (-1, p.__ip, 0);
3716 When attached to a type (including a @code{union} or a @code{struct}),
3717 this attribute means that variables of that type are meant to appear
3718 possibly unused. GCC will not produce a warning for any variables of
3719 that type, even if the variable appears to do nothing. This is often
3720 the case with lock or thread classes, which are usually defined and then
3721 not referenced, but contain constructors and destructors that have
3722 nontrivial bookkeeping functions.
3725 The @code{deprecated} attribute results in a warning if the type
3726 is used anywhere in the source file. This is useful when identifying
3727 types that are expected to be removed in a future version of a program.
3728 If possible, the warning also includes the location of the declaration
3729 of the deprecated type, to enable users to easily find further
3730 information about why the type is deprecated, or what they should do
3731 instead. Note that the warnings only occur for uses and then only
3732 if the type is being applied to an identifier that itself is not being
3733 declared as deprecated.
3736 typedef int T1 __attribute__ ((deprecated));
3740 typedef T1 T3 __attribute__ ((deprecated));
3741 T3 z __attribute__ ((deprecated));
3744 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3745 warning is issued for line 4 because T2 is not explicitly
3746 deprecated. Line 5 has no warning because T3 is explicitly
3747 deprecated. Similarly for line 6.
3749 The @code{deprecated} attribute can also be used for functions and
3750 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3753 Accesses to objects with types with this attribute are not subjected to
3754 type-based alias analysis, but are instead assumed to be able to alias
3755 any other type of objects, just like the @code{char} type. See
3756 @option{-fstrict-aliasing} for more information on aliasing issues.
3761 typedef short __attribute__((__may_alias__)) short_a;
3767 short_a *b = (short_a *) &a;
3771 if (a == 0x12345678)
3778 If you replaced @code{short_a} with @code{short} in the variable
3779 declaration, the above program would abort when compiled with
3780 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3781 above in recent GCC versions.
3784 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3785 applied to class, struct, union and enum types. Unlike other type
3786 attributes, the attribute must appear between the initial keyword and
3787 the name of the type; it cannot appear after the body of the type.
3789 Note that the type visibility is applied to vague linkage entities
3790 associated with the class (vtable, typeinfo node, etc.). In
3791 particular, if a class is thrown as an exception in one shared object
3792 and caught in another, the class must have default visibility.
3793 Otherwise the two shared objects will be unable to use the same
3794 typeinfo node and exception handling will break.
3796 @c APPLE LOCAL begin weak types 5954418
3798 In C++, attribute weak can be applied to a class to ensure that all
3799 non-hidden instances of the type are treated as the same type across
3800 shared library boundaries on platforms (such as darwin and arm aapcs)
3801 that can emit vtables and the type info meta data as non-comdat
3802 symbols. This is useful when the class has a key method and the
3803 translation unit that contains the key method is used in more than one
3804 shared library or in a shared library and the application. Doing this
3805 results in more expensive startup times. This attribute is inherited
3806 by subclasses, so it is only necessary to mark a base type. The
3807 typical use would be to mark any types used for throwing across shared
3808 library boundaries or those used in dynamic_cast operations across a
3809 shared library boundary.
3810 @c APPLE LOCAL end weak types 5954418
3812 @subsection ARM Type Attributes
3814 On those ARM targets that support @code{dllimport} (such as Symbian
3815 OS), you can use the @code{notshared} attribute to indicate that the
3816 virtual table and other similar data for a class should not be
3817 exported from a DLL@. For example:
3820 class __declspec(notshared) C @{
3822 __declspec(dllimport) C();
3826 __declspec(dllexport)
3830 In this code, @code{C::C} is exported from the current DLL, but the
3831 virtual table for @code{C} is not exported. (You can use
3832 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3833 most Symbian OS code uses @code{__declspec}.)
3835 @anchor{i386 Type Attributes}
3836 @subsection i386 Type Attributes
3838 Two attributes are currently defined for i386 configurations:
3839 @code{ms_struct} and @code{gcc_struct}
3843 @cindex @code{ms_struct}
3844 @cindex @code{gcc_struct}
3846 If @code{packed} is used on a structure, or if bit-fields are used
3847 it may be that the Microsoft ABI packs them differently
3848 than GCC would normally pack them. Particularly when moving packed
3849 data between functions compiled with GCC and the native Microsoft compiler
3850 (either via function call or as data in a file), it may be necessary to access
3853 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3854 compilers to match the native Microsoft compiler.
3857 To specify multiple attributes, separate them by commas within the
3858 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3861 @anchor{PowerPC Type Attributes}
3862 @subsection PowerPC Type Attributes
3864 Three attributes currently are defined for PowerPC configurations:
3865 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3867 For full documentation of the struct attributes please see the
3868 documentation in the @xref{i386 Type Attributes}, section.
3870 The @code{altivec} attribute allows one to declare AltiVec vector data
3871 types supported by the AltiVec Programming Interface Manual. The
3872 attribute requires an argument to specify one of three vector types:
3873 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3874 and @code{bool__} (always followed by unsigned).
3877 __attribute__((altivec(vector__)))
3878 __attribute__((altivec(pixel__))) unsigned short
3879 __attribute__((altivec(bool__))) unsigned
3882 These attributes mainly are intended to support the @code{__vector},
3883 @code{__pixel}, and @code{__bool} AltiVec keywords.
3885 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
3886 @node Label Attributes
3887 @section Specifying Attributes of Labels and Statements
3888 @cindex attribute of labels
3889 @cindex label attributes
3890 @cindex attribute of statements
3891 @cindex statement attributes
3893 The keyword @code{__attribute__} allows you to specify special
3894 attributes of labels and statements.
3896 Some attributes are currently defined generically for variables.
3897 Other attributes are defined for variables on particular target
3898 systems. Other attributes are available for functions
3899 (@pxref{Function Attributes}), types (@pxref{Type Attributes}) and
3900 variables (@pxref{Variable Attributes}).
3902 You may also specify attributes with @samp{__} preceding and following
3903 each keyword. This allows you to use them in header files without
3904 being concerned about a possible macro of the same name. For example,
3905 you may use @code{__aligned__} instead of @code{aligned}.
3907 @xref{Attribute Syntax}, for details of the exact syntax for using
3911 @cindex @code{aligned} attribute
3912 @item aligned (@var{alignment})
3913 This attribute specifies a minimum alignment for the label,
3914 measured in bytes. For example, the declaration:
3917 some_label: __attribute__((aligned(16)))
3921 requests the compiler to align the label, inserting @code{nop}s as necessary,
3922 to a 16-byte boundary.
3924 The alignment is only a request. The compiler will usually be able to
3925 honour it but sometimes the label will be eliminated by the compiler,
3926 in which case its alignment will be eliminated too.
3928 When applied to loops, the @code{aligned} attribute causes the loop to
3932 When attached to a label this attribute means that the label might not
3933 be used. GCC will not produce a warning for the label, even if the
3934 label doesn't seem to be referenced. This feature is intended for
3935 code generated by programs which contains labels that may be unused
3936 but which is compiled with @option{-Wall}. It would not normally be
3937 appropriate to use in it human-written code, though it could be useful
3938 in cases where the code that jumps to the label is contained within an
3939 @code{#ifdef} conditional.
3941 This attribute can only be applied to labels, not statements, because
3942 there is no warning if a statement is removed.
3945 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
3947 @section An Inline Function is As Fast As a Macro
3948 @cindex inline functions
3949 @cindex integrating function code
3951 @cindex macros, inline alternative
3953 By declaring a function inline, you can direct GCC to make
3954 calls to that function faster. One way GCC can achieve this is to
3955 integrate that function's code into the code for its callers. This
3956 makes execution faster by eliminating the function-call overhead; in
3957 addition, if any of the actual argument values are constant, their
3958 known values may permit simplifications at compile time so that not
3959 all of the inline function's code needs to be included. The effect on
3960 code size is less predictable; object code may be larger or smaller
3961 with function inlining, depending on the particular case. You can
3962 also direct GCC to try to integrate all ``simple enough'' functions
3963 into their callers with the option @option{-finline-functions}.
3965 GCC implements three different semantics of declaring a function
3966 inline. One is available with @option{-std=gnu89}, another when
3967 @option{-std=c99} or @option{-std=gnu99}, and the third is used when
3970 To declare a function inline, use the @code{inline} keyword in its
3971 declaration, like this:
3981 If you are writing a header file to be included in ISO C89 programs, write
3982 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3984 The three types of inlining behave similarly in two important cases:
3985 when the @code{inline} keyword is used on a @code{static} function,
3986 like the example above, and when a function is first declared without
3987 using the @code{inline} keyword and then is defined with
3988 @code{inline}, like this:
3991 extern int inc (int *a);
3999 In both of these common cases, the program behaves the same as if you
4000 had not used the @code{inline} keyword, except for its speed.
4002 @cindex inline functions, omission of
4003 @opindex fkeep-inline-functions
4004 When a function is both inline and @code{static}, if all calls to the
4005 function are integrated into the caller, and the function's address is
4006 never used, then the function's own assembler code is never referenced.
4007 In this case, GCC does not actually output assembler code for the
4008 function, unless you specify the option @option{-fkeep-inline-functions}.
4009 Some calls cannot be integrated for various reasons (in particular,
4010 calls that precede the function's definition cannot be integrated, and
4011 neither can recursive calls within the definition). If there is a
4012 nonintegrated call, then the function is compiled to assembler code as
4013 usual. The function must also be compiled as usual if the program
4014 refers to its address, because that can't be inlined.
4016 @cindex automatic @code{inline} for C++ member fns
4017 @cindex @code{inline} automatic for C++ member fns
4018 @cindex member fns, automatically @code{inline}
4019 @cindex C++ member fns, automatically @code{inline}
4020 @opindex fno-default-inline
4021 As required by ISO C++, GCC considers member functions defined within
4022 the body of a class to be marked inline even if they are
4023 not explicitly declared with the @code{inline} keyword. You can
4024 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4025 Options,,Options Controlling C++ Dialect}.
4027 GCC does not inline any functions when not optimizing unless you specify
4028 the @samp{always_inline} attribute for the function, like this:
4031 /* @r{Prototype.} */
4032 inline void foo (const char) __attribute__((always_inline));
4035 The remainder of this section is specific to GNU C89 inlining.
4037 @cindex non-static inline function
4038 When an inline function is not @code{static}, then the compiler must assume
4039 that there may be calls from other source files; since a global symbol can
4040 be defined only once in any program, the function must not be defined in
4041 the other source files, so the calls therein cannot be integrated.
4042 Therefore, a non-@code{static} inline function is always compiled on its
4043 own in the usual fashion.
4045 If you specify both @code{inline} and @code{extern} in the function
4046 definition, then the definition is used only for inlining. In no case
4047 is the function compiled on its own, not even if you refer to its
4048 address explicitly. Such an address becomes an external reference, as
4049 if you had only declared the function, and had not defined it.
4051 This combination of @code{inline} and @code{extern} has almost the
4052 effect of a macro. The way to use it is to put a function definition in
4053 a header file with these keywords, and put another copy of the
4054 definition (lacking @code{inline} and @code{extern}) in a library file.
4055 The definition in the header file will cause most calls to the function
4056 to be inlined. If any uses of the function remain, they will refer to
4057 the single copy in the library.
4060 @section Assembler Instructions with C Expression Operands
4061 @cindex extended @code{asm}
4062 @cindex @code{asm} expressions
4063 @cindex assembler instructions
4066 In an assembler instruction using @code{asm}, you can specify the
4067 operands of the instruction using C expressions. This means you need not
4068 guess which registers or memory locations will contain the data you want
4071 You must specify an assembler instruction template much like what
4072 appears in a machine description, plus an operand constraint string for
4075 For example, here is how to use the 68881's @code{fsinx} instruction:
4078 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4082 Here @code{angle} is the C expression for the input operand while
4083 @code{result} is that of the output operand. Each has @samp{"f"} as its
4084 operand constraint, saying that a floating point register is required.
4085 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4086 output operands' constraints must use @samp{=}. The constraints use the
4087 same language used in the machine description (@pxref{Constraints}).
4089 Each operand is described by an operand-constraint string followed by
4090 the C expression in parentheses. A colon separates the assembler
4091 template from the first output operand and another separates the last
4092 output operand from the first input, if any. Commas separate the
4093 operands within each group. The total number of operands is currently
4094 limited to 30; this limitation may be lifted in some future version of
4097 If there are no output operands but there are input operands, you must
4098 place two consecutive colons surrounding the place where the output
4101 As of GCC version 3.1, it is also possible to specify input and output
4102 operands using symbolic names which can be referenced within the
4103 assembler code. These names are specified inside square brackets
4104 preceding the constraint string, and can be referenced inside the
4105 assembler code using @code{%[@var{name}]} instead of a percentage sign
4106 followed by the operand number. Using named operands the above example
4110 asm ("fsinx %[angle],%[output]"
4111 : [output] "=f" (result)
4112 : [angle] "f" (angle));
4116 Note that the symbolic operand names have no relation whatsoever to
4117 other C identifiers. You may use any name you like, even those of
4118 existing C symbols, but you must ensure that no two operands within the same
4119 assembler construct use the same symbolic name.
4121 Output operand expressions must be lvalues; the compiler can check this.
4122 The input operands need not be lvalues. The compiler cannot check
4123 whether the operands have data types that are reasonable for the
4124 instruction being executed. It does not parse the assembler instruction
4125 template and does not know what it means or even whether it is valid
4126 assembler input. The extended @code{asm} feature is most often used for
4127 machine instructions the compiler itself does not know exist. If
4128 the output expression cannot be directly addressed (for example, it is a
4129 bit-field), your constraint must allow a register. In that case, GCC
4130 will use the register as the output of the @code{asm}, and then store
4131 that register into the output.
4133 The ordinary output operands must be write-only; GCC will assume that
4134 the values in these operands before the instruction are dead and need
4135 not be generated. Extended asm supports input-output or read-write
4136 operands. Use the constraint character @samp{+} to indicate such an
4137 operand and list it with the output operands. You should only use
4138 read-write operands when the constraints for the operand (or the
4139 operand in which only some of the bits are to be changed) allow a
4142 You may, as an alternative, logically split its function into two
4143 separate operands, one input operand and one write-only output
4144 operand. The connection between them is expressed by constraints
4145 which say they need to be in the same location when the instruction
4146 executes. You can use the same C expression for both operands, or
4147 different expressions. For example, here we write the (fictitious)
4148 @samp{combine} instruction with @code{bar} as its read-only source
4149 operand and @code{foo} as its read-write destination:
4152 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4156 The constraint @samp{"0"} for operand 1 says that it must occupy the
4157 same location as operand 0. A number in constraint is allowed only in
4158 an input operand and it must refer to an output operand.
4160 Only a number in the constraint can guarantee that one operand will be in
4161 the same place as another. The mere fact that @code{foo} is the value
4162 of both operands is not enough to guarantee that they will be in the
4163 same place in the generated assembler code. The following would not
4167 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4170 Various optimizations or reloading could cause operands 0 and 1 to be in
4171 different registers; GCC knows no reason not to do so. For example, the
4172 compiler might find a copy of the value of @code{foo} in one register and
4173 use it for operand 1, but generate the output operand 0 in a different
4174 register (copying it afterward to @code{foo}'s own address). Of course,
4175 since the register for operand 1 is not even mentioned in the assembler
4176 code, the result will not work, but GCC can't tell that.
4178 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4179 the operand number for a matching constraint. For example:
4182 asm ("cmoveq %1,%2,%[result]"
4183 : [result] "=r"(result)
4184 : "r" (test), "r"(new), "[result]"(old));
4187 Sometimes you need to make an @code{asm} operand be a specific register,
4188 but there's no matching constraint letter for that register @emph{by
4189 itself}. To force the operand into that register, use a local variable
4190 for the operand and specify the register in the variable declaration.
4191 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4192 register constraint letter that matches the register:
4195 register int *p1 asm ("r0") = @dots{};
4196 register int *p2 asm ("r1") = @dots{};
4197 register int *result asm ("r0");
4198 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4201 @anchor{Example of asm with clobbered asm reg}
4202 In the above example, beware that a register that is call-clobbered by
4203 the target ABI will be overwritten by any function call in the
4204 assignment, including library calls for arithmetic operators.
4205 Assuming it is a call-clobbered register, this may happen to @code{r0}
4206 above by the assignment to @code{p2}. If you have to use such a
4207 register, use temporary variables for expressions between the register
4212 register int *p1 asm ("r0") = @dots{};
4213 register int *p2 asm ("r1") = t1;
4214 register int *result asm ("r0");
4215 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4218 Some instructions clobber specific hard registers. To describe this,
4219 write a third colon after the input operands, followed by the names of
4220 the clobbered hard registers (given as strings). Here is a realistic
4221 example for the VAX:
4224 asm volatile ("movc3 %0,%1,%2"
4225 : /* @r{no outputs} */
4226 : "g" (from), "g" (to), "g" (count)
4227 : "r0", "r1", "r2", "r3", "r4", "r5");
4230 You may not write a clobber description in a way that overlaps with an
4231 input or output operand. For example, you may not have an operand
4232 describing a register class with one member if you mention that register
4233 in the clobber list. Variables declared to live in specific registers
4234 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4235 have no part mentioned in the clobber description.
4236 There is no way for you to specify that an input
4237 operand is modified without also specifying it as an output
4238 operand. Note that if all the output operands you specify are for this
4239 purpose (and hence unused), you will then also need to specify
4240 @code{volatile} for the @code{asm} construct, as described below, to
4241 prevent GCC from deleting the @code{asm} statement as unused.
4243 If you refer to a particular hardware register from the assembler code,
4244 you will probably have to list the register after the third colon to
4245 tell the compiler the register's value is modified. In some assemblers,
4246 the register names begin with @samp{%}; to produce one @samp{%} in the
4247 assembler code, you must write @samp{%%} in the input.
4249 If your assembler instruction can alter the condition code register, add
4250 @samp{cc} to the list of clobbered registers. GCC on some machines
4251 represents the condition codes as a specific hardware register;
4252 @samp{cc} serves to name this register. On other machines, the
4253 condition code is handled differently, and specifying @samp{cc} has no
4254 effect. But it is valid no matter what the machine.
4256 If your assembler instructions access memory in an unpredictable
4257 fashion, add @samp{memory} to the list of clobbered registers. This
4258 will cause GCC to not keep memory values cached in registers across the
4259 assembler instruction and not optimize stores or loads to that memory.
4260 You will also want to add the @code{volatile} keyword if the memory
4261 affected is not listed in the inputs or outputs of the @code{asm}, as
4262 the @samp{memory} clobber does not count as a side-effect of the
4263 @code{asm}. If you know how large the accessed memory is, you can add
4264 it as input or output but if this is not known, you should add
4265 @samp{memory}. As an example, if you access ten bytes of a string, you
4266 can use a memory input like:
4269 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4272 Note that in the following example the memory input is necessary,
4273 otherwise GCC might optimize the store to @code{x} away:
4280 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4281 "=&d" (r) : "a" (y), "m" (*y));
4286 You can put multiple assembler instructions together in a single
4287 @code{asm} template, separated by the characters normally used in assembly
4288 code for the system. A combination that works in most places is a newline
4289 to break the line, plus a tab character to move to the instruction field
4290 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4291 assembler allows semicolons as a line-breaking character. Note that some
4292 assembler dialects use semicolons to start a comment.
4293 The input operands are guaranteed not to use any of the clobbered
4294 registers, and neither will the output operands' addresses, so you can
4295 read and write the clobbered registers as many times as you like. Here
4296 is an example of multiple instructions in a template; it assumes the
4297 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4300 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4302 : "g" (from), "g" (to)
4306 Unless an output operand has the @samp{&} constraint modifier, GCC
4307 may allocate it in the same register as an unrelated input operand, on
4308 the assumption the inputs are consumed before the outputs are produced.
4309 This assumption may be false if the assembler code actually consists of
4310 more than one instruction. In such a case, use @samp{&} for each output
4311 operand that may not overlap an input. @xref{Modifiers}.
4313 If you want to test the condition code produced by an assembler
4314 instruction, you must include a branch and a label in the @code{asm}
4315 construct, as follows:
4318 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4324 This assumes your assembler supports local labels, as the GNU assembler
4325 and most Unix assemblers do.
4327 Speaking of labels, jumps from one @code{asm} to another are not
4328 supported. The compiler's optimizers do not know about these jumps, and
4329 therefore they cannot take account of them when deciding how to
4332 @cindex macros containing @code{asm}
4333 Usually the most convenient way to use these @code{asm} instructions is to
4334 encapsulate them in macros that look like functions. For example,
4338 (@{ double __value, __arg = (x); \
4339 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4344 Here the variable @code{__arg} is used to make sure that the instruction
4345 operates on a proper @code{double} value, and to accept only those
4346 arguments @code{x} which can convert automatically to a @code{double}.
4348 Another way to make sure the instruction operates on the correct data
4349 type is to use a cast in the @code{asm}. This is different from using a
4350 variable @code{__arg} in that it converts more different types. For
4351 example, if the desired type were @code{int}, casting the argument to
4352 @code{int} would accept a pointer with no complaint, while assigning the
4353 argument to an @code{int} variable named @code{__arg} would warn about
4354 using a pointer unless the caller explicitly casts it.
4356 If an @code{asm} has output operands, GCC assumes for optimization
4357 purposes the instruction has no side effects except to change the output
4358 operands. This does not mean instructions with a side effect cannot be
4359 used, but you must be careful, because the compiler may eliminate them
4360 if the output operands aren't used, or move them out of loops, or
4361 replace two with one if they constitute a common subexpression. Also,
4362 if your instruction does have a side effect on a variable that otherwise
4363 appears not to change, the old value of the variable may be reused later
4364 if it happens to be found in a register.
4366 You can prevent an @code{asm} instruction from being deleted
4367 by writing the keyword @code{volatile} after
4368 the @code{asm}. For example:
4371 #define get_and_set_priority(new) \
4373 asm volatile ("get_and_set_priority %0, %1" \
4374 : "=g" (__old) : "g" (new)); \
4379 The @code{volatile} keyword indicates that the instruction has
4380 important side-effects. GCC will not delete a volatile @code{asm} if
4381 it is reachable. (The instruction can still be deleted if GCC can
4382 prove that control-flow will never reach the location of the
4383 instruction.) Note that even a volatile @code{asm} instruction
4384 can be moved relative to other code, including across jump
4385 instructions. For example, on many targets there is a system
4386 register which can be set to control the rounding mode of
4387 floating point operations. You might try
4388 setting it with a volatile @code{asm}, like this PowerPC example:
4391 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4396 This will not work reliably, as the compiler may move the addition back
4397 before the volatile @code{asm}. To make it work you need to add an
4398 artificial dependency to the @code{asm} referencing a variable in the code
4399 you don't want moved, for example:
4402 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4406 Similarly, you can't expect a
4407 sequence of volatile @code{asm} instructions to remain perfectly
4408 consecutive. If you want consecutive output, use a single @code{asm}.
4409 Also, GCC will perform some optimizations across a volatile @code{asm}
4410 instruction; GCC does not ``forget everything'' when it encounters
4411 a volatile @code{asm} instruction the way some other compilers do.
4413 An @code{asm} instruction without any output operands will be treated
4414 identically to a volatile @code{asm} instruction.
4416 It is a natural idea to look for a way to give access to the condition
4417 code left by the assembler instruction. However, when we attempted to
4418 implement this, we found no way to make it work reliably. The problem
4419 is that output operands might need reloading, which would result in
4420 additional following ``store'' instructions. On most machines, these
4421 instructions would alter the condition code before there was time to
4422 test it. This problem doesn't arise for ordinary ``test'' and
4423 ``compare'' instructions because they don't have any output operands.
4425 For reasons similar to those described above, it is not possible to give
4426 an assembler instruction access to the condition code left by previous
4429 If you are writing a header file that should be includable in ISO C
4430 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4433 @subsection Size of an @code{asm}
4435 Some targets require that GCC track the size of each instruction used in
4436 order to generate correct code. Because the final length of an
4437 @code{asm} is only known by the assembler, GCC must make an estimate as
4438 to how big it will be. The estimate is formed by counting the number of
4439 statements in the pattern of the @code{asm} and multiplying that by the
4440 length of the longest instruction on that processor. Statements in the
4441 @code{asm} are identified by newline characters and whatever statement
4442 separator characters are supported by the assembler; on most processors
4443 this is the `@code{;}' character.
4445 Normally, GCC's estimate is perfectly adequate to ensure that correct
4446 code is generated, but it is possible to confuse the compiler if you use
4447 pseudo instructions or assembler macros that expand into multiple real
4448 instructions or if you use assembler directives that expand to more
4449 space in the object file than would be needed for a single instruction.
4450 If this happens then the assembler will produce a diagnostic saying that
4451 a label is unreachable.
4453 @subsection i386 floating point asm operands
4455 There are several rules on the usage of stack-like regs in
4456 asm_operands insns. These rules apply only to the operands that are
4461 Given a set of input regs that die in an asm_operands, it is
4462 necessary to know which are implicitly popped by the asm, and
4463 which must be explicitly popped by gcc.
4465 An input reg that is implicitly popped by the asm must be
4466 explicitly clobbered, unless it is constrained to match an
4470 For any input reg that is implicitly popped by an asm, it is
4471 necessary to know how to adjust the stack to compensate for the pop.
4472 If any non-popped input is closer to the top of the reg-stack than
4473 the implicitly popped reg, it would not be possible to know what the
4474 stack looked like---it's not clear how the rest of the stack ``slides
4477 All implicitly popped input regs must be closer to the top of
4478 the reg-stack than any input that is not implicitly popped.
4480 It is possible that if an input dies in an insn, reload might
4481 use the input reg for an output reload. Consider this example:
4484 asm ("foo" : "=t" (a) : "f" (b));
4487 This asm says that input B is not popped by the asm, and that
4488 the asm pushes a result onto the reg-stack, i.e., the stack is one
4489 deeper after the asm than it was before. But, it is possible that
4490 reload will think that it can use the same reg for both the input and
4491 the output, if input B dies in this insn.
4493 If any input operand uses the @code{f} constraint, all output reg
4494 constraints must use the @code{&} earlyclobber.
4496 The asm above would be written as
4499 asm ("foo" : "=&t" (a) : "f" (b));
4503 Some operands need to be in particular places on the stack. All
4504 output operands fall in this category---there is no other way to
4505 know which regs the outputs appear in unless the user indicates
4506 this in the constraints.
4508 Output operands must specifically indicate which reg an output
4509 appears in after an asm. @code{=f} is not allowed: the operand
4510 constraints must select a class with a single reg.
4513 Output operands may not be ``inserted'' between existing stack regs.
4514 Since no 387 opcode uses a read/write operand, all output operands
4515 are dead before the asm_operands, and are pushed by the asm_operands.
4516 It makes no sense to push anywhere but the top of the reg-stack.
4518 Output operands must start at the top of the reg-stack: output
4519 operands may not ``skip'' a reg.
4522 Some asm statements may need extra stack space for internal
4523 calculations. This can be guaranteed by clobbering stack registers
4524 unrelated to the inputs and outputs.
4528 Here are a couple of reasonable asms to want to write. This asm
4529 takes one input, which is internally popped, and produces two outputs.
4532 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4535 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4536 and replaces them with one output. The user must code the @code{st(1)}
4537 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4540 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4546 @section Controlling Names Used in Assembler Code
4547 @cindex assembler names for identifiers
4548 @cindex names used in assembler code
4549 @cindex identifiers, names in assembler code
4551 You can specify the name to be used in the assembler code for a C
4552 function or variable by writing the @code{asm} (or @code{__asm__})
4553 keyword after the declarator as follows:
4556 int foo asm ("myfoo") = 2;
4560 This specifies that the name to be used for the variable @code{foo} in
4561 the assembler code should be @samp{myfoo} rather than the usual
4564 On systems where an underscore is normally prepended to the name of a C
4565 function or variable, this feature allows you to define names for the
4566 linker that do not start with an underscore.
4568 It does not make sense to use this feature with a non-static local
4569 variable since such variables do not have assembler names. If you are
4570 trying to put the variable in a particular register, see @ref{Explicit
4571 Reg Vars}. GCC presently accepts such code with a warning, but will
4572 probably be changed to issue an error, rather than a warning, in the
4575 You cannot use @code{asm} in this way in a function @emph{definition}; but
4576 you can get the same effect by writing a declaration for the function
4577 before its definition and putting @code{asm} there, like this:
4580 extern func () asm ("FUNC");
4587 It is up to you to make sure that the assembler names you choose do not
4588 conflict with any other assembler symbols. Also, you must not use a
4589 register name; that would produce completely invalid assembler code. GCC
4590 does not as yet have the ability to store static variables in registers.
4591 Perhaps that will be added.
4593 @node Explicit Reg Vars
4594 @section Variables in Specified Registers
4595 @cindex explicit register variables
4596 @cindex variables in specified registers
4597 @cindex specified registers
4598 @cindex registers, global allocation
4600 GNU C allows you to put a few global variables into specified hardware
4601 registers. You can also specify the register in which an ordinary
4602 register variable should be allocated.
4606 Global register variables reserve registers throughout the program.
4607 This may be useful in programs such as programming language
4608 interpreters which have a couple of global variables that are accessed
4612 Local register variables in specific registers do not reserve the
4613 registers, except at the point where they are used as input or output
4614 operands in an @code{asm} statement and the @code{asm} statement itself is
4615 not deleted. The compiler's data flow analysis is capable of determining
4616 where the specified registers contain live values, and where they are
4617 available for other uses. Stores into local register variables may be deleted
4618 when they appear to be dead according to dataflow analysis. References
4619 to local register variables may be deleted or moved or simplified.
4621 These local variables are sometimes convenient for use with the extended
4622 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4623 output of the assembler instruction directly into a particular register.
4624 (This will work provided the register you specify fits the constraints
4625 specified for that operand in the @code{asm}.)
4633 @node Global Reg Vars
4634 @subsection Defining Global Register Variables
4635 @cindex global register variables
4636 @cindex registers, global variables in
4638 You can define a global register variable in GNU C like this:
4641 register int *foo asm ("a5");
4645 Here @code{a5} is the name of the register which should be used. Choose a
4646 register which is normally saved and restored by function calls on your
4647 machine, so that library routines will not clobber it.
4649 Naturally the register name is cpu-dependent, so you would need to
4650 conditionalize your program according to cpu type. The register
4651 @code{a5} would be a good choice on a 68000 for a variable of pointer
4652 type. On machines with register windows, be sure to choose a ``global''
4653 register that is not affected magically by the function call mechanism.
4655 In addition, operating systems on one type of cpu may differ in how they
4656 name the registers; then you would need additional conditionals. For
4657 example, some 68000 operating systems call this register @code{%a5}.
4659 Eventually there may be a way of asking the compiler to choose a register
4660 automatically, but first we need to figure out how it should choose and
4661 how to enable you to guide the choice. No solution is evident.
4663 Defining a global register variable in a certain register reserves that
4664 register entirely for this use, at least within the current compilation.
4665 The register will not be allocated for any other purpose in the functions
4666 in the current compilation. The register will not be saved and restored by
4667 these functions. Stores into this register are never deleted even if they
4668 would appear to be dead, but references may be deleted or moved or
4671 It is not safe to access the global register variables from signal
4672 handlers, or from more than one thread of control, because the system
4673 library routines may temporarily use the register for other things (unless
4674 you recompile them specially for the task at hand).
4676 @cindex @code{qsort}, and global register variables
4677 It is not safe for one function that uses a global register variable to
4678 call another such function @code{foo} by way of a third function
4679 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4680 different source file in which the variable wasn't declared). This is
4681 because @code{lose} might save the register and put some other value there.
4682 For example, you can't expect a global register variable to be available in
4683 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4684 might have put something else in that register. (If you are prepared to
4685 recompile @code{qsort} with the same global register variable, you can
4686 solve this problem.)
4688 If you want to recompile @code{qsort} or other source files which do not
4689 actually use your global register variable, so that they will not use that
4690 register for any other purpose, then it suffices to specify the compiler
4691 option @option{-ffixed-@var{reg}}. You need not actually add a global
4692 register declaration to their source code.
4694 A function which can alter the value of a global register variable cannot
4695 safely be called from a function compiled without this variable, because it
4696 could clobber the value the caller expects to find there on return.
4697 Therefore, the function which is the entry point into the part of the
4698 program that uses the global register variable must explicitly save and
4699 restore the value which belongs to its caller.
4701 @cindex register variable after @code{longjmp}
4702 @cindex global register after @code{longjmp}
4703 @cindex value after @code{longjmp}
4706 On most machines, @code{longjmp} will restore to each global register
4707 variable the value it had at the time of the @code{setjmp}. On some
4708 machines, however, @code{longjmp} will not change the value of global
4709 register variables. To be portable, the function that called @code{setjmp}
4710 should make other arrangements to save the values of the global register
4711 variables, and to restore them in a @code{longjmp}. This way, the same
4712 thing will happen regardless of what @code{longjmp} does.
4714 All global register variable declarations must precede all function
4715 definitions. If such a declaration could appear after function
4716 definitions, the declaration would be too late to prevent the register from
4717 being used for other purposes in the preceding functions.
4719 Global register variables may not have initial values, because an
4720 executable file has no means to supply initial contents for a register.
4722 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4723 registers, but certain library functions, such as @code{getwd}, as well
4724 as the subroutines for division and remainder, modify g3 and g4. g1 and
4725 g2 are local temporaries.
4727 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4728 Of course, it will not do to use more than a few of those.
4730 @node Local Reg Vars
4731 @subsection Specifying Registers for Local Variables
4732 @cindex local variables, specifying registers
4733 @cindex specifying registers for local variables
4734 @cindex registers for local variables
4736 You can define a local register variable with a specified register
4740 register int *foo asm ("a5");
4744 Here @code{a5} is the name of the register which should be used. Note
4745 that this is the same syntax used for defining global register
4746 variables, but for a local variable it would appear within a function.
4748 Naturally the register name is cpu-dependent, but this is not a
4749 problem, since specific registers are most often useful with explicit
4750 assembler instructions (@pxref{Extended Asm}). Both of these things
4751 generally require that you conditionalize your program according to
4754 In addition, operating systems on one type of cpu may differ in how they
4755 name the registers; then you would need additional conditionals. For
4756 example, some 68000 operating systems call this register @code{%a5}.
4758 Defining such a register variable does not reserve the register; it
4759 remains available for other uses in places where flow control determines
4760 the variable's value is not live.
4762 This option does not guarantee that GCC will generate code that has
4763 this variable in the register you specify at all times. You may not
4764 code an explicit reference to this register in the @emph{assembler
4765 instruction template} part of an @code{asm} statement and assume it will
4766 always refer to this variable. However, using the variable as an
4767 @code{asm} @emph{operand} guarantees that the specified register is used
4770 Stores into local register variables may be deleted when they appear to be dead
4771 according to dataflow analysis. References to local register variables may
4772 be deleted or moved or simplified.
4774 As for global register variables, it's recommended that you choose a
4775 register which is normally saved and restored by function calls on
4776 your machine, so that library routines will not clobber it. A common
4777 pitfall is to initialize multiple call-clobbered registers with
4778 arbitrary expressions, where a function call or library call for an
4779 arithmetic operator will overwrite a register value from a previous
4780 assignment, for example @code{r0} below:
4782 register int *p1 asm ("r0") = @dots{};
4783 register int *p2 asm ("r1") = @dots{};
4785 In those cases, a solution is to use a temporary variable for
4786 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4788 @node Alternate Keywords
4789 @section Alternate Keywords
4790 @cindex alternate keywords
4791 @cindex keywords, alternate
4793 @option{-ansi} and the various @option{-std} options disable certain
4794 keywords. This causes trouble when you want to use GNU C extensions, or
4795 a general-purpose header file that should be usable by all programs,
4796 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4797 @code{inline} are not available in programs compiled with
4798 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4799 program compiled with @option{-std=c99}). The ISO C99 keyword
4800 @code{restrict} is only available when @option{-std=gnu99} (which will
4801 eventually be the default) or @option{-std=c99} (or the equivalent
4802 @option{-std=iso9899:1999}) is used.
4804 The way to solve these problems is to put @samp{__} at the beginning and
4805 end of each problematical keyword. For example, use @code{__asm__}
4806 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4808 Other C compilers won't accept these alternative keywords; if you want to
4809 compile with another compiler, you can define the alternate keywords as
4810 macros to replace them with the customary keywords. It looks like this:
4818 @findex __extension__
4820 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4822 prevent such warnings within one expression by writing
4823 @code{__extension__} before the expression. @code{__extension__} has no
4824 effect aside from this.
4826 @node Incomplete Enums
4827 @section Incomplete @code{enum} Types
4829 You can define an @code{enum} tag without specifying its possible values.
4830 This results in an incomplete type, much like what you get if you write
4831 @code{struct foo} without describing the elements. A later declaration
4832 which does specify the possible values completes the type.
4834 You can't allocate variables or storage using the type while it is
4835 incomplete. However, you can work with pointers to that type.
4837 This extension may not be very useful, but it makes the handling of
4838 @code{enum} more consistent with the way @code{struct} and @code{union}
4841 This extension is not supported by GNU C++.
4843 @node Function Names
4844 @section Function Names as Strings
4845 @cindex @code{__func__} identifier
4846 @cindex @code{__FUNCTION__} identifier
4847 @cindex @code{__PRETTY_FUNCTION__} identifier
4849 GCC provides three magic variables which hold the name of the current
4850 function, as a string. The first of these is @code{__func__}, which
4851 is part of the C99 standard:
4854 The identifier @code{__func__} is implicitly declared by the translator
4855 as if, immediately following the opening brace of each function
4856 definition, the declaration
4859 static const char __func__[] = "function-name";
4862 appeared, where function-name is the name of the lexically-enclosing
4863 function. This name is the unadorned name of the function.
4866 @code{__FUNCTION__} is another name for @code{__func__}. Older
4867 versions of GCC recognize only this name. However, it is not
4868 standardized. For maximum portability, we recommend you use
4869 @code{__func__}, but provide a fallback definition with the
4873 #if __STDC_VERSION__ < 199901L
4875 # define __func__ __FUNCTION__
4877 # define __func__ "<unknown>"
4882 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4883 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4884 the type signature of the function as well as its bare name. For
4885 example, this program:
4889 extern int printf (char *, ...);
4896 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4897 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4915 __PRETTY_FUNCTION__ = void a::sub(int)
4918 These identifiers are not preprocessor macros. In GCC 3.3 and
4919 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4920 were treated as string literals; they could be used to initialize
4921 @code{char} arrays, and they could be concatenated with other string
4922 literals. GCC 3.4 and later treat them as variables, like
4923 @code{__func__}. In C++, @code{__FUNCTION__} and
4924 @code{__PRETTY_FUNCTION__} have always been variables.
4926 @node Return Address
4927 @section Getting the Return or Frame Address of a Function
4929 These functions may be used to get information about the callers of a
4932 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4933 This function returns the return address of the current function, or of
4934 one of its callers. The @var{level} argument is number of frames to
4935 scan up the call stack. A value of @code{0} yields the return address
4936 of the current function, a value of @code{1} yields the return address
4937 of the caller of the current function, and so forth. When inlining
4938 the expected behavior is that the function will return the address of
4939 the function that will be returned to. To work around this behavior use
4940 the @code{noinline} function attribute.
4942 The @var{level} argument must be a constant integer.
4944 On some machines it may be impossible to determine the return address of
4945 any function other than the current one; in such cases, or when the top
4946 of the stack has been reached, this function will return @code{0} or a
4947 random value. In addition, @code{__builtin_frame_address} may be used
4948 to determine if the top of the stack has been reached.
4950 This function should only be used with a nonzero argument for debugging
4954 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4955 This function is similar to @code{__builtin_return_address}, but it
4956 returns the address of the function frame rather than the return address
4957 of the function. Calling @code{__builtin_frame_address} with a value of
4958 @code{0} yields the frame address of the current function, a value of
4959 @code{1} yields the frame address of the caller of the current function,
4962 The frame is the area on the stack which holds local variables and saved
4963 registers. The frame address is normally the address of the first word
4964 pushed on to the stack by the function. However, the exact definition
4965 depends upon the processor and the calling convention. If the processor
4966 has a dedicated frame pointer register, and the function has a frame,
4967 then @code{__builtin_frame_address} will return the value of the frame
4970 On some machines it may be impossible to determine the frame address of
4971 any function other than the current one; in such cases, or when the top
4972 of the stack has been reached, this function will return @code{0} if
4973 the first frame pointer is properly initialized by the startup code.
4975 This function should only be used with a nonzero argument for debugging
4979 @node Vector Extensions
4980 @section Using vector instructions through built-in functions
4982 On some targets, the instruction set contains SIMD vector instructions that
4983 operate on multiple values contained in one large register at the same time.
4984 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4987 The first step in using these extensions is to provide the necessary data
4988 types. This should be done using an appropriate @code{typedef}:
4991 typedef int v4si __attribute__ ((vector_size (16)));
4994 The @code{int} type specifies the base type, while the attribute specifies
4995 the vector size for the variable, measured in bytes. For example, the
4996 declaration above causes the compiler to set the mode for the @code{v4si}
4997 type to be 16 bytes wide and divided into @code{int} sized units. For
4998 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4999 corresponding mode of @code{foo} will be @acronym{V4SI}.
5001 The @code{vector_size} attribute is only applicable to integral and
5002 float scalars, although arrays, pointers, and function return values
5003 are allowed in conjunction with this construct.
5005 All the basic integer types can be used as base types, both as signed
5006 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5007 @code{long long}. In addition, @code{float} and @code{double} can be
5008 used to build floating-point vector types.
5010 Specifying a combination that is not valid for the current architecture
5011 will cause GCC to synthesize the instructions using a narrower mode.
5012 For example, if you specify a variable of type @code{V4SI} and your
5013 architecture does not allow for this specific SIMD type, GCC will
5014 produce code that uses 4 @code{SIs}.
5016 The types defined in this manner can be used with a subset of normal C
5017 operations. Currently, GCC will allow using the following operators
5018 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5020 The operations behave like C++ @code{valarrays}. Addition is defined as
5021 the addition of the corresponding elements of the operands. For
5022 example, in the code below, each of the 4 elements in @var{a} will be
5023 added to the corresponding 4 elements in @var{b} and the resulting
5024 vector will be stored in @var{c}.
5027 typedef int v4si __attribute__ ((vector_size (16)));
5034 Subtraction, multiplication, division, and the logical operations
5035 operate in a similar manner. Likewise, the result of using the unary
5036 minus or complement operators on a vector type is a vector whose
5037 elements are the negative or complemented values of the corresponding
5038 elements in the operand.
5040 You can declare variables and use them in function calls and returns, as
5041 well as in assignments and some casts. You can specify a vector type as
5042 a return type for a function. Vector types can also be used as function
5043 arguments. It is possible to cast from one vector type to another,
5044 provided they are of the same size (in fact, you can also cast vectors
5045 to and from other datatypes of the same size).
5047 You cannot operate between vectors of different lengths or different
5048 signedness without a cast.
5050 A port that supports hardware vector operations, usually provides a set
5051 of built-in functions that can be used to operate on vectors. For
5052 example, a function to add two vectors and multiply the result by a
5053 third could look like this:
5056 v4si f (v4si a, v4si b, v4si c)
5058 v4si tmp = __builtin_addv4si (a, b);
5059 return __builtin_mulv4si (tmp, c);
5066 @findex __builtin_offsetof
5068 GCC implements for both C and C++ a syntactic extension to implement
5069 the @code{offsetof} macro.
5073 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5075 offsetof_member_designator:
5077 | offsetof_member_designator "." @code{identifier}
5078 | offsetof_member_designator "[" @code{expr} "]"
5081 This extension is sufficient such that
5084 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5087 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5088 may be dependent. In either case, @var{member} may consist of a single
5089 identifier, or a sequence of member accesses and array references.
5091 @node Atomic Builtins
5092 @section Built-in functions for atomic memory access
5094 The following builtins are intended to be compatible with those described
5095 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5096 section 7.4. As such, they depart from the normal GCC practice of using
5097 the ``__builtin_'' prefix, and further that they are overloaded such that
5098 they work on multiple types.
5100 The definition given in the Intel documentation allows only for the use of
5101 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5102 counterparts. GCC will allow any integral scalar or pointer type that is
5103 1, 2, 4 or 8 bytes in length.
5105 Not all operations are supported by all target processors. If a particular
5106 operation cannot be implemented on the target processor, a warning will be
5107 generated and a call an external function will be generated. The external
5108 function will carry the same name as the builtin, with an additional suffix
5109 @samp{_@var{n}} where @var{n} is the size of the data type.
5111 @c ??? Should we have a mechanism to suppress this warning? This is almost
5112 @c useful for implementing the operation under the control of an external
5115 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5116 no memory operand will be moved across the operation, either forward or
5117 backward. Further, instructions will be issued as necessary to prevent the
5118 processor from speculating loads across the operation and from queuing stores
5119 after the operation.
5121 All of the routines are are described in the Intel documentation to take
5122 ``an optional list of variables protected by the memory barrier''. It's
5123 not clear what is meant by that; it could mean that @emph{only} the
5124 following variables are protected, or it could mean that these variables
5125 should in addition be protected. At present GCC ignores this list and
5126 protects all variables which are globally accessible. If in the future
5127 we make some use of this list, an empty list will continue to mean all
5128 globally accessible variables.
5131 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5132 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5133 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5134 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5135 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5136 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5137 @findex __sync_fetch_and_add
5138 @findex __sync_fetch_and_sub
5139 @findex __sync_fetch_and_or
5140 @findex __sync_fetch_and_and
5141 @findex __sync_fetch_and_xor
5142 @findex __sync_fetch_and_nand
5143 These builtins perform the operation suggested by the name, and
5144 returns the value that had previously been in memory. That is,
5147 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5148 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5151 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5152 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5153 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5154 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5155 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5156 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5157 @findex __sync_add_and_fetch
5158 @findex __sync_sub_and_fetch
5159 @findex __sync_or_and_fetch
5160 @findex __sync_and_and_fetch
5161 @findex __sync_xor_and_fetch
5162 @findex __sync_nand_and_fetch
5163 These builtins perform the operation suggested by the name, and
5164 return the new value. That is,
5167 @{ *ptr @var{op}= value; return *ptr; @}
5168 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5171 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5172 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5173 @findex __sync_bool_compare_and_swap
5174 @findex __sync_val_compare_and_swap
5175 These builtins perform an atomic compare and swap. That is, if the current
5176 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5179 The ``bool'' version returns true if the comparison is successful and
5180 @var{newval} was written. The ``val'' version returns the contents
5181 of @code{*@var{ptr}} before the operation.
5183 @item __sync_synchronize (...)
5184 @findex __sync_synchronize
5185 This builtin issues a full memory barrier.
5187 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5188 @findex __sync_lock_test_and_set
5189 This builtin, as described by Intel, is not a traditional test-and-set
5190 operation, but rather an atomic exchange operation. It writes @var{value}
5191 into @code{*@var{ptr}}, and returns the previous contents of
5194 Many targets have only minimal support for such locks, and do not support
5195 a full exchange operation. In this case, a target may support reduced
5196 functionality here by which the @emph{only} valid value to store is the
5197 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5198 is implementation defined.
5200 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5201 This means that references after the builtin cannot move to (or be
5202 speculated to) before the builtin, but previous memory stores may not
5203 be globally visible yet, and previous memory loads may not yet be
5206 @item void __sync_lock_release (@var{type} *ptr, ...)
5207 @findex __sync_lock_release
5208 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5209 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5211 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5212 This means that all previous memory stores are globally visible, and all
5213 previous memory loads have been satisfied, but following memory reads
5214 are not prevented from being speculated to before the barrier.
5217 @node Object Size Checking
5218 @section Object Size Checking Builtins
5219 @findex __builtin_object_size
5220 @findex __builtin___memcpy_chk
5221 @findex __builtin___mempcpy_chk
5222 @findex __builtin___memmove_chk
5223 @findex __builtin___memset_chk
5224 @findex __builtin___strcpy_chk
5225 @findex __builtin___stpcpy_chk
5226 @findex __builtin___strncpy_chk
5227 @findex __builtin___strcat_chk
5228 @findex __builtin___strncat_chk
5229 @findex __builtin___sprintf_chk
5230 @findex __builtin___snprintf_chk
5231 @findex __builtin___vsprintf_chk
5232 @findex __builtin___vsnprintf_chk
5233 @findex __builtin___printf_chk
5234 @findex __builtin___vprintf_chk
5235 @findex __builtin___fprintf_chk
5236 @findex __builtin___vfprintf_chk
5238 GCC implements a limited buffer overflow protection mechanism
5239 that can prevent some buffer overflow attacks.
5241 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5242 is a built-in construct that returns a constant number of bytes from
5243 @var{ptr} to the end of the object @var{ptr} pointer points to
5244 (if known at compile time). @code{__builtin_object_size} never evaluates
5245 its arguments for side-effects. If there are any side-effects in them, it
5246 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5247 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5248 point to and all of them are known at compile time, the returned number
5249 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5250 0 and minimum if nonzero. If it is not possible to determine which objects
5251 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5252 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5253 for @var{type} 2 or 3.
5255 @var{type} is an integer constant from 0 to 3. If the least significant
5256 bit is clear, objects are whole variables, if it is set, a closest
5257 surrounding subobject is considered the object a pointer points to.
5258 The second bit determines if maximum or minimum of remaining bytes
5262 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5263 char *p = &var.buf1[1], *q = &var.b;
5265 /* Here the object p points to is var. */
5266 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5267 /* The subobject p points to is var.buf1. */
5268 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5269 /* The object q points to is var. */
5270 assert (__builtin_object_size (q, 0)
5271 == (char *) (&var + 1) - (char *) &var.b);
5272 /* The subobject q points to is var.b. */
5273 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5277 There are built-in functions added for many common string operation
5278 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5279 built-in is provided. This built-in has an additional last argument,
5280 which is the number of bytes remaining in object the @var{dest}
5281 argument points to or @code{(size_t) -1} if the size is not known.
5283 The built-in functions are optimized into the normal string functions
5284 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5285 it is known at compile time that the destination object will not
5286 be overflown. If the compiler can determine at compile time the
5287 object will be always overflown, it issues a warning.
5289 The intended use can be e.g.
5293 #define bos0(dest) __builtin_object_size (dest, 0)
5294 #define memcpy(dest, src, n) \
5295 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5299 /* It is unknown what object p points to, so this is optimized
5300 into plain memcpy - no checking is possible. */
5301 memcpy (p, "abcde", n);
5302 /* Destination is known and length too. It is known at compile
5303 time there will be no overflow. */
5304 memcpy (&buf[5], "abcde", 5);
5305 /* Destination is known, but the length is not known at compile time.
5306 This will result in __memcpy_chk call that can check for overflow
5308 memcpy (&buf[5], "abcde", n);
5309 /* Destination is known and it is known at compile time there will
5310 be overflow. There will be a warning and __memcpy_chk call that
5311 will abort the program at runtime. */
5312 memcpy (&buf[6], "abcde", 5);
5315 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5316 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5317 @code{strcat} and @code{strncat}.
5319 There are also checking built-in functions for formatted output functions.
5321 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5322 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5323 const char *fmt, ...);
5324 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5326 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5327 const char *fmt, va_list ap);
5330 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5331 etc. functions and can contain implementation specific flags on what
5332 additional security measures the checking function might take, such as
5333 handling @code{%n} differently.
5335 The @var{os} argument is the object size @var{s} points to, like in the
5336 other built-in functions. There is a small difference in the behavior
5337 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5338 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5339 the checking function is called with @var{os} argument set to
5342 In addition to this, there are checking built-in functions
5343 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5344 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5345 These have just one additional argument, @var{flag}, right before
5346 format string @var{fmt}. If the compiler is able to optimize them to
5347 @code{fputc} etc. functions, it will, otherwise the checking function
5348 should be called and the @var{flag} argument passed to it.
5350 @node Other Builtins
5351 @section Other built-in functions provided by GCC
5352 @cindex built-in functions
5353 @findex __builtin_isgreater
5354 @findex __builtin_isgreaterequal
5355 @findex __builtin_isless
5356 @findex __builtin_islessequal
5357 @findex __builtin_islessgreater
5358 @findex __builtin_isunordered
5359 @findex __builtin_powi
5360 @findex __builtin_powif
5361 @findex __builtin_powil
5519 @findex fprintf_unlocked
5521 @findex fputs_unlocked
5631 @findex printf_unlocked
5660 @findex significandf
5661 @findex significandl
5732 GCC provides a large number of built-in functions other than the ones
5733 mentioned above. Some of these are for internal use in the processing
5734 of exceptions or variable-length argument lists and will not be
5735 documented here because they may change from time to time; we do not
5736 recommend general use of these functions.
5738 The remaining functions are provided for optimization purposes.
5740 @opindex fno-builtin
5741 GCC includes built-in versions of many of the functions in the standard
5742 C library. The versions prefixed with @code{__builtin_} will always be
5743 treated as having the same meaning as the C library function even if you
5744 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5745 Many of these functions are only optimized in certain cases; if they are
5746 not optimized in a particular case, a call to the library function will
5751 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5752 @option{-std=c99}), the functions
5753 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5754 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5755 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5756 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5757 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5758 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5759 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5760 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5761 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5762 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5763 @code{significandf}, @code{significandl}, @code{significand},
5764 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5765 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5766 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5767 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5768 @code{ynl} and @code{yn}
5769 may be handled as built-in functions.
5770 All these functions have corresponding versions
5771 prefixed with @code{__builtin_}, which may be used even in strict C89
5774 The ISO C99 functions
5775 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5776 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5777 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5778 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5779 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5780 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5781 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5782 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5783 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5784 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5785 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5786 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5787 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5788 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5789 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5790 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5791 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5792 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5793 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5794 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5795 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5796 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5797 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5798 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5799 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5800 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5801 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5802 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5803 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5804 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5805 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5806 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5807 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5808 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5809 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5810 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5811 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5812 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5813 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5814 are handled as built-in functions
5815 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5817 There are also built-in versions of the ISO C99 functions
5818 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5819 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5820 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5821 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5822 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5823 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5824 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5825 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5826 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5827 that are recognized in any mode since ISO C90 reserves these names for
5828 the purpose to which ISO C99 puts them. All these functions have
5829 corresponding versions prefixed with @code{__builtin_}.
5831 The ISO C94 functions
5832 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5833 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5834 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5836 are handled as built-in functions
5837 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5839 The ISO C90 functions
5840 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5841 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5842 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5843 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5844 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5845 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5846 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5847 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5848 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5849 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5850 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5851 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5852 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5853 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5854 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5855 @code{vprintf} and @code{vsprintf}
5856 are all recognized as built-in functions unless
5857 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5858 is specified for an individual function). All of these functions have
5859 corresponding versions prefixed with @code{__builtin_}.
5861 GCC provides built-in versions of the ISO C99 floating point comparison
5862 macros that avoid raising exceptions for unordered operands. They have
5863 the same names as the standard macros ( @code{isgreater},
5864 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5865 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5866 prefixed. We intend for a library implementor to be able to simply
5867 @code{#define} each standard macro to its built-in equivalent.
5869 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5871 You can use the built-in function @code{__builtin_types_compatible_p} to
5872 determine whether two types are the same.
5874 This built-in function returns 1 if the unqualified versions of the
5875 types @var{type1} and @var{type2} (which are types, not expressions) are
5876 compatible, 0 otherwise. The result of this built-in function can be
5877 used in integer constant expressions.
5879 This built-in function ignores top level qualifiers (e.g., @code{const},
5880 @code{volatile}). For example, @code{int} is equivalent to @code{const
5883 The type @code{int[]} and @code{int[5]} are compatible. On the other
5884 hand, @code{int} and @code{char *} are not compatible, even if the size
5885 of their types, on the particular architecture are the same. Also, the
5886 amount of pointer indirection is taken into account when determining
5887 similarity. Consequently, @code{short *} is not similar to
5888 @code{short **}. Furthermore, two types that are typedefed are
5889 considered compatible if their underlying types are compatible.
5891 An @code{enum} type is not considered to be compatible with another
5892 @code{enum} type even if both are compatible with the same integer
5893 type; this is what the C standard specifies.
5894 For example, @code{enum @{foo, bar@}} is not similar to
5895 @code{enum @{hot, dog@}}.
5897 You would typically use this function in code whose execution varies
5898 depending on the arguments' types. For example:
5903 typeof (x) tmp = (x); \
5904 if (__builtin_types_compatible_p (typeof (x), long double)) \
5905 tmp = foo_long_double (tmp); \
5906 else if (__builtin_types_compatible_p (typeof (x), double)) \
5907 tmp = foo_double (tmp); \
5908 else if (__builtin_types_compatible_p (typeof (x), float)) \
5909 tmp = foo_float (tmp); \
5916 @emph{Note:} This construct is only available for C@.
5920 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5922 You can use the built-in function @code{__builtin_choose_expr} to
5923 evaluate code depending on the value of a constant expression. This
5924 built-in function returns @var{exp1} if @var{const_exp}, which is a
5925 constant expression that must be able to be determined at compile time,
5926 is nonzero. Otherwise it returns 0.
5928 This built-in function is analogous to the @samp{? :} operator in C,
5929 except that the expression returned has its type unaltered by promotion
5930 rules. Also, the built-in function does not evaluate the expression
5931 that was not chosen. For example, if @var{const_exp} evaluates to true,
5932 @var{exp2} is not evaluated even if it has side-effects.
5934 This built-in function can return an lvalue if the chosen argument is an
5937 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5938 type. Similarly, if @var{exp2} is returned, its return type is the same
5945 __builtin_choose_expr ( \
5946 __builtin_types_compatible_p (typeof (x), double), \
5948 __builtin_choose_expr ( \
5949 __builtin_types_compatible_p (typeof (x), float), \
5951 /* @r{The void expression results in a compile-time error} \
5952 @r{when assigning the result to something.} */ \
5956 @emph{Note:} This construct is only available for C@. Furthermore, the
5957 unused expression (@var{exp1} or @var{exp2} depending on the value of
5958 @var{const_exp}) may still generate syntax errors. This may change in
5963 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5964 You can use the built-in function @code{__builtin_constant_p} to
5965 determine if a value is known to be constant at compile-time and hence
5966 that GCC can perform constant-folding on expressions involving that
5967 value. The argument of the function is the value to test. The function
5968 returns the integer 1 if the argument is known to be a compile-time
5969 constant and 0 if it is not known to be a compile-time constant. A
5970 return of 0 does not indicate that the value is @emph{not} a constant,
5971 but merely that GCC cannot prove it is a constant with the specified
5972 value of the @option{-O} option.
5974 You would typically use this function in an embedded application where
5975 memory was a critical resource. If you have some complex calculation,
5976 you may want it to be folded if it involves constants, but need to call
5977 a function if it does not. For example:
5980 #define Scale_Value(X) \
5981 (__builtin_constant_p (X) \
5982 ? ((X) * SCALE + OFFSET) : Scale (X))
5985 You may use this built-in function in either a macro or an inline
5986 function. However, if you use it in an inlined function and pass an
5987 argument of the function as the argument to the built-in, GCC will
5988 never return 1 when you call the inline function with a string constant
5989 or compound literal (@pxref{Compound Literals}) and will not return 1
5990 when you pass a constant numeric value to the inline function unless you
5991 specify the @option{-O} option.
5993 You may also use @code{__builtin_constant_p} in initializers for static
5994 data. For instance, you can write
5997 static const int table[] = @{
5998 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6004 This is an acceptable initializer even if @var{EXPRESSION} is not a
6005 constant expression. GCC must be more conservative about evaluating the
6006 built-in in this case, because it has no opportunity to perform
6009 Previous versions of GCC did not accept this built-in in data
6010 initializers. The earliest version where it is completely safe is
6014 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6015 @opindex fprofile-arcs
6016 You may use @code{__builtin_expect} to provide the compiler with
6017 branch prediction information. In general, you should prefer to
6018 use actual profile feedback for this (@option{-fprofile-arcs}), as
6019 programmers are notoriously bad at predicting how their programs
6020 actually perform. However, there are applications in which this
6021 data is hard to collect.
6023 The return value is the value of @var{exp}, which should be an
6024 integral expression. The value of @var{c} must be a compile-time
6025 constant. The semantics of the built-in are that it is expected
6026 that @var{exp} == @var{c}. For example:
6029 if (__builtin_expect (x, 0))
6034 would indicate that we do not expect to call @code{foo}, since
6035 we expect @code{x} to be zero. Since you are limited to integral
6036 expressions for @var{exp}, you should use constructions such as
6039 if (__builtin_expect (ptr != NULL, 1))
6044 when testing pointer or floating-point values.
6047 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6048 This function is used to minimize cache-miss latency by moving data into
6049 a cache before it is accessed.
6050 You can insert calls to @code{__builtin_prefetch} into code for which
6051 you know addresses of data in memory that is likely to be accessed soon.
6052 If the target supports them, data prefetch instructions will be generated.
6053 If the prefetch is done early enough before the access then the data will
6054 be in the cache by the time it is accessed.
6056 The value of @var{addr} is the address of the memory to prefetch.
6057 There are two optional arguments, @var{rw} and @var{locality}.
6058 The value of @var{rw} is a compile-time constant one or zero; one
6059 means that the prefetch is preparing for a write to the memory address
6060 and zero, the default, means that the prefetch is preparing for a read.
6061 The value @var{locality} must be a compile-time constant integer between
6062 zero and three. A value of zero means that the data has no temporal
6063 locality, so it need not be left in the cache after the access. A value
6064 of three means that the data has a high degree of temporal locality and
6065 should be left in all levels of cache possible. Values of one and two
6066 mean, respectively, a low or moderate degree of temporal locality. The
6070 for (i = 0; i < n; i++)
6073 __builtin_prefetch (&a[i+j], 1, 1);
6074 __builtin_prefetch (&b[i+j], 0, 1);
6079 Data prefetch does not generate faults if @var{addr} is invalid, but
6080 the address expression itself must be valid. For example, a prefetch
6081 of @code{p->next} will not fault if @code{p->next} is not a valid
6082 address, but evaluation will fault if @code{p} is not a valid address.
6084 If the target does not support data prefetch, the address expression
6085 is evaluated if it includes side effects but no other code is generated
6086 and GCC does not issue a warning.
6089 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6090 Returns a positive infinity, if supported by the floating-point format,
6091 else @code{DBL_MAX}. This function is suitable for implementing the
6092 ISO C macro @code{HUGE_VAL}.
6095 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6096 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6099 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6100 Similar to @code{__builtin_huge_val}, except the return
6101 type is @code{long double}.
6104 @deftypefn {Built-in Function} double __builtin_inf (void)
6105 Similar to @code{__builtin_huge_val}, except a warning is generated
6106 if the target floating-point format does not support infinities.
6109 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6110 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6113 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6114 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6117 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6118 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6121 @deftypefn {Built-in Function} float __builtin_inff (void)
6122 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6123 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6126 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6127 Similar to @code{__builtin_inf}, except the return
6128 type is @code{long double}.
6131 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6132 This is an implementation of the ISO C99 function @code{nan}.
6134 Since ISO C99 defines this function in terms of @code{strtod}, which we
6135 do not implement, a description of the parsing is in order. The string
6136 is parsed as by @code{strtol}; that is, the base is recognized by
6137 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6138 in the significand such that the least significant bit of the number
6139 is at the least significant bit of the significand. The number is
6140 truncated to fit the significand field provided. The significand is
6141 forced to be a quiet NaN@.
6143 This function, if given a string literal all of which would have been
6144 consumed by strtol, is evaluated early enough that it is considered a
6145 compile-time constant.
6148 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6149 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6152 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6153 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6156 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6157 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6160 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6161 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6164 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6165 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6168 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6169 Similar to @code{__builtin_nan}, except the significand is forced
6170 to be a signaling NaN@. The @code{nans} function is proposed by
6171 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6174 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6175 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6178 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6179 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6182 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6183 Returns one plus the index of the least significant 1-bit of @var{x}, or
6184 if @var{x} is zero, returns zero.
6187 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6188 Returns the number of leading 0-bits in @var{x}, starting at the most
6189 significant bit position. If @var{x} is 0, the result is undefined.
6192 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6193 Returns the number of trailing 0-bits in @var{x}, starting at the least
6194 significant bit position. If @var{x} is 0, the result is undefined.
6197 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6198 Returns the number of 1-bits in @var{x}.
6201 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6202 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6206 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6207 Similar to @code{__builtin_ffs}, except the argument type is
6208 @code{unsigned long}.
6211 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6212 Similar to @code{__builtin_clz}, except the argument type is
6213 @code{unsigned long}.
6216 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6217 Similar to @code{__builtin_ctz}, except the argument type is
6218 @code{unsigned long}.
6221 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6222 Similar to @code{__builtin_popcount}, except the argument type is
6223 @code{unsigned long}.
6226 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6227 Similar to @code{__builtin_parity}, except the argument type is
6228 @code{unsigned long}.
6231 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6232 Similar to @code{__builtin_ffs}, except the argument type is
6233 @code{unsigned long long}.
6236 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6237 Similar to @code{__builtin_clz}, except the argument type is
6238 @code{unsigned long long}.
6241 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6242 Similar to @code{__builtin_ctz}, except the argument type is
6243 @code{unsigned long long}.
6246 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6247 Similar to @code{__builtin_popcount}, except the argument type is
6248 @code{unsigned long long}.
6251 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6252 Similar to @code{__builtin_parity}, except the argument type is
6253 @code{unsigned long long}.
6256 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6257 Returns the first argument raised to the power of the second. Unlike the
6258 @code{pow} function no guarantees about precision and rounding are made.
6261 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6262 Similar to @code{__builtin_powi}, except the argument and return types
6266 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6267 Similar to @code{__builtin_powi}, except the argument and return types
6268 are @code{long double}.
6271 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6272 Returns @var{x} with the order of the bytes reversed; for example,
6273 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6277 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6278 Similar to @code{__builtin_bswap32}, except the argument and return types
6282 @node Target Builtins
6283 @section Built-in Functions Specific to Particular Target Machines
6285 On some target machines, GCC supports many built-in functions specific
6286 to those machines. Generally these generate calls to specific machine
6287 instructions, but allow the compiler to schedule those calls.
6290 * Alpha Built-in Functions::
6291 * ARM Built-in Functions::
6292 * Blackfin Built-in Functions::
6293 * FR-V Built-in Functions::
6294 * X86 Built-in Functions::
6295 * MIPS DSP Built-in Functions::
6296 * MIPS Paired-Single Support::
6297 * PowerPC AltiVec Built-in Functions::
6298 * SPARC VIS Built-in Functions::
6301 @node Alpha Built-in Functions
6302 @subsection Alpha Built-in Functions
6304 These built-in functions are available for the Alpha family of
6305 processors, depending on the command-line switches used.
6307 The following built-in functions are always available. They
6308 all generate the machine instruction that is part of the name.
6311 long __builtin_alpha_implver (void)
6312 long __builtin_alpha_rpcc (void)
6313 long __builtin_alpha_amask (long)
6314 long __builtin_alpha_cmpbge (long, long)
6315 long __builtin_alpha_extbl (long, long)
6316 long __builtin_alpha_extwl (long, long)
6317 long __builtin_alpha_extll (long, long)
6318 long __builtin_alpha_extql (long, long)
6319 long __builtin_alpha_extwh (long, long)
6320 long __builtin_alpha_extlh (long, long)
6321 long __builtin_alpha_extqh (long, long)
6322 long __builtin_alpha_insbl (long, long)
6323 long __builtin_alpha_inswl (long, long)
6324 long __builtin_alpha_insll (long, long)
6325 long __builtin_alpha_insql (long, long)
6326 long __builtin_alpha_inswh (long, long)
6327 long __builtin_alpha_inslh (long, long)
6328 long __builtin_alpha_insqh (long, long)
6329 long __builtin_alpha_mskbl (long, long)
6330 long __builtin_alpha_mskwl (long, long)
6331 long __builtin_alpha_mskll (long, long)
6332 long __builtin_alpha_mskql (long, long)
6333 long __builtin_alpha_mskwh (long, long)
6334 long __builtin_alpha_msklh (long, long)
6335 long __builtin_alpha_mskqh (long, long)
6336 long __builtin_alpha_umulh (long, long)
6337 long __builtin_alpha_zap (long, long)
6338 long __builtin_alpha_zapnot (long, long)
6341 The following built-in functions are always with @option{-mmax}
6342 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6343 later. They all generate the machine instruction that is part
6347 long __builtin_alpha_pklb (long)
6348 long __builtin_alpha_pkwb (long)
6349 long __builtin_alpha_unpkbl (long)
6350 long __builtin_alpha_unpkbw (long)
6351 long __builtin_alpha_minub8 (long, long)
6352 long __builtin_alpha_minsb8 (long, long)
6353 long __builtin_alpha_minuw4 (long, long)
6354 long __builtin_alpha_minsw4 (long, long)
6355 long __builtin_alpha_maxub8 (long, long)
6356 long __builtin_alpha_maxsb8 (long, long)
6357 long __builtin_alpha_maxuw4 (long, long)
6358 long __builtin_alpha_maxsw4 (long, long)
6359 long __builtin_alpha_perr (long, long)
6362 The following built-in functions are always with @option{-mcix}
6363 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6364 later. They all generate the machine instruction that is part
6368 long __builtin_alpha_cttz (long)
6369 long __builtin_alpha_ctlz (long)
6370 long __builtin_alpha_ctpop (long)
6373 The following builtins are available on systems that use the OSF/1
6374 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6375 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6376 @code{rdval} and @code{wrval}.
6379 void *__builtin_thread_pointer (void)
6380 void __builtin_set_thread_pointer (void *)
6383 @node ARM Built-in Functions
6384 @subsection ARM Built-in Functions
6386 These built-in functions are available for the ARM family of
6387 processors, when the @option{-mcpu=iwmmxt} switch is used:
6390 typedef int v2si __attribute__ ((vector_size (8)));
6391 typedef short v4hi __attribute__ ((vector_size (8)));
6392 typedef char v8qi __attribute__ ((vector_size (8)));
6394 int __builtin_arm_getwcx (int)
6395 void __builtin_arm_setwcx (int, int)
6396 int __builtin_arm_textrmsb (v8qi, int)
6397 int __builtin_arm_textrmsh (v4hi, int)
6398 int __builtin_arm_textrmsw (v2si, int)
6399 int __builtin_arm_textrmub (v8qi, int)
6400 int __builtin_arm_textrmuh (v4hi, int)
6401 int __builtin_arm_textrmuw (v2si, int)
6402 v8qi __builtin_arm_tinsrb (v8qi, int)
6403 v4hi __builtin_arm_tinsrh (v4hi, int)
6404 v2si __builtin_arm_tinsrw (v2si, int)
6405 long long __builtin_arm_tmia (long long, int, int)
6406 long long __builtin_arm_tmiabb (long long, int, int)
6407 long long __builtin_arm_tmiabt (long long, int, int)
6408 long long __builtin_arm_tmiaph (long long, int, int)
6409 long long __builtin_arm_tmiatb (long long, int, int)
6410 long long __builtin_arm_tmiatt (long long, int, int)
6411 int __builtin_arm_tmovmskb (v8qi)
6412 int __builtin_arm_tmovmskh (v4hi)
6413 int __builtin_arm_tmovmskw (v2si)
6414 long long __builtin_arm_waccb (v8qi)
6415 long long __builtin_arm_wacch (v4hi)
6416 long long __builtin_arm_waccw (v2si)
6417 v8qi __builtin_arm_waddb (v8qi, v8qi)
6418 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6419 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6420 v4hi __builtin_arm_waddh (v4hi, v4hi)
6421 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6422 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6423 v2si __builtin_arm_waddw (v2si, v2si)
6424 v2si __builtin_arm_waddwss (v2si, v2si)
6425 v2si __builtin_arm_waddwus (v2si, v2si)
6426 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6427 long long __builtin_arm_wand(long long, long long)
6428 long long __builtin_arm_wandn (long long, long long)
6429 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6430 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6431 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6432 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6433 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6434 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6435 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6436 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6437 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6438 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6439 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6440 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6441 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6442 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6443 long long __builtin_arm_wmacsz (v4hi, v4hi)
6444 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6445 long long __builtin_arm_wmacuz (v4hi, v4hi)
6446 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6447 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6448 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6449 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6450 v2si __builtin_arm_wmaxsw (v2si, v2si)
6451 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6452 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6453 v2si __builtin_arm_wmaxuw (v2si, v2si)
6454 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6455 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6456 v2si __builtin_arm_wminsw (v2si, v2si)
6457 v8qi __builtin_arm_wminub (v8qi, v8qi)
6458 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6459 v2si __builtin_arm_wminuw (v2si, v2si)
6460 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6461 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6462 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6463 long long __builtin_arm_wor (long long, long long)
6464 v2si __builtin_arm_wpackdss (long long, long long)
6465 v2si __builtin_arm_wpackdus (long long, long long)
6466 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6467 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6468 v4hi __builtin_arm_wpackwss (v2si, v2si)
6469 v4hi __builtin_arm_wpackwus (v2si, v2si)
6470 long long __builtin_arm_wrord (long long, long long)
6471 long long __builtin_arm_wrordi (long long, int)
6472 v4hi __builtin_arm_wrorh (v4hi, long long)
6473 v4hi __builtin_arm_wrorhi (v4hi, int)
6474 v2si __builtin_arm_wrorw (v2si, long long)
6475 v2si __builtin_arm_wrorwi (v2si, int)
6476 v2si __builtin_arm_wsadb (v8qi, v8qi)
6477 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6478 v2si __builtin_arm_wsadh (v4hi, v4hi)
6479 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6480 v4hi __builtin_arm_wshufh (v4hi, int)
6481 long long __builtin_arm_wslld (long long, long long)
6482 long long __builtin_arm_wslldi (long long, int)
6483 v4hi __builtin_arm_wsllh (v4hi, long long)
6484 v4hi __builtin_arm_wsllhi (v4hi, int)
6485 v2si __builtin_arm_wsllw (v2si, long long)
6486 v2si __builtin_arm_wsllwi (v2si, int)
6487 long long __builtin_arm_wsrad (long long, long long)
6488 long long __builtin_arm_wsradi (long long, int)
6489 v4hi __builtin_arm_wsrah (v4hi, long long)
6490 v4hi __builtin_arm_wsrahi (v4hi, int)
6491 v2si __builtin_arm_wsraw (v2si, long long)
6492 v2si __builtin_arm_wsrawi (v2si, int)
6493 long long __builtin_arm_wsrld (long long, long long)
6494 long long __builtin_arm_wsrldi (long long, int)
6495 v4hi __builtin_arm_wsrlh (v4hi, long long)
6496 v4hi __builtin_arm_wsrlhi (v4hi, int)
6497 v2si __builtin_arm_wsrlw (v2si, long long)
6498 v2si __builtin_arm_wsrlwi (v2si, int)
6499 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6500 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6501 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6502 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6503 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6504 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6505 v2si __builtin_arm_wsubw (v2si, v2si)
6506 v2si __builtin_arm_wsubwss (v2si, v2si)
6507 v2si __builtin_arm_wsubwus (v2si, v2si)
6508 v4hi __builtin_arm_wunpckehsb (v8qi)
6509 v2si __builtin_arm_wunpckehsh (v4hi)
6510 long long __builtin_arm_wunpckehsw (v2si)
6511 v4hi __builtin_arm_wunpckehub (v8qi)
6512 v2si __builtin_arm_wunpckehuh (v4hi)
6513 long long __builtin_arm_wunpckehuw (v2si)
6514 v4hi __builtin_arm_wunpckelsb (v8qi)
6515 v2si __builtin_arm_wunpckelsh (v4hi)
6516 long long __builtin_arm_wunpckelsw (v2si)
6517 v4hi __builtin_arm_wunpckelub (v8qi)
6518 v2si __builtin_arm_wunpckeluh (v4hi)
6519 long long __builtin_arm_wunpckeluw (v2si)
6520 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6521 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6522 v2si __builtin_arm_wunpckihw (v2si, v2si)
6523 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6524 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6525 v2si __builtin_arm_wunpckilw (v2si, v2si)
6526 long long __builtin_arm_wxor (long long, long long)
6527 long long __builtin_arm_wzero ()
6530 @node Blackfin Built-in Functions
6531 @subsection Blackfin Built-in Functions
6533 Currently, there are two Blackfin-specific built-in functions. These are
6534 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6535 using inline assembly; by using these built-in functions the compiler can
6536 automatically add workarounds for hardware errata involving these
6537 instructions. These functions are named as follows:
6540 void __builtin_bfin_csync (void)
6541 void __builtin_bfin_ssync (void)
6544 @node FR-V Built-in Functions
6545 @subsection FR-V Built-in Functions
6547 GCC provides many FR-V-specific built-in functions. In general,
6548 these functions are intended to be compatible with those described
6549 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6550 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6551 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6552 pointer rather than by value.
6554 Most of the functions are named after specific FR-V instructions.
6555 Such functions are said to be ``directly mapped'' and are summarized
6556 here in tabular form.
6560 * Directly-mapped Integer Functions::
6561 * Directly-mapped Media Functions::
6562 * Raw read/write Functions::
6563 * Other Built-in Functions::
6566 @node Argument Types
6567 @subsubsection Argument Types
6569 The arguments to the built-in functions can be divided into three groups:
6570 register numbers, compile-time constants and run-time values. In order
6571 to make this classification clear at a glance, the arguments and return
6572 values are given the following pseudo types:
6574 @multitable @columnfractions .20 .30 .15 .35
6575 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6576 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6577 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6578 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6579 @item @code{uw2} @tab @code{unsigned long long} @tab No
6580 @tab an unsigned doubleword
6581 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6582 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6583 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6584 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6587 These pseudo types are not defined by GCC, they are simply a notational
6588 convenience used in this manual.
6590 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6591 and @code{sw2} are evaluated at run time. They correspond to
6592 register operands in the underlying FR-V instructions.
6594 @code{const} arguments represent immediate operands in the underlying
6595 FR-V instructions. They must be compile-time constants.
6597 @code{acc} arguments are evaluated at compile time and specify the number
6598 of an accumulator register. For example, an @code{acc} argument of 2
6599 will select the ACC2 register.
6601 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6602 number of an IACC register. See @pxref{Other Built-in Functions}
6605 @node Directly-mapped Integer Functions
6606 @subsubsection Directly-mapped Integer Functions
6608 The functions listed below map directly to FR-V I-type instructions.
6610 @multitable @columnfractions .45 .32 .23
6611 @item Function prototype @tab Example usage @tab Assembly output
6612 @item @code{sw1 __ADDSS (sw1, sw1)}
6613 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6614 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6615 @item @code{sw1 __SCAN (sw1, sw1)}
6616 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6617 @tab @code{SCAN @var{a},@var{b},@var{c}}
6618 @item @code{sw1 __SCUTSS (sw1)}
6619 @tab @code{@var{b} = __SCUTSS (@var{a})}
6620 @tab @code{SCUTSS @var{a},@var{b}}
6621 @item @code{sw1 __SLASS (sw1, sw1)}
6622 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6623 @tab @code{SLASS @var{a},@var{b},@var{c}}
6624 @item @code{void __SMASS (sw1, sw1)}
6625 @tab @code{__SMASS (@var{a}, @var{b})}
6626 @tab @code{SMASS @var{a},@var{b}}
6627 @item @code{void __SMSSS (sw1, sw1)}
6628 @tab @code{__SMSSS (@var{a}, @var{b})}
6629 @tab @code{SMSSS @var{a},@var{b}}
6630 @item @code{void __SMU (sw1, sw1)}
6631 @tab @code{__SMU (@var{a}, @var{b})}
6632 @tab @code{SMU @var{a},@var{b}}
6633 @item @code{sw2 __SMUL (sw1, sw1)}
6634 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6635 @tab @code{SMUL @var{a},@var{b},@var{c}}
6636 @item @code{sw1 __SUBSS (sw1, sw1)}
6637 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6638 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6639 @item @code{uw2 __UMUL (uw1, uw1)}
6640 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6641 @tab @code{UMUL @var{a},@var{b},@var{c}}
6644 @node Directly-mapped Media Functions
6645 @subsubsection Directly-mapped Media Functions
6647 The functions listed below map directly to FR-V M-type instructions.
6649 @multitable @columnfractions .45 .32 .23
6650 @item Function prototype @tab Example usage @tab Assembly output
6651 @item @code{uw1 __MABSHS (sw1)}
6652 @tab @code{@var{b} = __MABSHS (@var{a})}
6653 @tab @code{MABSHS @var{a},@var{b}}
6654 @item @code{void __MADDACCS (acc, acc)}
6655 @tab @code{__MADDACCS (@var{b}, @var{a})}
6656 @tab @code{MADDACCS @var{a},@var{b}}
6657 @item @code{sw1 __MADDHSS (sw1, sw1)}
6658 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6659 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6660 @item @code{uw1 __MADDHUS (uw1, uw1)}
6661 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6662 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6663 @item @code{uw1 __MAND (uw1, uw1)}
6664 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6665 @tab @code{MAND @var{a},@var{b},@var{c}}
6666 @item @code{void __MASACCS (acc, acc)}
6667 @tab @code{__MASACCS (@var{b}, @var{a})}
6668 @tab @code{MASACCS @var{a},@var{b}}
6669 @item @code{uw1 __MAVEH (uw1, uw1)}
6670 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6671 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6672 @item @code{uw2 __MBTOH (uw1)}
6673 @tab @code{@var{b} = __MBTOH (@var{a})}
6674 @tab @code{MBTOH @var{a},@var{b}}
6675 @item @code{void __MBTOHE (uw1 *, uw1)}
6676 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6677 @tab @code{MBTOHE @var{a},@var{b}}
6678 @item @code{void __MCLRACC (acc)}
6679 @tab @code{__MCLRACC (@var{a})}
6680 @tab @code{MCLRACC @var{a}}
6681 @item @code{void __MCLRACCA (void)}
6682 @tab @code{__MCLRACCA ()}
6683 @tab @code{MCLRACCA}
6684 @item @code{uw1 __Mcop1 (uw1, uw1)}
6685 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6686 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6687 @item @code{uw1 __Mcop2 (uw1, uw1)}
6688 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6689 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6690 @item @code{uw1 __MCPLHI (uw2, const)}
6691 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6692 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6693 @item @code{uw1 __MCPLI (uw2, const)}
6694 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6695 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6696 @item @code{void __MCPXIS (acc, sw1, sw1)}
6697 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6698 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6699 @item @code{void __MCPXIU (acc, uw1, uw1)}
6700 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6701 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6702 @item @code{void __MCPXRS (acc, sw1, sw1)}
6703 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6704 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6705 @item @code{void __MCPXRU (acc, uw1, uw1)}
6706 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6707 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6708 @item @code{uw1 __MCUT (acc, uw1)}
6709 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6710 @tab @code{MCUT @var{a},@var{b},@var{c}}
6711 @item @code{uw1 __MCUTSS (acc, sw1)}
6712 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6713 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6714 @item @code{void __MDADDACCS (acc, acc)}
6715 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6716 @tab @code{MDADDACCS @var{a},@var{b}}
6717 @item @code{void __MDASACCS (acc, acc)}
6718 @tab @code{__MDASACCS (@var{b}, @var{a})}
6719 @tab @code{MDASACCS @var{a},@var{b}}
6720 @item @code{uw2 __MDCUTSSI (acc, const)}
6721 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6722 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6723 @item @code{uw2 __MDPACKH (uw2, uw2)}
6724 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6725 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6726 @item @code{uw2 __MDROTLI (uw2, const)}
6727 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6728 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6729 @item @code{void __MDSUBACCS (acc, acc)}
6730 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6731 @tab @code{MDSUBACCS @var{a},@var{b}}
6732 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6733 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6734 @tab @code{MDUNPACKH @var{a},@var{b}}
6735 @item @code{uw2 __MEXPDHD (uw1, const)}
6736 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6737 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6738 @item @code{uw1 __MEXPDHW (uw1, const)}
6739 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6740 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6741 @item @code{uw1 __MHDSETH (uw1, const)}
6742 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6743 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6744 @item @code{sw1 __MHDSETS (const)}
6745 @tab @code{@var{b} = __MHDSETS (@var{a})}
6746 @tab @code{MHDSETS #@var{a},@var{b}}
6747 @item @code{uw1 __MHSETHIH (uw1, const)}
6748 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6749 @tab @code{MHSETHIH #@var{a},@var{b}}
6750 @item @code{sw1 __MHSETHIS (sw1, const)}
6751 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6752 @tab @code{MHSETHIS #@var{a},@var{b}}
6753 @item @code{uw1 __MHSETLOH (uw1, const)}
6754 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6755 @tab @code{MHSETLOH #@var{a},@var{b}}
6756 @item @code{sw1 __MHSETLOS (sw1, const)}
6757 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6758 @tab @code{MHSETLOS #@var{a},@var{b}}
6759 @item @code{uw1 __MHTOB (uw2)}
6760 @tab @code{@var{b} = __MHTOB (@var{a})}
6761 @tab @code{MHTOB @var{a},@var{b}}
6762 @item @code{void __MMACHS (acc, sw1, sw1)}
6763 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6764 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6765 @item @code{void __MMACHU (acc, uw1, uw1)}
6766 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6767 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6768 @item @code{void __MMRDHS (acc, sw1, sw1)}
6769 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6770 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6771 @item @code{void __MMRDHU (acc, uw1, uw1)}
6772 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6773 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6774 @item @code{void __MMULHS (acc, sw1, sw1)}
6775 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6776 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6777 @item @code{void __MMULHU (acc, uw1, uw1)}
6778 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6779 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6780 @item @code{void __MMULXHS (acc, sw1, sw1)}
6781 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6782 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6783 @item @code{void __MMULXHU (acc, uw1, uw1)}
6784 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6785 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6786 @item @code{uw1 __MNOT (uw1)}
6787 @tab @code{@var{b} = __MNOT (@var{a})}
6788 @tab @code{MNOT @var{a},@var{b}}
6789 @item @code{uw1 __MOR (uw1, uw1)}
6790 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6791 @tab @code{MOR @var{a},@var{b},@var{c}}
6792 @item @code{uw1 __MPACKH (uh, uh)}
6793 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6794 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6795 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6796 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6797 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6798 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6799 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6800 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6801 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6802 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6803 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6804 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6805 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6806 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6807 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6808 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6809 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6810 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6811 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6812 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6813 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6814 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6815 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6816 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6817 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6818 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6819 @item @code{void __MQMACHS (acc, sw2, sw2)}
6820 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6821 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6822 @item @code{void __MQMACHU (acc, uw2, uw2)}
6823 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6824 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6825 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6826 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6827 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6828 @item @code{void __MQMULHS (acc, sw2, sw2)}
6829 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6830 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6831 @item @code{void __MQMULHU (acc, uw2, uw2)}
6832 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6833 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6834 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6835 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6836 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6837 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6838 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6839 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6840 @item @code{sw2 __MQSATHS (sw2, sw2)}
6841 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6842 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6843 @item @code{uw2 __MQSLLHI (uw2, int)}
6844 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6845 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6846 @item @code{sw2 __MQSRAHI (sw2, int)}
6847 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6848 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6849 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6850 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6851 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6852 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6853 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6854 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6855 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6856 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6857 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6858 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6859 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6860 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6861 @item @code{uw1 __MRDACC (acc)}
6862 @tab @code{@var{b} = __MRDACC (@var{a})}
6863 @tab @code{MRDACC @var{a},@var{b}}
6864 @item @code{uw1 __MRDACCG (acc)}
6865 @tab @code{@var{b} = __MRDACCG (@var{a})}
6866 @tab @code{MRDACCG @var{a},@var{b}}
6867 @item @code{uw1 __MROTLI (uw1, const)}
6868 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6869 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6870 @item @code{uw1 __MROTRI (uw1, const)}
6871 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6872 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6873 @item @code{sw1 __MSATHS (sw1, sw1)}
6874 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6875 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6876 @item @code{uw1 __MSATHU (uw1, uw1)}
6877 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6878 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6879 @item @code{uw1 __MSLLHI (uw1, const)}
6880 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6881 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6882 @item @code{sw1 __MSRAHI (sw1, const)}
6883 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6884 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6885 @item @code{uw1 __MSRLHI (uw1, const)}
6886 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6887 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6888 @item @code{void __MSUBACCS (acc, acc)}
6889 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6890 @tab @code{MSUBACCS @var{a},@var{b}}
6891 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6892 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6893 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6894 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6895 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6896 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6897 @item @code{void __MTRAP (void)}
6898 @tab @code{__MTRAP ()}
6900 @item @code{uw2 __MUNPACKH (uw1)}
6901 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6902 @tab @code{MUNPACKH @var{a},@var{b}}
6903 @item @code{uw1 __MWCUT (uw2, uw1)}
6904 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6905 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6906 @item @code{void __MWTACC (acc, uw1)}
6907 @tab @code{__MWTACC (@var{b}, @var{a})}
6908 @tab @code{MWTACC @var{a},@var{b}}
6909 @item @code{void __MWTACCG (acc, uw1)}
6910 @tab @code{__MWTACCG (@var{b}, @var{a})}
6911 @tab @code{MWTACCG @var{a},@var{b}}
6912 @item @code{uw1 __MXOR (uw1, uw1)}
6913 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6914 @tab @code{MXOR @var{a},@var{b},@var{c}}
6917 @node Raw read/write Functions
6918 @subsubsection Raw read/write Functions
6920 This sections describes built-in functions related to read and write
6921 instructions to access memory. These functions generate
6922 @code{membar} instructions to flush the I/O load and stores where
6923 appropriate, as described in Fujitsu's manual described above.
6927 @item unsigned char __builtin_read8 (void *@var{data})
6928 @item unsigned short __builtin_read16 (void *@var{data})
6929 @item unsigned long __builtin_read32 (void *@var{data})
6930 @item unsigned long long __builtin_read64 (void *@var{data})
6932 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6933 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6934 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6935 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6938 @node Other Built-in Functions
6939 @subsubsection Other Built-in Functions
6941 This section describes built-in functions that are not named after
6942 a specific FR-V instruction.
6945 @item sw2 __IACCreadll (iacc @var{reg})
6946 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6947 for future expansion and must be 0.
6949 @item sw1 __IACCreadl (iacc @var{reg})
6950 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6951 Other values of @var{reg} are rejected as invalid.
6953 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6954 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6955 is reserved for future expansion and must be 0.
6957 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6958 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6959 is 1. Other values of @var{reg} are rejected as invalid.
6961 @item void __data_prefetch0 (const void *@var{x})
6962 Use the @code{dcpl} instruction to load the contents of address @var{x}
6963 into the data cache.
6965 @item void __data_prefetch (const void *@var{x})
6966 Use the @code{nldub} instruction to load the contents of address @var{x}
6967 into the data cache. The instruction will be issued in slot I1@.
6970 @node X86 Built-in Functions
6971 @subsection X86 Built-in Functions
6973 These built-in functions are available for the i386 and x86-64 family
6974 of computers, depending on the command-line switches used.
6976 Note that, if you specify command-line switches such as @option{-msse},
6977 the compiler could use the extended instruction sets even if the built-ins
6978 are not used explicitly in the program. For this reason, applications
6979 which perform runtime CPU detection must compile separate files for each
6980 supported architecture, using the appropriate flags. In particular,
6981 the file containing the CPU detection code should be compiled without
6984 The following machine modes are available for use with MMX built-in functions
6985 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6986 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6987 vector of eight 8-bit integers. Some of the built-in functions operate on
6988 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6990 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6991 of two 32-bit floating point values.
6993 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6994 floating point values. Some instructions use a vector of four 32-bit
6995 integers, these use @code{V4SI}. Finally, some instructions operate on an
6996 entire vector register, interpreting it as a 128-bit integer, these use mode
6999 The following built-in functions are made available by @option{-mmmx}.
7000 All of them generate the machine instruction that is part of the name.
7003 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7004 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7005 v2si __builtin_ia32_paddd (v2si, v2si)
7006 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7007 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7008 v2si __builtin_ia32_psubd (v2si, v2si)
7009 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7010 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7011 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7012 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7013 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7014 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7015 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7016 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7017 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7018 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7019 di __builtin_ia32_pand (di, di)
7020 di __builtin_ia32_pandn (di,di)
7021 di __builtin_ia32_por (di, di)
7022 di __builtin_ia32_pxor (di, di)
7023 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7024 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7025 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7026 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7027 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7028 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7029 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7030 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7031 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7032 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7033 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7034 v2si __builtin_ia32_punpckldq (v2si, v2si)
7035 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7036 v4hi __builtin_ia32_packssdw (v2si, v2si)
7037 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7040 The following built-in functions are made available either with
7041 @option{-msse}, or with a combination of @option{-m3dnow} and
7042 @option{-march=athlon}. All of them generate the machine
7043 instruction that is part of the name.
7046 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7047 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7048 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7049 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
7050 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7051 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7052 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7053 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7054 int __builtin_ia32_pextrw (v4hi, int)
7055 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7056 int __builtin_ia32_pmovmskb (v8qi)
7057 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7058 void __builtin_ia32_movntq (di *, di)
7059 void __builtin_ia32_sfence (void)
7062 The following built-in functions are available when @option{-msse} is used.
7063 All of them generate the machine instruction that is part of the name.
7066 int __builtin_ia32_comieq (v4sf, v4sf)
7067 int __builtin_ia32_comineq (v4sf, v4sf)
7068 int __builtin_ia32_comilt (v4sf, v4sf)
7069 int __builtin_ia32_comile (v4sf, v4sf)
7070 int __builtin_ia32_comigt (v4sf, v4sf)
7071 int __builtin_ia32_comige (v4sf, v4sf)
7072 int __builtin_ia32_ucomieq (v4sf, v4sf)
7073 int __builtin_ia32_ucomineq (v4sf, v4sf)
7074 int __builtin_ia32_ucomilt (v4sf, v4sf)
7075 int __builtin_ia32_ucomile (v4sf, v4sf)
7076 int __builtin_ia32_ucomigt (v4sf, v4sf)
7077 int __builtin_ia32_ucomige (v4sf, v4sf)
7078 v4sf __builtin_ia32_addps (v4sf, v4sf)
7079 v4sf __builtin_ia32_subps (v4sf, v4sf)
7080 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7081 v4sf __builtin_ia32_divps (v4sf, v4sf)
7082 v4sf __builtin_ia32_addss (v4sf, v4sf)
7083 v4sf __builtin_ia32_subss (v4sf, v4sf)
7084 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7085 v4sf __builtin_ia32_divss (v4sf, v4sf)
7086 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7087 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7088 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7089 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7090 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7091 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7092 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7093 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7094 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7095 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7096 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7097 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7098 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7099 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7100 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7101 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7102 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7103 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7104 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7105 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7106 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7107 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7108 v4sf __builtin_ia32_minps (v4sf, v4sf)
7109 v4sf __builtin_ia32_minss (v4sf, v4sf)
7110 v4sf __builtin_ia32_andps (v4sf, v4sf)
7111 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7112 v4sf __builtin_ia32_orps (v4sf, v4sf)
7113 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7114 v4sf __builtin_ia32_movss (v4sf, v4sf)
7115 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7116 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7117 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7118 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7119 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7120 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7121 v2si __builtin_ia32_cvtps2pi (v4sf)
7122 int __builtin_ia32_cvtss2si (v4sf)
7123 v2si __builtin_ia32_cvttps2pi (v4sf)
7124 int __builtin_ia32_cvttss2si (v4sf)
7125 v4sf __builtin_ia32_rcpps (v4sf)
7126 v4sf __builtin_ia32_rsqrtps (v4sf)
7127 v4sf __builtin_ia32_sqrtps (v4sf)
7128 v4sf __builtin_ia32_rcpss (v4sf)
7129 v4sf __builtin_ia32_rsqrtss (v4sf)
7130 v4sf __builtin_ia32_sqrtss (v4sf)
7131 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7132 void __builtin_ia32_movntps (float *, v4sf)
7133 int __builtin_ia32_movmskps (v4sf)
7136 The following built-in functions are available when @option{-msse} is used.
7139 @item v4sf __builtin_ia32_loadaps (float *)
7140 Generates the @code{movaps} machine instruction as a load from memory.
7141 @item void __builtin_ia32_storeaps (float *, v4sf)
7142 Generates the @code{movaps} machine instruction as a store to memory.
7143 @item v4sf __builtin_ia32_loadups (float *)
7144 Generates the @code{movups} machine instruction as a load from memory.
7145 @item void __builtin_ia32_storeups (float *, v4sf)
7146 Generates the @code{movups} machine instruction as a store to memory.
7147 @item v4sf __builtin_ia32_loadsss (float *)
7148 Generates the @code{movss} machine instruction as a load from memory.
7149 @item void __builtin_ia32_storess (float *, v4sf)
7150 Generates the @code{movss} machine instruction as a store to memory.
7151 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7152 Generates the @code{movhps} machine instruction as a load from memory.
7153 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7154 Generates the @code{movlps} machine instruction as a load from memory
7155 @item void __builtin_ia32_storehps (v4sf, v2si *)
7156 Generates the @code{movhps} machine instruction as a store to memory.
7157 @item void __builtin_ia32_storelps (v4sf, v2si *)
7158 Generates the @code{movlps} machine instruction as a store to memory.
7161 The following built-in functions are available when @option{-msse2} is used.
7162 All of them generate the machine instruction that is part of the name.
7165 int __builtin_ia32_comisdeq (v2df, v2df)
7166 int __builtin_ia32_comisdlt (v2df, v2df)
7167 int __builtin_ia32_comisdle (v2df, v2df)
7168 int __builtin_ia32_comisdgt (v2df, v2df)
7169 int __builtin_ia32_comisdge (v2df, v2df)
7170 int __builtin_ia32_comisdneq (v2df, v2df)
7171 int __builtin_ia32_ucomisdeq (v2df, v2df)
7172 int __builtin_ia32_ucomisdlt (v2df, v2df)
7173 int __builtin_ia32_ucomisdle (v2df, v2df)
7174 int __builtin_ia32_ucomisdgt (v2df, v2df)
7175 int __builtin_ia32_ucomisdge (v2df, v2df)
7176 int __builtin_ia32_ucomisdneq (v2df, v2df)
7177 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7178 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7179 v2df __builtin_ia32_cmplepd (v2df, v2df)
7180 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7181 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7182 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7183 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7184 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7185 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7186 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7187 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7188 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7189 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7190 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7191 v2df __builtin_ia32_cmplesd (v2df, v2df)
7192 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7193 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7194 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7195 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7196 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7197 v2di __builtin_ia32_paddq (v2di, v2di)
7198 v2di __builtin_ia32_psubq (v2di, v2di)
7199 v2df __builtin_ia32_addpd (v2df, v2df)
7200 v2df __builtin_ia32_subpd (v2df, v2df)
7201 v2df __builtin_ia32_mulpd (v2df, v2df)
7202 v2df __builtin_ia32_divpd (v2df, v2df)
7203 v2df __builtin_ia32_addsd (v2df, v2df)
7204 v2df __builtin_ia32_subsd (v2df, v2df)
7205 v2df __builtin_ia32_mulsd (v2df, v2df)
7206 v2df __builtin_ia32_divsd (v2df, v2df)
7207 v2df __builtin_ia32_minpd (v2df, v2df)
7208 v2df __builtin_ia32_maxpd (v2df, v2df)
7209 v2df __builtin_ia32_minsd (v2df, v2df)
7210 v2df __builtin_ia32_maxsd (v2df, v2df)
7211 v2df __builtin_ia32_andpd (v2df, v2df)
7212 v2df __builtin_ia32_andnpd (v2df, v2df)
7213 v2df __builtin_ia32_orpd (v2df, v2df)
7214 v2df __builtin_ia32_xorpd (v2df, v2df)
7215 v2df __builtin_ia32_movsd (v2df, v2df)
7216 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7217 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7218 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7219 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7220 v4si __builtin_ia32_paddd128 (v4si, v4si)
7221 v2di __builtin_ia32_paddq128 (v2di, v2di)
7222 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7223 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7224 v4si __builtin_ia32_psubd128 (v4si, v4si)
7225 v2di __builtin_ia32_psubq128 (v2di, v2di)
7226 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7227 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7228 v2di __builtin_ia32_pand128 (v2di, v2di)
7229 v2di __builtin_ia32_pandn128 (v2di, v2di)
7230 v2di __builtin_ia32_por128 (v2di, v2di)
7231 v2di __builtin_ia32_pxor128 (v2di, v2di)
7232 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7233 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7234 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7235 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7236 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7237 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7238 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7239 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7240 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7241 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7242 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7243 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7244 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7245 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7246 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7247 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7248 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7249 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7250 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7251 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7252 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7253 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7254 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7255 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7256 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7257 v2df __builtin_ia32_loadupd (double *)
7258 void __builtin_ia32_storeupd (double *, v2df)
7259 v2df __builtin_ia32_loadhpd (v2df, double *)
7260 v2df __builtin_ia32_loadlpd (v2df, double *)
7261 int __builtin_ia32_movmskpd (v2df)
7262 int __builtin_ia32_pmovmskb128 (v16qi)
7263 void __builtin_ia32_movnti (int *, int)
7264 void __builtin_ia32_movntpd (double *, v2df)
7265 void __builtin_ia32_movntdq (v2df *, v2df)
7266 v4si __builtin_ia32_pshufd (v4si, int)
7267 v8hi __builtin_ia32_pshuflw (v8hi, int)
7268 v8hi __builtin_ia32_pshufhw (v8hi, int)
7269 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7270 v2df __builtin_ia32_sqrtpd (v2df)
7271 v2df __builtin_ia32_sqrtsd (v2df)
7272 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7273 v2df __builtin_ia32_cvtdq2pd (v4si)
7274 v4sf __builtin_ia32_cvtdq2ps (v4si)
7275 v4si __builtin_ia32_cvtpd2dq (v2df)
7276 v2si __builtin_ia32_cvtpd2pi (v2df)
7277 v4sf __builtin_ia32_cvtpd2ps (v2df)
7278 v4si __builtin_ia32_cvttpd2dq (v2df)
7279 v2si __builtin_ia32_cvttpd2pi (v2df)
7280 v2df __builtin_ia32_cvtpi2pd (v2si)
7281 int __builtin_ia32_cvtsd2si (v2df)
7282 int __builtin_ia32_cvttsd2si (v2df)
7283 long long __builtin_ia32_cvtsd2si64 (v2df)
7284 long long __builtin_ia32_cvttsd2si64 (v2df)
7285 v4si __builtin_ia32_cvtps2dq (v4sf)
7286 v2df __builtin_ia32_cvtps2pd (v4sf)
7287 v4si __builtin_ia32_cvttps2dq (v4sf)
7288 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7289 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7290 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7291 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7292 void __builtin_ia32_clflush (const void *)
7293 void __builtin_ia32_lfence (void)
7294 void __builtin_ia32_mfence (void)
7295 v16qi __builtin_ia32_loaddqu (const char *)
7296 void __builtin_ia32_storedqu (char *, v16qi)
7297 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7298 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7299 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7300 v4si __builtin_ia32_pslld128 (v4si, v2di)
7301 v2di __builtin_ia32_psllq128 (v4si, v2di)
7302 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7303 v4si __builtin_ia32_psrld128 (v4si, v2di)
7304 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7305 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7306 v4si __builtin_ia32_psrad128 (v4si, v2di)
7307 v2di __builtin_ia32_pslldqi128 (v2di, int)
7308 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7309 v4si __builtin_ia32_pslldi128 (v4si, int)
7310 v2di __builtin_ia32_psllqi128 (v2di, int)
7311 v2di __builtin_ia32_psrldqi128 (v2di, int)
7312 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7313 v4si __builtin_ia32_psrldi128 (v4si, int)
7314 v2di __builtin_ia32_psrlqi128 (v2di, int)
7315 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7316 v4si __builtin_ia32_psradi128 (v4si, int)
7317 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7320 The following built-in functions are available when @option{-msse3} is used.
7321 All of them generate the machine instruction that is part of the name.
7324 v2df __builtin_ia32_addsubpd (v2df, v2df)
7325 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7326 v2df __builtin_ia32_haddpd (v2df, v2df)
7327 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7328 v2df __builtin_ia32_hsubpd (v2df, v2df)
7329 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7330 v16qi __builtin_ia32_lddqu (char const *)
7331 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7332 v2df __builtin_ia32_movddup (v2df)
7333 v4sf __builtin_ia32_movshdup (v4sf)
7334 v4sf __builtin_ia32_movsldup (v4sf)
7335 void __builtin_ia32_mwait (unsigned int, unsigned int)
7338 The following built-in functions are available when @option{-msse3} is used.
7341 @item v2df __builtin_ia32_loadddup (double const *)
7342 Generates the @code{movddup} machine instruction as a load from memory.
7345 The following built-in functions are available when @option{-mssse3} is used.
7346 All of them generate the machine instruction that is part of the name
7350 v2si __builtin_ia32_phaddd (v2si, v2si)
7351 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7352 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7353 v2si __builtin_ia32_phsubd (v2si, v2si)
7354 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7355 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7356 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7357 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7358 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7359 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7360 v2si __builtin_ia32_psignd (v2si, v2si)
7361 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7362 long long __builtin_ia32_palignr (long long, long long, int)
7363 v8qi __builtin_ia32_pabsb (v8qi)
7364 v2si __builtin_ia32_pabsd (v2si)
7365 v4hi __builtin_ia32_pabsw (v4hi)
7368 The following built-in functions are available when @option{-mssse3} is used.
7369 All of them generate the machine instruction that is part of the name
7373 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7374 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7375 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7376 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7377 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7378 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7379 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7380 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7381 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7382 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7383 v4si __builtin_ia32_psignd128 (v4si, v4si)
7384 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7385 v2di __builtin_ia32_palignr (v2di, v2di, int)
7386 v16qi __builtin_ia32_pabsb128 (v16qi)
7387 v4si __builtin_ia32_pabsd128 (v4si)
7388 v8hi __builtin_ia32_pabsw128 (v8hi)
7391 The following built-in functions are available when @option{-msse4a} is used.
7394 void _mm_stream_sd (double*,__m128d);
7395 Generates the @code{movntsd} machine instruction.
7396 void _mm_stream_ss (float*,__m128);
7397 Generates the @code{movntss} machine instruction.
7398 __m128i _mm_extract_si64 (__m128i, __m128i);
7399 Generates the @code{extrq} machine instruction with only SSE register operands.
7400 __m128i _mm_extracti_si64 (__m128i, int, int);
7401 Generates the @code{extrq} machine instruction with SSE register and immediate operands.
7402 __m128i _mm_insert_si64 (__m128i, __m128i);
7403 Generates the @code{insertq} machine instruction with only SSE register operands.
7404 __m128i _mm_inserti_si64 (__m128i, __m128i, int, int);
7405 Generates the @code{insertq} machine instruction with SSE register and immediate operands.
7408 The following built-in functions are available when @option{-m3dnow} is used.
7409 All of them generate the machine instruction that is part of the name.
7412 void __builtin_ia32_femms (void)
7413 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7414 v2si __builtin_ia32_pf2id (v2sf)
7415 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7416 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7417 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7418 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7419 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7420 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7421 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7422 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7423 v2sf __builtin_ia32_pfrcp (v2sf)
7424 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7425 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7426 v2sf __builtin_ia32_pfrsqrt (v2sf)
7427 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7428 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7429 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7430 v2sf __builtin_ia32_pi2fd (v2si)
7431 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7434 The following built-in functions are available when both @option{-m3dnow}
7435 and @option{-march=athlon} are used. All of them generate the machine
7436 instruction that is part of the name.
7439 v2si __builtin_ia32_pf2iw (v2sf)
7440 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7441 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7442 v2sf __builtin_ia32_pi2fw (v2si)
7443 v2sf __builtin_ia32_pswapdsf (v2sf)
7444 v2si __builtin_ia32_pswapdsi (v2si)
7447 @node MIPS DSP Built-in Functions
7448 @subsection MIPS DSP Built-in Functions
7450 The MIPS DSP Application-Specific Extension (ASE) includes new
7451 instructions that are designed to improve the performance of DSP and
7452 media applications. It provides instructions that operate on packed
7453 8-bit integer data, Q15 fractional data and Q31 fractional data.
7455 GCC supports MIPS DSP operations using both the generic
7456 vector extensions (@pxref{Vector Extensions}) and a collection of
7457 MIPS-specific built-in functions. Both kinds of support are
7458 enabled by the @option{-mdsp} command-line option.
7460 At present, GCC only provides support for operations on 32-bit
7461 vectors. The vector type associated with 8-bit integer data is
7462 usually called @code{v4i8} and the vector type associated with Q15 is
7463 usually called @code{v2q15}. They can be defined in C as follows:
7466 typedef char v4i8 __attribute__ ((vector_size(4)));
7467 typedef short v2q15 __attribute__ ((vector_size(4)));
7470 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7471 aggregates. For example:
7474 v4i8 a = @{1, 2, 3, 4@};
7476 b = (v4i8) @{5, 6, 7, 8@};
7478 v2q15 c = @{0x0fcb, 0x3a75@};
7480 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7483 @emph{Note:} The CPU's endianness determines the order in which values
7484 are packed. On little-endian targets, the first value is the least
7485 significant and the last value is the most significant. The opposite
7486 order applies to big-endian targets. For example, the code above will
7487 set the lowest byte of @code{a} to @code{1} on little-endian targets
7488 and @code{4} on big-endian targets.
7490 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7491 representation. As shown in this example, the integer representation
7492 of a Q15 value can be obtained by multiplying the fractional value by
7493 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7496 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7497 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7498 and @code{c} and @code{d} are @code{v2q15} values.
7500 @multitable @columnfractions .50 .50
7501 @item C code @tab MIPS instruction
7502 @item @code{a + b} @tab @code{addu.qb}
7503 @item @code{c + d} @tab @code{addq.ph}
7504 @item @code{a - b} @tab @code{subu.qb}
7505 @item @code{c - d} @tab @code{subq.ph}
7508 It is easier to describe the DSP built-in functions if we first define
7509 the following types:
7514 typedef long long a64;
7517 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7518 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7519 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7520 @code{long long}, but we use @code{a64} to indicate values that will
7521 be placed in one of the four DSP accumulators (@code{$ac0},
7522 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7524 Also, some built-in functions prefer or require immediate numbers as
7525 parameters, because the corresponding DSP instructions accept both immediate
7526 numbers and register operands, or accept immediate numbers only. The
7527 immediate parameters are listed as follows.
7535 imm_n32_31: -32 to 31.
7536 imm_n512_511: -512 to 511.
7539 The following built-in functions map directly to a particular MIPS DSP
7540 instruction. Please refer to the architecture specification
7541 for details on what each instruction does.
7544 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7545 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7546 q31 __builtin_mips_addq_s_w (q31, q31)
7547 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7548 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7549 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7550 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7551 q31 __builtin_mips_subq_s_w (q31, q31)
7552 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7553 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7554 i32 __builtin_mips_addsc (i32, i32)
7555 i32 __builtin_mips_addwc (i32, i32)
7556 i32 __builtin_mips_modsub (i32, i32)
7557 i32 __builtin_mips_raddu_w_qb (v4i8)
7558 v2q15 __builtin_mips_absq_s_ph (v2q15)
7559 q31 __builtin_mips_absq_s_w (q31)
7560 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7561 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7562 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7563 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7564 q31 __builtin_mips_preceq_w_phl (v2q15)
7565 q31 __builtin_mips_preceq_w_phr (v2q15)
7566 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7567 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7568 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7569 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7570 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7571 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7572 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7573 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7574 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7575 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7576 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7577 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7578 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7579 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7580 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7581 q31 __builtin_mips_shll_s_w (q31, i32)
7582 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7583 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7584 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7585 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7586 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7587 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7588 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7589 q31 __builtin_mips_shra_r_w (q31, i32)
7590 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7591 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7592 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7593 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7594 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7595 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7596 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7597 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7598 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7599 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7600 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7601 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7602 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7603 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7604 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7605 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7606 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7607 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7608 i32 __builtin_mips_bitrev (i32)
7609 i32 __builtin_mips_insv (i32, i32)
7610 v4i8 __builtin_mips_repl_qb (imm0_255)
7611 v4i8 __builtin_mips_repl_qb (i32)
7612 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7613 v2q15 __builtin_mips_repl_ph (i32)
7614 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7615 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7616 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7617 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7618 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7619 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7620 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7621 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7622 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7623 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7624 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7625 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7626 i32 __builtin_mips_extr_w (a64, imm0_31)
7627 i32 __builtin_mips_extr_w (a64, i32)
7628 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7629 i32 __builtin_mips_extr_s_h (a64, i32)
7630 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7631 i32 __builtin_mips_extr_rs_w (a64, i32)
7632 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7633 i32 __builtin_mips_extr_r_w (a64, i32)
7634 i32 __builtin_mips_extp (a64, imm0_31)
7635 i32 __builtin_mips_extp (a64, i32)
7636 i32 __builtin_mips_extpdp (a64, imm0_31)
7637 i32 __builtin_mips_extpdp (a64, i32)
7638 a64 __builtin_mips_shilo (a64, imm_n32_31)
7639 a64 __builtin_mips_shilo (a64, i32)
7640 a64 __builtin_mips_mthlip (a64, i32)
7641 void __builtin_mips_wrdsp (i32, imm0_63)
7642 i32 __builtin_mips_rddsp (imm0_63)
7643 i32 __builtin_mips_lbux (void *, i32)
7644 i32 __builtin_mips_lhx (void *, i32)
7645 i32 __builtin_mips_lwx (void *, i32)
7646 i32 __builtin_mips_bposge32 (void)
7649 @node MIPS Paired-Single Support
7650 @subsection MIPS Paired-Single Support
7652 The MIPS64 architecture includes a number of instructions that
7653 operate on pairs of single-precision floating-point values.
7654 Each pair is packed into a 64-bit floating-point register,
7655 with one element being designated the ``upper half'' and
7656 the other being designated the ``lower half''.
7658 GCC supports paired-single operations using both the generic
7659 vector extensions (@pxref{Vector Extensions}) and a collection of
7660 MIPS-specific built-in functions. Both kinds of support are
7661 enabled by the @option{-mpaired-single} command-line option.
7663 The vector type associated with paired-single values is usually
7664 called @code{v2sf}. It can be defined in C as follows:
7667 typedef float v2sf __attribute__ ((vector_size (8)));
7670 @code{v2sf} values are initialized in the same way as aggregates.
7674 v2sf a = @{1.5, 9.1@};
7677 b = (v2sf) @{e, f@};
7680 @emph{Note:} The CPU's endianness determines which value is stored in
7681 the upper half of a register and which value is stored in the lower half.
7682 On little-endian targets, the first value is the lower one and the second
7683 value is the upper one. The opposite order applies to big-endian targets.
7684 For example, the code above will set the lower half of @code{a} to
7685 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7688 * Paired-Single Arithmetic::
7689 * Paired-Single Built-in Functions::
7690 * MIPS-3D Built-in Functions::
7693 @node Paired-Single Arithmetic
7694 @subsubsection Paired-Single Arithmetic
7696 The table below lists the @code{v2sf} operations for which hardware
7697 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7698 values and @code{x} is an integral value.
7700 @multitable @columnfractions .50 .50
7701 @item C code @tab MIPS instruction
7702 @item @code{a + b} @tab @code{add.ps}
7703 @item @code{a - b} @tab @code{sub.ps}
7704 @item @code{-a} @tab @code{neg.ps}
7705 @item @code{a * b} @tab @code{mul.ps}
7706 @item @code{a * b + c} @tab @code{madd.ps}
7707 @item @code{a * b - c} @tab @code{msub.ps}
7708 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7709 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7710 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7713 Note that the multiply-accumulate instructions can be disabled
7714 using the command-line option @code{-mno-fused-madd}.
7716 @node Paired-Single Built-in Functions
7717 @subsubsection Paired-Single Built-in Functions
7719 The following paired-single functions map directly to a particular
7720 MIPS instruction. Please refer to the architecture specification
7721 for details on what each instruction does.
7724 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7725 Pair lower lower (@code{pll.ps}).
7727 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7728 Pair upper lower (@code{pul.ps}).
7730 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7731 Pair lower upper (@code{plu.ps}).
7733 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7734 Pair upper upper (@code{puu.ps}).
7736 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7737 Convert pair to paired single (@code{cvt.ps.s}).
7739 @item float __builtin_mips_cvt_s_pl (v2sf)
7740 Convert pair lower to single (@code{cvt.s.pl}).
7742 @item float __builtin_mips_cvt_s_pu (v2sf)
7743 Convert pair upper to single (@code{cvt.s.pu}).
7745 @item v2sf __builtin_mips_abs_ps (v2sf)
7746 Absolute value (@code{abs.ps}).
7748 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7749 Align variable (@code{alnv.ps}).
7751 @emph{Note:} The value of the third parameter must be 0 or 4
7752 modulo 8, otherwise the result will be unpredictable. Please read the
7753 instruction description for details.
7756 The following multi-instruction functions are also available.
7757 In each case, @var{cond} can be any of the 16 floating-point conditions:
7758 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7759 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7760 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7763 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7764 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7765 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7766 @code{movt.ps}/@code{movf.ps}).
7768 The @code{movt} functions return the value @var{x} computed by:
7771 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7772 mov.ps @var{x},@var{c}
7773 movt.ps @var{x},@var{d},@var{cc}
7776 The @code{movf} functions are similar but use @code{movf.ps} instead
7779 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7780 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7781 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7782 @code{bc1t}/@code{bc1f}).
7784 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7785 and return either the upper or lower half of the result. For example:
7789 if (__builtin_mips_upper_c_eq_ps (a, b))
7790 upper_halves_are_equal ();
7792 upper_halves_are_unequal ();
7794 if (__builtin_mips_lower_c_eq_ps (a, b))
7795 lower_halves_are_equal ();
7797 lower_halves_are_unequal ();
7801 @node MIPS-3D Built-in Functions
7802 @subsubsection MIPS-3D Built-in Functions
7804 The MIPS-3D Application-Specific Extension (ASE) includes additional
7805 paired-single instructions that are designed to improve the performance
7806 of 3D graphics operations. Support for these instructions is controlled
7807 by the @option{-mips3d} command-line option.
7809 The functions listed below map directly to a particular MIPS-3D
7810 instruction. Please refer to the architecture specification for
7811 more details on what each instruction does.
7814 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7815 Reduction add (@code{addr.ps}).
7817 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7818 Reduction multiply (@code{mulr.ps}).
7820 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7821 Convert paired single to paired word (@code{cvt.pw.ps}).
7823 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7824 Convert paired word to paired single (@code{cvt.ps.pw}).
7826 @item float __builtin_mips_recip1_s (float)
7827 @itemx double __builtin_mips_recip1_d (double)
7828 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7829 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7831 @item float __builtin_mips_recip2_s (float, float)
7832 @itemx double __builtin_mips_recip2_d (double, double)
7833 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7834 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7836 @item float __builtin_mips_rsqrt1_s (float)
7837 @itemx double __builtin_mips_rsqrt1_d (double)
7838 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7839 Reduced precision reciprocal square root (sequence step 1)
7840 (@code{rsqrt1.@var{fmt}}).
7842 @item float __builtin_mips_rsqrt2_s (float, float)
7843 @itemx double __builtin_mips_rsqrt2_d (double, double)
7844 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7845 Reduced precision reciprocal square root (sequence step 2)
7846 (@code{rsqrt2.@var{fmt}}).
7849 The following multi-instruction functions are also available.
7850 In each case, @var{cond} can be any of the 16 floating-point conditions:
7851 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7852 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7853 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7856 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7857 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7858 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7859 @code{bc1t}/@code{bc1f}).
7861 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7862 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7867 if (__builtin_mips_cabs_eq_s (a, b))
7873 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7874 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7875 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7876 @code{bc1t}/@code{bc1f}).
7878 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7879 and return either the upper or lower half of the result. For example:
7883 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7884 upper_halves_are_equal ();
7886 upper_halves_are_unequal ();
7888 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7889 lower_halves_are_equal ();
7891 lower_halves_are_unequal ();
7894 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7895 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7896 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7897 @code{movt.ps}/@code{movf.ps}).
7899 The @code{movt} functions return the value @var{x} computed by:
7902 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7903 mov.ps @var{x},@var{c}
7904 movt.ps @var{x},@var{d},@var{cc}
7907 The @code{movf} functions are similar but use @code{movf.ps} instead
7910 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7911 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7912 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7913 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7914 Comparison of two paired-single values
7915 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7916 @code{bc1any2t}/@code{bc1any2f}).
7918 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7919 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7920 result is true and the @code{all} forms return true if both results are true.
7925 if (__builtin_mips_any_c_eq_ps (a, b))
7930 if (__builtin_mips_all_c_eq_ps (a, b))
7936 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7937 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7938 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7939 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7940 Comparison of four paired-single values
7941 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7942 @code{bc1any4t}/@code{bc1any4f}).
7944 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7945 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7946 The @code{any} forms return true if any of the four results are true
7947 and the @code{all} forms return true if all four results are true.
7952 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7957 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7964 @node PowerPC AltiVec Built-in Functions
7965 @subsection PowerPC AltiVec Built-in Functions
7967 GCC provides an interface for the PowerPC family of processors to access
7968 the AltiVec operations described in Motorola's AltiVec Programming
7969 Interface Manual. The interface is made available by including
7970 @code{<altivec.h>} and using @option{-maltivec} and
7971 @option{-mabi=altivec}. The interface supports the following vector
7975 vector unsigned char
7979 vector unsigned short
7990 GCC's implementation of the high-level language interface available from
7991 C and C++ code differs from Motorola's documentation in several ways.
7996 A vector constant is a list of constant expressions within curly braces.
7999 A vector initializer requires no cast if the vector constant is of the
8000 same type as the variable it is initializing.
8003 If @code{signed} or @code{unsigned} is omitted, the signedness of the
8004 vector type is the default signedness of the base type. The default
8005 varies depending on the operating system, so a portable program should
8006 always specify the signedness.
8009 Compiling with @option{-maltivec} adds keywords @code{__vector},
8010 @code{__pixel}, and @code{__bool}. Macros @option{vector},
8011 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
8015 GCC allows using a @code{typedef} name as the type specifier for a
8019 For C, overloaded functions are implemented with macros so the following
8023 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
8026 Since @code{vec_add} is a macro, the vector constant in the example
8027 is treated as four separate arguments. Wrap the entire argument in
8028 parentheses for this to work.
8031 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
8032 Internally, GCC uses built-in functions to achieve the functionality in
8033 the aforementioned header file, but they are not supported and are
8034 subject to change without notice.
8036 The following interfaces are supported for the generic and specific
8037 AltiVec operations and the AltiVec predicates. In cases where there
8038 is a direct mapping between generic and specific operations, only the
8039 generic names are shown here, although the specific operations can also
8042 Arguments that are documented as @code{const int} require literal
8043 integral values within the range required for that operation.
8046 vector signed char vec_abs (vector signed char);
8047 vector signed short vec_abs (vector signed short);
8048 vector signed int vec_abs (vector signed int);
8049 vector float vec_abs (vector float);
8051 vector signed char vec_abss (vector signed char);
8052 vector signed short vec_abss (vector signed short);
8053 vector signed int vec_abss (vector signed int);
8055 vector signed char vec_add (vector bool char, vector signed char);
8056 vector signed char vec_add (vector signed char, vector bool char);
8057 vector signed char vec_add (vector signed char, vector signed char);
8058 vector unsigned char vec_add (vector bool char, vector unsigned char);
8059 vector unsigned char vec_add (vector unsigned char, vector bool char);
8060 vector unsigned char vec_add (vector unsigned char,
8061 vector unsigned char);
8062 vector signed short vec_add (vector bool short, vector signed short);
8063 vector signed short vec_add (vector signed short, vector bool short);
8064 vector signed short vec_add (vector signed short, vector signed short);
8065 vector unsigned short vec_add (vector bool short,
8066 vector unsigned short);
8067 vector unsigned short vec_add (vector unsigned short,
8069 vector unsigned short vec_add (vector unsigned short,
8070 vector unsigned short);
8071 vector signed int vec_add (vector bool int, vector signed int);
8072 vector signed int vec_add (vector signed int, vector bool int);
8073 vector signed int vec_add (vector signed int, vector signed int);
8074 vector unsigned int vec_add (vector bool int, vector unsigned int);
8075 vector unsigned int vec_add (vector unsigned int, vector bool int);
8076 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
8077 vector float vec_add (vector float, vector float);
8079 vector float vec_vaddfp (vector float, vector float);
8081 vector signed int vec_vadduwm (vector bool int, vector signed int);
8082 vector signed int vec_vadduwm (vector signed int, vector bool int);
8083 vector signed int vec_vadduwm (vector signed int, vector signed int);
8084 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
8085 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
8086 vector unsigned int vec_vadduwm (vector unsigned int,
8087 vector unsigned int);
8089 vector signed short vec_vadduhm (vector bool short,
8090 vector signed short);
8091 vector signed short vec_vadduhm (vector signed short,
8093 vector signed short vec_vadduhm (vector signed short,
8094 vector signed short);
8095 vector unsigned short vec_vadduhm (vector bool short,
8096 vector unsigned short);
8097 vector unsigned short vec_vadduhm (vector unsigned short,
8099 vector unsigned short vec_vadduhm (vector unsigned short,
8100 vector unsigned short);
8102 vector signed char vec_vaddubm (vector bool char, vector signed char);
8103 vector signed char vec_vaddubm (vector signed char, vector bool char);
8104 vector signed char vec_vaddubm (vector signed char, vector signed char);
8105 vector unsigned char vec_vaddubm (vector bool char,
8106 vector unsigned char);
8107 vector unsigned char vec_vaddubm (vector unsigned char,
8109 vector unsigned char vec_vaddubm (vector unsigned char,
8110 vector unsigned char);
8112 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
8114 vector unsigned char vec_adds (vector bool char, vector unsigned char);
8115 vector unsigned char vec_adds (vector unsigned char, vector bool char);
8116 vector unsigned char vec_adds (vector unsigned char,
8117 vector unsigned char);
8118 vector signed char vec_adds (vector bool char, vector signed char);
8119 vector signed char vec_adds (vector signed char, vector bool char);
8120 vector signed char vec_adds (vector signed char, vector signed char);
8121 vector unsigned short vec_adds (vector bool short,
8122 vector unsigned short);
8123 vector unsigned short vec_adds (vector unsigned short,
8125 vector unsigned short vec_adds (vector unsigned short,
8126 vector unsigned short);
8127 vector signed short vec_adds (vector bool short, vector signed short);
8128 vector signed short vec_adds (vector signed short, vector bool short);
8129 vector signed short vec_adds (vector signed short, vector signed short);
8130 vector unsigned int vec_adds (vector bool int, vector unsigned int);
8131 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8132 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8133 vector signed int vec_adds (vector bool int, vector signed int);
8134 vector signed int vec_adds (vector signed int, vector bool int);
8135 vector signed int vec_adds (vector signed int, vector signed int);
8137 vector signed int vec_vaddsws (vector bool int, vector signed int);
8138 vector signed int vec_vaddsws (vector signed int, vector bool int);
8139 vector signed int vec_vaddsws (vector signed int, vector signed int);
8141 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8142 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8143 vector unsigned int vec_vadduws (vector unsigned int,
8144 vector unsigned int);
8146 vector signed short vec_vaddshs (vector bool short,
8147 vector signed short);
8148 vector signed short vec_vaddshs (vector signed short,
8150 vector signed short vec_vaddshs (vector signed short,
8151 vector signed short);
8153 vector unsigned short vec_vadduhs (vector bool short,
8154 vector unsigned short);
8155 vector unsigned short vec_vadduhs (vector unsigned short,
8157 vector unsigned short vec_vadduhs (vector unsigned short,
8158 vector unsigned short);
8160 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8161 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8162 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8164 vector unsigned char vec_vaddubs (vector bool char,
8165 vector unsigned char);
8166 vector unsigned char vec_vaddubs (vector unsigned char,
8168 vector unsigned char vec_vaddubs (vector unsigned char,
8169 vector unsigned char);
8171 vector float vec_and (vector float, vector float);
8172 vector float vec_and (vector float, vector bool int);
8173 vector float vec_and (vector bool int, vector float);
8174 vector bool int vec_and (vector bool int, vector bool int);
8175 vector signed int vec_and (vector bool int, vector signed int);
8176 vector signed int vec_and (vector signed int, vector bool int);
8177 vector signed int vec_and (vector signed int, vector signed int);
8178 vector unsigned int vec_and (vector bool int, vector unsigned int);
8179 vector unsigned int vec_and (vector unsigned int, vector bool int);
8180 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8181 vector bool short vec_and (vector bool short, vector bool short);
8182 vector signed short vec_and (vector bool short, vector signed short);
8183 vector signed short vec_and (vector signed short, vector bool short);
8184 vector signed short vec_and (vector signed short, vector signed short);
8185 vector unsigned short vec_and (vector bool short,
8186 vector unsigned short);
8187 vector unsigned short vec_and (vector unsigned short,
8189 vector unsigned short vec_and (vector unsigned short,
8190 vector unsigned short);
8191 vector signed char vec_and (vector bool char, vector signed char);
8192 vector bool char vec_and (vector bool char, vector bool char);
8193 vector signed char vec_and (vector signed char, vector bool char);
8194 vector signed char vec_and (vector signed char, vector signed char);
8195 vector unsigned char vec_and (vector bool char, vector unsigned char);
8196 vector unsigned char vec_and (vector unsigned char, vector bool char);
8197 vector unsigned char vec_and (vector unsigned char,
8198 vector unsigned char);
8200 vector float vec_andc (vector float, vector float);
8201 vector float vec_andc (vector float, vector bool int);
8202 vector float vec_andc (vector bool int, vector float);
8203 vector bool int vec_andc (vector bool int, vector bool int);
8204 vector signed int vec_andc (vector bool int, vector signed int);
8205 vector signed int vec_andc (vector signed int, vector bool int);
8206 vector signed int vec_andc (vector signed int, vector signed int);
8207 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8208 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8209 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8210 vector bool short vec_andc (vector bool short, vector bool short);
8211 vector signed short vec_andc (vector bool short, vector signed short);
8212 vector signed short vec_andc (vector signed short, vector bool short);
8213 vector signed short vec_andc (vector signed short, vector signed short);
8214 vector unsigned short vec_andc (vector bool short,
8215 vector unsigned short);
8216 vector unsigned short vec_andc (vector unsigned short,
8218 vector unsigned short vec_andc (vector unsigned short,
8219 vector unsigned short);
8220 vector signed char vec_andc (vector bool char, vector signed char);
8221 vector bool char vec_andc (vector bool char, vector bool char);
8222 vector signed char vec_andc (vector signed char, vector bool char);
8223 vector signed char vec_andc (vector signed char, vector signed char);
8224 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8225 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8226 vector unsigned char vec_andc (vector unsigned char,
8227 vector unsigned char);
8229 vector unsigned char vec_avg (vector unsigned char,
8230 vector unsigned char);
8231 vector signed char vec_avg (vector signed char, vector signed char);
8232 vector unsigned short vec_avg (vector unsigned short,
8233 vector unsigned short);
8234 vector signed short vec_avg (vector signed short, vector signed short);
8235 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8236 vector signed int vec_avg (vector signed int, vector signed int);
8238 vector signed int vec_vavgsw (vector signed int, vector signed int);
8240 vector unsigned int vec_vavguw (vector unsigned int,
8241 vector unsigned int);
8243 vector signed short vec_vavgsh (vector signed short,
8244 vector signed short);
8246 vector unsigned short vec_vavguh (vector unsigned short,
8247 vector unsigned short);
8249 vector signed char vec_vavgsb (vector signed char, vector signed char);
8251 vector unsigned char vec_vavgub (vector unsigned char,
8252 vector unsigned char);
8254 vector float vec_ceil (vector float);
8256 vector signed int vec_cmpb (vector float, vector float);
8258 vector bool char vec_cmpeq (vector signed char, vector signed char);
8259 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8260 vector bool short vec_cmpeq (vector signed short, vector signed short);
8261 vector bool short vec_cmpeq (vector unsigned short,
8262 vector unsigned short);
8263 vector bool int vec_cmpeq (vector signed int, vector signed int);
8264 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8265 vector bool int vec_cmpeq (vector float, vector float);
8267 vector bool int vec_vcmpeqfp (vector float, vector float);
8269 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8270 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8272 vector bool short vec_vcmpequh (vector signed short,
8273 vector signed short);
8274 vector bool short vec_vcmpequh (vector unsigned short,
8275 vector unsigned short);
8277 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8278 vector bool char vec_vcmpequb (vector unsigned char,
8279 vector unsigned char);
8281 vector bool int vec_cmpge (vector float, vector float);
8283 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8284 vector bool char vec_cmpgt (vector signed char, vector signed char);
8285 vector bool short vec_cmpgt (vector unsigned short,
8286 vector unsigned short);
8287 vector bool short vec_cmpgt (vector signed short, vector signed short);
8288 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8289 vector bool int vec_cmpgt (vector signed int, vector signed int);
8290 vector bool int vec_cmpgt (vector float, vector float);
8292 vector bool int vec_vcmpgtfp (vector float, vector float);
8294 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8296 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8298 vector bool short vec_vcmpgtsh (vector signed short,
8299 vector signed short);
8301 vector bool short vec_vcmpgtuh (vector unsigned short,
8302 vector unsigned short);
8304 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8306 vector bool char vec_vcmpgtub (vector unsigned char,
8307 vector unsigned char);
8309 vector bool int vec_cmple (vector float, vector float);
8311 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8312 vector bool char vec_cmplt (vector signed char, vector signed char);
8313 vector bool short vec_cmplt (vector unsigned short,
8314 vector unsigned short);
8315 vector bool short vec_cmplt (vector signed short, vector signed short);
8316 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8317 vector bool int vec_cmplt (vector signed int, vector signed int);
8318 vector bool int vec_cmplt (vector float, vector float);
8320 vector float vec_ctf (vector unsigned int, const int);
8321 vector float vec_ctf (vector signed int, const int);
8323 vector float vec_vcfsx (vector signed int, const int);
8325 vector float vec_vcfux (vector unsigned int, const int);
8327 vector signed int vec_cts (vector float, const int);
8329 vector unsigned int vec_ctu (vector float, const int);
8331 void vec_dss (const int);
8333 void vec_dssall (void);
8335 void vec_dst (const vector unsigned char *, int, const int);
8336 void vec_dst (const vector signed char *, int, const int);
8337 void vec_dst (const vector bool char *, int, const int);
8338 void vec_dst (const vector unsigned short *, int, const int);
8339 void vec_dst (const vector signed short *, int, const int);
8340 void vec_dst (const vector bool short *, int, const int);
8341 void vec_dst (const vector pixel *, int, const int);
8342 void vec_dst (const vector unsigned int *, int, const int);
8343 void vec_dst (const vector signed int *, int, const int);
8344 void vec_dst (const vector bool int *, int, const int);
8345 void vec_dst (const vector float *, int, const int);
8346 void vec_dst (const unsigned char *, int, const int);
8347 void vec_dst (const signed char *, int, const int);
8348 void vec_dst (const unsigned short *, int, const int);
8349 void vec_dst (const short *, int, const int);
8350 void vec_dst (const unsigned int *, int, const int);
8351 void vec_dst (const int *, int, const int);
8352 void vec_dst (const unsigned long *, int, const int);
8353 void vec_dst (const long *, int, const int);
8354 void vec_dst (const float *, int, const int);
8356 void vec_dstst (const vector unsigned char *, int, const int);
8357 void vec_dstst (const vector signed char *, int, const int);
8358 void vec_dstst (const vector bool char *, int, const int);
8359 void vec_dstst (const vector unsigned short *, int, const int);
8360 void vec_dstst (const vector signed short *, int, const int);
8361 void vec_dstst (const vector bool short *, int, const int);
8362 void vec_dstst (const vector pixel *, int, const int);
8363 void vec_dstst (const vector unsigned int *, int, const int);
8364 void vec_dstst (const vector signed int *, int, const int);
8365 void vec_dstst (const vector bool int *, int, const int);
8366 void vec_dstst (const vector float *, int, const int);
8367 void vec_dstst (const unsigned char *, int, const int);
8368 void vec_dstst (const signed char *, int, const int);
8369 void vec_dstst (const unsigned short *, int, const int);
8370 void vec_dstst (const short *, int, const int);
8371 void vec_dstst (const unsigned int *, int, const int);
8372 void vec_dstst (const int *, int, const int);
8373 void vec_dstst (const unsigned long *, int, const int);
8374 void vec_dstst (const long *, int, const int);
8375 void vec_dstst (const float *, int, const int);
8377 void vec_dststt (const vector unsigned char *, int, const int);
8378 void vec_dststt (const vector signed char *, int, const int);
8379 void vec_dststt (const vector bool char *, int, const int);
8380 void vec_dststt (const vector unsigned short *, int, const int);
8381 void vec_dststt (const vector signed short *, int, const int);
8382 void vec_dststt (const vector bool short *, int, const int);
8383 void vec_dststt (const vector pixel *, int, const int);
8384 void vec_dststt (const vector unsigned int *, int, const int);
8385 void vec_dststt (const vector signed int *, int, const int);
8386 void vec_dststt (const vector bool int *, int, const int);
8387 void vec_dststt (const vector float *, int, const int);
8388 void vec_dststt (const unsigned char *, int, const int);
8389 void vec_dststt (const signed char *, int, const int);
8390 void vec_dststt (const unsigned short *, int, const int);
8391 void vec_dststt (const short *, int, const int);
8392 void vec_dststt (const unsigned int *, int, const int);
8393 void vec_dststt (const int *, int, const int);
8394 void vec_dststt (const unsigned long *, int, const int);
8395 void vec_dststt (const long *, int, const int);
8396 void vec_dststt (const float *, int, const int);
8398 void vec_dstt (const vector unsigned char *, int, const int);
8399 void vec_dstt (const vector signed char *, int, const int);
8400 void vec_dstt (const vector bool char *, int, const int);
8401 void vec_dstt (const vector unsigned short *, int, const int);
8402 void vec_dstt (const vector signed short *, int, const int);
8403 void vec_dstt (const vector bool short *, int, const int);
8404 void vec_dstt (const vector pixel *, int, const int);
8405 void vec_dstt (const vector unsigned int *, int, const int);
8406 void vec_dstt (const vector signed int *, int, const int);
8407 void vec_dstt (const vector bool int *, int, const int);
8408 void vec_dstt (const vector float *, int, const int);
8409 void vec_dstt (const unsigned char *, int, const int);
8410 void vec_dstt (const signed char *, int, const int);
8411 void vec_dstt (const unsigned short *, int, const int);
8412 void vec_dstt (const short *, int, const int);
8413 void vec_dstt (const unsigned int *, int, const int);
8414 void vec_dstt (const int *, int, const int);
8415 void vec_dstt (const unsigned long *, int, const int);
8416 void vec_dstt (const long *, int, const int);
8417 void vec_dstt (const float *, int, const int);
8419 vector float vec_expte (vector float);
8421 vector float vec_floor (vector float);
8423 vector float vec_ld (int, const vector float *);
8424 vector float vec_ld (int, const float *);
8425 vector bool int vec_ld (int, const vector bool int *);
8426 vector signed int vec_ld (int, const vector signed int *);
8427 vector signed int vec_ld (int, const int *);
8428 vector signed int vec_ld (int, const long *);
8429 vector unsigned int vec_ld (int, const vector unsigned int *);
8430 vector unsigned int vec_ld (int, const unsigned int *);
8431 vector unsigned int vec_ld (int, const unsigned long *);
8432 vector bool short vec_ld (int, const vector bool short *);
8433 vector pixel vec_ld (int, const vector pixel *);
8434 vector signed short vec_ld (int, const vector signed short *);
8435 vector signed short vec_ld (int, const short *);
8436 vector unsigned short vec_ld (int, const vector unsigned short *);
8437 vector unsigned short vec_ld (int, const unsigned short *);
8438 vector bool char vec_ld (int, const vector bool char *);
8439 vector signed char vec_ld (int, const vector signed char *);
8440 vector signed char vec_ld (int, const signed char *);
8441 vector unsigned char vec_ld (int, const vector unsigned char *);
8442 vector unsigned char vec_ld (int, const unsigned char *);
8444 vector signed char vec_lde (int, const signed char *);
8445 vector unsigned char vec_lde (int, const unsigned char *);
8446 vector signed short vec_lde (int, const short *);
8447 vector unsigned short vec_lde (int, const unsigned short *);
8448 vector float vec_lde (int, const float *);
8449 vector signed int vec_lde (int, const int *);
8450 vector unsigned int vec_lde (int, const unsigned int *);
8451 vector signed int vec_lde (int, const long *);
8452 vector unsigned int vec_lde (int, const unsigned long *);
8454 vector float vec_lvewx (int, float *);
8455 vector signed int vec_lvewx (int, int *);
8456 vector unsigned int vec_lvewx (int, unsigned int *);
8457 vector signed int vec_lvewx (int, long *);
8458 vector unsigned int vec_lvewx (int, unsigned long *);
8460 vector signed short vec_lvehx (int, short *);
8461 vector unsigned short vec_lvehx (int, unsigned short *);
8463 vector signed char vec_lvebx (int, char *);
8464 vector unsigned char vec_lvebx (int, unsigned char *);
8466 vector float vec_ldl (int, const vector float *);
8467 vector float vec_ldl (int, const float *);
8468 vector bool int vec_ldl (int, const vector bool int *);
8469 vector signed int vec_ldl (int, const vector signed int *);
8470 vector signed int vec_ldl (int, const int *);
8471 vector signed int vec_ldl (int, const long *);
8472 vector unsigned int vec_ldl (int, const vector unsigned int *);
8473 vector unsigned int vec_ldl (int, const unsigned int *);
8474 vector unsigned int vec_ldl (int, const unsigned long *);
8475 vector bool short vec_ldl (int, const vector bool short *);
8476 vector pixel vec_ldl (int, const vector pixel *);
8477 vector signed short vec_ldl (int, const vector signed short *);
8478 vector signed short vec_ldl (int, const short *);
8479 vector unsigned short vec_ldl (int, const vector unsigned short *);
8480 vector unsigned short vec_ldl (int, const unsigned short *);
8481 vector bool char vec_ldl (int, const vector bool char *);
8482 vector signed char vec_ldl (int, const vector signed char *);
8483 vector signed char vec_ldl (int, const signed char *);
8484 vector unsigned char vec_ldl (int, const vector unsigned char *);
8485 vector unsigned char vec_ldl (int, const unsigned char *);
8487 vector float vec_loge (vector float);
8489 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8490 vector unsigned char vec_lvsl (int, const volatile signed char *);
8491 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8492 vector unsigned char vec_lvsl (int, const volatile short *);
8493 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8494 vector unsigned char vec_lvsl (int, const volatile int *);
8495 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8496 vector unsigned char vec_lvsl (int, const volatile long *);
8497 vector unsigned char vec_lvsl (int, const volatile float *);
8499 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8500 vector unsigned char vec_lvsr (int, const volatile signed char *);
8501 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8502 vector unsigned char vec_lvsr (int, const volatile short *);
8503 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8504 vector unsigned char vec_lvsr (int, const volatile int *);
8505 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8506 vector unsigned char vec_lvsr (int, const volatile long *);
8507 vector unsigned char vec_lvsr (int, const volatile float *);
8509 vector float vec_madd (vector float, vector float, vector float);
8511 vector signed short vec_madds (vector signed short,
8512 vector signed short,
8513 vector signed short);
8515 vector unsigned char vec_max (vector bool char, vector unsigned char);
8516 vector unsigned char vec_max (vector unsigned char, vector bool char);
8517 vector unsigned char vec_max (vector unsigned char,
8518 vector unsigned char);
8519 vector signed char vec_max (vector bool char, vector signed char);
8520 vector signed char vec_max (vector signed char, vector bool char);
8521 vector signed char vec_max (vector signed char, vector signed char);
8522 vector unsigned short vec_max (vector bool short,
8523 vector unsigned short);
8524 vector unsigned short vec_max (vector unsigned short,
8526 vector unsigned short vec_max (vector unsigned short,
8527 vector unsigned short);
8528 vector signed short vec_max (vector bool short, vector signed short);
8529 vector signed short vec_max (vector signed short, vector bool short);
8530 vector signed short vec_max (vector signed short, vector signed short);
8531 vector unsigned int vec_max (vector bool int, vector unsigned int);
8532 vector unsigned int vec_max (vector unsigned int, vector bool int);
8533 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8534 vector signed int vec_max (vector bool int, vector signed int);
8535 vector signed int vec_max (vector signed int, vector bool int);
8536 vector signed int vec_max (vector signed int, vector signed int);
8537 vector float vec_max (vector float, vector float);
8539 vector float vec_vmaxfp (vector float, vector float);
8541 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8542 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8543 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8545 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8546 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8547 vector unsigned int vec_vmaxuw (vector unsigned int,
8548 vector unsigned int);
8550 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8551 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8552 vector signed short vec_vmaxsh (vector signed short,
8553 vector signed short);
8555 vector unsigned short vec_vmaxuh (vector bool short,
8556 vector unsigned short);
8557 vector unsigned short vec_vmaxuh (vector unsigned short,
8559 vector unsigned short vec_vmaxuh (vector unsigned short,
8560 vector unsigned short);
8562 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8563 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8564 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8566 vector unsigned char vec_vmaxub (vector bool char,
8567 vector unsigned char);
8568 vector unsigned char vec_vmaxub (vector unsigned char,
8570 vector unsigned char vec_vmaxub (vector unsigned char,
8571 vector unsigned char);
8573 vector bool char vec_mergeh (vector bool char, vector bool char);
8574 vector signed char vec_mergeh (vector signed char, vector signed char);
8575 vector unsigned char vec_mergeh (vector unsigned char,
8576 vector unsigned char);
8577 vector bool short vec_mergeh (vector bool short, vector bool short);
8578 vector pixel vec_mergeh (vector pixel, vector pixel);
8579 vector signed short vec_mergeh (vector signed short,
8580 vector signed short);
8581 vector unsigned short vec_mergeh (vector unsigned short,
8582 vector unsigned short);
8583 vector float vec_mergeh (vector float, vector float);
8584 vector bool int vec_mergeh (vector bool int, vector bool int);
8585 vector signed int vec_mergeh (vector signed int, vector signed int);
8586 vector unsigned int vec_mergeh (vector unsigned int,
8587 vector unsigned int);
8589 vector float vec_vmrghw (vector float, vector float);
8590 vector bool int vec_vmrghw (vector bool int, vector bool int);
8591 vector signed int vec_vmrghw (vector signed int, vector signed int);
8592 vector unsigned int vec_vmrghw (vector unsigned int,
8593 vector unsigned int);
8595 vector bool short vec_vmrghh (vector bool short, vector bool short);
8596 vector signed short vec_vmrghh (vector signed short,
8597 vector signed short);
8598 vector unsigned short vec_vmrghh (vector unsigned short,
8599 vector unsigned short);
8600 vector pixel vec_vmrghh (vector pixel, vector pixel);
8602 vector bool char vec_vmrghb (vector bool char, vector bool char);
8603 vector signed char vec_vmrghb (vector signed char, vector signed char);
8604 vector unsigned char vec_vmrghb (vector unsigned char,
8605 vector unsigned char);
8607 vector bool char vec_mergel (vector bool char, vector bool char);
8608 vector signed char vec_mergel (vector signed char, vector signed char);
8609 vector unsigned char vec_mergel (vector unsigned char,
8610 vector unsigned char);
8611 vector bool short vec_mergel (vector bool short, vector bool short);
8612 vector pixel vec_mergel (vector pixel, vector pixel);
8613 vector signed short vec_mergel (vector signed short,
8614 vector signed short);
8615 vector unsigned short vec_mergel (vector unsigned short,
8616 vector unsigned short);
8617 vector float vec_mergel (vector float, vector float);
8618 vector bool int vec_mergel (vector bool int, vector bool int);
8619 vector signed int vec_mergel (vector signed int, vector signed int);
8620 vector unsigned int vec_mergel (vector unsigned int,
8621 vector unsigned int);
8623 vector float vec_vmrglw (vector float, vector float);
8624 vector signed int vec_vmrglw (vector signed int, vector signed int);
8625 vector unsigned int vec_vmrglw (vector unsigned int,
8626 vector unsigned int);
8627 vector bool int vec_vmrglw (vector bool int, vector bool int);
8629 vector bool short vec_vmrglh (vector bool short, vector bool short);
8630 vector signed short vec_vmrglh (vector signed short,
8631 vector signed short);
8632 vector unsigned short vec_vmrglh (vector unsigned short,
8633 vector unsigned short);
8634 vector pixel vec_vmrglh (vector pixel, vector pixel);
8636 vector bool char vec_vmrglb (vector bool char, vector bool char);
8637 vector signed char vec_vmrglb (vector signed char, vector signed char);
8638 vector unsigned char vec_vmrglb (vector unsigned char,
8639 vector unsigned char);
8641 vector unsigned short vec_mfvscr (void);
8643 vector unsigned char vec_min (vector bool char, vector unsigned char);
8644 vector unsigned char vec_min (vector unsigned char, vector bool char);
8645 vector unsigned char vec_min (vector unsigned char,
8646 vector unsigned char);
8647 vector signed char vec_min (vector bool char, vector signed char);
8648 vector signed char vec_min (vector signed char, vector bool char);
8649 vector signed char vec_min (vector signed char, vector signed char);
8650 vector unsigned short vec_min (vector bool short,
8651 vector unsigned short);
8652 vector unsigned short vec_min (vector unsigned short,
8654 vector unsigned short vec_min (vector unsigned short,
8655 vector unsigned short);
8656 vector signed short vec_min (vector bool short, vector signed short);
8657 vector signed short vec_min (vector signed short, vector bool short);
8658 vector signed short vec_min (vector signed short, vector signed short);
8659 vector unsigned int vec_min (vector bool int, vector unsigned int);
8660 vector unsigned int vec_min (vector unsigned int, vector bool int);
8661 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8662 vector signed int vec_min (vector bool int, vector signed int);
8663 vector signed int vec_min (vector signed int, vector bool int);
8664 vector signed int vec_min (vector signed int, vector signed int);
8665 vector float vec_min (vector float, vector float);
8667 vector float vec_vminfp (vector float, vector float);
8669 vector signed int vec_vminsw (vector bool int, vector signed int);
8670 vector signed int vec_vminsw (vector signed int, vector bool int);
8671 vector signed int vec_vminsw (vector signed int, vector signed int);
8673 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8674 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8675 vector unsigned int vec_vminuw (vector unsigned int,
8676 vector unsigned int);
8678 vector signed short vec_vminsh (vector bool short, vector signed short);
8679 vector signed short vec_vminsh (vector signed short, vector bool short);
8680 vector signed short vec_vminsh (vector signed short,
8681 vector signed short);
8683 vector unsigned short vec_vminuh (vector bool short,
8684 vector unsigned short);
8685 vector unsigned short vec_vminuh (vector unsigned short,
8687 vector unsigned short vec_vminuh (vector unsigned short,
8688 vector unsigned short);
8690 vector signed char vec_vminsb (vector bool char, vector signed char);
8691 vector signed char vec_vminsb (vector signed char, vector bool char);
8692 vector signed char vec_vminsb (vector signed char, vector signed char);
8694 vector unsigned char vec_vminub (vector bool char,
8695 vector unsigned char);
8696 vector unsigned char vec_vminub (vector unsigned char,
8698 vector unsigned char vec_vminub (vector unsigned char,
8699 vector unsigned char);
8701 vector signed short vec_mladd (vector signed short,
8702 vector signed short,
8703 vector signed short);
8704 vector signed short vec_mladd (vector signed short,
8705 vector unsigned short,
8706 vector unsigned short);
8707 vector signed short vec_mladd (vector unsigned short,
8708 vector signed short,
8709 vector signed short);
8710 vector unsigned short vec_mladd (vector unsigned short,
8711 vector unsigned short,
8712 vector unsigned short);
8714 vector signed short vec_mradds (vector signed short,
8715 vector signed short,
8716 vector signed short);
8718 vector unsigned int vec_msum (vector unsigned char,
8719 vector unsigned char,
8720 vector unsigned int);
8721 vector signed int vec_msum (vector signed char,
8722 vector unsigned char,
8724 vector unsigned int vec_msum (vector unsigned short,
8725 vector unsigned short,
8726 vector unsigned int);
8727 vector signed int vec_msum (vector signed short,
8728 vector signed short,
8731 vector signed int vec_vmsumshm (vector signed short,
8732 vector signed short,
8735 vector unsigned int vec_vmsumuhm (vector unsigned short,
8736 vector unsigned short,
8737 vector unsigned int);
8739 vector signed int vec_vmsummbm (vector signed char,
8740 vector unsigned char,
8743 vector unsigned int vec_vmsumubm (vector unsigned char,
8744 vector unsigned char,
8745 vector unsigned int);
8747 vector unsigned int vec_msums (vector unsigned short,
8748 vector unsigned short,
8749 vector unsigned int);
8750 vector signed int vec_msums (vector signed short,
8751 vector signed short,
8754 vector signed int vec_vmsumshs (vector signed short,
8755 vector signed short,
8758 vector unsigned int vec_vmsumuhs (vector unsigned short,
8759 vector unsigned short,
8760 vector unsigned int);
8762 void vec_mtvscr (vector signed int);
8763 void vec_mtvscr (vector unsigned int);
8764 void vec_mtvscr (vector bool int);
8765 void vec_mtvscr (vector signed short);
8766 void vec_mtvscr (vector unsigned short);
8767 void vec_mtvscr (vector bool short);
8768 void vec_mtvscr (vector pixel);
8769 void vec_mtvscr (vector signed char);
8770 void vec_mtvscr (vector unsigned char);
8771 void vec_mtvscr (vector bool char);
8773 vector unsigned short vec_mule (vector unsigned char,
8774 vector unsigned char);
8775 vector signed short vec_mule (vector signed char,
8776 vector signed char);
8777 vector unsigned int vec_mule (vector unsigned short,
8778 vector unsigned short);
8779 vector signed int vec_mule (vector signed short, vector signed short);
8781 vector signed int vec_vmulesh (vector signed short,
8782 vector signed short);
8784 vector unsigned int vec_vmuleuh (vector unsigned short,
8785 vector unsigned short);
8787 vector signed short vec_vmulesb (vector signed char,
8788 vector signed char);
8790 vector unsigned short vec_vmuleub (vector unsigned char,
8791 vector unsigned char);
8793 vector unsigned short vec_mulo (vector unsigned char,
8794 vector unsigned char);
8795 vector signed short vec_mulo (vector signed char, vector signed char);
8796 vector unsigned int vec_mulo (vector unsigned short,
8797 vector unsigned short);
8798 vector signed int vec_mulo (vector signed short, vector signed short);
8800 vector signed int vec_vmulosh (vector signed short,
8801 vector signed short);
8803 vector unsigned int vec_vmulouh (vector unsigned short,
8804 vector unsigned short);
8806 vector signed short vec_vmulosb (vector signed char,
8807 vector signed char);
8809 vector unsigned short vec_vmuloub (vector unsigned char,
8810 vector unsigned char);
8812 vector float vec_nmsub (vector float, vector float, vector float);
8814 vector float vec_nor (vector float, vector float);
8815 vector signed int vec_nor (vector signed int, vector signed int);
8816 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8817 vector bool int vec_nor (vector bool int, vector bool int);
8818 vector signed short vec_nor (vector signed short, vector signed short);
8819 vector unsigned short vec_nor (vector unsigned short,
8820 vector unsigned short);
8821 vector bool short vec_nor (vector bool short, vector bool short);
8822 vector signed char vec_nor (vector signed char, vector signed char);
8823 vector unsigned char vec_nor (vector unsigned char,
8824 vector unsigned char);
8825 vector bool char vec_nor (vector bool char, vector bool char);
8827 vector float vec_or (vector float, vector float);
8828 vector float vec_or (vector float, vector bool int);
8829 vector float vec_or (vector bool int, vector float);
8830 vector bool int vec_or (vector bool int, vector bool int);
8831 vector signed int vec_or (vector bool int, vector signed int);
8832 vector signed int vec_or (vector signed int, vector bool int);
8833 vector signed int vec_or (vector signed int, vector signed int);
8834 vector unsigned int vec_or (vector bool int, vector unsigned int);
8835 vector unsigned int vec_or (vector unsigned int, vector bool int);
8836 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8837 vector bool short vec_or (vector bool short, vector bool short);
8838 vector signed short vec_or (vector bool short, vector signed short);
8839 vector signed short vec_or (vector signed short, vector bool short);
8840 vector signed short vec_or (vector signed short, vector signed short);
8841 vector unsigned short vec_or (vector bool short, vector unsigned short);
8842 vector unsigned short vec_or (vector unsigned short, vector bool short);
8843 vector unsigned short vec_or (vector unsigned short,
8844 vector unsigned short);
8845 vector signed char vec_or (vector bool char, vector signed char);
8846 vector bool char vec_or (vector bool char, vector bool char);
8847 vector signed char vec_or (vector signed char, vector bool char);
8848 vector signed char vec_or (vector signed char, vector signed char);
8849 vector unsigned char vec_or (vector bool char, vector unsigned char);
8850 vector unsigned char vec_or (vector unsigned char, vector bool char);
8851 vector unsigned char vec_or (vector unsigned char,
8852 vector unsigned char);
8854 vector signed char vec_pack (vector signed short, vector signed short);
8855 vector unsigned char vec_pack (vector unsigned short,
8856 vector unsigned short);
8857 vector bool char vec_pack (vector bool short, vector bool short);
8858 vector signed short vec_pack (vector signed int, vector signed int);
8859 vector unsigned short vec_pack (vector unsigned int,
8860 vector unsigned int);
8861 vector bool short vec_pack (vector bool int, vector bool int);
8863 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8864 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8865 vector unsigned short vec_vpkuwum (vector unsigned int,
8866 vector unsigned int);
8868 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8869 vector signed char vec_vpkuhum (vector signed short,
8870 vector signed short);
8871 vector unsigned char vec_vpkuhum (vector unsigned short,
8872 vector unsigned short);
8874 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8876 vector unsigned char vec_packs (vector unsigned short,
8877 vector unsigned short);
8878 vector signed char vec_packs (vector signed short, vector signed short);
8879 vector unsigned short vec_packs (vector unsigned int,
8880 vector unsigned int);
8881 vector signed short vec_packs (vector signed int, vector signed int);
8883 vector signed short vec_vpkswss (vector signed int, vector signed int);
8885 vector unsigned short vec_vpkuwus (vector unsigned int,
8886 vector unsigned int);
8888 vector signed char vec_vpkshss (vector signed short,
8889 vector signed short);
8891 vector unsigned char vec_vpkuhus (vector unsigned short,
8892 vector unsigned short);
8894 vector unsigned char vec_packsu (vector unsigned short,
8895 vector unsigned short);
8896 vector unsigned char vec_packsu (vector signed short,
8897 vector signed short);
8898 vector unsigned short vec_packsu (vector unsigned int,
8899 vector unsigned int);
8900 vector unsigned short vec_packsu (vector signed int, vector signed int);
8902 vector unsigned short vec_vpkswus (vector signed int,
8905 vector unsigned char vec_vpkshus (vector signed short,
8906 vector signed short);
8908 vector float vec_perm (vector float,
8910 vector unsigned char);
8911 vector signed int vec_perm (vector signed int,
8913 vector unsigned char);
8914 vector unsigned int vec_perm (vector unsigned int,
8915 vector unsigned int,
8916 vector unsigned char);
8917 vector bool int vec_perm (vector bool int,
8919 vector unsigned char);
8920 vector signed short vec_perm (vector signed short,
8921 vector signed short,
8922 vector unsigned char);
8923 vector unsigned short vec_perm (vector unsigned short,
8924 vector unsigned short,
8925 vector unsigned char);
8926 vector bool short vec_perm (vector bool short,
8928 vector unsigned char);
8929 vector pixel vec_perm (vector pixel,
8931 vector unsigned char);
8932 vector signed char vec_perm (vector signed char,
8934 vector unsigned char);
8935 vector unsigned char vec_perm (vector unsigned char,
8936 vector unsigned char,
8937 vector unsigned char);
8938 vector bool char vec_perm (vector bool char,
8940 vector unsigned char);
8942 vector float vec_re (vector float);
8944 vector signed char vec_rl (vector signed char,
8945 vector unsigned char);
8946 vector unsigned char vec_rl (vector unsigned char,
8947 vector unsigned char);
8948 vector signed short vec_rl (vector signed short, vector unsigned short);
8949 vector unsigned short vec_rl (vector unsigned short,
8950 vector unsigned short);
8951 vector signed int vec_rl (vector signed int, vector unsigned int);
8952 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8954 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8955 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8957 vector signed short vec_vrlh (vector signed short,
8958 vector unsigned short);
8959 vector unsigned short vec_vrlh (vector unsigned short,
8960 vector unsigned short);
8962 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8963 vector unsigned char vec_vrlb (vector unsigned char,
8964 vector unsigned char);
8966 vector float vec_round (vector float);
8968 vector float vec_rsqrte (vector float);
8970 vector float vec_sel (vector float, vector float, vector bool int);
8971 vector float vec_sel (vector float, vector float, vector unsigned int);
8972 vector signed int vec_sel (vector signed int,
8975 vector signed int vec_sel (vector signed int,
8977 vector unsigned int);
8978 vector unsigned int vec_sel (vector unsigned int,
8979 vector unsigned int,
8981 vector unsigned int vec_sel (vector unsigned int,
8982 vector unsigned int,
8983 vector unsigned int);
8984 vector bool int vec_sel (vector bool int,
8987 vector bool int vec_sel (vector bool int,
8989 vector unsigned int);
8990 vector signed short vec_sel (vector signed short,
8991 vector signed short,
8993 vector signed short vec_sel (vector signed short,
8994 vector signed short,
8995 vector unsigned short);
8996 vector unsigned short vec_sel (vector unsigned short,
8997 vector unsigned short,
8999 vector unsigned short vec_sel (vector unsigned short,
9000 vector unsigned short,
9001 vector unsigned short);
9002 vector bool short vec_sel (vector bool short,
9005 vector bool short vec_sel (vector bool short,
9007 vector unsigned short);
9008 vector signed char vec_sel (vector signed char,
9011 vector signed char vec_sel (vector signed char,
9013 vector unsigned char);
9014 vector unsigned char vec_sel (vector unsigned char,
9015 vector unsigned char,
9017 vector unsigned char vec_sel (vector unsigned char,
9018 vector unsigned char,
9019 vector unsigned char);
9020 vector bool char vec_sel (vector bool char,
9023 vector bool char vec_sel (vector bool char,
9025 vector unsigned char);
9027 vector signed char vec_sl (vector signed char,
9028 vector unsigned char);
9029 vector unsigned char vec_sl (vector unsigned char,
9030 vector unsigned char);
9031 vector signed short vec_sl (vector signed short, vector unsigned short);
9032 vector unsigned short vec_sl (vector unsigned short,
9033 vector unsigned short);
9034 vector signed int vec_sl (vector signed int, vector unsigned int);
9035 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
9037 vector signed int vec_vslw (vector signed int, vector unsigned int);
9038 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
9040 vector signed short vec_vslh (vector signed short,
9041 vector unsigned short);
9042 vector unsigned short vec_vslh (vector unsigned short,
9043 vector unsigned short);
9045 vector signed char vec_vslb (vector signed char, vector unsigned char);
9046 vector unsigned char vec_vslb (vector unsigned char,
9047 vector unsigned char);
9049 vector float vec_sld (vector float, vector float, const int);
9050 vector signed int vec_sld (vector signed int,
9053 vector unsigned int vec_sld (vector unsigned int,
9054 vector unsigned int,
9056 vector bool int vec_sld (vector bool int,
9059 vector signed short vec_sld (vector signed short,
9060 vector signed short,
9062 vector unsigned short vec_sld (vector unsigned short,
9063 vector unsigned short,
9065 vector bool short vec_sld (vector bool short,
9068 vector pixel vec_sld (vector pixel,
9071 vector signed char vec_sld (vector signed char,
9074 vector unsigned char vec_sld (vector unsigned char,
9075 vector unsigned char,
9077 vector bool char vec_sld (vector bool char,
9081 vector signed int vec_sll (vector signed int,
9082 vector unsigned int);
9083 vector signed int vec_sll (vector signed int,
9084 vector unsigned short);
9085 vector signed int vec_sll (vector signed int,
9086 vector unsigned char);
9087 vector unsigned int vec_sll (vector unsigned int,
9088 vector unsigned int);
9089 vector unsigned int vec_sll (vector unsigned int,
9090 vector unsigned short);
9091 vector unsigned int vec_sll (vector unsigned int,
9092 vector unsigned char);
9093 vector bool int vec_sll (vector bool int,
9094 vector unsigned int);
9095 vector bool int vec_sll (vector bool int,
9096 vector unsigned short);
9097 vector bool int vec_sll (vector bool int,
9098 vector unsigned char);
9099 vector signed short vec_sll (vector signed short,
9100 vector unsigned int);
9101 vector signed short vec_sll (vector signed short,
9102 vector unsigned short);
9103 vector signed short vec_sll (vector signed short,
9104 vector unsigned char);
9105 vector unsigned short vec_sll (vector unsigned short,
9106 vector unsigned int);
9107 vector unsigned short vec_sll (vector unsigned short,
9108 vector unsigned short);
9109 vector unsigned short vec_sll (vector unsigned short,
9110 vector unsigned char);
9111 vector bool short vec_sll (vector bool short, vector unsigned int);
9112 vector bool short vec_sll (vector bool short, vector unsigned short);
9113 vector bool short vec_sll (vector bool short, vector unsigned char);
9114 vector pixel vec_sll (vector pixel, vector unsigned int);
9115 vector pixel vec_sll (vector pixel, vector unsigned short);
9116 vector pixel vec_sll (vector pixel, vector unsigned char);
9117 vector signed char vec_sll (vector signed char, vector unsigned int);
9118 vector signed char vec_sll (vector signed char, vector unsigned short);
9119 vector signed char vec_sll (vector signed char, vector unsigned char);
9120 vector unsigned char vec_sll (vector unsigned char,
9121 vector unsigned int);
9122 vector unsigned char vec_sll (vector unsigned char,
9123 vector unsigned short);
9124 vector unsigned char vec_sll (vector unsigned char,
9125 vector unsigned char);
9126 vector bool char vec_sll (vector bool char, vector unsigned int);
9127 vector bool char vec_sll (vector bool char, vector unsigned short);
9128 vector bool char vec_sll (vector bool char, vector unsigned char);
9130 vector float vec_slo (vector float, vector signed char);
9131 vector float vec_slo (vector float, vector unsigned char);
9132 vector signed int vec_slo (vector signed int, vector signed char);
9133 vector signed int vec_slo (vector signed int, vector unsigned char);
9134 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9135 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9136 vector signed short vec_slo (vector signed short, vector signed char);
9137 vector signed short vec_slo (vector signed short, vector unsigned char);
9138 vector unsigned short vec_slo (vector unsigned short,
9139 vector signed char);
9140 vector unsigned short vec_slo (vector unsigned short,
9141 vector unsigned char);
9142 vector pixel vec_slo (vector pixel, vector signed char);
9143 vector pixel vec_slo (vector pixel, vector unsigned char);
9144 vector signed char vec_slo (vector signed char, vector signed char);
9145 vector signed char vec_slo (vector signed char, vector unsigned char);
9146 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9147 vector unsigned char vec_slo (vector unsigned char,
9148 vector unsigned char);
9150 vector signed char vec_splat (vector signed char, const int);
9151 vector unsigned char vec_splat (vector unsigned char, const int);
9152 vector bool char vec_splat (vector bool char, const int);
9153 vector signed short vec_splat (vector signed short, const int);
9154 vector unsigned short vec_splat (vector unsigned short, const int);
9155 vector bool short vec_splat (vector bool short, const int);
9156 vector pixel vec_splat (vector pixel, const int);
9157 vector float vec_splat (vector float, const int);
9158 vector signed int vec_splat (vector signed int, const int);
9159 vector unsigned int vec_splat (vector unsigned int, const int);
9160 vector bool int vec_splat (vector bool int, const int);
9162 vector float vec_vspltw (vector float, const int);
9163 vector signed int vec_vspltw (vector signed int, const int);
9164 vector unsigned int vec_vspltw (vector unsigned int, const int);
9165 vector bool int vec_vspltw (vector bool int, const int);
9167 vector bool short vec_vsplth (vector bool short, const int);
9168 vector signed short vec_vsplth (vector signed short, const int);
9169 vector unsigned short vec_vsplth (vector unsigned short, const int);
9170 vector pixel vec_vsplth (vector pixel, const int);
9172 vector signed char vec_vspltb (vector signed char, const int);
9173 vector unsigned char vec_vspltb (vector unsigned char, const int);
9174 vector bool char vec_vspltb (vector bool char, const int);
9176 vector signed char vec_splat_s8 (const int);
9178 vector signed short vec_splat_s16 (const int);
9180 vector signed int vec_splat_s32 (const int);
9182 vector unsigned char vec_splat_u8 (const int);
9184 vector unsigned short vec_splat_u16 (const int);
9186 vector unsigned int vec_splat_u32 (const int);
9188 vector signed char vec_sr (vector signed char, vector unsigned char);
9189 vector unsigned char vec_sr (vector unsigned char,
9190 vector unsigned char);
9191 vector signed short vec_sr (vector signed short,
9192 vector unsigned short);
9193 vector unsigned short vec_sr (vector unsigned short,
9194 vector unsigned short);
9195 vector signed int vec_sr (vector signed int, vector unsigned int);
9196 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9198 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9199 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9201 vector signed short vec_vsrh (vector signed short,
9202 vector unsigned short);
9203 vector unsigned short vec_vsrh (vector unsigned short,
9204 vector unsigned short);
9206 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9207 vector unsigned char vec_vsrb (vector unsigned char,
9208 vector unsigned char);
9210 vector signed char vec_sra (vector signed char, vector unsigned char);
9211 vector unsigned char vec_sra (vector unsigned char,
9212 vector unsigned char);
9213 vector signed short vec_sra (vector signed short,
9214 vector unsigned short);
9215 vector unsigned short vec_sra (vector unsigned short,
9216 vector unsigned short);
9217 vector signed int vec_sra (vector signed int, vector unsigned int);
9218 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9220 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9221 vector unsigned int vec_vsraw (vector unsigned int,
9222 vector unsigned int);
9224 vector signed short vec_vsrah (vector signed short,
9225 vector unsigned short);
9226 vector unsigned short vec_vsrah (vector unsigned short,
9227 vector unsigned short);
9229 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9230 vector unsigned char vec_vsrab (vector unsigned char,
9231 vector unsigned char);
9233 vector signed int vec_srl (vector signed int, vector unsigned int);
9234 vector signed int vec_srl (vector signed int, vector unsigned short);
9235 vector signed int vec_srl (vector signed int, vector unsigned char);
9236 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9237 vector unsigned int vec_srl (vector unsigned int,
9238 vector unsigned short);
9239 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9240 vector bool int vec_srl (vector bool int, vector unsigned int);
9241 vector bool int vec_srl (vector bool int, vector unsigned short);
9242 vector bool int vec_srl (vector bool int, vector unsigned char);
9243 vector signed short vec_srl (vector signed short, vector unsigned int);
9244 vector signed short vec_srl (vector signed short,
9245 vector unsigned short);
9246 vector signed short vec_srl (vector signed short, vector unsigned char);
9247 vector unsigned short vec_srl (vector unsigned short,
9248 vector unsigned int);
9249 vector unsigned short vec_srl (vector unsigned short,
9250 vector unsigned short);
9251 vector unsigned short vec_srl (vector unsigned short,
9252 vector unsigned char);
9253 vector bool short vec_srl (vector bool short, vector unsigned int);
9254 vector bool short vec_srl (vector bool short, vector unsigned short);
9255 vector bool short vec_srl (vector bool short, vector unsigned char);
9256 vector pixel vec_srl (vector pixel, vector unsigned int);
9257 vector pixel vec_srl (vector pixel, vector unsigned short);
9258 vector pixel vec_srl (vector pixel, vector unsigned char);
9259 vector signed char vec_srl (vector signed char, vector unsigned int);
9260 vector signed char vec_srl (vector signed char, vector unsigned short);
9261 vector signed char vec_srl (vector signed char, vector unsigned char);
9262 vector unsigned char vec_srl (vector unsigned char,
9263 vector unsigned int);
9264 vector unsigned char vec_srl (vector unsigned char,
9265 vector unsigned short);
9266 vector unsigned char vec_srl (vector unsigned char,
9267 vector unsigned char);
9268 vector bool char vec_srl (vector bool char, vector unsigned int);
9269 vector bool char vec_srl (vector bool char, vector unsigned short);
9270 vector bool char vec_srl (vector bool char, vector unsigned char);
9272 vector float vec_sro (vector float, vector signed char);
9273 vector float vec_sro (vector float, vector unsigned char);
9274 vector signed int vec_sro (vector signed int, vector signed char);
9275 vector signed int vec_sro (vector signed int, vector unsigned char);
9276 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9277 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9278 vector signed short vec_sro (vector signed short, vector signed char);
9279 vector signed short vec_sro (vector signed short, vector unsigned char);
9280 vector unsigned short vec_sro (vector unsigned short,
9281 vector signed char);
9282 vector unsigned short vec_sro (vector unsigned short,
9283 vector unsigned char);
9284 vector pixel vec_sro (vector pixel, vector signed char);
9285 vector pixel vec_sro (vector pixel, vector unsigned char);
9286 vector signed char vec_sro (vector signed char, vector signed char);
9287 vector signed char vec_sro (vector signed char, vector unsigned char);
9288 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9289 vector unsigned char vec_sro (vector unsigned char,
9290 vector unsigned char);
9292 void vec_st (vector float, int, vector float *);
9293 void vec_st (vector float, int, float *);
9294 void vec_st (vector signed int, int, vector signed int *);
9295 void vec_st (vector signed int, int, int *);
9296 void vec_st (vector unsigned int, int, vector unsigned int *);
9297 void vec_st (vector unsigned int, int, unsigned int *);
9298 void vec_st (vector bool int, int, vector bool int *);
9299 void vec_st (vector bool int, int, unsigned int *);
9300 void vec_st (vector bool int, int, int *);
9301 void vec_st (vector signed short, int, vector signed short *);
9302 void vec_st (vector signed short, int, short *);
9303 void vec_st (vector unsigned short, int, vector unsigned short *);
9304 void vec_st (vector unsigned short, int, unsigned short *);
9305 void vec_st (vector bool short, int, vector bool short *);
9306 void vec_st (vector bool short, int, unsigned short *);
9307 void vec_st (vector pixel, int, vector pixel *);
9308 void vec_st (vector pixel, int, unsigned short *);
9309 void vec_st (vector pixel, int, short *);
9310 void vec_st (vector bool short, int, short *);
9311 void vec_st (vector signed char, int, vector signed char *);
9312 void vec_st (vector signed char, int, signed char *);
9313 void vec_st (vector unsigned char, int, vector unsigned char *);
9314 void vec_st (vector unsigned char, int, unsigned char *);
9315 void vec_st (vector bool char, int, vector bool char *);
9316 void vec_st (vector bool char, int, unsigned char *);
9317 void vec_st (vector bool char, int, signed char *);
9319 void vec_ste (vector signed char, int, signed char *);
9320 void vec_ste (vector unsigned char, int, unsigned char *);
9321 void vec_ste (vector bool char, int, signed char *);
9322 void vec_ste (vector bool char, int, unsigned char *);
9323 void vec_ste (vector signed short, int, short *);
9324 void vec_ste (vector unsigned short, int, unsigned short *);
9325 void vec_ste (vector bool short, int, short *);
9326 void vec_ste (vector bool short, int, unsigned short *);
9327 void vec_ste (vector pixel, int, short *);
9328 void vec_ste (vector pixel, int, unsigned short *);
9329 void vec_ste (vector float, int, float *);
9330 void vec_ste (vector signed int, int, int *);
9331 void vec_ste (vector unsigned int, int, unsigned int *);
9332 void vec_ste (vector bool int, int, int *);
9333 void vec_ste (vector bool int, int, unsigned int *);
9335 void vec_stvewx (vector float, int, float *);
9336 void vec_stvewx (vector signed int, int, int *);
9337 void vec_stvewx (vector unsigned int, int, unsigned int *);
9338 void vec_stvewx (vector bool int, int, int *);
9339 void vec_stvewx (vector bool int, int, unsigned int *);
9341 void vec_stvehx (vector signed short, int, short *);
9342 void vec_stvehx (vector unsigned short, int, unsigned short *);
9343 void vec_stvehx (vector bool short, int, short *);
9344 void vec_stvehx (vector bool short, int, unsigned short *);
9345 void vec_stvehx (vector pixel, int, short *);
9346 void vec_stvehx (vector pixel, int, unsigned short *);
9348 void vec_stvebx (vector signed char, int, signed char *);
9349 void vec_stvebx (vector unsigned char, int, unsigned char *);
9350 void vec_stvebx (vector bool char, int, signed char *);
9351 void vec_stvebx (vector bool char, int, unsigned char *);
9353 void vec_stl (vector float, int, vector float *);
9354 void vec_stl (vector float, int, float *);
9355 void vec_stl (vector signed int, int, vector signed int *);
9356 void vec_stl (vector signed int, int, int *);
9357 void vec_stl (vector unsigned int, int, vector unsigned int *);
9358 void vec_stl (vector unsigned int, int, unsigned int *);
9359 void vec_stl (vector bool int, int, vector bool int *);
9360 void vec_stl (vector bool int, int, unsigned int *);
9361 void vec_stl (vector bool int, int, int *);
9362 void vec_stl (vector signed short, int, vector signed short *);
9363 void vec_stl (vector signed short, int, short *);
9364 void vec_stl (vector unsigned short, int, vector unsigned short *);
9365 void vec_stl (vector unsigned short, int, unsigned short *);
9366 void vec_stl (vector bool short, int, vector bool short *);
9367 void vec_stl (vector bool short, int, unsigned short *);
9368 void vec_stl (vector bool short, int, short *);
9369 void vec_stl (vector pixel, int, vector pixel *);
9370 void vec_stl (vector pixel, int, unsigned short *);
9371 void vec_stl (vector pixel, int, short *);
9372 void vec_stl (vector signed char, int, vector signed char *);
9373 void vec_stl (vector signed char, int, signed char *);
9374 void vec_stl (vector unsigned char, int, vector unsigned char *);
9375 void vec_stl (vector unsigned char, int, unsigned char *);
9376 void vec_stl (vector bool char, int, vector bool char *);
9377 void vec_stl (vector bool char, int, unsigned char *);
9378 void vec_stl (vector bool char, int, signed char *);
9380 vector signed char vec_sub (vector bool char, vector signed char);
9381 vector signed char vec_sub (vector signed char, vector bool char);
9382 vector signed char vec_sub (vector signed char, vector signed char);
9383 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9384 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9385 vector unsigned char vec_sub (vector unsigned char,
9386 vector unsigned char);
9387 vector signed short vec_sub (vector bool short, vector signed short);
9388 vector signed short vec_sub (vector signed short, vector bool short);
9389 vector signed short vec_sub (vector signed short, vector signed short);
9390 vector unsigned short vec_sub (vector bool short,
9391 vector unsigned short);
9392 vector unsigned short vec_sub (vector unsigned short,
9394 vector unsigned short vec_sub (vector unsigned short,
9395 vector unsigned short);
9396 vector signed int vec_sub (vector bool int, vector signed int);
9397 vector signed int vec_sub (vector signed int, vector bool int);
9398 vector signed int vec_sub (vector signed int, vector signed int);
9399 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9400 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9401 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9402 vector float vec_sub (vector float, vector float);
9404 vector float vec_vsubfp (vector float, vector float);
9406 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9407 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9408 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9409 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9410 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9411 vector unsigned int vec_vsubuwm (vector unsigned int,
9412 vector unsigned int);
9414 vector signed short vec_vsubuhm (vector bool short,
9415 vector signed short);
9416 vector signed short vec_vsubuhm (vector signed short,
9418 vector signed short vec_vsubuhm (vector signed short,
9419 vector signed short);
9420 vector unsigned short vec_vsubuhm (vector bool short,
9421 vector unsigned short);
9422 vector unsigned short vec_vsubuhm (vector unsigned short,
9424 vector unsigned short vec_vsubuhm (vector unsigned short,
9425 vector unsigned short);
9427 vector signed char vec_vsububm (vector bool char, vector signed char);
9428 vector signed char vec_vsububm (vector signed char, vector bool char);
9429 vector signed char vec_vsububm (vector signed char, vector signed char);
9430 vector unsigned char vec_vsububm (vector bool char,
9431 vector unsigned char);
9432 vector unsigned char vec_vsububm (vector unsigned char,
9434 vector unsigned char vec_vsububm (vector unsigned char,
9435 vector unsigned char);
9437 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9439 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9440 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9441 vector unsigned char vec_subs (vector unsigned char,
9442 vector unsigned char);
9443 vector signed char vec_subs (vector bool char, vector signed char);
9444 vector signed char vec_subs (vector signed char, vector bool char);
9445 vector signed char vec_subs (vector signed char, vector signed char);
9446 vector unsigned short vec_subs (vector bool short,
9447 vector unsigned short);
9448 vector unsigned short vec_subs (vector unsigned short,
9450 vector unsigned short vec_subs (vector unsigned short,
9451 vector unsigned short);
9452 vector signed short vec_subs (vector bool short, vector signed short);
9453 vector signed short vec_subs (vector signed short, vector bool short);
9454 vector signed short vec_subs (vector signed short, vector signed short);
9455 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9456 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9457 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9458 vector signed int vec_subs (vector bool int, vector signed int);
9459 vector signed int vec_subs (vector signed int, vector bool int);
9460 vector signed int vec_subs (vector signed int, vector signed int);
9462 vector signed int vec_vsubsws (vector bool int, vector signed int);
9463 vector signed int vec_vsubsws (vector signed int, vector bool int);
9464 vector signed int vec_vsubsws (vector signed int, vector signed int);
9466 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9467 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9468 vector unsigned int vec_vsubuws (vector unsigned int,
9469 vector unsigned int);
9471 vector signed short vec_vsubshs (vector bool short,
9472 vector signed short);
9473 vector signed short vec_vsubshs (vector signed short,
9475 vector signed short vec_vsubshs (vector signed short,
9476 vector signed short);
9478 vector unsigned short vec_vsubuhs (vector bool short,
9479 vector unsigned short);
9480 vector unsigned short vec_vsubuhs (vector unsigned short,
9482 vector unsigned short vec_vsubuhs (vector unsigned short,
9483 vector unsigned short);
9485 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9486 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9487 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9489 vector unsigned char vec_vsububs (vector bool char,
9490 vector unsigned char);
9491 vector unsigned char vec_vsububs (vector unsigned char,
9493 vector unsigned char vec_vsububs (vector unsigned char,
9494 vector unsigned char);
9496 vector unsigned int vec_sum4s (vector unsigned char,
9497 vector unsigned int);
9498 vector signed int vec_sum4s (vector signed char, vector signed int);
9499 vector signed int vec_sum4s (vector signed short, vector signed int);
9501 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9503 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9505 vector unsigned int vec_vsum4ubs (vector unsigned char,
9506 vector unsigned int);
9508 vector signed int vec_sum2s (vector signed int, vector signed int);
9510 vector signed int vec_sums (vector signed int, vector signed int);
9512 vector float vec_trunc (vector float);
9514 vector signed short vec_unpackh (vector signed char);
9515 vector bool short vec_unpackh (vector bool char);
9516 vector signed int vec_unpackh (vector signed short);
9517 vector bool int vec_unpackh (vector bool short);
9518 vector unsigned int vec_unpackh (vector pixel);
9520 vector bool int vec_vupkhsh (vector bool short);
9521 vector signed int vec_vupkhsh (vector signed short);
9523 vector unsigned int vec_vupkhpx (vector pixel);
9525 vector bool short vec_vupkhsb (vector bool char);
9526 vector signed short vec_vupkhsb (vector signed char);
9528 vector signed short vec_unpackl (vector signed char);
9529 vector bool short vec_unpackl (vector bool char);
9530 vector unsigned int vec_unpackl (vector pixel);
9531 vector signed int vec_unpackl (vector signed short);
9532 vector bool int vec_unpackl (vector bool short);
9534 vector unsigned int vec_vupklpx (vector pixel);
9536 vector bool int vec_vupklsh (vector bool short);
9537 vector signed int vec_vupklsh (vector signed short);
9539 vector bool short vec_vupklsb (vector bool char);
9540 vector signed short vec_vupklsb (vector signed char);
9542 vector float vec_xor (vector float, vector float);
9543 vector float vec_xor (vector float, vector bool int);
9544 vector float vec_xor (vector bool int, vector float);
9545 vector bool int vec_xor (vector bool int, vector bool int);
9546 vector signed int vec_xor (vector bool int, vector signed int);
9547 vector signed int vec_xor (vector signed int, vector bool int);
9548 vector signed int vec_xor (vector signed int, vector signed int);
9549 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9550 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9551 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9552 vector bool short vec_xor (vector bool short, vector bool short);
9553 vector signed short vec_xor (vector bool short, vector signed short);
9554 vector signed short vec_xor (vector signed short, vector bool short);
9555 vector signed short vec_xor (vector signed short, vector signed short);
9556 vector unsigned short vec_xor (vector bool short,
9557 vector unsigned short);
9558 vector unsigned short vec_xor (vector unsigned short,
9560 vector unsigned short vec_xor (vector unsigned short,
9561 vector unsigned short);
9562 vector signed char vec_xor (vector bool char, vector signed char);
9563 vector bool char vec_xor (vector bool char, vector bool char);
9564 vector signed char vec_xor (vector signed char, vector bool char);
9565 vector signed char vec_xor (vector signed char, vector signed char);
9566 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9567 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9568 vector unsigned char vec_xor (vector unsigned char,
9569 vector unsigned char);
9571 int vec_all_eq (vector signed char, vector bool char);
9572 int vec_all_eq (vector signed char, vector signed char);
9573 int vec_all_eq (vector unsigned char, vector bool char);
9574 int vec_all_eq (vector unsigned char, vector unsigned char);
9575 int vec_all_eq (vector bool char, vector bool char);
9576 int vec_all_eq (vector bool char, vector unsigned char);
9577 int vec_all_eq (vector bool char, vector signed char);
9578 int vec_all_eq (vector signed short, vector bool short);
9579 int vec_all_eq (vector signed short, vector signed short);
9580 int vec_all_eq (vector unsigned short, vector bool short);
9581 int vec_all_eq (vector unsigned short, vector unsigned short);
9582 int vec_all_eq (vector bool short, vector bool short);
9583 int vec_all_eq (vector bool short, vector unsigned short);
9584 int vec_all_eq (vector bool short, vector signed short);
9585 int vec_all_eq (vector pixel, vector pixel);
9586 int vec_all_eq (vector signed int, vector bool int);
9587 int vec_all_eq (vector signed int, vector signed int);
9588 int vec_all_eq (vector unsigned int, vector bool int);
9589 int vec_all_eq (vector unsigned int, vector unsigned int);
9590 int vec_all_eq (vector bool int, vector bool int);
9591 int vec_all_eq (vector bool int, vector unsigned int);
9592 int vec_all_eq (vector bool int, vector signed int);
9593 int vec_all_eq (vector float, vector float);
9595 int vec_all_ge (vector bool char, vector unsigned char);
9596 int vec_all_ge (vector unsigned char, vector bool char);
9597 int vec_all_ge (vector unsigned char, vector unsigned char);
9598 int vec_all_ge (vector bool char, vector signed char);
9599 int vec_all_ge (vector signed char, vector bool char);
9600 int vec_all_ge (vector signed char, vector signed char);
9601 int vec_all_ge (vector bool short, vector unsigned short);
9602 int vec_all_ge (vector unsigned short, vector bool short);
9603 int vec_all_ge (vector unsigned short, vector unsigned short);
9604 int vec_all_ge (vector signed short, vector signed short);
9605 int vec_all_ge (vector bool short, vector signed short);
9606 int vec_all_ge (vector signed short, vector bool short);
9607 int vec_all_ge (vector bool int, vector unsigned int);
9608 int vec_all_ge (vector unsigned int, vector bool int);
9609 int vec_all_ge (vector unsigned int, vector unsigned int);
9610 int vec_all_ge (vector bool int, vector signed int);
9611 int vec_all_ge (vector signed int, vector bool int);
9612 int vec_all_ge (vector signed int, vector signed int);
9613 int vec_all_ge (vector float, vector float);
9615 int vec_all_gt (vector bool char, vector unsigned char);
9616 int vec_all_gt (vector unsigned char, vector bool char);
9617 int vec_all_gt (vector unsigned char, vector unsigned char);
9618 int vec_all_gt (vector bool char, vector signed char);
9619 int vec_all_gt (vector signed char, vector bool char);
9620 int vec_all_gt (vector signed char, vector signed char);
9621 int vec_all_gt (vector bool short, vector unsigned short);
9622 int vec_all_gt (vector unsigned short, vector bool short);
9623 int vec_all_gt (vector unsigned short, vector unsigned short);
9624 int vec_all_gt (vector bool short, vector signed short);
9625 int vec_all_gt (vector signed short, vector bool short);
9626 int vec_all_gt (vector signed short, vector signed short);
9627 int vec_all_gt (vector bool int, vector unsigned int);
9628 int vec_all_gt (vector unsigned int, vector bool int);
9629 int vec_all_gt (vector unsigned int, vector unsigned int);
9630 int vec_all_gt (vector bool int, vector signed int);
9631 int vec_all_gt (vector signed int, vector bool int);
9632 int vec_all_gt (vector signed int, vector signed int);
9633 int vec_all_gt (vector float, vector float);
9635 int vec_all_in (vector float, vector float);
9637 int vec_all_le (vector bool char, vector unsigned char);
9638 int vec_all_le (vector unsigned char, vector bool char);
9639 int vec_all_le (vector unsigned char, vector unsigned char);
9640 int vec_all_le (vector bool char, vector signed char);
9641 int vec_all_le (vector signed char, vector bool char);
9642 int vec_all_le (vector signed char, vector signed char);
9643 int vec_all_le (vector bool short, vector unsigned short);
9644 int vec_all_le (vector unsigned short, vector bool short);
9645 int vec_all_le (vector unsigned short, vector unsigned short);
9646 int vec_all_le (vector bool short, vector signed short);
9647 int vec_all_le (vector signed short, vector bool short);
9648 int vec_all_le (vector signed short, vector signed short);
9649 int vec_all_le (vector bool int, vector unsigned int);
9650 int vec_all_le (vector unsigned int, vector bool int);
9651 int vec_all_le (vector unsigned int, vector unsigned int);
9652 int vec_all_le (vector bool int, vector signed int);
9653 int vec_all_le (vector signed int, vector bool int);
9654 int vec_all_le (vector signed int, vector signed int);
9655 int vec_all_le (vector float, vector float);
9657 int vec_all_lt (vector bool char, vector unsigned char);
9658 int vec_all_lt (vector unsigned char, vector bool char);
9659 int vec_all_lt (vector unsigned char, vector unsigned char);
9660 int vec_all_lt (vector bool char, vector signed char);
9661 int vec_all_lt (vector signed char, vector bool char);
9662 int vec_all_lt (vector signed char, vector signed char);
9663 int vec_all_lt (vector bool short, vector unsigned short);
9664 int vec_all_lt (vector unsigned short, vector bool short);
9665 int vec_all_lt (vector unsigned short, vector unsigned short);
9666 int vec_all_lt (vector bool short, vector signed short);
9667 int vec_all_lt (vector signed short, vector bool short);
9668 int vec_all_lt (vector signed short, vector signed short);
9669 int vec_all_lt (vector bool int, vector unsigned int);
9670 int vec_all_lt (vector unsigned int, vector bool int);
9671 int vec_all_lt (vector unsigned int, vector unsigned int);
9672 int vec_all_lt (vector bool int, vector signed int);
9673 int vec_all_lt (vector signed int, vector bool int);
9674 int vec_all_lt (vector signed int, vector signed int);
9675 int vec_all_lt (vector float, vector float);
9677 int vec_all_nan (vector float);
9679 int vec_all_ne (vector signed char, vector bool char);
9680 int vec_all_ne (vector signed char, vector signed char);
9681 int vec_all_ne (vector unsigned char, vector bool char);
9682 int vec_all_ne (vector unsigned char, vector unsigned char);
9683 int vec_all_ne (vector bool char, vector bool char);
9684 int vec_all_ne (vector bool char, vector unsigned char);
9685 int vec_all_ne (vector bool char, vector signed char);
9686 int vec_all_ne (vector signed short, vector bool short);
9687 int vec_all_ne (vector signed short, vector signed short);
9688 int vec_all_ne (vector unsigned short, vector bool short);
9689 int vec_all_ne (vector unsigned short, vector unsigned short);
9690 int vec_all_ne (vector bool short, vector bool short);
9691 int vec_all_ne (vector bool short, vector unsigned short);
9692 int vec_all_ne (vector bool short, vector signed short);
9693 int vec_all_ne (vector pixel, vector pixel);
9694 int vec_all_ne (vector signed int, vector bool int);
9695 int vec_all_ne (vector signed int, vector signed int);
9696 int vec_all_ne (vector unsigned int, vector bool int);
9697 int vec_all_ne (vector unsigned int, vector unsigned int);
9698 int vec_all_ne (vector bool int, vector bool int);
9699 int vec_all_ne (vector bool int, vector unsigned int);
9700 int vec_all_ne (vector bool int, vector signed int);
9701 int vec_all_ne (vector float, vector float);
9703 int vec_all_nge (vector float, vector float);
9705 int vec_all_ngt (vector float, vector float);
9707 int vec_all_nle (vector float, vector float);
9709 int vec_all_nlt (vector float, vector float);
9711 int vec_all_numeric (vector float);
9713 int vec_any_eq (vector signed char, vector bool char);
9714 int vec_any_eq (vector signed char, vector signed char);
9715 int vec_any_eq (vector unsigned char, vector bool char);
9716 int vec_any_eq (vector unsigned char, vector unsigned char);
9717 int vec_any_eq (vector bool char, vector bool char);
9718 int vec_any_eq (vector bool char, vector unsigned char);
9719 int vec_any_eq (vector bool char, vector signed char);
9720 int vec_any_eq (vector signed short, vector bool short);
9721 int vec_any_eq (vector signed short, vector signed short);
9722 int vec_any_eq (vector unsigned short, vector bool short);
9723 int vec_any_eq (vector unsigned short, vector unsigned short);
9724 int vec_any_eq (vector bool short, vector bool short);
9725 int vec_any_eq (vector bool short, vector unsigned short);
9726 int vec_any_eq (vector bool short, vector signed short);
9727 int vec_any_eq (vector pixel, vector pixel);
9728 int vec_any_eq (vector signed int, vector bool int);
9729 int vec_any_eq (vector signed int, vector signed int);
9730 int vec_any_eq (vector unsigned int, vector bool int);
9731 int vec_any_eq (vector unsigned int, vector unsigned int);
9732 int vec_any_eq (vector bool int, vector bool int);
9733 int vec_any_eq (vector bool int, vector unsigned int);
9734 int vec_any_eq (vector bool int, vector signed int);
9735 int vec_any_eq (vector float, vector float);
9737 int vec_any_ge (vector signed char, vector bool char);
9738 int vec_any_ge (vector unsigned char, vector bool char);
9739 int vec_any_ge (vector unsigned char, vector unsigned char);
9740 int vec_any_ge (vector signed char, vector signed char);
9741 int vec_any_ge (vector bool char, vector unsigned char);
9742 int vec_any_ge (vector bool char, vector signed char);
9743 int vec_any_ge (vector unsigned short, vector bool short);
9744 int vec_any_ge (vector unsigned short, vector unsigned short);
9745 int vec_any_ge (vector signed short, vector signed short);
9746 int vec_any_ge (vector signed short, vector bool short);
9747 int vec_any_ge (vector bool short, vector unsigned short);
9748 int vec_any_ge (vector bool short, vector signed short);
9749 int vec_any_ge (vector signed int, vector bool int);
9750 int vec_any_ge (vector unsigned int, vector bool int);
9751 int vec_any_ge (vector unsigned int, vector unsigned int);
9752 int vec_any_ge (vector signed int, vector signed int);
9753 int vec_any_ge (vector bool int, vector unsigned int);
9754 int vec_any_ge (vector bool int, vector signed int);
9755 int vec_any_ge (vector float, vector float);
9757 int vec_any_gt (vector bool char, vector unsigned char);
9758 int vec_any_gt (vector unsigned char, vector bool char);
9759 int vec_any_gt (vector unsigned char, vector unsigned char);
9760 int vec_any_gt (vector bool char, vector signed char);
9761 int vec_any_gt (vector signed char, vector bool char);
9762 int vec_any_gt (vector signed char, vector signed char);
9763 int vec_any_gt (vector bool short, vector unsigned short);
9764 int vec_any_gt (vector unsigned short, vector bool short);
9765 int vec_any_gt (vector unsigned short, vector unsigned short);
9766 int vec_any_gt (vector bool short, vector signed short);
9767 int vec_any_gt (vector signed short, vector bool short);
9768 int vec_any_gt (vector signed short, vector signed short);
9769 int vec_any_gt (vector bool int, vector unsigned int);
9770 int vec_any_gt (vector unsigned int, vector bool int);
9771 int vec_any_gt (vector unsigned int, vector unsigned int);
9772 int vec_any_gt (vector bool int, vector signed int);
9773 int vec_any_gt (vector signed int, vector bool int);
9774 int vec_any_gt (vector signed int, vector signed int);
9775 int vec_any_gt (vector float, vector float);
9777 int vec_any_le (vector bool char, vector unsigned char);
9778 int vec_any_le (vector unsigned char, vector bool char);
9779 int vec_any_le (vector unsigned char, vector unsigned char);
9780 int vec_any_le (vector bool char, vector signed char);
9781 int vec_any_le (vector signed char, vector bool char);
9782 int vec_any_le (vector signed char, vector signed char);
9783 int vec_any_le (vector bool short, vector unsigned short);
9784 int vec_any_le (vector unsigned short, vector bool short);
9785 int vec_any_le (vector unsigned short, vector unsigned short);
9786 int vec_any_le (vector bool short, vector signed short);
9787 int vec_any_le (vector signed short, vector bool short);
9788 int vec_any_le (vector signed short, vector signed short);
9789 int vec_any_le (vector bool int, vector unsigned int);
9790 int vec_any_le (vector unsigned int, vector bool int);
9791 int vec_any_le (vector unsigned int, vector unsigned int);
9792 int vec_any_le (vector bool int, vector signed int);
9793 int vec_any_le (vector signed int, vector bool int);
9794 int vec_any_le (vector signed int, vector signed int);
9795 int vec_any_le (vector float, vector float);
9797 int vec_any_lt (vector bool char, vector unsigned char);
9798 int vec_any_lt (vector unsigned char, vector bool char);
9799 int vec_any_lt (vector unsigned char, vector unsigned char);
9800 int vec_any_lt (vector bool char, vector signed char);
9801 int vec_any_lt (vector signed char, vector bool char);
9802 int vec_any_lt (vector signed char, vector signed char);
9803 int vec_any_lt (vector bool short, vector unsigned short);
9804 int vec_any_lt (vector unsigned short, vector bool short);
9805 int vec_any_lt (vector unsigned short, vector unsigned short);
9806 int vec_any_lt (vector bool short, vector signed short);
9807 int vec_any_lt (vector signed short, vector bool short);
9808 int vec_any_lt (vector signed short, vector signed short);
9809 int vec_any_lt (vector bool int, vector unsigned int);
9810 int vec_any_lt (vector unsigned int, vector bool int);
9811 int vec_any_lt (vector unsigned int, vector unsigned int);
9812 int vec_any_lt (vector bool int, vector signed int);
9813 int vec_any_lt (vector signed int, vector bool int);
9814 int vec_any_lt (vector signed int, vector signed int);
9815 int vec_any_lt (vector float, vector float);
9817 int vec_any_nan (vector float);
9819 int vec_any_ne (vector signed char, vector bool char);
9820 int vec_any_ne (vector signed char, vector signed char);
9821 int vec_any_ne (vector unsigned char, vector bool char);
9822 int vec_any_ne (vector unsigned char, vector unsigned char);
9823 int vec_any_ne (vector bool char, vector bool char);
9824 int vec_any_ne (vector bool char, vector unsigned char);
9825 int vec_any_ne (vector bool char, vector signed char);
9826 int vec_any_ne (vector signed short, vector bool short);
9827 int vec_any_ne (vector signed short, vector signed short);
9828 int vec_any_ne (vector unsigned short, vector bool short);
9829 int vec_any_ne (vector unsigned short, vector unsigned short);
9830 int vec_any_ne (vector bool short, vector bool short);
9831 int vec_any_ne (vector bool short, vector unsigned short);
9832 int vec_any_ne (vector bool short, vector signed short);
9833 int vec_any_ne (vector pixel, vector pixel);
9834 int vec_any_ne (vector signed int, vector bool int);
9835 int vec_any_ne (vector signed int, vector signed int);
9836 int vec_any_ne (vector unsigned int, vector bool int);
9837 int vec_any_ne (vector unsigned int, vector unsigned int);
9838 int vec_any_ne (vector bool int, vector bool int);
9839 int vec_any_ne (vector bool int, vector unsigned int);
9840 int vec_any_ne (vector bool int, vector signed int);
9841 int vec_any_ne (vector float, vector float);
9843 int vec_any_nge (vector float, vector float);
9845 int vec_any_ngt (vector float, vector float);
9847 int vec_any_nle (vector float, vector float);
9849 int vec_any_nlt (vector float, vector float);
9851 int vec_any_numeric (vector float);
9853 int vec_any_out (vector float, vector float);
9856 @node SPARC VIS Built-in Functions
9857 @subsection SPARC VIS Built-in Functions
9859 GCC supports SIMD operations on the SPARC using both the generic vector
9860 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9861 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9862 switch, the VIS extension is exposed as the following built-in functions:
9865 typedef int v2si __attribute__ ((vector_size (8)));
9866 typedef short v4hi __attribute__ ((vector_size (8)));
9867 typedef short v2hi __attribute__ ((vector_size (4)));
9868 typedef char v8qi __attribute__ ((vector_size (8)));
9869 typedef char v4qi __attribute__ ((vector_size (4)));
9871 void * __builtin_vis_alignaddr (void *, long);
9872 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9873 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9874 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9875 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9877 v4hi __builtin_vis_fexpand (v4qi);
9879 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9880 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9881 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9882 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9883 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9884 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9885 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9887 v4qi __builtin_vis_fpack16 (v4hi);
9888 v8qi __builtin_vis_fpack32 (v2si, v2si);
9889 v2hi __builtin_vis_fpackfix (v2si);
9890 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9892 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9895 @node Target Format Checks
9896 @section Format Checks Specific to Particular Target Machines
9898 For some target machines, GCC supports additional options to the
9900 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9903 * Solaris Format Checks::
9906 @node Solaris Format Checks
9907 @subsection Solaris Format Checks
9909 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9910 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9911 conversions, and the two-argument @code{%b} conversion for displaying
9912 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9915 @section Pragmas Accepted by GCC
9919 GCC supports several types of pragmas, primarily in order to compile
9920 code originally written for other compilers. Note that in general
9921 we do not recommend the use of pragmas; @xref{Function Attributes},
9922 for further explanation.
9927 * RS/6000 and PowerPC Pragmas::
9930 * Symbol-Renaming Pragmas::
9931 * Structure-Packing Pragmas::
9933 * Diagnostic Pragmas::
9934 * Visibility Pragmas::
9938 @subsection ARM Pragmas
9940 The ARM target defines pragmas for controlling the default addition of
9941 @code{long_call} and @code{short_call} attributes to functions.
9942 @xref{Function Attributes}, for information about the effects of these
9947 @cindex pragma, long_calls
9948 Set all subsequent functions to have the @code{long_call} attribute.
9951 @cindex pragma, no_long_calls
9952 Set all subsequent functions to have the @code{short_call} attribute.
9954 @item long_calls_off
9955 @cindex pragma, long_calls_off
9956 Do not affect the @code{long_call} or @code{short_call} attributes of
9957 subsequent functions.
9961 @subsection M32C Pragmas
9964 @item memregs @var{number}
9965 @cindex pragma, memregs
9966 Overrides the command line option @code{-memregs=} for the current
9967 file. Use with care! This pragma must be before any function in the
9968 file, and mixing different memregs values in different objects may
9969 make them incompatible. This pragma is useful when a
9970 performance-critical function uses a memreg for temporary values,
9971 as it may allow you to reduce the number of memregs used.
9975 @node RS/6000 and PowerPC Pragmas
9976 @subsection RS/6000 and PowerPC Pragmas
9978 The RS/6000 and PowerPC targets define one pragma for controlling
9979 whether or not the @code{longcall} attribute is added to function
9980 declarations by default. This pragma overrides the @option{-mlongcall}
9981 option, but not the @code{longcall} and @code{shortcall} attributes.
9982 @xref{RS/6000 and PowerPC Options}, for more information about when long
9983 calls are and are not necessary.
9987 @cindex pragma, longcall
9988 Apply the @code{longcall} attribute to all subsequent function
9992 Do not apply the @code{longcall} attribute to subsequent function
9996 @c Describe c4x pragmas here.
9997 @c Describe h8300 pragmas here.
9998 @c Describe sh pragmas here.
9999 @c Describe v850 pragmas here.
10001 @node Darwin Pragmas
10002 @subsection Darwin Pragmas
10004 The following pragmas are available for all architectures running the
10005 Darwin operating system. These are useful for compatibility with other
10009 @item mark @var{tokens}@dots{}
10010 @cindex pragma, mark
10011 This pragma is accepted, but has no effect.
10013 @item options align=@var{alignment}
10014 @cindex pragma, options align
10015 This pragma sets the alignment of fields in structures. The values of
10016 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
10017 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
10018 properly; to restore the previous setting, use @code{reset} for the
10021 @item segment @var{tokens}@dots{}
10022 @cindex pragma, segment
10023 This pragma is accepted, but has no effect.
10025 @item unused (@var{var} [, @var{var}]@dots{})
10026 @cindex pragma, unused
10027 This pragma declares variables to be possibly unused. GCC will not
10028 produce warnings for the listed variables. The effect is similar to
10029 that of the @code{unused} attribute, except that this pragma may appear
10030 anywhere within the variables' scopes.
10033 @node Solaris Pragmas
10034 @subsection Solaris Pragmas
10036 The Solaris target supports @code{#pragma redefine_extname}
10037 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
10038 @code{#pragma} directives for compatibility with the system compiler.
10041 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
10042 @cindex pragma, align
10044 Increase the minimum alignment of each @var{variable} to @var{alignment}.
10045 This is the same as GCC's @code{aligned} attribute @pxref{Variable
10046 Attributes}). Macro expansion occurs on the arguments to this pragma
10047 when compiling C. It does not currently occur when compiling C++, but
10048 this is a bug which may be fixed in a future release.
10050 @item fini (@var{function} [, @var{function}]...)
10051 @cindex pragma, fini
10053 This pragma causes each listed @var{function} to be called after
10054 main, or during shared module unloading, by adding a call to the
10055 @code{.fini} section.
10057 @item init (@var{function} [, @var{function}]...)
10058 @cindex pragma, init
10060 This pragma causes each listed @var{function} to be called during
10061 initialization (before @code{main}) or during shared module loading, by
10062 adding a call to the @code{.init} section.
10066 @node Symbol-Renaming Pragmas
10067 @subsection Symbol-Renaming Pragmas
10069 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
10070 supports two @code{#pragma} directives which change the name used in
10071 assembly for a given declaration. These pragmas are only available on
10072 platforms whose system headers need them. To get this effect on all
10073 platforms supported by GCC, use the asm labels extension (@pxref{Asm
10077 @item redefine_extname @var{oldname} @var{newname}
10078 @cindex pragma, redefine_extname
10080 This pragma gives the C function @var{oldname} the assembly symbol
10081 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
10082 will be defined if this pragma is available (currently only on
10085 @item extern_prefix @var{string}
10086 @cindex pragma, extern_prefix
10088 This pragma causes all subsequent external function and variable
10089 declarations to have @var{string} prepended to their assembly symbols.
10090 This effect may be terminated with another @code{extern_prefix} pragma
10091 whose argument is an empty string. The preprocessor macro
10092 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10093 available (currently only on Tru64 UNIX)@.
10096 These pragmas and the asm labels extension interact in a complicated
10097 manner. Here are some corner cases you may want to be aware of.
10100 @item Both pragmas silently apply only to declarations with external
10101 linkage. Asm labels do not have this restriction.
10103 @item In C++, both pragmas silently apply only to declarations with
10104 ``C'' linkage. Again, asm labels do not have this restriction.
10106 @item If any of the three ways of changing the assembly name of a
10107 declaration is applied to a declaration whose assembly name has
10108 already been determined (either by a previous use of one of these
10109 features, or because the compiler needed the assembly name in order to
10110 generate code), and the new name is different, a warning issues and
10111 the name does not change.
10113 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10114 always the C-language name.
10116 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10117 occurs with an asm label attached, the prefix is silently ignored for
10120 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10121 apply to the same declaration, whichever triggered first wins, and a
10122 warning issues if they contradict each other. (We would like to have
10123 @code{#pragma redefine_extname} always win, for consistency with asm
10124 labels, but if @code{#pragma extern_prefix} triggers first we have no
10125 way of knowing that that happened.)
10128 @node Structure-Packing Pragmas
10129 @subsection Structure-Packing Pragmas
10131 For compatibility with Win32, GCC supports a set of @code{#pragma}
10132 directives which change the maximum alignment of members of structures
10133 (other than zero-width bitfields), unions, and classes subsequently
10134 defined. The @var{n} value below always is required to be a small power
10135 of two and specifies the new alignment in bytes.
10138 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10139 @item @code{#pragma pack()} sets the alignment to the one that was in
10140 effect when compilation started (see also command line option
10141 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10142 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10143 setting on an internal stack and then optionally sets the new alignment.
10144 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10145 saved at the top of the internal stack (and removes that stack entry).
10146 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10147 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10148 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10149 @code{#pragma pack(pop)}.
10152 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10153 @code{#pragma} which lays out a structure as the documented
10154 @code{__attribute__ ((ms_struct))}.
10156 @item @code{#pragma ms_struct on} turns on the layout for structures
10158 @item @code{#pragma ms_struct off} turns off the layout for structures
10160 @item @code{#pragma ms_struct reset} goes back to the default layout.
10164 @subsection Weak Pragmas
10166 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10167 directives for declaring symbols to be weak, and defining weak
10171 @item #pragma weak @var{symbol}
10172 @cindex pragma, weak
10173 This pragma declares @var{symbol} to be weak, as if the declaration
10174 had the attribute of the same name. The pragma may appear before
10175 or after the declaration of @var{symbol}, but must appear before
10176 either its first use or its definition. It is not an error for
10177 @var{symbol} to never be defined at all.
10179 @item #pragma weak @var{symbol1} = @var{symbol2}
10180 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10181 It is an error if @var{symbol2} is not defined in the current
10185 @node Diagnostic Pragmas
10186 @subsection Diagnostic Pragmas
10188 GCC allows the user to selectively enable or disable certain types of
10189 diagnostics, and change the kind of the diagnostic. For example, a
10190 project's policy might require that all sources compile with
10191 @option{-Werror} but certain files might have exceptions allowing
10192 specific types of warnings. Or, a project might selectively enable
10193 diagnostics and treat them as errors depending on which preprocessor
10194 macros are defined.
10197 @item #pragma GCC diagnostic @var{kind} @var{option}
10198 @cindex pragma, diagnostic
10200 Modifies the disposition of a diagnostic. Note that not all
10201 diagnostics are modifiable; at the moment only warnings (normally
10202 controlled by @samp{-W...}) can be controlled, and not all of them.
10203 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10204 are controllable and which option controls them.
10206 @var{kind} is @samp{error} to treat this diagnostic as an error,
10207 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10208 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10209 @var{option} is a double quoted string which matches the command line
10213 #pragma GCC diagnostic warning "-Wformat"
10214 #pragma GCC diagnostic error "-Wformat"
10215 #pragma GCC diagnostic ignored "-Wformat"
10218 Note that these pragmas override any command line options. Also,
10219 while it is syntactically valid to put these pragmas anywhere in your
10220 sources, the only supported location for them is before any data or
10221 functions are defined. Doing otherwise may result in unpredictable
10222 results depending on how the optimizer manages your sources. If the
10223 same option is listed multiple times, the last one specified is the
10224 one that is in effect. This pragma is not intended to be a general
10225 purpose replacement for command line options, but for implementing
10226 strict control over project policies.
10230 @node Visibility Pragmas
10231 @subsection Visibility Pragmas
10234 @item #pragma GCC visibility push(@var{visibility})
10235 @itemx #pragma GCC visibility pop
10236 @cindex pragma, visibility
10238 This pragma allows the user to set the visibility for multiple
10239 declarations without having to give each a visibility attribute
10240 @xref{Function Attributes}, for more information about visibility and
10241 the attribute syntax.
10243 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10244 declarations. Class members and template specializations are not
10245 affected; if you want to override the visibility for a particular
10246 member or instantiation, you must use an attribute.
10250 @node Unnamed Fields
10251 @section Unnamed struct/union fields within structs/unions
10255 For compatibility with other compilers, GCC allows you to define
10256 a structure or union that contains, as fields, structures and unions
10257 without names. For example:
10270 In this example, the user would be able to access members of the unnamed
10271 union with code like @samp{foo.b}. Note that only unnamed structs and
10272 unions are allowed, you may not have, for example, an unnamed
10275 You must never create such structures that cause ambiguous field definitions.
10276 For example, this structure:
10287 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10288 Such constructs are not supported and must be avoided. In the future,
10289 such constructs may be detected and treated as compilation errors.
10291 @opindex fms-extensions
10292 Unless @option{-fms-extensions} is used, the unnamed field must be a
10293 structure or union definition without a tag (for example, @samp{struct
10294 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10295 also be a definition with a tag such as @samp{struct foo @{ int a;
10296 @};}, a reference to a previously defined structure or union such as
10297 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10298 previously defined structure or union type.
10301 @section Thread-Local Storage
10302 @cindex Thread-Local Storage
10303 @cindex @acronym{TLS}
10306 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10307 are allocated such that there is one instance of the variable per extant
10308 thread. The run-time model GCC uses to implement this originates
10309 in the IA-64 processor-specific ABI, but has since been migrated
10310 to other processors as well. It requires significant support from
10311 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10312 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10313 is not available everywhere.
10315 At the user level, the extension is visible with a new storage
10316 class keyword: @code{__thread}. For example:
10320 extern __thread struct state s;
10321 static __thread char *p;
10324 The @code{__thread} specifier may be used alone, with the @code{extern}
10325 or @code{static} specifiers, but with no other storage class specifier.
10326 When used with @code{extern} or @code{static}, @code{__thread} must appear
10327 immediately after the other storage class specifier.
10329 The @code{__thread} specifier may be applied to any global, file-scoped
10330 static, function-scoped static, or static data member of a class. It may
10331 not be applied to block-scoped automatic or non-static data member.
10333 When the address-of operator is applied to a thread-local variable, it is
10334 evaluated at run-time and returns the address of the current thread's
10335 instance of that variable. An address so obtained may be used by any
10336 thread. When a thread terminates, any pointers to thread-local variables
10337 in that thread become invalid.
10339 No static initialization may refer to the address of a thread-local variable.
10341 In C++, if an initializer is present for a thread-local variable, it must
10342 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10345 See @uref{http://people.redhat.com/drepper/tls.pdf,
10346 ELF Handling For Thread-Local Storage} for a detailed explanation of
10347 the four thread-local storage addressing models, and how the run-time
10348 is expected to function.
10351 * C99 Thread-Local Edits::
10352 * C++98 Thread-Local Edits::
10355 @node C99 Thread-Local Edits
10356 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10358 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10359 that document the exact semantics of the language extension.
10363 @cite{5.1.2 Execution environments}
10365 Add new text after paragraph 1
10368 Within either execution environment, a @dfn{thread} is a flow of
10369 control within a program. It is implementation defined whether
10370 or not there may be more than one thread associated with a program.
10371 It is implementation defined how threads beyond the first are
10372 created, the name and type of the function called at thread
10373 startup, and how threads may be terminated. However, objects
10374 with thread storage duration shall be initialized before thread
10379 @cite{6.2.4 Storage durations of objects}
10381 Add new text before paragraph 3
10384 An object whose identifier is declared with the storage-class
10385 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10386 Its lifetime is the entire execution of the thread, and its
10387 stored value is initialized only once, prior to thread startup.
10391 @cite{6.4.1 Keywords}
10393 Add @code{__thread}.
10396 @cite{6.7.1 Storage-class specifiers}
10398 Add @code{__thread} to the list of storage class specifiers in
10401 Change paragraph 2 to
10404 With the exception of @code{__thread}, at most one storage-class
10405 specifier may be given [@dots{}]. The @code{__thread} specifier may
10406 be used alone, or immediately following @code{extern} or
10410 Add new text after paragraph 6
10413 The declaration of an identifier for a variable that has
10414 block scope that specifies @code{__thread} shall also
10415 specify either @code{extern} or @code{static}.
10417 The @code{__thread} specifier shall be used only with
10422 @node C++98 Thread-Local Edits
10423 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10425 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10426 that document the exact semantics of the language extension.
10430 @b{[intro.execution]}
10432 New text after paragraph 4
10435 A @dfn{thread} is a flow of control within the abstract machine.
10436 It is implementation defined whether or not there may be more than
10440 New text after paragraph 7
10443 It is unspecified whether additional action must be taken to
10444 ensure when and whether side effects are visible to other threads.
10450 Add @code{__thread}.
10453 @b{[basic.start.main]}
10455 Add after paragraph 5
10458 The thread that begins execution at the @code{main} function is called
10459 the @dfn{main thread}. It is implementation defined how functions
10460 beginning threads other than the main thread are designated or typed.
10461 A function so designated, as well as the @code{main} function, is called
10462 a @dfn{thread startup function}. It is implementation defined what
10463 happens if a thread startup function returns. It is implementation
10464 defined what happens to other threads when any thread calls @code{exit}.
10468 @b{[basic.start.init]}
10470 Add after paragraph 4
10473 The storage for an object of thread storage duration shall be
10474 statically initialized before the first statement of the thread startup
10475 function. An object of thread storage duration shall not require
10476 dynamic initialization.
10480 @b{[basic.start.term]}
10482 Add after paragraph 3
10485 The type of an object with thread storage duration shall not have a
10486 non-trivial destructor, nor shall it be an array type whose elements
10487 (directly or indirectly) have non-trivial destructors.
10493 Add ``thread storage duration'' to the list in paragraph 1.
10498 Thread, static, and automatic storage durations are associated with
10499 objects introduced by declarations [@dots{}].
10502 Add @code{__thread} to the list of specifiers in paragraph 3.
10505 @b{[basic.stc.thread]}
10507 New section before @b{[basic.stc.static]}
10510 The keyword @code{__thread} applied to a non-local object gives the
10511 object thread storage duration.
10513 A local variable or class data member declared both @code{static}
10514 and @code{__thread} gives the variable or member thread storage
10519 @b{[basic.stc.static]}
10524 All objects which have neither thread storage duration, dynamic
10525 storage duration nor are local [@dots{}].
10531 Add @code{__thread} to the list in paragraph 1.
10536 With the exception of @code{__thread}, at most one
10537 @var{storage-class-specifier} shall appear in a given
10538 @var{decl-specifier-seq}. The @code{__thread} specifier may
10539 be used alone, or immediately following the @code{extern} or
10540 @code{static} specifiers. [@dots{}]
10543 Add after paragraph 5
10546 The @code{__thread} specifier can be applied only to the names of objects
10547 and to anonymous unions.
10553 Add after paragraph 6
10556 Non-@code{static} members shall not be @code{__thread}.
10560 @c APPLE LOCAL begin blocks 7205047 5811887
10566 Blocks is a language feature that allows one to create anonymous
10567 functions. The feature is also known as lambdas or closures in other
10568 languages. The feature is controlled by @option{-fblocks}.
10569 See @uref{http://developer.apple.com/mac/library/documentation/Cocoa/Conceptual/Blocks/Articles/00_Introduction.html} for additional details.
10570 @c APPLE LOCAL end blocks 7205047 5811887
10572 @node Binary constants
10573 @section Binary constants using the @samp{0b} prefix
10574 @cindex Binary constants using the @samp{0b} prefix
10576 Integer constants can be written as binary constants, consisting of a
10577 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
10578 @samp{0B}. This is particularly useful in environments that operate a
10579 lot on the bit-level (like microcontrollers).
10581 The following statements are identical:
10590 The type of these constants follows the same rules as for octal or
10591 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
10594 @node C++ Extensions
10595 @chapter Extensions to the C++ Language
10596 @cindex extensions, C++ language
10597 @cindex C++ language extensions
10599 The GNU compiler provides these extensions to the C++ language (and you
10600 can also use most of the C language extensions in your C++ programs). If you
10601 want to write code that checks whether these features are available, you can
10602 test for the GNU compiler the same way as for C programs: check for a
10603 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10604 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10605 Predefined Macros,cpp,The GNU C Preprocessor}).
10608 * Volatiles:: What constitutes an access to a volatile object.
10609 * Restricted Pointers:: C99 restricted pointers and references.
10610 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10611 * C++ Interface:: You can use a single C++ header file for both
10612 declarations and definitions.
10613 * Template Instantiation:: Methods for ensuring that exactly one copy of
10614 each needed template instantiation is emitted.
10615 * Bound member functions:: You can extract a function pointer to the
10616 method denoted by a @samp{->*} or @samp{.*} expression.
10617 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10618 * Namespace Association:: Strong using-directives for namespace association.
10619 * Java Exceptions:: Tweaking exception handling to work with Java.
10620 * Deprecated Features:: Things will disappear from g++.
10621 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10625 @section When is a Volatile Object Accessed?
10626 @cindex accessing volatiles
10627 @cindex volatile read
10628 @cindex volatile write
10629 @cindex volatile access
10631 Both the C and C++ standard have the concept of volatile objects. These
10632 are normally accessed by pointers and used for accessing hardware. The
10633 standards encourage compilers to refrain from optimizations concerning
10634 accesses to volatile objects. The C standard leaves it implementation
10635 defined as to what constitutes a volatile access. The C++ standard omits
10636 to specify this, except to say that C++ should behave in a similar manner
10637 to C with respect to volatiles, where possible. The minimum either
10638 standard specifies is that at a sequence point all previous accesses to
10639 volatile objects have stabilized and no subsequent accesses have
10640 occurred. Thus an implementation is free to reorder and combine
10641 volatile accesses which occur between sequence points, but cannot do so
10642 for accesses across a sequence point. The use of volatiles does not
10643 allow you to violate the restriction on updating objects multiple times
10644 within a sequence point.
10646 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10648 The behavior differs slightly between C and C++ in the non-obvious cases:
10651 volatile int *src = @var{somevalue};
10655 With C, such expressions are rvalues, and GCC interprets this either as a
10656 read of the volatile object being pointed to or only as request to evaluate
10657 the side-effects. The C++ standard specifies that such expressions do not
10658 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10659 object may be incomplete. The C++ standard does not specify explicitly
10660 that it is this lvalue to rvalue conversion which may be responsible for
10661 causing an access. However, there is reason to believe that it is,
10662 because otherwise certain simple expressions become undefined. However,
10663 because it would surprise most programmers, G++ treats dereferencing a
10664 pointer to volatile object of complete type when the value is unused as
10665 GCC would do for an equivalent type in C. When the object has incomplete
10666 type, G++ issues a warning; if you wish to force an error, you must
10667 force a conversion to rvalue with, for instance, a static cast.
10669 When using a reference to volatile, G++ does not treat equivalent
10670 expressions as accesses to volatiles, but instead issues a warning that
10671 no volatile is accessed. The rationale for this is that otherwise it
10672 becomes difficult to determine where volatile access occur, and not
10673 possible to ignore the return value from functions returning volatile
10674 references. Again, if you wish to force a read, cast the reference to
10677 @node Restricted Pointers
10678 @section Restricting Pointer Aliasing
10679 @cindex restricted pointers
10680 @cindex restricted references
10681 @cindex restricted this pointer
10683 As with the C front end, G++ understands the C99 feature of restricted pointers,
10684 specified with the @code{__restrict__}, or @code{__restrict} type
10685 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10686 language flag, @code{restrict} is not a keyword in C++.
10688 In addition to allowing restricted pointers, you can specify restricted
10689 references, which indicate that the reference is not aliased in the local
10693 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10700 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10701 @var{rref} refers to a (different) unaliased integer.
10703 You may also specify whether a member function's @var{this} pointer is
10704 unaliased by using @code{__restrict__} as a member function qualifier.
10707 void T::fn () __restrict__
10714 Within the body of @code{T::fn}, @var{this} will have the effective
10715 definition @code{T *__restrict__ const this}. Notice that the
10716 interpretation of a @code{__restrict__} member function qualifier is
10717 different to that of @code{const} or @code{volatile} qualifier, in that it
10718 is applied to the pointer rather than the object. This is consistent with
10719 other compilers which implement restricted pointers.
10721 As with all outermost parameter qualifiers, @code{__restrict__} is
10722 ignored in function definition matching. This means you only need to
10723 specify @code{__restrict__} in a function definition, rather than
10724 in a function prototype as well.
10726 @node Vague Linkage
10727 @section Vague Linkage
10728 @cindex vague linkage
10730 There are several constructs in C++ which require space in the object
10731 file but are not clearly tied to a single translation unit. We say that
10732 these constructs have ``vague linkage''. Typically such constructs are
10733 emitted wherever they are needed, though sometimes we can be more
10737 @item Inline Functions
10738 Inline functions are typically defined in a header file which can be
10739 included in many different compilations. Hopefully they can usually be
10740 inlined, but sometimes an out-of-line copy is necessary, if the address
10741 of the function is taken or if inlining fails. In general, we emit an
10742 out-of-line copy in all translation units where one is needed. As an
10743 exception, we only emit inline virtual functions with the vtable, since
10744 it will always require a copy.
10746 Local static variables and string constants used in an inline function
10747 are also considered to have vague linkage, since they must be shared
10748 between all inlined and out-of-line instances of the function.
10752 C++ virtual functions are implemented in most compilers using a lookup
10753 table, known as a vtable. The vtable contains pointers to the virtual
10754 functions provided by a class, and each object of the class contains a
10755 pointer to its vtable (or vtables, in some multiple-inheritance
10756 situations). If the class declares any non-inline, non-pure virtual
10757 functions, the first one is chosen as the ``key method'' for the class,
10758 and the vtable is only emitted in the translation unit where the key
10761 @emph{Note:} If the chosen key method is later defined as inline, the
10762 vtable will still be emitted in every translation unit which defines it.
10763 Make sure that any inline virtuals are declared inline in the class
10764 body, even if they are not defined there.
10766 @item type_info objects
10769 C++ requires information about types to be written out in order to
10770 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10771 For polymorphic classes (classes with virtual functions), the type_info
10772 object is written out along with the vtable so that @samp{dynamic_cast}
10773 can determine the dynamic type of a class object at runtime. For all
10774 other types, we write out the type_info object when it is used: when
10775 applying @samp{typeid} to an expression, throwing an object, or
10776 referring to a type in a catch clause or exception specification.
10778 @item Template Instantiations
10779 Most everything in this section also applies to template instantiations,
10780 but there are other options as well.
10781 @xref{Template Instantiation,,Where's the Template?}.
10785 When used with GNU ld version 2.8 or later on an ELF system such as
10786 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10787 these constructs will be discarded at link time. This is known as
10790 On targets that don't support COMDAT, but do support weak symbols, GCC
10791 will use them. This way one copy will override all the others, but
10792 the unused copies will still take up space in the executable.
10794 For targets which do not support either COMDAT or weak symbols,
10795 most entities with vague linkage will be emitted as local symbols to
10796 avoid duplicate definition errors from the linker. This will not happen
10797 for local statics in inlines, however, as having multiple copies will
10798 almost certainly break things.
10800 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10801 another way to control placement of these constructs.
10803 @node C++ Interface
10804 @section #pragma interface and implementation
10806 @cindex interface and implementation headers, C++
10807 @cindex C++ interface and implementation headers
10808 @cindex pragmas, interface and implementation
10810 @code{#pragma interface} and @code{#pragma implementation} provide the
10811 user with a way of explicitly directing the compiler to emit entities
10812 with vague linkage (and debugging information) in a particular
10815 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10816 most cases, because of COMDAT support and the ``key method'' heuristic
10817 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10818 program to grow due to unnecessary out-of-line copies of inline
10819 functions. Currently (3.4) the only benefit of these
10820 @code{#pragma}s is reduced duplication of debugging information, and
10821 that should be addressed soon on DWARF 2 targets with the use of
10825 @item #pragma interface
10826 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10827 @kindex #pragma interface
10828 Use this directive in @emph{header files} that define object classes, to save
10829 space in most of the object files that use those classes. Normally,
10830 local copies of certain information (backup copies of inline member
10831 functions, debugging information, and the internal tables that implement
10832 virtual functions) must be kept in each object file that includes class
10833 definitions. You can use this pragma to avoid such duplication. When a
10834 header file containing @samp{#pragma interface} is included in a
10835 compilation, this auxiliary information will not be generated (unless
10836 the main input source file itself uses @samp{#pragma implementation}).
10837 Instead, the object files will contain references to be resolved at link
10840 The second form of this directive is useful for the case where you have
10841 multiple headers with the same name in different directories. If you
10842 use this form, you must specify the same string to @samp{#pragma
10845 @item #pragma implementation
10846 @itemx #pragma implementation "@var{objects}.h"
10847 @kindex #pragma implementation
10848 Use this pragma in a @emph{main input file}, when you want full output from
10849 included header files to be generated (and made globally visible). The
10850 included header file, in turn, should use @samp{#pragma interface}.
10851 Backup copies of inline member functions, debugging information, and the
10852 internal tables used to implement virtual functions are all generated in
10853 implementation files.
10855 @cindex implied @code{#pragma implementation}
10856 @cindex @code{#pragma implementation}, implied
10857 @cindex naming convention, implementation headers
10858 If you use @samp{#pragma implementation} with no argument, it applies to
10859 an include file with the same basename@footnote{A file's @dfn{basename}
10860 was the name stripped of all leading path information and of trailing
10861 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10862 file. For example, in @file{allclass.cc}, giving just
10863 @samp{#pragma implementation}
10864 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10866 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10867 an implementation file whenever you would include it from
10868 @file{allclass.cc} even if you never specified @samp{#pragma
10869 implementation}. This was deemed to be more trouble than it was worth,
10870 however, and disabled.
10872 Use the string argument if you want a single implementation file to
10873 include code from multiple header files. (You must also use
10874 @samp{#include} to include the header file; @samp{#pragma
10875 implementation} only specifies how to use the file---it doesn't actually
10878 There is no way to split up the contents of a single header file into
10879 multiple implementation files.
10882 @cindex inlining and C++ pragmas
10883 @cindex C++ pragmas, effect on inlining
10884 @cindex pragmas in C++, effect on inlining
10885 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10886 effect on function inlining.
10888 If you define a class in a header file marked with @samp{#pragma
10889 interface}, the effect on an inline function defined in that class is
10890 similar to an explicit @code{extern} declaration---the compiler emits
10891 no code at all to define an independent version of the function. Its
10892 definition is used only for inlining with its callers.
10894 @opindex fno-implement-inlines
10895 Conversely, when you include the same header file in a main source file
10896 that declares it as @samp{#pragma implementation}, the compiler emits
10897 code for the function itself; this defines a version of the function
10898 that can be found via pointers (or by callers compiled without
10899 inlining). If all calls to the function can be inlined, you can avoid
10900 emitting the function by compiling with @option{-fno-implement-inlines}.
10901 If any calls were not inlined, you will get linker errors.
10903 @node Template Instantiation
10904 @section Where's the Template?
10905 @cindex template instantiation
10907 C++ templates are the first language feature to require more
10908 intelligence from the environment than one usually finds on a UNIX
10909 system. Somehow the compiler and linker have to make sure that each
10910 template instance occurs exactly once in the executable if it is needed,
10911 and not at all otherwise. There are two basic approaches to this
10912 problem, which are referred to as the Borland model and the Cfront model.
10915 @item Borland model
10916 Borland C++ solved the template instantiation problem by adding the code
10917 equivalent of common blocks to their linker; the compiler emits template
10918 instances in each translation unit that uses them, and the linker
10919 collapses them together. The advantage of this model is that the linker
10920 only has to consider the object files themselves; there is no external
10921 complexity to worry about. This disadvantage is that compilation time
10922 is increased because the template code is being compiled repeatedly.
10923 Code written for this model tends to include definitions of all
10924 templates in the header file, since they must be seen to be
10928 The AT&T C++ translator, Cfront, solved the template instantiation
10929 problem by creating the notion of a template repository, an
10930 automatically maintained place where template instances are stored. A
10931 more modern version of the repository works as follows: As individual
10932 object files are built, the compiler places any template definitions and
10933 instantiations encountered in the repository. At link time, the link
10934 wrapper adds in the objects in the repository and compiles any needed
10935 instances that were not previously emitted. The advantages of this
10936 model are more optimal compilation speed and the ability to use the
10937 system linker; to implement the Borland model a compiler vendor also
10938 needs to replace the linker. The disadvantages are vastly increased
10939 complexity, and thus potential for error; for some code this can be
10940 just as transparent, but in practice it can been very difficult to build
10941 multiple programs in one directory and one program in multiple
10942 directories. Code written for this model tends to separate definitions
10943 of non-inline member templates into a separate file, which should be
10944 compiled separately.
10947 When used with GNU ld version 2.8 or later on an ELF system such as
10948 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10949 Borland model. On other systems, G++ implements neither automatic
10952 A future version of G++ will support a hybrid model whereby the compiler
10953 will emit any instantiations for which the template definition is
10954 included in the compile, and store template definitions and
10955 instantiation context information into the object file for the rest.
10956 The link wrapper will extract that information as necessary and invoke
10957 the compiler to produce the remaining instantiations. The linker will
10958 then combine duplicate instantiations.
10960 In the mean time, you have the following options for dealing with
10961 template instantiations:
10966 Compile your template-using code with @option{-frepo}. The compiler will
10967 generate files with the extension @samp{.rpo} listing all of the
10968 template instantiations used in the corresponding object files which
10969 could be instantiated there; the link wrapper, @samp{collect2}, will
10970 then update the @samp{.rpo} files to tell the compiler where to place
10971 those instantiations and rebuild any affected object files. The
10972 link-time overhead is negligible after the first pass, as the compiler
10973 will continue to place the instantiations in the same files.
10975 This is your best option for application code written for the Borland
10976 model, as it will just work. Code written for the Cfront model will
10977 need to be modified so that the template definitions are available at
10978 one or more points of instantiation; usually this is as simple as adding
10979 @code{#include <tmethods.cc>} to the end of each template header.
10981 For library code, if you want the library to provide all of the template
10982 instantiations it needs, just try to link all of its object files
10983 together; the link will fail, but cause the instantiations to be
10984 generated as a side effect. Be warned, however, that this may cause
10985 conflicts if multiple libraries try to provide the same instantiations.
10986 For greater control, use explicit instantiation as described in the next
10990 @opindex fno-implicit-templates
10991 Compile your code with @option{-fno-implicit-templates} to disable the
10992 implicit generation of template instances, and explicitly instantiate
10993 all the ones you use. This approach requires more knowledge of exactly
10994 which instances you need than do the others, but it's less
10995 mysterious and allows greater control. You can scatter the explicit
10996 instantiations throughout your program, perhaps putting them in the
10997 translation units where the instances are used or the translation units
10998 that define the templates themselves; you can put all of the explicit
10999 instantiations you need into one big file; or you can create small files
11006 template class Foo<int>;
11007 template ostream& operator <<
11008 (ostream&, const Foo<int>&);
11011 for each of the instances you need, and create a template instantiation
11012 library from those.
11014 If you are using Cfront-model code, you can probably get away with not
11015 using @option{-fno-implicit-templates} when compiling files that don't
11016 @samp{#include} the member template definitions.
11018 If you use one big file to do the instantiations, you may want to
11019 compile it without @option{-fno-implicit-templates} so you get all of the
11020 instances required by your explicit instantiations (but not by any
11021 other files) without having to specify them as well.
11023 G++ has extended the template instantiation syntax given in the ISO
11024 standard to allow forward declaration of explicit instantiations
11025 (with @code{extern}), instantiation of the compiler support data for a
11026 template class (i.e.@: the vtable) without instantiating any of its
11027 members (with @code{inline}), and instantiation of only the static data
11028 members of a template class, without the support data or member
11029 functions (with (@code{static}):
11032 extern template int max (int, int);
11033 inline template class Foo<int>;
11034 static template class Foo<int>;
11038 Do nothing. Pretend G++ does implement automatic instantiation
11039 management. Code written for the Borland model will work fine, but
11040 each translation unit will contain instances of each of the templates it
11041 uses. In a large program, this can lead to an unacceptable amount of code
11045 @node Bound member functions
11046 @section Extracting the function pointer from a bound pointer to member function
11048 @cindex pointer to member function
11049 @cindex bound pointer to member function
11051 In C++, pointer to member functions (PMFs) are implemented using a wide
11052 pointer of sorts to handle all the possible call mechanisms; the PMF
11053 needs to store information about how to adjust the @samp{this} pointer,
11054 and if the function pointed to is virtual, where to find the vtable, and
11055 where in the vtable to look for the member function. If you are using
11056 PMFs in an inner loop, you should really reconsider that decision. If
11057 that is not an option, you can extract the pointer to the function that
11058 would be called for a given object/PMF pair and call it directly inside
11059 the inner loop, to save a bit of time.
11061 Note that you will still be paying the penalty for the call through a
11062 function pointer; on most modern architectures, such a call defeats the
11063 branch prediction features of the CPU@. This is also true of normal
11064 virtual function calls.
11066 The syntax for this extension is
11070 extern int (A::*fp)();
11071 typedef int (*fptr)(A *);
11073 fptr p = (fptr)(a.*fp);
11076 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
11077 no object is needed to obtain the address of the function. They can be
11078 converted to function pointers directly:
11081 fptr p1 = (fptr)(&A::foo);
11084 @opindex Wno-pmf-conversions
11085 You must specify @option{-Wno-pmf-conversions} to use this extension.
11087 @node C++ Attributes
11088 @section C++-Specific Variable, Function, and Type Attributes
11090 Some attributes only make sense for C++ programs.
11093 @item init_priority (@var{priority})
11094 @cindex init_priority attribute
11097 In Standard C++, objects defined at namespace scope are guaranteed to be
11098 initialized in an order in strict accordance with that of their definitions
11099 @emph{in a given translation unit}. No guarantee is made for initializations
11100 across translation units. However, GNU C++ allows users to control the
11101 order of initialization of objects defined at namespace scope with the
11102 @code{init_priority} attribute by specifying a relative @var{priority},
11103 a constant integral expression currently bounded between 101 and 65535
11104 inclusive. Lower numbers indicate a higher priority.
11106 In the following example, @code{A} would normally be created before
11107 @code{B}, but the @code{init_priority} attribute has reversed that order:
11110 Some_Class A __attribute__ ((init_priority (2000)));
11111 Some_Class B __attribute__ ((init_priority (543)));
11115 Note that the particular values of @var{priority} do not matter; only their
11118 @item java_interface
11119 @cindex java_interface attribute
11121 This type attribute informs C++ that the class is a Java interface. It may
11122 only be applied to classes declared within an @code{extern "Java"} block.
11123 Calls to methods declared in this interface will be dispatched using GCJ's
11124 interface table mechanism, instead of regular virtual table dispatch.
11128 See also @xref{Namespace Association}.
11130 @node Namespace Association
11131 @section Namespace Association
11133 @strong{Caution:} The semantics of this extension are not fully
11134 defined. Users should refrain from using this extension as its
11135 semantics may change subtly over time. It is possible that this
11136 extension will be removed in future versions of G++.
11138 A using-directive with @code{__attribute ((strong))} is stronger
11139 than a normal using-directive in two ways:
11143 Templates from the used namespace can be specialized and explicitly
11144 instantiated as though they were members of the using namespace.
11147 The using namespace is considered an associated namespace of all
11148 templates in the used namespace for purposes of argument-dependent
11152 The used namespace must be nested within the using namespace so that
11153 normal unqualified lookup works properly.
11155 This is useful for composing a namespace transparently from
11156 implementation namespaces. For example:
11161 template <class T> struct A @{ @};
11163 using namespace debug __attribute ((__strong__));
11164 template <> struct A<int> @{ @}; // @r{ok to specialize}
11166 template <class T> void f (A<T>);
11171 f (std::A<float>()); // @r{lookup finds} std::f
11176 @node Java Exceptions
11177 @section Java Exceptions
11179 The Java language uses a slightly different exception handling model
11180 from C++. Normally, GNU C++ will automatically detect when you are
11181 writing C++ code that uses Java exceptions, and handle them
11182 appropriately. However, if C++ code only needs to execute destructors
11183 when Java exceptions are thrown through it, GCC will guess incorrectly.
11184 Sample problematic code is:
11187 struct S @{ ~S(); @};
11188 extern void bar(); // @r{is written in Java, and may throw exceptions}
11197 The usual effect of an incorrect guess is a link failure, complaining of
11198 a missing routine called @samp{__gxx_personality_v0}.
11200 You can inform the compiler that Java exceptions are to be used in a
11201 translation unit, irrespective of what it might think, by writing
11202 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11203 @samp{#pragma} must appear before any functions that throw or catch
11204 exceptions, or run destructors when exceptions are thrown through them.
11206 You cannot mix Java and C++ exceptions in the same translation unit. It
11207 is believed to be safe to throw a C++ exception from one file through
11208 another file compiled for the Java exception model, or vice versa, but
11209 there may be bugs in this area.
11211 @node Deprecated Features
11212 @section Deprecated Features
11214 In the past, the GNU C++ compiler was extended to experiment with new
11215 features, at a time when the C++ language was still evolving. Now that
11216 the C++ standard is complete, some of those features are superseded by
11217 superior alternatives. Using the old features might cause a warning in
11218 some cases that the feature will be dropped in the future. In other
11219 cases, the feature might be gone already.
11221 While the list below is not exhaustive, it documents some of the options
11222 that are now deprecated:
11225 @item -fexternal-templates
11226 @itemx -falt-external-templates
11227 These are two of the many ways for G++ to implement template
11228 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11229 defines how template definitions have to be organized across
11230 implementation units. G++ has an implicit instantiation mechanism that
11231 should work just fine for standard-conforming code.
11233 @item -fstrict-prototype
11234 @itemx -fno-strict-prototype
11235 Previously it was possible to use an empty prototype parameter list to
11236 indicate an unspecified number of parameters (like C), rather than no
11237 parameters, as C++ demands. This feature has been removed, except where
11238 it is required for backwards compatibility @xref{Backwards Compatibility}.
11241 G++ allows a virtual function returning @samp{void *} to be overridden
11242 by one returning a different pointer type. This extension to the
11243 covariant return type rules is now deprecated and will be removed from a
11246 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11247 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11248 and will be removed in a future version. Code using these operators
11249 should be modified to use @code{std::min} and @code{std::max} instead.
11251 The named return value extension has been deprecated, and is now
11254 The use of initializer lists with new expressions has been deprecated,
11255 and is now removed from G++.
11257 Floating and complex non-type template parameters have been deprecated,
11258 and are now removed from G++.
11260 The implicit typename extension has been deprecated and is now
11263 The use of default arguments in function pointers, function typedefs
11264 and other places where they are not permitted by the standard is
11265 deprecated and will be removed from a future version of G++.
11267 G++ allows floating-point literals to appear in integral constant expressions,
11268 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11269 This extension is deprecated and will be removed from a future version.
11271 G++ allows static data members of const floating-point type to be declared
11272 with an initializer in a class definition. The standard only allows
11273 initializers for static members of const integral types and const
11274 enumeration types so this extension has been deprecated and will be removed
11275 from a future version.
11277 @node Backwards Compatibility
11278 @section Backwards Compatibility
11279 @cindex Backwards Compatibility
11280 @cindex ARM [Annotated C++ Reference Manual]
11282 Now that there is a definitive ISO standard C++, G++ has a specification
11283 to adhere to. The C++ language evolved over time, and features that
11284 used to be acceptable in previous drafts of the standard, such as the ARM
11285 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11286 compilation of C++ written to such drafts, G++ contains some backwards
11287 compatibilities. @emph{All such backwards compatibility features are
11288 liable to disappear in future versions of G++.} They should be considered
11289 deprecated @xref{Deprecated Features}.
11293 If a variable is declared at for scope, it used to remain in scope until
11294 the end of the scope which contained the for statement (rather than just
11295 within the for scope). G++ retains this, but issues a warning, if such a
11296 variable is accessed outside the for scope.
11298 @item Implicit C language
11299 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11300 scope to set the language. On such systems, all header files are
11301 implicitly scoped inside a C language scope. Also, an empty prototype
11302 @code{()} will be treated as an unspecified number of arguments, rather
11303 than no arguments, as C++ demands.