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 and Objective-C@. Most of them are
20 also available in C++. @xref{C++ Extensions,,Extensions to the
21 C++ Language}, for extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * 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 * Alignment:: Inquiring about the alignment of a type or variable.
62 * Inline:: Defining inline functions (as fast as macros).
63 * Extended Asm:: Assembler instructions with C expressions as operands.
64 (With them you can define ``built-in'' functions.)
65 * Constraints:: Constraints for asm operands
66 * Asm Labels:: Specifying the assembler name to use for a C symbol.
67 * Explicit Reg Vars:: Defining variables residing in specified registers.
68 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
69 * Incomplete Enums:: @code{enum foo;}, with details to follow.
70 * Function Names:: Printable strings which are the name of the current
72 * Return Address:: Getting the return or frame address of a function.
73 * Vector Extensions:: Using vector instructions through built-in functions.
74 * Offsetof:: Special syntax for implementing @code{offsetof}.
75 * Atomic Builtins:: Built-in functions for atomic memory access.
76 * Object Size Checking:: Built-in functions for limited buffer overflow
78 * Other Builtins:: Other built-in functions.
79 * Target Builtins:: Built-in functions specific to particular targets.
80 * Target Format Checks:: Format checks specific to particular targets.
81 * Pragmas:: Pragmas accepted by GCC.
82 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
83 * Thread-Local:: Per-thread variables.
87 @section Statements and Declarations in Expressions
88 @cindex statements inside expressions
89 @cindex declarations inside expressions
90 @cindex expressions containing statements
91 @cindex macros, statements in expressions
93 @c the above section title wrapped and causes an underfull hbox.. i
94 @c changed it from "within" to "in". --mew 4feb93
95 A compound statement enclosed in parentheses may appear as an expression
96 in GNU C@. This allows you to use loops, switches, and local variables
99 Recall that a compound statement is a sequence of statements surrounded
100 by braces; in this construct, parentheses go around the braces. For
104 (@{ int y = foo (); int z;
111 is a valid (though slightly more complex than necessary) expression
112 for the absolute value of @code{foo ()}.
114 The last thing in the compound statement should be an expression
115 followed by a semicolon; the value of this subexpression serves as the
116 value of the entire construct. (If you use some other kind of statement
117 last within the braces, the construct has type @code{void}, and thus
118 effectively no value.)
120 This feature is especially useful in making macro definitions ``safe'' (so
121 that they evaluate each operand exactly once). For example, the
122 ``maximum'' function is commonly defined as a macro in standard C as
126 #define max(a,b) ((a) > (b) ? (a) : (b))
130 @cindex side effects, macro argument
131 But this definition computes either @var{a} or @var{b} twice, with bad
132 results if the operand has side effects. In GNU C, if you know the
133 type of the operands (here taken as @code{int}), you can define
134 the macro safely as follows:
137 #define maxint(a,b) \
138 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
141 Embedded statements are not allowed in constant expressions, such as
142 the value of an enumeration constant, the width of a bit-field, or
143 the initial value of a static variable.
145 If you don't know the type of the operand, you can still do this, but you
146 must use @code{typeof} (@pxref{Typeof}).
148 In G++, the result value of a statement expression undergoes array and
149 function pointer decay, and is returned by value to the enclosing
150 expression. For instance, if @code{A} is a class, then
159 will construct a temporary @code{A} object to hold the result of the
160 statement expression, and that will be used to invoke @code{Foo}.
161 Therefore the @code{this} pointer observed by @code{Foo} will not be the
164 Any temporaries created within a statement within a statement expression
165 will be destroyed at the statement's end. This makes statement
166 expressions inside macros slightly different from function calls. In
167 the latter case temporaries introduced during argument evaluation will
168 be destroyed at the end of the statement that includes the function
169 call. In the statement expression case they will be destroyed during
170 the statement expression. For instance,
173 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
174 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
184 will have different places where temporaries are destroyed. For the
185 @code{macro} case, the temporary @code{X} will be destroyed just after
186 the initialization of @code{b}. In the @code{function} case that
187 temporary will be destroyed when the function returns.
189 These considerations mean that it is probably a bad idea to use
190 statement-expressions of this form in header files that are designed to
191 work with C++. (Note that some versions of the GNU C Library contained
192 header files using statement-expression that lead to precisely this
195 Jumping into a statement expression with @code{goto} or using a
196 @code{switch} statement outside the statement expression with a
197 @code{case} or @code{default} label inside the statement expression is
198 not permitted. Jumping into a statement expression with a computed
199 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
200 Jumping out of a statement expression is permitted, but if the
201 statement expression is part of a larger expression then it is
202 unspecified which other subexpressions of that expression have been
203 evaluated except where the language definition requires certain
204 subexpressions to be evaluated before or after the statement
205 expression. In any case, as with a function call the evaluation of a
206 statement expression is not interleaved with the evaluation of other
207 parts of the containing expression. For example,
210 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
214 will call @code{foo} and @code{bar1} and will not call @code{baz} but
215 may or may not call @code{bar2}. If @code{bar2} is called, it will be
216 called after @code{foo} and before @code{bar1}
219 @section Locally Declared Labels
221 @cindex macros, local labels
223 GCC allows you to declare @dfn{local labels} in any nested block
224 scope. A local label is just like an ordinary label, but you can
225 only reference it (with a @code{goto} statement, or by taking its
226 address) within the block in which it was declared.
228 A local label declaration looks like this:
231 __label__ @var{label};
238 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
241 Local label declarations must come at the beginning of the block,
242 before any ordinary declarations or statements.
244 The label declaration defines the label @emph{name}, but does not define
245 the label itself. You must do this in the usual way, with
246 @code{@var{label}:}, within the statements of the statement expression.
248 The local label feature is useful for complex macros. If a macro
249 contains nested loops, a @code{goto} can be useful for breaking out of
250 them. However, an ordinary label whose scope is the whole function
251 cannot be used: if the macro can be expanded several times in one
252 function, the label will be multiply defined in that function. A
253 local label avoids this problem. For example:
256 #define SEARCH(value, array, target) \
259 typeof (target) _SEARCH_target = (target); \
260 typeof (*(array)) *_SEARCH_array = (array); \
263 for (i = 0; i < max; i++) \
264 for (j = 0; j < max; j++) \
265 if (_SEARCH_array[i][j] == _SEARCH_target) \
266 @{ (value) = i; goto found; @} \
272 This could also be written using a statement-expression:
275 #define SEARCH(array, target) \
278 typeof (target) _SEARCH_target = (target); \
279 typeof (*(array)) *_SEARCH_array = (array); \
282 for (i = 0; i < max; i++) \
283 for (j = 0; j < max; j++) \
284 if (_SEARCH_array[i][j] == _SEARCH_target) \
285 @{ value = i; goto found; @} \
292 Local label declarations also make the labels they declare visible to
293 nested functions, if there are any. @xref{Nested Functions}, for details.
295 @node Labels as Values
296 @section Labels as Values
297 @cindex labels as values
298 @cindex computed gotos
299 @cindex goto with computed label
300 @cindex address of a label
302 You can get the address of a label defined in the current function
303 (or a containing function) with the unary operator @samp{&&}. The
304 value has type @code{void *}. This value is a constant and can be used
305 wherever a constant of that type is valid. For example:
313 To use these values, you need to be able to jump to one. This is done
314 with the computed goto statement@footnote{The analogous feature in
315 Fortran is called an assigned goto, but that name seems inappropriate in
316 C, where one can do more than simply store label addresses in label
317 variables.}, @code{goto *@var{exp};}. For example,
324 Any expression of type @code{void *} is allowed.
326 One way of using these constants is in initializing a static array that
327 will serve as a jump table:
330 static void *array[] = @{ &&foo, &&bar, &&hack @};
333 Then you can select a label with indexing, like this:
340 Note that this does not check whether the subscript is in bounds---array
341 indexing in C never does that.
343 Such an array of label values serves a purpose much like that of the
344 @code{switch} statement. The @code{switch} statement is cleaner, so
345 use that rather than an array unless the problem does not fit a
346 @code{switch} statement very well.
348 Another use of label values is in an interpreter for threaded code.
349 The labels within the interpreter function can be stored in the
350 threaded code for super-fast dispatching.
352 You may not use this mechanism to jump to code in a different function.
353 If you do that, totally unpredictable things will happen. The best way to
354 avoid this is to store the label address only in automatic variables and
355 never pass it as an argument.
357 An alternate way to write the above example is
360 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
362 goto *(&&foo + array[i]);
366 This is more friendly to code living in shared libraries, as it reduces
367 the number of dynamic relocations that are needed, and by consequence,
368 allows the data to be read-only.
370 @node Nested Functions
371 @section Nested Functions
372 @cindex nested functions
373 @cindex downward funargs
376 A @dfn{nested function} is a function defined inside another function.
377 (Nested functions are not supported for GNU C++.) The nested function's
378 name is local to the block where it is defined. For example, here we
379 define a nested function named @code{square}, and call it twice:
383 foo (double a, double b)
385 double square (double z) @{ return z * z; @}
387 return square (a) + square (b);
392 The nested function can access all the variables of the containing
393 function that are visible at the point of its definition. This is
394 called @dfn{lexical scoping}. For example, here we show a nested
395 function which uses an inherited variable named @code{offset}:
399 bar (int *array, int offset, int size)
401 int access (int *array, int index)
402 @{ return array[index + offset]; @}
405 for (i = 0; i < size; i++)
406 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
411 Nested function definitions are permitted within functions in the places
412 where variable definitions are allowed; that is, in any block, mixed
413 with the other declarations and statements in the block.
415 It is possible to call the nested function from outside the scope of its
416 name by storing its address or passing the address to another function:
419 hack (int *array, int size)
421 void store (int index, int value)
422 @{ array[index] = value; @}
424 intermediate (store, size);
428 Here, the function @code{intermediate} receives the address of
429 @code{store} as an argument. If @code{intermediate} calls @code{store},
430 the arguments given to @code{store} are used to store into @code{array}.
431 But this technique works only so long as the containing function
432 (@code{hack}, in this example) does not exit.
434 If you try to call the nested function through its address after the
435 containing function has exited, all hell will break loose. If you try
436 to call it after a containing scope level has exited, and if it refers
437 to some of the variables that are no longer in scope, you may be lucky,
438 but it's not wise to take the risk. If, however, the nested function
439 does not refer to anything that has gone out of scope, you should be
442 GCC implements taking the address of a nested function using a technique
443 called @dfn{trampolines}. A paper describing them is available as
446 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
448 A nested function can jump to a label inherited from a containing
449 function, provided the label was explicitly declared in the containing
450 function (@pxref{Local Labels}). Such a jump returns instantly to the
451 containing function, exiting the nested function which did the
452 @code{goto} and any intermediate functions as well. Here is an example:
456 bar (int *array, int offset, int size)
459 int access (int *array, int index)
463 return array[index + offset];
467 for (i = 0; i < size; i++)
468 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
472 /* @r{Control comes here from @code{access}
473 if it detects an error.} */
480 A nested function always has no linkage. Declaring one with
481 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
482 before its definition, use @code{auto} (which is otherwise meaningless
483 for function declarations).
486 bar (int *array, int offset, int size)
489 auto int access (int *, int);
491 int access (int *array, int index)
495 return array[index + offset];
501 @node Constructing Calls
502 @section Constructing Function Calls
503 @cindex constructing calls
504 @cindex forwarding calls
506 Using the built-in functions described below, you can record
507 the arguments a function received, and call another function
508 with the same arguments, without knowing the number or types
511 You can also record the return value of that function call,
512 and later return that value, without knowing what data type
513 the function tried to return (as long as your caller expects
516 However, these built-in functions may interact badly with some
517 sophisticated features or other extensions of the language. It
518 is, therefore, not recommended to use them outside very simple
519 functions acting as mere forwarders for their arguments.
521 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
522 This built-in function returns a pointer to data
523 describing how to perform a call with the same arguments as were passed
524 to the current function.
526 The function saves the arg pointer register, structure value address,
527 and all registers that might be used to pass arguments to a function
528 into a block of memory allocated on the stack. Then it returns the
529 address of that block.
532 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
533 This built-in function invokes @var{function}
534 with a copy of the parameters described by @var{arguments}
537 The value of @var{arguments} should be the value returned by
538 @code{__builtin_apply_args}. The argument @var{size} specifies the size
539 of the stack argument data, in bytes.
541 This function returns a pointer to data describing
542 how to return whatever value was returned by @var{function}. The data
543 is saved in a block of memory allocated on the stack.
545 It is not always simple to compute the proper value for @var{size}. The
546 value is used by @code{__builtin_apply} to compute the amount of data
547 that should be pushed on the stack and copied from the incoming argument
551 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
552 This built-in function returns the value described by @var{result} from
553 the containing function. You should specify, for @var{result}, a value
554 returned by @code{__builtin_apply}.
558 @section Referring to a Type with @code{typeof}
561 @cindex macros, types of arguments
563 Another way to refer to the type of an expression is with @code{typeof}.
564 The syntax of using of this keyword looks like @code{sizeof}, but the
565 construct acts semantically like a type name defined with @code{typedef}.
567 There are two ways of writing the argument to @code{typeof}: with an
568 expression or with a type. Here is an example with an expression:
575 This assumes that @code{x} is an array of pointers to functions;
576 the type described is that of the values of the functions.
578 Here is an example with a typename as the argument:
585 Here the type described is that of pointers to @code{int}.
587 If you are writing a header file that must work when included in ISO C
588 programs, write @code{__typeof__} instead of @code{typeof}.
589 @xref{Alternate Keywords}.
591 A @code{typeof}-construct can be used anywhere a typedef name could be
592 used. For example, you can use it in a declaration, in a cast, or inside
593 of @code{sizeof} or @code{typeof}.
595 @code{typeof} is often useful in conjunction with the
596 statements-within-expressions feature. Here is how the two together can
597 be used to define a safe ``maximum'' macro that operates on any
598 arithmetic type and evaluates each of its arguments exactly once:
602 (@{ typeof (a) _a = (a); \
603 typeof (b) _b = (b); \
604 _a > _b ? _a : _b; @})
607 @cindex underscores in variables in macros
608 @cindex @samp{_} in variables in macros
609 @cindex local variables in macros
610 @cindex variables, local, in macros
611 @cindex macros, local variables in
613 The reason for using names that start with underscores for the local
614 variables is to avoid conflicts with variable names that occur within the
615 expressions that are substituted for @code{a} and @code{b}. Eventually we
616 hope to design a new form of declaration syntax that allows you to declare
617 variables whose scopes start only after their initializers; this will be a
618 more reliable way to prevent such conflicts.
621 Some more examples of the use of @code{typeof}:
625 This declares @code{y} with the type of what @code{x} points to.
632 This declares @code{y} as an array of such values.
639 This declares @code{y} as an array of pointers to characters:
642 typeof (typeof (char *)[4]) y;
646 It is equivalent to the following traditional C declaration:
652 To see the meaning of the declaration using @code{typeof}, and why it
653 might be a useful way to write, rewrite it with these macros:
656 #define pointer(T) typeof(T *)
657 #define array(T, N) typeof(T [N])
661 Now the declaration can be rewritten this way:
664 array (pointer (char), 4) y;
668 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
669 pointers to @code{char}.
672 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
673 a more limited extension which permitted one to write
676 typedef @var{T} = @var{expr};
680 with the effect of declaring @var{T} to have the type of the expression
681 @var{expr}. This extension does not work with GCC 3 (versions between
682 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
683 relies on it should be rewritten to use @code{typeof}:
686 typedef typeof(@var{expr}) @var{T};
690 This will work with all versions of GCC@.
693 @section Conditionals with Omitted Operands
694 @cindex conditional expressions, extensions
695 @cindex omitted middle-operands
696 @cindex middle-operands, omitted
697 @cindex extensions, @code{?:}
698 @cindex @code{?:} extensions
700 The middle operand in a conditional expression may be omitted. Then
701 if the first operand is nonzero, its value is the value of the conditional
704 Therefore, the expression
711 has the value of @code{x} if that is nonzero; otherwise, the value of
714 This example is perfectly equivalent to
720 @cindex side effect in ?:
721 @cindex ?: side effect
723 In this simple case, the ability to omit the middle operand is not
724 especially useful. When it becomes useful is when the first operand does,
725 or may (if it is a macro argument), contain a side effect. Then repeating
726 the operand in the middle would perform the side effect twice. Omitting
727 the middle operand uses the value already computed without the undesirable
728 effects of recomputing it.
731 @section Double-Word Integers
732 @cindex @code{long long} data types
733 @cindex double-word arithmetic
734 @cindex multiprecision arithmetic
735 @cindex @code{LL} integer suffix
736 @cindex @code{ULL} integer suffix
738 ISO C99 supports data types for integers that are at least 64 bits wide,
739 and as an extension GCC supports them in C89 mode and in C++.
740 Simply write @code{long long int} for a signed integer, or
741 @code{unsigned long long int} for an unsigned integer. To make an
742 integer constant of type @code{long long int}, add the suffix @samp{LL}
743 to the integer. To make an integer constant of type @code{unsigned long
744 long int}, add the suffix @samp{ULL} to the integer.
746 You can use these types in arithmetic like any other integer types.
747 Addition, subtraction, and bitwise boolean operations on these types
748 are open-coded on all types of machines. Multiplication is open-coded
749 if the machine supports fullword-to-doubleword a widening multiply
750 instruction. Division and shifts are open-coded only on machines that
751 provide special support. The operations that are not open-coded use
752 special library routines that come with GCC@.
754 There may be pitfalls when you use @code{long long} types for function
755 arguments, unless you declare function prototypes. If a function
756 expects type @code{int} for its argument, and you pass a value of type
757 @code{long long int}, confusion will result because the caller and the
758 subroutine will disagree about the number of bytes for the argument.
759 Likewise, if the function expects @code{long long int} and you pass
760 @code{int}. The best way to avoid such problems is to use prototypes.
763 @section Complex Numbers
764 @cindex complex numbers
765 @cindex @code{_Complex} keyword
766 @cindex @code{__complex__} keyword
768 ISO C99 supports complex floating data types, and as an extension GCC
769 supports them in C89 mode and in C++, and supports complex integer data
770 types which are not part of ISO C99. You can declare complex types
771 using the keyword @code{_Complex}. As an extension, the older GNU
772 keyword @code{__complex__} is also supported.
774 For example, @samp{_Complex double x;} declares @code{x} as a
775 variable whose real part and imaginary part are both of type
776 @code{double}. @samp{_Complex short int y;} declares @code{y} to
777 have real and imaginary parts of type @code{short int}; this is not
778 likely to be useful, but it shows that the set of complex types is
781 To write a constant with a complex data type, use the suffix @samp{i} or
782 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
783 has type @code{_Complex float} and @code{3i} has type
784 @code{_Complex int}. Such a constant always has a pure imaginary
785 value, but you can form any complex value you like by adding one to a
786 real constant. This is a GNU extension; if you have an ISO C99
787 conforming C library (such as GNU libc), and want to construct complex
788 constants of floating type, you should include @code{<complex.h>} and
789 use the macros @code{I} or @code{_Complex_I} instead.
791 @cindex @code{__real__} keyword
792 @cindex @code{__imag__} keyword
793 To extract the real part of a complex-valued expression @var{exp}, write
794 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
795 extract the imaginary part. This is a GNU extension; for values of
796 floating type, you should use the ISO C99 functions @code{crealf},
797 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
798 @code{cimagl}, declared in @code{<complex.h>} and also provided as
799 built-in functions by GCC@.
801 @cindex complex conjugation
802 The operator @samp{~} performs complex conjugation when used on a value
803 with a complex type. This is a GNU extension; for values of
804 floating type, you should use the ISO C99 functions @code{conjf},
805 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
806 provided as built-in functions by GCC@.
808 GCC can allocate complex automatic variables in a noncontiguous
809 fashion; it's even possible for the real part to be in a register while
810 the imaginary part is on the stack (or vice-versa). Only the DWARF2
811 debug info format can represent this, so use of DWARF2 is recommended.
812 If you are using the stabs debug info format, GCC describes a noncontiguous
813 complex variable as if it were two separate variables of noncomplex type.
814 If the variable's actual name is @code{foo}, the two fictitious
815 variables are named @code{foo$real} and @code{foo$imag}. You can
816 examine and set these two fictitious variables with your debugger.
819 @section Decimal Floating Types
820 @cindex decimal floating types
821 @cindex @code{_Decimal32} data type
822 @cindex @code{_Decimal64} data type
823 @cindex @code{_Decimal128} data type
824 @cindex @code{df} integer suffix
825 @cindex @code{dd} integer suffix
826 @cindex @code{dl} integer suffix
827 @cindex @code{DF} integer suffix
828 @cindex @code{DD} integer suffix
829 @cindex @code{DL} integer suffix
831 As an extension, the GNU C compiler supports decimal floating types as
832 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
833 floating types in GCC will evolve as the draft technical report changes.
834 Calling conventions for any target might also change. Not all targets
835 support decimal floating types.
837 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
838 @code{_Decimal128}. They use a radix of ten, unlike the floating types
839 @code{float}, @code{double}, and @code{long double} whose radix is not
840 specified by the C standard but is usually two.
842 Support for decimal floating types includes the arithmetic operators
843 add, subtract, multiply, divide; unary arithmetic operators;
844 relational operators; equality operators; and conversions to and from
845 integer and other floating types. Use a suffix @samp{df} or
846 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
847 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
850 GCC support of decimal float as specified by the draft technical report
855 Translation time data type (TTDT) is not supported.
858 Characteristics of decimal floating types are defined in header file
859 @file{decfloat.h} rather than @file{float.h}.
862 When the value of a decimal floating type cannot be represented in the
863 integer type to which it is being converted, the result is undefined
864 rather than the result value specified by the draft technical report.
867 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
868 are supported by the DWARF2 debug information format.
874 ISO C99 supports floating-point numbers written not only in the usual
875 decimal notation, such as @code{1.55e1}, but also numbers such as
876 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
877 supports this in C89 mode (except in some cases when strictly
878 conforming) and in C++. In that format the
879 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
880 mandatory. The exponent is a decimal number that indicates the power of
881 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
888 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
889 is the same as @code{1.55e1}.
891 Unlike for floating-point numbers in the decimal notation the exponent
892 is always required in the hexadecimal notation. Otherwise the compiler
893 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
894 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
895 extension for floating-point constants of type @code{float}.
898 @section Arrays of Length Zero
899 @cindex arrays of length zero
900 @cindex zero-length arrays
901 @cindex length-zero arrays
902 @cindex flexible array members
904 Zero-length arrays are allowed in GNU C@. They are very useful as the
905 last element of a structure which is really a header for a variable-length
914 struct line *thisline = (struct line *)
915 malloc (sizeof (struct line) + this_length);
916 thisline->length = this_length;
919 In ISO C90, you would have to give @code{contents} a length of 1, which
920 means either you waste space or complicate the argument to @code{malloc}.
922 In ISO C99, you would use a @dfn{flexible array member}, which is
923 slightly different in syntax and semantics:
927 Flexible array members are written as @code{contents[]} without
931 Flexible array members have incomplete type, and so the @code{sizeof}
932 operator may not be applied. As a quirk of the original implementation
933 of zero-length arrays, @code{sizeof} evaluates to zero.
936 Flexible array members may only appear as the last member of a
937 @code{struct} that is otherwise non-empty.
940 A structure containing a flexible array member, or a union containing
941 such a structure (possibly recursively), may not be a member of a
942 structure or an element of an array. (However, these uses are
943 permitted by GCC as extensions.)
946 GCC versions before 3.0 allowed zero-length arrays to be statically
947 initialized, as if they were flexible arrays. In addition to those
948 cases that were useful, it also allowed initializations in situations
949 that would corrupt later data. Non-empty initialization of zero-length
950 arrays is now treated like any case where there are more initializer
951 elements than the array holds, in that a suitable warning about "excess
952 elements in array" is given, and the excess elements (all of them, in
953 this case) are ignored.
955 Instead GCC allows static initialization of flexible array members.
956 This is equivalent to defining a new structure containing the original
957 structure followed by an array of sufficient size to contain the data.
958 I.e.@: in the following, @code{f1} is constructed as if it were declared
964 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
967 struct f1 f1; int data[3];
968 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
972 The convenience of this extension is that @code{f1} has the desired
973 type, eliminating the need to consistently refer to @code{f2.f1}.
975 This has symmetry with normal static arrays, in that an array of
976 unknown size is also written with @code{[]}.
978 Of course, this extension only makes sense if the extra data comes at
979 the end of a top-level object, as otherwise we would be overwriting
980 data at subsequent offsets. To avoid undue complication and confusion
981 with initialization of deeply nested arrays, we simply disallow any
982 non-empty initialization except when the structure is the top-level
986 struct foo @{ int x; int y[]; @};
987 struct bar @{ struct foo z; @};
989 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
990 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
991 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
992 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
995 @node Empty Structures
996 @section Structures With No Members
997 @cindex empty structures
998 @cindex zero-size structures
1000 GCC permits a C structure to have no members:
1007 The structure will have size zero. In C++, empty structures are part
1008 of the language. G++ treats empty structures as if they had a single
1009 member of type @code{char}.
1011 @node Variable Length
1012 @section Arrays of Variable Length
1013 @cindex variable-length arrays
1014 @cindex arrays of variable length
1017 Variable-length automatic arrays are allowed in ISO C99, and as an
1018 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1019 implementation of variable-length arrays does not yet conform in detail
1020 to the ISO C99 standard.) These arrays are
1021 declared like any other automatic arrays, but with a length that is not
1022 a constant expression. The storage is allocated at the point of
1023 declaration and deallocated when the brace-level is exited. For
1028 concat_fopen (char *s1, char *s2, char *mode)
1030 char str[strlen (s1) + strlen (s2) + 1];
1033 return fopen (str, mode);
1037 @cindex scope of a variable length array
1038 @cindex variable-length array scope
1039 @cindex deallocating variable length arrays
1040 Jumping or breaking out of the scope of the array name deallocates the
1041 storage. Jumping into the scope is not allowed; you get an error
1044 @cindex @code{alloca} vs variable-length arrays
1045 You can use the function @code{alloca} to get an effect much like
1046 variable-length arrays. The function @code{alloca} is available in
1047 many other C implementations (but not in all). On the other hand,
1048 variable-length arrays are more elegant.
1050 There are other differences between these two methods. Space allocated
1051 with @code{alloca} exists until the containing @emph{function} returns.
1052 The space for a variable-length array is deallocated as soon as the array
1053 name's scope ends. (If you use both variable-length arrays and
1054 @code{alloca} in the same function, deallocation of a variable-length array
1055 will also deallocate anything more recently allocated with @code{alloca}.)
1057 You can also use variable-length arrays as arguments to functions:
1061 tester (int len, char data[len][len])
1067 The length of an array is computed once when the storage is allocated
1068 and is remembered for the scope of the array in case you access it with
1071 If you want to pass the array first and the length afterward, you can
1072 use a forward declaration in the parameter list---another GNU extension.
1076 tester (int len; char data[len][len], int len)
1082 @cindex parameter forward declaration
1083 The @samp{int len} before the semicolon is a @dfn{parameter forward
1084 declaration}, and it serves the purpose of making the name @code{len}
1085 known when the declaration of @code{data} is parsed.
1087 You can write any number of such parameter forward declarations in the
1088 parameter list. They can be separated by commas or semicolons, but the
1089 last one must end with a semicolon, which is followed by the ``real''
1090 parameter declarations. Each forward declaration must match a ``real''
1091 declaration in parameter name and data type. ISO C99 does not support
1092 parameter forward declarations.
1094 @node Variadic Macros
1095 @section Macros with a Variable Number of Arguments.
1096 @cindex variable number of arguments
1097 @cindex macro with variable arguments
1098 @cindex rest argument (in macro)
1099 @cindex variadic macros
1101 In the ISO C standard of 1999, a macro can be declared to accept a
1102 variable number of arguments much as a function can. The syntax for
1103 defining the macro is similar to that of a function. Here is an
1107 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1110 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1111 such a macro, it represents the zero or more tokens until the closing
1112 parenthesis that ends the invocation, including any commas. This set of
1113 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1114 wherever it appears. See the CPP manual for more information.
1116 GCC has long supported variadic macros, and used a different syntax that
1117 allowed you to give a name to the variable arguments just like any other
1118 argument. Here is an example:
1121 #define debug(format, args...) fprintf (stderr, format, args)
1124 This is in all ways equivalent to the ISO C example above, but arguably
1125 more readable and descriptive.
1127 GNU CPP has two further variadic macro extensions, and permits them to
1128 be used with either of the above forms of macro definition.
1130 In standard C, you are not allowed to leave the variable argument out
1131 entirely; but you are allowed to pass an empty argument. For example,
1132 this invocation is invalid in ISO C, because there is no comma after
1139 GNU CPP permits you to completely omit the variable arguments in this
1140 way. In the above examples, the compiler would complain, though since
1141 the expansion of the macro still has the extra comma after the format
1144 To help solve this problem, CPP behaves specially for variable arguments
1145 used with the token paste operator, @samp{##}. If instead you write
1148 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1151 and if the variable arguments are omitted or empty, the @samp{##}
1152 operator causes the preprocessor to remove the comma before it. If you
1153 do provide some variable arguments in your macro invocation, GNU CPP
1154 does not complain about the paste operation and instead places the
1155 variable arguments after the comma. Just like any other pasted macro
1156 argument, these arguments are not macro expanded.
1158 @node Escaped Newlines
1159 @section Slightly Looser Rules for Escaped Newlines
1160 @cindex escaped newlines
1161 @cindex newlines (escaped)
1163 Recently, the preprocessor has relaxed its treatment of escaped
1164 newlines. Previously, the newline had to immediately follow a
1165 backslash. The current implementation allows whitespace in the form
1166 of spaces, horizontal and vertical tabs, and form feeds between the
1167 backslash and the subsequent newline. The preprocessor issues a
1168 warning, but treats it as a valid escaped newline and combines the two
1169 lines to form a single logical line. This works within comments and
1170 tokens, as well as between tokens. Comments are @emph{not} treated as
1171 whitespace for the purposes of this relaxation, since they have not
1172 yet been replaced with spaces.
1175 @section Non-Lvalue Arrays May Have Subscripts
1176 @cindex subscripting
1177 @cindex arrays, non-lvalue
1179 @cindex subscripting and function values
1180 In ISO C99, arrays that are not lvalues still decay to pointers, and
1181 may be subscripted, although they may not be modified or used after
1182 the next sequence point and the unary @samp{&} operator may not be
1183 applied to them. As an extension, GCC allows such arrays to be
1184 subscripted in C89 mode, though otherwise they do not decay to
1185 pointers outside C99 mode. For example,
1186 this is valid in GNU C though not valid in C89:
1190 struct foo @{int a[4];@};
1196 return f().a[index];
1202 @section Arithmetic on @code{void}- and Function-Pointers
1203 @cindex void pointers, arithmetic
1204 @cindex void, size of pointer to
1205 @cindex function pointers, arithmetic
1206 @cindex function, size of pointer to
1208 In GNU C, addition and subtraction operations are supported on pointers to
1209 @code{void} and on pointers to functions. This is done by treating the
1210 size of a @code{void} or of a function as 1.
1212 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1213 and on function types, and returns 1.
1215 @opindex Wpointer-arith
1216 The option @option{-Wpointer-arith} requests a warning if these extensions
1220 @section Non-Constant Initializers
1221 @cindex initializers, non-constant
1222 @cindex non-constant initializers
1224 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1225 automatic variable are not required to be constant expressions in GNU C@.
1226 Here is an example of an initializer with run-time varying elements:
1229 foo (float f, float g)
1231 float beat_freqs[2] = @{ f-g, f+g @};
1236 @node Compound Literals
1237 @section Compound Literals
1238 @cindex constructor expressions
1239 @cindex initializations in expressions
1240 @cindex structures, constructor expression
1241 @cindex expressions, constructor
1242 @cindex compound literals
1243 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1245 ISO C99 supports compound literals. A compound literal looks like
1246 a cast containing an initializer. Its value is an object of the
1247 type specified in the cast, containing the elements specified in
1248 the initializer; it is an lvalue. As an extension, GCC supports
1249 compound literals in C89 mode and in C++.
1251 Usually, the specified type is a structure. Assume that
1252 @code{struct foo} and @code{structure} are declared as shown:
1255 struct foo @{int a; char b[2];@} structure;
1259 Here is an example of constructing a @code{struct foo} with a compound literal:
1262 structure = ((struct foo) @{x + y, 'a', 0@});
1266 This is equivalent to writing the following:
1270 struct foo temp = @{x + y, 'a', 0@};
1275 You can also construct an array. If all the elements of the compound literal
1276 are (made up of) simple constant expressions, suitable for use in
1277 initializers of objects of static storage duration, then the compound
1278 literal can be coerced to a pointer to its first element and used in
1279 such an initializer, as shown here:
1282 char **foo = (char *[]) @{ "x", "y", "z" @};
1285 Compound literals for scalar types and union types are is
1286 also allowed, but then the compound literal is equivalent
1289 As a GNU extension, GCC allows initialization of objects with static storage
1290 duration by compound literals (which is not possible in ISO C99, because
1291 the initializer is not a constant).
1292 It is handled as if the object was initialized only with the bracket
1293 enclosed list if the types of the compound literal and the object match.
1294 The initializer list of the compound literal must be constant.
1295 If the object being initialized has array type of unknown size, the size is
1296 determined by compound literal size.
1299 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1300 static int y[] = (int []) @{1, 2, 3@};
1301 static int z[] = (int [3]) @{1@};
1305 The above lines are equivalent to the following:
1307 static struct foo x = @{1, 'a', 'b'@};
1308 static int y[] = @{1, 2, 3@};
1309 static int z[] = @{1, 0, 0@};
1312 @node Designated Inits
1313 @section Designated Initializers
1314 @cindex initializers with labeled elements
1315 @cindex labeled elements in initializers
1316 @cindex case labels in initializers
1317 @cindex designated initializers
1319 Standard C89 requires the elements of an initializer to appear in a fixed
1320 order, the same as the order of the elements in the array or structure
1323 In ISO C99 you can give the elements in any order, specifying the array
1324 indices or structure field names they apply to, and GNU C allows this as
1325 an extension in C89 mode as well. This extension is not
1326 implemented in GNU C++.
1328 To specify an array index, write
1329 @samp{[@var{index}] =} before the element value. For example,
1332 int a[6] = @{ [4] = 29, [2] = 15 @};
1339 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1343 The index values must be constant expressions, even if the array being
1344 initialized is automatic.
1346 An alternative syntax for this which has been obsolete since GCC 2.5 but
1347 GCC still accepts is to write @samp{[@var{index}]} before the element
1348 value, with no @samp{=}.
1350 To initialize a range of elements to the same value, write
1351 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1352 extension. For example,
1355 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1359 If the value in it has side-effects, the side-effects will happen only once,
1360 not for each initialized field by the range initializer.
1363 Note that the length of the array is the highest value specified
1366 In a structure initializer, specify the name of a field to initialize
1367 with @samp{.@var{fieldname} =} before the element value. For example,
1368 given the following structure,
1371 struct point @{ int x, y; @};
1375 the following initialization
1378 struct point p = @{ .y = yvalue, .x = xvalue @};
1385 struct point p = @{ xvalue, yvalue @};
1388 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1389 @samp{@var{fieldname}:}, as shown here:
1392 struct point p = @{ y: yvalue, x: xvalue @};
1396 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1397 @dfn{designator}. You can also use a designator (or the obsolete colon
1398 syntax) when initializing a union, to specify which element of the union
1399 should be used. For example,
1402 union foo @{ int i; double d; @};
1404 union foo f = @{ .d = 4 @};
1408 will convert 4 to a @code{double} to store it in the union using
1409 the second element. By contrast, casting 4 to type @code{union foo}
1410 would store it into the union as the integer @code{i}, since it is
1411 an integer. (@xref{Cast to Union}.)
1413 You can combine this technique of naming elements with ordinary C
1414 initialization of successive elements. Each initializer element that
1415 does not have a designator applies to the next consecutive element of the
1416 array or structure. For example,
1419 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1426 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1429 Labeling the elements of an array initializer is especially useful
1430 when the indices are characters or belong to an @code{enum} type.
1435 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1436 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1439 @cindex designator lists
1440 You can also write a series of @samp{.@var{fieldname}} and
1441 @samp{[@var{index}]} designators before an @samp{=} to specify a
1442 nested subobject to initialize; the list is taken relative to the
1443 subobject corresponding to the closest surrounding brace pair. For
1444 example, with the @samp{struct point} declaration above:
1447 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1451 If the same field is initialized multiple times, it will have value from
1452 the last initialization. If any such overridden initialization has
1453 side-effect, it is unspecified whether the side-effect happens or not.
1454 Currently, GCC will discard them and issue a warning.
1457 @section Case Ranges
1459 @cindex ranges in case statements
1461 You can specify a range of consecutive values in a single @code{case} label,
1465 case @var{low} ... @var{high}:
1469 This has the same effect as the proper number of individual @code{case}
1470 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1472 This feature is especially useful for ranges of ASCII character codes:
1478 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1479 it may be parsed wrong when you use it with integer values. For example,
1494 @section Cast to a Union Type
1495 @cindex cast to a union
1496 @cindex union, casting to a
1498 A cast to union type is similar to other casts, except that the type
1499 specified is a union type. You can specify the type either with
1500 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1501 a constructor though, not a cast, and hence does not yield an lvalue like
1502 normal casts. (@xref{Compound Literals}.)
1504 The types that may be cast to the union type are those of the members
1505 of the union. Thus, given the following union and variables:
1508 union foo @{ int i; double d; @};
1514 both @code{x} and @code{y} can be cast to type @code{union foo}.
1516 Using the cast as the right-hand side of an assignment to a variable of
1517 union type is equivalent to storing in a member of the union:
1522 u = (union foo) x @equiv{} u.i = x
1523 u = (union foo) y @equiv{} u.d = y
1526 You can also use the union cast as a function argument:
1529 void hack (union foo);
1531 hack ((union foo) x);
1534 @node Mixed Declarations
1535 @section Mixed Declarations and Code
1536 @cindex mixed declarations and code
1537 @cindex declarations, mixed with code
1538 @cindex code, mixed with declarations
1540 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1541 within compound statements. As an extension, GCC also allows this in
1542 C89 mode. For example, you could do:
1551 Each identifier is visible from where it is declared until the end of
1552 the enclosing block.
1554 @node Function Attributes
1555 @section Declaring Attributes of Functions
1556 @cindex function attributes
1557 @cindex declaring attributes of functions
1558 @cindex functions that never return
1559 @cindex functions that return more than once
1560 @cindex functions that have no side effects
1561 @cindex functions in arbitrary sections
1562 @cindex functions that behave like malloc
1563 @cindex @code{volatile} applied to function
1564 @cindex @code{const} applied to function
1565 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1566 @cindex functions with non-null pointer arguments
1567 @cindex functions that are passed arguments in registers on the 386
1568 @cindex functions that pop the argument stack on the 386
1569 @cindex functions that do not pop the argument stack on the 386
1571 In GNU C, you declare certain things about functions called in your program
1572 which help the compiler optimize function calls and check your code more
1575 The keyword @code{__attribute__} allows you to specify special
1576 attributes when making a declaration. This keyword is followed by an
1577 attribute specification inside double parentheses. The following
1578 attributes are currently defined for functions on all targets:
1579 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1580 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1581 @code{format}, @code{format_arg}, @code{no_instrument_function},
1582 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1583 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1584 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1585 @code{gnu_inline} and @code{externally_visible}. Several other
1586 attributes are defined for functions on particular target systems. Other
1587 attributes, including @code{section} are supported for variables declarations
1588 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1590 You may also specify attributes with @samp{__} preceding and following
1591 each keyword. This allows you to use them in header files without
1592 being concerned about a possible macro of the same name. For example,
1593 you may use @code{__noreturn__} instead of @code{noreturn}.
1595 @xref{Attribute Syntax}, for details of the exact syntax for using
1599 @c Keep this table alphabetized by attribute name. Treat _ as space.
1601 @item alias ("@var{target}")
1602 @cindex @code{alias} attribute
1603 The @code{alias} attribute causes the declaration to be emitted as an
1604 alias for another symbol, which must be specified. For instance,
1607 void __f () @{ /* @r{Do something.} */; @}
1608 void f () __attribute__ ((weak, alias ("__f")));
1611 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1612 mangled name for the target must be used. It is an error if @samp{__f}
1613 is not defined in the same translation unit.
1615 Not all target machines support this attribute.
1618 @cindex @code{always_inline} function attribute
1619 Generally, functions are not inlined unless optimization is specified.
1620 For functions declared inline, this attribute inlines the function even
1621 if no optimization level was specified.
1624 @cindex @code{gnu_inline} function attribute
1625 This attribute should be used with a function which is also declared
1626 with the @code{inline} keyword. It directs GCC to treat the function
1627 as if it were defined in gnu89 mode even when compiling in C99 or
1630 If the function is declared @code{extern}, then this definition of the
1631 function is used only for inlining. In no case is the function
1632 compiled as a standalone function, not even if you take its address
1633 explicitly. Such an address becomes an external reference, as if you
1634 had only declared the function, and had not defined it. This has
1635 almost the effect of a macro. The way to use this is to put a
1636 function definition in a header file with this attribute, and put
1637 another copy of the function, without @code{extern}, in a library
1638 file. The definition in the header file will cause most calls to the
1639 function to be inlined. If any uses of the function remain, they will
1640 refer to the single copy in the library. Note that the two
1641 definitions of the functions need not be precisely the same, although
1642 if they do not have the same effect your program may behave oddly.
1644 If the function is neither @code{extern} nor @code{static}, then the
1645 function is compiled as a standalone function, as well as being
1646 inlined where possible.
1648 This is how GCC traditionally handled functions declared
1649 @code{inline}. Since ISO C99 specifies a different semantics for
1650 @code{inline}, this function attribute is provided as a transition
1651 measure and as a useful feature in its own right. This attribute is
1652 available in GCC 4.1.3 and later. It is available if either of the
1653 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1654 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1655 Function is As Fast As a Macro}.
1657 Note that since the first version of GCC to support C99 inline semantics
1658 is 4.3, earlier versions of GCC which accept this attribute effectively
1659 assume that it is always present, whether or not it is given explicitly.
1660 In versions prior to 4.3, the only effect of explicitly including it is
1661 to disable warnings about using inline functions in C99 mode.
1663 @cindex @code{flatten} function attribute
1665 Generally, inlining into a function is limited. For a function marked with
1666 this attribute, every call inside this function will be inlined, if possible.
1667 Whether the function itself is considered for inlining depends on its size and
1668 the current inlining parameters. The @code{flatten} attribute only works
1669 reliably in unit-at-a-time mode.
1672 @cindex functions that do pop the argument stack on the 386
1674 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1675 assume that the calling function will pop off the stack space used to
1676 pass arguments. This is
1677 useful to override the effects of the @option{-mrtd} switch.
1680 @cindex @code{const} function attribute
1681 Many functions do not examine any values except their arguments, and
1682 have no effects except the return value. Basically this is just slightly
1683 more strict class than the @code{pure} attribute below, since function is not
1684 allowed to read global memory.
1686 @cindex pointer arguments
1687 Note that a function that has pointer arguments and examines the data
1688 pointed to must @emph{not} be declared @code{const}. Likewise, a
1689 function that calls a non-@code{const} function usually must not be
1690 @code{const}. It does not make sense for a @code{const} function to
1693 The attribute @code{const} is not implemented in GCC versions earlier
1694 than 2.5. An alternative way to declare that a function has no side
1695 effects, which works in the current version and in some older versions,
1699 typedef int intfn ();
1701 extern const intfn square;
1704 This approach does not work in GNU C++ from 2.6.0 on, since the language
1705 specifies that the @samp{const} must be attached to the return value.
1709 @cindex @code{constructor} function attribute
1710 @cindex @code{destructor} function attribute
1711 The @code{constructor} attribute causes the function to be called
1712 automatically before execution enters @code{main ()}. Similarly, the
1713 @code{destructor} attribute causes the function to be called
1714 automatically after @code{main ()} has completed or @code{exit ()} has
1715 been called. Functions with these attributes are useful for
1716 initializing data that will be used implicitly during the execution of
1719 These attributes are not currently implemented for Objective-C@.
1722 @cindex @code{deprecated} attribute.
1723 The @code{deprecated} attribute results in a warning if the function
1724 is used anywhere in the source file. This is useful when identifying
1725 functions that are expected to be removed in a future version of a
1726 program. The warning also includes the location of the declaration
1727 of the deprecated function, to enable users to easily find further
1728 information about why the function is deprecated, or what they should
1729 do instead. Note that the warnings only occurs for uses:
1732 int old_fn () __attribute__ ((deprecated));
1734 int (*fn_ptr)() = old_fn;
1737 results in a warning on line 3 but not line 2.
1739 The @code{deprecated} attribute can also be used for variables and
1740 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1743 @cindex @code{__declspec(dllexport)}
1744 On Microsoft Windows targets and Symbian OS targets the
1745 @code{dllexport} attribute causes the compiler to provide a global
1746 pointer to a pointer in a DLL, so that it can be referenced with the
1747 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1748 name is formed by combining @code{_imp__} and the function or variable
1751 You can use @code{__declspec(dllexport)} as a synonym for
1752 @code{__attribute__ ((dllexport))} for compatibility with other
1755 On systems that support the @code{visibility} attribute, this
1756 attribute also implies ``default'' visibility, unless a
1757 @code{visibility} attribute is explicitly specified. You should avoid
1758 the use of @code{dllexport} with ``hidden'' or ``internal''
1759 visibility; in the future GCC may issue an error for those cases.
1761 Currently, the @code{dllexport} attribute is ignored for inlined
1762 functions, unless the @option{-fkeep-inline-functions} flag has been
1763 used. The attribute is also ignored for undefined symbols.
1765 When applied to C++ classes, the attribute marks defined non-inlined
1766 member functions and static data members as exports. Static consts
1767 initialized in-class are not marked unless they are also defined
1770 For Microsoft Windows targets there are alternative methods for
1771 including the symbol in the DLL's export table such as using a
1772 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1773 the @option{--export-all} linker flag.
1776 @cindex @code{__declspec(dllimport)}
1777 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1778 attribute causes the compiler to reference a function or variable via
1779 a global pointer to a pointer that is set up by the DLL exporting the
1780 symbol. The attribute implies @code{extern} storage. On Microsoft
1781 Windows targets, the pointer name is formed by combining @code{_imp__}
1782 and the function or variable name.
1784 You can use @code{__declspec(dllimport)} as a synonym for
1785 @code{__attribute__ ((dllimport))} for compatibility with other
1788 Currently, the attribute is ignored for inlined functions. If the
1789 attribute is applied to a symbol @emph{definition}, an error is reported.
1790 If a symbol previously declared @code{dllimport} is later defined, the
1791 attribute is ignored in subsequent references, and a warning is emitted.
1792 The attribute is also overridden by a subsequent declaration as
1795 When applied to C++ classes, the attribute marks non-inlined
1796 member functions and static data members as imports. However, the
1797 attribute is ignored for virtual methods to allow creation of vtables
1800 On the SH Symbian OS target the @code{dllimport} attribute also has
1801 another affect---it can cause the vtable and run-time type information
1802 for a class to be exported. This happens when the class has a
1803 dllimport'ed constructor or a non-inline, non-pure virtual function
1804 and, for either of those two conditions, the class also has a inline
1805 constructor or destructor and has a key function that is defined in
1806 the current translation unit.
1808 For Microsoft Windows based targets the use of the @code{dllimport}
1809 attribute on functions is not necessary, but provides a small
1810 performance benefit by eliminating a thunk in the DLL@. The use of the
1811 @code{dllimport} attribute on imported variables was required on older
1812 versions of the GNU linker, but can now be avoided by passing the
1813 @option{--enable-auto-import} switch to the GNU linker. As with
1814 functions, using the attribute for a variable eliminates a thunk in
1817 One drawback to using this attribute is that a pointer to a function
1818 or variable marked as @code{dllimport} cannot be used as a constant
1819 address. On Microsoft Windows targets, the attribute can be disabled
1820 for functions by setting the @option{-mnop-fun-dllimport} flag.
1823 @cindex eight bit data on the H8/300, H8/300H, and H8S
1824 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1825 variable should be placed into the eight bit data section.
1826 The compiler will generate more efficient code for certain operations
1827 on data in the eight bit data area. Note the eight bit data area is limited to
1830 You must use GAS and GLD from GNU binutils version 2.7 or later for
1831 this attribute to work correctly.
1833 @item exception_handler
1834 @cindex exception handler functions on the Blackfin processor
1835 Use this attribute on the Blackfin to indicate that the specified function
1836 is an exception handler. The compiler will generate function entry and
1837 exit sequences suitable for use in an exception handler when this
1838 attribute is present.
1841 @cindex functions which handle memory bank switching
1842 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1843 use a calling convention that takes care of switching memory banks when
1844 entering and leaving a function. This calling convention is also the
1845 default when using the @option{-mlong-calls} option.
1847 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1848 to call and return from a function.
1850 On 68HC11 the compiler will generate a sequence of instructions
1851 to invoke a board-specific routine to switch the memory bank and call the
1852 real function. The board-specific routine simulates a @code{call}.
1853 At the end of a function, it will jump to a board-specific routine
1854 instead of using @code{rts}. The board-specific return routine simulates
1858 @cindex functions that pop the argument stack on the 386
1859 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1860 pass the first argument (if of integral type) in the register ECX and
1861 the second argument (if of integral type) in the register EDX@. Subsequent
1862 and other typed arguments are passed on the stack. The called function will
1863 pop the arguments off the stack. If the number of arguments is variable all
1864 arguments are pushed on the stack.
1866 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1867 @cindex @code{format} function attribute
1869 The @code{format} attribute specifies that a function takes @code{printf},
1870 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1871 should be type-checked against a format string. For example, the
1876 my_printf (void *my_object, const char *my_format, ...)
1877 __attribute__ ((format (printf, 2, 3)));
1881 causes the compiler to check the arguments in calls to @code{my_printf}
1882 for consistency with the @code{printf} style format string argument
1885 The parameter @var{archetype} determines how the format string is
1886 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1887 or @code{strfmon}. (You can also use @code{__printf__},
1888 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1889 parameter @var{string-index} specifies which argument is the format
1890 string argument (starting from 1), while @var{first-to-check} is the
1891 number of the first argument to check against the format string. For
1892 functions where the arguments are not available to be checked (such as
1893 @code{vprintf}), specify the third parameter as zero. In this case the
1894 compiler only checks the format string for consistency. For
1895 @code{strftime} formats, the third parameter is required to be zero.
1896 Since non-static C++ methods have an implicit @code{this} argument, the
1897 arguments of such methods should be counted from two, not one, when
1898 giving values for @var{string-index} and @var{first-to-check}.
1900 In the example above, the format string (@code{my_format}) is the second
1901 argument of the function @code{my_print}, and the arguments to check
1902 start with the third argument, so the correct parameters for the format
1903 attribute are 2 and 3.
1905 @opindex ffreestanding
1906 @opindex fno-builtin
1907 The @code{format} attribute allows you to identify your own functions
1908 which take format strings as arguments, so that GCC can check the
1909 calls to these functions for errors. The compiler always (unless
1910 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1911 for the standard library functions @code{printf}, @code{fprintf},
1912 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1913 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1914 warnings are requested (using @option{-Wformat}), so there is no need to
1915 modify the header file @file{stdio.h}. In C99 mode, the functions
1916 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1917 @code{vsscanf} are also checked. Except in strictly conforming C
1918 standard modes, the X/Open function @code{strfmon} is also checked as
1919 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1920 @xref{C Dialect Options,,Options Controlling C Dialect}.
1922 The target may provide additional types of format checks.
1923 @xref{Target Format Checks,,Format Checks Specific to Particular
1926 @item format_arg (@var{string-index})
1927 @cindex @code{format_arg} function attribute
1928 @opindex Wformat-nonliteral
1929 The @code{format_arg} attribute specifies that a function takes a format
1930 string for a @code{printf}, @code{scanf}, @code{strftime} or
1931 @code{strfmon} style function and modifies it (for example, to translate
1932 it into another language), so the result can be passed to a
1933 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1934 function (with the remaining arguments to the format function the same
1935 as they would have been for the unmodified string). For example, the
1940 my_dgettext (char *my_domain, const char *my_format)
1941 __attribute__ ((format_arg (2)));
1945 causes the compiler to check the arguments in calls to a @code{printf},
1946 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1947 format string argument is a call to the @code{my_dgettext} function, for
1948 consistency with the format string argument @code{my_format}. If the
1949 @code{format_arg} attribute had not been specified, all the compiler
1950 could tell in such calls to format functions would be that the format
1951 string argument is not constant; this would generate a warning when
1952 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1953 without the attribute.
1955 The parameter @var{string-index} specifies which argument is the format
1956 string argument (starting from one). Since non-static C++ methods have
1957 an implicit @code{this} argument, the arguments of such methods should
1958 be counted from two.
1960 The @code{format-arg} attribute allows you to identify your own
1961 functions which modify format strings, so that GCC can check the
1962 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1963 type function whose operands are a call to one of your own function.
1964 The compiler always treats @code{gettext}, @code{dgettext}, and
1965 @code{dcgettext} in this manner except when strict ISO C support is
1966 requested by @option{-ansi} or an appropriate @option{-std} option, or
1967 @option{-ffreestanding} or @option{-fno-builtin}
1968 is used. @xref{C Dialect Options,,Options
1969 Controlling C Dialect}.
1971 @item function_vector
1972 @cindex calling functions through the function vector on the H8/300 processors
1973 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1974 function should be called through the function vector. Calling a
1975 function through the function vector will reduce code size, however;
1976 the function vector has a limited size (maximum 128 entries on the H8/300
1977 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1979 You must use GAS and GLD from GNU binutils version 2.7 or later for
1980 this attribute to work correctly.
1983 @cindex interrupt handler functions
1984 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
1985 ports to indicate that the specified function is an interrupt handler.
1986 The compiler will generate function entry and exit sequences suitable
1987 for use in an interrupt handler when this attribute is present.
1989 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1990 SH processors can be specified via the @code{interrupt_handler} attribute.
1992 Note, on the AVR, interrupts will be enabled inside the function.
1994 Note, for the ARM, you can specify the kind of interrupt to be handled by
1995 adding an optional parameter to the interrupt attribute like this:
1998 void f () __attribute__ ((interrupt ("IRQ")));
2001 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2003 @item interrupt_handler
2004 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2005 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2006 indicate that the specified function is an interrupt handler. The compiler
2007 will generate function entry and exit sequences suitable for use in an
2008 interrupt handler when this attribute is present.
2011 @cindex User stack pointer in interrupts on the Blackfin
2012 When used together with @code{interrupt_handler}, @code{exception_handler}
2013 or @code{nmi_handler}, code will be generated to load the stack pointer
2014 from the USP register in the function prologue.
2016 @item long_call/short_call
2017 @cindex indirect calls on ARM
2018 This attribute specifies how a particular function is called on
2019 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2020 command line switch and @code{#pragma long_calls} settings. The
2021 @code{long_call} attribute indicates that the function might be far
2022 away from the call site and require a different (more expensive)
2023 calling sequence. The @code{short_call} attribute always places
2024 the offset to the function from the call site into the @samp{BL}
2025 instruction directly.
2027 @item longcall/shortcall
2028 @cindex functions called via pointer on the RS/6000 and PowerPC
2029 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2030 indicates that the function might be far away from the call site and
2031 require a different (more expensive) calling sequence. The
2032 @code{shortcall} attribute indicates that the function is always close
2033 enough for the shorter calling sequence to be used. These attributes
2034 override both the @option{-mlongcall} switch and, on the RS/6000 and
2035 PowerPC, the @code{#pragma longcall} setting.
2037 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2038 calls are necessary.
2041 @cindex indirect calls on MIPS
2042 This attribute specifies how a particular function is called on MIPS@.
2043 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2044 command line switch. This attribute causes the compiler to always call
2045 the function by first loading its address into a register, and then using
2046 the contents of that register.
2049 @cindex @code{malloc} attribute
2050 The @code{malloc} attribute is used to tell the compiler that a function
2051 may be treated as if any non-@code{NULL} pointer it returns cannot
2052 alias any other pointer valid when the function returns.
2053 This will often improve optimization.
2054 Standard functions with this property include @code{malloc} and
2055 @code{calloc}. @code{realloc}-like functions have this property as
2056 long as the old pointer is never referred to (including comparing it
2057 to the new pointer) after the function returns a non-@code{NULL}
2060 @item model (@var{model-name})
2061 @cindex function addressability on the M32R/D
2062 @cindex variable addressability on the IA-64
2064 On the M32R/D, use this attribute to set the addressability of an
2065 object, and of the code generated for a function. The identifier
2066 @var{model-name} is one of @code{small}, @code{medium}, or
2067 @code{large}, representing each of the code models.
2069 Small model objects live in the lower 16MB of memory (so that their
2070 addresses can be loaded with the @code{ld24} instruction), and are
2071 callable with the @code{bl} instruction.
2073 Medium model objects may live anywhere in the 32-bit address space (the
2074 compiler will generate @code{seth/add3} instructions to load their addresses),
2075 and are callable with the @code{bl} instruction.
2077 Large model objects may live anywhere in the 32-bit address space (the
2078 compiler will generate @code{seth/add3} instructions to load their addresses),
2079 and may not be reachable with the @code{bl} instruction (the compiler will
2080 generate the much slower @code{seth/add3/jl} instruction sequence).
2082 On IA-64, use this attribute to set the addressability of an object.
2083 At present, the only supported identifier for @var{model-name} is
2084 @code{small}, indicating addressability via ``small'' (22-bit)
2085 addresses (so that their addresses can be loaded with the @code{addl}
2086 instruction). Caveat: such addressing is by definition not position
2087 independent and hence this attribute must not be used for objects
2088 defined by shared libraries.
2091 @cindex function without a prologue/epilogue code
2092 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2093 specified function does not need prologue/epilogue sequences generated by
2094 the compiler. It is up to the programmer to provide these sequences.
2097 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2098 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2099 use the normal calling convention based on @code{jsr} and @code{rts}.
2100 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2104 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2105 Use this attribute together with @code{interrupt_handler},
2106 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2107 entry code should enable nested interrupts or exceptions.
2110 @cindex NMI handler functions on the Blackfin processor
2111 Use this attribute on the Blackfin to indicate that the specified function
2112 is an NMI handler. The compiler will generate function entry and
2113 exit sequences suitable for use in an NMI handler when this
2114 attribute is present.
2116 @item no_instrument_function
2117 @cindex @code{no_instrument_function} function attribute
2118 @opindex finstrument-functions
2119 If @option{-finstrument-functions} is given, profiling function calls will
2120 be generated at entry and exit of most user-compiled functions.
2121 Functions with this attribute will not be so instrumented.
2124 @cindex @code{noinline} function attribute
2125 This function attribute prevents a function from being considered for
2128 @item nonnull (@var{arg-index}, @dots{})
2129 @cindex @code{nonnull} function attribute
2130 The @code{nonnull} attribute specifies that some function parameters should
2131 be non-null pointers. For instance, the declaration:
2135 my_memcpy (void *dest, const void *src, size_t len)
2136 __attribute__((nonnull (1, 2)));
2140 causes the compiler to check that, in calls to @code{my_memcpy},
2141 arguments @var{dest} and @var{src} are non-null. If the compiler
2142 determines that a null pointer is passed in an argument slot marked
2143 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2144 is issued. The compiler may also choose to make optimizations based
2145 on the knowledge that certain function arguments will not be null.
2147 If no argument index list is given to the @code{nonnull} attribute,
2148 all pointer arguments are marked as non-null. To illustrate, the
2149 following declaration is equivalent to the previous example:
2153 my_memcpy (void *dest, const void *src, size_t len)
2154 __attribute__((nonnull));
2158 @cindex @code{noreturn} function attribute
2159 A few standard library functions, such as @code{abort} and @code{exit},
2160 cannot return. GCC knows this automatically. Some programs define
2161 their own functions that never return. You can declare them
2162 @code{noreturn} to tell the compiler this fact. For example,
2166 void fatal () __attribute__ ((noreturn));
2169 fatal (/* @r{@dots{}} */)
2171 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2177 The @code{noreturn} keyword tells the compiler to assume that
2178 @code{fatal} cannot return. It can then optimize without regard to what
2179 would happen if @code{fatal} ever did return. This makes slightly
2180 better code. More importantly, it helps avoid spurious warnings of
2181 uninitialized variables.
2183 The @code{noreturn} keyword does not affect the exceptional path when that
2184 applies: a @code{noreturn}-marked function may still return to the caller
2185 by throwing an exception or calling @code{longjmp}.
2187 Do not assume that registers saved by the calling function are
2188 restored before calling the @code{noreturn} function.
2190 It does not make sense for a @code{noreturn} function to have a return
2191 type other than @code{void}.
2193 The attribute @code{noreturn} is not implemented in GCC versions
2194 earlier than 2.5. An alternative way to declare that a function does
2195 not return, which works in the current version and in some older
2196 versions, is as follows:
2199 typedef void voidfn ();
2201 volatile voidfn fatal;
2204 This approach does not work in GNU C++.
2207 @cindex @code{nothrow} function attribute
2208 The @code{nothrow} attribute is used to inform the compiler that a
2209 function cannot throw an exception. For example, most functions in
2210 the standard C library can be guaranteed not to throw an exception
2211 with the notable exceptions of @code{qsort} and @code{bsearch} that
2212 take function pointer arguments. The @code{nothrow} attribute is not
2213 implemented in GCC versions earlier than 3.3.
2216 @cindex @code{pure} function attribute
2217 Many functions have no effects except the return value and their
2218 return value depends only on the parameters and/or global variables.
2219 Such a function can be subject
2220 to common subexpression elimination and loop optimization just as an
2221 arithmetic operator would be. These functions should be declared
2222 with the attribute @code{pure}. For example,
2225 int square (int) __attribute__ ((pure));
2229 says that the hypothetical function @code{square} is safe to call
2230 fewer times than the program says.
2232 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2233 Interesting non-pure functions are functions with infinite loops or those
2234 depending on volatile memory or other system resource, that may change between
2235 two consecutive calls (such as @code{feof} in a multithreading environment).
2237 The attribute @code{pure} is not implemented in GCC versions earlier
2240 @item regparm (@var{number})
2241 @cindex @code{regparm} attribute
2242 @cindex functions that are passed arguments in registers on the 386
2243 On the Intel 386, the @code{regparm} attribute causes the compiler to
2244 pass arguments number one to @var{number} if they are of integral type
2245 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2246 take a variable number of arguments will continue to be passed all of their
2247 arguments on the stack.
2249 Beware that on some ELF systems this attribute is unsuitable for
2250 global functions in shared libraries with lazy binding (which is the
2251 default). Lazy binding will send the first call via resolving code in
2252 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2253 per the standard calling conventions. Solaris 8 is affected by this.
2254 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2255 safe since the loaders there save all registers. (Lazy binding can be
2256 disabled with the linker or the loader if desired, to avoid the
2260 @cindex @code{sseregparm} attribute
2261 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2262 causes the compiler to pass up to 3 floating point arguments in
2263 SSE registers instead of on the stack. Functions that take a
2264 variable number of arguments will continue to pass all of their
2265 floating point arguments on the stack.
2267 @item force_align_arg_pointer
2268 @cindex @code{force_align_arg_pointer} attribute
2269 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2270 applied to individual function definitions, generating an alternate
2271 prologue and epilogue that realigns the runtime stack. This supports
2272 mixing legacy codes that run with a 4-byte aligned stack with modern
2273 codes that keep a 16-byte stack for SSE compatibility. The alternate
2274 prologue and epilogue are slower and bigger than the regular ones, and
2275 the alternate prologue requires a scratch register; this lowers the
2276 number of registers available if used in conjunction with the
2277 @code{regparm} attribute. The @code{force_align_arg_pointer}
2278 attribute is incompatible with nested functions; this is considered a
2282 @cindex @code{returns_twice} attribute
2283 The @code{returns_twice} attribute tells the compiler that a function may
2284 return more than one time. The compiler will ensure that all registers
2285 are dead before calling such a function and will emit a warning about
2286 the variables that may be clobbered after the second return from the
2287 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2288 The @code{longjmp}-like counterpart of such function, if any, might need
2289 to be marked with the @code{noreturn} attribute.
2292 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2293 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2294 all registers except the stack pointer should be saved in the prologue
2295 regardless of whether they are used or not.
2297 @item section ("@var{section-name}")
2298 @cindex @code{section} function attribute
2299 Normally, the compiler places the code it generates in the @code{text} section.
2300 Sometimes, however, you need additional sections, or you need certain
2301 particular functions to appear in special sections. The @code{section}
2302 attribute specifies that a function lives in a particular section.
2303 For example, the declaration:
2306 extern void foobar (void) __attribute__ ((section ("bar")));
2310 puts the function @code{foobar} in the @code{bar} section.
2312 Some file formats do not support arbitrary sections so the @code{section}
2313 attribute is not available on all platforms.
2314 If you need to map the entire contents of a module to a particular
2315 section, consider using the facilities of the linker instead.
2318 @cindex @code{sentinel} function attribute
2319 This function attribute ensures that a parameter in a function call is
2320 an explicit @code{NULL}. The attribute is only valid on variadic
2321 functions. By default, the sentinel is located at position zero, the
2322 last parameter of the function call. If an optional integer position
2323 argument P is supplied to the attribute, the sentinel must be located at
2324 position P counting backwards from the end of the argument list.
2327 __attribute__ ((sentinel))
2329 __attribute__ ((sentinel(0)))
2332 The attribute is automatically set with a position of 0 for the built-in
2333 functions @code{execl} and @code{execlp}. The built-in function
2334 @code{execle} has the attribute set with a position of 1.
2336 A valid @code{NULL} in this context is defined as zero with any pointer
2337 type. If your system defines the @code{NULL} macro with an integer type
2338 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2339 with a copy that redefines NULL appropriately.
2341 The warnings for missing or incorrect sentinels are enabled with
2345 See long_call/short_call.
2348 See longcall/shortcall.
2351 @cindex signal handler functions on the AVR processors
2352 Use this attribute on the AVR to indicate that the specified
2353 function is a signal handler. The compiler will generate function
2354 entry and exit sequences suitable for use in a signal handler when this
2355 attribute is present. Interrupts will be disabled inside the function.
2358 Use this attribute on the SH to indicate an @code{interrupt_handler}
2359 function should switch to an alternate stack. It expects a string
2360 argument that names a global variable holding the address of the
2365 void f () __attribute__ ((interrupt_handler,
2366 sp_switch ("alt_stack")));
2370 @cindex functions that pop the argument stack on the 386
2371 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2372 assume that the called function will pop off the stack space used to
2373 pass arguments, unless it takes a variable number of arguments.
2376 @cindex tiny data section on the H8/300H and H8S
2377 Use this attribute on the H8/300H and H8S to indicate that the specified
2378 variable should be placed into the tiny data section.
2379 The compiler will generate more efficient code for loads and stores
2380 on data in the tiny data section. Note the tiny data area is limited to
2381 slightly under 32kbytes of data.
2384 Use this attribute on the SH for an @code{interrupt_handler} to return using
2385 @code{trapa} instead of @code{rte}. This attribute expects an integer
2386 argument specifying the trap number to be used.
2389 @cindex @code{unused} attribute.
2390 This attribute, attached to a function, means that the function is meant
2391 to be possibly unused. GCC will not produce a warning for this
2395 @cindex @code{used} attribute.
2396 This attribute, attached to a function, means that code must be emitted
2397 for the function even if it appears that the function is not referenced.
2398 This is useful, for example, when the function is referenced only in
2401 @item visibility ("@var{visibility_type}")
2402 @cindex @code{visibility} attribute
2403 This attribute affects the linkage of the declaration to which it is attached.
2404 There are four supported @var{visibility_type} values: default,
2405 hidden, protected or internal visibility.
2408 void __attribute__ ((visibility ("protected")))
2409 f () @{ /* @r{Do something.} */; @}
2410 int i __attribute__ ((visibility ("hidden")));
2413 The possible values of @var{visibility_type} correspond to the
2414 visibility settings in the ELF gABI.
2417 @c keep this list of visibilities in alphabetical order.
2420 Default visibility is the normal case for the object file format.
2421 This value is available for the visibility attribute to override other
2422 options that may change the assumed visibility of entities.
2424 On ELF, default visibility means that the declaration is visible to other
2425 modules and, in shared libraries, means that the declared entity may be
2428 On Darwin, default visibility means that the declaration is visible to
2431 Default visibility corresponds to ``external linkage'' in the language.
2434 Hidden visibility indicates that the entity declared will have a new
2435 form of linkage, which we'll call ``hidden linkage''. Two
2436 declarations of an object with hidden linkage refer to the same object
2437 if they are in the same shared object.
2440 Internal visibility is like hidden visibility, but with additional
2441 processor specific semantics. Unless otherwise specified by the
2442 psABI, GCC defines internal visibility to mean that a function is
2443 @emph{never} called from another module. Compare this with hidden
2444 functions which, while they cannot be referenced directly by other
2445 modules, can be referenced indirectly via function pointers. By
2446 indicating that a function cannot be called from outside the module,
2447 GCC may for instance omit the load of a PIC register since it is known
2448 that the calling function loaded the correct value.
2451 Protected visibility is like default visibility except that it
2452 indicates that references within the defining module will bind to the
2453 definition in that module. That is, the declared entity cannot be
2454 overridden by another module.
2458 All visibilities are supported on many, but not all, ELF targets
2459 (supported when the assembler supports the @samp{.visibility}
2460 pseudo-op). Default visibility is supported everywhere. Hidden
2461 visibility is supported on Darwin targets.
2463 The visibility attribute should be applied only to declarations which
2464 would otherwise have external linkage. The attribute should be applied
2465 consistently, so that the same entity should not be declared with
2466 different settings of the attribute.
2468 In C++, the visibility attribute applies to types as well as functions
2469 and objects, because in C++ types have linkage. A class must not have
2470 greater visibility than its non-static data member types and bases,
2471 and class members default to the visibility of their class. Also, a
2472 declaration without explicit visibility is limited to the visibility
2475 In C++, you can mark member functions and static member variables of a
2476 class with the visibility attribute. This is useful if if you know a
2477 particular method or static member variable should only be used from
2478 one shared object; then you can mark it hidden while the rest of the
2479 class has default visibility. Care must be taken to avoid breaking
2480 the One Definition Rule; for example, it is usually not useful to mark
2481 an inline method as hidden without marking the whole class as hidden.
2483 A C++ namespace declaration can also have the visibility attribute.
2484 This attribute applies only to the particular namespace body, not to
2485 other definitions of the same namespace; it is equivalent to using
2486 @samp{#pragma GCC visibility} before and after the namespace
2487 definition (@pxref{Visibility Pragmas}).
2489 In C++, if a template argument has limited visibility, this
2490 restriction is implicitly propagated to the template instantiation.
2491 Otherwise, template instantiations and specializations default to the
2492 visibility of their template.
2494 If both the template and enclosing class have explicit visibility, the
2495 visibility from the template is used.
2497 @item warn_unused_result
2498 @cindex @code{warn_unused_result} attribute
2499 The @code{warn_unused_result} attribute causes a warning to be emitted
2500 if a caller of the function with this attribute does not use its
2501 return value. This is useful for functions where not checking
2502 the result is either a security problem or always a bug, such as
2506 int fn () __attribute__ ((warn_unused_result));
2509 if (fn () < 0) return -1;
2515 results in warning on line 5.
2518 @cindex @code{weak} attribute
2519 The @code{weak} attribute causes the declaration to be emitted as a weak
2520 symbol rather than a global. This is primarily useful in defining
2521 library functions which can be overridden in user code, though it can
2522 also be used with non-function declarations. Weak symbols are supported
2523 for ELF targets, and also for a.out targets when using the GNU assembler
2527 @itemx weakref ("@var{target}")
2528 @cindex @code{weakref} attribute
2529 The @code{weakref} attribute marks a declaration as a weak reference.
2530 Without arguments, it should be accompanied by an @code{alias} attribute
2531 naming the target symbol. Optionally, the @var{target} may be given as
2532 an argument to @code{weakref} itself. In either case, @code{weakref}
2533 implicitly marks the declaration as @code{weak}. Without a
2534 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2535 @code{weakref} is equivalent to @code{weak}.
2538 static int x() __attribute__ ((weakref ("y")));
2539 /* is equivalent to... */
2540 static int x() __attribute__ ((weak, weakref, alias ("y")));
2542 static int x() __attribute__ ((weakref));
2543 static int x() __attribute__ ((alias ("y")));
2546 A weak reference is an alias that does not by itself require a
2547 definition to be given for the target symbol. If the target symbol is
2548 only referenced through weak references, then the becomes a @code{weak}
2549 undefined symbol. If it is directly referenced, however, then such
2550 strong references prevail, and a definition will be required for the
2551 symbol, not necessarily in the same translation unit.
2553 The effect is equivalent to moving all references to the alias to a
2554 separate translation unit, renaming the alias to the aliased symbol,
2555 declaring it as weak, compiling the two separate translation units and
2556 performing a reloadable link on them.
2558 At present, a declaration to which @code{weakref} is attached can
2559 only be @code{static}.
2561 @item externally_visible
2562 @cindex @code{externally_visible} attribute.
2563 This attribute, attached to a global variable or function nullify
2564 effect of @option{-fwhole-program} command line option, so the object
2565 remain visible outside the current compilation unit
2569 You can specify multiple attributes in a declaration by separating them
2570 by commas within the double parentheses or by immediately following an
2571 attribute declaration with another attribute declaration.
2573 @cindex @code{#pragma}, reason for not using
2574 @cindex pragma, reason for not using
2575 Some people object to the @code{__attribute__} feature, suggesting that
2576 ISO C's @code{#pragma} should be used instead. At the time
2577 @code{__attribute__} was designed, there were two reasons for not doing
2582 It is impossible to generate @code{#pragma} commands from a macro.
2585 There is no telling what the same @code{#pragma} might mean in another
2589 These two reasons applied to almost any application that might have been
2590 proposed for @code{#pragma}. It was basically a mistake to use
2591 @code{#pragma} for @emph{anything}.
2593 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2594 to be generated from macros. In addition, a @code{#pragma GCC}
2595 namespace is now in use for GCC-specific pragmas. However, it has been
2596 found convenient to use @code{__attribute__} to achieve a natural
2597 attachment of attributes to their corresponding declarations, whereas
2598 @code{#pragma GCC} is of use for constructs that do not naturally form
2599 part of the grammar. @xref{Other Directives,,Miscellaneous
2600 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2602 @node Attribute Syntax
2603 @section Attribute Syntax
2604 @cindex attribute syntax
2606 This section describes the syntax with which @code{__attribute__} may be
2607 used, and the constructs to which attribute specifiers bind, for the C
2608 language. Some details may vary for C++ and Objective-C@. Because of
2609 infelicities in the grammar for attributes, some forms described here
2610 may not be successfully parsed in all cases.
2612 There are some problems with the semantics of attributes in C++. For
2613 example, there are no manglings for attributes, although they may affect
2614 code generation, so problems may arise when attributed types are used in
2615 conjunction with templates or overloading. Similarly, @code{typeid}
2616 does not distinguish between types with different attributes. Support
2617 for attributes in C++ may be restricted in future to attributes on
2618 declarations only, but not on nested declarators.
2620 @xref{Function Attributes}, for details of the semantics of attributes
2621 applying to functions. @xref{Variable Attributes}, for details of the
2622 semantics of attributes applying to variables. @xref{Type Attributes},
2623 for details of the semantics of attributes applying to structure, union
2624 and enumerated types.
2626 An @dfn{attribute specifier} is of the form
2627 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2628 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2629 each attribute is one of the following:
2633 Empty. Empty attributes are ignored.
2636 A word (which may be an identifier such as @code{unused}, or a reserved
2637 word such as @code{const}).
2640 A word, followed by, in parentheses, parameters for the attribute.
2641 These parameters take one of the following forms:
2645 An identifier. For example, @code{mode} attributes use this form.
2648 An identifier followed by a comma and a non-empty comma-separated list
2649 of expressions. For example, @code{format} attributes use this form.
2652 A possibly empty comma-separated list of expressions. For example,
2653 @code{format_arg} attributes use this form with the list being a single
2654 integer constant expression, and @code{alias} attributes use this form
2655 with the list being a single string constant.
2659 An @dfn{attribute specifier list} is a sequence of one or more attribute
2660 specifiers, not separated by any other tokens.
2662 In GNU C, an attribute specifier list may appear after the colon following a
2663 label, other than a @code{case} or @code{default} label. The only
2664 attribute it makes sense to use after a label is @code{unused}. This
2665 feature is intended for code generated by programs which contains labels
2666 that may be unused but which is compiled with @option{-Wall}. It would
2667 not normally be appropriate to use in it human-written code, though it
2668 could be useful in cases where the code that jumps to the label is
2669 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2670 such placement of attribute lists, as it is permissible for a
2671 declaration, which could begin with an attribute list, to be labelled in
2672 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2673 does not arise there.
2675 An attribute specifier list may appear as part of a @code{struct},
2676 @code{union} or @code{enum} specifier. It may go either immediately
2677 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2678 the closing brace. The former syntax is preferred.
2679 Where attribute specifiers follow the closing brace, they are considered
2680 to relate to the structure, union or enumerated type defined, not to any
2681 enclosing declaration the type specifier appears in, and the type
2682 defined is not complete until after the attribute specifiers.
2683 @c Otherwise, there would be the following problems: a shift/reduce
2684 @c conflict between attributes binding the struct/union/enum and
2685 @c binding to the list of specifiers/qualifiers; and "aligned"
2686 @c attributes could use sizeof for the structure, but the size could be
2687 @c changed later by "packed" attributes.
2689 Otherwise, an attribute specifier appears as part of a declaration,
2690 counting declarations of unnamed parameters and type names, and relates
2691 to that declaration (which may be nested in another declaration, for
2692 example in the case of a parameter declaration), or to a particular declarator
2693 within a declaration. Where an
2694 attribute specifier is applied to a parameter declared as a function or
2695 an array, it should apply to the function or array rather than the
2696 pointer to which the parameter is implicitly converted, but this is not
2697 yet correctly implemented.
2699 Any list of specifiers and qualifiers at the start of a declaration may
2700 contain attribute specifiers, whether or not such a list may in that
2701 context contain storage class specifiers. (Some attributes, however,
2702 are essentially in the nature of storage class specifiers, and only make
2703 sense where storage class specifiers may be used; for example,
2704 @code{section}.) There is one necessary limitation to this syntax: the
2705 first old-style parameter declaration in a function definition cannot
2706 begin with an attribute specifier, because such an attribute applies to
2707 the function instead by syntax described below (which, however, is not
2708 yet implemented in this case). In some other cases, attribute
2709 specifiers are permitted by this grammar but not yet supported by the
2710 compiler. All attribute specifiers in this place relate to the
2711 declaration as a whole. In the obsolescent usage where a type of
2712 @code{int} is implied by the absence of type specifiers, such a list of
2713 specifiers and qualifiers may be an attribute specifier list with no
2714 other specifiers or qualifiers.
2716 At present, the first parameter in a function prototype must have some
2717 type specifier which is not an attribute specifier; this resolves an
2718 ambiguity in the interpretation of @code{void f(int
2719 (__attribute__((foo)) x))}, but is subject to change. At present, if
2720 the parentheses of a function declarator contain only attributes then
2721 those attributes are ignored, rather than yielding an error or warning
2722 or implying a single parameter of type int, but this is subject to
2725 An attribute specifier list may appear immediately before a declarator
2726 (other than the first) in a comma-separated list of declarators in a
2727 declaration of more than one identifier using a single list of
2728 specifiers and qualifiers. Such attribute specifiers apply
2729 only to the identifier before whose declarator they appear. For
2733 __attribute__((noreturn)) void d0 (void),
2734 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2739 the @code{noreturn} attribute applies to all the functions
2740 declared; the @code{format} attribute only applies to @code{d1}.
2742 An attribute specifier list may appear immediately before the comma,
2743 @code{=} or semicolon terminating the declaration of an identifier other
2744 than a function definition. At present, such attribute specifiers apply
2745 to the declared object or function, but in future they may attach to the
2746 outermost adjacent declarator. In simple cases there is no difference,
2747 but, for example, in
2750 void (****f)(void) __attribute__((noreturn));
2754 at present the @code{noreturn} attribute applies to @code{f}, which
2755 causes a warning since @code{f} is not a function, but in future it may
2756 apply to the function @code{****f}. The precise semantics of what
2757 attributes in such cases will apply to are not yet specified. Where an
2758 assembler name for an object or function is specified (@pxref{Asm
2759 Labels}), at present the attribute must follow the @code{asm}
2760 specification; in future, attributes before the @code{asm} specification
2761 may apply to the adjacent declarator, and those after it to the declared
2764 An attribute specifier list may, in future, be permitted to appear after
2765 the declarator in a function definition (before any old-style parameter
2766 declarations or the function body).
2768 Attribute specifiers may be mixed with type qualifiers appearing inside
2769 the @code{[]} of a parameter array declarator, in the C99 construct by
2770 which such qualifiers are applied to the pointer to which the array is
2771 implicitly converted. Such attribute specifiers apply to the pointer,
2772 not to the array, but at present this is not implemented and they are
2775 An attribute specifier list may appear at the start of a nested
2776 declarator. At present, there are some limitations in this usage: the
2777 attributes correctly apply to the declarator, but for most individual
2778 attributes the semantics this implies are not implemented.
2779 When attribute specifiers follow the @code{*} of a pointer
2780 declarator, they may be mixed with any type qualifiers present.
2781 The following describes the formal semantics of this syntax. It will make the
2782 most sense if you are familiar with the formal specification of
2783 declarators in the ISO C standard.
2785 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2786 D1}, where @code{T} contains declaration specifiers that specify a type
2787 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2788 contains an identifier @var{ident}. The type specified for @var{ident}
2789 for derived declarators whose type does not include an attribute
2790 specifier is as in the ISO C standard.
2792 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2793 and the declaration @code{T D} specifies the type
2794 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2795 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2796 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2798 If @code{D1} has the form @code{*
2799 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2800 declaration @code{T D} specifies the type
2801 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2802 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2803 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2809 void (__attribute__((noreturn)) ****f) (void);
2813 specifies the type ``pointer to pointer to pointer to pointer to
2814 non-returning function returning @code{void}''. As another example,
2817 char *__attribute__((aligned(8))) *f;
2821 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2822 Note again that this does not work with most attributes; for example,
2823 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2824 is not yet supported.
2826 For compatibility with existing code written for compiler versions that
2827 did not implement attributes on nested declarators, some laxity is
2828 allowed in the placing of attributes. If an attribute that only applies
2829 to types is applied to a declaration, it will be treated as applying to
2830 the type of that declaration. If an attribute that only applies to
2831 declarations is applied to the type of a declaration, it will be treated
2832 as applying to that declaration; and, for compatibility with code
2833 placing the attributes immediately before the identifier declared, such
2834 an attribute applied to a function return type will be treated as
2835 applying to the function type, and such an attribute applied to an array
2836 element type will be treated as applying to the array type. If an
2837 attribute that only applies to function types is applied to a
2838 pointer-to-function type, it will be treated as applying to the pointer
2839 target type; if such an attribute is applied to a function return type
2840 that is not a pointer-to-function type, it will be treated as applying
2841 to the function type.
2843 @node Function Prototypes
2844 @section Prototypes and Old-Style Function Definitions
2845 @cindex function prototype declarations
2846 @cindex old-style function definitions
2847 @cindex promotion of formal parameters
2849 GNU C extends ISO C to allow a function prototype to override a later
2850 old-style non-prototype definition. Consider the following example:
2853 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2860 /* @r{Prototype function declaration.} */
2861 int isroot P((uid_t));
2863 /* @r{Old-style function definition.} */
2865 isroot (x) /* @r{??? lossage here ???} */
2872 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2873 not allow this example, because subword arguments in old-style
2874 non-prototype definitions are promoted. Therefore in this example the
2875 function definition's argument is really an @code{int}, which does not
2876 match the prototype argument type of @code{short}.
2878 This restriction of ISO C makes it hard to write code that is portable
2879 to traditional C compilers, because the programmer does not know
2880 whether the @code{uid_t} type is @code{short}, @code{int}, or
2881 @code{long}. Therefore, in cases like these GNU C allows a prototype
2882 to override a later old-style definition. More precisely, in GNU C, a
2883 function prototype argument type overrides the argument type specified
2884 by a later old-style definition if the former type is the same as the
2885 latter type before promotion. Thus in GNU C the above example is
2886 equivalent to the following:
2899 GNU C++ does not support old-style function definitions, so this
2900 extension is irrelevant.
2903 @section C++ Style Comments
2905 @cindex C++ comments
2906 @cindex comments, C++ style
2908 In GNU C, you may use C++ style comments, which start with @samp{//} and
2909 continue until the end of the line. Many other C implementations allow
2910 such comments, and they are included in the 1999 C standard. However,
2911 C++ style comments are not recognized if you specify an @option{-std}
2912 option specifying a version of ISO C before C99, or @option{-ansi}
2913 (equivalent to @option{-std=c89}).
2916 @section Dollar Signs in Identifier Names
2918 @cindex dollar signs in identifier names
2919 @cindex identifier names, dollar signs in
2921 In GNU C, you may normally use dollar signs in identifier names.
2922 This is because many traditional C implementations allow such identifiers.
2923 However, dollar signs in identifiers are not supported on a few target
2924 machines, typically because the target assembler does not allow them.
2926 @node Character Escapes
2927 @section The Character @key{ESC} in Constants
2929 You can use the sequence @samp{\e} in a string or character constant to
2930 stand for the ASCII character @key{ESC}.
2933 @section Inquiring on Alignment of Types or Variables
2935 @cindex type alignment
2936 @cindex variable alignment
2938 The keyword @code{__alignof__} allows you to inquire about how an object
2939 is aligned, or the minimum alignment usually required by a type. Its
2940 syntax is just like @code{sizeof}.
2942 For example, if the target machine requires a @code{double} value to be
2943 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2944 This is true on many RISC machines. On more traditional machine
2945 designs, @code{__alignof__ (double)} is 4 or even 2.
2947 Some machines never actually require alignment; they allow reference to any
2948 data type even at an odd address. For these machines, @code{__alignof__}
2949 reports the @emph{recommended} alignment of a type.
2951 If the operand of @code{__alignof__} is an lvalue rather than a type,
2952 its value is the required alignment for its type, taking into account
2953 any minimum alignment specified with GCC's @code{__attribute__}
2954 extension (@pxref{Variable Attributes}). For example, after this
2958 struct foo @{ int x; char y; @} foo1;
2962 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2963 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2965 It is an error to ask for the alignment of an incomplete type.
2967 @node Variable Attributes
2968 @section Specifying Attributes of Variables
2969 @cindex attribute of variables
2970 @cindex variable attributes
2972 The keyword @code{__attribute__} allows you to specify special
2973 attributes of variables or structure fields. This keyword is followed
2974 by an attribute specification inside double parentheses. Some
2975 attributes are currently defined generically for variables.
2976 Other attributes are defined for variables on particular target
2977 systems. Other attributes are available for functions
2978 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2979 Other front ends might define more attributes
2980 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2982 You may also specify attributes with @samp{__} preceding and following
2983 each keyword. This allows you to use them in header files without
2984 being concerned about a possible macro of the same name. For example,
2985 you may use @code{__aligned__} instead of @code{aligned}.
2987 @xref{Attribute Syntax}, for details of the exact syntax for using
2991 @cindex @code{aligned} attribute
2992 @item aligned (@var{alignment})
2993 This attribute specifies a minimum alignment for the variable or
2994 structure field, measured in bytes. For example, the declaration:
2997 int x __attribute__ ((aligned (16))) = 0;
3001 causes the compiler to allocate the global variable @code{x} on a
3002 16-byte boundary. On a 68040, this could be used in conjunction with
3003 an @code{asm} expression to access the @code{move16} instruction which
3004 requires 16-byte aligned operands.
3006 You can also specify the alignment of structure fields. For example, to
3007 create a double-word aligned @code{int} pair, you could write:
3010 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3014 This is an alternative to creating a union with a @code{double} member
3015 that forces the union to be double-word aligned.
3017 As in the preceding examples, you can explicitly specify the alignment
3018 (in bytes) that you wish the compiler to use for a given variable or
3019 structure field. Alternatively, you can leave out the alignment factor
3020 and just ask the compiler to align a variable or field to the maximum
3021 useful alignment for the target machine you are compiling for. For
3022 example, you could write:
3025 short array[3] __attribute__ ((aligned));
3028 Whenever you leave out the alignment factor in an @code{aligned} attribute
3029 specification, the compiler automatically sets the alignment for the declared
3030 variable or field to the largest alignment which is ever used for any data
3031 type on the target machine you are compiling for. Doing this can often make
3032 copy operations more efficient, because the compiler can use whatever
3033 instructions copy the biggest chunks of memory when performing copies to
3034 or from the variables or fields that you have aligned this way.
3036 The @code{aligned} attribute can only increase the alignment; but you
3037 can decrease it by specifying @code{packed} as well. See below.
3039 Note that the effectiveness of @code{aligned} attributes may be limited
3040 by inherent limitations in your linker. On many systems, the linker is
3041 only able to arrange for variables to be aligned up to a certain maximum
3042 alignment. (For some linkers, the maximum supported alignment may
3043 be very very small.) If your linker is only able to align variables
3044 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3045 in an @code{__attribute__} will still only provide you with 8 byte
3046 alignment. See your linker documentation for further information.
3048 @item cleanup (@var{cleanup_function})
3049 @cindex @code{cleanup} attribute
3050 The @code{cleanup} attribute runs a function when the variable goes
3051 out of scope. This attribute can only be applied to auto function
3052 scope variables; it may not be applied to parameters or variables
3053 with static storage duration. The function must take one parameter,
3054 a pointer to a type compatible with the variable. The return value
3055 of the function (if any) is ignored.
3057 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3058 will be run during the stack unwinding that happens during the
3059 processing of the exception. Note that the @code{cleanup} attribute
3060 does not allow the exception to be caught, only to perform an action.
3061 It is undefined what happens if @var{cleanup_function} does not
3066 @cindex @code{common} attribute
3067 @cindex @code{nocommon} attribute
3070 The @code{common} attribute requests GCC to place a variable in
3071 ``common'' storage. The @code{nocommon} attribute requests the
3072 opposite---to allocate space for it directly.
3074 These attributes override the default chosen by the
3075 @option{-fno-common} and @option{-fcommon} flags respectively.
3078 @cindex @code{deprecated} attribute
3079 The @code{deprecated} attribute results in a warning if the variable
3080 is used anywhere in the source file. This is useful when identifying
3081 variables that are expected to be removed in a future version of a
3082 program. The warning also includes the location of the declaration
3083 of the deprecated variable, to enable users to easily find further
3084 information about why the variable is deprecated, or what they should
3085 do instead. Note that the warning only occurs for uses:
3088 extern int old_var __attribute__ ((deprecated));
3090 int new_fn () @{ return old_var; @}
3093 results in a warning on line 3 but not line 2.
3095 The @code{deprecated} attribute can also be used for functions and
3096 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3098 @item mode (@var{mode})
3099 @cindex @code{mode} attribute
3100 This attribute specifies the data type for the declaration---whichever
3101 type corresponds to the mode @var{mode}. This in effect lets you
3102 request an integer or floating point type according to its width.
3104 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3105 indicate the mode corresponding to a one-byte integer, @samp{word} or
3106 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3107 or @samp{__pointer__} for the mode used to represent pointers.
3110 @cindex @code{packed} attribute
3111 The @code{packed} attribute specifies that a variable or structure field
3112 should have the smallest possible alignment---one byte for a variable,
3113 and one bit for a field, unless you specify a larger value with the
3114 @code{aligned} attribute.
3116 Here is a structure in which the field @code{x} is packed, so that it
3117 immediately follows @code{a}:
3123 int x[2] __attribute__ ((packed));
3127 @item section ("@var{section-name}")
3128 @cindex @code{section} variable attribute
3129 Normally, the compiler places the objects it generates in sections like
3130 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3131 or you need certain particular variables to appear in special sections,
3132 for example to map to special hardware. The @code{section}
3133 attribute specifies that a variable (or function) lives in a particular
3134 section. For example, this small program uses several specific section names:
3137 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3138 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3139 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3140 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3144 /* @r{Initialize stack pointer} */
3145 init_sp (stack + sizeof (stack));
3147 /* @r{Initialize initialized data} */
3148 memcpy (&init_data, &data, &edata - &data);
3150 /* @r{Turn on the serial ports} */
3157 Use the @code{section} attribute with an @emph{initialized} definition
3158 of a @emph{global} variable, as shown in the example. GCC issues
3159 a warning and otherwise ignores the @code{section} attribute in
3160 uninitialized variable declarations.
3162 You may only use the @code{section} attribute with a fully initialized
3163 global definition because of the way linkers work. The linker requires
3164 each object be defined once, with the exception that uninitialized
3165 variables tentatively go in the @code{common} (or @code{bss}) section
3166 and can be multiply ``defined''. You can force a variable to be
3167 initialized with the @option{-fno-common} flag or the @code{nocommon}
3170 Some file formats do not support arbitrary sections so the @code{section}
3171 attribute is not available on all platforms.
3172 If you need to map the entire contents of a module to a particular
3173 section, consider using the facilities of the linker instead.
3176 @cindex @code{shared} variable attribute
3177 On Microsoft Windows, in addition to putting variable definitions in a named
3178 section, the section can also be shared among all running copies of an
3179 executable or DLL@. For example, this small program defines shared data
3180 by putting it in a named section @code{shared} and marking the section
3184 int foo __attribute__((section ("shared"), shared)) = 0;
3189 /* @r{Read and write foo. All running
3190 copies see the same value.} */
3196 You may only use the @code{shared} attribute along with @code{section}
3197 attribute with a fully initialized global definition because of the way
3198 linkers work. See @code{section} attribute for more information.
3200 The @code{shared} attribute is only available on Microsoft Windows@.
3202 @item tls_model ("@var{tls_model}")
3203 @cindex @code{tls_model} attribute
3204 The @code{tls_model} attribute sets thread-local storage model
3205 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3206 overriding @option{-ftls-model=} command line switch on a per-variable
3208 The @var{tls_model} argument should be one of @code{global-dynamic},
3209 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3211 Not all targets support this attribute.
3214 This attribute, attached to a variable, means that the variable is meant
3215 to be possibly unused. GCC will not produce a warning for this
3219 This attribute, attached to a variable, means that the variable must be
3220 emitted even if it appears that the variable is not referenced.
3222 @item vector_size (@var{bytes})
3223 This attribute specifies the vector size for the variable, measured in
3224 bytes. For example, the declaration:
3227 int foo __attribute__ ((vector_size (16)));
3231 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3232 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3233 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3235 This attribute is only applicable to integral and float scalars,
3236 although arrays, pointers, and function return values are allowed in
3237 conjunction with this construct.
3239 Aggregates with this attribute are invalid, even if they are of the same
3240 size as a corresponding scalar. For example, the declaration:
3243 struct S @{ int a; @};
3244 struct S __attribute__ ((vector_size (16))) foo;
3248 is invalid even if the size of the structure is the same as the size of
3252 The @code{selectany} attribute causes an initialized global variable to
3253 have link-once semantics. When multiple definitions of the variable are
3254 encountered by the linker, the first is selected and the remainder are
3255 discarded. Following usage by the Microsoft compiler, the linker is told
3256 @emph{not} to warn about size or content differences of the multiple
3259 Although the primary usage of this attribute is for POD types, the
3260 attribute can also be applied to global C++ objects that are initialized
3261 by a constructor. In this case, the static initialization and destruction
3262 code for the object is emitted in each translation defining the object,
3263 but the calls to the constructor and destructor are protected by a
3264 link-once guard variable.
3266 The @code{selectany} attribute is only available on Microsoft Windows
3267 targets. You can use @code{__declspec (selectany)} as a synonym for
3268 @code{__attribute__ ((selectany))} for compatibility with other
3272 The @code{weak} attribute is described in @xref{Function Attributes}.
3275 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3278 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3282 @subsection M32R/D Variable Attributes
3284 One attribute is currently defined for the M32R/D@.
3287 @item model (@var{model-name})
3288 @cindex variable addressability on the M32R/D
3289 Use this attribute on the M32R/D to set the addressability of an object.
3290 The identifier @var{model-name} is one of @code{small}, @code{medium},
3291 or @code{large}, representing each of the code models.
3293 Small model objects live in the lower 16MB of memory (so that their
3294 addresses can be loaded with the @code{ld24} instruction).
3296 Medium and large model objects may live anywhere in the 32-bit address space
3297 (the compiler will generate @code{seth/add3} instructions to load their
3301 @anchor{i386 Variable Attributes}
3302 @subsection i386 Variable Attributes
3304 Two attributes are currently defined for i386 configurations:
3305 @code{ms_struct} and @code{gcc_struct}
3310 @cindex @code{ms_struct} attribute
3311 @cindex @code{gcc_struct} attribute
3313 If @code{packed} is used on a structure, or if bit-fields are used
3314 it may be that the Microsoft ABI packs them differently
3315 than GCC would normally pack them. Particularly when moving packed
3316 data between functions compiled with GCC and the native Microsoft compiler
3317 (either via function call or as data in a file), it may be necessary to access
3320 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3321 compilers to match the native Microsoft compiler.
3323 The Microsoft structure layout algorithm is fairly simple with the exception
3324 of the bitfield packing:
3326 The padding and alignment of members of structures and whether a bit field
3327 can straddle a storage-unit boundary
3330 @item Structure members are stored sequentially in the order in which they are
3331 declared: the first member has the lowest memory address and the last member
3334 @item Every data object has an alignment-requirement. The alignment-requirement
3335 for all data except structures, unions, and arrays is either the size of the
3336 object or the current packing size (specified with either the aligned attribute
3337 or the pack pragma), whichever is less. For structures, unions, and arrays,
3338 the alignment-requirement is the largest alignment-requirement of its members.
3339 Every object is allocated an offset so that:
3341 offset % alignment-requirement == 0
3343 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3344 unit if the integral types are the same size and if the next bit field fits
3345 into the current allocation unit without crossing the boundary imposed by the
3346 common alignment requirements of the bit fields.
3349 Handling of zero-length bitfields:
3351 MSVC interprets zero-length bitfields in the following ways:
3354 @item If a zero-length bitfield is inserted between two bitfields that would
3355 normally be coalesced, the bitfields will not be coalesced.
3362 unsigned long bf_1 : 12;
3364 unsigned long bf_2 : 12;
3368 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3369 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3371 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3372 alignment of the zero-length bitfield is greater than the member that follows it,
3373 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3393 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3394 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3395 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3398 Taking this into account, it is important to note the following:
3401 @item If a zero-length bitfield follows a normal bitfield, the type of the
3402 zero-length bitfield may affect the alignment of the structure as whole. For
3403 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3404 normal bitfield, and is of type short.
3406 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3407 still affect the alignment of the structure:
3417 Here, @code{t4} will take up 4 bytes.
3420 @item Zero-length bitfields following non-bitfield members are ignored:
3431 Here, @code{t5} will take up 2 bytes.
3435 @subsection PowerPC Variable Attributes
3437 Three attributes currently are defined for PowerPC configurations:
3438 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3440 For full documentation of the struct attributes please see the
3441 documentation in the @xref{i386 Variable Attributes}, section.
3443 For documentation of @code{altivec} attribute please see the
3444 documentation in the @xref{PowerPC Type Attributes}, section.
3446 @subsection Xstormy16 Variable Attributes
3448 One attribute is currently defined for xstormy16 configurations:
3453 @cindex @code{below100} attribute
3455 If a variable has the @code{below100} attribute (@code{BELOW100} is
3456 allowed also), GCC will place the variable in the first 0x100 bytes of
3457 memory and use special opcodes to access it. Such variables will be
3458 placed in either the @code{.bss_below100} section or the
3459 @code{.data_below100} section.
3463 @node Type Attributes
3464 @section Specifying Attributes of Types
3465 @cindex attribute of types
3466 @cindex type attributes
3468 The keyword @code{__attribute__} allows you to specify special
3469 attributes of @code{struct} and @code{union} types when you define
3470 such types. This keyword is followed by an attribute specification
3471 inside double parentheses. Seven attributes are currently defined for
3472 types: @code{aligned}, @code{packed}, @code{transparent_union},
3473 @code{unused}, @code{deprecated}, @code{visibility}, and
3474 @code{may_alias}. Other attributes are defined for functions
3475 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3478 You may also specify any one of these attributes with @samp{__}
3479 preceding and following its keyword. This allows you to use these
3480 attributes in header files without being concerned about a possible
3481 macro of the same name. For example, you may use @code{__aligned__}
3482 instead of @code{aligned}.
3484 You may specify type attributes either in a @code{typedef} declaration
3485 or in an enum, struct or union type declaration or definition.
3487 For an enum, struct or union type, you may specify attributes either
3488 between the enum, struct or union tag and the name of the type, or
3489 just past the closing curly brace of the @emph{definition}. The
3490 former syntax is preferred.
3492 @xref{Attribute Syntax}, for details of the exact syntax for using
3496 @cindex @code{aligned} attribute
3497 @item aligned (@var{alignment})
3498 This attribute specifies a minimum alignment (in bytes) for variables
3499 of the specified type. For example, the declarations:
3502 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3503 typedef int more_aligned_int __attribute__ ((aligned (8)));
3507 force the compiler to insure (as far as it can) that each variable whose
3508 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3509 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3510 variables of type @code{struct S} aligned to 8-byte boundaries allows
3511 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3512 store) instructions when copying one variable of type @code{struct S} to
3513 another, thus improving run-time efficiency.
3515 Note that the alignment of any given @code{struct} or @code{union} type
3516 is required by the ISO C standard to be at least a perfect multiple of
3517 the lowest common multiple of the alignments of all of the members of
3518 the @code{struct} or @code{union} in question. This means that you @emph{can}
3519 effectively adjust the alignment of a @code{struct} or @code{union}
3520 type by attaching an @code{aligned} attribute to any one of the members
3521 of such a type, but the notation illustrated in the example above is a
3522 more obvious, intuitive, and readable way to request the compiler to
3523 adjust the alignment of an entire @code{struct} or @code{union} type.
3525 As in the preceding example, you can explicitly specify the alignment
3526 (in bytes) that you wish the compiler to use for a given @code{struct}
3527 or @code{union} type. Alternatively, you can leave out the alignment factor
3528 and just ask the compiler to align a type to the maximum
3529 useful alignment for the target machine you are compiling for. For
3530 example, you could write:
3533 struct S @{ short f[3]; @} __attribute__ ((aligned));
3536 Whenever you leave out the alignment factor in an @code{aligned}
3537 attribute specification, the compiler automatically sets the alignment
3538 for the type to the largest alignment which is ever used for any data
3539 type on the target machine you are compiling for. Doing this can often
3540 make copy operations more efficient, because the compiler can use
3541 whatever instructions copy the biggest chunks of memory when performing
3542 copies to or from the variables which have types that you have aligned
3545 In the example above, if the size of each @code{short} is 2 bytes, then
3546 the size of the entire @code{struct S} type is 6 bytes. The smallest
3547 power of two which is greater than or equal to that is 8, so the
3548 compiler sets the alignment for the entire @code{struct S} type to 8
3551 Note that although you can ask the compiler to select a time-efficient
3552 alignment for a given type and then declare only individual stand-alone
3553 objects of that type, the compiler's ability to select a time-efficient
3554 alignment is primarily useful only when you plan to create arrays of
3555 variables having the relevant (efficiently aligned) type. If you
3556 declare or use arrays of variables of an efficiently-aligned type, then
3557 it is likely that your program will also be doing pointer arithmetic (or
3558 subscripting, which amounts to the same thing) on pointers to the
3559 relevant type, and the code that the compiler generates for these
3560 pointer arithmetic operations will often be more efficient for
3561 efficiently-aligned types than for other types.
3563 The @code{aligned} attribute can only increase the alignment; but you
3564 can decrease it by specifying @code{packed} as well. See below.
3566 Note that the effectiveness of @code{aligned} attributes may be limited
3567 by inherent limitations in your linker. On many systems, the linker is
3568 only able to arrange for variables to be aligned up to a certain maximum
3569 alignment. (For some linkers, the maximum supported alignment may
3570 be very very small.) If your linker is only able to align variables
3571 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3572 in an @code{__attribute__} will still only provide you with 8 byte
3573 alignment. See your linker documentation for further information.
3576 This attribute, attached to @code{struct} or @code{union} type
3577 definition, specifies that each member (other than zero-width bitfields)
3578 of the structure or union is placed to minimize the memory required. When
3579 attached to an @code{enum} definition, it indicates that the smallest
3580 integral type should be used.
3582 @opindex fshort-enums
3583 Specifying this attribute for @code{struct} and @code{union} types is
3584 equivalent to specifying the @code{packed} attribute on each of the
3585 structure or union members. Specifying the @option{-fshort-enums}
3586 flag on the line is equivalent to specifying the @code{packed}
3587 attribute on all @code{enum} definitions.
3589 In the following example @code{struct my_packed_struct}'s members are
3590 packed closely together, but the internal layout of its @code{s} member
3591 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3595 struct my_unpacked_struct
3601 struct __attribute__ ((__packed__)) my_packed_struct
3605 struct my_unpacked_struct s;
3609 You may only specify this attribute on the definition of a @code{enum},
3610 @code{struct} or @code{union}, not on a @code{typedef} which does not
3611 also define the enumerated type, structure or union.
3613 @item transparent_union
3614 This attribute, attached to a @code{union} type definition, indicates
3615 that any function parameter having that union type causes calls to that
3616 function to be treated in a special way.
3618 First, the argument corresponding to a transparent union type can be of
3619 any type in the union; no cast is required. Also, if the union contains
3620 a pointer type, the corresponding argument can be a null pointer
3621 constant or a void pointer expression; and if the union contains a void
3622 pointer type, the corresponding argument can be any pointer expression.
3623 If the union member type is a pointer, qualifiers like @code{const} on
3624 the referenced type must be respected, just as with normal pointer
3627 Second, the argument is passed to the function using the calling
3628 conventions of the first member of the transparent union, not the calling
3629 conventions of the union itself. All members of the union must have the
3630 same machine representation; this is necessary for this argument passing
3633 Transparent unions are designed for library functions that have multiple
3634 interfaces for compatibility reasons. For example, suppose the
3635 @code{wait} function must accept either a value of type @code{int *} to
3636 comply with Posix, or a value of type @code{union wait *} to comply with
3637 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3638 @code{wait} would accept both kinds of arguments, but it would also
3639 accept any other pointer type and this would make argument type checking
3640 less useful. Instead, @code{<sys/wait.h>} might define the interface
3648 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3650 pid_t wait (wait_status_ptr_t);
3653 This interface allows either @code{int *} or @code{union wait *}
3654 arguments to be passed, using the @code{int *} calling convention.
3655 The program can call @code{wait} with arguments of either type:
3658 int w1 () @{ int w; return wait (&w); @}
3659 int w2 () @{ union wait w; return wait (&w); @}
3662 With this interface, @code{wait}'s implementation might look like this:
3665 pid_t wait (wait_status_ptr_t p)
3667 return waitpid (-1, p.__ip, 0);
3672 When attached to a type (including a @code{union} or a @code{struct}),
3673 this attribute means that variables of that type are meant to appear
3674 possibly unused. GCC will not produce a warning for any variables of
3675 that type, even if the variable appears to do nothing. This is often
3676 the case with lock or thread classes, which are usually defined and then
3677 not referenced, but contain constructors and destructors that have
3678 nontrivial bookkeeping functions.
3681 The @code{deprecated} attribute results in a warning if the type
3682 is used anywhere in the source file. This is useful when identifying
3683 types that are expected to be removed in a future version of a program.
3684 If possible, the warning also includes the location of the declaration
3685 of the deprecated type, to enable users to easily find further
3686 information about why the type is deprecated, or what they should do
3687 instead. Note that the warnings only occur for uses and then only
3688 if the type is being applied to an identifier that itself is not being
3689 declared as deprecated.
3692 typedef int T1 __attribute__ ((deprecated));
3696 typedef T1 T3 __attribute__ ((deprecated));
3697 T3 z __attribute__ ((deprecated));
3700 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3701 warning is issued for line 4 because T2 is not explicitly
3702 deprecated. Line 5 has no warning because T3 is explicitly
3703 deprecated. Similarly for line 6.
3705 The @code{deprecated} attribute can also be used for functions and
3706 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3709 Accesses to objects with types with this attribute are not subjected to
3710 type-based alias analysis, but are instead assumed to be able to alias
3711 any other type of objects, just like the @code{char} type. See
3712 @option{-fstrict-aliasing} for more information on aliasing issues.
3717 typedef short __attribute__((__may_alias__)) short_a;
3723 short_a *b = (short_a *) &a;
3727 if (a == 0x12345678)
3734 If you replaced @code{short_a} with @code{short} in the variable
3735 declaration, the above program would abort when compiled with
3736 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3737 above in recent GCC versions.
3740 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3741 applied to class, struct, union and enum types. Unlike other type
3742 attributes, the attribute must appear between the initial keyword and
3743 the name of the type; it cannot appear after the body of the type.
3745 Note that the type visibility is applied to vague linkage entities
3746 associated with the class (vtable, typeinfo node, etc.). In
3747 particular, if a class is thrown as an exception in one shared object
3748 and caught in another, the class must have default visibility.
3749 Otherwise the two shared objects will be unable to use the same
3750 typeinfo node and exception handling will break.
3752 @subsection ARM Type Attributes
3754 On those ARM targets that support @code{dllimport} (such as Symbian
3755 OS), you can use the @code{notshared} attribute to indicate that the
3756 virtual table and other similar data for a class should not be
3757 exported from a DLL@. For example:
3760 class __declspec(notshared) C @{
3762 __declspec(dllimport) C();
3766 __declspec(dllexport)
3770 In this code, @code{C::C} is exported from the current DLL, but the
3771 virtual table for @code{C} is not exported. (You can use
3772 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3773 most Symbian OS code uses @code{__declspec}.)
3775 @anchor{i386 Type Attributes}
3776 @subsection i386 Type Attributes
3778 Two attributes are currently defined for i386 configurations:
3779 @code{ms_struct} and @code{gcc_struct}
3783 @cindex @code{ms_struct}
3784 @cindex @code{gcc_struct}
3786 If @code{packed} is used on a structure, or if bit-fields are used
3787 it may be that the Microsoft ABI packs them differently
3788 than GCC would normally pack them. Particularly when moving packed
3789 data between functions compiled with GCC and the native Microsoft compiler
3790 (either via function call or as data in a file), it may be necessary to access
3793 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3794 compilers to match the native Microsoft compiler.
3797 To specify multiple attributes, separate them by commas within the
3798 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3801 @anchor{PowerPC Type Attributes}
3802 @subsection PowerPC Type Attributes
3804 Three attributes currently are defined for PowerPC configurations:
3805 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3807 For full documentation of the struct attributes please see the
3808 documentation in the @xref{i386 Type Attributes}, section.
3810 The @code{altivec} attribute allows one to declare AltiVec vector data
3811 types supported by the AltiVec Programming Interface Manual. The
3812 attribute requires an argument to specify one of three vector types:
3813 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3814 and @code{bool__} (always followed by unsigned).
3817 __attribute__((altivec(vector__)))
3818 __attribute__((altivec(pixel__))) unsigned short
3819 __attribute__((altivec(bool__))) unsigned
3822 These attributes mainly are intended to support the @code{__vector},
3823 @code{__pixel}, and @code{__bool} AltiVec keywords.
3826 @section An Inline Function is As Fast As a Macro
3827 @cindex inline functions
3828 @cindex integrating function code
3830 @cindex macros, inline alternative
3832 By declaring a function inline, you can direct GCC to make
3833 calls to that function faster. One way GCC can achieve this is to
3834 integrate that function's code into the code for its callers. This
3835 makes execution faster by eliminating the function-call overhead; in
3836 addition, if any of the actual argument values are constant, their
3837 known values may permit simplifications at compile time so that not
3838 all of the inline function's code needs to be included. The effect on
3839 code size is less predictable; object code may be larger or smaller
3840 with function inlining, depending on the particular case. You can
3841 also direct GCC to try to integrate all ``simple enough'' functions
3842 into their callers with the option @option{-finline-functions}.
3844 GCC implements three different semantics of declaring a function
3845 inline. One is available with @option{-std=gnu89}, another when
3846 @option{-std=c99} or @option{-std=gnu99}, and the third is used when
3849 To declare a function inline, use the @code{inline} keyword in its
3850 declaration, like this:
3860 If you are writing a header file to be included in ISO C89 programs, write
3861 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3863 The three types of inlining behave similarly in two important cases:
3864 when the @code{inline} keyword is used on a @code{static} function,
3865 like the example above, and when a function is first declared without
3866 using the @code{inline} keyword and then is defined with
3867 @code{inline}, like this:
3870 extern int inc (int *a);
3878 In both of these common cases, the program behaves the same as if you
3879 had not used the @code{inline} keyword, except for its speed.
3881 @cindex inline functions, omission of
3882 @opindex fkeep-inline-functions
3883 When a function is both inline and @code{static}, if all calls to the
3884 function are integrated into the caller, and the function's address is
3885 never used, then the function's own assembler code is never referenced.
3886 In this case, GCC does not actually output assembler code for the
3887 function, unless you specify the option @option{-fkeep-inline-functions}.
3888 Some calls cannot be integrated for various reasons (in particular,
3889 calls that precede the function's definition cannot be integrated, and
3890 neither can recursive calls within the definition). If there is a
3891 nonintegrated call, then the function is compiled to assembler code as
3892 usual. The function must also be compiled as usual if the program
3893 refers to its address, because that can't be inlined.
3895 @cindex automatic @code{inline} for C++ member fns
3896 @cindex @code{inline} automatic for C++ member fns
3897 @cindex member fns, automatically @code{inline}
3898 @cindex C++ member fns, automatically @code{inline}
3899 @opindex fno-default-inline
3900 As required by ISO C++, GCC considers member functions defined within
3901 the body of a class to be marked inline even if they are
3902 not explicitly declared with the @code{inline} keyword. You can
3903 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
3904 Options,,Options Controlling C++ Dialect}.
3906 GCC does not inline any functions when not optimizing unless you specify
3907 the @samp{always_inline} attribute for the function, like this:
3910 /* @r{Prototype.} */
3911 inline void foo (const char) __attribute__((always_inline));
3914 The remainder of this section is specific to GNU C89 inlining.
3916 @cindex non-static inline function
3917 When an inline function is not @code{static}, then the compiler must assume
3918 that there may be calls from other source files; since a global symbol can
3919 be defined only once in any program, the function must not be defined in
3920 the other source files, so the calls therein cannot be integrated.
3921 Therefore, a non-@code{static} inline function is always compiled on its
3922 own in the usual fashion.
3924 If you specify both @code{inline} and @code{extern} in the function
3925 definition, then the definition is used only for inlining. In no case
3926 is the function compiled on its own, not even if you refer to its
3927 address explicitly. Such an address becomes an external reference, as
3928 if you had only declared the function, and had not defined it.
3930 This combination of @code{inline} and @code{extern} has almost the
3931 effect of a macro. The way to use it is to put a function definition in
3932 a header file with these keywords, and put another copy of the
3933 definition (lacking @code{inline} and @code{extern}) in a library file.
3934 The definition in the header file will cause most calls to the function
3935 to be inlined. If any uses of the function remain, they will refer to
3936 the single copy in the library.
3939 @section Assembler Instructions with C Expression Operands
3940 @cindex extended @code{asm}
3941 @cindex @code{asm} expressions
3942 @cindex assembler instructions
3945 In an assembler instruction using @code{asm}, you can specify the
3946 operands of the instruction using C expressions. This means you need not
3947 guess which registers or memory locations will contain the data you want
3950 You must specify an assembler instruction template much like what
3951 appears in a machine description, plus an operand constraint string for
3954 For example, here is how to use the 68881's @code{fsinx} instruction:
3957 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3961 Here @code{angle} is the C expression for the input operand while
3962 @code{result} is that of the output operand. Each has @samp{"f"} as its
3963 operand constraint, saying that a floating point register is required.
3964 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3965 output operands' constraints must use @samp{=}. The constraints use the
3966 same language used in the machine description (@pxref{Constraints}).
3968 Each operand is described by an operand-constraint string followed by
3969 the C expression in parentheses. A colon separates the assembler
3970 template from the first output operand and another separates the last
3971 output operand from the first input, if any. Commas separate the
3972 operands within each group. The total number of operands is currently
3973 limited to 30; this limitation may be lifted in some future version of
3976 If there are no output operands but there are input operands, you must
3977 place two consecutive colons surrounding the place where the output
3980 As of GCC version 3.1, it is also possible to specify input and output
3981 operands using symbolic names which can be referenced within the
3982 assembler code. These names are specified inside square brackets
3983 preceding the constraint string, and can be referenced inside the
3984 assembler code using @code{%[@var{name}]} instead of a percentage sign
3985 followed by the operand number. Using named operands the above example
3989 asm ("fsinx %[angle],%[output]"
3990 : [output] "=f" (result)
3991 : [angle] "f" (angle));
3995 Note that the symbolic operand names have no relation whatsoever to
3996 other C identifiers. You may use any name you like, even those of
3997 existing C symbols, but you must ensure that no two operands within the same
3998 assembler construct use the same symbolic name.
4000 Output operand expressions must be lvalues; the compiler can check this.
4001 The input operands need not be lvalues. The compiler cannot check
4002 whether the operands have data types that are reasonable for the
4003 instruction being executed. It does not parse the assembler instruction
4004 template and does not know what it means or even whether it is valid
4005 assembler input. The extended @code{asm} feature is most often used for
4006 machine instructions the compiler itself does not know exist. If
4007 the output expression cannot be directly addressed (for example, it is a
4008 bit-field), your constraint must allow a register. In that case, GCC
4009 will use the register as the output of the @code{asm}, and then store
4010 that register into the output.
4012 The ordinary output operands must be write-only; GCC will assume that
4013 the values in these operands before the instruction are dead and need
4014 not be generated. Extended asm supports input-output or read-write
4015 operands. Use the constraint character @samp{+} to indicate such an
4016 operand and list it with the output operands. You should only use
4017 read-write operands when the constraints for the operand (or the
4018 operand in which only some of the bits are to be changed) allow a
4021 You may, as an alternative, logically split its function into two
4022 separate operands, one input operand and one write-only output
4023 operand. The connection between them is expressed by constraints
4024 which say they need to be in the same location when the instruction
4025 executes. You can use the same C expression for both operands, or
4026 different expressions. For example, here we write the (fictitious)
4027 @samp{combine} instruction with @code{bar} as its read-only source
4028 operand and @code{foo} as its read-write destination:
4031 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4035 The constraint @samp{"0"} for operand 1 says that it must occupy the
4036 same location as operand 0. A number in constraint is allowed only in
4037 an input operand and it must refer to an output operand.
4039 Only a number in the constraint can guarantee that one operand will be in
4040 the same place as another. The mere fact that @code{foo} is the value
4041 of both operands is not enough to guarantee that they will be in the
4042 same place in the generated assembler code. The following would not
4046 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4049 Various optimizations or reloading could cause operands 0 and 1 to be in
4050 different registers; GCC knows no reason not to do so. For example, the
4051 compiler might find a copy of the value of @code{foo} in one register and
4052 use it for operand 1, but generate the output operand 0 in a different
4053 register (copying it afterward to @code{foo}'s own address). Of course,
4054 since the register for operand 1 is not even mentioned in the assembler
4055 code, the result will not work, but GCC can't tell that.
4057 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4058 the operand number for a matching constraint. For example:
4061 asm ("cmoveq %1,%2,%[result]"
4062 : [result] "=r"(result)
4063 : "r" (test), "r"(new), "[result]"(old));
4066 Sometimes you need to make an @code{asm} operand be a specific register,
4067 but there's no matching constraint letter for that register @emph{by
4068 itself}. To force the operand into that register, use a local variable
4069 for the operand and specify the register in the variable declaration.
4070 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4071 register constraint letter that matches the register:
4074 register int *p1 asm ("r0") = @dots{};
4075 register int *p2 asm ("r1") = @dots{};
4076 register int *result asm ("r0");
4077 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4080 @anchor{Example of asm with clobbered asm reg}
4081 In the above example, beware that a register that is call-clobbered by
4082 the target ABI will be overwritten by any function call in the
4083 assignment, including library calls for arithmetic operators.
4084 Assuming it is a call-clobbered register, this may happen to @code{r0}
4085 above by the assignment to @code{p2}. If you have to use such a
4086 register, use temporary variables for expressions between the register
4091 register int *p1 asm ("r0") = @dots{};
4092 register int *p2 asm ("r1") = t1;
4093 register int *result asm ("r0");
4094 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4097 Some instructions clobber specific hard registers. To describe this,
4098 write a third colon after the input operands, followed by the names of
4099 the clobbered hard registers (given as strings). Here is a realistic
4100 example for the VAX:
4103 asm volatile ("movc3 %0,%1,%2"
4104 : /* @r{no outputs} */
4105 : "g" (from), "g" (to), "g" (count)
4106 : "r0", "r1", "r2", "r3", "r4", "r5");
4109 You may not write a clobber description in a way that overlaps with an
4110 input or output operand. For example, you may not have an operand
4111 describing a register class with one member if you mention that register
4112 in the clobber list. Variables declared to live in specific registers
4113 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4114 have no part mentioned in the clobber description.
4115 There is no way for you to specify that an input
4116 operand is modified without also specifying it as an output
4117 operand. Note that if all the output operands you specify are for this
4118 purpose (and hence unused), you will then also need to specify
4119 @code{volatile} for the @code{asm} construct, as described below, to
4120 prevent GCC from deleting the @code{asm} statement as unused.
4122 If you refer to a particular hardware register from the assembler code,
4123 you will probably have to list the register after the third colon to
4124 tell the compiler the register's value is modified. In some assemblers,
4125 the register names begin with @samp{%}; to produce one @samp{%} in the
4126 assembler code, you must write @samp{%%} in the input.
4128 If your assembler instruction can alter the condition code register, add
4129 @samp{cc} to the list of clobbered registers. GCC on some machines
4130 represents the condition codes as a specific hardware register;
4131 @samp{cc} serves to name this register. On other machines, the
4132 condition code is handled differently, and specifying @samp{cc} has no
4133 effect. But it is valid no matter what the machine.
4135 If your assembler instructions access memory in an unpredictable
4136 fashion, add @samp{memory} to the list of clobbered registers. This
4137 will cause GCC to not keep memory values cached in registers across the
4138 assembler instruction and not optimize stores or loads to that memory.
4139 You will also want to add the @code{volatile} keyword if the memory
4140 affected is not listed in the inputs or outputs of the @code{asm}, as
4141 the @samp{memory} clobber does not count as a side-effect of the
4142 @code{asm}. If you know how large the accessed memory is, you can add
4143 it as input or output but if this is not known, you should add
4144 @samp{memory}. As an example, if you access ten bytes of a string, you
4145 can use a memory input like:
4148 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4151 Note that in the following example the memory input is necessary,
4152 otherwise GCC might optimize the store to @code{x} away:
4159 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4160 "=&d" (r) : "a" (y), "m" (*y));
4165 You can put multiple assembler instructions together in a single
4166 @code{asm} template, separated by the characters normally used in assembly
4167 code for the system. A combination that works in most places is a newline
4168 to break the line, plus a tab character to move to the instruction field
4169 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4170 assembler allows semicolons as a line-breaking character. Note that some
4171 assembler dialects use semicolons to start a comment.
4172 The input operands are guaranteed not to use any of the clobbered
4173 registers, and neither will the output operands' addresses, so you can
4174 read and write the clobbered registers as many times as you like. Here
4175 is an example of multiple instructions in a template; it assumes the
4176 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4179 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4181 : "g" (from), "g" (to)
4185 Unless an output operand has the @samp{&} constraint modifier, GCC
4186 may allocate it in the same register as an unrelated input operand, on
4187 the assumption the inputs are consumed before the outputs are produced.
4188 This assumption may be false if the assembler code actually consists of
4189 more than one instruction. In such a case, use @samp{&} for each output
4190 operand that may not overlap an input. @xref{Modifiers}.
4192 If you want to test the condition code produced by an assembler
4193 instruction, you must include a branch and a label in the @code{asm}
4194 construct, as follows:
4197 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4203 This assumes your assembler supports local labels, as the GNU assembler
4204 and most Unix assemblers do.
4206 Speaking of labels, jumps from one @code{asm} to another are not
4207 supported. The compiler's optimizers do not know about these jumps, and
4208 therefore they cannot take account of them when deciding how to
4211 @cindex macros containing @code{asm}
4212 Usually the most convenient way to use these @code{asm} instructions is to
4213 encapsulate them in macros that look like functions. For example,
4217 (@{ double __value, __arg = (x); \
4218 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4223 Here the variable @code{__arg} is used to make sure that the instruction
4224 operates on a proper @code{double} value, and to accept only those
4225 arguments @code{x} which can convert automatically to a @code{double}.
4227 Another way to make sure the instruction operates on the correct data
4228 type is to use a cast in the @code{asm}. This is different from using a
4229 variable @code{__arg} in that it converts more different types. For
4230 example, if the desired type were @code{int}, casting the argument to
4231 @code{int} would accept a pointer with no complaint, while assigning the
4232 argument to an @code{int} variable named @code{__arg} would warn about
4233 using a pointer unless the caller explicitly casts it.
4235 If an @code{asm} has output operands, GCC assumes for optimization
4236 purposes the instruction has no side effects except to change the output
4237 operands. This does not mean instructions with a side effect cannot be
4238 used, but you must be careful, because the compiler may eliminate them
4239 if the output operands aren't used, or move them out of loops, or
4240 replace two with one if they constitute a common subexpression. Also,
4241 if your instruction does have a side effect on a variable that otherwise
4242 appears not to change, the old value of the variable may be reused later
4243 if it happens to be found in a register.
4245 You can prevent an @code{asm} instruction from being deleted
4246 by writing the keyword @code{volatile} after
4247 the @code{asm}. For example:
4250 #define get_and_set_priority(new) \
4252 asm volatile ("get_and_set_priority %0, %1" \
4253 : "=g" (__old) : "g" (new)); \
4258 The @code{volatile} keyword indicates that the instruction has
4259 important side-effects. GCC will not delete a volatile @code{asm} if
4260 it is reachable. (The instruction can still be deleted if GCC can
4261 prove that control-flow will never reach the location of the
4262 instruction.) Note that even a volatile @code{asm} instruction
4263 can be moved relative to other code, including across jump
4264 instructions. For example, on many targets there is a system
4265 register which can be set to control the rounding mode of
4266 floating point operations. You might try
4267 setting it with a volatile @code{asm}, like this PowerPC example:
4270 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4275 This will not work reliably, as the compiler may move the addition back
4276 before the volatile @code{asm}. To make it work you need to add an
4277 artificial dependency to the @code{asm} referencing a variable in the code
4278 you don't want moved, for example:
4281 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4285 Similarly, you can't expect a
4286 sequence of volatile @code{asm} instructions to remain perfectly
4287 consecutive. If you want consecutive output, use a single @code{asm}.
4288 Also, GCC will perform some optimizations across a volatile @code{asm}
4289 instruction; GCC does not ``forget everything'' when it encounters
4290 a volatile @code{asm} instruction the way some other compilers do.
4292 An @code{asm} instruction without any output operands will be treated
4293 identically to a volatile @code{asm} instruction.
4295 It is a natural idea to look for a way to give access to the condition
4296 code left by the assembler instruction. However, when we attempted to
4297 implement this, we found no way to make it work reliably. The problem
4298 is that output operands might need reloading, which would result in
4299 additional following ``store'' instructions. On most machines, these
4300 instructions would alter the condition code before there was time to
4301 test it. This problem doesn't arise for ordinary ``test'' and
4302 ``compare'' instructions because they don't have any output operands.
4304 For reasons similar to those described above, it is not possible to give
4305 an assembler instruction access to the condition code left by previous
4308 If you are writing a header file that should be includable in ISO C
4309 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4312 @subsection Size of an @code{asm}
4314 Some targets require that GCC track the size of each instruction used in
4315 order to generate correct code. Because the final length of an
4316 @code{asm} is only known by the assembler, GCC must make an estimate as
4317 to how big it will be. The estimate is formed by counting the number of
4318 statements in the pattern of the @code{asm} and multiplying that by the
4319 length of the longest instruction on that processor. Statements in the
4320 @code{asm} are identified by newline characters and whatever statement
4321 separator characters are supported by the assembler; on most processors
4322 this is the `@code{;}' character.
4324 Normally, GCC's estimate is perfectly adequate to ensure that correct
4325 code is generated, but it is possible to confuse the compiler if you use
4326 pseudo instructions or assembler macros that expand into multiple real
4327 instructions or if you use assembler directives that expand to more
4328 space in the object file than would be needed for a single instruction.
4329 If this happens then the assembler will produce a diagnostic saying that
4330 a label is unreachable.
4332 @subsection i386 floating point asm operands
4334 There are several rules on the usage of stack-like regs in
4335 asm_operands insns. These rules apply only to the operands that are
4340 Given a set of input regs that die in an asm_operands, it is
4341 necessary to know which are implicitly popped by the asm, and
4342 which must be explicitly popped by gcc.
4344 An input reg that is implicitly popped by the asm must be
4345 explicitly clobbered, unless it is constrained to match an
4349 For any input reg that is implicitly popped by an asm, it is
4350 necessary to know how to adjust the stack to compensate for the pop.
4351 If any non-popped input is closer to the top of the reg-stack than
4352 the implicitly popped reg, it would not be possible to know what the
4353 stack looked like---it's not clear how the rest of the stack ``slides
4356 All implicitly popped input regs must be closer to the top of
4357 the reg-stack than any input that is not implicitly popped.
4359 It is possible that if an input dies in an insn, reload might
4360 use the input reg for an output reload. Consider this example:
4363 asm ("foo" : "=t" (a) : "f" (b));
4366 This asm says that input B is not popped by the asm, and that
4367 the asm pushes a result onto the reg-stack, i.e., the stack is one
4368 deeper after the asm than it was before. But, it is possible that
4369 reload will think that it can use the same reg for both the input and
4370 the output, if input B dies in this insn.
4372 If any input operand uses the @code{f} constraint, all output reg
4373 constraints must use the @code{&} earlyclobber.
4375 The asm above would be written as
4378 asm ("foo" : "=&t" (a) : "f" (b));
4382 Some operands need to be in particular places on the stack. All
4383 output operands fall in this category---there is no other way to
4384 know which regs the outputs appear in unless the user indicates
4385 this in the constraints.
4387 Output operands must specifically indicate which reg an output
4388 appears in after an asm. @code{=f} is not allowed: the operand
4389 constraints must select a class with a single reg.
4392 Output operands may not be ``inserted'' between existing stack regs.
4393 Since no 387 opcode uses a read/write operand, all output operands
4394 are dead before the asm_operands, and are pushed by the asm_operands.
4395 It makes no sense to push anywhere but the top of the reg-stack.
4397 Output operands must start at the top of the reg-stack: output
4398 operands may not ``skip'' a reg.
4401 Some asm statements may need extra stack space for internal
4402 calculations. This can be guaranteed by clobbering stack registers
4403 unrelated to the inputs and outputs.
4407 Here are a couple of reasonable asms to want to write. This asm
4408 takes one input, which is internally popped, and produces two outputs.
4411 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4414 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4415 and replaces them with one output. The user must code the @code{st(1)}
4416 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4419 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4425 @section Controlling Names Used in Assembler Code
4426 @cindex assembler names for identifiers
4427 @cindex names used in assembler code
4428 @cindex identifiers, names in assembler code
4430 You can specify the name to be used in the assembler code for a C
4431 function or variable by writing the @code{asm} (or @code{__asm__})
4432 keyword after the declarator as follows:
4435 int foo asm ("myfoo") = 2;
4439 This specifies that the name to be used for the variable @code{foo} in
4440 the assembler code should be @samp{myfoo} rather than the usual
4443 On systems where an underscore is normally prepended to the name of a C
4444 function or variable, this feature allows you to define names for the
4445 linker that do not start with an underscore.
4447 It does not make sense to use this feature with a non-static local
4448 variable since such variables do not have assembler names. If you are
4449 trying to put the variable in a particular register, see @ref{Explicit
4450 Reg Vars}. GCC presently accepts such code with a warning, but will
4451 probably be changed to issue an error, rather than a warning, in the
4454 You cannot use @code{asm} in this way in a function @emph{definition}; but
4455 you can get the same effect by writing a declaration for the function
4456 before its definition and putting @code{asm} there, like this:
4459 extern func () asm ("FUNC");
4466 It is up to you to make sure that the assembler names you choose do not
4467 conflict with any other assembler symbols. Also, you must not use a
4468 register name; that would produce completely invalid assembler code. GCC
4469 does not as yet have the ability to store static variables in registers.
4470 Perhaps that will be added.
4472 @node Explicit Reg Vars
4473 @section Variables in Specified Registers
4474 @cindex explicit register variables
4475 @cindex variables in specified registers
4476 @cindex specified registers
4477 @cindex registers, global allocation
4479 GNU C allows you to put a few global variables into specified hardware
4480 registers. You can also specify the register in which an ordinary
4481 register variable should be allocated.
4485 Global register variables reserve registers throughout the program.
4486 This may be useful in programs such as programming language
4487 interpreters which have a couple of global variables that are accessed
4491 Local register variables in specific registers do not reserve the
4492 registers, except at the point where they are used as input or output
4493 operands in an @code{asm} statement and the @code{asm} statement itself is
4494 not deleted. The compiler's data flow analysis is capable of determining
4495 where the specified registers contain live values, and where they are
4496 available for other uses. Stores into local register variables may be deleted
4497 when they appear to be dead according to dataflow analysis. References
4498 to local register variables may be deleted or moved or simplified.
4500 These local variables are sometimes convenient for use with the extended
4501 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4502 output of the assembler instruction directly into a particular register.
4503 (This will work provided the register you specify fits the constraints
4504 specified for that operand in the @code{asm}.)
4512 @node Global Reg Vars
4513 @subsection Defining Global Register Variables
4514 @cindex global register variables
4515 @cindex registers, global variables in
4517 You can define a global register variable in GNU C like this:
4520 register int *foo asm ("a5");
4524 Here @code{a5} is the name of the register which should be used. Choose a
4525 register which is normally saved and restored by function calls on your
4526 machine, so that library routines will not clobber it.
4528 Naturally the register name is cpu-dependent, so you would need to
4529 conditionalize your program according to cpu type. The register
4530 @code{a5} would be a good choice on a 68000 for a variable of pointer
4531 type. On machines with register windows, be sure to choose a ``global''
4532 register that is not affected magically by the function call mechanism.
4534 In addition, operating systems on one type of cpu may differ in how they
4535 name the registers; then you would need additional conditionals. For
4536 example, some 68000 operating systems call this register @code{%a5}.
4538 Eventually there may be a way of asking the compiler to choose a register
4539 automatically, but first we need to figure out how it should choose and
4540 how to enable you to guide the choice. No solution is evident.
4542 Defining a global register variable in a certain register reserves that
4543 register entirely for this use, at least within the current compilation.
4544 The register will not be allocated for any other purpose in the functions
4545 in the current compilation. The register will not be saved and restored by
4546 these functions. Stores into this register are never deleted even if they
4547 would appear to be dead, but references may be deleted or moved or
4550 It is not safe to access the global register variables from signal
4551 handlers, or from more than one thread of control, because the system
4552 library routines may temporarily use the register for other things (unless
4553 you recompile them specially for the task at hand).
4555 @cindex @code{qsort}, and global register variables
4556 It is not safe for one function that uses a global register variable to
4557 call another such function @code{foo} by way of a third function
4558 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4559 different source file in which the variable wasn't declared). This is
4560 because @code{lose} might save the register and put some other value there.
4561 For example, you can't expect a global register variable to be available in
4562 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4563 might have put something else in that register. (If you are prepared to
4564 recompile @code{qsort} with the same global register variable, you can
4565 solve this problem.)
4567 If you want to recompile @code{qsort} or other source files which do not
4568 actually use your global register variable, so that they will not use that
4569 register for any other purpose, then it suffices to specify the compiler
4570 option @option{-ffixed-@var{reg}}. You need not actually add a global
4571 register declaration to their source code.
4573 A function which can alter the value of a global register variable cannot
4574 safely be called from a function compiled without this variable, because it
4575 could clobber the value the caller expects to find there on return.
4576 Therefore, the function which is the entry point into the part of the
4577 program that uses the global register variable must explicitly save and
4578 restore the value which belongs to its caller.
4580 @cindex register variable after @code{longjmp}
4581 @cindex global register after @code{longjmp}
4582 @cindex value after @code{longjmp}
4585 On most machines, @code{longjmp} will restore to each global register
4586 variable the value it had at the time of the @code{setjmp}. On some
4587 machines, however, @code{longjmp} will not change the value of global
4588 register variables. To be portable, the function that called @code{setjmp}
4589 should make other arrangements to save the values of the global register
4590 variables, and to restore them in a @code{longjmp}. This way, the same
4591 thing will happen regardless of what @code{longjmp} does.
4593 All global register variable declarations must precede all function
4594 definitions. If such a declaration could appear after function
4595 definitions, the declaration would be too late to prevent the register from
4596 being used for other purposes in the preceding functions.
4598 Global register variables may not have initial values, because an
4599 executable file has no means to supply initial contents for a register.
4601 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4602 registers, but certain library functions, such as @code{getwd}, as well
4603 as the subroutines for division and remainder, modify g3 and g4. g1 and
4604 g2 are local temporaries.
4606 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4607 Of course, it will not do to use more than a few of those.
4609 @node Local Reg Vars
4610 @subsection Specifying Registers for Local Variables
4611 @cindex local variables, specifying registers
4612 @cindex specifying registers for local variables
4613 @cindex registers for local variables
4615 You can define a local register variable with a specified register
4619 register int *foo asm ("a5");
4623 Here @code{a5} is the name of the register which should be used. Note
4624 that this is the same syntax used for defining global register
4625 variables, but for a local variable it would appear within a function.
4627 Naturally the register name is cpu-dependent, but this is not a
4628 problem, since specific registers are most often useful with explicit
4629 assembler instructions (@pxref{Extended Asm}). Both of these things
4630 generally require that you conditionalize your program according to
4633 In addition, operating systems on one type of cpu may differ in how they
4634 name the registers; then you would need additional conditionals. For
4635 example, some 68000 operating systems call this register @code{%a5}.
4637 Defining such a register variable does not reserve the register; it
4638 remains available for other uses in places where flow control determines
4639 the variable's value is not live.
4641 This option does not guarantee that GCC will generate code that has
4642 this variable in the register you specify at all times. You may not
4643 code an explicit reference to this register in the @emph{assembler
4644 instruction template} part of an @code{asm} statement and assume it will
4645 always refer to this variable. However, using the variable as an
4646 @code{asm} @emph{operand} guarantees that the specified register is used
4649 Stores into local register variables may be deleted when they appear to be dead
4650 according to dataflow analysis. References to local register variables may
4651 be deleted or moved or simplified.
4653 As for global register variables, it's recommended that you choose a
4654 register which is normally saved and restored by function calls on
4655 your machine, so that library routines will not clobber it. A common
4656 pitfall is to initialize multiple call-clobbered registers with
4657 arbitrary expressions, where a function call or library call for an
4658 arithmetic operator will overwrite a register value from a previous
4659 assignment, for example @code{r0} below:
4661 register int *p1 asm ("r0") = @dots{};
4662 register int *p2 asm ("r1") = @dots{};
4664 In those cases, a solution is to use a temporary variable for
4665 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4667 @node Alternate Keywords
4668 @section Alternate Keywords
4669 @cindex alternate keywords
4670 @cindex keywords, alternate
4672 @option{-ansi} and the various @option{-std} options disable certain
4673 keywords. This causes trouble when you want to use GNU C extensions, or
4674 a general-purpose header file that should be usable by all programs,
4675 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4676 @code{inline} are not available in programs compiled with
4677 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4678 program compiled with @option{-std=c99}). The ISO C99 keyword
4679 @code{restrict} is only available when @option{-std=gnu99} (which will
4680 eventually be the default) or @option{-std=c99} (or the equivalent
4681 @option{-std=iso9899:1999}) is used.
4683 The way to solve these problems is to put @samp{__} at the beginning and
4684 end of each problematical keyword. For example, use @code{__asm__}
4685 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4687 Other C compilers won't accept these alternative keywords; if you want to
4688 compile with another compiler, you can define the alternate keywords as
4689 macros to replace them with the customary keywords. It looks like this:
4697 @findex __extension__
4699 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4701 prevent such warnings within one expression by writing
4702 @code{__extension__} before the expression. @code{__extension__} has no
4703 effect aside from this.
4705 @node Incomplete Enums
4706 @section Incomplete @code{enum} Types
4708 You can define an @code{enum} tag without specifying its possible values.
4709 This results in an incomplete type, much like what you get if you write
4710 @code{struct foo} without describing the elements. A later declaration
4711 which does specify the possible values completes the type.
4713 You can't allocate variables or storage using the type while it is
4714 incomplete. However, you can work with pointers to that type.
4716 This extension may not be very useful, but it makes the handling of
4717 @code{enum} more consistent with the way @code{struct} and @code{union}
4720 This extension is not supported by GNU C++.
4722 @node Function Names
4723 @section Function Names as Strings
4724 @cindex @code{__func__} identifier
4725 @cindex @code{__FUNCTION__} identifier
4726 @cindex @code{__PRETTY_FUNCTION__} identifier
4728 GCC provides three magic variables which hold the name of the current
4729 function, as a string. The first of these is @code{__func__}, which
4730 is part of the C99 standard:
4733 The identifier @code{__func__} is implicitly declared by the translator
4734 as if, immediately following the opening brace of each function
4735 definition, the declaration
4738 static const char __func__[] = "function-name";
4741 appeared, where function-name is the name of the lexically-enclosing
4742 function. This name is the unadorned name of the function.
4745 @code{__FUNCTION__} is another name for @code{__func__}. Older
4746 versions of GCC recognize only this name. However, it is not
4747 standardized. For maximum portability, we recommend you use
4748 @code{__func__}, but provide a fallback definition with the
4752 #if __STDC_VERSION__ < 199901L
4754 # define __func__ __FUNCTION__
4756 # define __func__ "<unknown>"
4761 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4762 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4763 the type signature of the function as well as its bare name. For
4764 example, this program:
4768 extern int printf (char *, ...);
4775 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4776 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4794 __PRETTY_FUNCTION__ = void a::sub(int)
4797 These identifiers are not preprocessor macros. In GCC 3.3 and
4798 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4799 were treated as string literals; they could be used to initialize
4800 @code{char} arrays, and they could be concatenated with other string
4801 literals. GCC 3.4 and later treat them as variables, like
4802 @code{__func__}. In C++, @code{__FUNCTION__} and
4803 @code{__PRETTY_FUNCTION__} have always been variables.
4805 @node Return Address
4806 @section Getting the Return or Frame Address of a Function
4808 These functions may be used to get information about the callers of a
4811 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4812 This function returns the return address of the current function, or of
4813 one of its callers. The @var{level} argument is number of frames to
4814 scan up the call stack. A value of @code{0} yields the return address
4815 of the current function, a value of @code{1} yields the return address
4816 of the caller of the current function, and so forth. When inlining
4817 the expected behavior is that the function will return the address of
4818 the function that will be returned to. To work around this behavior use
4819 the @code{noinline} function attribute.
4821 The @var{level} argument must be a constant integer.
4823 On some machines it may be impossible to determine the return address of
4824 any function other than the current one; in such cases, or when the top
4825 of the stack has been reached, this function will return @code{0} or a
4826 random value. In addition, @code{__builtin_frame_address} may be used
4827 to determine if the top of the stack has been reached.
4829 This function should only be used with a nonzero argument for debugging
4833 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4834 This function is similar to @code{__builtin_return_address}, but it
4835 returns the address of the function frame rather than the return address
4836 of the function. Calling @code{__builtin_frame_address} with a value of
4837 @code{0} yields the frame address of the current function, a value of
4838 @code{1} yields the frame address of the caller of the current function,
4841 The frame is the area on the stack which holds local variables and saved
4842 registers. The frame address is normally the address of the first word
4843 pushed on to the stack by the function. However, the exact definition
4844 depends upon the processor and the calling convention. If the processor
4845 has a dedicated frame pointer register, and the function has a frame,
4846 then @code{__builtin_frame_address} will return the value of the frame
4849 On some machines it may be impossible to determine the frame address of
4850 any function other than the current one; in such cases, or when the top
4851 of the stack has been reached, this function will return @code{0} if
4852 the first frame pointer is properly initialized by the startup code.
4854 This function should only be used with a nonzero argument for debugging
4858 @node Vector Extensions
4859 @section Using vector instructions through built-in functions
4861 On some targets, the instruction set contains SIMD vector instructions that
4862 operate on multiple values contained in one large register at the same time.
4863 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4866 The first step in using these extensions is to provide the necessary data
4867 types. This should be done using an appropriate @code{typedef}:
4870 typedef int v4si __attribute__ ((vector_size (16)));
4873 The @code{int} type specifies the base type, while the attribute specifies
4874 the vector size for the variable, measured in bytes. For example, the
4875 declaration above causes the compiler to set the mode for the @code{v4si}
4876 type to be 16 bytes wide and divided into @code{int} sized units. For
4877 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4878 corresponding mode of @code{foo} will be @acronym{V4SI}.
4880 The @code{vector_size} attribute is only applicable to integral and
4881 float scalars, although arrays, pointers, and function return values
4882 are allowed in conjunction with this construct.
4884 All the basic integer types can be used as base types, both as signed
4885 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4886 @code{long long}. In addition, @code{float} and @code{double} can be
4887 used to build floating-point vector types.
4889 Specifying a combination that is not valid for the current architecture
4890 will cause GCC to synthesize the instructions using a narrower mode.
4891 For example, if you specify a variable of type @code{V4SI} and your
4892 architecture does not allow for this specific SIMD type, GCC will
4893 produce code that uses 4 @code{SIs}.
4895 The types defined in this manner can be used with a subset of normal C
4896 operations. Currently, GCC will allow using the following operators
4897 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4899 The operations behave like C++ @code{valarrays}. Addition is defined as
4900 the addition of the corresponding elements of the operands. For
4901 example, in the code below, each of the 4 elements in @var{a} will be
4902 added to the corresponding 4 elements in @var{b} and the resulting
4903 vector will be stored in @var{c}.
4906 typedef int v4si __attribute__ ((vector_size (16)));
4913 Subtraction, multiplication, division, and the logical operations
4914 operate in a similar manner. Likewise, the result of using the unary
4915 minus or complement operators on a vector type is a vector whose
4916 elements are the negative or complemented values of the corresponding
4917 elements in the operand.
4919 You can declare variables and use them in function calls and returns, as
4920 well as in assignments and some casts. You can specify a vector type as
4921 a return type for a function. Vector types can also be used as function
4922 arguments. It is possible to cast from one vector type to another,
4923 provided they are of the same size (in fact, you can also cast vectors
4924 to and from other datatypes of the same size).
4926 You cannot operate between vectors of different lengths or different
4927 signedness without a cast.
4929 A port that supports hardware vector operations, usually provides a set
4930 of built-in functions that can be used to operate on vectors. For
4931 example, a function to add two vectors and multiply the result by a
4932 third could look like this:
4935 v4si f (v4si a, v4si b, v4si c)
4937 v4si tmp = __builtin_addv4si (a, b);
4938 return __builtin_mulv4si (tmp, c);
4945 @findex __builtin_offsetof
4947 GCC implements for both C and C++ a syntactic extension to implement
4948 the @code{offsetof} macro.
4952 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4954 offsetof_member_designator:
4956 | offsetof_member_designator "." @code{identifier}
4957 | offsetof_member_designator "[" @code{expr} "]"
4960 This extension is sufficient such that
4963 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4966 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4967 may be dependent. In either case, @var{member} may consist of a single
4968 identifier, or a sequence of member accesses and array references.
4970 @node Atomic Builtins
4971 @section Built-in functions for atomic memory access
4973 The following builtins are intended to be compatible with those described
4974 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
4975 section 7.4. As such, they depart from the normal GCC practice of using
4976 the ``__builtin_'' prefix, and further that they are overloaded such that
4977 they work on multiple types.
4979 The definition given in the Intel documentation allows only for the use of
4980 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
4981 counterparts. GCC will allow any integral scalar or pointer type that is
4982 1, 2, 4 or 8 bytes in length.
4984 Not all operations are supported by all target processors. If a particular
4985 operation cannot be implemented on the target processor, a warning will be
4986 generated and a call an external function will be generated. The external
4987 function will carry the same name as the builtin, with an additional suffix
4988 @samp{_@var{n}} where @var{n} is the size of the data type.
4990 @c ??? Should we have a mechanism to suppress this warning? This is almost
4991 @c useful for implementing the operation under the control of an external
4994 In most cases, these builtins are considered a @dfn{full barrier}. That is,
4995 no memory operand will be moved across the operation, either forward or
4996 backward. Further, instructions will be issued as necessary to prevent the
4997 processor from speculating loads across the operation and from queuing stores
4998 after the operation.
5000 All of the routines are are described in the Intel documentation to take
5001 ``an optional list of variables protected by the memory barrier''. It's
5002 not clear what is meant by that; it could mean that @emph{only} the
5003 following variables are protected, or it could mean that these variables
5004 should in addition be protected. At present GCC ignores this list and
5005 protects all variables which are globally accessible. If in the future
5006 we make some use of this list, an empty list will continue to mean all
5007 globally accessible variables.
5010 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5011 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5012 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5013 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5014 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5015 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5016 @findex __sync_fetch_and_add
5017 @findex __sync_fetch_and_sub
5018 @findex __sync_fetch_and_or
5019 @findex __sync_fetch_and_and
5020 @findex __sync_fetch_and_xor
5021 @findex __sync_fetch_and_nand
5022 These builtins perform the operation suggested by the name, and
5023 returns the value that had previously been in memory. That is,
5026 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5027 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5030 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5031 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5032 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5033 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5034 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5035 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5036 @findex __sync_add_and_fetch
5037 @findex __sync_sub_and_fetch
5038 @findex __sync_or_and_fetch
5039 @findex __sync_and_and_fetch
5040 @findex __sync_xor_and_fetch
5041 @findex __sync_nand_and_fetch
5042 These builtins perform the operation suggested by the name, and
5043 return the new value. That is,
5046 @{ *ptr @var{op}= value; return *ptr; @}
5047 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5050 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5051 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5052 @findex __sync_bool_compare_and_swap
5053 @findex __sync_val_compare_and_swap
5054 These builtins perform an atomic compare and swap. That is, if the current
5055 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5058 The ``bool'' version returns true if the comparison is successful and
5059 @var{newval} was written. The ``val'' version returns the contents
5060 of @code{*@var{ptr}} before the operation.
5062 @item __sync_synchronize (...)
5063 @findex __sync_synchronize
5064 This builtin issues a full memory barrier.
5066 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5067 @findex __sync_lock_test_and_set
5068 This builtin, as described by Intel, is not a traditional test-and-set
5069 operation, but rather an atomic exchange operation. It writes @var{value}
5070 into @code{*@var{ptr}}, and returns the previous contents of
5073 Many targets have only minimal support for such locks, and do not support
5074 a full exchange operation. In this case, a target may support reduced
5075 functionality here by which the @emph{only} valid value to store is the
5076 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5077 is implementation defined.
5079 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5080 This means that references after the builtin cannot move to (or be
5081 speculated to) before the builtin, but previous memory stores may not
5082 be globally visible yet, and previous memory loads may not yet be
5085 @item void __sync_lock_release (@var{type} *ptr, ...)
5086 @findex __sync_lock_release
5087 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5088 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5090 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5091 This means that all previous memory stores are globally visible, and all
5092 previous memory loads have been satisfied, but following memory reads
5093 are not prevented from being speculated to before the barrier.
5096 @node Object Size Checking
5097 @section Object Size Checking Builtins
5098 @findex __builtin_object_size
5099 @findex __builtin___memcpy_chk
5100 @findex __builtin___mempcpy_chk
5101 @findex __builtin___memmove_chk
5102 @findex __builtin___memset_chk
5103 @findex __builtin___strcpy_chk
5104 @findex __builtin___stpcpy_chk
5105 @findex __builtin___strncpy_chk
5106 @findex __builtin___strcat_chk
5107 @findex __builtin___strncat_chk
5108 @findex __builtin___sprintf_chk
5109 @findex __builtin___snprintf_chk
5110 @findex __builtin___vsprintf_chk
5111 @findex __builtin___vsnprintf_chk
5112 @findex __builtin___printf_chk
5113 @findex __builtin___vprintf_chk
5114 @findex __builtin___fprintf_chk
5115 @findex __builtin___vfprintf_chk
5117 GCC implements a limited buffer overflow protection mechanism
5118 that can prevent some buffer overflow attacks.
5120 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5121 is a built-in construct that returns a constant number of bytes from
5122 @var{ptr} to the end of the object @var{ptr} pointer points to
5123 (if known at compile time). @code{__builtin_object_size} never evaluates
5124 its arguments for side-effects. If there are any side-effects in them, it
5125 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5126 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5127 point to and all of them are known at compile time, the returned number
5128 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5129 0 and minimum if nonzero. If it is not possible to determine which objects
5130 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5131 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5132 for @var{type} 2 or 3.
5134 @var{type} is an integer constant from 0 to 3. If the least significant
5135 bit is clear, objects are whole variables, if it is set, a closest
5136 surrounding subobject is considered the object a pointer points to.
5137 The second bit determines if maximum or minimum of remaining bytes
5141 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5142 char *p = &var.buf1[1], *q = &var.b;
5144 /* Here the object p points to is var. */
5145 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5146 /* The subobject p points to is var.buf1. */
5147 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5148 /* The object q points to is var. */
5149 assert (__builtin_object_size (q, 0)
5150 == (char *) (&var + 1) - (char *) &var.b);
5151 /* The subobject q points to is var.b. */
5152 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5156 There are built-in functions added for many common string operation
5157 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5158 built-in is provided. This built-in has an additional last argument,
5159 which is the number of bytes remaining in object the @var{dest}
5160 argument points to or @code{(size_t) -1} if the size is not known.
5162 The built-in functions are optimized into the normal string functions
5163 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5164 it is known at compile time that the destination object will not
5165 be overflown. If the compiler can determine at compile time the
5166 object will be always overflown, it issues a warning.
5168 The intended use can be e.g.
5172 #define bos0(dest) __builtin_object_size (dest, 0)
5173 #define memcpy(dest, src, n) \
5174 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5178 /* It is unknown what object p points to, so this is optimized
5179 into plain memcpy - no checking is possible. */
5180 memcpy (p, "abcde", n);
5181 /* Destination is known and length too. It is known at compile
5182 time there will be no overflow. */
5183 memcpy (&buf[5], "abcde", 5);
5184 /* Destination is known, but the length is not known at compile time.
5185 This will result in __memcpy_chk call that can check for overflow
5187 memcpy (&buf[5], "abcde", n);
5188 /* Destination is known and it is known at compile time there will
5189 be overflow. There will be a warning and __memcpy_chk call that
5190 will abort the program at runtime. */
5191 memcpy (&buf[6], "abcde", 5);
5194 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5195 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5196 @code{strcat} and @code{strncat}.
5198 There are also checking built-in functions for formatted output functions.
5200 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5201 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5202 const char *fmt, ...);
5203 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5205 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5206 const char *fmt, va_list ap);
5209 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5210 etc. functions and can contain implementation specific flags on what
5211 additional security measures the checking function might take, such as
5212 handling @code{%n} differently.
5214 The @var{os} argument is the object size @var{s} points to, like in the
5215 other built-in functions. There is a small difference in the behavior
5216 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5217 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5218 the checking function is called with @var{os} argument set to
5221 In addition to this, there are checking built-in functions
5222 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5223 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5224 These have just one additional argument, @var{flag}, right before
5225 format string @var{fmt}. If the compiler is able to optimize them to
5226 @code{fputc} etc. functions, it will, otherwise the checking function
5227 should be called and the @var{flag} argument passed to it.
5229 @node Other Builtins
5230 @section Other built-in functions provided by GCC
5231 @cindex built-in functions
5232 @findex __builtin_isgreater
5233 @findex __builtin_isgreaterequal
5234 @findex __builtin_isless
5235 @findex __builtin_islessequal
5236 @findex __builtin_islessgreater
5237 @findex __builtin_isunordered
5238 @findex __builtin_powi
5239 @findex __builtin_powif
5240 @findex __builtin_powil
5398 @findex fprintf_unlocked
5400 @findex fputs_unlocked
5510 @findex printf_unlocked
5539 @findex significandf
5540 @findex significandl
5611 GCC provides a large number of built-in functions other than the ones
5612 mentioned above. Some of these are for internal use in the processing
5613 of exceptions or variable-length argument lists and will not be
5614 documented here because they may change from time to time; we do not
5615 recommend general use of these functions.
5617 The remaining functions are provided for optimization purposes.
5619 @opindex fno-builtin
5620 GCC includes built-in versions of many of the functions in the standard
5621 C library. The versions prefixed with @code{__builtin_} will always be
5622 treated as having the same meaning as the C library function even if you
5623 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5624 Many of these functions are only optimized in certain cases; if they are
5625 not optimized in a particular case, a call to the library function will
5630 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5631 @option{-std=c99}), the functions
5632 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5633 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5634 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5635 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5636 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5637 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5638 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5639 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5640 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5641 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5642 @code{significandf}, @code{significandl}, @code{significand},
5643 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5644 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5645 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5646 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5647 @code{ynl} and @code{yn}
5648 may be handled as built-in functions.
5649 All these functions have corresponding versions
5650 prefixed with @code{__builtin_}, which may be used even in strict C89
5653 The ISO C99 functions
5654 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5655 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5656 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5657 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5658 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5659 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5660 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5661 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5662 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5663 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5664 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5665 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5666 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5667 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5668 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5669 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5670 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5671 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5672 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5673 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5674 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5675 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5676 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5677 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5678 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5679 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5680 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5681 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5682 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5683 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5684 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5685 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5686 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5687 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5688 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5689 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5690 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5691 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5692 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5693 are handled as built-in functions
5694 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5696 There are also built-in versions of the ISO C99 functions
5697 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5698 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5699 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5700 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5701 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5702 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5703 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5704 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5705 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5706 that are recognized in any mode since ISO C90 reserves these names for
5707 the purpose to which ISO C99 puts them. All these functions have
5708 corresponding versions prefixed with @code{__builtin_}.
5710 The ISO C94 functions
5711 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5712 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5713 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5715 are handled as built-in functions
5716 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5718 The ISO C90 functions
5719 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5720 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5721 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5722 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5723 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5724 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5725 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5726 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5727 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5728 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5729 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5730 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5731 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5732 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5733 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5734 @code{vprintf} and @code{vsprintf}
5735 are all recognized as built-in functions unless
5736 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5737 is specified for an individual function). All of these functions have
5738 corresponding versions prefixed with @code{__builtin_}.
5740 GCC provides built-in versions of the ISO C99 floating point comparison
5741 macros that avoid raising exceptions for unordered operands. They have
5742 the same names as the standard macros ( @code{isgreater},
5743 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5744 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5745 prefixed. We intend for a library implementor to be able to simply
5746 @code{#define} each standard macro to its built-in equivalent.
5748 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5750 You can use the built-in function @code{__builtin_types_compatible_p} to
5751 determine whether two types are the same.
5753 This built-in function returns 1 if the unqualified versions of the
5754 types @var{type1} and @var{type2} (which are types, not expressions) are
5755 compatible, 0 otherwise. The result of this built-in function can be
5756 used in integer constant expressions.
5758 This built-in function ignores top level qualifiers (e.g., @code{const},
5759 @code{volatile}). For example, @code{int} is equivalent to @code{const
5762 The type @code{int[]} and @code{int[5]} are compatible. On the other
5763 hand, @code{int} and @code{char *} are not compatible, even if the size
5764 of their types, on the particular architecture are the same. Also, the
5765 amount of pointer indirection is taken into account when determining
5766 similarity. Consequently, @code{short *} is not similar to
5767 @code{short **}. Furthermore, two types that are typedefed are
5768 considered compatible if their underlying types are compatible.
5770 An @code{enum} type is not considered to be compatible with another
5771 @code{enum} type even if both are compatible with the same integer
5772 type; this is what the C standard specifies.
5773 For example, @code{enum @{foo, bar@}} is not similar to
5774 @code{enum @{hot, dog@}}.
5776 You would typically use this function in code whose execution varies
5777 depending on the arguments' types. For example:
5782 typeof (x) tmp = (x); \
5783 if (__builtin_types_compatible_p (typeof (x), long double)) \
5784 tmp = foo_long_double (tmp); \
5785 else if (__builtin_types_compatible_p (typeof (x), double)) \
5786 tmp = foo_double (tmp); \
5787 else if (__builtin_types_compatible_p (typeof (x), float)) \
5788 tmp = foo_float (tmp); \
5795 @emph{Note:} This construct is only available for C@.
5799 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5801 You can use the built-in function @code{__builtin_choose_expr} to
5802 evaluate code depending on the value of a constant expression. This
5803 built-in function returns @var{exp1} if @var{const_exp}, which is a
5804 constant expression that must be able to be determined at compile time,
5805 is nonzero. Otherwise it returns 0.
5807 This built-in function is analogous to the @samp{? :} operator in C,
5808 except that the expression returned has its type unaltered by promotion
5809 rules. Also, the built-in function does not evaluate the expression
5810 that was not chosen. For example, if @var{const_exp} evaluates to true,
5811 @var{exp2} is not evaluated even if it has side-effects.
5813 This built-in function can return an lvalue if the chosen argument is an
5816 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5817 type. Similarly, if @var{exp2} is returned, its return type is the same
5824 __builtin_choose_expr ( \
5825 __builtin_types_compatible_p (typeof (x), double), \
5827 __builtin_choose_expr ( \
5828 __builtin_types_compatible_p (typeof (x), float), \
5830 /* @r{The void expression results in a compile-time error} \
5831 @r{when assigning the result to something.} */ \
5835 @emph{Note:} This construct is only available for C@. Furthermore, the
5836 unused expression (@var{exp1} or @var{exp2} depending on the value of
5837 @var{const_exp}) may still generate syntax errors. This may change in
5842 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5843 You can use the built-in function @code{__builtin_constant_p} to
5844 determine if a value is known to be constant at compile-time and hence
5845 that GCC can perform constant-folding on expressions involving that
5846 value. The argument of the function is the value to test. The function
5847 returns the integer 1 if the argument is known to be a compile-time
5848 constant and 0 if it is not known to be a compile-time constant. A
5849 return of 0 does not indicate that the value is @emph{not} a constant,
5850 but merely that GCC cannot prove it is a constant with the specified
5851 value of the @option{-O} option.
5853 You would typically use this function in an embedded application where
5854 memory was a critical resource. If you have some complex calculation,
5855 you may want it to be folded if it involves constants, but need to call
5856 a function if it does not. For example:
5859 #define Scale_Value(X) \
5860 (__builtin_constant_p (X) \
5861 ? ((X) * SCALE + OFFSET) : Scale (X))
5864 You may use this built-in function in either a macro or an inline
5865 function. However, if you use it in an inlined function and pass an
5866 argument of the function as the argument to the built-in, GCC will
5867 never return 1 when you call the inline function with a string constant
5868 or compound literal (@pxref{Compound Literals}) and will not return 1
5869 when you pass a constant numeric value to the inline function unless you
5870 specify the @option{-O} option.
5872 You may also use @code{__builtin_constant_p} in initializers for static
5873 data. For instance, you can write
5876 static const int table[] = @{
5877 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5883 This is an acceptable initializer even if @var{EXPRESSION} is not a
5884 constant expression. GCC must be more conservative about evaluating the
5885 built-in in this case, because it has no opportunity to perform
5888 Previous versions of GCC did not accept this built-in in data
5889 initializers. The earliest version where it is completely safe is
5893 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5894 @opindex fprofile-arcs
5895 You may use @code{__builtin_expect} to provide the compiler with
5896 branch prediction information. In general, you should prefer to
5897 use actual profile feedback for this (@option{-fprofile-arcs}), as
5898 programmers are notoriously bad at predicting how their programs
5899 actually perform. However, there are applications in which this
5900 data is hard to collect.
5902 The return value is the value of @var{exp}, which should be an
5903 integral expression. The value of @var{c} must be a compile-time
5904 constant. The semantics of the built-in are that it is expected
5905 that @var{exp} == @var{c}. For example:
5908 if (__builtin_expect (x, 0))
5913 would indicate that we do not expect to call @code{foo}, since
5914 we expect @code{x} to be zero. Since you are limited to integral
5915 expressions for @var{exp}, you should use constructions such as
5918 if (__builtin_expect (ptr != NULL, 1))
5923 when testing pointer or floating-point values.
5926 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5927 This function is used to minimize cache-miss latency by moving data into
5928 a cache before it is accessed.
5929 You can insert calls to @code{__builtin_prefetch} into code for which
5930 you know addresses of data in memory that is likely to be accessed soon.
5931 If the target supports them, data prefetch instructions will be generated.
5932 If the prefetch is done early enough before the access then the data will
5933 be in the cache by the time it is accessed.
5935 The value of @var{addr} is the address of the memory to prefetch.
5936 There are two optional arguments, @var{rw} and @var{locality}.
5937 The value of @var{rw} is a compile-time constant one or zero; one
5938 means that the prefetch is preparing for a write to the memory address
5939 and zero, the default, means that the prefetch is preparing for a read.
5940 The value @var{locality} must be a compile-time constant integer between
5941 zero and three. A value of zero means that the data has no temporal
5942 locality, so it need not be left in the cache after the access. A value
5943 of three means that the data has a high degree of temporal locality and
5944 should be left in all levels of cache possible. Values of one and two
5945 mean, respectively, a low or moderate degree of temporal locality. The
5949 for (i = 0; i < n; i++)
5952 __builtin_prefetch (&a[i+j], 1, 1);
5953 __builtin_prefetch (&b[i+j], 0, 1);
5958 Data prefetch does not generate faults if @var{addr} is invalid, but
5959 the address expression itself must be valid. For example, a prefetch
5960 of @code{p->next} will not fault if @code{p->next} is not a valid
5961 address, but evaluation will fault if @code{p} is not a valid address.
5963 If the target does not support data prefetch, the address expression
5964 is evaluated if it includes side effects but no other code is generated
5965 and GCC does not issue a warning.
5968 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5969 Returns a positive infinity, if supported by the floating-point format,
5970 else @code{DBL_MAX}. This function is suitable for implementing the
5971 ISO C macro @code{HUGE_VAL}.
5974 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5975 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5978 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5979 Similar to @code{__builtin_huge_val}, except the return
5980 type is @code{long double}.
5983 @deftypefn {Built-in Function} double __builtin_inf (void)
5984 Similar to @code{__builtin_huge_val}, except a warning is generated
5985 if the target floating-point format does not support infinities.
5988 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
5989 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
5992 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
5993 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
5996 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
5997 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6000 @deftypefn {Built-in Function} float __builtin_inff (void)
6001 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6002 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6005 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6006 Similar to @code{__builtin_inf}, except the return
6007 type is @code{long double}.
6010 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6011 This is an implementation of the ISO C99 function @code{nan}.
6013 Since ISO C99 defines this function in terms of @code{strtod}, which we
6014 do not implement, a description of the parsing is in order. The string
6015 is parsed as by @code{strtol}; that is, the base is recognized by
6016 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6017 in the significand such that the least significant bit of the number
6018 is at the least significant bit of the significand. The number is
6019 truncated to fit the significand field provided. The significand is
6020 forced to be a quiet NaN@.
6022 This function, if given a string literal all of which would have been
6023 consumed by strtol, is evaluated early enough that it is considered a
6024 compile-time constant.
6027 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6028 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6031 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6032 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6035 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6036 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6039 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6040 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6043 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6044 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6047 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6048 Similar to @code{__builtin_nan}, except the significand is forced
6049 to be a signaling NaN@. The @code{nans} function is proposed by
6050 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6053 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6054 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6057 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6058 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6061 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6062 Returns one plus the index of the least significant 1-bit of @var{x}, or
6063 if @var{x} is zero, returns zero.
6066 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6067 Returns the number of leading 0-bits in @var{x}, starting at the most
6068 significant bit position. If @var{x} is 0, the result is undefined.
6071 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6072 Returns the number of trailing 0-bits in @var{x}, starting at the least
6073 significant bit position. If @var{x} is 0, the result is undefined.
6076 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6077 Returns the number of 1-bits in @var{x}.
6080 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6081 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6085 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6086 Similar to @code{__builtin_ffs}, except the argument type is
6087 @code{unsigned long}.
6090 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6091 Similar to @code{__builtin_clz}, except the argument type is
6092 @code{unsigned long}.
6095 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6096 Similar to @code{__builtin_ctz}, except the argument type is
6097 @code{unsigned long}.
6100 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6101 Similar to @code{__builtin_popcount}, except the argument type is
6102 @code{unsigned long}.
6105 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6106 Similar to @code{__builtin_parity}, except the argument type is
6107 @code{unsigned long}.
6110 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6111 Similar to @code{__builtin_ffs}, except the argument type is
6112 @code{unsigned long long}.
6115 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6116 Similar to @code{__builtin_clz}, except the argument type is
6117 @code{unsigned long long}.
6120 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6121 Similar to @code{__builtin_ctz}, except the argument type is
6122 @code{unsigned long long}.
6125 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6126 Similar to @code{__builtin_popcount}, except the argument type is
6127 @code{unsigned long long}.
6130 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6131 Similar to @code{__builtin_parity}, except the argument type is
6132 @code{unsigned long long}.
6135 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6136 Returns the first argument raised to the power of the second. Unlike the
6137 @code{pow} function no guarantees about precision and rounding are made.
6140 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6141 Similar to @code{__builtin_powi}, except the argument and return types
6145 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6146 Similar to @code{__builtin_powi}, except the argument and return types
6147 are @code{long double}.
6151 @node Target Builtins
6152 @section Built-in Functions Specific to Particular Target Machines
6154 On some target machines, GCC supports many built-in functions specific
6155 to those machines. Generally these generate calls to specific machine
6156 instructions, but allow the compiler to schedule those calls.
6159 * Alpha Built-in Functions::
6160 * ARM Built-in Functions::
6161 * Blackfin Built-in Functions::
6162 * FR-V Built-in Functions::
6163 * X86 Built-in Functions::
6164 * MIPS DSP Built-in Functions::
6165 * MIPS Paired-Single Support::
6166 * PowerPC AltiVec Built-in Functions::
6167 * SPARC VIS Built-in Functions::
6170 @node Alpha Built-in Functions
6171 @subsection Alpha Built-in Functions
6173 These built-in functions are available for the Alpha family of
6174 processors, depending on the command-line switches used.
6176 The following built-in functions are always available. They
6177 all generate the machine instruction that is part of the name.
6180 long __builtin_alpha_implver (void)
6181 long __builtin_alpha_rpcc (void)
6182 long __builtin_alpha_amask (long)
6183 long __builtin_alpha_cmpbge (long, long)
6184 long __builtin_alpha_extbl (long, long)
6185 long __builtin_alpha_extwl (long, long)
6186 long __builtin_alpha_extll (long, long)
6187 long __builtin_alpha_extql (long, long)
6188 long __builtin_alpha_extwh (long, long)
6189 long __builtin_alpha_extlh (long, long)
6190 long __builtin_alpha_extqh (long, long)
6191 long __builtin_alpha_insbl (long, long)
6192 long __builtin_alpha_inswl (long, long)
6193 long __builtin_alpha_insll (long, long)
6194 long __builtin_alpha_insql (long, long)
6195 long __builtin_alpha_inswh (long, long)
6196 long __builtin_alpha_inslh (long, long)
6197 long __builtin_alpha_insqh (long, long)
6198 long __builtin_alpha_mskbl (long, long)
6199 long __builtin_alpha_mskwl (long, long)
6200 long __builtin_alpha_mskll (long, long)
6201 long __builtin_alpha_mskql (long, long)
6202 long __builtin_alpha_mskwh (long, long)
6203 long __builtin_alpha_msklh (long, long)
6204 long __builtin_alpha_mskqh (long, long)
6205 long __builtin_alpha_umulh (long, long)
6206 long __builtin_alpha_zap (long, long)
6207 long __builtin_alpha_zapnot (long, long)
6210 The following built-in functions are always with @option{-mmax}
6211 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6212 later. They all generate the machine instruction that is part
6216 long __builtin_alpha_pklb (long)
6217 long __builtin_alpha_pkwb (long)
6218 long __builtin_alpha_unpkbl (long)
6219 long __builtin_alpha_unpkbw (long)
6220 long __builtin_alpha_minub8 (long, long)
6221 long __builtin_alpha_minsb8 (long, long)
6222 long __builtin_alpha_minuw4 (long, long)
6223 long __builtin_alpha_minsw4 (long, long)
6224 long __builtin_alpha_maxub8 (long, long)
6225 long __builtin_alpha_maxsb8 (long, long)
6226 long __builtin_alpha_maxuw4 (long, long)
6227 long __builtin_alpha_maxsw4 (long, long)
6228 long __builtin_alpha_perr (long, long)
6231 The following built-in functions are always with @option{-mcix}
6232 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6233 later. They all generate the machine instruction that is part
6237 long __builtin_alpha_cttz (long)
6238 long __builtin_alpha_ctlz (long)
6239 long __builtin_alpha_ctpop (long)
6242 The following builtins are available on systems that use the OSF/1
6243 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6244 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6245 @code{rdval} and @code{wrval}.
6248 void *__builtin_thread_pointer (void)
6249 void __builtin_set_thread_pointer (void *)
6252 @node ARM Built-in Functions
6253 @subsection ARM Built-in Functions
6255 These built-in functions are available for the ARM family of
6256 processors, when the @option{-mcpu=iwmmxt} switch is used:
6259 typedef int v2si __attribute__ ((vector_size (8)));
6260 typedef short v4hi __attribute__ ((vector_size (8)));
6261 typedef char v8qi __attribute__ ((vector_size (8)));
6263 int __builtin_arm_getwcx (int)
6264 void __builtin_arm_setwcx (int, int)
6265 int __builtin_arm_textrmsb (v8qi, int)
6266 int __builtin_arm_textrmsh (v4hi, int)
6267 int __builtin_arm_textrmsw (v2si, int)
6268 int __builtin_arm_textrmub (v8qi, int)
6269 int __builtin_arm_textrmuh (v4hi, int)
6270 int __builtin_arm_textrmuw (v2si, int)
6271 v8qi __builtin_arm_tinsrb (v8qi, int)
6272 v4hi __builtin_arm_tinsrh (v4hi, int)
6273 v2si __builtin_arm_tinsrw (v2si, int)
6274 long long __builtin_arm_tmia (long long, int, int)
6275 long long __builtin_arm_tmiabb (long long, int, int)
6276 long long __builtin_arm_tmiabt (long long, int, int)
6277 long long __builtin_arm_tmiaph (long long, int, int)
6278 long long __builtin_arm_tmiatb (long long, int, int)
6279 long long __builtin_arm_tmiatt (long long, int, int)
6280 int __builtin_arm_tmovmskb (v8qi)
6281 int __builtin_arm_tmovmskh (v4hi)
6282 int __builtin_arm_tmovmskw (v2si)
6283 long long __builtin_arm_waccb (v8qi)
6284 long long __builtin_arm_wacch (v4hi)
6285 long long __builtin_arm_waccw (v2si)
6286 v8qi __builtin_arm_waddb (v8qi, v8qi)
6287 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6288 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6289 v4hi __builtin_arm_waddh (v4hi, v4hi)
6290 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6291 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6292 v2si __builtin_arm_waddw (v2si, v2si)
6293 v2si __builtin_arm_waddwss (v2si, v2si)
6294 v2si __builtin_arm_waddwus (v2si, v2si)
6295 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6296 long long __builtin_arm_wand(long long, long long)
6297 long long __builtin_arm_wandn (long long, long long)
6298 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6299 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6300 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6301 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6302 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6303 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6304 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6305 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6306 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6307 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6308 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6309 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6310 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6311 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6312 long long __builtin_arm_wmacsz (v4hi, v4hi)
6313 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6314 long long __builtin_arm_wmacuz (v4hi, v4hi)
6315 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6316 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6317 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6318 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6319 v2si __builtin_arm_wmaxsw (v2si, v2si)
6320 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6321 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6322 v2si __builtin_arm_wmaxuw (v2si, v2si)
6323 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6324 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6325 v2si __builtin_arm_wminsw (v2si, v2si)
6326 v8qi __builtin_arm_wminub (v8qi, v8qi)
6327 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6328 v2si __builtin_arm_wminuw (v2si, v2si)
6329 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6330 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6331 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6332 long long __builtin_arm_wor (long long, long long)
6333 v2si __builtin_arm_wpackdss (long long, long long)
6334 v2si __builtin_arm_wpackdus (long long, long long)
6335 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6336 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6337 v4hi __builtin_arm_wpackwss (v2si, v2si)
6338 v4hi __builtin_arm_wpackwus (v2si, v2si)
6339 long long __builtin_arm_wrord (long long, long long)
6340 long long __builtin_arm_wrordi (long long, int)
6341 v4hi __builtin_arm_wrorh (v4hi, long long)
6342 v4hi __builtin_arm_wrorhi (v4hi, int)
6343 v2si __builtin_arm_wrorw (v2si, long long)
6344 v2si __builtin_arm_wrorwi (v2si, int)
6345 v2si __builtin_arm_wsadb (v8qi, v8qi)
6346 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6347 v2si __builtin_arm_wsadh (v4hi, v4hi)
6348 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6349 v4hi __builtin_arm_wshufh (v4hi, int)
6350 long long __builtin_arm_wslld (long long, long long)
6351 long long __builtin_arm_wslldi (long long, int)
6352 v4hi __builtin_arm_wsllh (v4hi, long long)
6353 v4hi __builtin_arm_wsllhi (v4hi, int)
6354 v2si __builtin_arm_wsllw (v2si, long long)
6355 v2si __builtin_arm_wsllwi (v2si, int)
6356 long long __builtin_arm_wsrad (long long, long long)
6357 long long __builtin_arm_wsradi (long long, int)
6358 v4hi __builtin_arm_wsrah (v4hi, long long)
6359 v4hi __builtin_arm_wsrahi (v4hi, int)
6360 v2si __builtin_arm_wsraw (v2si, long long)
6361 v2si __builtin_arm_wsrawi (v2si, int)
6362 long long __builtin_arm_wsrld (long long, long long)
6363 long long __builtin_arm_wsrldi (long long, int)
6364 v4hi __builtin_arm_wsrlh (v4hi, long long)
6365 v4hi __builtin_arm_wsrlhi (v4hi, int)
6366 v2si __builtin_arm_wsrlw (v2si, long long)
6367 v2si __builtin_arm_wsrlwi (v2si, int)
6368 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6369 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6370 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6371 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6372 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6373 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6374 v2si __builtin_arm_wsubw (v2si, v2si)
6375 v2si __builtin_arm_wsubwss (v2si, v2si)
6376 v2si __builtin_arm_wsubwus (v2si, v2si)
6377 v4hi __builtin_arm_wunpckehsb (v8qi)
6378 v2si __builtin_arm_wunpckehsh (v4hi)
6379 long long __builtin_arm_wunpckehsw (v2si)
6380 v4hi __builtin_arm_wunpckehub (v8qi)
6381 v2si __builtin_arm_wunpckehuh (v4hi)
6382 long long __builtin_arm_wunpckehuw (v2si)
6383 v4hi __builtin_arm_wunpckelsb (v8qi)
6384 v2si __builtin_arm_wunpckelsh (v4hi)
6385 long long __builtin_arm_wunpckelsw (v2si)
6386 v4hi __builtin_arm_wunpckelub (v8qi)
6387 v2si __builtin_arm_wunpckeluh (v4hi)
6388 long long __builtin_arm_wunpckeluw (v2si)
6389 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6390 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6391 v2si __builtin_arm_wunpckihw (v2si, v2si)
6392 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6393 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6394 v2si __builtin_arm_wunpckilw (v2si, v2si)
6395 long long __builtin_arm_wxor (long long, long long)
6396 long long __builtin_arm_wzero ()
6399 @node Blackfin Built-in Functions
6400 @subsection Blackfin Built-in Functions
6402 Currently, there are two Blackfin-specific built-in functions. These are
6403 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6404 using inline assembly; by using these built-in functions the compiler can
6405 automatically add workarounds for hardware errata involving these
6406 instructions. These functions are named as follows:
6409 void __builtin_bfin_csync (void)
6410 void __builtin_bfin_ssync (void)
6413 @node FR-V Built-in Functions
6414 @subsection FR-V Built-in Functions
6416 GCC provides many FR-V-specific built-in functions. In general,
6417 these functions are intended to be compatible with those described
6418 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6419 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6420 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6421 pointer rather than by value.
6423 Most of the functions are named after specific FR-V instructions.
6424 Such functions are said to be ``directly mapped'' and are summarized
6425 here in tabular form.
6429 * Directly-mapped Integer Functions::
6430 * Directly-mapped Media Functions::
6431 * Raw read/write Functions::
6432 * Other Built-in Functions::
6435 @node Argument Types
6436 @subsubsection Argument Types
6438 The arguments to the built-in functions can be divided into three groups:
6439 register numbers, compile-time constants and run-time values. In order
6440 to make this classification clear at a glance, the arguments and return
6441 values are given the following pseudo types:
6443 @multitable @columnfractions .20 .30 .15 .35
6444 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6445 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6446 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6447 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6448 @item @code{uw2} @tab @code{unsigned long long} @tab No
6449 @tab an unsigned doubleword
6450 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6451 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6452 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6453 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6456 These pseudo types are not defined by GCC, they are simply a notational
6457 convenience used in this manual.
6459 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6460 and @code{sw2} are evaluated at run time. They correspond to
6461 register operands in the underlying FR-V instructions.
6463 @code{const} arguments represent immediate operands in the underlying
6464 FR-V instructions. They must be compile-time constants.
6466 @code{acc} arguments are evaluated at compile time and specify the number
6467 of an accumulator register. For example, an @code{acc} argument of 2
6468 will select the ACC2 register.
6470 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6471 number of an IACC register. See @pxref{Other Built-in Functions}
6474 @node Directly-mapped Integer Functions
6475 @subsubsection Directly-mapped Integer Functions
6477 The functions listed below map directly to FR-V I-type instructions.
6479 @multitable @columnfractions .45 .32 .23
6480 @item Function prototype @tab Example usage @tab Assembly output
6481 @item @code{sw1 __ADDSS (sw1, sw1)}
6482 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6483 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6484 @item @code{sw1 __SCAN (sw1, sw1)}
6485 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6486 @tab @code{SCAN @var{a},@var{b},@var{c}}
6487 @item @code{sw1 __SCUTSS (sw1)}
6488 @tab @code{@var{b} = __SCUTSS (@var{a})}
6489 @tab @code{SCUTSS @var{a},@var{b}}
6490 @item @code{sw1 __SLASS (sw1, sw1)}
6491 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6492 @tab @code{SLASS @var{a},@var{b},@var{c}}
6493 @item @code{void __SMASS (sw1, sw1)}
6494 @tab @code{__SMASS (@var{a}, @var{b})}
6495 @tab @code{SMASS @var{a},@var{b}}
6496 @item @code{void __SMSSS (sw1, sw1)}
6497 @tab @code{__SMSSS (@var{a}, @var{b})}
6498 @tab @code{SMSSS @var{a},@var{b}}
6499 @item @code{void __SMU (sw1, sw1)}
6500 @tab @code{__SMU (@var{a}, @var{b})}
6501 @tab @code{SMU @var{a},@var{b}}
6502 @item @code{sw2 __SMUL (sw1, sw1)}
6503 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6504 @tab @code{SMUL @var{a},@var{b},@var{c}}
6505 @item @code{sw1 __SUBSS (sw1, sw1)}
6506 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6507 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6508 @item @code{uw2 __UMUL (uw1, uw1)}
6509 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6510 @tab @code{UMUL @var{a},@var{b},@var{c}}
6513 @node Directly-mapped Media Functions
6514 @subsubsection Directly-mapped Media Functions
6516 The functions listed below map directly to FR-V M-type instructions.
6518 @multitable @columnfractions .45 .32 .23
6519 @item Function prototype @tab Example usage @tab Assembly output
6520 @item @code{uw1 __MABSHS (sw1)}
6521 @tab @code{@var{b} = __MABSHS (@var{a})}
6522 @tab @code{MABSHS @var{a},@var{b}}
6523 @item @code{void __MADDACCS (acc, acc)}
6524 @tab @code{__MADDACCS (@var{b}, @var{a})}
6525 @tab @code{MADDACCS @var{a},@var{b}}
6526 @item @code{sw1 __MADDHSS (sw1, sw1)}
6527 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6528 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6529 @item @code{uw1 __MADDHUS (uw1, uw1)}
6530 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6531 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6532 @item @code{uw1 __MAND (uw1, uw1)}
6533 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6534 @tab @code{MAND @var{a},@var{b},@var{c}}
6535 @item @code{void __MASACCS (acc, acc)}
6536 @tab @code{__MASACCS (@var{b}, @var{a})}
6537 @tab @code{MASACCS @var{a},@var{b}}
6538 @item @code{uw1 __MAVEH (uw1, uw1)}
6539 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6540 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6541 @item @code{uw2 __MBTOH (uw1)}
6542 @tab @code{@var{b} = __MBTOH (@var{a})}
6543 @tab @code{MBTOH @var{a},@var{b}}
6544 @item @code{void __MBTOHE (uw1 *, uw1)}
6545 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6546 @tab @code{MBTOHE @var{a},@var{b}}
6547 @item @code{void __MCLRACC (acc)}
6548 @tab @code{__MCLRACC (@var{a})}
6549 @tab @code{MCLRACC @var{a}}
6550 @item @code{void __MCLRACCA (void)}
6551 @tab @code{__MCLRACCA ()}
6552 @tab @code{MCLRACCA}
6553 @item @code{uw1 __Mcop1 (uw1, uw1)}
6554 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6555 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6556 @item @code{uw1 __Mcop2 (uw1, uw1)}
6557 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6558 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6559 @item @code{uw1 __MCPLHI (uw2, const)}
6560 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6561 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6562 @item @code{uw1 __MCPLI (uw2, const)}
6563 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6564 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6565 @item @code{void __MCPXIS (acc, sw1, sw1)}
6566 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6567 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6568 @item @code{void __MCPXIU (acc, uw1, uw1)}
6569 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6570 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6571 @item @code{void __MCPXRS (acc, sw1, sw1)}
6572 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6573 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6574 @item @code{void __MCPXRU (acc, uw1, uw1)}
6575 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6576 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6577 @item @code{uw1 __MCUT (acc, uw1)}
6578 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6579 @tab @code{MCUT @var{a},@var{b},@var{c}}
6580 @item @code{uw1 __MCUTSS (acc, sw1)}
6581 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6582 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6583 @item @code{void __MDADDACCS (acc, acc)}
6584 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6585 @tab @code{MDADDACCS @var{a},@var{b}}
6586 @item @code{void __MDASACCS (acc, acc)}
6587 @tab @code{__MDASACCS (@var{b}, @var{a})}
6588 @tab @code{MDASACCS @var{a},@var{b}}
6589 @item @code{uw2 __MDCUTSSI (acc, const)}
6590 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6591 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6592 @item @code{uw2 __MDPACKH (uw2, uw2)}
6593 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6594 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6595 @item @code{uw2 __MDROTLI (uw2, const)}
6596 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6597 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6598 @item @code{void __MDSUBACCS (acc, acc)}
6599 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6600 @tab @code{MDSUBACCS @var{a},@var{b}}
6601 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6602 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6603 @tab @code{MDUNPACKH @var{a},@var{b}}
6604 @item @code{uw2 __MEXPDHD (uw1, const)}
6605 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6606 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6607 @item @code{uw1 __MEXPDHW (uw1, const)}
6608 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6609 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6610 @item @code{uw1 __MHDSETH (uw1, const)}
6611 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6612 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6613 @item @code{sw1 __MHDSETS (const)}
6614 @tab @code{@var{b} = __MHDSETS (@var{a})}
6615 @tab @code{MHDSETS #@var{a},@var{b}}
6616 @item @code{uw1 __MHSETHIH (uw1, const)}
6617 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6618 @tab @code{MHSETHIH #@var{a},@var{b}}
6619 @item @code{sw1 __MHSETHIS (sw1, const)}
6620 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6621 @tab @code{MHSETHIS #@var{a},@var{b}}
6622 @item @code{uw1 __MHSETLOH (uw1, const)}
6623 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6624 @tab @code{MHSETLOH #@var{a},@var{b}}
6625 @item @code{sw1 __MHSETLOS (sw1, const)}
6626 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6627 @tab @code{MHSETLOS #@var{a},@var{b}}
6628 @item @code{uw1 __MHTOB (uw2)}
6629 @tab @code{@var{b} = __MHTOB (@var{a})}
6630 @tab @code{MHTOB @var{a},@var{b}}
6631 @item @code{void __MMACHS (acc, sw1, sw1)}
6632 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6633 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6634 @item @code{void __MMACHU (acc, uw1, uw1)}
6635 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6636 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6637 @item @code{void __MMRDHS (acc, sw1, sw1)}
6638 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6639 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6640 @item @code{void __MMRDHU (acc, uw1, uw1)}
6641 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6642 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6643 @item @code{void __MMULHS (acc, sw1, sw1)}
6644 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6645 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6646 @item @code{void __MMULHU (acc, uw1, uw1)}
6647 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6648 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6649 @item @code{void __MMULXHS (acc, sw1, sw1)}
6650 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6651 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6652 @item @code{void __MMULXHU (acc, uw1, uw1)}
6653 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6654 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6655 @item @code{uw1 __MNOT (uw1)}
6656 @tab @code{@var{b} = __MNOT (@var{a})}
6657 @tab @code{MNOT @var{a},@var{b}}
6658 @item @code{uw1 __MOR (uw1, uw1)}
6659 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6660 @tab @code{MOR @var{a},@var{b},@var{c}}
6661 @item @code{uw1 __MPACKH (uh, uh)}
6662 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6663 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6664 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6665 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6666 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6667 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6668 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6669 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6670 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6671 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6672 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6673 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6674 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6675 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6676 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6677 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6678 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6679 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6680 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6681 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6682 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6683 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6684 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6685 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6686 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6687 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6688 @item @code{void __MQMACHS (acc, sw2, sw2)}
6689 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6690 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6691 @item @code{void __MQMACHU (acc, uw2, uw2)}
6692 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6693 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6694 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6695 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6696 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6697 @item @code{void __MQMULHS (acc, sw2, sw2)}
6698 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6699 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6700 @item @code{void __MQMULHU (acc, uw2, uw2)}
6701 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6702 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6703 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6704 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6705 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6706 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6707 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6708 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6709 @item @code{sw2 __MQSATHS (sw2, sw2)}
6710 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6711 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6712 @item @code{uw2 __MQSLLHI (uw2, int)}
6713 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6714 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6715 @item @code{sw2 __MQSRAHI (sw2, int)}
6716 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6717 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6718 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6719 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6720 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6721 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6722 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6723 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6724 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6725 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6726 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6727 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6728 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6729 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6730 @item @code{uw1 __MRDACC (acc)}
6731 @tab @code{@var{b} = __MRDACC (@var{a})}
6732 @tab @code{MRDACC @var{a},@var{b}}
6733 @item @code{uw1 __MRDACCG (acc)}
6734 @tab @code{@var{b} = __MRDACCG (@var{a})}
6735 @tab @code{MRDACCG @var{a},@var{b}}
6736 @item @code{uw1 __MROTLI (uw1, const)}
6737 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6738 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6739 @item @code{uw1 __MROTRI (uw1, const)}
6740 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6741 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6742 @item @code{sw1 __MSATHS (sw1, sw1)}
6743 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6744 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6745 @item @code{uw1 __MSATHU (uw1, uw1)}
6746 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6747 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6748 @item @code{uw1 __MSLLHI (uw1, const)}
6749 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6750 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6751 @item @code{sw1 __MSRAHI (sw1, const)}
6752 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6753 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6754 @item @code{uw1 __MSRLHI (uw1, const)}
6755 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6756 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6757 @item @code{void __MSUBACCS (acc, acc)}
6758 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6759 @tab @code{MSUBACCS @var{a},@var{b}}
6760 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6761 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6762 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6763 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6764 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6765 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6766 @item @code{void __MTRAP (void)}
6767 @tab @code{__MTRAP ()}
6769 @item @code{uw2 __MUNPACKH (uw1)}
6770 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6771 @tab @code{MUNPACKH @var{a},@var{b}}
6772 @item @code{uw1 __MWCUT (uw2, uw1)}
6773 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6774 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6775 @item @code{void __MWTACC (acc, uw1)}
6776 @tab @code{__MWTACC (@var{b}, @var{a})}
6777 @tab @code{MWTACC @var{a},@var{b}}
6778 @item @code{void __MWTACCG (acc, uw1)}
6779 @tab @code{__MWTACCG (@var{b}, @var{a})}
6780 @tab @code{MWTACCG @var{a},@var{b}}
6781 @item @code{uw1 __MXOR (uw1, uw1)}
6782 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6783 @tab @code{MXOR @var{a},@var{b},@var{c}}
6786 @node Raw read/write Functions
6787 @subsubsection Raw read/write Functions
6789 This sections describes built-in functions related to read and write
6790 instructions to access memory. These functions generate
6791 @code{membar} instructions to flush the I/O load and stores where
6792 appropriate, as described in Fujitsu's manual described above.
6796 @item unsigned char __builtin_read8 (void *@var{data})
6797 @item unsigned short __builtin_read16 (void *@var{data})
6798 @item unsigned long __builtin_read32 (void *@var{data})
6799 @item unsigned long long __builtin_read64 (void *@var{data})
6801 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6802 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6803 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6804 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6807 @node Other Built-in Functions
6808 @subsubsection Other Built-in Functions
6810 This section describes built-in functions that are not named after
6811 a specific FR-V instruction.
6814 @item sw2 __IACCreadll (iacc @var{reg})
6815 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6816 for future expansion and must be 0.
6818 @item sw1 __IACCreadl (iacc @var{reg})
6819 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6820 Other values of @var{reg} are rejected as invalid.
6822 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6823 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6824 is reserved for future expansion and must be 0.
6826 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6827 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6828 is 1. Other values of @var{reg} are rejected as invalid.
6830 @item void __data_prefetch0 (const void *@var{x})
6831 Use the @code{dcpl} instruction to load the contents of address @var{x}
6832 into the data cache.
6834 @item void __data_prefetch (const void *@var{x})
6835 Use the @code{nldub} instruction to load the contents of address @var{x}
6836 into the data cache. The instruction will be issued in slot I1@.
6839 @node X86 Built-in Functions
6840 @subsection X86 Built-in Functions
6842 These built-in functions are available for the i386 and x86-64 family
6843 of computers, depending on the command-line switches used.
6845 Note that, if you specify command-line switches such as @option{-msse},
6846 the compiler could use the extended instruction sets even if the built-ins
6847 are not used explicitly in the program. For this reason, applications
6848 which perform runtime CPU detection must compile separate files for each
6849 supported architecture, using the appropriate flags. In particular,
6850 the file containing the CPU detection code should be compiled without
6853 The following machine modes are available for use with MMX built-in functions
6854 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6855 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6856 vector of eight 8-bit integers. Some of the built-in functions operate on
6857 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6859 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6860 of two 32-bit floating point values.
6862 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6863 floating point values. Some instructions use a vector of four 32-bit
6864 integers, these use @code{V4SI}. Finally, some instructions operate on an
6865 entire vector register, interpreting it as a 128-bit integer, these use mode
6868 The following built-in functions are made available by @option{-mmmx}.
6869 All of them generate the machine instruction that is part of the name.
6872 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6873 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6874 v2si __builtin_ia32_paddd (v2si, v2si)
6875 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6876 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6877 v2si __builtin_ia32_psubd (v2si, v2si)
6878 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6879 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6880 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6881 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6882 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6883 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6884 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6885 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6886 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6887 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6888 di __builtin_ia32_pand (di, di)
6889 di __builtin_ia32_pandn (di,di)
6890 di __builtin_ia32_por (di, di)
6891 di __builtin_ia32_pxor (di, di)
6892 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6893 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6894 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6895 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6896 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6897 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6898 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6899 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6900 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6901 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6902 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6903 v2si __builtin_ia32_punpckldq (v2si, v2si)
6904 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6905 v4hi __builtin_ia32_packssdw (v2si, v2si)
6906 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6909 The following built-in functions are made available either with
6910 @option{-msse}, or with a combination of @option{-m3dnow} and
6911 @option{-march=athlon}. All of them generate the machine
6912 instruction that is part of the name.
6915 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6916 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6917 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6918 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6919 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6920 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6921 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6922 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6923 int __builtin_ia32_pextrw (v4hi, int)
6924 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6925 int __builtin_ia32_pmovmskb (v8qi)
6926 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6927 void __builtin_ia32_movntq (di *, di)
6928 void __builtin_ia32_sfence (void)
6931 The following built-in functions are available when @option{-msse} is used.
6932 All of them generate the machine instruction that is part of the name.
6935 int __builtin_ia32_comieq (v4sf, v4sf)
6936 int __builtin_ia32_comineq (v4sf, v4sf)
6937 int __builtin_ia32_comilt (v4sf, v4sf)
6938 int __builtin_ia32_comile (v4sf, v4sf)
6939 int __builtin_ia32_comigt (v4sf, v4sf)
6940 int __builtin_ia32_comige (v4sf, v4sf)
6941 int __builtin_ia32_ucomieq (v4sf, v4sf)
6942 int __builtin_ia32_ucomineq (v4sf, v4sf)
6943 int __builtin_ia32_ucomilt (v4sf, v4sf)
6944 int __builtin_ia32_ucomile (v4sf, v4sf)
6945 int __builtin_ia32_ucomigt (v4sf, v4sf)
6946 int __builtin_ia32_ucomige (v4sf, v4sf)
6947 v4sf __builtin_ia32_addps (v4sf, v4sf)
6948 v4sf __builtin_ia32_subps (v4sf, v4sf)
6949 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6950 v4sf __builtin_ia32_divps (v4sf, v4sf)
6951 v4sf __builtin_ia32_addss (v4sf, v4sf)
6952 v4sf __builtin_ia32_subss (v4sf, v4sf)
6953 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6954 v4sf __builtin_ia32_divss (v4sf, v4sf)
6955 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6956 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6957 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6958 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6959 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6960 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6961 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6962 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6963 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6964 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6965 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6966 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6967 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6968 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6969 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6970 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6971 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6972 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6973 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6974 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6975 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6976 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6977 v4sf __builtin_ia32_minps (v4sf, v4sf)
6978 v4sf __builtin_ia32_minss (v4sf, v4sf)
6979 v4sf __builtin_ia32_andps (v4sf, v4sf)
6980 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6981 v4sf __builtin_ia32_orps (v4sf, v4sf)
6982 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6983 v4sf __builtin_ia32_movss (v4sf, v4sf)
6984 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6985 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6986 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6987 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6988 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6989 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6990 v2si __builtin_ia32_cvtps2pi (v4sf)
6991 int __builtin_ia32_cvtss2si (v4sf)
6992 v2si __builtin_ia32_cvttps2pi (v4sf)
6993 int __builtin_ia32_cvttss2si (v4sf)
6994 v4sf __builtin_ia32_rcpps (v4sf)
6995 v4sf __builtin_ia32_rsqrtps (v4sf)
6996 v4sf __builtin_ia32_sqrtps (v4sf)
6997 v4sf __builtin_ia32_rcpss (v4sf)
6998 v4sf __builtin_ia32_rsqrtss (v4sf)
6999 v4sf __builtin_ia32_sqrtss (v4sf)
7000 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7001 void __builtin_ia32_movntps (float *, v4sf)
7002 int __builtin_ia32_movmskps (v4sf)
7005 The following built-in functions are available when @option{-msse} is used.
7008 @item v4sf __builtin_ia32_loadaps (float *)
7009 Generates the @code{movaps} machine instruction as a load from memory.
7010 @item void __builtin_ia32_storeaps (float *, v4sf)
7011 Generates the @code{movaps} machine instruction as a store to memory.
7012 @item v4sf __builtin_ia32_loadups (float *)
7013 Generates the @code{movups} machine instruction as a load from memory.
7014 @item void __builtin_ia32_storeups (float *, v4sf)
7015 Generates the @code{movups} machine instruction as a store to memory.
7016 @item v4sf __builtin_ia32_loadsss (float *)
7017 Generates the @code{movss} machine instruction as a load from memory.
7018 @item void __builtin_ia32_storess (float *, v4sf)
7019 Generates the @code{movss} machine instruction as a store to memory.
7020 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7021 Generates the @code{movhps} machine instruction as a load from memory.
7022 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7023 Generates the @code{movlps} machine instruction as a load from memory
7024 @item void __builtin_ia32_storehps (v4sf, v2si *)
7025 Generates the @code{movhps} machine instruction as a store to memory.
7026 @item void __builtin_ia32_storelps (v4sf, v2si *)
7027 Generates the @code{movlps} machine instruction as a store to memory.
7030 The following built-in functions are available when @option{-msse2} is used.
7031 All of them generate the machine instruction that is part of the name.
7034 int __builtin_ia32_comisdeq (v2df, v2df)
7035 int __builtin_ia32_comisdlt (v2df, v2df)
7036 int __builtin_ia32_comisdle (v2df, v2df)
7037 int __builtin_ia32_comisdgt (v2df, v2df)
7038 int __builtin_ia32_comisdge (v2df, v2df)
7039 int __builtin_ia32_comisdneq (v2df, v2df)
7040 int __builtin_ia32_ucomisdeq (v2df, v2df)
7041 int __builtin_ia32_ucomisdlt (v2df, v2df)
7042 int __builtin_ia32_ucomisdle (v2df, v2df)
7043 int __builtin_ia32_ucomisdgt (v2df, v2df)
7044 int __builtin_ia32_ucomisdge (v2df, v2df)
7045 int __builtin_ia32_ucomisdneq (v2df, v2df)
7046 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7047 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7048 v2df __builtin_ia32_cmplepd (v2df, v2df)
7049 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7050 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7051 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7052 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7053 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7054 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7055 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7056 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7057 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7058 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7059 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7060 v2df __builtin_ia32_cmplesd (v2df, v2df)
7061 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7062 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7063 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7064 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7065 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7066 v2di __builtin_ia32_paddq (v2di, v2di)
7067 v2di __builtin_ia32_psubq (v2di, v2di)
7068 v2df __builtin_ia32_addpd (v2df, v2df)
7069 v2df __builtin_ia32_subpd (v2df, v2df)
7070 v2df __builtin_ia32_mulpd (v2df, v2df)
7071 v2df __builtin_ia32_divpd (v2df, v2df)
7072 v2df __builtin_ia32_addsd (v2df, v2df)
7073 v2df __builtin_ia32_subsd (v2df, v2df)
7074 v2df __builtin_ia32_mulsd (v2df, v2df)
7075 v2df __builtin_ia32_divsd (v2df, v2df)
7076 v2df __builtin_ia32_minpd (v2df, v2df)
7077 v2df __builtin_ia32_maxpd (v2df, v2df)
7078 v2df __builtin_ia32_minsd (v2df, v2df)
7079 v2df __builtin_ia32_maxsd (v2df, v2df)
7080 v2df __builtin_ia32_andpd (v2df, v2df)
7081 v2df __builtin_ia32_andnpd (v2df, v2df)
7082 v2df __builtin_ia32_orpd (v2df, v2df)
7083 v2df __builtin_ia32_xorpd (v2df, v2df)
7084 v2df __builtin_ia32_movsd (v2df, v2df)
7085 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7086 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7087 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7088 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7089 v4si __builtin_ia32_paddd128 (v4si, v4si)
7090 v2di __builtin_ia32_paddq128 (v2di, v2di)
7091 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7092 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7093 v4si __builtin_ia32_psubd128 (v4si, v4si)
7094 v2di __builtin_ia32_psubq128 (v2di, v2di)
7095 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7096 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7097 v2di __builtin_ia32_pand128 (v2di, v2di)
7098 v2di __builtin_ia32_pandn128 (v2di, v2di)
7099 v2di __builtin_ia32_por128 (v2di, v2di)
7100 v2di __builtin_ia32_pxor128 (v2di, v2di)
7101 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7102 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7103 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7104 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7105 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7106 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7107 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7108 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7109 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7110 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7111 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7112 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7113 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7114 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7115 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7116 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7117 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7118 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7119 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7120 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7121 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7122 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7123 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7124 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7125 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7126 v2df __builtin_ia32_loadupd (double *)
7127 void __builtin_ia32_storeupd (double *, v2df)
7128 v2df __builtin_ia32_loadhpd (v2df, double *)
7129 v2df __builtin_ia32_loadlpd (v2df, double *)
7130 int __builtin_ia32_movmskpd (v2df)
7131 int __builtin_ia32_pmovmskb128 (v16qi)
7132 void __builtin_ia32_movnti (int *, int)
7133 void __builtin_ia32_movntpd (double *, v2df)
7134 void __builtin_ia32_movntdq (v2df *, v2df)
7135 v4si __builtin_ia32_pshufd (v4si, int)
7136 v8hi __builtin_ia32_pshuflw (v8hi, int)
7137 v8hi __builtin_ia32_pshufhw (v8hi, int)
7138 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7139 v2df __builtin_ia32_sqrtpd (v2df)
7140 v2df __builtin_ia32_sqrtsd (v2df)
7141 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7142 v2df __builtin_ia32_cvtdq2pd (v4si)
7143 v4sf __builtin_ia32_cvtdq2ps (v4si)
7144 v4si __builtin_ia32_cvtpd2dq (v2df)
7145 v2si __builtin_ia32_cvtpd2pi (v2df)
7146 v4sf __builtin_ia32_cvtpd2ps (v2df)
7147 v4si __builtin_ia32_cvttpd2dq (v2df)
7148 v2si __builtin_ia32_cvttpd2pi (v2df)
7149 v2df __builtin_ia32_cvtpi2pd (v2si)
7150 int __builtin_ia32_cvtsd2si (v2df)
7151 int __builtin_ia32_cvttsd2si (v2df)
7152 long long __builtin_ia32_cvtsd2si64 (v2df)
7153 long long __builtin_ia32_cvttsd2si64 (v2df)
7154 v4si __builtin_ia32_cvtps2dq (v4sf)
7155 v2df __builtin_ia32_cvtps2pd (v4sf)
7156 v4si __builtin_ia32_cvttps2dq (v4sf)
7157 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7158 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7159 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7160 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7161 void __builtin_ia32_clflush (const void *)
7162 void __builtin_ia32_lfence (void)
7163 void __builtin_ia32_mfence (void)
7164 v16qi __builtin_ia32_loaddqu (const char *)
7165 void __builtin_ia32_storedqu (char *, v16qi)
7166 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7167 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7168 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7169 v4si __builtin_ia32_pslld128 (v4si, v2di)
7170 v2di __builtin_ia32_psllq128 (v4si, v2di)
7171 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7172 v4si __builtin_ia32_psrld128 (v4si, v2di)
7173 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7174 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7175 v4si __builtin_ia32_psrad128 (v4si, v2di)
7176 v2di __builtin_ia32_pslldqi128 (v2di, int)
7177 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7178 v4si __builtin_ia32_pslldi128 (v4si, int)
7179 v2di __builtin_ia32_psllqi128 (v2di, int)
7180 v2di __builtin_ia32_psrldqi128 (v2di, int)
7181 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7182 v4si __builtin_ia32_psrldi128 (v4si, int)
7183 v2di __builtin_ia32_psrlqi128 (v2di, int)
7184 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7185 v4si __builtin_ia32_psradi128 (v4si, int)
7186 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7189 The following built-in functions are available when @option{-msse3} is used.
7190 All of them generate the machine instruction that is part of the name.
7193 v2df __builtin_ia32_addsubpd (v2df, v2df)
7194 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7195 v2df __builtin_ia32_haddpd (v2df, v2df)
7196 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7197 v2df __builtin_ia32_hsubpd (v2df, v2df)
7198 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7199 v16qi __builtin_ia32_lddqu (char const *)
7200 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7201 v2df __builtin_ia32_movddup (v2df)
7202 v4sf __builtin_ia32_movshdup (v4sf)
7203 v4sf __builtin_ia32_movsldup (v4sf)
7204 void __builtin_ia32_mwait (unsigned int, unsigned int)
7207 The following built-in functions are available when @option{-msse3} is used.
7210 @item v2df __builtin_ia32_loadddup (double const *)
7211 Generates the @code{movddup} machine instruction as a load from memory.
7214 The following built-in functions are available when @option{-m3dnow} is used.
7215 All of them generate the machine instruction that is part of the name.
7218 void __builtin_ia32_femms (void)
7219 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7220 v2si __builtin_ia32_pf2id (v2sf)
7221 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7222 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7223 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7224 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7225 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7226 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7227 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7228 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7229 v2sf __builtin_ia32_pfrcp (v2sf)
7230 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7231 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7232 v2sf __builtin_ia32_pfrsqrt (v2sf)
7233 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7234 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7235 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7236 v2sf __builtin_ia32_pi2fd (v2si)
7237 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7240 The following built-in functions are available when both @option{-m3dnow}
7241 and @option{-march=athlon} are used. All of them generate the machine
7242 instruction that is part of the name.
7245 v2si __builtin_ia32_pf2iw (v2sf)
7246 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7247 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7248 v2sf __builtin_ia32_pi2fw (v2si)
7249 v2sf __builtin_ia32_pswapdsf (v2sf)
7250 v2si __builtin_ia32_pswapdsi (v2si)
7253 @node MIPS DSP Built-in Functions
7254 @subsection MIPS DSP Built-in Functions
7256 The MIPS DSP Application-Specific Extension (ASE) includes new
7257 instructions that are designed to improve the performance of DSP and
7258 media applications. It provides instructions that operate on packed
7259 8-bit integer data, Q15 fractional data and Q31 fractional data.
7261 GCC supports MIPS DSP operations using both the generic
7262 vector extensions (@pxref{Vector Extensions}) and a collection of
7263 MIPS-specific built-in functions. Both kinds of support are
7264 enabled by the @option{-mdsp} command-line option.
7266 At present, GCC only provides support for operations on 32-bit
7267 vectors. The vector type associated with 8-bit integer data is
7268 usually called @code{v4i8} and the vector type associated with Q15 is
7269 usually called @code{v2q15}. They can be defined in C as follows:
7272 typedef char v4i8 __attribute__ ((vector_size(4)));
7273 typedef short v2q15 __attribute__ ((vector_size(4)));
7276 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7277 aggregates. For example:
7280 v4i8 a = @{1, 2, 3, 4@};
7282 b = (v4i8) @{5, 6, 7, 8@};
7284 v2q15 c = @{0x0fcb, 0x3a75@};
7286 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7289 @emph{Note:} The CPU's endianness determines the order in which values
7290 are packed. On little-endian targets, the first value is the least
7291 significant and the last value is the most significant. The opposite
7292 order applies to big-endian targets. For example, the code above will
7293 set the lowest byte of @code{a} to @code{1} on little-endian targets
7294 and @code{4} on big-endian targets.
7296 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7297 representation. As shown in this example, the integer representation
7298 of a Q15 value can be obtained by multiplying the fractional value by
7299 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7302 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7303 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7304 and @code{c} and @code{d} are @code{v2q15} values.
7306 @multitable @columnfractions .50 .50
7307 @item C code @tab MIPS instruction
7308 @item @code{a + b} @tab @code{addu.qb}
7309 @item @code{c + d} @tab @code{addq.ph}
7310 @item @code{a - b} @tab @code{subu.qb}
7311 @item @code{c - d} @tab @code{subq.ph}
7314 It is easier to describe the DSP built-in functions if we first define
7315 the following types:
7320 typedef long long a64;
7323 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7324 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7325 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7326 @code{long long}, but we use @code{a64} to indicate values that will
7327 be placed in one of the four DSP accumulators (@code{$ac0},
7328 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7330 Also, some built-in functions prefer or require immediate numbers as
7331 parameters, because the corresponding DSP instructions accept both immediate
7332 numbers and register operands, or accept immediate numbers only. The
7333 immediate parameters are listed as follows.
7341 imm_n32_31: -32 to 31.
7342 imm_n512_511: -512 to 511.
7345 The following built-in functions map directly to a particular MIPS DSP
7346 instruction. Please refer to the architecture specification
7347 for details on what each instruction does.
7350 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7351 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7352 q31 __builtin_mips_addq_s_w (q31, q31)
7353 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7354 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7355 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7356 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7357 q31 __builtin_mips_subq_s_w (q31, q31)
7358 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7359 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7360 i32 __builtin_mips_addsc (i32, i32)
7361 i32 __builtin_mips_addwc (i32, i32)
7362 i32 __builtin_mips_modsub (i32, i32)
7363 i32 __builtin_mips_raddu_w_qb (v4i8)
7364 v2q15 __builtin_mips_absq_s_ph (v2q15)
7365 q31 __builtin_mips_absq_s_w (q31)
7366 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7367 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7368 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7369 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7370 q31 __builtin_mips_preceq_w_phl (v2q15)
7371 q31 __builtin_mips_preceq_w_phr (v2q15)
7372 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7373 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7374 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7375 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7376 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7377 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7378 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7379 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7380 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7381 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7382 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7383 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7384 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7385 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7386 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7387 q31 __builtin_mips_shll_s_w (q31, i32)
7388 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7389 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7390 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7391 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7392 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7393 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7394 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7395 q31 __builtin_mips_shra_r_w (q31, i32)
7396 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7397 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7398 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7399 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7400 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7401 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7402 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7403 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7404 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7405 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7406 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7407 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7408 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7409 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7410 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7411 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7412 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7413 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7414 i32 __builtin_mips_bitrev (i32)
7415 i32 __builtin_mips_insv (i32, i32)
7416 v4i8 __builtin_mips_repl_qb (imm0_255)
7417 v4i8 __builtin_mips_repl_qb (i32)
7418 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7419 v2q15 __builtin_mips_repl_ph (i32)
7420 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7421 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7422 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7423 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7424 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7425 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7426 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7427 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7428 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7429 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7430 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7431 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7432 i32 __builtin_mips_extr_w (a64, imm0_31)
7433 i32 __builtin_mips_extr_w (a64, i32)
7434 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7435 i32 __builtin_mips_extr_s_h (a64, i32)
7436 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7437 i32 __builtin_mips_extr_rs_w (a64, i32)
7438 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7439 i32 __builtin_mips_extr_r_w (a64, i32)
7440 i32 __builtin_mips_extp (a64, imm0_31)
7441 i32 __builtin_mips_extp (a64, i32)
7442 i32 __builtin_mips_extpdp (a64, imm0_31)
7443 i32 __builtin_mips_extpdp (a64, i32)
7444 a64 __builtin_mips_shilo (a64, imm_n32_31)
7445 a64 __builtin_mips_shilo (a64, i32)
7446 a64 __builtin_mips_mthlip (a64, i32)
7447 void __builtin_mips_wrdsp (i32, imm0_63)
7448 i32 __builtin_mips_rddsp (imm0_63)
7449 i32 __builtin_mips_lbux (void *, i32)
7450 i32 __builtin_mips_lhx (void *, i32)
7451 i32 __builtin_mips_lwx (void *, i32)
7452 i32 __builtin_mips_bposge32 (void)
7455 @node MIPS Paired-Single Support
7456 @subsection MIPS Paired-Single Support
7458 The MIPS64 architecture includes a number of instructions that
7459 operate on pairs of single-precision floating-point values.
7460 Each pair is packed into a 64-bit floating-point register,
7461 with one element being designated the ``upper half'' and
7462 the other being designated the ``lower half''.
7464 GCC supports paired-single operations using both the generic
7465 vector extensions (@pxref{Vector Extensions}) and a collection of
7466 MIPS-specific built-in functions. Both kinds of support are
7467 enabled by the @option{-mpaired-single} command-line option.
7469 The vector type associated with paired-single values is usually
7470 called @code{v2sf}. It can be defined in C as follows:
7473 typedef float v2sf __attribute__ ((vector_size (8)));
7476 @code{v2sf} values are initialized in the same way as aggregates.
7480 v2sf a = @{1.5, 9.1@};
7483 b = (v2sf) @{e, f@};
7486 @emph{Note:} The CPU's endianness determines which value is stored in
7487 the upper half of a register and which value is stored in the lower half.
7488 On little-endian targets, the first value is the lower one and the second
7489 value is the upper one. The opposite order applies to big-endian targets.
7490 For example, the code above will set the lower half of @code{a} to
7491 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7494 * Paired-Single Arithmetic::
7495 * Paired-Single Built-in Functions::
7496 * MIPS-3D Built-in Functions::
7499 @node Paired-Single Arithmetic
7500 @subsubsection Paired-Single Arithmetic
7502 The table below lists the @code{v2sf} operations for which hardware
7503 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7504 values and @code{x} is an integral value.
7506 @multitable @columnfractions .50 .50
7507 @item C code @tab MIPS instruction
7508 @item @code{a + b} @tab @code{add.ps}
7509 @item @code{a - b} @tab @code{sub.ps}
7510 @item @code{-a} @tab @code{neg.ps}
7511 @item @code{a * b} @tab @code{mul.ps}
7512 @item @code{a * b + c} @tab @code{madd.ps}
7513 @item @code{a * b - c} @tab @code{msub.ps}
7514 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7515 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7516 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7519 Note that the multiply-accumulate instructions can be disabled
7520 using the command-line option @code{-mno-fused-madd}.
7522 @node Paired-Single Built-in Functions
7523 @subsubsection Paired-Single Built-in Functions
7525 The following paired-single functions map directly to a particular
7526 MIPS instruction. Please refer to the architecture specification
7527 for details on what each instruction does.
7530 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7531 Pair lower lower (@code{pll.ps}).
7533 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7534 Pair upper lower (@code{pul.ps}).
7536 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7537 Pair lower upper (@code{plu.ps}).
7539 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7540 Pair upper upper (@code{puu.ps}).
7542 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7543 Convert pair to paired single (@code{cvt.ps.s}).
7545 @item float __builtin_mips_cvt_s_pl (v2sf)
7546 Convert pair lower to single (@code{cvt.s.pl}).
7548 @item float __builtin_mips_cvt_s_pu (v2sf)
7549 Convert pair upper to single (@code{cvt.s.pu}).
7551 @item v2sf __builtin_mips_abs_ps (v2sf)
7552 Absolute value (@code{abs.ps}).
7554 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7555 Align variable (@code{alnv.ps}).
7557 @emph{Note:} The value of the third parameter must be 0 or 4
7558 modulo 8, otherwise the result will be unpredictable. Please read the
7559 instruction description for details.
7562 The following multi-instruction functions are also available.
7563 In each case, @var{cond} can be any of the 16 floating-point conditions:
7564 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7565 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7566 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7569 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7570 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7571 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7572 @code{movt.ps}/@code{movf.ps}).
7574 The @code{movt} functions return the value @var{x} computed by:
7577 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7578 mov.ps @var{x},@var{c}
7579 movt.ps @var{x},@var{d},@var{cc}
7582 The @code{movf} functions are similar but use @code{movf.ps} instead
7585 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7586 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7587 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7588 @code{bc1t}/@code{bc1f}).
7590 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7591 and return either the upper or lower half of the result. For example:
7595 if (__builtin_mips_upper_c_eq_ps (a, b))
7596 upper_halves_are_equal ();
7598 upper_halves_are_unequal ();
7600 if (__builtin_mips_lower_c_eq_ps (a, b))
7601 lower_halves_are_equal ();
7603 lower_halves_are_unequal ();
7607 @node MIPS-3D Built-in Functions
7608 @subsubsection MIPS-3D Built-in Functions
7610 The MIPS-3D Application-Specific Extension (ASE) includes additional
7611 paired-single instructions that are designed to improve the performance
7612 of 3D graphics operations. Support for these instructions is controlled
7613 by the @option{-mips3d} command-line option.
7615 The functions listed below map directly to a particular MIPS-3D
7616 instruction. Please refer to the architecture specification for
7617 more details on what each instruction does.
7620 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7621 Reduction add (@code{addr.ps}).
7623 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7624 Reduction multiply (@code{mulr.ps}).
7626 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7627 Convert paired single to paired word (@code{cvt.pw.ps}).
7629 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7630 Convert paired word to paired single (@code{cvt.ps.pw}).
7632 @item float __builtin_mips_recip1_s (float)
7633 @itemx double __builtin_mips_recip1_d (double)
7634 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7635 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7637 @item float __builtin_mips_recip2_s (float, float)
7638 @itemx double __builtin_mips_recip2_d (double, double)
7639 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7640 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7642 @item float __builtin_mips_rsqrt1_s (float)
7643 @itemx double __builtin_mips_rsqrt1_d (double)
7644 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7645 Reduced precision reciprocal square root (sequence step 1)
7646 (@code{rsqrt1.@var{fmt}}).
7648 @item float __builtin_mips_rsqrt2_s (float, float)
7649 @itemx double __builtin_mips_rsqrt2_d (double, double)
7650 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7651 Reduced precision reciprocal square root (sequence step 2)
7652 (@code{rsqrt2.@var{fmt}}).
7655 The following multi-instruction functions are also available.
7656 In each case, @var{cond} can be any of the 16 floating-point conditions:
7657 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7658 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7659 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7662 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7663 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7664 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7665 @code{bc1t}/@code{bc1f}).
7667 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7668 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7673 if (__builtin_mips_cabs_eq_s (a, b))
7679 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7680 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7681 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7682 @code{bc1t}/@code{bc1f}).
7684 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7685 and return either the upper or lower half of the result. For example:
7689 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7690 upper_halves_are_equal ();
7692 upper_halves_are_unequal ();
7694 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7695 lower_halves_are_equal ();
7697 lower_halves_are_unequal ();
7700 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7701 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7702 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7703 @code{movt.ps}/@code{movf.ps}).
7705 The @code{movt} functions return the value @var{x} computed by:
7708 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7709 mov.ps @var{x},@var{c}
7710 movt.ps @var{x},@var{d},@var{cc}
7713 The @code{movf} functions are similar but use @code{movf.ps} instead
7716 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7717 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7718 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7719 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7720 Comparison of two paired-single values
7721 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7722 @code{bc1any2t}/@code{bc1any2f}).
7724 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7725 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7726 result is true and the @code{all} forms return true if both results are true.
7731 if (__builtin_mips_any_c_eq_ps (a, b))
7736 if (__builtin_mips_all_c_eq_ps (a, b))
7742 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7743 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7744 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7745 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7746 Comparison of four paired-single values
7747 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7748 @code{bc1any4t}/@code{bc1any4f}).
7750 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7751 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7752 The @code{any} forms return true if any of the four results are true
7753 and the @code{all} forms return true if all four results are true.
7758 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7763 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7770 @node PowerPC AltiVec Built-in Functions
7771 @subsection PowerPC AltiVec Built-in Functions
7773 GCC provides an interface for the PowerPC family of processors to access
7774 the AltiVec operations described in Motorola's AltiVec Programming
7775 Interface Manual. The interface is made available by including
7776 @code{<altivec.h>} and using @option{-maltivec} and
7777 @option{-mabi=altivec}. The interface supports the following vector
7781 vector unsigned char
7785 vector unsigned short
7796 GCC's implementation of the high-level language interface available from
7797 C and C++ code differs from Motorola's documentation in several ways.
7802 A vector constant is a list of constant expressions within curly braces.
7805 A vector initializer requires no cast if the vector constant is of the
7806 same type as the variable it is initializing.
7809 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7810 vector type is the default signedness of the base type. The default
7811 varies depending on the operating system, so a portable program should
7812 always specify the signedness.
7815 Compiling with @option{-maltivec} adds keywords @code{__vector},
7816 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7817 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7821 GCC allows using a @code{typedef} name as the type specifier for a
7825 For C, overloaded functions are implemented with macros so the following
7829 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7832 Since @code{vec_add} is a macro, the vector constant in the example
7833 is treated as four separate arguments. Wrap the entire argument in
7834 parentheses for this to work.
7837 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
7838 Internally, GCC uses built-in functions to achieve the functionality in
7839 the aforementioned header file, but they are not supported and are
7840 subject to change without notice.
7842 The following interfaces are supported for the generic and specific
7843 AltiVec operations and the AltiVec predicates. In cases where there
7844 is a direct mapping between generic and specific operations, only the
7845 generic names are shown here, although the specific operations can also
7848 Arguments that are documented as @code{const int} require literal
7849 integral values within the range required for that operation.
7852 vector signed char vec_abs (vector signed char);
7853 vector signed short vec_abs (vector signed short);
7854 vector signed int vec_abs (vector signed int);
7855 vector float vec_abs (vector float);
7857 vector signed char vec_abss (vector signed char);
7858 vector signed short vec_abss (vector signed short);
7859 vector signed int vec_abss (vector signed int);
7861 vector signed char vec_add (vector bool char, vector signed char);
7862 vector signed char vec_add (vector signed char, vector bool char);
7863 vector signed char vec_add (vector signed char, vector signed char);
7864 vector unsigned char vec_add (vector bool char, vector unsigned char);
7865 vector unsigned char vec_add (vector unsigned char, vector bool char);
7866 vector unsigned char vec_add (vector unsigned char,
7867 vector unsigned char);
7868 vector signed short vec_add (vector bool short, vector signed short);
7869 vector signed short vec_add (vector signed short, vector bool short);
7870 vector signed short vec_add (vector signed short, vector signed short);
7871 vector unsigned short vec_add (vector bool short,
7872 vector unsigned short);
7873 vector unsigned short vec_add (vector unsigned short,
7875 vector unsigned short vec_add (vector unsigned short,
7876 vector unsigned short);
7877 vector signed int vec_add (vector bool int, vector signed int);
7878 vector signed int vec_add (vector signed int, vector bool int);
7879 vector signed int vec_add (vector signed int, vector signed int);
7880 vector unsigned int vec_add (vector bool int, vector unsigned int);
7881 vector unsigned int vec_add (vector unsigned int, vector bool int);
7882 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
7883 vector float vec_add (vector float, vector float);
7885 vector float vec_vaddfp (vector float, vector float);
7887 vector signed int vec_vadduwm (vector bool int, vector signed int);
7888 vector signed int vec_vadduwm (vector signed int, vector bool int);
7889 vector signed int vec_vadduwm (vector signed int, vector signed int);
7890 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
7891 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
7892 vector unsigned int vec_vadduwm (vector unsigned int,
7893 vector unsigned int);
7895 vector signed short vec_vadduhm (vector bool short,
7896 vector signed short);
7897 vector signed short vec_vadduhm (vector signed short,
7899 vector signed short vec_vadduhm (vector signed short,
7900 vector signed short);
7901 vector unsigned short vec_vadduhm (vector bool short,
7902 vector unsigned short);
7903 vector unsigned short vec_vadduhm (vector unsigned short,
7905 vector unsigned short vec_vadduhm (vector unsigned short,
7906 vector unsigned short);
7908 vector signed char vec_vaddubm (vector bool char, vector signed char);
7909 vector signed char vec_vaddubm (vector signed char, vector bool char);
7910 vector signed char vec_vaddubm (vector signed char, vector signed char);
7911 vector unsigned char vec_vaddubm (vector bool char,
7912 vector unsigned char);
7913 vector unsigned char vec_vaddubm (vector unsigned char,
7915 vector unsigned char vec_vaddubm (vector unsigned char,
7916 vector unsigned char);
7918 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
7920 vector unsigned char vec_adds (vector bool char, vector unsigned char);
7921 vector unsigned char vec_adds (vector unsigned char, vector bool char);
7922 vector unsigned char vec_adds (vector unsigned char,
7923 vector unsigned char);
7924 vector signed char vec_adds (vector bool char, vector signed char);
7925 vector signed char vec_adds (vector signed char, vector bool char);
7926 vector signed char vec_adds (vector signed char, vector signed char);
7927 vector unsigned short vec_adds (vector bool short,
7928 vector unsigned short);
7929 vector unsigned short vec_adds (vector unsigned short,
7931 vector unsigned short vec_adds (vector unsigned short,
7932 vector unsigned short);
7933 vector signed short vec_adds (vector bool short, vector signed short);
7934 vector signed short vec_adds (vector signed short, vector bool short);
7935 vector signed short vec_adds (vector signed short, vector signed short);
7936 vector unsigned int vec_adds (vector bool int, vector unsigned int);
7937 vector unsigned int vec_adds (vector unsigned int, vector bool int);
7938 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
7939 vector signed int vec_adds (vector bool int, vector signed int);
7940 vector signed int vec_adds (vector signed int, vector bool int);
7941 vector signed int vec_adds (vector signed int, vector signed int);
7943 vector signed int vec_vaddsws (vector bool int, vector signed int);
7944 vector signed int vec_vaddsws (vector signed int, vector bool int);
7945 vector signed int vec_vaddsws (vector signed int, vector signed int);
7947 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
7948 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
7949 vector unsigned int vec_vadduws (vector unsigned int,
7950 vector unsigned int);
7952 vector signed short vec_vaddshs (vector bool short,
7953 vector signed short);
7954 vector signed short vec_vaddshs (vector signed short,
7956 vector signed short vec_vaddshs (vector signed short,
7957 vector signed short);
7959 vector unsigned short vec_vadduhs (vector bool short,
7960 vector unsigned short);
7961 vector unsigned short vec_vadduhs (vector unsigned short,
7963 vector unsigned short vec_vadduhs (vector unsigned short,
7964 vector unsigned short);
7966 vector signed char vec_vaddsbs (vector bool char, vector signed char);
7967 vector signed char vec_vaddsbs (vector signed char, vector bool char);
7968 vector signed char vec_vaddsbs (vector signed char, vector signed char);
7970 vector unsigned char vec_vaddubs (vector bool char,
7971 vector unsigned char);
7972 vector unsigned char vec_vaddubs (vector unsigned char,
7974 vector unsigned char vec_vaddubs (vector unsigned char,
7975 vector unsigned char);
7977 vector float vec_and (vector float, vector float);
7978 vector float vec_and (vector float, vector bool int);
7979 vector float vec_and (vector bool int, vector float);
7980 vector bool int vec_and (vector bool int, vector bool int);
7981 vector signed int vec_and (vector bool int, vector signed int);
7982 vector signed int vec_and (vector signed int, vector bool int);
7983 vector signed int vec_and (vector signed int, vector signed int);
7984 vector unsigned int vec_and (vector bool int, vector unsigned int);
7985 vector unsigned int vec_and (vector unsigned int, vector bool int);
7986 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
7987 vector bool short vec_and (vector bool short, vector bool short);
7988 vector signed short vec_and (vector bool short, vector signed short);
7989 vector signed short vec_and (vector signed short, vector bool short);
7990 vector signed short vec_and (vector signed short, vector signed short);
7991 vector unsigned short vec_and (vector bool short,
7992 vector unsigned short);
7993 vector unsigned short vec_and (vector unsigned short,
7995 vector unsigned short vec_and (vector unsigned short,
7996 vector unsigned short);
7997 vector signed char vec_and (vector bool char, vector signed char);
7998 vector bool char vec_and (vector bool char, vector bool char);
7999 vector signed char vec_and (vector signed char, vector bool char);
8000 vector signed char vec_and (vector signed char, vector signed char);
8001 vector unsigned char vec_and (vector bool char, vector unsigned char);
8002 vector unsigned char vec_and (vector unsigned char, vector bool char);
8003 vector unsigned char vec_and (vector unsigned char,
8004 vector unsigned char);
8006 vector float vec_andc (vector float, vector float);
8007 vector float vec_andc (vector float, vector bool int);
8008 vector float vec_andc (vector bool int, vector float);
8009 vector bool int vec_andc (vector bool int, vector bool int);
8010 vector signed int vec_andc (vector bool int, vector signed int);
8011 vector signed int vec_andc (vector signed int, vector bool int);
8012 vector signed int vec_andc (vector signed int, vector signed int);
8013 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8014 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8015 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8016 vector bool short vec_andc (vector bool short, vector bool short);
8017 vector signed short vec_andc (vector bool short, vector signed short);
8018 vector signed short vec_andc (vector signed short, vector bool short);
8019 vector signed short vec_andc (vector signed short, vector signed short);
8020 vector unsigned short vec_andc (vector bool short,
8021 vector unsigned short);
8022 vector unsigned short vec_andc (vector unsigned short,
8024 vector unsigned short vec_andc (vector unsigned short,
8025 vector unsigned short);
8026 vector signed char vec_andc (vector bool char, vector signed char);
8027 vector bool char vec_andc (vector bool char, vector bool char);
8028 vector signed char vec_andc (vector signed char, vector bool char);
8029 vector signed char vec_andc (vector signed char, vector signed char);
8030 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8031 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8032 vector unsigned char vec_andc (vector unsigned char,
8033 vector unsigned char);
8035 vector unsigned char vec_avg (vector unsigned char,
8036 vector unsigned char);
8037 vector signed char vec_avg (vector signed char, vector signed char);
8038 vector unsigned short vec_avg (vector unsigned short,
8039 vector unsigned short);
8040 vector signed short vec_avg (vector signed short, vector signed short);
8041 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8042 vector signed int vec_avg (vector signed int, vector signed int);
8044 vector signed int vec_vavgsw (vector signed int, vector signed int);
8046 vector unsigned int vec_vavguw (vector unsigned int,
8047 vector unsigned int);
8049 vector signed short vec_vavgsh (vector signed short,
8050 vector signed short);
8052 vector unsigned short vec_vavguh (vector unsigned short,
8053 vector unsigned short);
8055 vector signed char vec_vavgsb (vector signed char, vector signed char);
8057 vector unsigned char vec_vavgub (vector unsigned char,
8058 vector unsigned char);
8060 vector float vec_ceil (vector float);
8062 vector signed int vec_cmpb (vector float, vector float);
8064 vector bool char vec_cmpeq (vector signed char, vector signed char);
8065 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8066 vector bool short vec_cmpeq (vector signed short, vector signed short);
8067 vector bool short vec_cmpeq (vector unsigned short,
8068 vector unsigned short);
8069 vector bool int vec_cmpeq (vector signed int, vector signed int);
8070 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8071 vector bool int vec_cmpeq (vector float, vector float);
8073 vector bool int vec_vcmpeqfp (vector float, vector float);
8075 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8076 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8078 vector bool short vec_vcmpequh (vector signed short,
8079 vector signed short);
8080 vector bool short vec_vcmpequh (vector unsigned short,
8081 vector unsigned short);
8083 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8084 vector bool char vec_vcmpequb (vector unsigned char,
8085 vector unsigned char);
8087 vector bool int vec_cmpge (vector float, vector float);
8089 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8090 vector bool char vec_cmpgt (vector signed char, vector signed char);
8091 vector bool short vec_cmpgt (vector unsigned short,
8092 vector unsigned short);
8093 vector bool short vec_cmpgt (vector signed short, vector signed short);
8094 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8095 vector bool int vec_cmpgt (vector signed int, vector signed int);
8096 vector bool int vec_cmpgt (vector float, vector float);
8098 vector bool int vec_vcmpgtfp (vector float, vector float);
8100 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8102 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8104 vector bool short vec_vcmpgtsh (vector signed short,
8105 vector signed short);
8107 vector bool short vec_vcmpgtuh (vector unsigned short,
8108 vector unsigned short);
8110 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8112 vector bool char vec_vcmpgtub (vector unsigned char,
8113 vector unsigned char);
8115 vector bool int vec_cmple (vector float, vector float);
8117 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8118 vector bool char vec_cmplt (vector signed char, vector signed char);
8119 vector bool short vec_cmplt (vector unsigned short,
8120 vector unsigned short);
8121 vector bool short vec_cmplt (vector signed short, vector signed short);
8122 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8123 vector bool int vec_cmplt (vector signed int, vector signed int);
8124 vector bool int vec_cmplt (vector float, vector float);
8126 vector float vec_ctf (vector unsigned int, const int);
8127 vector float vec_ctf (vector signed int, const int);
8129 vector float vec_vcfsx (vector signed int, const int);
8131 vector float vec_vcfux (vector unsigned int, const int);
8133 vector signed int vec_cts (vector float, const int);
8135 vector unsigned int vec_ctu (vector float, const int);
8137 void vec_dss (const int);
8139 void vec_dssall (void);
8141 void vec_dst (const vector unsigned char *, int, const int);
8142 void vec_dst (const vector signed char *, int, const int);
8143 void vec_dst (const vector bool char *, int, const int);
8144 void vec_dst (const vector unsigned short *, int, const int);
8145 void vec_dst (const vector signed short *, int, const int);
8146 void vec_dst (const vector bool short *, int, const int);
8147 void vec_dst (const vector pixel *, int, const int);
8148 void vec_dst (const vector unsigned int *, int, const int);
8149 void vec_dst (const vector signed int *, int, const int);
8150 void vec_dst (const vector bool int *, int, const int);
8151 void vec_dst (const vector float *, int, const int);
8152 void vec_dst (const unsigned char *, int, const int);
8153 void vec_dst (const signed char *, int, const int);
8154 void vec_dst (const unsigned short *, int, const int);
8155 void vec_dst (const short *, int, const int);
8156 void vec_dst (const unsigned int *, int, const int);
8157 void vec_dst (const int *, int, const int);
8158 void vec_dst (const unsigned long *, int, const int);
8159 void vec_dst (const long *, int, const int);
8160 void vec_dst (const float *, int, const int);
8162 void vec_dstst (const vector unsigned char *, int, const int);
8163 void vec_dstst (const vector signed char *, int, const int);
8164 void vec_dstst (const vector bool char *, int, const int);
8165 void vec_dstst (const vector unsigned short *, int, const int);
8166 void vec_dstst (const vector signed short *, int, const int);
8167 void vec_dstst (const vector bool short *, int, const int);
8168 void vec_dstst (const vector pixel *, int, const int);
8169 void vec_dstst (const vector unsigned int *, int, const int);
8170 void vec_dstst (const vector signed int *, int, const int);
8171 void vec_dstst (const vector bool int *, int, const int);
8172 void vec_dstst (const vector float *, int, const int);
8173 void vec_dstst (const unsigned char *, int, const int);
8174 void vec_dstst (const signed char *, int, const int);
8175 void vec_dstst (const unsigned short *, int, const int);
8176 void vec_dstst (const short *, int, const int);
8177 void vec_dstst (const unsigned int *, int, const int);
8178 void vec_dstst (const int *, int, const int);
8179 void vec_dstst (const unsigned long *, int, const int);
8180 void vec_dstst (const long *, int, const int);
8181 void vec_dstst (const float *, int, const int);
8183 void vec_dststt (const vector unsigned char *, int, const int);
8184 void vec_dststt (const vector signed char *, int, const int);
8185 void vec_dststt (const vector bool char *, int, const int);
8186 void vec_dststt (const vector unsigned short *, int, const int);
8187 void vec_dststt (const vector signed short *, int, const int);
8188 void vec_dststt (const vector bool short *, int, const int);
8189 void vec_dststt (const vector pixel *, int, const int);
8190 void vec_dststt (const vector unsigned int *, int, const int);
8191 void vec_dststt (const vector signed int *, int, const int);
8192 void vec_dststt (const vector bool int *, int, const int);
8193 void vec_dststt (const vector float *, int, const int);
8194 void vec_dststt (const unsigned char *, int, const int);
8195 void vec_dststt (const signed char *, int, const int);
8196 void vec_dststt (const unsigned short *, int, const int);
8197 void vec_dststt (const short *, int, const int);
8198 void vec_dststt (const unsigned int *, int, const int);
8199 void vec_dststt (const int *, int, const int);
8200 void vec_dststt (const unsigned long *, int, const int);
8201 void vec_dststt (const long *, int, const int);
8202 void vec_dststt (const float *, int, const int);
8204 void vec_dstt (const vector unsigned char *, int, const int);
8205 void vec_dstt (const vector signed char *, int, const int);
8206 void vec_dstt (const vector bool char *, int, const int);
8207 void vec_dstt (const vector unsigned short *, int, const int);
8208 void vec_dstt (const vector signed short *, int, const int);
8209 void vec_dstt (const vector bool short *, int, const int);
8210 void vec_dstt (const vector pixel *, int, const int);
8211 void vec_dstt (const vector unsigned int *, int, const int);
8212 void vec_dstt (const vector signed int *, int, const int);
8213 void vec_dstt (const vector bool int *, int, const int);
8214 void vec_dstt (const vector float *, int, const int);
8215 void vec_dstt (const unsigned char *, int, const int);
8216 void vec_dstt (const signed char *, int, const int);
8217 void vec_dstt (const unsigned short *, int, const int);
8218 void vec_dstt (const short *, int, const int);
8219 void vec_dstt (const unsigned int *, int, const int);
8220 void vec_dstt (const int *, int, const int);
8221 void vec_dstt (const unsigned long *, int, const int);
8222 void vec_dstt (const long *, int, const int);
8223 void vec_dstt (const float *, int, const int);
8225 vector float vec_expte (vector float);
8227 vector float vec_floor (vector float);
8229 vector float vec_ld (int, const vector float *);
8230 vector float vec_ld (int, const float *);
8231 vector bool int vec_ld (int, const vector bool int *);
8232 vector signed int vec_ld (int, const vector signed int *);
8233 vector signed int vec_ld (int, const int *);
8234 vector signed int vec_ld (int, const long *);
8235 vector unsigned int vec_ld (int, const vector unsigned int *);
8236 vector unsigned int vec_ld (int, const unsigned int *);
8237 vector unsigned int vec_ld (int, const unsigned long *);
8238 vector bool short vec_ld (int, const vector bool short *);
8239 vector pixel vec_ld (int, const vector pixel *);
8240 vector signed short vec_ld (int, const vector signed short *);
8241 vector signed short vec_ld (int, const short *);
8242 vector unsigned short vec_ld (int, const vector unsigned short *);
8243 vector unsigned short vec_ld (int, const unsigned short *);
8244 vector bool char vec_ld (int, const vector bool char *);
8245 vector signed char vec_ld (int, const vector signed char *);
8246 vector signed char vec_ld (int, const signed char *);
8247 vector unsigned char vec_ld (int, const vector unsigned char *);
8248 vector unsigned char vec_ld (int, const unsigned char *);
8250 vector signed char vec_lde (int, const signed char *);
8251 vector unsigned char vec_lde (int, const unsigned char *);
8252 vector signed short vec_lde (int, const short *);
8253 vector unsigned short vec_lde (int, const unsigned short *);
8254 vector float vec_lde (int, const float *);
8255 vector signed int vec_lde (int, const int *);
8256 vector unsigned int vec_lde (int, const unsigned int *);
8257 vector signed int vec_lde (int, const long *);
8258 vector unsigned int vec_lde (int, const unsigned long *);
8260 vector float vec_lvewx (int, float *);
8261 vector signed int vec_lvewx (int, int *);
8262 vector unsigned int vec_lvewx (int, unsigned int *);
8263 vector signed int vec_lvewx (int, long *);
8264 vector unsigned int vec_lvewx (int, unsigned long *);
8266 vector signed short vec_lvehx (int, short *);
8267 vector unsigned short vec_lvehx (int, unsigned short *);
8269 vector signed char vec_lvebx (int, char *);
8270 vector unsigned char vec_lvebx (int, unsigned char *);
8272 vector float vec_ldl (int, const vector float *);
8273 vector float vec_ldl (int, const float *);
8274 vector bool int vec_ldl (int, const vector bool int *);
8275 vector signed int vec_ldl (int, const vector signed int *);
8276 vector signed int vec_ldl (int, const int *);
8277 vector signed int vec_ldl (int, const long *);
8278 vector unsigned int vec_ldl (int, const vector unsigned int *);
8279 vector unsigned int vec_ldl (int, const unsigned int *);
8280 vector unsigned int vec_ldl (int, const unsigned long *);
8281 vector bool short vec_ldl (int, const vector bool short *);
8282 vector pixel vec_ldl (int, const vector pixel *);
8283 vector signed short vec_ldl (int, const vector signed short *);
8284 vector signed short vec_ldl (int, const short *);
8285 vector unsigned short vec_ldl (int, const vector unsigned short *);
8286 vector unsigned short vec_ldl (int, const unsigned short *);
8287 vector bool char vec_ldl (int, const vector bool char *);
8288 vector signed char vec_ldl (int, const vector signed char *);
8289 vector signed char vec_ldl (int, const signed char *);
8290 vector unsigned char vec_ldl (int, const vector unsigned char *);
8291 vector unsigned char vec_ldl (int, const unsigned char *);
8293 vector float vec_loge (vector float);
8295 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8296 vector unsigned char vec_lvsl (int, const volatile signed char *);
8297 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8298 vector unsigned char vec_lvsl (int, const volatile short *);
8299 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8300 vector unsigned char vec_lvsl (int, const volatile int *);
8301 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8302 vector unsigned char vec_lvsl (int, const volatile long *);
8303 vector unsigned char vec_lvsl (int, const volatile float *);
8305 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8306 vector unsigned char vec_lvsr (int, const volatile signed char *);
8307 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8308 vector unsigned char vec_lvsr (int, const volatile short *);
8309 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8310 vector unsigned char vec_lvsr (int, const volatile int *);
8311 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8312 vector unsigned char vec_lvsr (int, const volatile long *);
8313 vector unsigned char vec_lvsr (int, const volatile float *);
8315 vector float vec_madd (vector float, vector float, vector float);
8317 vector signed short vec_madds (vector signed short,
8318 vector signed short,
8319 vector signed short);
8321 vector unsigned char vec_max (vector bool char, vector unsigned char);
8322 vector unsigned char vec_max (vector unsigned char, vector bool char);
8323 vector unsigned char vec_max (vector unsigned char,
8324 vector unsigned char);
8325 vector signed char vec_max (vector bool char, vector signed char);
8326 vector signed char vec_max (vector signed char, vector bool char);
8327 vector signed char vec_max (vector signed char, vector signed char);
8328 vector unsigned short vec_max (vector bool short,
8329 vector unsigned short);
8330 vector unsigned short vec_max (vector unsigned short,
8332 vector unsigned short vec_max (vector unsigned short,
8333 vector unsigned short);
8334 vector signed short vec_max (vector bool short, vector signed short);
8335 vector signed short vec_max (vector signed short, vector bool short);
8336 vector signed short vec_max (vector signed short, vector signed short);
8337 vector unsigned int vec_max (vector bool int, vector unsigned int);
8338 vector unsigned int vec_max (vector unsigned int, vector bool int);
8339 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8340 vector signed int vec_max (vector bool int, vector signed int);
8341 vector signed int vec_max (vector signed int, vector bool int);
8342 vector signed int vec_max (vector signed int, vector signed int);
8343 vector float vec_max (vector float, vector float);
8345 vector float vec_vmaxfp (vector float, vector float);
8347 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8348 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8349 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8351 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8352 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8353 vector unsigned int vec_vmaxuw (vector unsigned int,
8354 vector unsigned int);
8356 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8357 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8358 vector signed short vec_vmaxsh (vector signed short,
8359 vector signed short);
8361 vector unsigned short vec_vmaxuh (vector bool short,
8362 vector unsigned short);
8363 vector unsigned short vec_vmaxuh (vector unsigned short,
8365 vector unsigned short vec_vmaxuh (vector unsigned short,
8366 vector unsigned short);
8368 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8369 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8370 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8372 vector unsigned char vec_vmaxub (vector bool char,
8373 vector unsigned char);
8374 vector unsigned char vec_vmaxub (vector unsigned char,
8376 vector unsigned char vec_vmaxub (vector unsigned char,
8377 vector unsigned char);
8379 vector bool char vec_mergeh (vector bool char, vector bool char);
8380 vector signed char vec_mergeh (vector signed char, vector signed char);
8381 vector unsigned char vec_mergeh (vector unsigned char,
8382 vector unsigned char);
8383 vector bool short vec_mergeh (vector bool short, vector bool short);
8384 vector pixel vec_mergeh (vector pixel, vector pixel);
8385 vector signed short vec_mergeh (vector signed short,
8386 vector signed short);
8387 vector unsigned short vec_mergeh (vector unsigned short,
8388 vector unsigned short);
8389 vector float vec_mergeh (vector float, vector float);
8390 vector bool int vec_mergeh (vector bool int, vector bool int);
8391 vector signed int vec_mergeh (vector signed int, vector signed int);
8392 vector unsigned int vec_mergeh (vector unsigned int,
8393 vector unsigned int);
8395 vector float vec_vmrghw (vector float, vector float);
8396 vector bool int vec_vmrghw (vector bool int, vector bool int);
8397 vector signed int vec_vmrghw (vector signed int, vector signed int);
8398 vector unsigned int vec_vmrghw (vector unsigned int,
8399 vector unsigned int);
8401 vector bool short vec_vmrghh (vector bool short, vector bool short);
8402 vector signed short vec_vmrghh (vector signed short,
8403 vector signed short);
8404 vector unsigned short vec_vmrghh (vector unsigned short,
8405 vector unsigned short);
8406 vector pixel vec_vmrghh (vector pixel, vector pixel);
8408 vector bool char vec_vmrghb (vector bool char, vector bool char);
8409 vector signed char vec_vmrghb (vector signed char, vector signed char);
8410 vector unsigned char vec_vmrghb (vector unsigned char,
8411 vector unsigned char);
8413 vector bool char vec_mergel (vector bool char, vector bool char);
8414 vector signed char vec_mergel (vector signed char, vector signed char);
8415 vector unsigned char vec_mergel (vector unsigned char,
8416 vector unsigned char);
8417 vector bool short vec_mergel (vector bool short, vector bool short);
8418 vector pixel vec_mergel (vector pixel, vector pixel);
8419 vector signed short vec_mergel (vector signed short,
8420 vector signed short);
8421 vector unsigned short vec_mergel (vector unsigned short,
8422 vector unsigned short);
8423 vector float vec_mergel (vector float, vector float);
8424 vector bool int vec_mergel (vector bool int, vector bool int);
8425 vector signed int vec_mergel (vector signed int, vector signed int);
8426 vector unsigned int vec_mergel (vector unsigned int,
8427 vector unsigned int);
8429 vector float vec_vmrglw (vector float, vector float);
8430 vector signed int vec_vmrglw (vector signed int, vector signed int);
8431 vector unsigned int vec_vmrglw (vector unsigned int,
8432 vector unsigned int);
8433 vector bool int vec_vmrglw (vector bool int, vector bool int);
8435 vector bool short vec_vmrglh (vector bool short, vector bool short);
8436 vector signed short vec_vmrglh (vector signed short,
8437 vector signed short);
8438 vector unsigned short vec_vmrglh (vector unsigned short,
8439 vector unsigned short);
8440 vector pixel vec_vmrglh (vector pixel, vector pixel);
8442 vector bool char vec_vmrglb (vector bool char, vector bool char);
8443 vector signed char vec_vmrglb (vector signed char, vector signed char);
8444 vector unsigned char vec_vmrglb (vector unsigned char,
8445 vector unsigned char);
8447 vector unsigned short vec_mfvscr (void);
8449 vector unsigned char vec_min (vector bool char, vector unsigned char);
8450 vector unsigned char vec_min (vector unsigned char, vector bool char);
8451 vector unsigned char vec_min (vector unsigned char,
8452 vector unsigned char);
8453 vector signed char vec_min (vector bool char, vector signed char);
8454 vector signed char vec_min (vector signed char, vector bool char);
8455 vector signed char vec_min (vector signed char, vector signed char);
8456 vector unsigned short vec_min (vector bool short,
8457 vector unsigned short);
8458 vector unsigned short vec_min (vector unsigned short,
8460 vector unsigned short vec_min (vector unsigned short,
8461 vector unsigned short);
8462 vector signed short vec_min (vector bool short, vector signed short);
8463 vector signed short vec_min (vector signed short, vector bool short);
8464 vector signed short vec_min (vector signed short, vector signed short);
8465 vector unsigned int vec_min (vector bool int, vector unsigned int);
8466 vector unsigned int vec_min (vector unsigned int, vector bool int);
8467 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8468 vector signed int vec_min (vector bool int, vector signed int);
8469 vector signed int vec_min (vector signed int, vector bool int);
8470 vector signed int vec_min (vector signed int, vector signed int);
8471 vector float vec_min (vector float, vector float);
8473 vector float vec_vminfp (vector float, vector float);
8475 vector signed int vec_vminsw (vector bool int, vector signed int);
8476 vector signed int vec_vminsw (vector signed int, vector bool int);
8477 vector signed int vec_vminsw (vector signed int, vector signed int);
8479 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8480 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8481 vector unsigned int vec_vminuw (vector unsigned int,
8482 vector unsigned int);
8484 vector signed short vec_vminsh (vector bool short, vector signed short);
8485 vector signed short vec_vminsh (vector signed short, vector bool short);
8486 vector signed short vec_vminsh (vector signed short,
8487 vector signed short);
8489 vector unsigned short vec_vminuh (vector bool short,
8490 vector unsigned short);
8491 vector unsigned short vec_vminuh (vector unsigned short,
8493 vector unsigned short vec_vminuh (vector unsigned short,
8494 vector unsigned short);
8496 vector signed char vec_vminsb (vector bool char, vector signed char);
8497 vector signed char vec_vminsb (vector signed char, vector bool char);
8498 vector signed char vec_vminsb (vector signed char, vector signed char);
8500 vector unsigned char vec_vminub (vector bool char,
8501 vector unsigned char);
8502 vector unsigned char vec_vminub (vector unsigned char,
8504 vector unsigned char vec_vminub (vector unsigned char,
8505 vector unsigned char);
8507 vector signed short vec_mladd (vector signed short,
8508 vector signed short,
8509 vector signed short);
8510 vector signed short vec_mladd (vector signed short,
8511 vector unsigned short,
8512 vector unsigned short);
8513 vector signed short vec_mladd (vector unsigned short,
8514 vector signed short,
8515 vector signed short);
8516 vector unsigned short vec_mladd (vector unsigned short,
8517 vector unsigned short,
8518 vector unsigned short);
8520 vector signed short vec_mradds (vector signed short,
8521 vector signed short,
8522 vector signed short);
8524 vector unsigned int vec_msum (vector unsigned char,
8525 vector unsigned char,
8526 vector unsigned int);
8527 vector signed int vec_msum (vector signed char,
8528 vector unsigned char,
8530 vector unsigned int vec_msum (vector unsigned short,
8531 vector unsigned short,
8532 vector unsigned int);
8533 vector signed int vec_msum (vector signed short,
8534 vector signed short,
8537 vector signed int vec_vmsumshm (vector signed short,
8538 vector signed short,
8541 vector unsigned int vec_vmsumuhm (vector unsigned short,
8542 vector unsigned short,
8543 vector unsigned int);
8545 vector signed int vec_vmsummbm (vector signed char,
8546 vector unsigned char,
8549 vector unsigned int vec_vmsumubm (vector unsigned char,
8550 vector unsigned char,
8551 vector unsigned int);
8553 vector unsigned int vec_msums (vector unsigned short,
8554 vector unsigned short,
8555 vector unsigned int);
8556 vector signed int vec_msums (vector signed short,
8557 vector signed short,
8560 vector signed int vec_vmsumshs (vector signed short,
8561 vector signed short,
8564 vector unsigned int vec_vmsumuhs (vector unsigned short,
8565 vector unsigned short,
8566 vector unsigned int);
8568 void vec_mtvscr (vector signed int);
8569 void vec_mtvscr (vector unsigned int);
8570 void vec_mtvscr (vector bool int);
8571 void vec_mtvscr (vector signed short);
8572 void vec_mtvscr (vector unsigned short);
8573 void vec_mtvscr (vector bool short);
8574 void vec_mtvscr (vector pixel);
8575 void vec_mtvscr (vector signed char);
8576 void vec_mtvscr (vector unsigned char);
8577 void vec_mtvscr (vector bool char);
8579 vector unsigned short vec_mule (vector unsigned char,
8580 vector unsigned char);
8581 vector signed short vec_mule (vector signed char,
8582 vector signed char);
8583 vector unsigned int vec_mule (vector unsigned short,
8584 vector unsigned short);
8585 vector signed int vec_mule (vector signed short, vector signed short);
8587 vector signed int vec_vmulesh (vector signed short,
8588 vector signed short);
8590 vector unsigned int vec_vmuleuh (vector unsigned short,
8591 vector unsigned short);
8593 vector signed short vec_vmulesb (vector signed char,
8594 vector signed char);
8596 vector unsigned short vec_vmuleub (vector unsigned char,
8597 vector unsigned char);
8599 vector unsigned short vec_mulo (vector unsigned char,
8600 vector unsigned char);
8601 vector signed short vec_mulo (vector signed char, vector signed char);
8602 vector unsigned int vec_mulo (vector unsigned short,
8603 vector unsigned short);
8604 vector signed int vec_mulo (vector signed short, vector signed short);
8606 vector signed int vec_vmulosh (vector signed short,
8607 vector signed short);
8609 vector unsigned int vec_vmulouh (vector unsigned short,
8610 vector unsigned short);
8612 vector signed short vec_vmulosb (vector signed char,
8613 vector signed char);
8615 vector unsigned short vec_vmuloub (vector unsigned char,
8616 vector unsigned char);
8618 vector float vec_nmsub (vector float, vector float, vector float);
8620 vector float vec_nor (vector float, vector float);
8621 vector signed int vec_nor (vector signed int, vector signed int);
8622 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8623 vector bool int vec_nor (vector bool int, vector bool int);
8624 vector signed short vec_nor (vector signed short, vector signed short);
8625 vector unsigned short vec_nor (vector unsigned short,
8626 vector unsigned short);
8627 vector bool short vec_nor (vector bool short, vector bool short);
8628 vector signed char vec_nor (vector signed char, vector signed char);
8629 vector unsigned char vec_nor (vector unsigned char,
8630 vector unsigned char);
8631 vector bool char vec_nor (vector bool char, vector bool char);
8633 vector float vec_or (vector float, vector float);
8634 vector float vec_or (vector float, vector bool int);
8635 vector float vec_or (vector bool int, vector float);
8636 vector bool int vec_or (vector bool int, vector bool int);
8637 vector signed int vec_or (vector bool int, vector signed int);
8638 vector signed int vec_or (vector signed int, vector bool int);
8639 vector signed int vec_or (vector signed int, vector signed int);
8640 vector unsigned int vec_or (vector bool int, vector unsigned int);
8641 vector unsigned int vec_or (vector unsigned int, vector bool int);
8642 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8643 vector bool short vec_or (vector bool short, vector bool short);
8644 vector signed short vec_or (vector bool short, vector signed short);
8645 vector signed short vec_or (vector signed short, vector bool short);
8646 vector signed short vec_or (vector signed short, vector signed short);
8647 vector unsigned short vec_or (vector bool short, vector unsigned short);
8648 vector unsigned short vec_or (vector unsigned short, vector bool short);
8649 vector unsigned short vec_or (vector unsigned short,
8650 vector unsigned short);
8651 vector signed char vec_or (vector bool char, vector signed char);
8652 vector bool char vec_or (vector bool char, vector bool char);
8653 vector signed char vec_or (vector signed char, vector bool char);
8654 vector signed char vec_or (vector signed char, vector signed char);
8655 vector unsigned char vec_or (vector bool char, vector unsigned char);
8656 vector unsigned char vec_or (vector unsigned char, vector bool char);
8657 vector unsigned char vec_or (vector unsigned char,
8658 vector unsigned char);
8660 vector signed char vec_pack (vector signed short, vector signed short);
8661 vector unsigned char vec_pack (vector unsigned short,
8662 vector unsigned short);
8663 vector bool char vec_pack (vector bool short, vector bool short);
8664 vector signed short vec_pack (vector signed int, vector signed int);
8665 vector unsigned short vec_pack (vector unsigned int,
8666 vector unsigned int);
8667 vector bool short vec_pack (vector bool int, vector bool int);
8669 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8670 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8671 vector unsigned short vec_vpkuwum (vector unsigned int,
8672 vector unsigned int);
8674 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8675 vector signed char vec_vpkuhum (vector signed short,
8676 vector signed short);
8677 vector unsigned char vec_vpkuhum (vector unsigned short,
8678 vector unsigned short);
8680 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8682 vector unsigned char vec_packs (vector unsigned short,
8683 vector unsigned short);
8684 vector signed char vec_packs (vector signed short, vector signed short);
8685 vector unsigned short vec_packs (vector unsigned int,
8686 vector unsigned int);
8687 vector signed short vec_packs (vector signed int, vector signed int);
8689 vector signed short vec_vpkswss (vector signed int, vector signed int);
8691 vector unsigned short vec_vpkuwus (vector unsigned int,
8692 vector unsigned int);
8694 vector signed char vec_vpkshss (vector signed short,
8695 vector signed short);
8697 vector unsigned char vec_vpkuhus (vector unsigned short,
8698 vector unsigned short);
8700 vector unsigned char vec_packsu (vector unsigned short,
8701 vector unsigned short);
8702 vector unsigned char vec_packsu (vector signed short,
8703 vector signed short);
8704 vector unsigned short vec_packsu (vector unsigned int,
8705 vector unsigned int);
8706 vector unsigned short vec_packsu (vector signed int, vector signed int);
8708 vector unsigned short vec_vpkswus (vector signed int,
8711 vector unsigned char vec_vpkshus (vector signed short,
8712 vector signed short);
8714 vector float vec_perm (vector float,
8716 vector unsigned char);
8717 vector signed int vec_perm (vector signed int,
8719 vector unsigned char);
8720 vector unsigned int vec_perm (vector unsigned int,
8721 vector unsigned int,
8722 vector unsigned char);
8723 vector bool int vec_perm (vector bool int,
8725 vector unsigned char);
8726 vector signed short vec_perm (vector signed short,
8727 vector signed short,
8728 vector unsigned char);
8729 vector unsigned short vec_perm (vector unsigned short,
8730 vector unsigned short,
8731 vector unsigned char);
8732 vector bool short vec_perm (vector bool short,
8734 vector unsigned char);
8735 vector pixel vec_perm (vector pixel,
8737 vector unsigned char);
8738 vector signed char vec_perm (vector signed char,
8740 vector unsigned char);
8741 vector unsigned char vec_perm (vector unsigned char,
8742 vector unsigned char,
8743 vector unsigned char);
8744 vector bool char vec_perm (vector bool char,
8746 vector unsigned char);
8748 vector float vec_re (vector float);
8750 vector signed char vec_rl (vector signed char,
8751 vector unsigned char);
8752 vector unsigned char vec_rl (vector unsigned char,
8753 vector unsigned char);
8754 vector signed short vec_rl (vector signed short, vector unsigned short);
8755 vector unsigned short vec_rl (vector unsigned short,
8756 vector unsigned short);
8757 vector signed int vec_rl (vector signed int, vector unsigned int);
8758 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8760 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8761 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8763 vector signed short vec_vrlh (vector signed short,
8764 vector unsigned short);
8765 vector unsigned short vec_vrlh (vector unsigned short,
8766 vector unsigned short);
8768 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8769 vector unsigned char vec_vrlb (vector unsigned char,
8770 vector unsigned char);
8772 vector float vec_round (vector float);
8774 vector float vec_rsqrte (vector float);
8776 vector float vec_sel (vector float, vector float, vector bool int);
8777 vector float vec_sel (vector float, vector float, vector unsigned int);
8778 vector signed int vec_sel (vector signed int,
8781 vector signed int vec_sel (vector signed int,
8783 vector unsigned int);
8784 vector unsigned int vec_sel (vector unsigned int,
8785 vector unsigned int,
8787 vector unsigned int vec_sel (vector unsigned int,
8788 vector unsigned int,
8789 vector unsigned int);
8790 vector bool int vec_sel (vector bool int,
8793 vector bool int vec_sel (vector bool int,
8795 vector unsigned int);
8796 vector signed short vec_sel (vector signed short,
8797 vector signed short,
8799 vector signed short vec_sel (vector signed short,
8800 vector signed short,
8801 vector unsigned short);
8802 vector unsigned short vec_sel (vector unsigned short,
8803 vector unsigned short,
8805 vector unsigned short vec_sel (vector unsigned short,
8806 vector unsigned short,
8807 vector unsigned short);
8808 vector bool short vec_sel (vector bool short,
8811 vector bool short vec_sel (vector bool short,
8813 vector unsigned short);
8814 vector signed char vec_sel (vector signed char,
8817 vector signed char vec_sel (vector signed char,
8819 vector unsigned char);
8820 vector unsigned char vec_sel (vector unsigned char,
8821 vector unsigned char,
8823 vector unsigned char vec_sel (vector unsigned char,
8824 vector unsigned char,
8825 vector unsigned char);
8826 vector bool char vec_sel (vector bool char,
8829 vector bool char vec_sel (vector bool char,
8831 vector unsigned char);
8833 vector signed char vec_sl (vector signed char,
8834 vector unsigned char);
8835 vector unsigned char vec_sl (vector unsigned char,
8836 vector unsigned char);
8837 vector signed short vec_sl (vector signed short, vector unsigned short);
8838 vector unsigned short vec_sl (vector unsigned short,
8839 vector unsigned short);
8840 vector signed int vec_sl (vector signed int, vector unsigned int);
8841 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
8843 vector signed int vec_vslw (vector signed int, vector unsigned int);
8844 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
8846 vector signed short vec_vslh (vector signed short,
8847 vector unsigned short);
8848 vector unsigned short vec_vslh (vector unsigned short,
8849 vector unsigned short);
8851 vector signed char vec_vslb (vector signed char, vector unsigned char);
8852 vector unsigned char vec_vslb (vector unsigned char,
8853 vector unsigned char);
8855 vector float vec_sld (vector float, vector float, const int);
8856 vector signed int vec_sld (vector signed int,
8859 vector unsigned int vec_sld (vector unsigned int,
8860 vector unsigned int,
8862 vector bool int vec_sld (vector bool int,
8865 vector signed short vec_sld (vector signed short,
8866 vector signed short,
8868 vector unsigned short vec_sld (vector unsigned short,
8869 vector unsigned short,
8871 vector bool short vec_sld (vector bool short,
8874 vector pixel vec_sld (vector pixel,
8877 vector signed char vec_sld (vector signed char,
8880 vector unsigned char vec_sld (vector unsigned char,
8881 vector unsigned char,
8883 vector bool char vec_sld (vector bool char,
8887 vector signed int vec_sll (vector signed int,
8888 vector unsigned int);
8889 vector signed int vec_sll (vector signed int,
8890 vector unsigned short);
8891 vector signed int vec_sll (vector signed int,
8892 vector unsigned char);
8893 vector unsigned int vec_sll (vector unsigned int,
8894 vector unsigned int);
8895 vector unsigned int vec_sll (vector unsigned int,
8896 vector unsigned short);
8897 vector unsigned int vec_sll (vector unsigned int,
8898 vector unsigned char);
8899 vector bool int vec_sll (vector bool int,
8900 vector unsigned int);
8901 vector bool int vec_sll (vector bool int,
8902 vector unsigned short);
8903 vector bool int vec_sll (vector bool int,
8904 vector unsigned char);
8905 vector signed short vec_sll (vector signed short,
8906 vector unsigned int);
8907 vector signed short vec_sll (vector signed short,
8908 vector unsigned short);
8909 vector signed short vec_sll (vector signed short,
8910 vector unsigned char);
8911 vector unsigned short vec_sll (vector unsigned short,
8912 vector unsigned int);
8913 vector unsigned short vec_sll (vector unsigned short,
8914 vector unsigned short);
8915 vector unsigned short vec_sll (vector unsigned short,
8916 vector unsigned char);
8917 vector bool short vec_sll (vector bool short, vector unsigned int);
8918 vector bool short vec_sll (vector bool short, vector unsigned short);
8919 vector bool short vec_sll (vector bool short, vector unsigned char);
8920 vector pixel vec_sll (vector pixel, vector unsigned int);
8921 vector pixel vec_sll (vector pixel, vector unsigned short);
8922 vector pixel vec_sll (vector pixel, vector unsigned char);
8923 vector signed char vec_sll (vector signed char, vector unsigned int);
8924 vector signed char vec_sll (vector signed char, vector unsigned short);
8925 vector signed char vec_sll (vector signed char, vector unsigned char);
8926 vector unsigned char vec_sll (vector unsigned char,
8927 vector unsigned int);
8928 vector unsigned char vec_sll (vector unsigned char,
8929 vector unsigned short);
8930 vector unsigned char vec_sll (vector unsigned char,
8931 vector unsigned char);
8932 vector bool char vec_sll (vector bool char, vector unsigned int);
8933 vector bool char vec_sll (vector bool char, vector unsigned short);
8934 vector bool char vec_sll (vector bool char, vector unsigned char);
8936 vector float vec_slo (vector float, vector signed char);
8937 vector float vec_slo (vector float, vector unsigned char);
8938 vector signed int vec_slo (vector signed int, vector signed char);
8939 vector signed int vec_slo (vector signed int, vector unsigned char);
8940 vector unsigned int vec_slo (vector unsigned int, vector signed char);
8941 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
8942 vector signed short vec_slo (vector signed short, vector signed char);
8943 vector signed short vec_slo (vector signed short, vector unsigned char);
8944 vector unsigned short vec_slo (vector unsigned short,
8945 vector signed char);
8946 vector unsigned short vec_slo (vector unsigned short,
8947 vector unsigned char);
8948 vector pixel vec_slo (vector pixel, vector signed char);
8949 vector pixel vec_slo (vector pixel, vector unsigned char);
8950 vector signed char vec_slo (vector signed char, vector signed char);
8951 vector signed char vec_slo (vector signed char, vector unsigned char);
8952 vector unsigned char vec_slo (vector unsigned char, vector signed char);
8953 vector unsigned char vec_slo (vector unsigned char,
8954 vector unsigned char);
8956 vector signed char vec_splat (vector signed char, const int);
8957 vector unsigned char vec_splat (vector unsigned char, const int);
8958 vector bool char vec_splat (vector bool char, const int);
8959 vector signed short vec_splat (vector signed short, const int);
8960 vector unsigned short vec_splat (vector unsigned short, const int);
8961 vector bool short vec_splat (vector bool short, const int);
8962 vector pixel vec_splat (vector pixel, const int);
8963 vector float vec_splat (vector float, const int);
8964 vector signed int vec_splat (vector signed int, const int);
8965 vector unsigned int vec_splat (vector unsigned int, const int);
8966 vector bool int vec_splat (vector bool int, const int);
8968 vector float vec_vspltw (vector float, const int);
8969 vector signed int vec_vspltw (vector signed int, const int);
8970 vector unsigned int vec_vspltw (vector unsigned int, const int);
8971 vector bool int vec_vspltw (vector bool int, const int);
8973 vector bool short vec_vsplth (vector bool short, const int);
8974 vector signed short vec_vsplth (vector signed short, const int);
8975 vector unsigned short vec_vsplth (vector unsigned short, const int);
8976 vector pixel vec_vsplth (vector pixel, const int);
8978 vector signed char vec_vspltb (vector signed char, const int);
8979 vector unsigned char vec_vspltb (vector unsigned char, const int);
8980 vector bool char vec_vspltb (vector bool char, const int);
8982 vector signed char vec_splat_s8 (const int);
8984 vector signed short vec_splat_s16 (const int);
8986 vector signed int vec_splat_s32 (const int);
8988 vector unsigned char vec_splat_u8 (const int);
8990 vector unsigned short vec_splat_u16 (const int);
8992 vector unsigned int vec_splat_u32 (const int);
8994 vector signed char vec_sr (vector signed char, vector unsigned char);
8995 vector unsigned char vec_sr (vector unsigned char,
8996 vector unsigned char);
8997 vector signed short vec_sr (vector signed short,
8998 vector unsigned short);
8999 vector unsigned short vec_sr (vector unsigned short,
9000 vector unsigned short);
9001 vector signed int vec_sr (vector signed int, vector unsigned int);
9002 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9004 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9005 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9007 vector signed short vec_vsrh (vector signed short,
9008 vector unsigned short);
9009 vector unsigned short vec_vsrh (vector unsigned short,
9010 vector unsigned short);
9012 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9013 vector unsigned char vec_vsrb (vector unsigned char,
9014 vector unsigned char);
9016 vector signed char vec_sra (vector signed char, vector unsigned char);
9017 vector unsigned char vec_sra (vector unsigned char,
9018 vector unsigned char);
9019 vector signed short vec_sra (vector signed short,
9020 vector unsigned short);
9021 vector unsigned short vec_sra (vector unsigned short,
9022 vector unsigned short);
9023 vector signed int vec_sra (vector signed int, vector unsigned int);
9024 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9026 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9027 vector unsigned int vec_vsraw (vector unsigned int,
9028 vector unsigned int);
9030 vector signed short vec_vsrah (vector signed short,
9031 vector unsigned short);
9032 vector unsigned short vec_vsrah (vector unsigned short,
9033 vector unsigned short);
9035 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9036 vector unsigned char vec_vsrab (vector unsigned char,
9037 vector unsigned char);
9039 vector signed int vec_srl (vector signed int, vector unsigned int);
9040 vector signed int vec_srl (vector signed int, vector unsigned short);
9041 vector signed int vec_srl (vector signed int, vector unsigned char);
9042 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9043 vector unsigned int vec_srl (vector unsigned int,
9044 vector unsigned short);
9045 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9046 vector bool int vec_srl (vector bool int, vector unsigned int);
9047 vector bool int vec_srl (vector bool int, vector unsigned short);
9048 vector bool int vec_srl (vector bool int, vector unsigned char);
9049 vector signed short vec_srl (vector signed short, vector unsigned int);
9050 vector signed short vec_srl (vector signed short,
9051 vector unsigned short);
9052 vector signed short vec_srl (vector signed short, vector unsigned char);
9053 vector unsigned short vec_srl (vector unsigned short,
9054 vector unsigned int);
9055 vector unsigned short vec_srl (vector unsigned short,
9056 vector unsigned short);
9057 vector unsigned short vec_srl (vector unsigned short,
9058 vector unsigned char);
9059 vector bool short vec_srl (vector bool short, vector unsigned int);
9060 vector bool short vec_srl (vector bool short, vector unsigned short);
9061 vector bool short vec_srl (vector bool short, vector unsigned char);
9062 vector pixel vec_srl (vector pixel, vector unsigned int);
9063 vector pixel vec_srl (vector pixel, vector unsigned short);
9064 vector pixel vec_srl (vector pixel, vector unsigned char);
9065 vector signed char vec_srl (vector signed char, vector unsigned int);
9066 vector signed char vec_srl (vector signed char, vector unsigned short);
9067 vector signed char vec_srl (vector signed char, vector unsigned char);
9068 vector unsigned char vec_srl (vector unsigned char,
9069 vector unsigned int);
9070 vector unsigned char vec_srl (vector unsigned char,
9071 vector unsigned short);
9072 vector unsigned char vec_srl (vector unsigned char,
9073 vector unsigned char);
9074 vector bool char vec_srl (vector bool char, vector unsigned int);
9075 vector bool char vec_srl (vector bool char, vector unsigned short);
9076 vector bool char vec_srl (vector bool char, vector unsigned char);
9078 vector float vec_sro (vector float, vector signed char);
9079 vector float vec_sro (vector float, vector unsigned char);
9080 vector signed int vec_sro (vector signed int, vector signed char);
9081 vector signed int vec_sro (vector signed int, vector unsigned char);
9082 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9083 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9084 vector signed short vec_sro (vector signed short, vector signed char);
9085 vector signed short vec_sro (vector signed short, vector unsigned char);
9086 vector unsigned short vec_sro (vector unsigned short,
9087 vector signed char);
9088 vector unsigned short vec_sro (vector unsigned short,
9089 vector unsigned char);
9090 vector pixel vec_sro (vector pixel, vector signed char);
9091 vector pixel vec_sro (vector pixel, vector unsigned char);
9092 vector signed char vec_sro (vector signed char, vector signed char);
9093 vector signed char vec_sro (vector signed char, vector unsigned char);
9094 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9095 vector unsigned char vec_sro (vector unsigned char,
9096 vector unsigned char);
9098 void vec_st (vector float, int, vector float *);
9099 void vec_st (vector float, int, float *);
9100 void vec_st (vector signed int, int, vector signed int *);
9101 void vec_st (vector signed int, int, int *);
9102 void vec_st (vector unsigned int, int, vector unsigned int *);
9103 void vec_st (vector unsigned int, int, unsigned int *);
9104 void vec_st (vector bool int, int, vector bool int *);
9105 void vec_st (vector bool int, int, unsigned int *);
9106 void vec_st (vector bool int, int, int *);
9107 void vec_st (vector signed short, int, vector signed short *);
9108 void vec_st (vector signed short, int, short *);
9109 void vec_st (vector unsigned short, int, vector unsigned short *);
9110 void vec_st (vector unsigned short, int, unsigned short *);
9111 void vec_st (vector bool short, int, vector bool short *);
9112 void vec_st (vector bool short, int, unsigned short *);
9113 void vec_st (vector pixel, int, vector pixel *);
9114 void vec_st (vector pixel, int, unsigned short *);
9115 void vec_st (vector pixel, int, short *);
9116 void vec_st (vector bool short, int, short *);
9117 void vec_st (vector signed char, int, vector signed char *);
9118 void vec_st (vector signed char, int, signed char *);
9119 void vec_st (vector unsigned char, int, vector unsigned char *);
9120 void vec_st (vector unsigned char, int, unsigned char *);
9121 void vec_st (vector bool char, int, vector bool char *);
9122 void vec_st (vector bool char, int, unsigned char *);
9123 void vec_st (vector bool char, int, signed char *);
9125 void vec_ste (vector signed char, int, signed char *);
9126 void vec_ste (vector unsigned char, int, unsigned char *);
9127 void vec_ste (vector bool char, int, signed char *);
9128 void vec_ste (vector bool char, int, unsigned char *);
9129 void vec_ste (vector signed short, int, short *);
9130 void vec_ste (vector unsigned short, int, unsigned short *);
9131 void vec_ste (vector bool short, int, short *);
9132 void vec_ste (vector bool short, int, unsigned short *);
9133 void vec_ste (vector pixel, int, short *);
9134 void vec_ste (vector pixel, int, unsigned short *);
9135 void vec_ste (vector float, int, float *);
9136 void vec_ste (vector signed int, int, int *);
9137 void vec_ste (vector unsigned int, int, unsigned int *);
9138 void vec_ste (vector bool int, int, int *);
9139 void vec_ste (vector bool int, int, unsigned int *);
9141 void vec_stvewx (vector float, int, float *);
9142 void vec_stvewx (vector signed int, int, int *);
9143 void vec_stvewx (vector unsigned int, int, unsigned int *);
9144 void vec_stvewx (vector bool int, int, int *);
9145 void vec_stvewx (vector bool int, int, unsigned int *);
9147 void vec_stvehx (vector signed short, int, short *);
9148 void vec_stvehx (vector unsigned short, int, unsigned short *);
9149 void vec_stvehx (vector bool short, int, short *);
9150 void vec_stvehx (vector bool short, int, unsigned short *);
9151 void vec_stvehx (vector pixel, int, short *);
9152 void vec_stvehx (vector pixel, int, unsigned short *);
9154 void vec_stvebx (vector signed char, int, signed char *);
9155 void vec_stvebx (vector unsigned char, int, unsigned char *);
9156 void vec_stvebx (vector bool char, int, signed char *);
9157 void vec_stvebx (vector bool char, int, unsigned char *);
9159 void vec_stl (vector float, int, vector float *);
9160 void vec_stl (vector float, int, float *);
9161 void vec_stl (vector signed int, int, vector signed int *);
9162 void vec_stl (vector signed int, int, int *);
9163 void vec_stl (vector unsigned int, int, vector unsigned int *);
9164 void vec_stl (vector unsigned int, int, unsigned int *);
9165 void vec_stl (vector bool int, int, vector bool int *);
9166 void vec_stl (vector bool int, int, unsigned int *);
9167 void vec_stl (vector bool int, int, int *);
9168 void vec_stl (vector signed short, int, vector signed short *);
9169 void vec_stl (vector signed short, int, short *);
9170 void vec_stl (vector unsigned short, int, vector unsigned short *);
9171 void vec_stl (vector unsigned short, int, unsigned short *);
9172 void vec_stl (vector bool short, int, vector bool short *);
9173 void vec_stl (vector bool short, int, unsigned short *);
9174 void vec_stl (vector bool short, int, short *);
9175 void vec_stl (vector pixel, int, vector pixel *);
9176 void vec_stl (vector pixel, int, unsigned short *);
9177 void vec_stl (vector pixel, int, short *);
9178 void vec_stl (vector signed char, int, vector signed char *);
9179 void vec_stl (vector signed char, int, signed char *);
9180 void vec_stl (vector unsigned char, int, vector unsigned char *);
9181 void vec_stl (vector unsigned char, int, unsigned char *);
9182 void vec_stl (vector bool char, int, vector bool char *);
9183 void vec_stl (vector bool char, int, unsigned char *);
9184 void vec_stl (vector bool char, int, signed char *);
9186 vector signed char vec_sub (vector bool char, vector signed char);
9187 vector signed char vec_sub (vector signed char, vector bool char);
9188 vector signed char vec_sub (vector signed char, vector signed char);
9189 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9190 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9191 vector unsigned char vec_sub (vector unsigned char,
9192 vector unsigned char);
9193 vector signed short vec_sub (vector bool short, vector signed short);
9194 vector signed short vec_sub (vector signed short, vector bool short);
9195 vector signed short vec_sub (vector signed short, vector signed short);
9196 vector unsigned short vec_sub (vector bool short,
9197 vector unsigned short);
9198 vector unsigned short vec_sub (vector unsigned short,
9200 vector unsigned short vec_sub (vector unsigned short,
9201 vector unsigned short);
9202 vector signed int vec_sub (vector bool int, vector signed int);
9203 vector signed int vec_sub (vector signed int, vector bool int);
9204 vector signed int vec_sub (vector signed int, vector signed int);
9205 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9206 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9207 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9208 vector float vec_sub (vector float, vector float);
9210 vector float vec_vsubfp (vector float, vector float);
9212 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9213 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9214 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9215 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9216 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9217 vector unsigned int vec_vsubuwm (vector unsigned int,
9218 vector unsigned int);
9220 vector signed short vec_vsubuhm (vector bool short,
9221 vector signed short);
9222 vector signed short vec_vsubuhm (vector signed short,
9224 vector signed short vec_vsubuhm (vector signed short,
9225 vector signed short);
9226 vector unsigned short vec_vsubuhm (vector bool short,
9227 vector unsigned short);
9228 vector unsigned short vec_vsubuhm (vector unsigned short,
9230 vector unsigned short vec_vsubuhm (vector unsigned short,
9231 vector unsigned short);
9233 vector signed char vec_vsububm (vector bool char, vector signed char);
9234 vector signed char vec_vsububm (vector signed char, vector bool char);
9235 vector signed char vec_vsububm (vector signed char, vector signed char);
9236 vector unsigned char vec_vsububm (vector bool char,
9237 vector unsigned char);
9238 vector unsigned char vec_vsububm (vector unsigned char,
9240 vector unsigned char vec_vsububm (vector unsigned char,
9241 vector unsigned char);
9243 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9245 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9246 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9247 vector unsigned char vec_subs (vector unsigned char,
9248 vector unsigned char);
9249 vector signed char vec_subs (vector bool char, vector signed char);
9250 vector signed char vec_subs (vector signed char, vector bool char);
9251 vector signed char vec_subs (vector signed char, vector signed char);
9252 vector unsigned short vec_subs (vector bool short,
9253 vector unsigned short);
9254 vector unsigned short vec_subs (vector unsigned short,
9256 vector unsigned short vec_subs (vector unsigned short,
9257 vector unsigned short);
9258 vector signed short vec_subs (vector bool short, vector signed short);
9259 vector signed short vec_subs (vector signed short, vector bool short);
9260 vector signed short vec_subs (vector signed short, vector signed short);
9261 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9262 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9263 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9264 vector signed int vec_subs (vector bool int, vector signed int);
9265 vector signed int vec_subs (vector signed int, vector bool int);
9266 vector signed int vec_subs (vector signed int, vector signed int);
9268 vector signed int vec_vsubsws (vector bool int, vector signed int);
9269 vector signed int vec_vsubsws (vector signed int, vector bool int);
9270 vector signed int vec_vsubsws (vector signed int, vector signed int);
9272 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9273 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9274 vector unsigned int vec_vsubuws (vector unsigned int,
9275 vector unsigned int);
9277 vector signed short vec_vsubshs (vector bool short,
9278 vector signed short);
9279 vector signed short vec_vsubshs (vector signed short,
9281 vector signed short vec_vsubshs (vector signed short,
9282 vector signed short);
9284 vector unsigned short vec_vsubuhs (vector bool short,
9285 vector unsigned short);
9286 vector unsigned short vec_vsubuhs (vector unsigned short,
9288 vector unsigned short vec_vsubuhs (vector unsigned short,
9289 vector unsigned short);
9291 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9292 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9293 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9295 vector unsigned char vec_vsububs (vector bool char,
9296 vector unsigned char);
9297 vector unsigned char vec_vsububs (vector unsigned char,
9299 vector unsigned char vec_vsububs (vector unsigned char,
9300 vector unsigned char);
9302 vector unsigned int vec_sum4s (vector unsigned char,
9303 vector unsigned int);
9304 vector signed int vec_sum4s (vector signed char, vector signed int);
9305 vector signed int vec_sum4s (vector signed short, vector signed int);
9307 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9309 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9311 vector unsigned int vec_vsum4ubs (vector unsigned char,
9312 vector unsigned int);
9314 vector signed int vec_sum2s (vector signed int, vector signed int);
9316 vector signed int vec_sums (vector signed int, vector signed int);
9318 vector float vec_trunc (vector float);
9320 vector signed short vec_unpackh (vector signed char);
9321 vector bool short vec_unpackh (vector bool char);
9322 vector signed int vec_unpackh (vector signed short);
9323 vector bool int vec_unpackh (vector bool short);
9324 vector unsigned int vec_unpackh (vector pixel);
9326 vector bool int vec_vupkhsh (vector bool short);
9327 vector signed int vec_vupkhsh (vector signed short);
9329 vector unsigned int vec_vupkhpx (vector pixel);
9331 vector bool short vec_vupkhsb (vector bool char);
9332 vector signed short vec_vupkhsb (vector signed char);
9334 vector signed short vec_unpackl (vector signed char);
9335 vector bool short vec_unpackl (vector bool char);
9336 vector unsigned int vec_unpackl (vector pixel);
9337 vector signed int vec_unpackl (vector signed short);
9338 vector bool int vec_unpackl (vector bool short);
9340 vector unsigned int vec_vupklpx (vector pixel);
9342 vector bool int vec_vupklsh (vector bool short);
9343 vector signed int vec_vupklsh (vector signed short);
9345 vector bool short vec_vupklsb (vector bool char);
9346 vector signed short vec_vupklsb (vector signed char);
9348 vector float vec_xor (vector float, vector float);
9349 vector float vec_xor (vector float, vector bool int);
9350 vector float vec_xor (vector bool int, vector float);
9351 vector bool int vec_xor (vector bool int, vector bool int);
9352 vector signed int vec_xor (vector bool int, vector signed int);
9353 vector signed int vec_xor (vector signed int, vector bool int);
9354 vector signed int vec_xor (vector signed int, vector signed int);
9355 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9356 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9357 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9358 vector bool short vec_xor (vector bool short, vector bool short);
9359 vector signed short vec_xor (vector bool short, vector signed short);
9360 vector signed short vec_xor (vector signed short, vector bool short);
9361 vector signed short vec_xor (vector signed short, vector signed short);
9362 vector unsigned short vec_xor (vector bool short,
9363 vector unsigned short);
9364 vector unsigned short vec_xor (vector unsigned short,
9366 vector unsigned short vec_xor (vector unsigned short,
9367 vector unsigned short);
9368 vector signed char vec_xor (vector bool char, vector signed char);
9369 vector bool char vec_xor (vector bool char, vector bool char);
9370 vector signed char vec_xor (vector signed char, vector bool char);
9371 vector signed char vec_xor (vector signed char, vector signed char);
9372 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9373 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9374 vector unsigned char vec_xor (vector unsigned char,
9375 vector unsigned char);
9377 int vec_all_eq (vector signed char, vector bool char);
9378 int vec_all_eq (vector signed char, vector signed char);
9379 int vec_all_eq (vector unsigned char, vector bool char);
9380 int vec_all_eq (vector unsigned char, vector unsigned char);
9381 int vec_all_eq (vector bool char, vector bool char);
9382 int vec_all_eq (vector bool char, vector unsigned char);
9383 int vec_all_eq (vector bool char, vector signed char);
9384 int vec_all_eq (vector signed short, vector bool short);
9385 int vec_all_eq (vector signed short, vector signed short);
9386 int vec_all_eq (vector unsigned short, vector bool short);
9387 int vec_all_eq (vector unsigned short, vector unsigned short);
9388 int vec_all_eq (vector bool short, vector bool short);
9389 int vec_all_eq (vector bool short, vector unsigned short);
9390 int vec_all_eq (vector bool short, vector signed short);
9391 int vec_all_eq (vector pixel, vector pixel);
9392 int vec_all_eq (vector signed int, vector bool int);
9393 int vec_all_eq (vector signed int, vector signed int);
9394 int vec_all_eq (vector unsigned int, vector bool int);
9395 int vec_all_eq (vector unsigned int, vector unsigned int);
9396 int vec_all_eq (vector bool int, vector bool int);
9397 int vec_all_eq (vector bool int, vector unsigned int);
9398 int vec_all_eq (vector bool int, vector signed int);
9399 int vec_all_eq (vector float, vector float);
9401 int vec_all_ge (vector bool char, vector unsigned char);
9402 int vec_all_ge (vector unsigned char, vector bool char);
9403 int vec_all_ge (vector unsigned char, vector unsigned char);
9404 int vec_all_ge (vector bool char, vector signed char);
9405 int vec_all_ge (vector signed char, vector bool char);
9406 int vec_all_ge (vector signed char, vector signed char);
9407 int vec_all_ge (vector bool short, vector unsigned short);
9408 int vec_all_ge (vector unsigned short, vector bool short);
9409 int vec_all_ge (vector unsigned short, vector unsigned short);
9410 int vec_all_ge (vector signed short, vector signed short);
9411 int vec_all_ge (vector bool short, vector signed short);
9412 int vec_all_ge (vector signed short, vector bool short);
9413 int vec_all_ge (vector bool int, vector unsigned int);
9414 int vec_all_ge (vector unsigned int, vector bool int);
9415 int vec_all_ge (vector unsigned int, vector unsigned int);
9416 int vec_all_ge (vector bool int, vector signed int);
9417 int vec_all_ge (vector signed int, vector bool int);
9418 int vec_all_ge (vector signed int, vector signed int);
9419 int vec_all_ge (vector float, vector float);
9421 int vec_all_gt (vector bool char, vector unsigned char);
9422 int vec_all_gt (vector unsigned char, vector bool char);
9423 int vec_all_gt (vector unsigned char, vector unsigned char);
9424 int vec_all_gt (vector bool char, vector signed char);
9425 int vec_all_gt (vector signed char, vector bool char);
9426 int vec_all_gt (vector signed char, vector signed char);
9427 int vec_all_gt (vector bool short, vector unsigned short);
9428 int vec_all_gt (vector unsigned short, vector bool short);
9429 int vec_all_gt (vector unsigned short, vector unsigned short);
9430 int vec_all_gt (vector bool short, vector signed short);
9431 int vec_all_gt (vector signed short, vector bool short);
9432 int vec_all_gt (vector signed short, vector signed short);
9433 int vec_all_gt (vector bool int, vector unsigned int);
9434 int vec_all_gt (vector unsigned int, vector bool int);
9435 int vec_all_gt (vector unsigned int, vector unsigned int);
9436 int vec_all_gt (vector bool int, vector signed int);
9437 int vec_all_gt (vector signed int, vector bool int);
9438 int vec_all_gt (vector signed int, vector signed int);
9439 int vec_all_gt (vector float, vector float);
9441 int vec_all_in (vector float, vector float);
9443 int vec_all_le (vector bool char, vector unsigned char);
9444 int vec_all_le (vector unsigned char, vector bool char);
9445 int vec_all_le (vector unsigned char, vector unsigned char);
9446 int vec_all_le (vector bool char, vector signed char);
9447 int vec_all_le (vector signed char, vector bool char);
9448 int vec_all_le (vector signed char, vector signed char);
9449 int vec_all_le (vector bool short, vector unsigned short);
9450 int vec_all_le (vector unsigned short, vector bool short);
9451 int vec_all_le (vector unsigned short, vector unsigned short);
9452 int vec_all_le (vector bool short, vector signed short);
9453 int vec_all_le (vector signed short, vector bool short);
9454 int vec_all_le (vector signed short, vector signed short);
9455 int vec_all_le (vector bool int, vector unsigned int);
9456 int vec_all_le (vector unsigned int, vector bool int);
9457 int vec_all_le (vector unsigned int, vector unsigned int);
9458 int vec_all_le (vector bool int, vector signed int);
9459 int vec_all_le (vector signed int, vector bool int);
9460 int vec_all_le (vector signed int, vector signed int);
9461 int vec_all_le (vector float, vector float);
9463 int vec_all_lt (vector bool char, vector unsigned char);
9464 int vec_all_lt (vector unsigned char, vector bool char);
9465 int vec_all_lt (vector unsigned char, vector unsigned char);
9466 int vec_all_lt (vector bool char, vector signed char);
9467 int vec_all_lt (vector signed char, vector bool char);
9468 int vec_all_lt (vector signed char, vector signed char);
9469 int vec_all_lt (vector bool short, vector unsigned short);
9470 int vec_all_lt (vector unsigned short, vector bool short);
9471 int vec_all_lt (vector unsigned short, vector unsigned short);
9472 int vec_all_lt (vector bool short, vector signed short);
9473 int vec_all_lt (vector signed short, vector bool short);
9474 int vec_all_lt (vector signed short, vector signed short);
9475 int vec_all_lt (vector bool int, vector unsigned int);
9476 int vec_all_lt (vector unsigned int, vector bool int);
9477 int vec_all_lt (vector unsigned int, vector unsigned int);
9478 int vec_all_lt (vector bool int, vector signed int);
9479 int vec_all_lt (vector signed int, vector bool int);
9480 int vec_all_lt (vector signed int, vector signed int);
9481 int vec_all_lt (vector float, vector float);
9483 int vec_all_nan (vector float);
9485 int vec_all_ne (vector signed char, vector bool char);
9486 int vec_all_ne (vector signed char, vector signed char);
9487 int vec_all_ne (vector unsigned char, vector bool char);
9488 int vec_all_ne (vector unsigned char, vector unsigned char);
9489 int vec_all_ne (vector bool char, vector bool char);
9490 int vec_all_ne (vector bool char, vector unsigned char);
9491 int vec_all_ne (vector bool char, vector signed char);
9492 int vec_all_ne (vector signed short, vector bool short);
9493 int vec_all_ne (vector signed short, vector signed short);
9494 int vec_all_ne (vector unsigned short, vector bool short);
9495 int vec_all_ne (vector unsigned short, vector unsigned short);
9496 int vec_all_ne (vector bool short, vector bool short);
9497 int vec_all_ne (vector bool short, vector unsigned short);
9498 int vec_all_ne (vector bool short, vector signed short);
9499 int vec_all_ne (vector pixel, vector pixel);
9500 int vec_all_ne (vector signed int, vector bool int);
9501 int vec_all_ne (vector signed int, vector signed int);
9502 int vec_all_ne (vector unsigned int, vector bool int);
9503 int vec_all_ne (vector unsigned int, vector unsigned int);
9504 int vec_all_ne (vector bool int, vector bool int);
9505 int vec_all_ne (vector bool int, vector unsigned int);
9506 int vec_all_ne (vector bool int, vector signed int);
9507 int vec_all_ne (vector float, vector float);
9509 int vec_all_nge (vector float, vector float);
9511 int vec_all_ngt (vector float, vector float);
9513 int vec_all_nle (vector float, vector float);
9515 int vec_all_nlt (vector float, vector float);
9517 int vec_all_numeric (vector float);
9519 int vec_any_eq (vector signed char, vector bool char);
9520 int vec_any_eq (vector signed char, vector signed char);
9521 int vec_any_eq (vector unsigned char, vector bool char);
9522 int vec_any_eq (vector unsigned char, vector unsigned char);
9523 int vec_any_eq (vector bool char, vector bool char);
9524 int vec_any_eq (vector bool char, vector unsigned char);
9525 int vec_any_eq (vector bool char, vector signed char);
9526 int vec_any_eq (vector signed short, vector bool short);
9527 int vec_any_eq (vector signed short, vector signed short);
9528 int vec_any_eq (vector unsigned short, vector bool short);
9529 int vec_any_eq (vector unsigned short, vector unsigned short);
9530 int vec_any_eq (vector bool short, vector bool short);
9531 int vec_any_eq (vector bool short, vector unsigned short);
9532 int vec_any_eq (vector bool short, vector signed short);
9533 int vec_any_eq (vector pixel, vector pixel);
9534 int vec_any_eq (vector signed int, vector bool int);
9535 int vec_any_eq (vector signed int, vector signed int);
9536 int vec_any_eq (vector unsigned int, vector bool int);
9537 int vec_any_eq (vector unsigned int, vector unsigned int);
9538 int vec_any_eq (vector bool int, vector bool int);
9539 int vec_any_eq (vector bool int, vector unsigned int);
9540 int vec_any_eq (vector bool int, vector signed int);
9541 int vec_any_eq (vector float, vector float);
9543 int vec_any_ge (vector signed char, vector bool char);
9544 int vec_any_ge (vector unsigned char, vector bool char);
9545 int vec_any_ge (vector unsigned char, vector unsigned char);
9546 int vec_any_ge (vector signed char, vector signed char);
9547 int vec_any_ge (vector bool char, vector unsigned char);
9548 int vec_any_ge (vector bool char, vector signed char);
9549 int vec_any_ge (vector unsigned short, vector bool short);
9550 int vec_any_ge (vector unsigned short, vector unsigned short);
9551 int vec_any_ge (vector signed short, vector signed short);
9552 int vec_any_ge (vector signed short, vector bool short);
9553 int vec_any_ge (vector bool short, vector unsigned short);
9554 int vec_any_ge (vector bool short, vector signed short);
9555 int vec_any_ge (vector signed int, vector bool int);
9556 int vec_any_ge (vector unsigned int, vector bool int);
9557 int vec_any_ge (vector unsigned int, vector unsigned int);
9558 int vec_any_ge (vector signed int, vector signed int);
9559 int vec_any_ge (vector bool int, vector unsigned int);
9560 int vec_any_ge (vector bool int, vector signed int);
9561 int vec_any_ge (vector float, vector float);
9563 int vec_any_gt (vector bool char, vector unsigned char);
9564 int vec_any_gt (vector unsigned char, vector bool char);
9565 int vec_any_gt (vector unsigned char, vector unsigned char);
9566 int vec_any_gt (vector bool char, vector signed char);
9567 int vec_any_gt (vector signed char, vector bool char);
9568 int vec_any_gt (vector signed char, vector signed char);
9569 int vec_any_gt (vector bool short, vector unsigned short);
9570 int vec_any_gt (vector unsigned short, vector bool short);
9571 int vec_any_gt (vector unsigned short, vector unsigned short);
9572 int vec_any_gt (vector bool short, vector signed short);
9573 int vec_any_gt (vector signed short, vector bool short);
9574 int vec_any_gt (vector signed short, vector signed short);
9575 int vec_any_gt (vector bool int, vector unsigned int);
9576 int vec_any_gt (vector unsigned int, vector bool int);
9577 int vec_any_gt (vector unsigned int, vector unsigned int);
9578 int vec_any_gt (vector bool int, vector signed int);
9579 int vec_any_gt (vector signed int, vector bool int);
9580 int vec_any_gt (vector signed int, vector signed int);
9581 int vec_any_gt (vector float, vector float);
9583 int vec_any_le (vector bool char, vector unsigned char);
9584 int vec_any_le (vector unsigned char, vector bool char);
9585 int vec_any_le (vector unsigned char, vector unsigned char);
9586 int vec_any_le (vector bool char, vector signed char);
9587 int vec_any_le (vector signed char, vector bool char);
9588 int vec_any_le (vector signed char, vector signed char);
9589 int vec_any_le (vector bool short, vector unsigned short);
9590 int vec_any_le (vector unsigned short, vector bool short);
9591 int vec_any_le (vector unsigned short, vector unsigned short);
9592 int vec_any_le (vector bool short, vector signed short);
9593 int vec_any_le (vector signed short, vector bool short);
9594 int vec_any_le (vector signed short, vector signed short);
9595 int vec_any_le (vector bool int, vector unsigned int);
9596 int vec_any_le (vector unsigned int, vector bool int);
9597 int vec_any_le (vector unsigned int, vector unsigned int);
9598 int vec_any_le (vector bool int, vector signed int);
9599 int vec_any_le (vector signed int, vector bool int);
9600 int vec_any_le (vector signed int, vector signed int);
9601 int vec_any_le (vector float, vector float);
9603 int vec_any_lt (vector bool char, vector unsigned char);
9604 int vec_any_lt (vector unsigned char, vector bool char);
9605 int vec_any_lt (vector unsigned char, vector unsigned char);
9606 int vec_any_lt (vector bool char, vector signed char);
9607 int vec_any_lt (vector signed char, vector bool char);
9608 int vec_any_lt (vector signed char, vector signed char);
9609 int vec_any_lt (vector bool short, vector unsigned short);
9610 int vec_any_lt (vector unsigned short, vector bool short);
9611 int vec_any_lt (vector unsigned short, vector unsigned short);
9612 int vec_any_lt (vector bool short, vector signed short);
9613 int vec_any_lt (vector signed short, vector bool short);
9614 int vec_any_lt (vector signed short, vector signed short);
9615 int vec_any_lt (vector bool int, vector unsigned int);
9616 int vec_any_lt (vector unsigned int, vector bool int);
9617 int vec_any_lt (vector unsigned int, vector unsigned int);
9618 int vec_any_lt (vector bool int, vector signed int);
9619 int vec_any_lt (vector signed int, vector bool int);
9620 int vec_any_lt (vector signed int, vector signed int);
9621 int vec_any_lt (vector float, vector float);
9623 int vec_any_nan (vector float);
9625 int vec_any_ne (vector signed char, vector bool char);
9626 int vec_any_ne (vector signed char, vector signed char);
9627 int vec_any_ne (vector unsigned char, vector bool char);
9628 int vec_any_ne (vector unsigned char, vector unsigned char);
9629 int vec_any_ne (vector bool char, vector bool char);
9630 int vec_any_ne (vector bool char, vector unsigned char);
9631 int vec_any_ne (vector bool char, vector signed char);
9632 int vec_any_ne (vector signed short, vector bool short);
9633 int vec_any_ne (vector signed short, vector signed short);
9634 int vec_any_ne (vector unsigned short, vector bool short);
9635 int vec_any_ne (vector unsigned short, vector unsigned short);
9636 int vec_any_ne (vector bool short, vector bool short);
9637 int vec_any_ne (vector bool short, vector unsigned short);
9638 int vec_any_ne (vector bool short, vector signed short);
9639 int vec_any_ne (vector pixel, vector pixel);
9640 int vec_any_ne (vector signed int, vector bool int);
9641 int vec_any_ne (vector signed int, vector signed int);
9642 int vec_any_ne (vector unsigned int, vector bool int);
9643 int vec_any_ne (vector unsigned int, vector unsigned int);
9644 int vec_any_ne (vector bool int, vector bool int);
9645 int vec_any_ne (vector bool int, vector unsigned int);
9646 int vec_any_ne (vector bool int, vector signed int);
9647 int vec_any_ne (vector float, vector float);
9649 int vec_any_nge (vector float, vector float);
9651 int vec_any_ngt (vector float, vector float);
9653 int vec_any_nle (vector float, vector float);
9655 int vec_any_nlt (vector float, vector float);
9657 int vec_any_numeric (vector float);
9659 int vec_any_out (vector float, vector float);
9662 @node SPARC VIS Built-in Functions
9663 @subsection SPARC VIS Built-in Functions
9665 GCC supports SIMD operations on the SPARC using both the generic vector
9666 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9667 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9668 switch, the VIS extension is exposed as the following built-in functions:
9671 typedef int v2si __attribute__ ((vector_size (8)));
9672 typedef short v4hi __attribute__ ((vector_size (8)));
9673 typedef short v2hi __attribute__ ((vector_size (4)));
9674 typedef char v8qi __attribute__ ((vector_size (8)));
9675 typedef char v4qi __attribute__ ((vector_size (4)));
9677 void * __builtin_vis_alignaddr (void *, long);
9678 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9679 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9680 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9681 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9683 v4hi __builtin_vis_fexpand (v4qi);
9685 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9686 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9687 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9688 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9689 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9690 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9691 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9693 v4qi __builtin_vis_fpack16 (v4hi);
9694 v8qi __builtin_vis_fpack32 (v2si, v2si);
9695 v2hi __builtin_vis_fpackfix (v2si);
9696 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9698 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9701 @node Target Format Checks
9702 @section Format Checks Specific to Particular Target Machines
9704 For some target machines, GCC supports additional options to the
9706 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9709 * Solaris Format Checks::
9712 @node Solaris Format Checks
9713 @subsection Solaris Format Checks
9715 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9716 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9717 conversions, and the two-argument @code{%b} conversion for displaying
9718 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9721 @section Pragmas Accepted by GCC
9725 GCC supports several types of pragmas, primarily in order to compile
9726 code originally written for other compilers. Note that in general
9727 we do not recommend the use of pragmas; @xref{Function Attributes},
9728 for further explanation.
9733 * RS/6000 and PowerPC Pragmas::
9736 * Symbol-Renaming Pragmas::
9737 * Structure-Packing Pragmas::
9739 * Diagnostic Pragmas::
9740 * Visibility Pragmas::
9744 @subsection ARM Pragmas
9746 The ARM target defines pragmas for controlling the default addition of
9747 @code{long_call} and @code{short_call} attributes to functions.
9748 @xref{Function Attributes}, for information about the effects of these
9753 @cindex pragma, long_calls
9754 Set all subsequent functions to have the @code{long_call} attribute.
9757 @cindex pragma, no_long_calls
9758 Set all subsequent functions to have the @code{short_call} attribute.
9760 @item long_calls_off
9761 @cindex pragma, long_calls_off
9762 Do not affect the @code{long_call} or @code{short_call} attributes of
9763 subsequent functions.
9767 @subsection M32C Pragmas
9770 @item memregs @var{number}
9771 @cindex pragma, memregs
9772 Overrides the command line option @code{-memregs=} for the current
9773 file. Use with care! This pragma must be before any function in the
9774 file, and mixing different memregs values in different objects may
9775 make them incompatible. This pragma is useful when a
9776 performance-critical function uses a memreg for temporary values,
9777 as it may allow you to reduce the number of memregs used.
9781 @node RS/6000 and PowerPC Pragmas
9782 @subsection RS/6000 and PowerPC Pragmas
9784 The RS/6000 and PowerPC targets define one pragma for controlling
9785 whether or not the @code{longcall} attribute is added to function
9786 declarations by default. This pragma overrides the @option{-mlongcall}
9787 option, but not the @code{longcall} and @code{shortcall} attributes.
9788 @xref{RS/6000 and PowerPC Options}, for more information about when long
9789 calls are and are not necessary.
9793 @cindex pragma, longcall
9794 Apply the @code{longcall} attribute to all subsequent function
9798 Do not apply the @code{longcall} attribute to subsequent function
9802 @c Describe c4x pragmas here.
9803 @c Describe h8300 pragmas here.
9804 @c Describe sh pragmas here.
9805 @c Describe v850 pragmas here.
9807 @node Darwin Pragmas
9808 @subsection Darwin Pragmas
9810 The following pragmas are available for all architectures running the
9811 Darwin operating system. These are useful for compatibility with other
9815 @item mark @var{tokens}@dots{}
9816 @cindex pragma, mark
9817 This pragma is accepted, but has no effect.
9819 @item options align=@var{alignment}
9820 @cindex pragma, options align
9821 This pragma sets the alignment of fields in structures. The values of
9822 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9823 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9824 properly; to restore the previous setting, use @code{reset} for the
9827 @item segment @var{tokens}@dots{}
9828 @cindex pragma, segment
9829 This pragma is accepted, but has no effect.
9831 @item unused (@var{var} [, @var{var}]@dots{})
9832 @cindex pragma, unused
9833 This pragma declares variables to be possibly unused. GCC will not
9834 produce warnings for the listed variables. The effect is similar to
9835 that of the @code{unused} attribute, except that this pragma may appear
9836 anywhere within the variables' scopes.
9839 @node Solaris Pragmas
9840 @subsection Solaris Pragmas
9842 The Solaris target supports @code{#pragma redefine_extname}
9843 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
9844 @code{#pragma} directives for compatibility with the system compiler.
9847 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
9848 @cindex pragma, align
9850 Increase the minimum alignment of each @var{variable} to @var{alignment}.
9851 This is the same as GCC's @code{aligned} attribute @pxref{Variable
9852 Attributes}). Macro expansion occurs on the arguments to this pragma
9853 when compiling C and Objective-C. It does not currently occur when
9854 compiling C++, but this is a bug which may be fixed in a future
9857 @item fini (@var{function} [, @var{function}]...)
9858 @cindex pragma, fini
9860 This pragma causes each listed @var{function} to be called after
9861 main, or during shared module unloading, by adding a call to the
9862 @code{.fini} section.
9864 @item init (@var{function} [, @var{function}]...)
9865 @cindex pragma, init
9867 This pragma causes each listed @var{function} to be called during
9868 initialization (before @code{main}) or during shared module loading, by
9869 adding a call to the @code{.init} section.
9873 @node Symbol-Renaming Pragmas
9874 @subsection Symbol-Renaming Pragmas
9876 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
9877 supports two @code{#pragma} directives which change the name used in
9878 assembly for a given declaration. These pragmas are only available on
9879 platforms whose system headers need them. To get this effect on all
9880 platforms supported by GCC, use the asm labels extension (@pxref{Asm
9884 @item redefine_extname @var{oldname} @var{newname}
9885 @cindex pragma, redefine_extname
9887 This pragma gives the C function @var{oldname} the assembly symbol
9888 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
9889 will be defined if this pragma is available (currently only on
9892 @item extern_prefix @var{string}
9893 @cindex pragma, extern_prefix
9895 This pragma causes all subsequent external function and variable
9896 declarations to have @var{string} prepended to their assembly symbols.
9897 This effect may be terminated with another @code{extern_prefix} pragma
9898 whose argument is an empty string. The preprocessor macro
9899 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
9900 available (currently only on Tru64 UNIX)@.
9903 These pragmas and the asm labels extension interact in a complicated
9904 manner. Here are some corner cases you may want to be aware of.
9907 @item Both pragmas silently apply only to declarations with external
9908 linkage. Asm labels do not have this restriction.
9910 @item In C++, both pragmas silently apply only to declarations with
9911 ``C'' linkage. Again, asm labels do not have this restriction.
9913 @item If any of the three ways of changing the assembly name of a
9914 declaration is applied to a declaration whose assembly name has
9915 already been determined (either by a previous use of one of these
9916 features, or because the compiler needed the assembly name in order to
9917 generate code), and the new name is different, a warning issues and
9918 the name does not change.
9920 @item The @var{oldname} used by @code{#pragma redefine_extname} is
9921 always the C-language name.
9923 @item If @code{#pragma extern_prefix} is in effect, and a declaration
9924 occurs with an asm label attached, the prefix is silently ignored for
9927 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
9928 apply to the same declaration, whichever triggered first wins, and a
9929 warning issues if they contradict each other. (We would like to have
9930 @code{#pragma redefine_extname} always win, for consistency with asm
9931 labels, but if @code{#pragma extern_prefix} triggers first we have no
9932 way of knowing that that happened.)
9935 @node Structure-Packing Pragmas
9936 @subsection Structure-Packing Pragmas
9938 For compatibility with Win32, GCC supports a set of @code{#pragma}
9939 directives which change the maximum alignment of members of structures
9940 (other than zero-width bitfields), unions, and classes subsequently
9941 defined. The @var{n} value below always is required to be a small power
9942 of two and specifies the new alignment in bytes.
9945 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
9946 @item @code{#pragma pack()} sets the alignment to the one that was in
9947 effect when compilation started (see also command line option
9948 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
9949 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
9950 setting on an internal stack and then optionally sets the new alignment.
9951 @item @code{#pragma pack(pop)} restores the alignment setting to the one
9952 saved at the top of the internal stack (and removes that stack entry).
9953 Note that @code{#pragma pack([@var{n}])} does not influence this internal
9954 stack; thus it is possible to have @code{#pragma pack(push)} followed by
9955 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
9956 @code{#pragma pack(pop)}.
9959 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
9960 @code{#pragma} which lays out a structure as the documented
9961 @code{__attribute__ ((ms_struct))}.
9963 @item @code{#pragma ms_struct on} turns on the layout for structures
9965 @item @code{#pragma ms_struct off} turns off the layout for structures
9967 @item @code{#pragma ms_struct reset} goes back to the default layout.
9971 @subsection Weak Pragmas
9973 For compatibility with SVR4, GCC supports a set of @code{#pragma}
9974 directives for declaring symbols to be weak, and defining weak
9978 @item #pragma weak @var{symbol}
9979 @cindex pragma, weak
9980 This pragma declares @var{symbol} to be weak, as if the declaration
9981 had the attribute of the same name. The pragma may appear before
9982 or after the declaration of @var{symbol}, but must appear before
9983 either its first use or its definition. It is not an error for
9984 @var{symbol} to never be defined at all.
9986 @item #pragma weak @var{symbol1} = @var{symbol2}
9987 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
9988 It is an error if @var{symbol2} is not defined in the current
9992 @node Diagnostic Pragmas
9993 @subsection Diagnostic Pragmas
9995 GCC allows the user to selectively enable or disable certain types of
9996 diagnostics, and change the kind of the diagnostic. For example, a
9997 project's policy might require that all sources compile with
9998 @option{-Werror} but certain files might have exceptions allowing
9999 specific types of warnings. Or, a project might selectively enable
10000 diagnostics and treat them as errors depending on which preprocessor
10001 macros are defined.
10004 @item #pragma GCC diagnostic @var{kind} @var{option}
10005 @cindex pragma, diagnostic
10007 Modifies the disposition of a diagnostic. Note that not all
10008 diagnostics are modifiable; at the moment only warnings (normally
10009 controlled by @samp{-W...}) can be controlled, and not all of them.
10010 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10011 are controllable and which option controls them.
10013 @var{kind} is @samp{error} to treat this diagnostic as an error,
10014 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10015 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10016 @var{option} is a double quoted string which matches the command line
10020 #pragma GCC diagnostic warning "-Wformat"
10021 #pragma GCC diagnostic error "-Wformat"
10022 #pragma GCC diagnostic ignored "-Wformat"
10025 Note that these pragmas override any command line options. Also,
10026 while it is syntactically valid to put these pragmas anywhere in your
10027 sources, the only supported location for them is before any data or
10028 functions are defined. Doing otherwise may result in unpredictable
10029 results depending on how the optimizer manages your sources. If the
10030 same option is listed multiple times, the last one specified is the
10031 one that is in effect. This pragma is not intended to be a general
10032 purpose replacement for command line options, but for implementing
10033 strict control over project policies.
10037 @node Visibility Pragmas
10038 @subsection Visibility Pragmas
10041 @item #pragma GCC visibility push(@var{visibility})
10042 @itemx #pragma GCC visibility pop
10043 @cindex pragma, visibility
10045 This pragma allows the user to set the visibility for multiple
10046 declarations without having to give each a visibility attribute
10047 @xref{Function Attributes}, for more information about visibility and
10048 the attribute syntax.
10050 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10051 declarations. Class members and template specializations are not
10052 affected; if you want to override the visibility for a particular
10053 member or instantiation, you must use an attribute.
10057 @node Unnamed Fields
10058 @section Unnamed struct/union fields within structs/unions
10062 For compatibility with other compilers, GCC allows you to define
10063 a structure or union that contains, as fields, structures and unions
10064 without names. For example:
10077 In this example, the user would be able to access members of the unnamed
10078 union with code like @samp{foo.b}. Note that only unnamed structs and
10079 unions are allowed, you may not have, for example, an unnamed
10082 You must never create such structures that cause ambiguous field definitions.
10083 For example, this structure:
10094 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10095 Such constructs are not supported and must be avoided. In the future,
10096 such constructs may be detected and treated as compilation errors.
10098 @opindex fms-extensions
10099 Unless @option{-fms-extensions} is used, the unnamed field must be a
10100 structure or union definition without a tag (for example, @samp{struct
10101 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10102 also be a definition with a tag such as @samp{struct foo @{ int a;
10103 @};}, a reference to a previously defined structure or union such as
10104 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10105 previously defined structure or union type.
10108 @section Thread-Local Storage
10109 @cindex Thread-Local Storage
10110 @cindex @acronym{TLS}
10113 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10114 are allocated such that there is one instance of the variable per extant
10115 thread. The run-time model GCC uses to implement this originates
10116 in the IA-64 processor-specific ABI, but has since been migrated
10117 to other processors as well. It requires significant support from
10118 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10119 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10120 is not available everywhere.
10122 At the user level, the extension is visible with a new storage
10123 class keyword: @code{__thread}. For example:
10127 extern __thread struct state s;
10128 static __thread char *p;
10131 The @code{__thread} specifier may be used alone, with the @code{extern}
10132 or @code{static} specifiers, but with no other storage class specifier.
10133 When used with @code{extern} or @code{static}, @code{__thread} must appear
10134 immediately after the other storage class specifier.
10136 The @code{__thread} specifier may be applied to any global, file-scoped
10137 static, function-scoped static, or static data member of a class. It may
10138 not be applied to block-scoped automatic or non-static data member.
10140 When the address-of operator is applied to a thread-local variable, it is
10141 evaluated at run-time and returns the address of the current thread's
10142 instance of that variable. An address so obtained may be used by any
10143 thread. When a thread terminates, any pointers to thread-local variables
10144 in that thread become invalid.
10146 No static initialization may refer to the address of a thread-local variable.
10148 In C++, if an initializer is present for a thread-local variable, it must
10149 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10152 See @uref{http://people.redhat.com/drepper/tls.pdf,
10153 ELF Handling For Thread-Local Storage} for a detailed explanation of
10154 the four thread-local storage addressing models, and how the run-time
10155 is expected to function.
10158 * C99 Thread-Local Edits::
10159 * C++98 Thread-Local Edits::
10162 @node C99 Thread-Local Edits
10163 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10165 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10166 that document the exact semantics of the language extension.
10170 @cite{5.1.2 Execution environments}
10172 Add new text after paragraph 1
10175 Within either execution environment, a @dfn{thread} is a flow of
10176 control within a program. It is implementation defined whether
10177 or not there may be more than one thread associated with a program.
10178 It is implementation defined how threads beyond the first are
10179 created, the name and type of the function called at thread
10180 startup, and how threads may be terminated. However, objects
10181 with thread storage duration shall be initialized before thread
10186 @cite{6.2.4 Storage durations of objects}
10188 Add new text before paragraph 3
10191 An object whose identifier is declared with the storage-class
10192 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10193 Its lifetime is the entire execution of the thread, and its
10194 stored value is initialized only once, prior to thread startup.
10198 @cite{6.4.1 Keywords}
10200 Add @code{__thread}.
10203 @cite{6.7.1 Storage-class specifiers}
10205 Add @code{__thread} to the list of storage class specifiers in
10208 Change paragraph 2 to
10211 With the exception of @code{__thread}, at most one storage-class
10212 specifier may be given [@dots{}]. The @code{__thread} specifier may
10213 be used alone, or immediately following @code{extern} or
10217 Add new text after paragraph 6
10220 The declaration of an identifier for a variable that has
10221 block scope that specifies @code{__thread} shall also
10222 specify either @code{extern} or @code{static}.
10224 The @code{__thread} specifier shall be used only with
10229 @node C++98 Thread-Local Edits
10230 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10232 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10233 that document the exact semantics of the language extension.
10237 @b{[intro.execution]}
10239 New text after paragraph 4
10242 A @dfn{thread} is a flow of control within the abstract machine.
10243 It is implementation defined whether or not there may be more than
10247 New text after paragraph 7
10250 It is unspecified whether additional action must be taken to
10251 ensure when and whether side effects are visible to other threads.
10257 Add @code{__thread}.
10260 @b{[basic.start.main]}
10262 Add after paragraph 5
10265 The thread that begins execution at the @code{main} function is called
10266 the @dfn{main thread}. It is implementation defined how functions
10267 beginning threads other than the main thread are designated or typed.
10268 A function so designated, as well as the @code{main} function, is called
10269 a @dfn{thread startup function}. It is implementation defined what
10270 happens if a thread startup function returns. It is implementation
10271 defined what happens to other threads when any thread calls @code{exit}.
10275 @b{[basic.start.init]}
10277 Add after paragraph 4
10280 The storage for an object of thread storage duration shall be
10281 statically initialized before the first statement of the thread startup
10282 function. An object of thread storage duration shall not require
10283 dynamic initialization.
10287 @b{[basic.start.term]}
10289 Add after paragraph 3
10292 The type of an object with thread storage duration shall not have a
10293 non-trivial destructor, nor shall it be an array type whose elements
10294 (directly or indirectly) have non-trivial destructors.
10300 Add ``thread storage duration'' to the list in paragraph 1.
10305 Thread, static, and automatic storage durations are associated with
10306 objects introduced by declarations [@dots{}].
10309 Add @code{__thread} to the list of specifiers in paragraph 3.
10312 @b{[basic.stc.thread]}
10314 New section before @b{[basic.stc.static]}
10317 The keyword @code{__thread} applied to a non-local object gives the
10318 object thread storage duration.
10320 A local variable or class data member declared both @code{static}
10321 and @code{__thread} gives the variable or member thread storage
10326 @b{[basic.stc.static]}
10331 All objects which have neither thread storage duration, dynamic
10332 storage duration nor are local [@dots{}].
10338 Add @code{__thread} to the list in paragraph 1.
10343 With the exception of @code{__thread}, at most one
10344 @var{storage-class-specifier} shall appear in a given
10345 @var{decl-specifier-seq}. The @code{__thread} specifier may
10346 be used alone, or immediately following the @code{extern} or
10347 @code{static} specifiers. [@dots{}]
10350 Add after paragraph 5
10353 The @code{__thread} specifier can be applied only to the names of objects
10354 and to anonymous unions.
10360 Add after paragraph 6
10363 Non-@code{static} members shall not be @code{__thread}.
10367 @node C++ Extensions
10368 @chapter Extensions to the C++ Language
10369 @cindex extensions, C++ language
10370 @cindex C++ language extensions
10372 The GNU compiler provides these extensions to the C++ language (and you
10373 can also use most of the C language extensions in your C++ programs). If you
10374 want to write code that checks whether these features are available, you can
10375 test for the GNU compiler the same way as for C programs: check for a
10376 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10377 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10378 Predefined Macros,cpp,The GNU C Preprocessor}).
10381 * Volatiles:: What constitutes an access to a volatile object.
10382 * Restricted Pointers:: C99 restricted pointers and references.
10383 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10384 * C++ Interface:: You can use a single C++ header file for both
10385 declarations and definitions.
10386 * Template Instantiation:: Methods for ensuring that exactly one copy of
10387 each needed template instantiation is emitted.
10388 * Bound member functions:: You can extract a function pointer to the
10389 method denoted by a @samp{->*} or @samp{.*} expression.
10390 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10391 * Namespace Association:: Strong using-directives for namespace association.
10392 * Java Exceptions:: Tweaking exception handling to work with Java.
10393 * Deprecated Features:: Things will disappear from g++.
10394 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10398 @section When is a Volatile Object Accessed?
10399 @cindex accessing volatiles
10400 @cindex volatile read
10401 @cindex volatile write
10402 @cindex volatile access
10404 Both the C and C++ standard have the concept of volatile objects. These
10405 are normally accessed by pointers and used for accessing hardware. The
10406 standards encourage compilers to refrain from optimizations concerning
10407 accesses to volatile objects. The C standard leaves it implementation
10408 defined as to what constitutes a volatile access. The C++ standard omits
10409 to specify this, except to say that C++ should behave in a similar manner
10410 to C with respect to volatiles, where possible. The minimum either
10411 standard specifies is that at a sequence point all previous accesses to
10412 volatile objects have stabilized and no subsequent accesses have
10413 occurred. Thus an implementation is free to reorder and combine
10414 volatile accesses which occur between sequence points, but cannot do so
10415 for accesses across a sequence point. The use of volatiles does not
10416 allow you to violate the restriction on updating objects multiple times
10417 within a sequence point.
10419 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10421 The behavior differs slightly between C and C++ in the non-obvious cases:
10424 volatile int *src = @var{somevalue};
10428 With C, such expressions are rvalues, and GCC interprets this either as a
10429 read of the volatile object being pointed to or only as request to evaluate
10430 the side-effects. The C++ standard specifies that such expressions do not
10431 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10432 object may be incomplete. The C++ standard does not specify explicitly
10433 that it is this lvalue to rvalue conversion which may be responsible for
10434 causing an access. However, there is reason to believe that it is,
10435 because otherwise certain simple expressions become undefined. However,
10436 because it would surprise most programmers, G++ treats dereferencing a
10437 pointer to volatile object of complete type when the value is unused as
10438 GCC would do for an equivalent type in C. When the object has incomplete
10439 type, G++ issues a warning; if you wish to force an error, you must
10440 force a conversion to rvalue with, for instance, a static cast.
10442 When using a reference to volatile, G++ does not treat equivalent
10443 expressions as accesses to volatiles, but instead issues a warning that
10444 no volatile is accessed. The rationale for this is that otherwise it
10445 becomes difficult to determine where volatile access occur, and not
10446 possible to ignore the return value from functions returning volatile
10447 references. Again, if you wish to force a read, cast the reference to
10450 @node Restricted Pointers
10451 @section Restricting Pointer Aliasing
10452 @cindex restricted pointers
10453 @cindex restricted references
10454 @cindex restricted this pointer
10456 As with the C front end, G++ understands the C99 feature of restricted pointers,
10457 specified with the @code{__restrict__}, or @code{__restrict} type
10458 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10459 language flag, @code{restrict} is not a keyword in C++.
10461 In addition to allowing restricted pointers, you can specify restricted
10462 references, which indicate that the reference is not aliased in the local
10466 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10473 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10474 @var{rref} refers to a (different) unaliased integer.
10476 You may also specify whether a member function's @var{this} pointer is
10477 unaliased by using @code{__restrict__} as a member function qualifier.
10480 void T::fn () __restrict__
10487 Within the body of @code{T::fn}, @var{this} will have the effective
10488 definition @code{T *__restrict__ const this}. Notice that the
10489 interpretation of a @code{__restrict__} member function qualifier is
10490 different to that of @code{const} or @code{volatile} qualifier, in that it
10491 is applied to the pointer rather than the object. This is consistent with
10492 other compilers which implement restricted pointers.
10494 As with all outermost parameter qualifiers, @code{__restrict__} is
10495 ignored in function definition matching. This means you only need to
10496 specify @code{__restrict__} in a function definition, rather than
10497 in a function prototype as well.
10499 @node Vague Linkage
10500 @section Vague Linkage
10501 @cindex vague linkage
10503 There are several constructs in C++ which require space in the object
10504 file but are not clearly tied to a single translation unit. We say that
10505 these constructs have ``vague linkage''. Typically such constructs are
10506 emitted wherever they are needed, though sometimes we can be more
10510 @item Inline Functions
10511 Inline functions are typically defined in a header file which can be
10512 included in many different compilations. Hopefully they can usually be
10513 inlined, but sometimes an out-of-line copy is necessary, if the address
10514 of the function is taken or if inlining fails. In general, we emit an
10515 out-of-line copy in all translation units where one is needed. As an
10516 exception, we only emit inline virtual functions with the vtable, since
10517 it will always require a copy.
10519 Local static variables and string constants used in an inline function
10520 are also considered to have vague linkage, since they must be shared
10521 between all inlined and out-of-line instances of the function.
10525 C++ virtual functions are implemented in most compilers using a lookup
10526 table, known as a vtable. The vtable contains pointers to the virtual
10527 functions provided by a class, and each object of the class contains a
10528 pointer to its vtable (or vtables, in some multiple-inheritance
10529 situations). If the class declares any non-inline, non-pure virtual
10530 functions, the first one is chosen as the ``key method'' for the class,
10531 and the vtable is only emitted in the translation unit where the key
10534 @emph{Note:} If the chosen key method is later defined as inline, the
10535 vtable will still be emitted in every translation unit which defines it.
10536 Make sure that any inline virtuals are declared inline in the class
10537 body, even if they are not defined there.
10539 @item type_info objects
10542 C++ requires information about types to be written out in order to
10543 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10544 For polymorphic classes (classes with virtual functions), the type_info
10545 object is written out along with the vtable so that @samp{dynamic_cast}
10546 can determine the dynamic type of a class object at runtime. For all
10547 other types, we write out the type_info object when it is used: when
10548 applying @samp{typeid} to an expression, throwing an object, or
10549 referring to a type in a catch clause or exception specification.
10551 @item Template Instantiations
10552 Most everything in this section also applies to template instantiations,
10553 but there are other options as well.
10554 @xref{Template Instantiation,,Where's the Template?}.
10558 When used with GNU ld version 2.8 or later on an ELF system such as
10559 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10560 these constructs will be discarded at link time. This is known as
10563 On targets that don't support COMDAT, but do support weak symbols, GCC
10564 will use them. This way one copy will override all the others, but
10565 the unused copies will still take up space in the executable.
10567 For targets which do not support either COMDAT or weak symbols,
10568 most entities with vague linkage will be emitted as local symbols to
10569 avoid duplicate definition errors from the linker. This will not happen
10570 for local statics in inlines, however, as having multiple copies will
10571 almost certainly break things.
10573 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10574 another way to control placement of these constructs.
10576 @node C++ Interface
10577 @section #pragma interface and implementation
10579 @cindex interface and implementation headers, C++
10580 @cindex C++ interface and implementation headers
10581 @cindex pragmas, interface and implementation
10583 @code{#pragma interface} and @code{#pragma implementation} provide the
10584 user with a way of explicitly directing the compiler to emit entities
10585 with vague linkage (and debugging information) in a particular
10588 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10589 most cases, because of COMDAT support and the ``key method'' heuristic
10590 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10591 program to grow due to unnecessary out-of-line copies of inline
10592 functions. Currently (3.4) the only benefit of these
10593 @code{#pragma}s is reduced duplication of debugging information, and
10594 that should be addressed soon on DWARF 2 targets with the use of
10598 @item #pragma interface
10599 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10600 @kindex #pragma interface
10601 Use this directive in @emph{header files} that define object classes, to save
10602 space in most of the object files that use those classes. Normally,
10603 local copies of certain information (backup copies of inline member
10604 functions, debugging information, and the internal tables that implement
10605 virtual functions) must be kept in each object file that includes class
10606 definitions. You can use this pragma to avoid such duplication. When a
10607 header file containing @samp{#pragma interface} is included in a
10608 compilation, this auxiliary information will not be generated (unless
10609 the main input source file itself uses @samp{#pragma implementation}).
10610 Instead, the object files will contain references to be resolved at link
10613 The second form of this directive is useful for the case where you have
10614 multiple headers with the same name in different directories. If you
10615 use this form, you must specify the same string to @samp{#pragma
10618 @item #pragma implementation
10619 @itemx #pragma implementation "@var{objects}.h"
10620 @kindex #pragma implementation
10621 Use this pragma in a @emph{main input file}, when you want full output from
10622 included header files to be generated (and made globally visible). The
10623 included header file, in turn, should use @samp{#pragma interface}.
10624 Backup copies of inline member functions, debugging information, and the
10625 internal tables used to implement virtual functions are all generated in
10626 implementation files.
10628 @cindex implied @code{#pragma implementation}
10629 @cindex @code{#pragma implementation}, implied
10630 @cindex naming convention, implementation headers
10631 If you use @samp{#pragma implementation} with no argument, it applies to
10632 an include file with the same basename@footnote{A file's @dfn{basename}
10633 was the name stripped of all leading path information and of trailing
10634 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10635 file. For example, in @file{allclass.cc}, giving just
10636 @samp{#pragma implementation}
10637 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10639 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10640 an implementation file whenever you would include it from
10641 @file{allclass.cc} even if you never specified @samp{#pragma
10642 implementation}. This was deemed to be more trouble than it was worth,
10643 however, and disabled.
10645 Use the string argument if you want a single implementation file to
10646 include code from multiple header files. (You must also use
10647 @samp{#include} to include the header file; @samp{#pragma
10648 implementation} only specifies how to use the file---it doesn't actually
10651 There is no way to split up the contents of a single header file into
10652 multiple implementation files.
10655 @cindex inlining and C++ pragmas
10656 @cindex C++ pragmas, effect on inlining
10657 @cindex pragmas in C++, effect on inlining
10658 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10659 effect on function inlining.
10661 If you define a class in a header file marked with @samp{#pragma
10662 interface}, the effect on an inline function defined in that class is
10663 similar to an explicit @code{extern} declaration---the compiler emits
10664 no code at all to define an independent version of the function. Its
10665 definition is used only for inlining with its callers.
10667 @opindex fno-implement-inlines
10668 Conversely, when you include the same header file in a main source file
10669 that declares it as @samp{#pragma implementation}, the compiler emits
10670 code for the function itself; this defines a version of the function
10671 that can be found via pointers (or by callers compiled without
10672 inlining). If all calls to the function can be inlined, you can avoid
10673 emitting the function by compiling with @option{-fno-implement-inlines}.
10674 If any calls were not inlined, you will get linker errors.
10676 @node Template Instantiation
10677 @section Where's the Template?
10678 @cindex template instantiation
10680 C++ templates are the first language feature to require more
10681 intelligence from the environment than one usually finds on a UNIX
10682 system. Somehow the compiler and linker have to make sure that each
10683 template instance occurs exactly once in the executable if it is needed,
10684 and not at all otherwise. There are two basic approaches to this
10685 problem, which are referred to as the Borland model and the Cfront model.
10688 @item Borland model
10689 Borland C++ solved the template instantiation problem by adding the code
10690 equivalent of common blocks to their linker; the compiler emits template
10691 instances in each translation unit that uses them, and the linker
10692 collapses them together. The advantage of this model is that the linker
10693 only has to consider the object files themselves; there is no external
10694 complexity to worry about. This disadvantage is that compilation time
10695 is increased because the template code is being compiled repeatedly.
10696 Code written for this model tends to include definitions of all
10697 templates in the header file, since they must be seen to be
10701 The AT&T C++ translator, Cfront, solved the template instantiation
10702 problem by creating the notion of a template repository, an
10703 automatically maintained place where template instances are stored. A
10704 more modern version of the repository works as follows: As individual
10705 object files are built, the compiler places any template definitions and
10706 instantiations encountered in the repository. At link time, the link
10707 wrapper adds in the objects in the repository and compiles any needed
10708 instances that were not previously emitted. The advantages of this
10709 model are more optimal compilation speed and the ability to use the
10710 system linker; to implement the Borland model a compiler vendor also
10711 needs to replace the linker. The disadvantages are vastly increased
10712 complexity, and thus potential for error; for some code this can be
10713 just as transparent, but in practice it can been very difficult to build
10714 multiple programs in one directory and one program in multiple
10715 directories. Code written for this model tends to separate definitions
10716 of non-inline member templates into a separate file, which should be
10717 compiled separately.
10720 When used with GNU ld version 2.8 or later on an ELF system such as
10721 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10722 Borland model. On other systems, G++ implements neither automatic
10725 A future version of G++ will support a hybrid model whereby the compiler
10726 will emit any instantiations for which the template definition is
10727 included in the compile, and store template definitions and
10728 instantiation context information into the object file for the rest.
10729 The link wrapper will extract that information as necessary and invoke
10730 the compiler to produce the remaining instantiations. The linker will
10731 then combine duplicate instantiations.
10733 In the mean time, you have the following options for dealing with
10734 template instantiations:
10739 Compile your template-using code with @option{-frepo}. The compiler will
10740 generate files with the extension @samp{.rpo} listing all of the
10741 template instantiations used in the corresponding object files which
10742 could be instantiated there; the link wrapper, @samp{collect2}, will
10743 then update the @samp{.rpo} files to tell the compiler where to place
10744 those instantiations and rebuild any affected object files. The
10745 link-time overhead is negligible after the first pass, as the compiler
10746 will continue to place the instantiations in the same files.
10748 This is your best option for application code written for the Borland
10749 model, as it will just work. Code written for the Cfront model will
10750 need to be modified so that the template definitions are available at
10751 one or more points of instantiation; usually this is as simple as adding
10752 @code{#include <tmethods.cc>} to the end of each template header.
10754 For library code, if you want the library to provide all of the template
10755 instantiations it needs, just try to link all of its object files
10756 together; the link will fail, but cause the instantiations to be
10757 generated as a side effect. Be warned, however, that this may cause
10758 conflicts if multiple libraries try to provide the same instantiations.
10759 For greater control, use explicit instantiation as described in the next
10763 @opindex fno-implicit-templates
10764 Compile your code with @option{-fno-implicit-templates} to disable the
10765 implicit generation of template instances, and explicitly instantiate
10766 all the ones you use. This approach requires more knowledge of exactly
10767 which instances you need than do the others, but it's less
10768 mysterious and allows greater control. You can scatter the explicit
10769 instantiations throughout your program, perhaps putting them in the
10770 translation units where the instances are used or the translation units
10771 that define the templates themselves; you can put all of the explicit
10772 instantiations you need into one big file; or you can create small files
10779 template class Foo<int>;
10780 template ostream& operator <<
10781 (ostream&, const Foo<int>&);
10784 for each of the instances you need, and create a template instantiation
10785 library from those.
10787 If you are using Cfront-model code, you can probably get away with not
10788 using @option{-fno-implicit-templates} when compiling files that don't
10789 @samp{#include} the member template definitions.
10791 If you use one big file to do the instantiations, you may want to
10792 compile it without @option{-fno-implicit-templates} so you get all of the
10793 instances required by your explicit instantiations (but not by any
10794 other files) without having to specify them as well.
10796 G++ has extended the template instantiation syntax given in the ISO
10797 standard to allow forward declaration of explicit instantiations
10798 (with @code{extern}), instantiation of the compiler support data for a
10799 template class (i.e.@: the vtable) without instantiating any of its
10800 members (with @code{inline}), and instantiation of only the static data
10801 members of a template class, without the support data or member
10802 functions (with (@code{static}):
10805 extern template int max (int, int);
10806 inline template class Foo<int>;
10807 static template class Foo<int>;
10811 Do nothing. Pretend G++ does implement automatic instantiation
10812 management. Code written for the Borland model will work fine, but
10813 each translation unit will contain instances of each of the templates it
10814 uses. In a large program, this can lead to an unacceptable amount of code
10818 @node Bound member functions
10819 @section Extracting the function pointer from a bound pointer to member function
10821 @cindex pointer to member function
10822 @cindex bound pointer to member function
10824 In C++, pointer to member functions (PMFs) are implemented using a wide
10825 pointer of sorts to handle all the possible call mechanisms; the PMF
10826 needs to store information about how to adjust the @samp{this} pointer,
10827 and if the function pointed to is virtual, where to find the vtable, and
10828 where in the vtable to look for the member function. If you are using
10829 PMFs in an inner loop, you should really reconsider that decision. If
10830 that is not an option, you can extract the pointer to the function that
10831 would be called for a given object/PMF pair and call it directly inside
10832 the inner loop, to save a bit of time.
10834 Note that you will still be paying the penalty for the call through a
10835 function pointer; on most modern architectures, such a call defeats the
10836 branch prediction features of the CPU@. This is also true of normal
10837 virtual function calls.
10839 The syntax for this extension is
10843 extern int (A::*fp)();
10844 typedef int (*fptr)(A *);
10846 fptr p = (fptr)(a.*fp);
10849 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
10850 no object is needed to obtain the address of the function. They can be
10851 converted to function pointers directly:
10854 fptr p1 = (fptr)(&A::foo);
10857 @opindex Wno-pmf-conversions
10858 You must specify @option{-Wno-pmf-conversions} to use this extension.
10860 @node C++ Attributes
10861 @section C++-Specific Variable, Function, and Type Attributes
10863 Some attributes only make sense for C++ programs.
10866 @item init_priority (@var{priority})
10867 @cindex init_priority attribute
10870 In Standard C++, objects defined at namespace scope are guaranteed to be
10871 initialized in an order in strict accordance with that of their definitions
10872 @emph{in a given translation unit}. No guarantee is made for initializations
10873 across translation units. However, GNU C++ allows users to control the
10874 order of initialization of objects defined at namespace scope with the
10875 @code{init_priority} attribute by specifying a relative @var{priority},
10876 a constant integral expression currently bounded between 101 and 65535
10877 inclusive. Lower numbers indicate a higher priority.
10879 In the following example, @code{A} would normally be created before
10880 @code{B}, but the @code{init_priority} attribute has reversed that order:
10883 Some_Class A __attribute__ ((init_priority (2000)));
10884 Some_Class B __attribute__ ((init_priority (543)));
10888 Note that the particular values of @var{priority} do not matter; only their
10891 @item java_interface
10892 @cindex java_interface attribute
10894 This type attribute informs C++ that the class is a Java interface. It may
10895 only be applied to classes declared within an @code{extern "Java"} block.
10896 Calls to methods declared in this interface will be dispatched using GCJ's
10897 interface table mechanism, instead of regular virtual table dispatch.
10901 See also @xref{Namespace Association}.
10903 @node Namespace Association
10904 @section Namespace Association
10906 @strong{Caution:} The semantics of this extension are not fully
10907 defined. Users should refrain from using this extension as its
10908 semantics may change subtly over time. It is possible that this
10909 extension will be removed in future versions of G++.
10911 A using-directive with @code{__attribute ((strong))} is stronger
10912 than a normal using-directive in two ways:
10916 Templates from the used namespace can be specialized and explicitly
10917 instantiated as though they were members of the using namespace.
10920 The using namespace is considered an associated namespace of all
10921 templates in the used namespace for purposes of argument-dependent
10925 The used namespace must be nested within the using namespace so that
10926 normal unqualified lookup works properly.
10928 This is useful for composing a namespace transparently from
10929 implementation namespaces. For example:
10934 template <class T> struct A @{ @};
10936 using namespace debug __attribute ((__strong__));
10937 template <> struct A<int> @{ @}; // @r{ok to specialize}
10939 template <class T> void f (A<T>);
10944 f (std::A<float>()); // @r{lookup finds} std::f
10949 @node Java Exceptions
10950 @section Java Exceptions
10952 The Java language uses a slightly different exception handling model
10953 from C++. Normally, GNU C++ will automatically detect when you are
10954 writing C++ code that uses Java exceptions, and handle them
10955 appropriately. However, if C++ code only needs to execute destructors
10956 when Java exceptions are thrown through it, GCC will guess incorrectly.
10957 Sample problematic code is:
10960 struct S @{ ~S(); @};
10961 extern void bar(); // @r{is written in Java, and may throw exceptions}
10970 The usual effect of an incorrect guess is a link failure, complaining of
10971 a missing routine called @samp{__gxx_personality_v0}.
10973 You can inform the compiler that Java exceptions are to be used in a
10974 translation unit, irrespective of what it might think, by writing
10975 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
10976 @samp{#pragma} must appear before any functions that throw or catch
10977 exceptions, or run destructors when exceptions are thrown through them.
10979 You cannot mix Java and C++ exceptions in the same translation unit. It
10980 is believed to be safe to throw a C++ exception from one file through
10981 another file compiled for the Java exception model, or vice versa, but
10982 there may be bugs in this area.
10984 @node Deprecated Features
10985 @section Deprecated Features
10987 In the past, the GNU C++ compiler was extended to experiment with new
10988 features, at a time when the C++ language was still evolving. Now that
10989 the C++ standard is complete, some of those features are superseded by
10990 superior alternatives. Using the old features might cause a warning in
10991 some cases that the feature will be dropped in the future. In other
10992 cases, the feature might be gone already.
10994 While the list below is not exhaustive, it documents some of the options
10995 that are now deprecated:
10998 @item -fexternal-templates
10999 @itemx -falt-external-templates
11000 These are two of the many ways for G++ to implement template
11001 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11002 defines how template definitions have to be organized across
11003 implementation units. G++ has an implicit instantiation mechanism that
11004 should work just fine for standard-conforming code.
11006 @item -fstrict-prototype
11007 @itemx -fno-strict-prototype
11008 Previously it was possible to use an empty prototype parameter list to
11009 indicate an unspecified number of parameters (like C), rather than no
11010 parameters, as C++ demands. This feature has been removed, except where
11011 it is required for backwards compatibility @xref{Backwards Compatibility}.
11014 G++ allows a virtual function returning @samp{void *} to be overridden
11015 by one returning a different pointer type. This extension to the
11016 covariant return type rules is now deprecated and will be removed from a
11019 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11020 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11021 and will be removed in a future version. Code using these operators
11022 should be modified to use @code{std::min} and @code{std::max} instead.
11024 The named return value extension has been deprecated, and is now
11027 The use of initializer lists with new expressions has been deprecated,
11028 and is now removed from G++.
11030 Floating and complex non-type template parameters have been deprecated,
11031 and are now removed from G++.
11033 The implicit typename extension has been deprecated and is now
11036 The use of default arguments in function pointers, function typedefs
11037 and other places where they are not permitted by the standard is
11038 deprecated and will be removed from a future version of G++.
11040 G++ allows floating-point literals to appear in integral constant expressions,
11041 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11042 This extension is deprecated and will be removed from a future version.
11044 G++ allows static data members of const floating-point type to be declared
11045 with an initializer in a class definition. The standard only allows
11046 initializers for static members of const integral types and const
11047 enumeration types so this extension has been deprecated and will be removed
11048 from a future version.
11050 @node Backwards Compatibility
11051 @section Backwards Compatibility
11052 @cindex Backwards Compatibility
11053 @cindex ARM [Annotated C++ Reference Manual]
11055 Now that there is a definitive ISO standard C++, G++ has a specification
11056 to adhere to. The C++ language evolved over time, and features that
11057 used to be acceptable in previous drafts of the standard, such as the ARM
11058 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11059 compilation of C++ written to such drafts, G++ contains some backwards
11060 compatibilities. @emph{All such backwards compatibility features are
11061 liable to disappear in future versions of G++.} They should be considered
11062 deprecated @xref{Deprecated Features}.
11066 If a variable is declared at for scope, it used to remain in scope until
11067 the end of the scope which contained the for statement (rather than just
11068 within the for scope). G++ retains this, but issues a warning, if such a
11069 variable is accessed outside the for scope.
11071 @item Implicit C language
11072 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11073 scope to set the language. On such systems, all header files are
11074 implicitly scoped inside a C language scope. Also, an empty prototype
11075 @code{()} will be treated as an unspecified number of arguments, rather
11076 than no arguments, as C++ demands.