1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C. Most of them are also available
20 in C++. @xref{C++ Extensions,,Extensions to the C++ Language}, for
21 extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Decimal Float:: Decimal Floating Types.
37 * Hex Floats:: Hexadecimal floating-point constants.
38 * Zero Length:: Zero-length arrays.
39 * Variable Length:: Arrays whose length is computed at run time.
40 * Empty Structures:: Structures with no members.
41 * Variadic Macros:: Macros with a variable number of arguments.
42 * Escaped Newlines:: Slightly looser rules for escaped newlines.
43 * Subscripting:: Any array can be subscripted, even if not an lvalue.
44 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
45 * Initializers:: Non-constant initializers.
46 * Compound Literals:: Compound literals give structures, unions
48 * Designated Inits:: Labeling elements of initializers.
49 * Cast to Union:: Casting to union type from any member of the union.
50 * Case Ranges:: `case 1 ... 9' and such.
51 * Mixed Declarations:: Mixing declarations and code.
52 * Function Attributes:: Declaring that functions have no side effects,
53 or that they can never return.
54 * Attribute Syntax:: Formal syntax for attributes.
55 * Function Prototypes:: Prototype declarations and old-style definitions.
56 * C++ Comments:: C++ comments are recognized.
57 * Dollar Signs:: Dollar sign is allowed in identifiers.
58 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * 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.
84 * Binary constants:: Binary constants using the @samp{0b} prefix.
88 @section Statements and Declarations in Expressions
89 @cindex statements inside expressions
90 @cindex declarations inside expressions
91 @cindex expressions containing statements
92 @cindex macros, statements in expressions
94 @c the above section title wrapped and causes an underfull hbox.. i
95 @c changed it from "within" to "in". --mew 4feb93
96 A compound statement enclosed in parentheses may appear as an expression
97 in GNU C@. This allows you to use loops, switches, and local variables
100 Recall that a compound statement is a sequence of statements surrounded
101 by braces; in this construct, parentheses go around the braces. For
105 (@{ int y = foo (); int z;
112 is a valid (though slightly more complex than necessary) expression
113 for the absolute value of @code{foo ()}.
115 The last thing in the compound statement should be an expression
116 followed by a semicolon; the value of this subexpression serves as the
117 value of the entire construct. (If you use some other kind of statement
118 last within the braces, the construct has type @code{void}, and thus
119 effectively no value.)
121 This feature is especially useful in making macro definitions ``safe'' (so
122 that they evaluate each operand exactly once). For example, the
123 ``maximum'' function is commonly defined as a macro in standard C as
127 #define max(a,b) ((a) > (b) ? (a) : (b))
131 @cindex side effects, macro argument
132 But this definition computes either @var{a} or @var{b} twice, with bad
133 results if the operand has side effects. In GNU C, if you know the
134 type of the operands (here taken as @code{int}), you can define
135 the macro safely as follows:
138 #define maxint(a,b) \
139 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
142 Embedded statements are not allowed in constant expressions, such as
143 the value of an enumeration constant, the width of a bit-field, or
144 the initial value of a static variable.
146 If you don't know the type of the operand, you can still do this, but you
147 must use @code{typeof} (@pxref{Typeof}).
149 In G++, the result value of a statement expression undergoes array and
150 function pointer decay, and is returned by value to the enclosing
151 expression. For instance, if @code{A} is a class, then
160 will construct a temporary @code{A} object to hold the result of the
161 statement expression, and that will be used to invoke @code{Foo}.
162 Therefore the @code{this} pointer observed by @code{Foo} will not be the
165 Any temporaries created within a statement within a statement expression
166 will be destroyed at the statement's end. This makes statement
167 expressions inside macros slightly different from function calls. In
168 the latter case temporaries introduced during argument evaluation will
169 be destroyed at the end of the statement that includes the function
170 call. In the statement expression case they will be destroyed during
171 the statement expression. For instance,
174 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
175 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
185 will have different places where temporaries are destroyed. For the
186 @code{macro} case, the temporary @code{X} will be destroyed just after
187 the initialization of @code{b}. In the @code{function} case that
188 temporary will be destroyed when the function returns.
190 These considerations mean that it is probably a bad idea to use
191 statement-expressions of this form in header files that are designed to
192 work with C++. (Note that some versions of the GNU C Library contained
193 header files using statement-expression that lead to precisely this
196 Jumping into a statement expression with @code{goto} or using a
197 @code{switch} statement outside the statement expression with a
198 @code{case} or @code{default} label inside the statement expression is
199 not permitted. Jumping into a statement expression with a computed
200 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
201 Jumping out of a statement expression is permitted, but if the
202 statement expression is part of a larger expression then it is
203 unspecified which other subexpressions of that expression have been
204 evaluated except where the language definition requires certain
205 subexpressions to be evaluated before or after the statement
206 expression. In any case, as with a function call the evaluation of a
207 statement expression is not interleaved with the evaluation of other
208 parts of the containing expression. For example,
211 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
215 will call @code{foo} and @code{bar1} and will not call @code{baz} but
216 may or may not call @code{bar2}. If @code{bar2} is called, it will be
217 called after @code{foo} and before @code{bar1}
220 @section Locally Declared Labels
222 @cindex macros, local labels
224 GCC allows you to declare @dfn{local labels} in any nested block
225 scope. A local label is just like an ordinary label, but you can
226 only reference it (with a @code{goto} statement, or by taking its
227 address) within the block in which it was declared.
229 A local label declaration looks like this:
232 __label__ @var{label};
239 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
242 Local label declarations must come at the beginning of the block,
243 before any ordinary declarations or statements.
245 The label declaration defines the label @emph{name}, but does not define
246 the label itself. You must do this in the usual way, with
247 @code{@var{label}:}, within the statements of the statement expression.
249 The local label feature is useful for complex macros. If a macro
250 contains nested loops, a @code{goto} can be useful for breaking out of
251 them. However, an ordinary label whose scope is the whole function
252 cannot be used: if the macro can be expanded several times in one
253 function, the label will be multiply defined in that function. A
254 local label avoids this problem. For example:
257 #define SEARCH(value, array, target) \
260 typeof (target) _SEARCH_target = (target); \
261 typeof (*(array)) *_SEARCH_array = (array); \
264 for (i = 0; i < max; i++) \
265 for (j = 0; j < max; j++) \
266 if (_SEARCH_array[i][j] == _SEARCH_target) \
267 @{ (value) = i; goto found; @} \
273 This could also be written using a statement-expression:
276 #define SEARCH(array, target) \
279 typeof (target) _SEARCH_target = (target); \
280 typeof (*(array)) *_SEARCH_array = (array); \
283 for (i = 0; i < max; i++) \
284 for (j = 0; j < max; j++) \
285 if (_SEARCH_array[i][j] == _SEARCH_target) \
286 @{ value = i; goto found; @} \
293 Local label declarations also make the labels they declare visible to
294 nested functions, if there are any. @xref{Nested Functions}, for details.
296 @node Labels as Values
297 @section Labels as Values
298 @cindex labels as values
299 @cindex computed gotos
300 @cindex goto with computed label
301 @cindex address of a label
303 You can get the address of a label defined in the current function
304 (or a containing function) with the unary operator @samp{&&}. The
305 value has type @code{void *}. This value is a constant and can be used
306 wherever a constant of that type is valid. For example:
314 To use these values, you need to be able to jump to one. This is done
315 with the computed goto statement@footnote{The analogous feature in
316 Fortran is called an assigned goto, but that name seems inappropriate in
317 C, where one can do more than simply store label addresses in label
318 variables.}, @code{goto *@var{exp};}. For example,
325 Any expression of type @code{void *} is allowed.
327 One way of using these constants is in initializing a static array that
328 will serve as a jump table:
331 static void *array[] = @{ &&foo, &&bar, &&hack @};
334 Then you can select a label with indexing, like this:
341 Note that this does not check whether the subscript is in bounds---array
342 indexing in C never does that.
344 Such an array of label values serves a purpose much like that of the
345 @code{switch} statement. The @code{switch} statement is cleaner, so
346 use that rather than an array unless the problem does not fit a
347 @code{switch} statement very well.
349 Another use of label values is in an interpreter for threaded code.
350 The labels within the interpreter function can be stored in the
351 threaded code for super-fast dispatching.
353 You may not use this mechanism to jump to code in a different function.
354 If you do that, totally unpredictable things will happen. The best way to
355 avoid this is to store the label address only in automatic variables and
356 never pass it as an argument.
358 An alternate way to write the above example is
361 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
363 goto *(&&foo + array[i]);
367 This is more friendly to code living in shared libraries, as it reduces
368 the number of dynamic relocations that are needed, and by consequence,
369 allows the data to be read-only.
371 @node Nested Functions
372 @section Nested Functions
373 @cindex nested functions
374 @cindex downward funargs
377 A @dfn{nested function} is a function defined inside another function.
378 (Nested functions are not supported for GNU C++.) The nested function's
379 name is local to the block where it is defined. For example, here we
380 define a nested function named @code{square}, and call it twice:
384 foo (double a, double b)
386 double square (double z) @{ return z * z; @}
388 return square (a) + square (b);
393 The nested function can access all the variables of the containing
394 function that are visible at the point of its definition. This is
395 called @dfn{lexical scoping}. For example, here we show a nested
396 function which uses an inherited variable named @code{offset}:
400 bar (int *array, int offset, int size)
402 int access (int *array, int index)
403 @{ return array[index + offset]; @}
406 for (i = 0; i < size; i++)
407 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
412 Nested function definitions are permitted within functions in the places
413 where variable definitions are allowed; that is, in any block, mixed
414 with the other declarations and statements in the block.
416 It is possible to call the nested function from outside the scope of its
417 name by storing its address or passing the address to another function:
420 hack (int *array, int size)
422 void store (int index, int value)
423 @{ array[index] = value; @}
425 intermediate (store, size);
429 Here, the function @code{intermediate} receives the address of
430 @code{store} as an argument. If @code{intermediate} calls @code{store},
431 the arguments given to @code{store} are used to store into @code{array}.
432 But this technique works only so long as the containing function
433 (@code{hack}, in this example) does not exit.
435 If you try to call the nested function through its address after the
436 containing function has exited, all hell will break loose. If you try
437 to call it after a containing scope level has exited, and if it refers
438 to some of the variables that are no longer in scope, you may be lucky,
439 but it's not wise to take the risk. If, however, the nested function
440 does not refer to anything that has gone out of scope, you should be
443 GCC implements taking the address of a nested function using a technique
444 called @dfn{trampolines}. A paper describing them is available as
447 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
449 A nested function can jump to a label inherited from a containing
450 function, provided the label was explicitly declared in the containing
451 function (@pxref{Local Labels}). Such a jump returns instantly to the
452 containing function, exiting the nested function which did the
453 @code{goto} and any intermediate functions as well. Here is an example:
457 bar (int *array, int offset, int size)
460 int access (int *array, int index)
464 return array[index + offset];
468 for (i = 0; i < size; i++)
469 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
473 /* @r{Control comes here from @code{access}
474 if it detects an error.} */
481 A nested function always has no linkage. Declaring one with
482 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
483 before its definition, use @code{auto} (which is otherwise meaningless
484 for function declarations).
487 bar (int *array, int offset, int size)
490 auto int access (int *, int);
492 int access (int *array, int index)
496 return array[index + offset];
502 @node Constructing Calls
503 @section Constructing Function Calls
504 @cindex constructing calls
505 @cindex forwarding calls
507 Using the built-in functions described below, you can record
508 the arguments a function received, and call another function
509 with the same arguments, without knowing the number or types
512 You can also record the return value of that function call,
513 and later return that value, without knowing what data type
514 the function tried to return (as long as your caller expects
517 However, these built-in functions may interact badly with some
518 sophisticated features or other extensions of the language. It
519 is, therefore, not recommended to use them outside very simple
520 functions acting as mere forwarders for their arguments.
522 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
523 This built-in function returns a pointer to data
524 describing how to perform a call with the same arguments as were passed
525 to the current function.
527 The function saves the arg pointer register, structure value address,
528 and all registers that might be used to pass arguments to a function
529 into a block of memory allocated on the stack. Then it returns the
530 address of that block.
533 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
534 This built-in function invokes @var{function}
535 with a copy of the parameters described by @var{arguments}
538 The value of @var{arguments} should be the value returned by
539 @code{__builtin_apply_args}. The argument @var{size} specifies the size
540 of the stack argument data, in bytes.
542 This function returns a pointer to data describing
543 how to return whatever value was returned by @var{function}. The data
544 is saved in a block of memory allocated on the stack.
546 It is not always simple to compute the proper value for @var{size}. The
547 value is used by @code{__builtin_apply} to compute the amount of data
548 that should be pushed on the stack and copied from the incoming argument
552 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
553 This built-in function returns the value described by @var{result} from
554 the containing function. You should specify, for @var{result}, a value
555 returned by @code{__builtin_apply}.
559 @section Referring to a Type with @code{typeof}
562 @cindex macros, types of arguments
564 Another way to refer to the type of an expression is with @code{typeof}.
565 The syntax of using of this keyword looks like @code{sizeof}, but the
566 construct acts semantically like a type name defined with @code{typedef}.
568 There are two ways of writing the argument to @code{typeof}: with an
569 expression or with a type. Here is an example with an expression:
576 This assumes that @code{x} is an array of pointers to functions;
577 the type described is that of the values of the functions.
579 Here is an example with a typename as the argument:
586 Here the type described is that of pointers to @code{int}.
588 If you are writing a header file that must work when included in ISO C
589 programs, write @code{__typeof__} instead of @code{typeof}.
590 @xref{Alternate Keywords}.
592 A @code{typeof}-construct can be used anywhere a typedef name could be
593 used. For example, you can use it in a declaration, in a cast, or inside
594 of @code{sizeof} or @code{typeof}.
596 @code{typeof} is often useful in conjunction with the
597 statements-within-expressions feature. Here is how the two together can
598 be used to define a safe ``maximum'' macro that operates on any
599 arithmetic type and evaluates each of its arguments exactly once:
603 (@{ typeof (a) _a = (a); \
604 typeof (b) _b = (b); \
605 _a > _b ? _a : _b; @})
608 @cindex underscores in variables in macros
609 @cindex @samp{_} in variables in macros
610 @cindex local variables in macros
611 @cindex variables, local, in macros
612 @cindex macros, local variables in
614 The reason for using names that start with underscores for the local
615 variables is to avoid conflicts with variable names that occur within the
616 expressions that are substituted for @code{a} and @code{b}. Eventually we
617 hope to design a new form of declaration syntax that allows you to declare
618 variables whose scopes start only after their initializers; this will be a
619 more reliable way to prevent such conflicts.
622 Some more examples of the use of @code{typeof}:
626 This declares @code{y} with the type of what @code{x} points to.
633 This declares @code{y} as an array of such values.
640 This declares @code{y} as an array of pointers to characters:
643 typeof (typeof (char *)[4]) y;
647 It is equivalent to the following traditional C declaration:
653 To see the meaning of the declaration using @code{typeof}, and why it
654 might be a useful way to write, rewrite it with these macros:
657 #define pointer(T) typeof(T *)
658 #define array(T, N) typeof(T [N])
662 Now the declaration can be rewritten this way:
665 array (pointer (char), 4) y;
669 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
670 pointers to @code{char}.
673 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
674 a more limited extension which permitted one to write
677 typedef @var{T} = @var{expr};
681 with the effect of declaring @var{T} to have the type of the expression
682 @var{expr}. This extension does not work with GCC 3 (versions between
683 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
684 relies on it should be rewritten to use @code{typeof}:
687 typedef typeof(@var{expr}) @var{T};
691 This will work with all versions of GCC@.
694 @section Conditionals with Omitted Operands
695 @cindex conditional expressions, extensions
696 @cindex omitted middle-operands
697 @cindex middle-operands, omitted
698 @cindex extensions, @code{?:}
699 @cindex @code{?:} extensions
701 The middle operand in a conditional expression may be omitted. Then
702 if the first operand is nonzero, its value is the value of the conditional
705 Therefore, the expression
712 has the value of @code{x} if that is nonzero; otherwise, the value of
715 This example is perfectly equivalent to
721 @cindex side effect in ?:
722 @cindex ?: side effect
724 In this simple case, the ability to omit the middle operand is not
725 especially useful. When it becomes useful is when the first operand does,
726 or may (if it is a macro argument), contain a side effect. Then repeating
727 the operand in the middle would perform the side effect twice. Omitting
728 the middle operand uses the value already computed without the undesirable
729 effects of recomputing it.
732 @section Double-Word Integers
733 @cindex @code{long long} data types
734 @cindex double-word arithmetic
735 @cindex multiprecision arithmetic
736 @cindex @code{LL} integer suffix
737 @cindex @code{ULL} integer suffix
739 ISO C99 supports data types for integers that are at least 64 bits wide,
740 and as an extension GCC supports them in C89 mode and in C++.
741 Simply write @code{long long int} for a signed integer, or
742 @code{unsigned long long int} for an unsigned integer. To make an
743 integer constant of type @code{long long int}, add the suffix @samp{LL}
744 to the integer. To make an integer constant of type @code{unsigned long
745 long int}, add the suffix @samp{ULL} to the integer.
747 You can use these types in arithmetic like any other integer types.
748 Addition, subtraction, and bitwise boolean operations on these types
749 are open-coded on all types of machines. Multiplication is open-coded
750 if the machine supports fullword-to-doubleword a widening multiply
751 instruction. Division and shifts are open-coded only on machines that
752 provide special support. The operations that are not open-coded use
753 special library routines that come with GCC@.
755 There may be pitfalls when you use @code{long long} types for function
756 arguments, unless you declare function prototypes. If a function
757 expects type @code{int} for its argument, and you pass a value of type
758 @code{long long int}, confusion will result because the caller and the
759 subroutine will disagree about the number of bytes for the argument.
760 Likewise, if the function expects @code{long long int} and you pass
761 @code{int}. The best way to avoid such problems is to use prototypes.
764 @section Complex Numbers
765 @cindex complex numbers
766 @cindex @code{_Complex} keyword
767 @cindex @code{__complex__} keyword
769 ISO C99 supports complex floating data types, and as an extension GCC
770 supports them in C89 mode and in C++, and supports complex integer data
771 types which are not part of ISO C99. You can declare complex types
772 using the keyword @code{_Complex}. As an extension, the older GNU
773 keyword @code{__complex__} is also supported.
775 For example, @samp{_Complex double x;} declares @code{x} as a
776 variable whose real part and imaginary part are both of type
777 @code{double}. @samp{_Complex short int y;} declares @code{y} to
778 have real and imaginary parts of type @code{short int}; this is not
779 likely to be useful, but it shows that the set of complex types is
782 To write a constant with a complex data type, use the suffix @samp{i} or
783 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
784 has type @code{_Complex float} and @code{3i} has type
785 @code{_Complex int}. Such a constant always has a pure imaginary
786 value, but you can form any complex value you like by adding one to a
787 real constant. This is a GNU extension; if you have an ISO C99
788 conforming C library (such as GNU libc), and want to construct complex
789 constants of floating type, you should include @code{<complex.h>} and
790 use the macros @code{I} or @code{_Complex_I} instead.
792 @cindex @code{__real__} keyword
793 @cindex @code{__imag__} keyword
794 To extract the real part of a complex-valued expression @var{exp}, write
795 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
796 extract the imaginary part. This is a GNU extension; for values of
797 floating type, you should use the ISO C99 functions @code{crealf},
798 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
799 @code{cimagl}, declared in @code{<complex.h>} and also provided as
800 built-in functions by GCC@.
802 @cindex complex conjugation
803 The operator @samp{~} performs complex conjugation when used on a value
804 with a complex type. This is a GNU extension; for values of
805 floating type, you should use the ISO C99 functions @code{conjf},
806 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
807 provided as built-in functions by GCC@.
809 GCC can allocate complex automatic variables in a noncontiguous
810 fashion; it's even possible for the real part to be in a register while
811 the imaginary part is on the stack (or vice-versa). Only the DWARF2
812 debug info format can represent this, so use of DWARF2 is recommended.
813 If you are using the stabs debug info format, GCC describes a noncontiguous
814 complex variable as if it were two separate variables of noncomplex type.
815 If the variable's actual name is @code{foo}, the two fictitious
816 variables are named @code{foo$real} and @code{foo$imag}. You can
817 examine and set these two fictitious variables with your debugger.
820 @section Decimal Floating Types
821 @cindex decimal floating types
822 @cindex @code{_Decimal32} data type
823 @cindex @code{_Decimal64} data type
824 @cindex @code{_Decimal128} data type
825 @cindex @code{df} integer suffix
826 @cindex @code{dd} integer suffix
827 @cindex @code{dl} integer suffix
828 @cindex @code{DF} integer suffix
829 @cindex @code{DD} integer suffix
830 @cindex @code{DL} integer suffix
832 As an extension, the GNU C compiler supports decimal floating types as
833 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
834 floating types in GCC will evolve as the draft technical report changes.
835 Calling conventions for any target might also change. Not all targets
836 support decimal floating types.
838 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
839 @code{_Decimal128}. They use a radix of ten, unlike the floating types
840 @code{float}, @code{double}, and @code{long double} whose radix is not
841 specified by the C standard but is usually two.
843 Support for decimal floating types includes the arithmetic operators
844 add, subtract, multiply, divide; unary arithmetic operators;
845 relational operators; equality operators; and conversions to and from
846 integer and other floating types. Use a suffix @samp{df} or
847 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
848 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
851 GCC support of decimal float as specified by the draft technical report
856 Translation time data type (TTDT) is not supported.
859 Characteristics of decimal floating types are defined in header file
860 @file{decfloat.h} rather than @file{float.h}.
863 When the value of a decimal floating type cannot be represented in the
864 integer type to which it is being converted, the result is undefined
865 rather than the result value specified by the draft technical report.
868 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
869 are supported by the DWARF2 debug information format.
875 ISO C99 supports floating-point numbers written not only in the usual
876 decimal notation, such as @code{1.55e1}, but also numbers such as
877 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
878 supports this in C89 mode (except in some cases when strictly
879 conforming) and in C++. In that format the
880 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
881 mandatory. The exponent is a decimal number that indicates the power of
882 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
889 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
890 is the same as @code{1.55e1}.
892 Unlike for floating-point numbers in the decimal notation the exponent
893 is always required in the hexadecimal notation. Otherwise the compiler
894 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
895 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
896 extension for floating-point constants of type @code{float}.
899 @section Arrays of Length Zero
900 @cindex arrays of length zero
901 @cindex zero-length arrays
902 @cindex length-zero arrays
903 @cindex flexible array members
905 Zero-length arrays are allowed in GNU C@. They are very useful as the
906 last element of a structure which is really a header for a variable-length
915 struct line *thisline = (struct line *)
916 malloc (sizeof (struct line) + this_length);
917 thisline->length = this_length;
920 In ISO C90, you would have to give @code{contents} a length of 1, which
921 means either you waste space or complicate the argument to @code{malloc}.
923 In ISO C99, you would use a @dfn{flexible array member}, which is
924 slightly different in syntax and semantics:
928 Flexible array members are written as @code{contents[]} without
932 Flexible array members have incomplete type, and so the @code{sizeof}
933 operator may not be applied. As a quirk of the original implementation
934 of zero-length arrays, @code{sizeof} evaluates to zero.
937 Flexible array members may only appear as the last member of a
938 @code{struct} that is otherwise non-empty.
941 A structure containing a flexible array member, or a union containing
942 such a structure (possibly recursively), may not be a member of a
943 structure or an element of an array. (However, these uses are
944 permitted by GCC as extensions.)
947 GCC versions before 3.0 allowed zero-length arrays to be statically
948 initialized, as if they were flexible arrays. In addition to those
949 cases that were useful, it also allowed initializations in situations
950 that would corrupt later data. Non-empty initialization of zero-length
951 arrays is now treated like any case where there are more initializer
952 elements than the array holds, in that a suitable warning about "excess
953 elements in array" is given, and the excess elements (all of them, in
954 this case) are ignored.
956 Instead GCC allows static initialization of flexible array members.
957 This is equivalent to defining a new structure containing the original
958 structure followed by an array of sufficient size to contain the data.
959 I.e.@: in the following, @code{f1} is constructed as if it were declared
965 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
968 struct f1 f1; int data[3];
969 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
973 The convenience of this extension is that @code{f1} has the desired
974 type, eliminating the need to consistently refer to @code{f2.f1}.
976 This has symmetry with normal static arrays, in that an array of
977 unknown size is also written with @code{[]}.
979 Of course, this extension only makes sense if the extra data comes at
980 the end of a top-level object, as otherwise we would be overwriting
981 data at subsequent offsets. To avoid undue complication and confusion
982 with initialization of deeply nested arrays, we simply disallow any
983 non-empty initialization except when the structure is the top-level
987 struct foo @{ int x; int y[]; @};
988 struct bar @{ struct foo z; @};
990 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
991 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
992 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
993 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
996 @node Empty Structures
997 @section Structures With No Members
998 @cindex empty structures
999 @cindex zero-size structures
1001 GCC permits a C structure to have no members:
1008 The structure will have size zero. In C++, empty structures are part
1009 of the language. G++ treats empty structures as if they had a single
1010 member of type @code{char}.
1012 @node Variable Length
1013 @section Arrays of Variable Length
1014 @cindex variable-length arrays
1015 @cindex arrays of variable length
1018 Variable-length automatic arrays are allowed in ISO C99, and as an
1019 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1020 implementation of variable-length arrays does not yet conform in detail
1021 to the ISO C99 standard.) These arrays are
1022 declared like any other automatic arrays, but with a length that is not
1023 a constant expression. The storage is allocated at the point of
1024 declaration and deallocated when the brace-level is exited. For
1029 concat_fopen (char *s1, char *s2, char *mode)
1031 char str[strlen (s1) + strlen (s2) + 1];
1034 return fopen (str, mode);
1038 @cindex scope of a variable length array
1039 @cindex variable-length array scope
1040 @cindex deallocating variable length arrays
1041 Jumping or breaking out of the scope of the array name deallocates the
1042 storage. Jumping into the scope is not allowed; you get an error
1045 @cindex @code{alloca} vs variable-length arrays
1046 You can use the function @code{alloca} to get an effect much like
1047 variable-length arrays. The function @code{alloca} is available in
1048 many other C implementations (but not in all). On the other hand,
1049 variable-length arrays are more elegant.
1051 There are other differences between these two methods. Space allocated
1052 with @code{alloca} exists until the containing @emph{function} returns.
1053 The space for a variable-length array is deallocated as soon as the array
1054 name's scope ends. (If you use both variable-length arrays and
1055 @code{alloca} in the same function, deallocation of a variable-length array
1056 will also deallocate anything more recently allocated with @code{alloca}.)
1058 You can also use variable-length arrays as arguments to functions:
1062 tester (int len, char data[len][len])
1068 The length of an array is computed once when the storage is allocated
1069 and is remembered for the scope of the array in case you access it with
1072 If you want to pass the array first and the length afterward, you can
1073 use a forward declaration in the parameter list---another GNU extension.
1077 tester (int len; char data[len][len], int len)
1083 @cindex parameter forward declaration
1084 The @samp{int len} before the semicolon is a @dfn{parameter forward
1085 declaration}, and it serves the purpose of making the name @code{len}
1086 known when the declaration of @code{data} is parsed.
1088 You can write any number of such parameter forward declarations in the
1089 parameter list. They can be separated by commas or semicolons, but the
1090 last one must end with a semicolon, which is followed by the ``real''
1091 parameter declarations. Each forward declaration must match a ``real''
1092 declaration in parameter name and data type. ISO C99 does not support
1093 parameter forward declarations.
1095 @node Variadic Macros
1096 @section Macros with a Variable Number of Arguments.
1097 @cindex variable number of arguments
1098 @cindex macro with variable arguments
1099 @cindex rest argument (in macro)
1100 @cindex variadic macros
1102 In the ISO C standard of 1999, a macro can be declared to accept a
1103 variable number of arguments much as a function can. The syntax for
1104 defining the macro is similar to that of a function. Here is an
1108 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1111 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1112 such a macro, it represents the zero or more tokens until the closing
1113 parenthesis that ends the invocation, including any commas. This set of
1114 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1115 wherever it appears. See the CPP manual for more information.
1117 GCC has long supported variadic macros, and used a different syntax that
1118 allowed you to give a name to the variable arguments just like any other
1119 argument. Here is an example:
1122 #define debug(format, args...) fprintf (stderr, format, args)
1125 This is in all ways equivalent to the ISO C example above, but arguably
1126 more readable and descriptive.
1128 GNU CPP has two further variadic macro extensions, and permits them to
1129 be used with either of the above forms of macro definition.
1131 In standard C, you are not allowed to leave the variable argument out
1132 entirely; but you are allowed to pass an empty argument. For example,
1133 this invocation is invalid in ISO C, because there is no comma after
1140 GNU CPP permits you to completely omit the variable arguments in this
1141 way. In the above examples, the compiler would complain, though since
1142 the expansion of the macro still has the extra comma after the format
1145 To help solve this problem, CPP behaves specially for variable arguments
1146 used with the token paste operator, @samp{##}. If instead you write
1149 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1152 and if the variable arguments are omitted or empty, the @samp{##}
1153 operator causes the preprocessor to remove the comma before it. If you
1154 do provide some variable arguments in your macro invocation, GNU CPP
1155 does not complain about the paste operation and instead places the
1156 variable arguments after the comma. Just like any other pasted macro
1157 argument, these arguments are not macro expanded.
1159 @node Escaped Newlines
1160 @section Slightly Looser Rules for Escaped Newlines
1161 @cindex escaped newlines
1162 @cindex newlines (escaped)
1164 Recently, the preprocessor has relaxed its treatment of escaped
1165 newlines. Previously, the newline had to immediately follow a
1166 backslash. The current implementation allows whitespace in the form
1167 of spaces, horizontal and vertical tabs, and form feeds between the
1168 backslash and the subsequent newline. The preprocessor issues a
1169 warning, but treats it as a valid escaped newline and combines the two
1170 lines to form a single logical line. This works within comments and
1171 tokens, as well as between tokens. Comments are @emph{not} treated as
1172 whitespace for the purposes of this relaxation, since they have not
1173 yet been replaced with spaces.
1176 @section Non-Lvalue Arrays May Have Subscripts
1177 @cindex subscripting
1178 @cindex arrays, non-lvalue
1180 @cindex subscripting and function values
1181 In ISO C99, arrays that are not lvalues still decay to pointers, and
1182 may be subscripted, although they may not be modified or used after
1183 the next sequence point and the unary @samp{&} operator may not be
1184 applied to them. As an extension, GCC allows such arrays to be
1185 subscripted in C89 mode, though otherwise they do not decay to
1186 pointers outside C99 mode. For example,
1187 this is valid in GNU C though not valid in C89:
1191 struct foo @{int a[4];@};
1197 return f().a[index];
1203 @section Arithmetic on @code{void}- and Function-Pointers
1204 @cindex void pointers, arithmetic
1205 @cindex void, size of pointer to
1206 @cindex function pointers, arithmetic
1207 @cindex function, size of pointer to
1209 In GNU C, addition and subtraction operations are supported on pointers to
1210 @code{void} and on pointers to functions. This is done by treating the
1211 size of a @code{void} or of a function as 1.
1213 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1214 and on function types, and returns 1.
1216 @opindex Wpointer-arith
1217 The option @option{-Wpointer-arith} requests a warning if these extensions
1221 @section Non-Constant Initializers
1222 @cindex initializers, non-constant
1223 @cindex non-constant initializers
1225 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1226 automatic variable are not required to be constant expressions in GNU C@.
1227 Here is an example of an initializer with run-time varying elements:
1230 foo (float f, float g)
1232 float beat_freqs[2] = @{ f-g, f+g @};
1237 @node Compound Literals
1238 @section Compound Literals
1239 @cindex constructor expressions
1240 @cindex initializations in expressions
1241 @cindex structures, constructor expression
1242 @cindex expressions, constructor
1243 @cindex compound literals
1244 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1246 ISO C99 supports compound literals. A compound literal looks like
1247 a cast containing an initializer. Its value is an object of the
1248 type specified in the cast, containing the elements specified in
1249 the initializer; it is an lvalue. As an extension, GCC supports
1250 compound literals in C89 mode and in C++.
1252 Usually, the specified type is a structure. Assume that
1253 @code{struct foo} and @code{structure} are declared as shown:
1256 struct foo @{int a; char b[2];@} structure;
1260 Here is an example of constructing a @code{struct foo} with a compound literal:
1263 structure = ((struct foo) @{x + y, 'a', 0@});
1267 This is equivalent to writing the following:
1271 struct foo temp = @{x + y, 'a', 0@};
1276 You can also construct an array. If all the elements of the compound literal
1277 are (made up of) simple constant expressions, suitable for use in
1278 initializers of objects of static storage duration, then the compound
1279 literal can be coerced to a pointer to its first element and used in
1280 such an initializer, as shown here:
1283 char **foo = (char *[]) @{ "x", "y", "z" @};
1286 Compound literals for scalar types and union types are is
1287 also allowed, but then the compound literal is equivalent
1290 As a GNU extension, GCC allows initialization of objects with static storage
1291 duration by compound literals (which is not possible in ISO C99, because
1292 the initializer is not a constant).
1293 It is handled as if the object was initialized only with the bracket
1294 enclosed list if the types of the compound literal and the object match.
1295 The initializer list of the compound literal must be constant.
1296 If the object being initialized has array type of unknown size, the size is
1297 determined by compound literal size.
1300 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1301 static int y[] = (int []) @{1, 2, 3@};
1302 static int z[] = (int [3]) @{1@};
1306 The above lines are equivalent to the following:
1308 static struct foo x = @{1, 'a', 'b'@};
1309 static int y[] = @{1, 2, 3@};
1310 static int z[] = @{1, 0, 0@};
1313 @node Designated Inits
1314 @section Designated Initializers
1315 @cindex initializers with labeled elements
1316 @cindex labeled elements in initializers
1317 @cindex case labels in initializers
1318 @cindex designated initializers
1320 Standard C89 requires the elements of an initializer to appear in a fixed
1321 order, the same as the order of the elements in the array or structure
1324 In ISO C99 you can give the elements in any order, specifying the array
1325 indices or structure field names they apply to, and GNU C allows this as
1326 an extension in C89 mode as well. This extension is not
1327 implemented in GNU C++.
1329 To specify an array index, write
1330 @samp{[@var{index}] =} before the element value. For example,
1333 int a[6] = @{ [4] = 29, [2] = 15 @};
1340 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1344 The index values must be constant expressions, even if the array being
1345 initialized is automatic.
1347 An alternative syntax for this which has been obsolete since GCC 2.5 but
1348 GCC still accepts is to write @samp{[@var{index}]} before the element
1349 value, with no @samp{=}.
1351 To initialize a range of elements to the same value, write
1352 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1353 extension. For example,
1356 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1360 If the value in it has side-effects, the side-effects will happen only once,
1361 not for each initialized field by the range initializer.
1364 Note that the length of the array is the highest value specified
1367 In a structure initializer, specify the name of a field to initialize
1368 with @samp{.@var{fieldname} =} before the element value. For example,
1369 given the following structure,
1372 struct point @{ int x, y; @};
1376 the following initialization
1379 struct point p = @{ .y = yvalue, .x = xvalue @};
1386 struct point p = @{ xvalue, yvalue @};
1389 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1390 @samp{@var{fieldname}:}, as shown here:
1393 struct point p = @{ y: yvalue, x: xvalue @};
1397 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1398 @dfn{designator}. You can also use a designator (or the obsolete colon
1399 syntax) when initializing a union, to specify which element of the union
1400 should be used. For example,
1403 union foo @{ int i; double d; @};
1405 union foo f = @{ .d = 4 @};
1409 will convert 4 to a @code{double} to store it in the union using
1410 the second element. By contrast, casting 4 to type @code{union foo}
1411 would store it into the union as the integer @code{i}, since it is
1412 an integer. (@xref{Cast to Union}.)
1414 You can combine this technique of naming elements with ordinary C
1415 initialization of successive elements. Each initializer element that
1416 does not have a designator applies to the next consecutive element of the
1417 array or structure. For example,
1420 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1427 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1430 Labeling the elements of an array initializer is especially useful
1431 when the indices are characters or belong to an @code{enum} type.
1436 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1437 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1440 @cindex designator lists
1441 You can also write a series of @samp{.@var{fieldname}} and
1442 @samp{[@var{index}]} designators before an @samp{=} to specify a
1443 nested subobject to initialize; the list is taken relative to the
1444 subobject corresponding to the closest surrounding brace pair. For
1445 example, with the @samp{struct point} declaration above:
1448 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1452 If the same field is initialized multiple times, it will have value from
1453 the last initialization. If any such overridden initialization has
1454 side-effect, it is unspecified whether the side-effect happens or not.
1455 Currently, GCC will discard them and issue a warning.
1458 @section Case Ranges
1460 @cindex ranges in case statements
1462 You can specify a range of consecutive values in a single @code{case} label,
1466 case @var{low} ... @var{high}:
1470 This has the same effect as the proper number of individual @code{case}
1471 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1473 This feature is especially useful for ranges of ASCII character codes:
1479 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1480 it may be parsed wrong when you use it with integer values. For example,
1495 @section Cast to a Union Type
1496 @cindex cast to a union
1497 @cindex union, casting to a
1499 A cast to union type is similar to other casts, except that the type
1500 specified is a union type. You can specify the type either with
1501 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1502 a constructor though, not a cast, and hence does not yield an lvalue like
1503 normal casts. (@xref{Compound Literals}.)
1505 The types that may be cast to the union type are those of the members
1506 of the union. Thus, given the following union and variables:
1509 union foo @{ int i; double d; @};
1515 both @code{x} and @code{y} can be cast to type @code{union foo}.
1517 Using the cast as the right-hand side of an assignment to a variable of
1518 union type is equivalent to storing in a member of the union:
1523 u = (union foo) x @equiv{} u.i = x
1524 u = (union foo) y @equiv{} u.d = y
1527 You can also use the union cast as a function argument:
1530 void hack (union foo);
1532 hack ((union foo) x);
1535 @node Mixed Declarations
1536 @section Mixed Declarations and Code
1537 @cindex mixed declarations and code
1538 @cindex declarations, mixed with code
1539 @cindex code, mixed with declarations
1541 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1542 within compound statements. As an extension, GCC also allows this in
1543 C89 mode. For example, you could do:
1552 Each identifier is visible from where it is declared until the end of
1553 the enclosing block.
1555 @node Function Attributes
1556 @section Declaring Attributes of Functions
1557 @cindex function attributes
1558 @cindex declaring attributes of functions
1559 @cindex functions that never return
1560 @cindex functions that return more than once
1561 @cindex functions that have no side effects
1562 @cindex functions in arbitrary sections
1563 @cindex functions that behave like malloc
1564 @cindex @code{volatile} applied to function
1565 @cindex @code{const} applied to function
1566 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1567 @cindex functions with non-null pointer arguments
1568 @cindex functions that are passed arguments in registers on the 386
1569 @cindex functions that pop the argument stack on the 386
1570 @cindex functions that do not pop the argument stack on the 386
1572 In GNU C, you declare certain things about functions called in your program
1573 which help the compiler optimize function calls and check your code more
1576 The keyword @code{__attribute__} allows you to specify special
1577 attributes when making a declaration. This keyword is followed by an
1578 attribute specification inside double parentheses. The following
1579 attributes are currently defined for functions on all targets:
1580 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1581 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1582 @code{format}, @code{format_arg}, @code{no_instrument_function},
1583 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1584 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1585 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1586 @code{gnu_inline} and @code{externally_visible}. Several other
1587 attributes are defined for functions on particular target systems. Other
1588 attributes, including @code{section} are supported for variables declarations
1589 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1591 You may also specify attributes with @samp{__} preceding and following
1592 each keyword. This allows you to use them in header files without
1593 being concerned about a possible macro of the same name. For example,
1594 you may use @code{__noreturn__} instead of @code{noreturn}.
1596 @xref{Attribute Syntax}, for details of the exact syntax for using
1600 @c Keep this table alphabetized by attribute name. Treat _ as space.
1602 @item alias ("@var{target}")
1603 @cindex @code{alias} attribute
1604 The @code{alias} attribute causes the declaration to be emitted as an
1605 alias for another symbol, which must be specified. For instance,
1608 void __f () @{ /* @r{Do something.} */; @}
1609 void f () __attribute__ ((weak, alias ("__f")));
1612 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1613 mangled name for the target must be used. It is an error if @samp{__f}
1614 is not defined in the same translation unit.
1616 Not all target machines support this attribute.
1619 @cindex @code{always_inline} function attribute
1620 Generally, functions are not inlined unless optimization is specified.
1621 For functions declared inline, this attribute inlines the function even
1622 if no optimization level was specified.
1625 @cindex @code{gnu_inline} function attribute
1626 This attribute should be used with a function which is also declared
1627 with the @code{inline} keyword. It directs GCC to treat the function
1628 as if it were defined in gnu89 mode even when compiling in C99 or
1631 If the function is declared @code{extern}, then this definition of the
1632 function is used only for inlining. In no case is the function
1633 compiled as a standalone function, not even if you take its address
1634 explicitly. Such an address becomes an external reference, as if you
1635 had only declared the function, and had not defined it. This has
1636 almost the effect of a macro. The way to use this is to put a
1637 function definition in a header file with this attribute, and put
1638 another copy of the function, without @code{extern}, in a library
1639 file. The definition in the header file will cause most calls to the
1640 function to be inlined. If any uses of the function remain, they will
1641 refer to the single copy in the library. Note that the two
1642 definitions of the functions need not be precisely the same, although
1643 if they do not have the same effect your program may behave oddly.
1645 If the function is neither @code{extern} nor @code{static}, then the
1646 function is compiled as a standalone function, as well as being
1647 inlined where possible.
1649 This is how GCC traditionally handled functions declared
1650 @code{inline}. Since ISO C99 specifies a different semantics for
1651 @code{inline}, this function attribute is provided as a transition
1652 measure and as a useful feature in its own right. This attribute is
1653 available in GCC 4.1.3 and later. It is available if either of the
1654 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1655 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1656 Function is As Fast As a Macro}.
1658 Note that since the first version of GCC to support C99 inline semantics
1659 is 4.3, earlier versions of GCC which accept this attribute effectively
1660 assume that it is always present, whether or not it is given explicitly.
1661 In versions prior to 4.3, the only effect of explicitly including it is
1662 to disable warnings about using inline functions in C99 mode.
1664 @cindex @code{flatten} function attribute
1666 Generally, inlining into a function is limited. For a function marked with
1667 this attribute, every call inside this function will be inlined, if possible.
1668 Whether the function itself is considered for inlining depends on its size and
1669 the current inlining parameters. The @code{flatten} attribute only works
1670 reliably in unit-at-a-time mode.
1673 @cindex functions that do pop the argument stack on the 386
1675 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1676 assume that the calling function will pop off the stack space used to
1677 pass arguments. This is
1678 useful to override the effects of the @option{-mrtd} switch.
1681 @cindex @code{const} function attribute
1682 Many functions do not examine any values except their arguments, and
1683 have no effects except the return value. Basically this is just slightly
1684 more strict class than the @code{pure} attribute below, since function is not
1685 allowed to read global memory.
1687 @cindex pointer arguments
1688 Note that a function that has pointer arguments and examines the data
1689 pointed to must @emph{not} be declared @code{const}. Likewise, a
1690 function that calls a non-@code{const} function usually must not be
1691 @code{const}. It does not make sense for a @code{const} function to
1694 The attribute @code{const} is not implemented in GCC versions earlier
1695 than 2.5. An alternative way to declare that a function has no side
1696 effects, which works in the current version and in some older versions,
1700 typedef int intfn ();
1702 extern const intfn square;
1705 This approach does not work in GNU C++ from 2.6.0 on, since the language
1706 specifies that the @samp{const} must be attached to the return value.
1710 @cindex @code{constructor} function attribute
1711 @cindex @code{destructor} function attribute
1712 The @code{constructor} attribute causes the function to be called
1713 automatically before execution enters @code{main ()}. Similarly, the
1714 @code{destructor} attribute causes the function to be called
1715 automatically after @code{main ()} has completed or @code{exit ()} has
1716 been called. Functions with these attributes are useful for
1717 initializing data that will be used implicitly during the execution of
1721 @cindex @code{deprecated} attribute.
1722 The @code{deprecated} attribute results in a warning if the function
1723 is used anywhere in the source file. This is useful when identifying
1724 functions that are expected to be removed in a future version of a
1725 program. The warning also includes the location of the declaration
1726 of the deprecated function, to enable users to easily find further
1727 information about why the function is deprecated, or what they should
1728 do instead. Note that the warnings only occurs for uses:
1731 int old_fn () __attribute__ ((deprecated));
1733 int (*fn_ptr)() = old_fn;
1736 results in a warning on line 3 but not line 2.
1738 The @code{deprecated} attribute can also be used for variables and
1739 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1742 @cindex @code{__declspec(dllexport)}
1743 On Microsoft Windows targets and Symbian OS targets the
1744 @code{dllexport} attribute causes the compiler to provide a global
1745 pointer to a pointer in a DLL, so that it can be referenced with the
1746 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1747 name is formed by combining @code{_imp__} and the function or variable
1750 You can use @code{__declspec(dllexport)} as a synonym for
1751 @code{__attribute__ ((dllexport))} for compatibility with other
1754 On systems that support the @code{visibility} attribute, this
1755 attribute also implies ``default'' visibility, unless a
1756 @code{visibility} attribute is explicitly specified. You should avoid
1757 the use of @code{dllexport} with ``hidden'' or ``internal''
1758 visibility; in the future GCC may issue an error for those cases.
1760 Currently, the @code{dllexport} attribute is ignored for inlined
1761 functions, unless the @option{-fkeep-inline-functions} flag has been
1762 used. The attribute is also ignored for undefined symbols.
1764 When applied to C++ classes, the attribute marks defined non-inlined
1765 member functions and static data members as exports. Static consts
1766 initialized in-class are not marked unless they are also defined
1769 For Microsoft Windows targets there are alternative methods for
1770 including the symbol in the DLL's export table such as using a
1771 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1772 the @option{--export-all} linker flag.
1775 @cindex @code{__declspec(dllimport)}
1776 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1777 attribute causes the compiler to reference a function or variable via
1778 a global pointer to a pointer that is set up by the DLL exporting the
1779 symbol. The attribute implies @code{extern} storage. On Microsoft
1780 Windows targets, the pointer name is formed by combining @code{_imp__}
1781 and the function or variable name.
1783 You can use @code{__declspec(dllimport)} as a synonym for
1784 @code{__attribute__ ((dllimport))} for compatibility with other
1787 Currently, the attribute is ignored for inlined functions. If the
1788 attribute is applied to a symbol @emph{definition}, an error is reported.
1789 If a symbol previously declared @code{dllimport} is later defined, the
1790 attribute is ignored in subsequent references, and a warning is emitted.
1791 The attribute is also overridden by a subsequent declaration as
1794 When applied to C++ classes, the attribute marks non-inlined
1795 member functions and static data members as imports. However, the
1796 attribute is ignored for virtual methods to allow creation of vtables
1799 On the SH Symbian OS target the @code{dllimport} attribute also has
1800 another affect---it can cause the vtable and run-time type information
1801 for a class to be exported. This happens when the class has a
1802 dllimport'ed constructor or a non-inline, non-pure virtual function
1803 and, for either of those two conditions, the class also has a inline
1804 constructor or destructor and has a key function that is defined in
1805 the current translation unit.
1807 For Microsoft Windows based targets the use of the @code{dllimport}
1808 attribute on functions is not necessary, but provides a small
1809 performance benefit by eliminating a thunk in the DLL@. The use of the
1810 @code{dllimport} attribute on imported variables was required on older
1811 versions of the GNU linker, but can now be avoided by passing the
1812 @option{--enable-auto-import} switch to the GNU linker. As with
1813 functions, using the attribute for a variable eliminates a thunk in
1816 One drawback to using this attribute is that a pointer to a function
1817 or variable marked as @code{dllimport} cannot be used as a constant
1818 address. On Microsoft Windows targets, the attribute can be disabled
1819 for functions by setting the @option{-mnop-fun-dllimport} flag.
1822 @cindex eight bit data on the H8/300, H8/300H, and H8S
1823 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1824 variable should be placed into the eight bit data section.
1825 The compiler will generate more efficient code for certain operations
1826 on data in the eight bit data area. Note the eight bit data area is limited to
1829 You must use GAS and GLD from GNU binutils version 2.7 or later for
1830 this attribute to work correctly.
1832 @item exception_handler
1833 @cindex exception handler functions on the Blackfin processor
1834 Use this attribute on the Blackfin to indicate that the specified function
1835 is an exception handler. The compiler will generate function entry and
1836 exit sequences suitable for use in an exception handler when this
1837 attribute is present.
1840 @cindex functions which handle memory bank switching
1841 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1842 use a calling convention that takes care of switching memory banks when
1843 entering and leaving a function. This calling convention is also the
1844 default when using the @option{-mlong-calls} option.
1846 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1847 to call and return from a function.
1849 On 68HC11 the compiler will generate a sequence of instructions
1850 to invoke a board-specific routine to switch the memory bank and call the
1851 real function. The board-specific routine simulates a @code{call}.
1852 At the end of a function, it will jump to a board-specific routine
1853 instead of using @code{rts}. The board-specific return routine simulates
1857 @cindex functions that pop the argument stack on the 386
1858 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1859 pass the first argument (if of integral type) in the register ECX and
1860 the second argument (if of integral type) in the register EDX@. Subsequent
1861 and other typed arguments are passed on the stack. The called function will
1862 pop the arguments off the stack. If the number of arguments is variable all
1863 arguments are pushed on the stack.
1865 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1866 @cindex @code{format} function attribute
1868 The @code{format} attribute specifies that a function takes @code{printf},
1869 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1870 should be type-checked against a format string. For example, the
1875 my_printf (void *my_object, const char *my_format, ...)
1876 __attribute__ ((format (printf, 2, 3)));
1880 causes the compiler to check the arguments in calls to @code{my_printf}
1881 for consistency with the @code{printf} style format string argument
1884 The parameter @var{archetype} determines how the format string is
1885 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1886 or @code{strfmon}. (You can also use @code{__printf__},
1887 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1888 parameter @var{string-index} specifies which argument is the format
1889 string argument (starting from 1), while @var{first-to-check} is the
1890 number of the first argument to check against the format string. For
1891 functions where the arguments are not available to be checked (such as
1892 @code{vprintf}), specify the third parameter as zero. In this case the
1893 compiler only checks the format string for consistency. For
1894 @code{strftime} formats, the third parameter is required to be zero.
1895 Since non-static C++ methods have an implicit @code{this} argument, the
1896 arguments of such methods should be counted from two, not one, when
1897 giving values for @var{string-index} and @var{first-to-check}.
1899 In the example above, the format string (@code{my_format}) is the second
1900 argument of the function @code{my_print}, and the arguments to check
1901 start with the third argument, so the correct parameters for the format
1902 attribute are 2 and 3.
1904 @opindex ffreestanding
1905 @opindex fno-builtin
1906 The @code{format} attribute allows you to identify your own functions
1907 which take format strings as arguments, so that GCC can check the
1908 calls to these functions for errors. The compiler always (unless
1909 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1910 for the standard library functions @code{printf}, @code{fprintf},
1911 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1912 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1913 warnings are requested (using @option{-Wformat}), so there is no need to
1914 modify the header file @file{stdio.h}. In C99 mode, the functions
1915 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1916 @code{vsscanf} are also checked. Except in strictly conforming C
1917 standard modes, the X/Open function @code{strfmon} is also checked as
1918 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1919 @xref{C Dialect Options,,Options Controlling C Dialect}.
1921 The target may provide additional types of format checks.
1922 @xref{Target Format Checks,,Format Checks Specific to Particular
1925 @item format_arg (@var{string-index})
1926 @cindex @code{format_arg} function attribute
1927 @opindex Wformat-nonliteral
1928 The @code{format_arg} attribute specifies that a function takes a format
1929 string for a @code{printf}, @code{scanf}, @code{strftime} or
1930 @code{strfmon} style function and modifies it (for example, to translate
1931 it into another language), so the result can be passed to a
1932 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1933 function (with the remaining arguments to the format function the same
1934 as they would have been for the unmodified string). For example, the
1939 my_dgettext (char *my_domain, const char *my_format)
1940 __attribute__ ((format_arg (2)));
1944 causes the compiler to check the arguments in calls to a @code{printf},
1945 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1946 format string argument is a call to the @code{my_dgettext} function, for
1947 consistency with the format string argument @code{my_format}. If the
1948 @code{format_arg} attribute had not been specified, all the compiler
1949 could tell in such calls to format functions would be that the format
1950 string argument is not constant; this would generate a warning when
1951 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1952 without the attribute.
1954 The parameter @var{string-index} specifies which argument is the format
1955 string argument (starting from one). Since non-static C++ methods have
1956 an implicit @code{this} argument, the arguments of such methods should
1957 be counted from two.
1959 The @code{format-arg} attribute allows you to identify your own
1960 functions which modify format strings, so that GCC can check the
1961 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1962 type function whose operands are a call to one of your own function.
1963 The compiler always treats @code{gettext}, @code{dgettext}, and
1964 @code{dcgettext} in this manner except when strict ISO C support is
1965 requested by @option{-ansi} or an appropriate @option{-std} option, or
1966 @option{-ffreestanding} or @option{-fno-builtin}
1967 is used. @xref{C Dialect Options,,Options
1968 Controlling C Dialect}.
1970 @item function_vector
1971 @cindex calling functions through the function vector on the H8/300 processors
1972 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1973 function should be called through the function vector. Calling a
1974 function through the function vector will reduce code size, however;
1975 the function vector has a limited size (maximum 128 entries on the H8/300
1976 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1978 You must use GAS and GLD from GNU binutils version 2.7 or later for
1979 this attribute to work correctly.
1982 @cindex interrupt handler functions
1983 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
1984 ports to indicate that the specified function is an interrupt handler.
1985 The compiler will generate function entry and exit sequences suitable
1986 for use in an interrupt handler when this attribute is present.
1988 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1989 SH processors can be specified via the @code{interrupt_handler} attribute.
1991 Note, on the AVR, interrupts will be enabled inside the function.
1993 Note, for the ARM, you can specify the kind of interrupt to be handled by
1994 adding an optional parameter to the interrupt attribute like this:
1997 void f () __attribute__ ((interrupt ("IRQ")));
2000 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2002 @item interrupt_handler
2003 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2004 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2005 indicate that the specified function is an interrupt handler. The compiler
2006 will generate function entry and exit sequences suitable for use in an
2007 interrupt handler when this attribute is present.
2010 @cindex User stack pointer in interrupts on the Blackfin
2011 When used together with @code{interrupt_handler}, @code{exception_handler}
2012 or @code{nmi_handler}, code will be generated to load the stack pointer
2013 from the USP register in the function prologue.
2015 @item long_call/short_call
2016 @cindex indirect calls on ARM
2017 This attribute specifies how a particular function is called on
2018 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2019 command line switch and @code{#pragma long_calls} settings. The
2020 @code{long_call} attribute indicates that the function might be far
2021 away from the call site and require a different (more expensive)
2022 calling sequence. The @code{short_call} attribute always places
2023 the offset to the function from the call site into the @samp{BL}
2024 instruction directly.
2026 @item longcall/shortcall
2027 @cindex functions called via pointer on the RS/6000 and PowerPC
2028 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2029 indicates that the function might be far away from the call site and
2030 require a different (more expensive) calling sequence. The
2031 @code{shortcall} attribute indicates that the function is always close
2032 enough for the shorter calling sequence to be used. These attributes
2033 override both the @option{-mlongcall} switch and, on the RS/6000 and
2034 PowerPC, the @code{#pragma longcall} setting.
2036 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2037 calls are necessary.
2040 @cindex indirect calls on MIPS
2041 This attribute specifies how a particular function is called on MIPS@.
2042 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2043 command line switch. This attribute causes the compiler to always call
2044 the function by first loading its address into a register, and then using
2045 the contents of that register.
2048 @cindex @code{malloc} attribute
2049 The @code{malloc} attribute is used to tell the compiler that a function
2050 may be treated as if any non-@code{NULL} pointer it returns cannot
2051 alias any other pointer valid when the function returns.
2052 This will often improve optimization.
2053 Standard functions with this property include @code{malloc} and
2054 @code{calloc}. @code{realloc}-like functions have this property as
2055 long as the old pointer is never referred to (including comparing it
2056 to the new pointer) after the function returns a non-@code{NULL}
2059 @item model (@var{model-name})
2060 @cindex function addressability on the M32R/D
2061 @cindex variable addressability on the IA-64
2063 On the M32R/D, use this attribute to set the addressability of an
2064 object, and of the code generated for a function. The identifier
2065 @var{model-name} is one of @code{small}, @code{medium}, or
2066 @code{large}, representing each of the code models.
2068 Small model objects live in the lower 16MB of memory (so that their
2069 addresses can be loaded with the @code{ld24} instruction), and are
2070 callable with the @code{bl} instruction.
2072 Medium model objects may live anywhere in the 32-bit address space (the
2073 compiler will generate @code{seth/add3} instructions to load their addresses),
2074 and are callable with the @code{bl} instruction.
2076 Large model objects may live anywhere in the 32-bit address space (the
2077 compiler will generate @code{seth/add3} instructions to load their addresses),
2078 and may not be reachable with the @code{bl} instruction (the compiler will
2079 generate the much slower @code{seth/add3/jl} instruction sequence).
2081 On IA-64, use this attribute to set the addressability of an object.
2082 At present, the only supported identifier for @var{model-name} is
2083 @code{small}, indicating addressability via ``small'' (22-bit)
2084 addresses (so that their addresses can be loaded with the @code{addl}
2085 instruction). Caveat: such addressing is by definition not position
2086 independent and hence this attribute must not be used for objects
2087 defined by shared libraries.
2090 @cindex function without a prologue/epilogue code
2091 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2092 specified function does not need prologue/epilogue sequences generated by
2093 the compiler. It is up to the programmer to provide these sequences.
2096 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2097 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2098 use the normal calling convention based on @code{jsr} and @code{rts}.
2099 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2103 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2104 Use this attribute together with @code{interrupt_handler},
2105 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2106 entry code should enable nested interrupts or exceptions.
2109 @cindex NMI handler functions on the Blackfin processor
2110 Use this attribute on the Blackfin to indicate that the specified function
2111 is an NMI handler. The compiler will generate function entry and
2112 exit sequences suitable for use in an NMI handler when this
2113 attribute is present.
2115 @item no_instrument_function
2116 @cindex @code{no_instrument_function} function attribute
2117 @opindex finstrument-functions
2118 If @option{-finstrument-functions} is given, profiling function calls will
2119 be generated at entry and exit of most user-compiled functions.
2120 Functions with this attribute will not be so instrumented.
2123 @cindex @code{noinline} function attribute
2124 This function attribute prevents a function from being considered for
2127 @item nonnull (@var{arg-index}, @dots{})
2128 @cindex @code{nonnull} function attribute
2129 The @code{nonnull} attribute specifies that some function parameters should
2130 be non-null pointers. For instance, the declaration:
2134 my_memcpy (void *dest, const void *src, size_t len)
2135 __attribute__((nonnull (1, 2)));
2139 causes the compiler to check that, in calls to @code{my_memcpy},
2140 arguments @var{dest} and @var{src} are non-null. If the compiler
2141 determines that a null pointer is passed in an argument slot marked
2142 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2143 is issued. The compiler may also choose to make optimizations based
2144 on the knowledge that certain function arguments will not be null.
2146 If no argument index list is given to the @code{nonnull} attribute,
2147 all pointer arguments are marked as non-null. To illustrate, the
2148 following declaration is equivalent to the previous example:
2152 my_memcpy (void *dest, const void *src, size_t len)
2153 __attribute__((nonnull));
2157 @cindex @code{noreturn} function attribute
2158 A few standard library functions, such as @code{abort} and @code{exit},
2159 cannot return. GCC knows this automatically. Some programs define
2160 their own functions that never return. You can declare them
2161 @code{noreturn} to tell the compiler this fact. For example,
2165 void fatal () __attribute__ ((noreturn));
2168 fatal (/* @r{@dots{}} */)
2170 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2176 The @code{noreturn} keyword tells the compiler to assume that
2177 @code{fatal} cannot return. It can then optimize without regard to what
2178 would happen if @code{fatal} ever did return. This makes slightly
2179 better code. More importantly, it helps avoid spurious warnings of
2180 uninitialized variables.
2182 The @code{noreturn} keyword does not affect the exceptional path when that
2183 applies: a @code{noreturn}-marked function may still return to the caller
2184 by throwing an exception or calling @code{longjmp}.
2186 Do not assume that registers saved by the calling function are
2187 restored before calling the @code{noreturn} function.
2189 It does not make sense for a @code{noreturn} function to have a return
2190 type other than @code{void}.
2192 The attribute @code{noreturn} is not implemented in GCC versions
2193 earlier than 2.5. An alternative way to declare that a function does
2194 not return, which works in the current version and in some older
2195 versions, is as follows:
2198 typedef void voidfn ();
2200 volatile voidfn fatal;
2203 This approach does not work in GNU C++.
2206 @cindex @code{nothrow} function attribute
2207 The @code{nothrow} attribute is used to inform the compiler that a
2208 function cannot throw an exception. For example, most functions in
2209 the standard C library can be guaranteed not to throw an exception
2210 with the notable exceptions of @code{qsort} and @code{bsearch} that
2211 take function pointer arguments. The @code{nothrow} attribute is not
2212 implemented in GCC versions earlier than 3.3.
2215 @cindex @code{pure} function attribute
2216 Many functions have no effects except the return value and their
2217 return value depends only on the parameters and/or global variables.
2218 Such a function can be subject
2219 to common subexpression elimination and loop optimization just as an
2220 arithmetic operator would be. These functions should be declared
2221 with the attribute @code{pure}. For example,
2224 int square (int) __attribute__ ((pure));
2228 says that the hypothetical function @code{square} is safe to call
2229 fewer times than the program says.
2231 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2232 Interesting non-pure functions are functions with infinite loops or those
2233 depending on volatile memory or other system resource, that may change between
2234 two consecutive calls (such as @code{feof} in a multithreading environment).
2236 The attribute @code{pure} is not implemented in GCC versions earlier
2239 @item regparm (@var{number})
2240 @cindex @code{regparm} attribute
2241 @cindex functions that are passed arguments in registers on the 386
2242 On the Intel 386, the @code{regparm} attribute causes the compiler to
2243 pass arguments number one to @var{number} if they are of integral type
2244 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2245 take a variable number of arguments will continue to be passed all of their
2246 arguments on the stack.
2248 Beware that on some ELF systems this attribute is unsuitable for
2249 global functions in shared libraries with lazy binding (which is the
2250 default). Lazy binding will send the first call via resolving code in
2251 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2252 per the standard calling conventions. Solaris 8 is affected by this.
2253 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2254 safe since the loaders there save all registers. (Lazy binding can be
2255 disabled with the linker or the loader if desired, to avoid the
2259 @cindex @code{sseregparm} attribute
2260 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2261 causes the compiler to pass up to 3 floating point arguments in
2262 SSE registers instead of on the stack. Functions that take a
2263 variable number of arguments will continue to pass all of their
2264 floating point arguments on the stack.
2266 @item force_align_arg_pointer
2267 @cindex @code{force_align_arg_pointer} attribute
2268 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2269 applied to individual function definitions, generating an alternate
2270 prologue and epilogue that realigns the runtime stack. This supports
2271 mixing legacy codes that run with a 4-byte aligned stack with modern
2272 codes that keep a 16-byte stack for SSE compatibility. The alternate
2273 prologue and epilogue are slower and bigger than the regular ones, and
2274 the alternate prologue requires a scratch register; this lowers the
2275 number of registers available if used in conjunction with the
2276 @code{regparm} attribute. The @code{force_align_arg_pointer}
2277 attribute is incompatible with nested functions; this is considered a
2281 @cindex @code{returns_twice} attribute
2282 The @code{returns_twice} attribute tells the compiler that a function may
2283 return more than one time. The compiler will ensure that all registers
2284 are dead before calling such a function and will emit a warning about
2285 the variables that may be clobbered after the second return from the
2286 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2287 The @code{longjmp}-like counterpart of such function, if any, might need
2288 to be marked with the @code{noreturn} attribute.
2291 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2292 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2293 all registers except the stack pointer should be saved in the prologue
2294 regardless of whether they are used or not.
2296 @item section ("@var{section-name}")
2297 @cindex @code{section} function attribute
2298 Normally, the compiler places the code it generates in the @code{text} section.
2299 Sometimes, however, you need additional sections, or you need certain
2300 particular functions to appear in special sections. The @code{section}
2301 attribute specifies that a function lives in a particular section.
2302 For example, the declaration:
2305 extern void foobar (void) __attribute__ ((section ("bar")));
2309 puts the function @code{foobar} in the @code{bar} section.
2311 Some file formats do not support arbitrary sections so the @code{section}
2312 attribute is not available on all platforms.
2313 If you need to map the entire contents of a module to a particular
2314 section, consider using the facilities of the linker instead.
2317 @cindex @code{sentinel} function attribute
2318 This function attribute ensures that a parameter in a function call is
2319 an explicit @code{NULL}. The attribute is only valid on variadic
2320 functions. By default, the sentinel is located at position zero, the
2321 last parameter of the function call. If an optional integer position
2322 argument P is supplied to the attribute, the sentinel must be located at
2323 position P counting backwards from the end of the argument list.
2326 __attribute__ ((sentinel))
2328 __attribute__ ((sentinel(0)))
2331 The attribute is automatically set with a position of 0 for the built-in
2332 functions @code{execl} and @code{execlp}. The built-in function
2333 @code{execle} has the attribute set with a position of 1.
2335 A valid @code{NULL} in this context is defined as zero with any pointer
2336 type. If your system defines the @code{NULL} macro with an integer type
2337 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2338 with a copy that redefines NULL appropriately.
2340 The warnings for missing or incorrect sentinels are enabled with
2344 See long_call/short_call.
2347 See longcall/shortcall.
2350 @cindex signal handler functions on the AVR processors
2351 Use this attribute on the AVR to indicate that the specified
2352 function is a signal handler. The compiler will generate function
2353 entry and exit sequences suitable for use in a signal handler when this
2354 attribute is present. Interrupts will be disabled inside the function.
2357 Use this attribute on the SH to indicate an @code{interrupt_handler}
2358 function should switch to an alternate stack. It expects a string
2359 argument that names a global variable holding the address of the
2364 void f () __attribute__ ((interrupt_handler,
2365 sp_switch ("alt_stack")));
2369 @cindex functions that pop the argument stack on the 386
2370 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2371 assume that the called function will pop off the stack space used to
2372 pass arguments, unless it takes a variable number of arguments.
2375 @cindex tiny data section on the H8/300H and H8S
2376 Use this attribute on the H8/300H and H8S to indicate that the specified
2377 variable should be placed into the tiny data section.
2378 The compiler will generate more efficient code for loads and stores
2379 on data in the tiny data section. Note the tiny data area is limited to
2380 slightly under 32kbytes of data.
2383 Use this attribute on the SH for an @code{interrupt_handler} to return using
2384 @code{trapa} instead of @code{rte}. This attribute expects an integer
2385 argument specifying the trap number to be used.
2388 @cindex @code{unused} attribute.
2389 This attribute, attached to a function, means that the function is meant
2390 to be possibly unused. GCC will not produce a warning for this
2394 @cindex @code{used} attribute.
2395 This attribute, attached to a function, means that code must be emitted
2396 for the function even if it appears that the function is not referenced.
2397 This is useful, for example, when the function is referenced only in
2400 @item visibility ("@var{visibility_type}")
2401 @cindex @code{visibility} attribute
2402 This attribute affects the linkage of the declaration to which it is attached.
2403 There are four supported @var{visibility_type} values: default,
2404 hidden, protected or internal visibility.
2407 void __attribute__ ((visibility ("protected")))
2408 f () @{ /* @r{Do something.} */; @}
2409 int i __attribute__ ((visibility ("hidden")));
2412 The possible values of @var{visibility_type} correspond to the
2413 visibility settings in the ELF gABI.
2416 @c keep this list of visibilities in alphabetical order.
2419 Default visibility is the normal case for the object file format.
2420 This value is available for the visibility attribute to override other
2421 options that may change the assumed visibility of entities.
2423 On ELF, default visibility means that the declaration is visible to other
2424 modules and, in shared libraries, means that the declared entity may be
2427 On Darwin, default visibility means that the declaration is visible to
2430 Default visibility corresponds to ``external linkage'' in the language.
2433 Hidden visibility indicates that the entity declared will have a new
2434 form of linkage, which we'll call ``hidden linkage''. Two
2435 declarations of an object with hidden linkage refer to the same object
2436 if they are in the same shared object.
2439 Internal visibility is like hidden visibility, but with additional
2440 processor specific semantics. Unless otherwise specified by the
2441 psABI, GCC defines internal visibility to mean that a function is
2442 @emph{never} called from another module. Compare this with hidden
2443 functions which, while they cannot be referenced directly by other
2444 modules, can be referenced indirectly via function pointers. By
2445 indicating that a function cannot be called from outside the module,
2446 GCC may for instance omit the load of a PIC register since it is known
2447 that the calling function loaded the correct value.
2450 Protected visibility is like default visibility except that it
2451 indicates that references within the defining module will bind to the
2452 definition in that module. That is, the declared entity cannot be
2453 overridden by another module.
2457 All visibilities are supported on many, but not all, ELF targets
2458 (supported when the assembler supports the @samp{.visibility}
2459 pseudo-op). Default visibility is supported everywhere. Hidden
2460 visibility is supported on Darwin targets.
2462 The visibility attribute should be applied only to declarations which
2463 would otherwise have external linkage. The attribute should be applied
2464 consistently, so that the same entity should not be declared with
2465 different settings of the attribute.
2467 In C++, the visibility attribute applies to types as well as functions
2468 and objects, because in C++ types have linkage. A class must not have
2469 greater visibility than its non-static data member types and bases,
2470 and class members default to the visibility of their class. Also, a
2471 declaration without explicit visibility is limited to the visibility
2474 In C++, you can mark member functions and static member variables of a
2475 class with the visibility attribute. This is useful if if you know a
2476 particular method or static member variable should only be used from
2477 one shared object; then you can mark it hidden while the rest of the
2478 class has default visibility. Care must be taken to avoid breaking
2479 the One Definition Rule; for example, it is usually not useful to mark
2480 an inline method as hidden without marking the whole class as hidden.
2482 A C++ namespace declaration can also have the visibility attribute.
2483 This attribute applies only to the particular namespace body, not to
2484 other definitions of the same namespace; it is equivalent to using
2485 @samp{#pragma GCC visibility} before and after the namespace
2486 definition (@pxref{Visibility Pragmas}).
2488 In C++, if a template argument has limited visibility, this
2489 restriction is implicitly propagated to the template instantiation.
2490 Otherwise, template instantiations and specializations default to the
2491 visibility of their template.
2493 If both the template and enclosing class have explicit visibility, the
2494 visibility from the template is used.
2496 @item warn_unused_result
2497 @cindex @code{warn_unused_result} attribute
2498 The @code{warn_unused_result} attribute causes a warning to be emitted
2499 if a caller of the function with this attribute does not use its
2500 return value. This is useful for functions where not checking
2501 the result is either a security problem or always a bug, such as
2505 int fn () __attribute__ ((warn_unused_result));
2508 if (fn () < 0) return -1;
2514 results in warning on line 5.
2517 @cindex @code{weak} attribute
2518 The @code{weak} attribute causes the declaration to be emitted as a weak
2519 symbol rather than a global. This is primarily useful in defining
2520 library functions which can be overridden in user code, though it can
2521 also be used with non-function declarations. Weak symbols are supported
2522 for ELF targets, and also for a.out targets when using the GNU assembler
2526 @itemx weakref ("@var{target}")
2527 @cindex @code{weakref} attribute
2528 The @code{weakref} attribute marks a declaration as a weak reference.
2529 Without arguments, it should be accompanied by an @code{alias} attribute
2530 naming the target symbol. Optionally, the @var{target} may be given as
2531 an argument to @code{weakref} itself. In either case, @code{weakref}
2532 implicitly marks the declaration as @code{weak}. Without a
2533 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2534 @code{weakref} is equivalent to @code{weak}.
2537 static int x() __attribute__ ((weakref ("y")));
2538 /* is equivalent to... */
2539 static int x() __attribute__ ((weak, weakref, alias ("y")));
2541 static int x() __attribute__ ((weakref));
2542 static int x() __attribute__ ((alias ("y")));
2545 A weak reference is an alias that does not by itself require a
2546 definition to be given for the target symbol. If the target symbol is
2547 only referenced through weak references, then the becomes a @code{weak}
2548 undefined symbol. If it is directly referenced, however, then such
2549 strong references prevail, and a definition will be required for the
2550 symbol, not necessarily in the same translation unit.
2552 The effect is equivalent to moving all references to the alias to a
2553 separate translation unit, renaming the alias to the aliased symbol,
2554 declaring it as weak, compiling the two separate translation units and
2555 performing a reloadable link on them.
2557 At present, a declaration to which @code{weakref} is attached can
2558 only be @code{static}.
2560 @item externally_visible
2561 @cindex @code{externally_visible} attribute.
2562 This attribute, attached to a global variable or function nullify
2563 effect of @option{-fwhole-program} command line option, so the object
2564 remain visible outside the current compilation unit
2568 You can specify multiple attributes in a declaration by separating them
2569 by commas within the double parentheses or by immediately following an
2570 attribute declaration with another attribute declaration.
2572 @cindex @code{#pragma}, reason for not using
2573 @cindex pragma, reason for not using
2574 Some people object to the @code{__attribute__} feature, suggesting that
2575 ISO C's @code{#pragma} should be used instead. At the time
2576 @code{__attribute__} was designed, there were two reasons for not doing
2581 It is impossible to generate @code{#pragma} commands from a macro.
2584 There is no telling what the same @code{#pragma} might mean in another
2588 These two reasons applied to almost any application that might have been
2589 proposed for @code{#pragma}. It was basically a mistake to use
2590 @code{#pragma} for @emph{anything}.
2592 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2593 to be generated from macros. In addition, a @code{#pragma GCC}
2594 namespace is now in use for GCC-specific pragmas. However, it has been
2595 found convenient to use @code{__attribute__} to achieve a natural
2596 attachment of attributes to their corresponding declarations, whereas
2597 @code{#pragma GCC} is of use for constructs that do not naturally form
2598 part of the grammar. @xref{Other Directives,,Miscellaneous
2599 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2601 @node Attribute Syntax
2602 @section Attribute Syntax
2603 @cindex attribute syntax
2605 This section describes the syntax with which @code{__attribute__} may be
2606 used, and the constructs to which attribute specifiers bind, for the C
2607 language. Some details may vary for C++. Because of infelicities in
2608 the grammar for attributes, some forms described here may not be
2609 successfully parsed in all cases.
2611 There are some problems with the semantics of attributes in C++. For
2612 example, there are no manglings for attributes, although they may affect
2613 code generation, so problems may arise when attributed types are used in
2614 conjunction with templates or overloading. Similarly, @code{typeid}
2615 does not distinguish between types with different attributes. Support
2616 for attributes in C++ may be restricted in future to attributes on
2617 declarations only, but not on nested declarators.
2619 @xref{Function Attributes}, for details of the semantics of attributes
2620 applying to functions. @xref{Variable Attributes}, for details of the
2621 semantics of attributes applying to variables. @xref{Type Attributes},
2622 for details of the semantics of attributes applying to structure, union
2623 and enumerated types.
2625 An @dfn{attribute specifier} is of the form
2626 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2627 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2628 each attribute is one of the following:
2632 Empty. Empty attributes are ignored.
2635 A word (which may be an identifier such as @code{unused}, or a reserved
2636 word such as @code{const}).
2639 A word, followed by, in parentheses, parameters for the attribute.
2640 These parameters take one of the following forms:
2644 An identifier. For example, @code{mode} attributes use this form.
2647 An identifier followed by a comma and a non-empty comma-separated list
2648 of expressions. For example, @code{format} attributes use this form.
2651 A possibly empty comma-separated list of expressions. For example,
2652 @code{format_arg} attributes use this form with the list being a single
2653 integer constant expression, and @code{alias} attributes use this form
2654 with the list being a single string constant.
2658 An @dfn{attribute specifier list} is a sequence of one or more attribute
2659 specifiers, not separated by any other tokens.
2661 In GNU C, an attribute specifier list may appear after the colon following a
2662 label, other than a @code{case} or @code{default} label. The only
2663 attribute it makes sense to use after a label is @code{unused}. This
2664 feature is intended for code generated by programs which contains labels
2665 that may be unused but which is compiled with @option{-Wall}. It would
2666 not normally be appropriate to use in it human-written code, though it
2667 could be useful in cases where the code that jumps to the label is
2668 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2669 such placement of attribute lists, as it is permissible for a
2670 declaration, which could begin with an attribute list, to be labelled in
2671 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2672 does not arise there.
2674 An attribute specifier list may appear as part of a @code{struct},
2675 @code{union} or @code{enum} specifier. It may go either immediately
2676 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2677 the closing brace. The former syntax is preferred.
2678 Where attribute specifiers follow the closing brace, they are considered
2679 to relate to the structure, union or enumerated type defined, not to any
2680 enclosing declaration the type specifier appears in, and the type
2681 defined is not complete until after the attribute specifiers.
2682 @c Otherwise, there would be the following problems: a shift/reduce
2683 @c conflict between attributes binding the struct/union/enum and
2684 @c binding to the list of specifiers/qualifiers; and "aligned"
2685 @c attributes could use sizeof for the structure, but the size could be
2686 @c changed later by "packed" attributes.
2688 Otherwise, an attribute specifier appears as part of a declaration,
2689 counting declarations of unnamed parameters and type names, and relates
2690 to that declaration (which may be nested in another declaration, for
2691 example in the case of a parameter declaration), or to a particular declarator
2692 within a declaration. Where an
2693 attribute specifier is applied to a parameter declared as a function or
2694 an array, it should apply to the function or array rather than the
2695 pointer to which the parameter is implicitly converted, but this is not
2696 yet correctly implemented.
2698 Any list of specifiers and qualifiers at the start of a declaration may
2699 contain attribute specifiers, whether or not such a list may in that
2700 context contain storage class specifiers. (Some attributes, however,
2701 are essentially in the nature of storage class specifiers, and only make
2702 sense where storage class specifiers may be used; for example,
2703 @code{section}.) There is one necessary limitation to this syntax: the
2704 first old-style parameter declaration in a function definition cannot
2705 begin with an attribute specifier, because such an attribute applies to
2706 the function instead by syntax described below (which, however, is not
2707 yet implemented in this case). In some other cases, attribute
2708 specifiers are permitted by this grammar but not yet supported by the
2709 compiler. All attribute specifiers in this place relate to the
2710 declaration as a whole. In the obsolescent usage where a type of
2711 @code{int} is implied by the absence of type specifiers, such a list of
2712 specifiers and qualifiers may be an attribute specifier list with no
2713 other specifiers or qualifiers.
2715 At present, the first parameter in a function prototype must have some
2716 type specifier which is not an attribute specifier; this resolves an
2717 ambiguity in the interpretation of @code{void f(int
2718 (__attribute__((foo)) x))}, but is subject to change. At present, if
2719 the parentheses of a function declarator contain only attributes then
2720 those attributes are ignored, rather than yielding an error or warning
2721 or implying a single parameter of type int, but this is subject to
2724 An attribute specifier list may appear immediately before a declarator
2725 (other than the first) in a comma-separated list of declarators in a
2726 declaration of more than one identifier using a single list of
2727 specifiers and qualifiers. Such attribute specifiers apply
2728 only to the identifier before whose declarator they appear. For
2732 __attribute__((noreturn)) void d0 (void),
2733 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2738 the @code{noreturn} attribute applies to all the functions
2739 declared; the @code{format} attribute only applies to @code{d1}.
2741 An attribute specifier list may appear immediately before the comma,
2742 @code{=} or semicolon terminating the declaration of an identifier other
2743 than a function definition. At present, such attribute specifiers apply
2744 to the declared object or function, but in future they may attach to the
2745 outermost adjacent declarator. In simple cases there is no difference,
2746 but, for example, in
2749 void (****f)(void) __attribute__((noreturn));
2753 at present the @code{noreturn} attribute applies to @code{f}, which
2754 causes a warning since @code{f} is not a function, but in future it may
2755 apply to the function @code{****f}. The precise semantics of what
2756 attributes in such cases will apply to are not yet specified. Where an
2757 assembler name for an object or function is specified (@pxref{Asm
2758 Labels}), at present the attribute must follow the @code{asm}
2759 specification; in future, attributes before the @code{asm} specification
2760 may apply to the adjacent declarator, and those after it to the declared
2763 An attribute specifier list may, in future, be permitted to appear after
2764 the declarator in a function definition (before any old-style parameter
2765 declarations or the function body).
2767 Attribute specifiers may be mixed with type qualifiers appearing inside
2768 the @code{[]} of a parameter array declarator, in the C99 construct by
2769 which such qualifiers are applied to the pointer to which the array is
2770 implicitly converted. Such attribute specifiers apply to the pointer,
2771 not to the array, but at present this is not implemented and they are
2774 An attribute specifier list may appear at the start of a nested
2775 declarator. At present, there are some limitations in this usage: the
2776 attributes correctly apply to the declarator, but for most individual
2777 attributes the semantics this implies are not implemented.
2778 When attribute specifiers follow the @code{*} of a pointer
2779 declarator, they may be mixed with any type qualifiers present.
2780 The following describes the formal semantics of this syntax. It will make the
2781 most sense if you are familiar with the formal specification of
2782 declarators in the ISO C standard.
2784 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2785 D1}, where @code{T} contains declaration specifiers that specify a type
2786 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2787 contains an identifier @var{ident}. The type specified for @var{ident}
2788 for derived declarators whose type does not include an attribute
2789 specifier is as in the ISO C standard.
2791 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2792 and the declaration @code{T D} specifies the type
2793 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2794 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2795 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2797 If @code{D1} has the form @code{*
2798 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2799 declaration @code{T D} specifies the type
2800 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2801 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2802 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2808 void (__attribute__((noreturn)) ****f) (void);
2812 specifies the type ``pointer to pointer to pointer to pointer to
2813 non-returning function returning @code{void}''. As another example,
2816 char *__attribute__((aligned(8))) *f;
2820 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2821 Note again that this does not work with most attributes; for example,
2822 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2823 is not yet supported.
2825 For compatibility with existing code written for compiler versions that
2826 did not implement attributes on nested declarators, some laxity is
2827 allowed in the placing of attributes. If an attribute that only applies
2828 to types is applied to a declaration, it will be treated as applying to
2829 the type of that declaration. If an attribute that only applies to
2830 declarations is applied to the type of a declaration, it will be treated
2831 as applying to that declaration; and, for compatibility with code
2832 placing the attributes immediately before the identifier declared, such
2833 an attribute applied to a function return type will be treated as
2834 applying to the function type, and such an attribute applied to an array
2835 element type will be treated as applying to the array type. If an
2836 attribute that only applies to function types is applied to a
2837 pointer-to-function type, it will be treated as applying to the pointer
2838 target type; if such an attribute is applied to a function return type
2839 that is not a pointer-to-function type, it will be treated as applying
2840 to the function type.
2842 @node Function Prototypes
2843 @section Prototypes and Old-Style Function Definitions
2844 @cindex function prototype declarations
2845 @cindex old-style function definitions
2846 @cindex promotion of formal parameters
2848 GNU C extends ISO C to allow a function prototype to override a later
2849 old-style non-prototype definition. Consider the following example:
2852 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2859 /* @r{Prototype function declaration.} */
2860 int isroot P((uid_t));
2862 /* @r{Old-style function definition.} */
2864 isroot (x) /* @r{??? lossage here ???} */
2871 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2872 not allow this example, because subword arguments in old-style
2873 non-prototype definitions are promoted. Therefore in this example the
2874 function definition's argument is really an @code{int}, which does not
2875 match the prototype argument type of @code{short}.
2877 This restriction of ISO C makes it hard to write code that is portable
2878 to traditional C compilers, because the programmer does not know
2879 whether the @code{uid_t} type is @code{short}, @code{int}, or
2880 @code{long}. Therefore, in cases like these GNU C allows a prototype
2881 to override a later old-style definition. More precisely, in GNU C, a
2882 function prototype argument type overrides the argument type specified
2883 by a later old-style definition if the former type is the same as the
2884 latter type before promotion. Thus in GNU C the above example is
2885 equivalent to the following:
2898 GNU C++ does not support old-style function definitions, so this
2899 extension is irrelevant.
2902 @section C++ Style Comments
2904 @cindex C++ comments
2905 @cindex comments, C++ style
2907 In GNU C, you may use C++ style comments, which start with @samp{//} and
2908 continue until the end of the line. Many other C implementations allow
2909 such comments, and they are included in the 1999 C standard. However,
2910 C++ style comments are not recognized if you specify an @option{-std}
2911 option specifying a version of ISO C before C99, or @option{-ansi}
2912 (equivalent to @option{-std=c89}).
2915 @section Dollar Signs in Identifier Names
2917 @cindex dollar signs in identifier names
2918 @cindex identifier names, dollar signs in
2920 In GNU C, you may normally use dollar signs in identifier names.
2921 This is because many traditional C implementations allow such identifiers.
2922 However, dollar signs in identifiers are not supported on a few target
2923 machines, typically because the target assembler does not allow them.
2925 @node Character Escapes
2926 @section The Character @key{ESC} in Constants
2928 You can use the sequence @samp{\e} in a string or character constant to
2929 stand for the ASCII character @key{ESC}.
2932 @section Inquiring on Alignment of Types or Variables
2934 @cindex type alignment
2935 @cindex variable alignment
2937 The keyword @code{__alignof__} allows you to inquire about how an object
2938 is aligned, or the minimum alignment usually required by a type. Its
2939 syntax is just like @code{sizeof}.
2941 For example, if the target machine requires a @code{double} value to be
2942 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2943 This is true on many RISC machines. On more traditional machine
2944 designs, @code{__alignof__ (double)} is 4 or even 2.
2946 Some machines never actually require alignment; they allow reference to any
2947 data type even at an odd address. For these machines, @code{__alignof__}
2948 reports the @emph{recommended} alignment of a type.
2950 If the operand of @code{__alignof__} is an lvalue rather than a type,
2951 its value is the required alignment for its type, taking into account
2952 any minimum alignment specified with GCC's @code{__attribute__}
2953 extension (@pxref{Variable Attributes}). For example, after this
2957 struct foo @{ int x; char y; @} foo1;
2961 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2962 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2964 It is an error to ask for the alignment of an incomplete type.
2966 @node Variable Attributes
2967 @section Specifying Attributes of Variables
2968 @cindex attribute of variables
2969 @cindex variable attributes
2971 The keyword @code{__attribute__} allows you to specify special
2972 attributes of variables or structure fields. This keyword is followed
2973 by an attribute specification inside double parentheses. Some
2974 attributes are currently defined generically for variables.
2975 Other attributes are defined for variables on particular target
2976 systems. Other attributes are available for functions
2977 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2978 Other front ends might define more attributes
2979 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2981 You may also specify attributes with @samp{__} preceding and following
2982 each keyword. This allows you to use them in header files without
2983 being concerned about a possible macro of the same name. For example,
2984 you may use @code{__aligned__} instead of @code{aligned}.
2986 @xref{Attribute Syntax}, for details of the exact syntax for using
2990 @cindex @code{aligned} attribute
2991 @item aligned (@var{alignment})
2992 This attribute specifies a minimum alignment for the variable or
2993 structure field, measured in bytes. For example, the declaration:
2996 int x __attribute__ ((aligned (16))) = 0;
3000 causes the compiler to allocate the global variable @code{x} on a
3001 16-byte boundary. On a 68040, this could be used in conjunction with
3002 an @code{asm} expression to access the @code{move16} instruction which
3003 requires 16-byte aligned operands.
3005 You can also specify the alignment of structure fields. For example, to
3006 create a double-word aligned @code{int} pair, you could write:
3009 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3013 This is an alternative to creating a union with a @code{double} member
3014 that forces the union to be double-word aligned.
3016 As in the preceding examples, you can explicitly specify the alignment
3017 (in bytes) that you wish the compiler to use for a given variable or
3018 structure field. Alternatively, you can leave out the alignment factor
3019 and just ask the compiler to align a variable or field to the maximum
3020 useful alignment for the target machine you are compiling for. For
3021 example, you could write:
3024 short array[3] __attribute__ ((aligned));
3027 Whenever you leave out the alignment factor in an @code{aligned} attribute
3028 specification, the compiler automatically sets the alignment for the declared
3029 variable or field to the largest alignment which is ever used for any data
3030 type on the target machine you are compiling for. Doing this can often make
3031 copy operations more efficient, because the compiler can use whatever
3032 instructions copy the biggest chunks of memory when performing copies to
3033 or from the variables or fields that you have aligned this way.
3035 The @code{aligned} attribute can only increase the alignment; but you
3036 can decrease it by specifying @code{packed} as well. See below.
3038 Note that the effectiveness of @code{aligned} attributes may be limited
3039 by inherent limitations in your linker. On many systems, the linker is
3040 only able to arrange for variables to be aligned up to a certain maximum
3041 alignment. (For some linkers, the maximum supported alignment may
3042 be very very small.) If your linker is only able to align variables
3043 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3044 in an @code{__attribute__} will still only provide you with 8 byte
3045 alignment. See your linker documentation for further information.
3047 @item cleanup (@var{cleanup_function})
3048 @cindex @code{cleanup} attribute
3049 The @code{cleanup} attribute runs a function when the variable goes
3050 out of scope. This attribute can only be applied to auto function
3051 scope variables; it may not be applied to parameters or variables
3052 with static storage duration. The function must take one parameter,
3053 a pointer to a type compatible with the variable. The return value
3054 of the function (if any) is ignored.
3056 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3057 will be run during the stack unwinding that happens during the
3058 processing of the exception. Note that the @code{cleanup} attribute
3059 does not allow the exception to be caught, only to perform an action.
3060 It is undefined what happens if @var{cleanup_function} does not
3065 @cindex @code{common} attribute
3066 @cindex @code{nocommon} attribute
3069 The @code{common} attribute requests GCC to place a variable in
3070 ``common'' storage. The @code{nocommon} attribute requests the
3071 opposite---to allocate space for it directly.
3073 These attributes override the default chosen by the
3074 @option{-fno-common} and @option{-fcommon} flags respectively.
3077 @cindex @code{deprecated} attribute
3078 The @code{deprecated} attribute results in a warning if the variable
3079 is used anywhere in the source file. This is useful when identifying
3080 variables that are expected to be removed in a future version of a
3081 program. The warning also includes the location of the declaration
3082 of the deprecated variable, to enable users to easily find further
3083 information about why the variable is deprecated, or what they should
3084 do instead. Note that the warning only occurs for uses:
3087 extern int old_var __attribute__ ((deprecated));
3089 int new_fn () @{ return old_var; @}
3092 results in a warning on line 3 but not line 2.
3094 The @code{deprecated} attribute can also be used for functions and
3095 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3097 @item mode (@var{mode})
3098 @cindex @code{mode} attribute
3099 This attribute specifies the data type for the declaration---whichever
3100 type corresponds to the mode @var{mode}. This in effect lets you
3101 request an integer or floating point type according to its width.
3103 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3104 indicate the mode corresponding to a one-byte integer, @samp{word} or
3105 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3106 or @samp{__pointer__} for the mode used to represent pointers.
3109 @cindex @code{packed} attribute
3110 The @code{packed} attribute specifies that a variable or structure field
3111 should have the smallest possible alignment---one byte for a variable,
3112 and one bit for a field, unless you specify a larger value with the
3113 @code{aligned} attribute.
3115 Here is a structure in which the field @code{x} is packed, so that it
3116 immediately follows @code{a}:
3122 int x[2] __attribute__ ((packed));
3126 @item section ("@var{section-name}")
3127 @cindex @code{section} variable attribute
3128 Normally, the compiler places the objects it generates in sections like
3129 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3130 or you need certain particular variables to appear in special sections,
3131 for example to map to special hardware. The @code{section}
3132 attribute specifies that a variable (or function) lives in a particular
3133 section. For example, this small program uses several specific section names:
3136 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3137 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3138 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3139 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3143 /* @r{Initialize stack pointer} */
3144 init_sp (stack + sizeof (stack));
3146 /* @r{Initialize initialized data} */
3147 memcpy (&init_data, &data, &edata - &data);
3149 /* @r{Turn on the serial ports} */
3156 Use the @code{section} attribute with an @emph{initialized} definition
3157 of a @emph{global} variable, as shown in the example. GCC issues
3158 a warning and otherwise ignores the @code{section} attribute in
3159 uninitialized variable declarations.
3161 You may only use the @code{section} attribute with a fully initialized
3162 global definition because of the way linkers work. The linker requires
3163 each object be defined once, with the exception that uninitialized
3164 variables tentatively go in the @code{common} (or @code{bss}) section
3165 and can be multiply ``defined''. You can force a variable to be
3166 initialized with the @option{-fno-common} flag or the @code{nocommon}
3169 Some file formats do not support arbitrary sections so the @code{section}
3170 attribute is not available on all platforms.
3171 If you need to map the entire contents of a module to a particular
3172 section, consider using the facilities of the linker instead.
3175 @cindex @code{shared} variable attribute
3176 On Microsoft Windows, in addition to putting variable definitions in a named
3177 section, the section can also be shared among all running copies of an
3178 executable or DLL@. For example, this small program defines shared data
3179 by putting it in a named section @code{shared} and marking the section
3183 int foo __attribute__((section ("shared"), shared)) = 0;
3188 /* @r{Read and write foo. All running
3189 copies see the same value.} */
3195 You may only use the @code{shared} attribute along with @code{section}
3196 attribute with a fully initialized global definition because of the way
3197 linkers work. See @code{section} attribute for more information.
3199 The @code{shared} attribute is only available on Microsoft Windows@.
3201 @item tls_model ("@var{tls_model}")
3202 @cindex @code{tls_model} attribute
3203 The @code{tls_model} attribute sets thread-local storage model
3204 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3205 overriding @option{-ftls-model=} command line switch on a per-variable
3207 The @var{tls_model} argument should be one of @code{global-dynamic},
3208 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3210 Not all targets support this attribute.
3213 This attribute, attached to a variable, means that the variable is meant
3214 to be possibly unused. GCC will not produce a warning for this
3218 This attribute, attached to a variable, means that the variable must be
3219 emitted even if it appears that the variable is not referenced.
3221 @item vector_size (@var{bytes})
3222 This attribute specifies the vector size for the variable, measured in
3223 bytes. For example, the declaration:
3226 int foo __attribute__ ((vector_size (16)));
3230 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3231 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3232 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3234 This attribute is only applicable to integral and float scalars,
3235 although arrays, pointers, and function return values are allowed in
3236 conjunction with this construct.
3238 Aggregates with this attribute are invalid, even if they are of the same
3239 size as a corresponding scalar. For example, the declaration:
3242 struct S @{ int a; @};
3243 struct S __attribute__ ((vector_size (16))) foo;
3247 is invalid even if the size of the structure is the same as the size of
3251 The @code{selectany} attribute causes an initialized global variable to
3252 have link-once semantics. When multiple definitions of the variable are
3253 encountered by the linker, the first is selected and the remainder are
3254 discarded. Following usage by the Microsoft compiler, the linker is told
3255 @emph{not} to warn about size or content differences of the multiple
3258 Although the primary usage of this attribute is for POD types, the
3259 attribute can also be applied to global C++ objects that are initialized
3260 by a constructor. In this case, the static initialization and destruction
3261 code for the object is emitted in each translation defining the object,
3262 but the calls to the constructor and destructor are protected by a
3263 link-once guard variable.
3265 The @code{selectany} attribute is only available on Microsoft Windows
3266 targets. You can use @code{__declspec (selectany)} as a synonym for
3267 @code{__attribute__ ((selectany))} for compatibility with other
3271 The @code{weak} attribute is described in @xref{Function Attributes}.
3274 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3277 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3281 @subsection M32R/D Variable Attributes
3283 One attribute is currently defined for the M32R/D@.
3286 @item model (@var{model-name})
3287 @cindex variable addressability on the M32R/D
3288 Use this attribute on the M32R/D to set the addressability of an object.
3289 The identifier @var{model-name} is one of @code{small}, @code{medium},
3290 or @code{large}, representing each of the code models.
3292 Small model objects live in the lower 16MB of memory (so that their
3293 addresses can be loaded with the @code{ld24} instruction).
3295 Medium and large model objects may live anywhere in the 32-bit address space
3296 (the compiler will generate @code{seth/add3} instructions to load their
3300 @anchor{i386 Variable Attributes}
3301 @subsection i386 Variable Attributes
3303 Two attributes are currently defined for i386 configurations:
3304 @code{ms_struct} and @code{gcc_struct}
3309 @cindex @code{ms_struct} attribute
3310 @cindex @code{gcc_struct} attribute
3312 If @code{packed} is used on a structure, or if bit-fields are used
3313 it may be that the Microsoft ABI packs them differently
3314 than GCC would normally pack them. Particularly when moving packed
3315 data between functions compiled with GCC and the native Microsoft compiler
3316 (either via function call or as data in a file), it may be necessary to access
3319 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3320 compilers to match the native Microsoft compiler.
3322 The Microsoft structure layout algorithm is fairly simple with the exception
3323 of the bitfield packing:
3325 The padding and alignment of members of structures and whether a bit field
3326 can straddle a storage-unit boundary
3329 @item Structure members are stored sequentially in the order in which they are
3330 declared: the first member has the lowest memory address and the last member
3333 @item Every data object has an alignment-requirement. The alignment-requirement
3334 for all data except structures, unions, and arrays is either the size of the
3335 object or the current packing size (specified with either the aligned attribute
3336 or the pack pragma), whichever is less. For structures, unions, and arrays,
3337 the alignment-requirement is the largest alignment-requirement of its members.
3338 Every object is allocated an offset so that:
3340 offset % alignment-requirement == 0
3342 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3343 unit if the integral types are the same size and if the next bit field fits
3344 into the current allocation unit without crossing the boundary imposed by the
3345 common alignment requirements of the bit fields.
3348 Handling of zero-length bitfields:
3350 MSVC interprets zero-length bitfields in the following ways:
3353 @item If a zero-length bitfield is inserted between two bitfields that would
3354 normally be coalesced, the bitfields will not be coalesced.
3361 unsigned long bf_1 : 12;
3363 unsigned long bf_2 : 12;
3367 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3368 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3370 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3371 alignment of the zero-length bitfield is greater than the member that follows it,
3372 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3392 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3393 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3394 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3397 Taking this into account, it is important to note the following:
3400 @item If a zero-length bitfield follows a normal bitfield, the type of the
3401 zero-length bitfield may affect the alignment of the structure as whole. For
3402 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3403 normal bitfield, and is of type short.
3405 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3406 still affect the alignment of the structure:
3416 Here, @code{t4} will take up 4 bytes.
3419 @item Zero-length bitfields following non-bitfield members are ignored:
3430 Here, @code{t5} will take up 2 bytes.
3434 @subsection PowerPC Variable Attributes
3436 Three attributes currently are defined for PowerPC configurations:
3437 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3439 For full documentation of the struct attributes please see the
3440 documentation in the @xref{i386 Variable Attributes}, section.
3442 For documentation of @code{altivec} attribute please see the
3443 documentation in the @xref{PowerPC Type Attributes}, section.
3445 @subsection Xstormy16 Variable Attributes
3447 One attribute is currently defined for xstormy16 configurations:
3452 @cindex @code{below100} attribute
3454 If a variable has the @code{below100} attribute (@code{BELOW100} is
3455 allowed also), GCC will place the variable in the first 0x100 bytes of
3456 memory and use special opcodes to access it. Such variables will be
3457 placed in either the @code{.bss_below100} section or the
3458 @code{.data_below100} section.
3462 @node Type Attributes
3463 @section Specifying Attributes of Types
3464 @cindex attribute of types
3465 @cindex type attributes
3467 The keyword @code{__attribute__} allows you to specify special
3468 attributes of @code{struct} and @code{union} types when you define
3469 such types. This keyword is followed by an attribute specification
3470 inside double parentheses. Seven attributes are currently defined for
3471 types: @code{aligned}, @code{packed}, @code{transparent_union},
3472 @code{unused}, @code{deprecated}, @code{visibility}, and
3473 @code{may_alias}. Other attributes are defined for functions
3474 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3477 You may also specify any one of these attributes with @samp{__}
3478 preceding and following its keyword. This allows you to use these
3479 attributes in header files without being concerned about a possible
3480 macro of the same name. For example, you may use @code{__aligned__}
3481 instead of @code{aligned}.
3483 You may specify type attributes either in a @code{typedef} declaration
3484 or in an enum, struct or union type declaration or definition.
3486 For an enum, struct or union type, you may specify attributes either
3487 between the enum, struct or union tag and the name of the type, or
3488 just past the closing curly brace of the @emph{definition}. The
3489 former syntax is preferred.
3491 @xref{Attribute Syntax}, for details of the exact syntax for using
3495 @cindex @code{aligned} attribute
3496 @item aligned (@var{alignment})
3497 This attribute specifies a minimum alignment (in bytes) for variables
3498 of the specified type. For example, the declarations:
3501 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3502 typedef int more_aligned_int __attribute__ ((aligned (8)));
3506 force the compiler to insure (as far as it can) that each variable whose
3507 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3508 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3509 variables of type @code{struct S} aligned to 8-byte boundaries allows
3510 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3511 store) instructions when copying one variable of type @code{struct S} to
3512 another, thus improving run-time efficiency.
3514 Note that the alignment of any given @code{struct} or @code{union} type
3515 is required by the ISO C standard to be at least a perfect multiple of
3516 the lowest common multiple of the alignments of all of the members of
3517 the @code{struct} or @code{union} in question. This means that you @emph{can}
3518 effectively adjust the alignment of a @code{struct} or @code{union}
3519 type by attaching an @code{aligned} attribute to any one of the members
3520 of such a type, but the notation illustrated in the example above is a
3521 more obvious, intuitive, and readable way to request the compiler to
3522 adjust the alignment of an entire @code{struct} or @code{union} type.
3524 As in the preceding example, you can explicitly specify the alignment
3525 (in bytes) that you wish the compiler to use for a given @code{struct}
3526 or @code{union} type. Alternatively, you can leave out the alignment factor
3527 and just ask the compiler to align a type to the maximum
3528 useful alignment for the target machine you are compiling for. For
3529 example, you could write:
3532 struct S @{ short f[3]; @} __attribute__ ((aligned));
3535 Whenever you leave out the alignment factor in an @code{aligned}
3536 attribute specification, the compiler automatically sets the alignment
3537 for the type to the largest alignment which is ever used for any data
3538 type on the target machine you are compiling for. Doing this can often
3539 make copy operations more efficient, because the compiler can use
3540 whatever instructions copy the biggest chunks of memory when performing
3541 copies to or from the variables which have types that you have aligned
3544 In the example above, if the size of each @code{short} is 2 bytes, then
3545 the size of the entire @code{struct S} type is 6 bytes. The smallest
3546 power of two which is greater than or equal to that is 8, so the
3547 compiler sets the alignment for the entire @code{struct S} type to 8
3550 Note that although you can ask the compiler to select a time-efficient
3551 alignment for a given type and then declare only individual stand-alone
3552 objects of that type, the compiler's ability to select a time-efficient
3553 alignment is primarily useful only when you plan to create arrays of
3554 variables having the relevant (efficiently aligned) type. If you
3555 declare or use arrays of variables of an efficiently-aligned type, then
3556 it is likely that your program will also be doing pointer arithmetic (or
3557 subscripting, which amounts to the same thing) on pointers to the
3558 relevant type, and the code that the compiler generates for these
3559 pointer arithmetic operations will often be more efficient for
3560 efficiently-aligned types than for other types.
3562 The @code{aligned} attribute can only increase the alignment; but you
3563 can decrease it by specifying @code{packed} as well. See below.
3565 Note that the effectiveness of @code{aligned} attributes may be limited
3566 by inherent limitations in your linker. On many systems, the linker is
3567 only able to arrange for variables to be aligned up to a certain maximum
3568 alignment. (For some linkers, the maximum supported alignment may
3569 be very very small.) If your linker is only able to align variables
3570 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3571 in an @code{__attribute__} will still only provide you with 8 byte
3572 alignment. See your linker documentation for further information.
3575 This attribute, attached to @code{struct} or @code{union} type
3576 definition, specifies that each member (other than zero-width bitfields)
3577 of the structure or union is placed to minimize the memory required. When
3578 attached to an @code{enum} definition, it indicates that the smallest
3579 integral type should be used.
3581 @opindex fshort-enums
3582 Specifying this attribute for @code{struct} and @code{union} types is
3583 equivalent to specifying the @code{packed} attribute on each of the
3584 structure or union members. Specifying the @option{-fshort-enums}
3585 flag on the line is equivalent to specifying the @code{packed}
3586 attribute on all @code{enum} definitions.
3588 In the following example @code{struct my_packed_struct}'s members are
3589 packed closely together, but the internal layout of its @code{s} member
3590 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3594 struct my_unpacked_struct
3600 struct __attribute__ ((__packed__)) my_packed_struct
3604 struct my_unpacked_struct s;
3608 You may only specify this attribute on the definition of a @code{enum},
3609 @code{struct} or @code{union}, not on a @code{typedef} which does not
3610 also define the enumerated type, structure or union.
3612 @item transparent_union
3613 This attribute, attached to a @code{union} type definition, indicates
3614 that any function parameter having that union type causes calls to that
3615 function to be treated in a special way.
3617 First, the argument corresponding to a transparent union type can be of
3618 any type in the union; no cast is required. Also, if the union contains
3619 a pointer type, the corresponding argument can be a null pointer
3620 constant or a void pointer expression; and if the union contains a void
3621 pointer type, the corresponding argument can be any pointer expression.
3622 If the union member type is a pointer, qualifiers like @code{const} on
3623 the referenced type must be respected, just as with normal pointer
3626 Second, the argument is passed to the function using the calling
3627 conventions of the first member of the transparent union, not the calling
3628 conventions of the union itself. All members of the union must have the
3629 same machine representation; this is necessary for this argument passing
3632 Transparent unions are designed for library functions that have multiple
3633 interfaces for compatibility reasons. For example, suppose the
3634 @code{wait} function must accept either a value of type @code{int *} to
3635 comply with Posix, or a value of type @code{union wait *} to comply with
3636 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3637 @code{wait} would accept both kinds of arguments, but it would also
3638 accept any other pointer type and this would make argument type checking
3639 less useful. Instead, @code{<sys/wait.h>} might define the interface
3647 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3649 pid_t wait (wait_status_ptr_t);
3652 This interface allows either @code{int *} or @code{union wait *}
3653 arguments to be passed, using the @code{int *} calling convention.
3654 The program can call @code{wait} with arguments of either type:
3657 int w1 () @{ int w; return wait (&w); @}
3658 int w2 () @{ union wait w; return wait (&w); @}
3661 With this interface, @code{wait}'s implementation might look like this:
3664 pid_t wait (wait_status_ptr_t p)
3666 return waitpid (-1, p.__ip, 0);
3671 When attached to a type (including a @code{union} or a @code{struct}),
3672 this attribute means that variables of that type are meant to appear
3673 possibly unused. GCC will not produce a warning for any variables of
3674 that type, even if the variable appears to do nothing. This is often
3675 the case with lock or thread classes, which are usually defined and then
3676 not referenced, but contain constructors and destructors that have
3677 nontrivial bookkeeping functions.
3680 The @code{deprecated} attribute results in a warning if the type
3681 is used anywhere in the source file. This is useful when identifying
3682 types that are expected to be removed in a future version of a program.
3683 If possible, the warning also includes the location of the declaration
3684 of the deprecated type, to enable users to easily find further
3685 information about why the type is deprecated, or what they should do
3686 instead. Note that the warnings only occur for uses and then only
3687 if the type is being applied to an identifier that itself is not being
3688 declared as deprecated.
3691 typedef int T1 __attribute__ ((deprecated));
3695 typedef T1 T3 __attribute__ ((deprecated));
3696 T3 z __attribute__ ((deprecated));
3699 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3700 warning is issued for line 4 because T2 is not explicitly
3701 deprecated. Line 5 has no warning because T3 is explicitly
3702 deprecated. Similarly for line 6.
3704 The @code{deprecated} attribute can also be used for functions and
3705 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3708 Accesses to objects with types with this attribute are not subjected to
3709 type-based alias analysis, but are instead assumed to be able to alias
3710 any other type of objects, just like the @code{char} type. See
3711 @option{-fstrict-aliasing} for more information on aliasing issues.
3716 typedef short __attribute__((__may_alias__)) short_a;
3722 short_a *b = (short_a *) &a;
3726 if (a == 0x12345678)
3733 If you replaced @code{short_a} with @code{short} in the variable
3734 declaration, the above program would abort when compiled with
3735 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3736 above in recent GCC versions.
3739 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3740 applied to class, struct, union and enum types. Unlike other type
3741 attributes, the attribute must appear between the initial keyword and
3742 the name of the type; it cannot appear after the body of the type.
3744 Note that the type visibility is applied to vague linkage entities
3745 associated with the class (vtable, typeinfo node, etc.). In
3746 particular, if a class is thrown as an exception in one shared object
3747 and caught in another, the class must have default visibility.
3748 Otherwise the two shared objects will be unable to use the same
3749 typeinfo node and exception handling will break.
3751 @subsection ARM Type Attributes
3753 On those ARM targets that support @code{dllimport} (such as Symbian
3754 OS), you can use the @code{notshared} attribute to indicate that the
3755 virtual table and other similar data for a class should not be
3756 exported from a DLL@. For example:
3759 class __declspec(notshared) C @{
3761 __declspec(dllimport) C();
3765 __declspec(dllexport)
3769 In this code, @code{C::C} is exported from the current DLL, but the
3770 virtual table for @code{C} is not exported. (You can use
3771 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3772 most Symbian OS code uses @code{__declspec}.)
3774 @anchor{i386 Type Attributes}
3775 @subsection i386 Type Attributes
3777 Two attributes are currently defined for i386 configurations:
3778 @code{ms_struct} and @code{gcc_struct}
3782 @cindex @code{ms_struct}
3783 @cindex @code{gcc_struct}
3785 If @code{packed} is used on a structure, or if bit-fields are used
3786 it may be that the Microsoft ABI packs them differently
3787 than GCC would normally pack them. Particularly when moving packed
3788 data between functions compiled with GCC and the native Microsoft compiler
3789 (either via function call or as data in a file), it may be necessary to access
3792 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3793 compilers to match the native Microsoft compiler.
3796 To specify multiple attributes, separate them by commas within the
3797 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3800 @anchor{PowerPC Type Attributes}
3801 @subsection PowerPC Type Attributes
3803 Three attributes currently are defined for PowerPC configurations:
3804 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3806 For full documentation of the struct attributes please see the
3807 documentation in the @xref{i386 Type Attributes}, section.
3809 The @code{altivec} attribute allows one to declare AltiVec vector data
3810 types supported by the AltiVec Programming Interface Manual. The
3811 attribute requires an argument to specify one of three vector types:
3812 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3813 and @code{bool__} (always followed by unsigned).
3816 __attribute__((altivec(vector__)))
3817 __attribute__((altivec(pixel__))) unsigned short
3818 __attribute__((altivec(bool__))) unsigned
3821 These attributes mainly are intended to support the @code{__vector},
3822 @code{__pixel}, and @code{__bool} AltiVec keywords.
3825 @section An Inline Function is As Fast As a Macro
3826 @cindex inline functions
3827 @cindex integrating function code
3829 @cindex macros, inline alternative
3831 By declaring a function inline, you can direct GCC to make
3832 calls to that function faster. One way GCC can achieve this is to
3833 integrate that function's code into the code for its callers. This
3834 makes execution faster by eliminating the function-call overhead; in
3835 addition, if any of the actual argument values are constant, their
3836 known values may permit simplifications at compile time so that not
3837 all of the inline function's code needs to be included. The effect on
3838 code size is less predictable; object code may be larger or smaller
3839 with function inlining, depending on the particular case. You can
3840 also direct GCC to try to integrate all ``simple enough'' functions
3841 into their callers with the option @option{-finline-functions}.
3843 GCC implements three different semantics of declaring a function
3844 inline. One is available with @option{-std=gnu89}, another when
3845 @option{-std=c99} or @option{-std=gnu99}, and the third is used when
3848 To declare a function inline, use the @code{inline} keyword in its
3849 declaration, like this:
3859 If you are writing a header file to be included in ISO C89 programs, write
3860 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3862 The three types of inlining behave similarly in two important cases:
3863 when the @code{inline} keyword is used on a @code{static} function,
3864 like the example above, and when a function is first declared without
3865 using the @code{inline} keyword and then is defined with
3866 @code{inline}, like this:
3869 extern int inc (int *a);
3877 In both of these common cases, the program behaves the same as if you
3878 had not used the @code{inline} keyword, except for its speed.
3880 @cindex inline functions, omission of
3881 @opindex fkeep-inline-functions
3882 When a function is both inline and @code{static}, if all calls to the
3883 function are integrated into the caller, and the function's address is
3884 never used, then the function's own assembler code is never referenced.
3885 In this case, GCC does not actually output assembler code for the
3886 function, unless you specify the option @option{-fkeep-inline-functions}.
3887 Some calls cannot be integrated for various reasons (in particular,
3888 calls that precede the function's definition cannot be integrated, and
3889 neither can recursive calls within the definition). If there is a
3890 nonintegrated call, then the function is compiled to assembler code as
3891 usual. The function must also be compiled as usual if the program
3892 refers to its address, because that can't be inlined.
3894 @cindex automatic @code{inline} for C++ member fns
3895 @cindex @code{inline} automatic for C++ member fns
3896 @cindex member fns, automatically @code{inline}
3897 @cindex C++ member fns, automatically @code{inline}
3898 @opindex fno-default-inline
3899 As required by ISO C++, GCC considers member functions defined within
3900 the body of a class to be marked inline even if they are
3901 not explicitly declared with the @code{inline} keyword. You can
3902 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
3903 Options,,Options Controlling C++ Dialect}.
3905 GCC does not inline any functions when not optimizing unless you specify
3906 the @samp{always_inline} attribute for the function, like this:
3909 /* @r{Prototype.} */
3910 inline void foo (const char) __attribute__((always_inline));
3913 The remainder of this section is specific to GNU C89 inlining.
3915 @cindex non-static inline function
3916 When an inline function is not @code{static}, then the compiler must assume
3917 that there may be calls from other source files; since a global symbol can
3918 be defined only once in any program, the function must not be defined in
3919 the other source files, so the calls therein cannot be integrated.
3920 Therefore, a non-@code{static} inline function is always compiled on its
3921 own in the usual fashion.
3923 If you specify both @code{inline} and @code{extern} in the function
3924 definition, then the definition is used only for inlining. In no case
3925 is the function compiled on its own, not even if you refer to its
3926 address explicitly. Such an address becomes an external reference, as
3927 if you had only declared the function, and had not defined it.
3929 This combination of @code{inline} and @code{extern} has almost the
3930 effect of a macro. The way to use it is to put a function definition in
3931 a header file with these keywords, and put another copy of the
3932 definition (lacking @code{inline} and @code{extern}) in a library file.
3933 The definition in the header file will cause most calls to the function
3934 to be inlined. If any uses of the function remain, they will refer to
3935 the single copy in the library.
3938 @section Assembler Instructions with C Expression Operands
3939 @cindex extended @code{asm}
3940 @cindex @code{asm} expressions
3941 @cindex assembler instructions
3944 In an assembler instruction using @code{asm}, you can specify the
3945 operands of the instruction using C expressions. This means you need not
3946 guess which registers or memory locations will contain the data you want
3949 You must specify an assembler instruction template much like what
3950 appears in a machine description, plus an operand constraint string for
3953 For example, here is how to use the 68881's @code{fsinx} instruction:
3956 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3960 Here @code{angle} is the C expression for the input operand while
3961 @code{result} is that of the output operand. Each has @samp{"f"} as its
3962 operand constraint, saying that a floating point register is required.
3963 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3964 output operands' constraints must use @samp{=}. The constraints use the
3965 same language used in the machine description (@pxref{Constraints}).
3967 Each operand is described by an operand-constraint string followed by
3968 the C expression in parentheses. A colon separates the assembler
3969 template from the first output operand and another separates the last
3970 output operand from the first input, if any. Commas separate the
3971 operands within each group. The total number of operands is currently
3972 limited to 30; this limitation may be lifted in some future version of
3975 If there are no output operands but there are input operands, you must
3976 place two consecutive colons surrounding the place where the output
3979 As of GCC version 3.1, it is also possible to specify input and output
3980 operands using symbolic names which can be referenced within the
3981 assembler code. These names are specified inside square brackets
3982 preceding the constraint string, and can be referenced inside the
3983 assembler code using @code{%[@var{name}]} instead of a percentage sign
3984 followed by the operand number. Using named operands the above example
3988 asm ("fsinx %[angle],%[output]"
3989 : [output] "=f" (result)
3990 : [angle] "f" (angle));
3994 Note that the symbolic operand names have no relation whatsoever to
3995 other C identifiers. You may use any name you like, even those of
3996 existing C symbols, but you must ensure that no two operands within the same
3997 assembler construct use the same symbolic name.
3999 Output operand expressions must be lvalues; the compiler can check this.
4000 The input operands need not be lvalues. The compiler cannot check
4001 whether the operands have data types that are reasonable for the
4002 instruction being executed. It does not parse the assembler instruction
4003 template and does not know what it means or even whether it is valid
4004 assembler input. The extended @code{asm} feature is most often used for
4005 machine instructions the compiler itself does not know exist. If
4006 the output expression cannot be directly addressed (for example, it is a
4007 bit-field), your constraint must allow a register. In that case, GCC
4008 will use the register as the output of the @code{asm}, and then store
4009 that register into the output.
4011 The ordinary output operands must be write-only; GCC will assume that
4012 the values in these operands before the instruction are dead and need
4013 not be generated. Extended asm supports input-output or read-write
4014 operands. Use the constraint character @samp{+} to indicate such an
4015 operand and list it with the output operands. You should only use
4016 read-write operands when the constraints for the operand (or the
4017 operand in which only some of the bits are to be changed) allow a
4020 You may, as an alternative, logically split its function into two
4021 separate operands, one input operand and one write-only output
4022 operand. The connection between them is expressed by constraints
4023 which say they need to be in the same location when the instruction
4024 executes. You can use the same C expression for both operands, or
4025 different expressions. For example, here we write the (fictitious)
4026 @samp{combine} instruction with @code{bar} as its read-only source
4027 operand and @code{foo} as its read-write destination:
4030 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4034 The constraint @samp{"0"} for operand 1 says that it must occupy the
4035 same location as operand 0. A number in constraint is allowed only in
4036 an input operand and it must refer to an output operand.
4038 Only a number in the constraint can guarantee that one operand will be in
4039 the same place as another. The mere fact that @code{foo} is the value
4040 of both operands is not enough to guarantee that they will be in the
4041 same place in the generated assembler code. The following would not
4045 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4048 Various optimizations or reloading could cause operands 0 and 1 to be in
4049 different registers; GCC knows no reason not to do so. For example, the
4050 compiler might find a copy of the value of @code{foo} in one register and
4051 use it for operand 1, but generate the output operand 0 in a different
4052 register (copying it afterward to @code{foo}'s own address). Of course,
4053 since the register for operand 1 is not even mentioned in the assembler
4054 code, the result will not work, but GCC can't tell that.
4056 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4057 the operand number for a matching constraint. For example:
4060 asm ("cmoveq %1,%2,%[result]"
4061 : [result] "=r"(result)
4062 : "r" (test), "r"(new), "[result]"(old));
4065 Sometimes you need to make an @code{asm} operand be a specific register,
4066 but there's no matching constraint letter for that register @emph{by
4067 itself}. To force the operand into that register, use a local variable
4068 for the operand and specify the register in the variable declaration.
4069 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4070 register constraint letter that matches the register:
4073 register int *p1 asm ("r0") = @dots{};
4074 register int *p2 asm ("r1") = @dots{};
4075 register int *result asm ("r0");
4076 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4079 @anchor{Example of asm with clobbered asm reg}
4080 In the above example, beware that a register that is call-clobbered by
4081 the target ABI will be overwritten by any function call in the
4082 assignment, including library calls for arithmetic operators.
4083 Assuming it is a call-clobbered register, this may happen to @code{r0}
4084 above by the assignment to @code{p2}. If you have to use such a
4085 register, use temporary variables for expressions between the register
4090 register int *p1 asm ("r0") = @dots{};
4091 register int *p2 asm ("r1") = t1;
4092 register int *result asm ("r0");
4093 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4096 Some instructions clobber specific hard registers. To describe this,
4097 write a third colon after the input operands, followed by the names of
4098 the clobbered hard registers (given as strings). Here is a realistic
4099 example for the VAX:
4102 asm volatile ("movc3 %0,%1,%2"
4103 : /* @r{no outputs} */
4104 : "g" (from), "g" (to), "g" (count)
4105 : "r0", "r1", "r2", "r3", "r4", "r5");
4108 You may not write a clobber description in a way that overlaps with an
4109 input or output operand. For example, you may not have an operand
4110 describing a register class with one member if you mention that register
4111 in the clobber list. Variables declared to live in specific registers
4112 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4113 have no part mentioned in the clobber description.
4114 There is no way for you to specify that an input
4115 operand is modified without also specifying it as an output
4116 operand. Note that if all the output operands you specify are for this
4117 purpose (and hence unused), you will then also need to specify
4118 @code{volatile} for the @code{asm} construct, as described below, to
4119 prevent GCC from deleting the @code{asm} statement as unused.
4121 If you refer to a particular hardware register from the assembler code,
4122 you will probably have to list the register after the third colon to
4123 tell the compiler the register's value is modified. In some assemblers,
4124 the register names begin with @samp{%}; to produce one @samp{%} in the
4125 assembler code, you must write @samp{%%} in the input.
4127 If your assembler instruction can alter the condition code register, add
4128 @samp{cc} to the list of clobbered registers. GCC on some machines
4129 represents the condition codes as a specific hardware register;
4130 @samp{cc} serves to name this register. On other machines, the
4131 condition code is handled differently, and specifying @samp{cc} has no
4132 effect. But it is valid no matter what the machine.
4134 If your assembler instructions access memory in an unpredictable
4135 fashion, add @samp{memory} to the list of clobbered registers. This
4136 will cause GCC to not keep memory values cached in registers across the
4137 assembler instruction and not optimize stores or loads to that memory.
4138 You will also want to add the @code{volatile} keyword if the memory
4139 affected is not listed in the inputs or outputs of the @code{asm}, as
4140 the @samp{memory} clobber does not count as a side-effect of the
4141 @code{asm}. If you know how large the accessed memory is, you can add
4142 it as input or output but if this is not known, you should add
4143 @samp{memory}. As an example, if you access ten bytes of a string, you
4144 can use a memory input like:
4147 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4150 Note that in the following example the memory input is necessary,
4151 otherwise GCC might optimize the store to @code{x} away:
4158 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4159 "=&d" (r) : "a" (y), "m" (*y));
4164 You can put multiple assembler instructions together in a single
4165 @code{asm} template, separated by the characters normally used in assembly
4166 code for the system. A combination that works in most places is a newline
4167 to break the line, plus a tab character to move to the instruction field
4168 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4169 assembler allows semicolons as a line-breaking character. Note that some
4170 assembler dialects use semicolons to start a comment.
4171 The input operands are guaranteed not to use any of the clobbered
4172 registers, and neither will the output operands' addresses, so you can
4173 read and write the clobbered registers as many times as you like. Here
4174 is an example of multiple instructions in a template; it assumes the
4175 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4178 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4180 : "g" (from), "g" (to)
4184 Unless an output operand has the @samp{&} constraint modifier, GCC
4185 may allocate it in the same register as an unrelated input operand, on
4186 the assumption the inputs are consumed before the outputs are produced.
4187 This assumption may be false if the assembler code actually consists of
4188 more than one instruction. In such a case, use @samp{&} for each output
4189 operand that may not overlap an input. @xref{Modifiers}.
4191 If you want to test the condition code produced by an assembler
4192 instruction, you must include a branch and a label in the @code{asm}
4193 construct, as follows:
4196 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4202 This assumes your assembler supports local labels, as the GNU assembler
4203 and most Unix assemblers do.
4205 Speaking of labels, jumps from one @code{asm} to another are not
4206 supported. The compiler's optimizers do not know about these jumps, and
4207 therefore they cannot take account of them when deciding how to
4210 @cindex macros containing @code{asm}
4211 Usually the most convenient way to use these @code{asm} instructions is to
4212 encapsulate them in macros that look like functions. For example,
4216 (@{ double __value, __arg = (x); \
4217 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4222 Here the variable @code{__arg} is used to make sure that the instruction
4223 operates on a proper @code{double} value, and to accept only those
4224 arguments @code{x} which can convert automatically to a @code{double}.
4226 Another way to make sure the instruction operates on the correct data
4227 type is to use a cast in the @code{asm}. This is different from using a
4228 variable @code{__arg} in that it converts more different types. For
4229 example, if the desired type were @code{int}, casting the argument to
4230 @code{int} would accept a pointer with no complaint, while assigning the
4231 argument to an @code{int} variable named @code{__arg} would warn about
4232 using a pointer unless the caller explicitly casts it.
4234 If an @code{asm} has output operands, GCC assumes for optimization
4235 purposes the instruction has no side effects except to change the output
4236 operands. This does not mean instructions with a side effect cannot be
4237 used, but you must be careful, because the compiler may eliminate them
4238 if the output operands aren't used, or move them out of loops, or
4239 replace two with one if they constitute a common subexpression. Also,
4240 if your instruction does have a side effect on a variable that otherwise
4241 appears not to change, the old value of the variable may be reused later
4242 if it happens to be found in a register.
4244 You can prevent an @code{asm} instruction from being deleted
4245 by writing the keyword @code{volatile} after
4246 the @code{asm}. For example:
4249 #define get_and_set_priority(new) \
4251 asm volatile ("get_and_set_priority %0, %1" \
4252 : "=g" (__old) : "g" (new)); \
4257 The @code{volatile} keyword indicates that the instruction has
4258 important side-effects. GCC will not delete a volatile @code{asm} if
4259 it is reachable. (The instruction can still be deleted if GCC can
4260 prove that control-flow will never reach the location of the
4261 instruction.) Note that even a volatile @code{asm} instruction
4262 can be moved relative to other code, including across jump
4263 instructions. For example, on many targets there is a system
4264 register which can be set to control the rounding mode of
4265 floating point operations. You might try
4266 setting it with a volatile @code{asm}, like this PowerPC example:
4269 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4274 This will not work reliably, as the compiler may move the addition back
4275 before the volatile @code{asm}. To make it work you need to add an
4276 artificial dependency to the @code{asm} referencing a variable in the code
4277 you don't want moved, for example:
4280 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4284 Similarly, you can't expect a
4285 sequence of volatile @code{asm} instructions to remain perfectly
4286 consecutive. If you want consecutive output, use a single @code{asm}.
4287 Also, GCC will perform some optimizations across a volatile @code{asm}
4288 instruction; GCC does not ``forget everything'' when it encounters
4289 a volatile @code{asm} instruction the way some other compilers do.
4291 An @code{asm} instruction without any output operands will be treated
4292 identically to a volatile @code{asm} instruction.
4294 It is a natural idea to look for a way to give access to the condition
4295 code left by the assembler instruction. However, when we attempted to
4296 implement this, we found no way to make it work reliably. The problem
4297 is that output operands might need reloading, which would result in
4298 additional following ``store'' instructions. On most machines, these
4299 instructions would alter the condition code before there was time to
4300 test it. This problem doesn't arise for ordinary ``test'' and
4301 ``compare'' instructions because they don't have any output operands.
4303 For reasons similar to those described above, it is not possible to give
4304 an assembler instruction access to the condition code left by previous
4307 If you are writing a header file that should be includable in ISO C
4308 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4311 @subsection Size of an @code{asm}
4313 Some targets require that GCC track the size of each instruction used in
4314 order to generate correct code. Because the final length of an
4315 @code{asm} is only known by the assembler, GCC must make an estimate as
4316 to how big it will be. The estimate is formed by counting the number of
4317 statements in the pattern of the @code{asm} and multiplying that by the
4318 length of the longest instruction on that processor. Statements in the
4319 @code{asm} are identified by newline characters and whatever statement
4320 separator characters are supported by the assembler; on most processors
4321 this is the `@code{;}' character.
4323 Normally, GCC's estimate is perfectly adequate to ensure that correct
4324 code is generated, but it is possible to confuse the compiler if you use
4325 pseudo instructions or assembler macros that expand into multiple real
4326 instructions or if you use assembler directives that expand to more
4327 space in the object file than would be needed for a single instruction.
4328 If this happens then the assembler will produce a diagnostic saying that
4329 a label is unreachable.
4331 @subsection i386 floating point asm operands
4333 There are several rules on the usage of stack-like regs in
4334 asm_operands insns. These rules apply only to the operands that are
4339 Given a set of input regs that die in an asm_operands, it is
4340 necessary to know which are implicitly popped by the asm, and
4341 which must be explicitly popped by gcc.
4343 An input reg that is implicitly popped by the asm must be
4344 explicitly clobbered, unless it is constrained to match an
4348 For any input reg that is implicitly popped by an asm, it is
4349 necessary to know how to adjust the stack to compensate for the pop.
4350 If any non-popped input is closer to the top of the reg-stack than
4351 the implicitly popped reg, it would not be possible to know what the
4352 stack looked like---it's not clear how the rest of the stack ``slides
4355 All implicitly popped input regs must be closer to the top of
4356 the reg-stack than any input that is not implicitly popped.
4358 It is possible that if an input dies in an insn, reload might
4359 use the input reg for an output reload. Consider this example:
4362 asm ("foo" : "=t" (a) : "f" (b));
4365 This asm says that input B is not popped by the asm, and that
4366 the asm pushes a result onto the reg-stack, i.e., the stack is one
4367 deeper after the asm than it was before. But, it is possible that
4368 reload will think that it can use the same reg for both the input and
4369 the output, if input B dies in this insn.
4371 If any input operand uses the @code{f} constraint, all output reg
4372 constraints must use the @code{&} earlyclobber.
4374 The asm above would be written as
4377 asm ("foo" : "=&t" (a) : "f" (b));
4381 Some operands need to be in particular places on the stack. All
4382 output operands fall in this category---there is no other way to
4383 know which regs the outputs appear in unless the user indicates
4384 this in the constraints.
4386 Output operands must specifically indicate which reg an output
4387 appears in after an asm. @code{=f} is not allowed: the operand
4388 constraints must select a class with a single reg.
4391 Output operands may not be ``inserted'' between existing stack regs.
4392 Since no 387 opcode uses a read/write operand, all output operands
4393 are dead before the asm_operands, and are pushed by the asm_operands.
4394 It makes no sense to push anywhere but the top of the reg-stack.
4396 Output operands must start at the top of the reg-stack: output
4397 operands may not ``skip'' a reg.
4400 Some asm statements may need extra stack space for internal
4401 calculations. This can be guaranteed by clobbering stack registers
4402 unrelated to the inputs and outputs.
4406 Here are a couple of reasonable asms to want to write. This asm
4407 takes one input, which is internally popped, and produces two outputs.
4410 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4413 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4414 and replaces them with one output. The user must code the @code{st(1)}
4415 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4418 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4424 @section Controlling Names Used in Assembler Code
4425 @cindex assembler names for identifiers
4426 @cindex names used in assembler code
4427 @cindex identifiers, names in assembler code
4429 You can specify the name to be used in the assembler code for a C
4430 function or variable by writing the @code{asm} (or @code{__asm__})
4431 keyword after the declarator as follows:
4434 int foo asm ("myfoo") = 2;
4438 This specifies that the name to be used for the variable @code{foo} in
4439 the assembler code should be @samp{myfoo} rather than the usual
4442 On systems where an underscore is normally prepended to the name of a C
4443 function or variable, this feature allows you to define names for the
4444 linker that do not start with an underscore.
4446 It does not make sense to use this feature with a non-static local
4447 variable since such variables do not have assembler names. If you are
4448 trying to put the variable in a particular register, see @ref{Explicit
4449 Reg Vars}. GCC presently accepts such code with a warning, but will
4450 probably be changed to issue an error, rather than a warning, in the
4453 You cannot use @code{asm} in this way in a function @emph{definition}; but
4454 you can get the same effect by writing a declaration for the function
4455 before its definition and putting @code{asm} there, like this:
4458 extern func () asm ("FUNC");
4465 It is up to you to make sure that the assembler names you choose do not
4466 conflict with any other assembler symbols. Also, you must not use a
4467 register name; that would produce completely invalid assembler code. GCC
4468 does not as yet have the ability to store static variables in registers.
4469 Perhaps that will be added.
4471 @node Explicit Reg Vars
4472 @section Variables in Specified Registers
4473 @cindex explicit register variables
4474 @cindex variables in specified registers
4475 @cindex specified registers
4476 @cindex registers, global allocation
4478 GNU C allows you to put a few global variables into specified hardware
4479 registers. You can also specify the register in which an ordinary
4480 register variable should be allocated.
4484 Global register variables reserve registers throughout the program.
4485 This may be useful in programs such as programming language
4486 interpreters which have a couple of global variables that are accessed
4490 Local register variables in specific registers do not reserve the
4491 registers, except at the point where they are used as input or output
4492 operands in an @code{asm} statement and the @code{asm} statement itself is
4493 not deleted. The compiler's data flow analysis is capable of determining
4494 where the specified registers contain live values, and where they are
4495 available for other uses. Stores into local register variables may be deleted
4496 when they appear to be dead according to dataflow analysis. References
4497 to local register variables may be deleted or moved or simplified.
4499 These local variables are sometimes convenient for use with the extended
4500 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4501 output of the assembler instruction directly into a particular register.
4502 (This will work provided the register you specify fits the constraints
4503 specified for that operand in the @code{asm}.)
4511 @node Global Reg Vars
4512 @subsection Defining Global Register Variables
4513 @cindex global register variables
4514 @cindex registers, global variables in
4516 You can define a global register variable in GNU C like this:
4519 register int *foo asm ("a5");
4523 Here @code{a5} is the name of the register which should be used. Choose a
4524 register which is normally saved and restored by function calls on your
4525 machine, so that library routines will not clobber it.
4527 Naturally the register name is cpu-dependent, so you would need to
4528 conditionalize your program according to cpu type. The register
4529 @code{a5} would be a good choice on a 68000 for a variable of pointer
4530 type. On machines with register windows, be sure to choose a ``global''
4531 register that is not affected magically by the function call mechanism.
4533 In addition, operating systems on one type of cpu may differ in how they
4534 name the registers; then you would need additional conditionals. For
4535 example, some 68000 operating systems call this register @code{%a5}.
4537 Eventually there may be a way of asking the compiler to choose a register
4538 automatically, but first we need to figure out how it should choose and
4539 how to enable you to guide the choice. No solution is evident.
4541 Defining a global register variable in a certain register reserves that
4542 register entirely for this use, at least within the current compilation.
4543 The register will not be allocated for any other purpose in the functions
4544 in the current compilation. The register will not be saved and restored by
4545 these functions. Stores into this register are never deleted even if they
4546 would appear to be dead, but references may be deleted or moved or
4549 It is not safe to access the global register variables from signal
4550 handlers, or from more than one thread of control, because the system
4551 library routines may temporarily use the register for other things (unless
4552 you recompile them specially for the task at hand).
4554 @cindex @code{qsort}, and global register variables
4555 It is not safe for one function that uses a global register variable to
4556 call another such function @code{foo} by way of a third function
4557 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4558 different source file in which the variable wasn't declared). This is
4559 because @code{lose} might save the register and put some other value there.
4560 For example, you can't expect a global register variable to be available in
4561 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4562 might have put something else in that register. (If you are prepared to
4563 recompile @code{qsort} with the same global register variable, you can
4564 solve this problem.)
4566 If you want to recompile @code{qsort} or other source files which do not
4567 actually use your global register variable, so that they will not use that
4568 register for any other purpose, then it suffices to specify the compiler
4569 option @option{-ffixed-@var{reg}}. You need not actually add a global
4570 register declaration to their source code.
4572 A function which can alter the value of a global register variable cannot
4573 safely be called from a function compiled without this variable, because it
4574 could clobber the value the caller expects to find there on return.
4575 Therefore, the function which is the entry point into the part of the
4576 program that uses the global register variable must explicitly save and
4577 restore the value which belongs to its caller.
4579 @cindex register variable after @code{longjmp}
4580 @cindex global register after @code{longjmp}
4581 @cindex value after @code{longjmp}
4584 On most machines, @code{longjmp} will restore to each global register
4585 variable the value it had at the time of the @code{setjmp}. On some
4586 machines, however, @code{longjmp} will not change the value of global
4587 register variables. To be portable, the function that called @code{setjmp}
4588 should make other arrangements to save the values of the global register
4589 variables, and to restore them in a @code{longjmp}. This way, the same
4590 thing will happen regardless of what @code{longjmp} does.
4592 All global register variable declarations must precede all function
4593 definitions. If such a declaration could appear after function
4594 definitions, the declaration would be too late to prevent the register from
4595 being used for other purposes in the preceding functions.
4597 Global register variables may not have initial values, because an
4598 executable file has no means to supply initial contents for a register.
4600 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4601 registers, but certain library functions, such as @code{getwd}, as well
4602 as the subroutines for division and remainder, modify g3 and g4. g1 and
4603 g2 are local temporaries.
4605 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4606 Of course, it will not do to use more than a few of those.
4608 @node Local Reg Vars
4609 @subsection Specifying Registers for Local Variables
4610 @cindex local variables, specifying registers
4611 @cindex specifying registers for local variables
4612 @cindex registers for local variables
4614 You can define a local register variable with a specified register
4618 register int *foo asm ("a5");
4622 Here @code{a5} is the name of the register which should be used. Note
4623 that this is the same syntax used for defining global register
4624 variables, but for a local variable it would appear within a function.
4626 Naturally the register name is cpu-dependent, but this is not a
4627 problem, since specific registers are most often useful with explicit
4628 assembler instructions (@pxref{Extended Asm}). Both of these things
4629 generally require that you conditionalize your program according to
4632 In addition, operating systems on one type of cpu may differ in how they
4633 name the registers; then you would need additional conditionals. For
4634 example, some 68000 operating systems call this register @code{%a5}.
4636 Defining such a register variable does not reserve the register; it
4637 remains available for other uses in places where flow control determines
4638 the variable's value is not live.
4640 This option does not guarantee that GCC will generate code that has
4641 this variable in the register you specify at all times. You may not
4642 code an explicit reference to this register in the @emph{assembler
4643 instruction template} part of an @code{asm} statement and assume it will
4644 always refer to this variable. However, using the variable as an
4645 @code{asm} @emph{operand} guarantees that the specified register is used
4648 Stores into local register variables may be deleted when they appear to be dead
4649 according to dataflow analysis. References to local register variables may
4650 be deleted or moved or simplified.
4652 As for global register variables, it's recommended that you choose a
4653 register which is normally saved and restored by function calls on
4654 your machine, so that library routines will not clobber it. A common
4655 pitfall is to initialize multiple call-clobbered registers with
4656 arbitrary expressions, where a function call or library call for an
4657 arithmetic operator will overwrite a register value from a previous
4658 assignment, for example @code{r0} below:
4660 register int *p1 asm ("r0") = @dots{};
4661 register int *p2 asm ("r1") = @dots{};
4663 In those cases, a solution is to use a temporary variable for
4664 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4666 @node Alternate Keywords
4667 @section Alternate Keywords
4668 @cindex alternate keywords
4669 @cindex keywords, alternate
4671 @option{-ansi} and the various @option{-std} options disable certain
4672 keywords. This causes trouble when you want to use GNU C extensions, or
4673 a general-purpose header file that should be usable by all programs,
4674 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4675 @code{inline} are not available in programs compiled with
4676 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4677 program compiled with @option{-std=c99}). The ISO C99 keyword
4678 @code{restrict} is only available when @option{-std=gnu99} (which will
4679 eventually be the default) or @option{-std=c99} (or the equivalent
4680 @option{-std=iso9899:1999}) is used.
4682 The way to solve these problems is to put @samp{__} at the beginning and
4683 end of each problematical keyword. For example, use @code{__asm__}
4684 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4686 Other C compilers won't accept these alternative keywords; if you want to
4687 compile with another compiler, you can define the alternate keywords as
4688 macros to replace them with the customary keywords. It looks like this:
4696 @findex __extension__
4698 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4700 prevent such warnings within one expression by writing
4701 @code{__extension__} before the expression. @code{__extension__} has no
4702 effect aside from this.
4704 @node Incomplete Enums
4705 @section Incomplete @code{enum} Types
4707 You can define an @code{enum} tag without specifying its possible values.
4708 This results in an incomplete type, much like what you get if you write
4709 @code{struct foo} without describing the elements. A later declaration
4710 which does specify the possible values completes the type.
4712 You can't allocate variables or storage using the type while it is
4713 incomplete. However, you can work with pointers to that type.
4715 This extension may not be very useful, but it makes the handling of
4716 @code{enum} more consistent with the way @code{struct} and @code{union}
4719 This extension is not supported by GNU C++.
4721 @node Function Names
4722 @section Function Names as Strings
4723 @cindex @code{__func__} identifier
4724 @cindex @code{__FUNCTION__} identifier
4725 @cindex @code{__PRETTY_FUNCTION__} identifier
4727 GCC provides three magic variables which hold the name of the current
4728 function, as a string. The first of these is @code{__func__}, which
4729 is part of the C99 standard:
4732 The identifier @code{__func__} is implicitly declared by the translator
4733 as if, immediately following the opening brace of each function
4734 definition, the declaration
4737 static const char __func__[] = "function-name";
4740 appeared, where function-name is the name of the lexically-enclosing
4741 function. This name is the unadorned name of the function.
4744 @code{__FUNCTION__} is another name for @code{__func__}. Older
4745 versions of GCC recognize only this name. However, it is not
4746 standardized. For maximum portability, we recommend you use
4747 @code{__func__}, but provide a fallback definition with the
4751 #if __STDC_VERSION__ < 199901L
4753 # define __func__ __FUNCTION__
4755 # define __func__ "<unknown>"
4760 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4761 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4762 the type signature of the function as well as its bare name. For
4763 example, this program:
4767 extern int printf (char *, ...);
4774 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4775 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4793 __PRETTY_FUNCTION__ = void a::sub(int)
4796 These identifiers are not preprocessor macros. In GCC 3.3 and
4797 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4798 were treated as string literals; they could be used to initialize
4799 @code{char} arrays, and they could be concatenated with other string
4800 literals. GCC 3.4 and later treat them as variables, like
4801 @code{__func__}. In C++, @code{__FUNCTION__} and
4802 @code{__PRETTY_FUNCTION__} have always been variables.
4804 @node Return Address
4805 @section Getting the Return or Frame Address of a Function
4807 These functions may be used to get information about the callers of a
4810 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4811 This function returns the return address of the current function, or of
4812 one of its callers. The @var{level} argument is number of frames to
4813 scan up the call stack. A value of @code{0} yields the return address
4814 of the current function, a value of @code{1} yields the return address
4815 of the caller of the current function, and so forth. When inlining
4816 the expected behavior is that the function will return the address of
4817 the function that will be returned to. To work around this behavior use
4818 the @code{noinline} function attribute.
4820 The @var{level} argument must be a constant integer.
4822 On some machines it may be impossible to determine the return address of
4823 any function other than the current one; in such cases, or when the top
4824 of the stack has been reached, this function will return @code{0} or a
4825 random value. In addition, @code{__builtin_frame_address} may be used
4826 to determine if the top of the stack has been reached.
4828 This function should only be used with a nonzero argument for debugging
4832 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4833 This function is similar to @code{__builtin_return_address}, but it
4834 returns the address of the function frame rather than the return address
4835 of the function. Calling @code{__builtin_frame_address} with a value of
4836 @code{0} yields the frame address of the current function, a value of
4837 @code{1} yields the frame address of the caller of the current function,
4840 The frame is the area on the stack which holds local variables and saved
4841 registers. The frame address is normally the address of the first word
4842 pushed on to the stack by the function. However, the exact definition
4843 depends upon the processor and the calling convention. If the processor
4844 has a dedicated frame pointer register, and the function has a frame,
4845 then @code{__builtin_frame_address} will return the value of the frame
4848 On some machines it may be impossible to determine the frame address of
4849 any function other than the current one; in such cases, or when the top
4850 of the stack has been reached, this function will return @code{0} if
4851 the first frame pointer is properly initialized by the startup code.
4853 This function should only be used with a nonzero argument for debugging
4857 @node Vector Extensions
4858 @section Using vector instructions through built-in functions
4860 On some targets, the instruction set contains SIMD vector instructions that
4861 operate on multiple values contained in one large register at the same time.
4862 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4865 The first step in using these extensions is to provide the necessary data
4866 types. This should be done using an appropriate @code{typedef}:
4869 typedef int v4si __attribute__ ((vector_size (16)));
4872 The @code{int} type specifies the base type, while the attribute specifies
4873 the vector size for the variable, measured in bytes. For example, the
4874 declaration above causes the compiler to set the mode for the @code{v4si}
4875 type to be 16 bytes wide and divided into @code{int} sized units. For
4876 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4877 corresponding mode of @code{foo} will be @acronym{V4SI}.
4879 The @code{vector_size} attribute is only applicable to integral and
4880 float scalars, although arrays, pointers, and function return values
4881 are allowed in conjunction with this construct.
4883 All the basic integer types can be used as base types, both as signed
4884 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4885 @code{long long}. In addition, @code{float} and @code{double} can be
4886 used to build floating-point vector types.
4888 Specifying a combination that is not valid for the current architecture
4889 will cause GCC to synthesize the instructions using a narrower mode.
4890 For example, if you specify a variable of type @code{V4SI} and your
4891 architecture does not allow for this specific SIMD type, GCC will
4892 produce code that uses 4 @code{SIs}.
4894 The types defined in this manner can be used with a subset of normal C
4895 operations. Currently, GCC will allow using the following operators
4896 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4898 The operations behave like C++ @code{valarrays}. Addition is defined as
4899 the addition of the corresponding elements of the operands. For
4900 example, in the code below, each of the 4 elements in @var{a} will be
4901 added to the corresponding 4 elements in @var{b} and the resulting
4902 vector will be stored in @var{c}.
4905 typedef int v4si __attribute__ ((vector_size (16)));
4912 Subtraction, multiplication, division, and the logical operations
4913 operate in a similar manner. Likewise, the result of using the unary
4914 minus or complement operators on a vector type is a vector whose
4915 elements are the negative or complemented values of the corresponding
4916 elements in the operand.
4918 You can declare variables and use them in function calls and returns, as
4919 well as in assignments and some casts. You can specify a vector type as
4920 a return type for a function. Vector types can also be used as function
4921 arguments. It is possible to cast from one vector type to another,
4922 provided they are of the same size (in fact, you can also cast vectors
4923 to and from other datatypes of the same size).
4925 You cannot operate between vectors of different lengths or different
4926 signedness without a cast.
4928 A port that supports hardware vector operations, usually provides a set
4929 of built-in functions that can be used to operate on vectors. For
4930 example, a function to add two vectors and multiply the result by a
4931 third could look like this:
4934 v4si f (v4si a, v4si b, v4si c)
4936 v4si tmp = __builtin_addv4si (a, b);
4937 return __builtin_mulv4si (tmp, c);
4944 @findex __builtin_offsetof
4946 GCC implements for both C and C++ a syntactic extension to implement
4947 the @code{offsetof} macro.
4951 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4953 offsetof_member_designator:
4955 | offsetof_member_designator "." @code{identifier}
4956 | offsetof_member_designator "[" @code{expr} "]"
4959 This extension is sufficient such that
4962 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4965 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4966 may be dependent. In either case, @var{member} may consist of a single
4967 identifier, or a sequence of member accesses and array references.
4969 @node Atomic Builtins
4970 @section Built-in functions for atomic memory access
4972 The following builtins are intended to be compatible with those described
4973 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
4974 section 7.4. As such, they depart from the normal GCC practice of using
4975 the ``__builtin_'' prefix, and further that they are overloaded such that
4976 they work on multiple types.
4978 The definition given in the Intel documentation allows only for the use of
4979 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
4980 counterparts. GCC will allow any integral scalar or pointer type that is
4981 1, 2, 4 or 8 bytes in length.
4983 Not all operations are supported by all target processors. If a particular
4984 operation cannot be implemented on the target processor, a warning will be
4985 generated and a call an external function will be generated. The external
4986 function will carry the same name as the builtin, with an additional suffix
4987 @samp{_@var{n}} where @var{n} is the size of the data type.
4989 @c ??? Should we have a mechanism to suppress this warning? This is almost
4990 @c useful for implementing the operation under the control of an external
4993 In most cases, these builtins are considered a @dfn{full barrier}. That is,
4994 no memory operand will be moved across the operation, either forward or
4995 backward. Further, instructions will be issued as necessary to prevent the
4996 processor from speculating loads across the operation and from queuing stores
4997 after the operation.
4999 All of the routines are are described in the Intel documentation to take
5000 ``an optional list of variables protected by the memory barrier''. It's
5001 not clear what is meant by that; it could mean that @emph{only} the
5002 following variables are protected, or it could mean that these variables
5003 should in addition be protected. At present GCC ignores this list and
5004 protects all variables which are globally accessible. If in the future
5005 we make some use of this list, an empty list will continue to mean all
5006 globally accessible variables.
5009 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5010 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5011 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5012 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5013 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5014 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5015 @findex __sync_fetch_and_add
5016 @findex __sync_fetch_and_sub
5017 @findex __sync_fetch_and_or
5018 @findex __sync_fetch_and_and
5019 @findex __sync_fetch_and_xor
5020 @findex __sync_fetch_and_nand
5021 These builtins perform the operation suggested by the name, and
5022 returns the value that had previously been in memory. That is,
5025 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5026 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5029 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5030 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5031 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5032 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5033 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5034 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5035 @findex __sync_add_and_fetch
5036 @findex __sync_sub_and_fetch
5037 @findex __sync_or_and_fetch
5038 @findex __sync_and_and_fetch
5039 @findex __sync_xor_and_fetch
5040 @findex __sync_nand_and_fetch
5041 These builtins perform the operation suggested by the name, and
5042 return the new value. That is,
5045 @{ *ptr @var{op}= value; return *ptr; @}
5046 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5049 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5050 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5051 @findex __sync_bool_compare_and_swap
5052 @findex __sync_val_compare_and_swap
5053 These builtins perform an atomic compare and swap. That is, if the current
5054 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5057 The ``bool'' version returns true if the comparison is successful and
5058 @var{newval} was written. The ``val'' version returns the contents
5059 of @code{*@var{ptr}} before the operation.
5061 @item __sync_synchronize (...)
5062 @findex __sync_synchronize
5063 This builtin issues a full memory barrier.
5065 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5066 @findex __sync_lock_test_and_set
5067 This builtin, as described by Intel, is not a traditional test-and-set
5068 operation, but rather an atomic exchange operation. It writes @var{value}
5069 into @code{*@var{ptr}}, and returns the previous contents of
5072 Many targets have only minimal support for such locks, and do not support
5073 a full exchange operation. In this case, a target may support reduced
5074 functionality here by which the @emph{only} valid value to store is the
5075 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5076 is implementation defined.
5078 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5079 This means that references after the builtin cannot move to (or be
5080 speculated to) before the builtin, but previous memory stores may not
5081 be globally visible yet, and previous memory loads may not yet be
5084 @item void __sync_lock_release (@var{type} *ptr, ...)
5085 @findex __sync_lock_release
5086 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5087 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5089 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5090 This means that all previous memory stores are globally visible, and all
5091 previous memory loads have been satisfied, but following memory reads
5092 are not prevented from being speculated to before the barrier.
5095 @node Object Size Checking
5096 @section Object Size Checking Builtins
5097 @findex __builtin_object_size
5098 @findex __builtin___memcpy_chk
5099 @findex __builtin___mempcpy_chk
5100 @findex __builtin___memmove_chk
5101 @findex __builtin___memset_chk
5102 @findex __builtin___strcpy_chk
5103 @findex __builtin___stpcpy_chk
5104 @findex __builtin___strncpy_chk
5105 @findex __builtin___strcat_chk
5106 @findex __builtin___strncat_chk
5107 @findex __builtin___sprintf_chk
5108 @findex __builtin___snprintf_chk
5109 @findex __builtin___vsprintf_chk
5110 @findex __builtin___vsnprintf_chk
5111 @findex __builtin___printf_chk
5112 @findex __builtin___vprintf_chk
5113 @findex __builtin___fprintf_chk
5114 @findex __builtin___vfprintf_chk
5116 GCC implements a limited buffer overflow protection mechanism
5117 that can prevent some buffer overflow attacks.
5119 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5120 is a built-in construct that returns a constant number of bytes from
5121 @var{ptr} to the end of the object @var{ptr} pointer points to
5122 (if known at compile time). @code{__builtin_object_size} never evaluates
5123 its arguments for side-effects. If there are any side-effects in them, it
5124 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5125 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5126 point to and all of them are known at compile time, the returned number
5127 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5128 0 and minimum if nonzero. If it is not possible to determine which objects
5129 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5130 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5131 for @var{type} 2 or 3.
5133 @var{type} is an integer constant from 0 to 3. If the least significant
5134 bit is clear, objects are whole variables, if it is set, a closest
5135 surrounding subobject is considered the object a pointer points to.
5136 The second bit determines if maximum or minimum of remaining bytes
5140 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5141 char *p = &var.buf1[1], *q = &var.b;
5143 /* Here the object p points to is var. */
5144 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5145 /* The subobject p points to is var.buf1. */
5146 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5147 /* The object q points to is var. */
5148 assert (__builtin_object_size (q, 0)
5149 == (char *) (&var + 1) - (char *) &var.b);
5150 /* The subobject q points to is var.b. */
5151 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5155 There are built-in functions added for many common string operation
5156 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5157 built-in is provided. This built-in has an additional last argument,
5158 which is the number of bytes remaining in object the @var{dest}
5159 argument points to or @code{(size_t) -1} if the size is not known.
5161 The built-in functions are optimized into the normal string functions
5162 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5163 it is known at compile time that the destination object will not
5164 be overflown. If the compiler can determine at compile time the
5165 object will be always overflown, it issues a warning.
5167 The intended use can be e.g.
5171 #define bos0(dest) __builtin_object_size (dest, 0)
5172 #define memcpy(dest, src, n) \
5173 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5177 /* It is unknown what object p points to, so this is optimized
5178 into plain memcpy - no checking is possible. */
5179 memcpy (p, "abcde", n);
5180 /* Destination is known and length too. It is known at compile
5181 time there will be no overflow. */
5182 memcpy (&buf[5], "abcde", 5);
5183 /* Destination is known, but the length is not known at compile time.
5184 This will result in __memcpy_chk call that can check for overflow
5186 memcpy (&buf[5], "abcde", n);
5187 /* Destination is known and it is known at compile time there will
5188 be overflow. There will be a warning and __memcpy_chk call that
5189 will abort the program at runtime. */
5190 memcpy (&buf[6], "abcde", 5);
5193 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5194 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5195 @code{strcat} and @code{strncat}.
5197 There are also checking built-in functions for formatted output functions.
5199 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5200 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5201 const char *fmt, ...);
5202 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5204 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5205 const char *fmt, va_list ap);
5208 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5209 etc. functions and can contain implementation specific flags on what
5210 additional security measures the checking function might take, such as
5211 handling @code{%n} differently.
5213 The @var{os} argument is the object size @var{s} points to, like in the
5214 other built-in functions. There is a small difference in the behavior
5215 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5216 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5217 the checking function is called with @var{os} argument set to
5220 In addition to this, there are checking built-in functions
5221 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5222 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5223 These have just one additional argument, @var{flag}, right before
5224 format string @var{fmt}. If the compiler is able to optimize them to
5225 @code{fputc} etc. functions, it will, otherwise the checking function
5226 should be called and the @var{flag} argument passed to it.
5228 @node Other Builtins
5229 @section Other built-in functions provided by GCC
5230 @cindex built-in functions
5231 @findex __builtin_isgreater
5232 @findex __builtin_isgreaterequal
5233 @findex __builtin_isless
5234 @findex __builtin_islessequal
5235 @findex __builtin_islessgreater
5236 @findex __builtin_isunordered
5237 @findex __builtin_powi
5238 @findex __builtin_powif
5239 @findex __builtin_powil
5397 @findex fprintf_unlocked
5399 @findex fputs_unlocked
5509 @findex printf_unlocked
5538 @findex significandf
5539 @findex significandl
5610 GCC provides a large number of built-in functions other than the ones
5611 mentioned above. Some of these are for internal use in the processing
5612 of exceptions or variable-length argument lists and will not be
5613 documented here because they may change from time to time; we do not
5614 recommend general use of these functions.
5616 The remaining functions are provided for optimization purposes.
5618 @opindex fno-builtin
5619 GCC includes built-in versions of many of the functions in the standard
5620 C library. The versions prefixed with @code{__builtin_} will always be
5621 treated as having the same meaning as the C library function even if you
5622 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5623 Many of these functions are only optimized in certain cases; if they are
5624 not optimized in a particular case, a call to the library function will
5629 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5630 @option{-std=c99}), the functions
5631 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5632 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5633 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5634 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5635 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5636 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5637 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5638 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5639 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5640 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5641 @code{significandf}, @code{significandl}, @code{significand},
5642 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5643 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5644 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5645 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5646 @code{ynl} and @code{yn}
5647 may be handled as built-in functions.
5648 All these functions have corresponding versions
5649 prefixed with @code{__builtin_}, which may be used even in strict C89
5652 The ISO C99 functions
5653 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5654 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5655 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5656 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5657 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5658 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5659 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5660 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5661 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5662 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5663 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5664 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5665 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5666 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5667 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5668 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5669 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5670 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5671 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5672 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5673 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5674 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5675 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5676 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5677 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5678 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5679 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5680 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5681 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5682 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5683 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5684 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5685 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5686 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5687 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5688 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5689 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5690 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5691 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5692 are handled as built-in functions
5693 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5695 There are also built-in versions of the ISO C99 functions
5696 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5697 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5698 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5699 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5700 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5701 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5702 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5703 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5704 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5705 that are recognized in any mode since ISO C90 reserves these names for
5706 the purpose to which ISO C99 puts them. All these functions have
5707 corresponding versions prefixed with @code{__builtin_}.
5709 The ISO C94 functions
5710 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5711 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5712 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5714 are handled as built-in functions
5715 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5717 The ISO C90 functions
5718 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5719 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5720 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5721 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5722 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5723 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5724 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5725 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5726 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5727 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5728 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5729 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5730 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5731 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5732 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5733 @code{vprintf} and @code{vsprintf}
5734 are all recognized as built-in functions unless
5735 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5736 is specified for an individual function). All of these functions have
5737 corresponding versions prefixed with @code{__builtin_}.
5739 GCC provides built-in versions of the ISO C99 floating point comparison
5740 macros that avoid raising exceptions for unordered operands. They have
5741 the same names as the standard macros ( @code{isgreater},
5742 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5743 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5744 prefixed. We intend for a library implementor to be able to simply
5745 @code{#define} each standard macro to its built-in equivalent.
5747 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5749 You can use the built-in function @code{__builtin_types_compatible_p} to
5750 determine whether two types are the same.
5752 This built-in function returns 1 if the unqualified versions of the
5753 types @var{type1} and @var{type2} (which are types, not expressions) are
5754 compatible, 0 otherwise. The result of this built-in function can be
5755 used in integer constant expressions.
5757 This built-in function ignores top level qualifiers (e.g., @code{const},
5758 @code{volatile}). For example, @code{int} is equivalent to @code{const
5761 The type @code{int[]} and @code{int[5]} are compatible. On the other
5762 hand, @code{int} and @code{char *} are not compatible, even if the size
5763 of their types, on the particular architecture are the same. Also, the
5764 amount of pointer indirection is taken into account when determining
5765 similarity. Consequently, @code{short *} is not similar to
5766 @code{short **}. Furthermore, two types that are typedefed are
5767 considered compatible if their underlying types are compatible.
5769 An @code{enum} type is not considered to be compatible with another
5770 @code{enum} type even if both are compatible with the same integer
5771 type; this is what the C standard specifies.
5772 For example, @code{enum @{foo, bar@}} is not similar to
5773 @code{enum @{hot, dog@}}.
5775 You would typically use this function in code whose execution varies
5776 depending on the arguments' types. For example:
5781 typeof (x) tmp = (x); \
5782 if (__builtin_types_compatible_p (typeof (x), long double)) \
5783 tmp = foo_long_double (tmp); \
5784 else if (__builtin_types_compatible_p (typeof (x), double)) \
5785 tmp = foo_double (tmp); \
5786 else if (__builtin_types_compatible_p (typeof (x), float)) \
5787 tmp = foo_float (tmp); \
5794 @emph{Note:} This construct is only available for C@.
5798 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5800 You can use the built-in function @code{__builtin_choose_expr} to
5801 evaluate code depending on the value of a constant expression. This
5802 built-in function returns @var{exp1} if @var{const_exp}, which is a
5803 constant expression that must be able to be determined at compile time,
5804 is nonzero. Otherwise it returns 0.
5806 This built-in function is analogous to the @samp{? :} operator in C,
5807 except that the expression returned has its type unaltered by promotion
5808 rules. Also, the built-in function does not evaluate the expression
5809 that was not chosen. For example, if @var{const_exp} evaluates to true,
5810 @var{exp2} is not evaluated even if it has side-effects.
5812 This built-in function can return an lvalue if the chosen argument is an
5815 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5816 type. Similarly, if @var{exp2} is returned, its return type is the same
5823 __builtin_choose_expr ( \
5824 __builtin_types_compatible_p (typeof (x), double), \
5826 __builtin_choose_expr ( \
5827 __builtin_types_compatible_p (typeof (x), float), \
5829 /* @r{The void expression results in a compile-time error} \
5830 @r{when assigning the result to something.} */ \
5834 @emph{Note:} This construct is only available for C@. Furthermore, the
5835 unused expression (@var{exp1} or @var{exp2} depending on the value of
5836 @var{const_exp}) may still generate syntax errors. This may change in
5841 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5842 You can use the built-in function @code{__builtin_constant_p} to
5843 determine if a value is known to be constant at compile-time and hence
5844 that GCC can perform constant-folding on expressions involving that
5845 value. The argument of the function is the value to test. The function
5846 returns the integer 1 if the argument is known to be a compile-time
5847 constant and 0 if it is not known to be a compile-time constant. A
5848 return of 0 does not indicate that the value is @emph{not} a constant,
5849 but merely that GCC cannot prove it is a constant with the specified
5850 value of the @option{-O} option.
5852 You would typically use this function in an embedded application where
5853 memory was a critical resource. If you have some complex calculation,
5854 you may want it to be folded if it involves constants, but need to call
5855 a function if it does not. For example:
5858 #define Scale_Value(X) \
5859 (__builtin_constant_p (X) \
5860 ? ((X) * SCALE + OFFSET) : Scale (X))
5863 You may use this built-in function in either a macro or an inline
5864 function. However, if you use it in an inlined function and pass an
5865 argument of the function as the argument to the built-in, GCC will
5866 never return 1 when you call the inline function with a string constant
5867 or compound literal (@pxref{Compound Literals}) and will not return 1
5868 when you pass a constant numeric value to the inline function unless you
5869 specify the @option{-O} option.
5871 You may also use @code{__builtin_constant_p} in initializers for static
5872 data. For instance, you can write
5875 static const int table[] = @{
5876 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5882 This is an acceptable initializer even if @var{EXPRESSION} is not a
5883 constant expression. GCC must be more conservative about evaluating the
5884 built-in in this case, because it has no opportunity to perform
5887 Previous versions of GCC did not accept this built-in in data
5888 initializers. The earliest version where it is completely safe is
5892 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5893 @opindex fprofile-arcs
5894 You may use @code{__builtin_expect} to provide the compiler with
5895 branch prediction information. In general, you should prefer to
5896 use actual profile feedback for this (@option{-fprofile-arcs}), as
5897 programmers are notoriously bad at predicting how their programs
5898 actually perform. However, there are applications in which this
5899 data is hard to collect.
5901 The return value is the value of @var{exp}, which should be an
5902 integral expression. The value of @var{c} must be a compile-time
5903 constant. The semantics of the built-in are that it is expected
5904 that @var{exp} == @var{c}. For example:
5907 if (__builtin_expect (x, 0))
5912 would indicate that we do not expect to call @code{foo}, since
5913 we expect @code{x} to be zero. Since you are limited to integral
5914 expressions for @var{exp}, you should use constructions such as
5917 if (__builtin_expect (ptr != NULL, 1))
5922 when testing pointer or floating-point values.
5925 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5926 This function is used to minimize cache-miss latency by moving data into
5927 a cache before it is accessed.
5928 You can insert calls to @code{__builtin_prefetch} into code for which
5929 you know addresses of data in memory that is likely to be accessed soon.
5930 If the target supports them, data prefetch instructions will be generated.
5931 If the prefetch is done early enough before the access then the data will
5932 be in the cache by the time it is accessed.
5934 The value of @var{addr} is the address of the memory to prefetch.
5935 There are two optional arguments, @var{rw} and @var{locality}.
5936 The value of @var{rw} is a compile-time constant one or zero; one
5937 means that the prefetch is preparing for a write to the memory address
5938 and zero, the default, means that the prefetch is preparing for a read.
5939 The value @var{locality} must be a compile-time constant integer between
5940 zero and three. A value of zero means that the data has no temporal
5941 locality, so it need not be left in the cache after the access. A value
5942 of three means that the data has a high degree of temporal locality and
5943 should be left in all levels of cache possible. Values of one and two
5944 mean, respectively, a low or moderate degree of temporal locality. The
5948 for (i = 0; i < n; i++)
5951 __builtin_prefetch (&a[i+j], 1, 1);
5952 __builtin_prefetch (&b[i+j], 0, 1);
5957 Data prefetch does not generate faults if @var{addr} is invalid, but
5958 the address expression itself must be valid. For example, a prefetch
5959 of @code{p->next} will not fault if @code{p->next} is not a valid
5960 address, but evaluation will fault if @code{p} is not a valid address.
5962 If the target does not support data prefetch, the address expression
5963 is evaluated if it includes side effects but no other code is generated
5964 and GCC does not issue a warning.
5967 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5968 Returns a positive infinity, if supported by the floating-point format,
5969 else @code{DBL_MAX}. This function is suitable for implementing the
5970 ISO C macro @code{HUGE_VAL}.
5973 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5974 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5977 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5978 Similar to @code{__builtin_huge_val}, except the return
5979 type is @code{long double}.
5982 @deftypefn {Built-in Function} double __builtin_inf (void)
5983 Similar to @code{__builtin_huge_val}, except a warning is generated
5984 if the target floating-point format does not support infinities.
5987 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
5988 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
5991 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
5992 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
5995 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
5996 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
5999 @deftypefn {Built-in Function} float __builtin_inff (void)
6000 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6001 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6004 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6005 Similar to @code{__builtin_inf}, except the return
6006 type is @code{long double}.
6009 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6010 This is an implementation of the ISO C99 function @code{nan}.
6012 Since ISO C99 defines this function in terms of @code{strtod}, which we
6013 do not implement, a description of the parsing is in order. The string
6014 is parsed as by @code{strtol}; that is, the base is recognized by
6015 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6016 in the significand such that the least significant bit of the number
6017 is at the least significant bit of the significand. The number is
6018 truncated to fit the significand field provided. The significand is
6019 forced to be a quiet NaN@.
6021 This function, if given a string literal all of which would have been
6022 consumed by strtol, is evaluated early enough that it is considered a
6023 compile-time constant.
6026 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6027 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6030 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6031 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6034 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6035 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6038 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6039 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6042 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6043 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6046 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6047 Similar to @code{__builtin_nan}, except the significand is forced
6048 to be a signaling NaN@. The @code{nans} function is proposed by
6049 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6052 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6053 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6056 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6057 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6060 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6061 Returns one plus the index of the least significant 1-bit of @var{x}, or
6062 if @var{x} is zero, returns zero.
6065 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6066 Returns the number of leading 0-bits in @var{x}, starting at the most
6067 significant bit position. If @var{x} is 0, the result is undefined.
6070 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6071 Returns the number of trailing 0-bits in @var{x}, starting at the least
6072 significant bit position. If @var{x} is 0, the result is undefined.
6075 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6076 Returns the number of 1-bits in @var{x}.
6079 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6080 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6084 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6085 Similar to @code{__builtin_ffs}, except the argument type is
6086 @code{unsigned long}.
6089 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6090 Similar to @code{__builtin_clz}, except the argument type is
6091 @code{unsigned long}.
6094 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6095 Similar to @code{__builtin_ctz}, except the argument type is
6096 @code{unsigned long}.
6099 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6100 Similar to @code{__builtin_popcount}, except the argument type is
6101 @code{unsigned long}.
6104 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6105 Similar to @code{__builtin_parity}, except the argument type is
6106 @code{unsigned long}.
6109 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6110 Similar to @code{__builtin_ffs}, except the argument type is
6111 @code{unsigned long long}.
6114 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6115 Similar to @code{__builtin_clz}, except the argument type is
6116 @code{unsigned long long}.
6119 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6120 Similar to @code{__builtin_ctz}, except the argument type is
6121 @code{unsigned long long}.
6124 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6125 Similar to @code{__builtin_popcount}, except the argument type is
6126 @code{unsigned long long}.
6129 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6130 Similar to @code{__builtin_parity}, except the argument type is
6131 @code{unsigned long long}.
6134 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6135 Returns the first argument raised to the power of the second. Unlike the
6136 @code{pow} function no guarantees about precision and rounding are made.
6139 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6140 Similar to @code{__builtin_powi}, except the argument and return types
6144 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6145 Similar to @code{__builtin_powi}, except the argument and return types
6146 are @code{long double}.
6150 @node Target Builtins
6151 @section Built-in Functions Specific to Particular Target Machines
6153 On some target machines, GCC supports many built-in functions specific
6154 to those machines. Generally these generate calls to specific machine
6155 instructions, but allow the compiler to schedule those calls.
6158 * Alpha Built-in Functions::
6159 * ARM Built-in Functions::
6160 * Blackfin Built-in Functions::
6161 * FR-V Built-in Functions::
6162 * X86 Built-in Functions::
6163 * MIPS DSP Built-in Functions::
6164 * MIPS Paired-Single Support::
6165 * PowerPC AltiVec Built-in Functions::
6166 * SPARC VIS Built-in Functions::
6169 @node Alpha Built-in Functions
6170 @subsection Alpha Built-in Functions
6172 These built-in functions are available for the Alpha family of
6173 processors, depending on the command-line switches used.
6175 The following built-in functions are always available. They
6176 all generate the machine instruction that is part of the name.
6179 long __builtin_alpha_implver (void)
6180 long __builtin_alpha_rpcc (void)
6181 long __builtin_alpha_amask (long)
6182 long __builtin_alpha_cmpbge (long, long)
6183 long __builtin_alpha_extbl (long, long)
6184 long __builtin_alpha_extwl (long, long)
6185 long __builtin_alpha_extll (long, long)
6186 long __builtin_alpha_extql (long, long)
6187 long __builtin_alpha_extwh (long, long)
6188 long __builtin_alpha_extlh (long, long)
6189 long __builtin_alpha_extqh (long, long)
6190 long __builtin_alpha_insbl (long, long)
6191 long __builtin_alpha_inswl (long, long)
6192 long __builtin_alpha_insll (long, long)
6193 long __builtin_alpha_insql (long, long)
6194 long __builtin_alpha_inswh (long, long)
6195 long __builtin_alpha_inslh (long, long)
6196 long __builtin_alpha_insqh (long, long)
6197 long __builtin_alpha_mskbl (long, long)
6198 long __builtin_alpha_mskwl (long, long)
6199 long __builtin_alpha_mskll (long, long)
6200 long __builtin_alpha_mskql (long, long)
6201 long __builtin_alpha_mskwh (long, long)
6202 long __builtin_alpha_msklh (long, long)
6203 long __builtin_alpha_mskqh (long, long)
6204 long __builtin_alpha_umulh (long, long)
6205 long __builtin_alpha_zap (long, long)
6206 long __builtin_alpha_zapnot (long, long)
6209 The following built-in functions are always with @option{-mmax}
6210 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6211 later. They all generate the machine instruction that is part
6215 long __builtin_alpha_pklb (long)
6216 long __builtin_alpha_pkwb (long)
6217 long __builtin_alpha_unpkbl (long)
6218 long __builtin_alpha_unpkbw (long)
6219 long __builtin_alpha_minub8 (long, long)
6220 long __builtin_alpha_minsb8 (long, long)
6221 long __builtin_alpha_minuw4 (long, long)
6222 long __builtin_alpha_minsw4 (long, long)
6223 long __builtin_alpha_maxub8 (long, long)
6224 long __builtin_alpha_maxsb8 (long, long)
6225 long __builtin_alpha_maxuw4 (long, long)
6226 long __builtin_alpha_maxsw4 (long, long)
6227 long __builtin_alpha_perr (long, long)
6230 The following built-in functions are always with @option{-mcix}
6231 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6232 later. They all generate the machine instruction that is part
6236 long __builtin_alpha_cttz (long)
6237 long __builtin_alpha_ctlz (long)
6238 long __builtin_alpha_ctpop (long)
6241 The following builtins are available on systems that use the OSF/1
6242 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6243 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6244 @code{rdval} and @code{wrval}.
6247 void *__builtin_thread_pointer (void)
6248 void __builtin_set_thread_pointer (void *)
6251 @node ARM Built-in Functions
6252 @subsection ARM Built-in Functions
6254 These built-in functions are available for the ARM family of
6255 processors, when the @option{-mcpu=iwmmxt} switch is used:
6258 typedef int v2si __attribute__ ((vector_size (8)));
6259 typedef short v4hi __attribute__ ((vector_size (8)));
6260 typedef char v8qi __attribute__ ((vector_size (8)));
6262 int __builtin_arm_getwcx (int)
6263 void __builtin_arm_setwcx (int, int)
6264 int __builtin_arm_textrmsb (v8qi, int)
6265 int __builtin_arm_textrmsh (v4hi, int)
6266 int __builtin_arm_textrmsw (v2si, int)
6267 int __builtin_arm_textrmub (v8qi, int)
6268 int __builtin_arm_textrmuh (v4hi, int)
6269 int __builtin_arm_textrmuw (v2si, int)
6270 v8qi __builtin_arm_tinsrb (v8qi, int)
6271 v4hi __builtin_arm_tinsrh (v4hi, int)
6272 v2si __builtin_arm_tinsrw (v2si, int)
6273 long long __builtin_arm_tmia (long long, int, int)
6274 long long __builtin_arm_tmiabb (long long, int, int)
6275 long long __builtin_arm_tmiabt (long long, int, int)
6276 long long __builtin_arm_tmiaph (long long, int, int)
6277 long long __builtin_arm_tmiatb (long long, int, int)
6278 long long __builtin_arm_tmiatt (long long, int, int)
6279 int __builtin_arm_tmovmskb (v8qi)
6280 int __builtin_arm_tmovmskh (v4hi)
6281 int __builtin_arm_tmovmskw (v2si)
6282 long long __builtin_arm_waccb (v8qi)
6283 long long __builtin_arm_wacch (v4hi)
6284 long long __builtin_arm_waccw (v2si)
6285 v8qi __builtin_arm_waddb (v8qi, v8qi)
6286 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6287 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6288 v4hi __builtin_arm_waddh (v4hi, v4hi)
6289 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6290 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6291 v2si __builtin_arm_waddw (v2si, v2si)
6292 v2si __builtin_arm_waddwss (v2si, v2si)
6293 v2si __builtin_arm_waddwus (v2si, v2si)
6294 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6295 long long __builtin_arm_wand(long long, long long)
6296 long long __builtin_arm_wandn (long long, long long)
6297 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6298 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6299 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6300 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6301 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6302 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6303 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6304 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6305 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6306 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6307 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6308 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6309 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6310 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6311 long long __builtin_arm_wmacsz (v4hi, v4hi)
6312 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6313 long long __builtin_arm_wmacuz (v4hi, v4hi)
6314 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6315 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6316 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6317 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6318 v2si __builtin_arm_wmaxsw (v2si, v2si)
6319 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6320 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6321 v2si __builtin_arm_wmaxuw (v2si, v2si)
6322 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6323 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6324 v2si __builtin_arm_wminsw (v2si, v2si)
6325 v8qi __builtin_arm_wminub (v8qi, v8qi)
6326 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6327 v2si __builtin_arm_wminuw (v2si, v2si)
6328 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6329 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6330 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6331 long long __builtin_arm_wor (long long, long long)
6332 v2si __builtin_arm_wpackdss (long long, long long)
6333 v2si __builtin_arm_wpackdus (long long, long long)
6334 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6335 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6336 v4hi __builtin_arm_wpackwss (v2si, v2si)
6337 v4hi __builtin_arm_wpackwus (v2si, v2si)
6338 long long __builtin_arm_wrord (long long, long long)
6339 long long __builtin_arm_wrordi (long long, int)
6340 v4hi __builtin_arm_wrorh (v4hi, long long)
6341 v4hi __builtin_arm_wrorhi (v4hi, int)
6342 v2si __builtin_arm_wrorw (v2si, long long)
6343 v2si __builtin_arm_wrorwi (v2si, int)
6344 v2si __builtin_arm_wsadb (v8qi, v8qi)
6345 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6346 v2si __builtin_arm_wsadh (v4hi, v4hi)
6347 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6348 v4hi __builtin_arm_wshufh (v4hi, int)
6349 long long __builtin_arm_wslld (long long, long long)
6350 long long __builtin_arm_wslldi (long long, int)
6351 v4hi __builtin_arm_wsllh (v4hi, long long)
6352 v4hi __builtin_arm_wsllhi (v4hi, int)
6353 v2si __builtin_arm_wsllw (v2si, long long)
6354 v2si __builtin_arm_wsllwi (v2si, int)
6355 long long __builtin_arm_wsrad (long long, long long)
6356 long long __builtin_arm_wsradi (long long, int)
6357 v4hi __builtin_arm_wsrah (v4hi, long long)
6358 v4hi __builtin_arm_wsrahi (v4hi, int)
6359 v2si __builtin_arm_wsraw (v2si, long long)
6360 v2si __builtin_arm_wsrawi (v2si, int)
6361 long long __builtin_arm_wsrld (long long, long long)
6362 long long __builtin_arm_wsrldi (long long, int)
6363 v4hi __builtin_arm_wsrlh (v4hi, long long)
6364 v4hi __builtin_arm_wsrlhi (v4hi, int)
6365 v2si __builtin_arm_wsrlw (v2si, long long)
6366 v2si __builtin_arm_wsrlwi (v2si, int)
6367 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6368 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6369 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6370 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6371 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6372 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6373 v2si __builtin_arm_wsubw (v2si, v2si)
6374 v2si __builtin_arm_wsubwss (v2si, v2si)
6375 v2si __builtin_arm_wsubwus (v2si, v2si)
6376 v4hi __builtin_arm_wunpckehsb (v8qi)
6377 v2si __builtin_arm_wunpckehsh (v4hi)
6378 long long __builtin_arm_wunpckehsw (v2si)
6379 v4hi __builtin_arm_wunpckehub (v8qi)
6380 v2si __builtin_arm_wunpckehuh (v4hi)
6381 long long __builtin_arm_wunpckehuw (v2si)
6382 v4hi __builtin_arm_wunpckelsb (v8qi)
6383 v2si __builtin_arm_wunpckelsh (v4hi)
6384 long long __builtin_arm_wunpckelsw (v2si)
6385 v4hi __builtin_arm_wunpckelub (v8qi)
6386 v2si __builtin_arm_wunpckeluh (v4hi)
6387 long long __builtin_arm_wunpckeluw (v2si)
6388 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6389 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6390 v2si __builtin_arm_wunpckihw (v2si, v2si)
6391 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6392 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6393 v2si __builtin_arm_wunpckilw (v2si, v2si)
6394 long long __builtin_arm_wxor (long long, long long)
6395 long long __builtin_arm_wzero ()
6398 @node Blackfin Built-in Functions
6399 @subsection Blackfin Built-in Functions
6401 Currently, there are two Blackfin-specific built-in functions. These are
6402 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6403 using inline assembly; by using these built-in functions the compiler can
6404 automatically add workarounds for hardware errata involving these
6405 instructions. These functions are named as follows:
6408 void __builtin_bfin_csync (void)
6409 void __builtin_bfin_ssync (void)
6412 @node FR-V Built-in Functions
6413 @subsection FR-V Built-in Functions
6415 GCC provides many FR-V-specific built-in functions. In general,
6416 these functions are intended to be compatible with those described
6417 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6418 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6419 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6420 pointer rather than by value.
6422 Most of the functions are named after specific FR-V instructions.
6423 Such functions are said to be ``directly mapped'' and are summarized
6424 here in tabular form.
6428 * Directly-mapped Integer Functions::
6429 * Directly-mapped Media Functions::
6430 * Raw read/write Functions::
6431 * Other Built-in Functions::
6434 @node Argument Types
6435 @subsubsection Argument Types
6437 The arguments to the built-in functions can be divided into three groups:
6438 register numbers, compile-time constants and run-time values. In order
6439 to make this classification clear at a glance, the arguments and return
6440 values are given the following pseudo types:
6442 @multitable @columnfractions .20 .30 .15 .35
6443 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6444 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6445 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6446 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6447 @item @code{uw2} @tab @code{unsigned long long} @tab No
6448 @tab an unsigned doubleword
6449 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6450 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6451 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6452 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6455 These pseudo types are not defined by GCC, they are simply a notational
6456 convenience used in this manual.
6458 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6459 and @code{sw2} are evaluated at run time. They correspond to
6460 register operands in the underlying FR-V instructions.
6462 @code{const} arguments represent immediate operands in the underlying
6463 FR-V instructions. They must be compile-time constants.
6465 @code{acc} arguments are evaluated at compile time and specify the number
6466 of an accumulator register. For example, an @code{acc} argument of 2
6467 will select the ACC2 register.
6469 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6470 number of an IACC register. See @pxref{Other Built-in Functions}
6473 @node Directly-mapped Integer Functions
6474 @subsubsection Directly-mapped Integer Functions
6476 The functions listed below map directly to FR-V I-type instructions.
6478 @multitable @columnfractions .45 .32 .23
6479 @item Function prototype @tab Example usage @tab Assembly output
6480 @item @code{sw1 __ADDSS (sw1, sw1)}
6481 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6482 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6483 @item @code{sw1 __SCAN (sw1, sw1)}
6484 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6485 @tab @code{SCAN @var{a},@var{b},@var{c}}
6486 @item @code{sw1 __SCUTSS (sw1)}
6487 @tab @code{@var{b} = __SCUTSS (@var{a})}
6488 @tab @code{SCUTSS @var{a},@var{b}}
6489 @item @code{sw1 __SLASS (sw1, sw1)}
6490 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6491 @tab @code{SLASS @var{a},@var{b},@var{c}}
6492 @item @code{void __SMASS (sw1, sw1)}
6493 @tab @code{__SMASS (@var{a}, @var{b})}
6494 @tab @code{SMASS @var{a},@var{b}}
6495 @item @code{void __SMSSS (sw1, sw1)}
6496 @tab @code{__SMSSS (@var{a}, @var{b})}
6497 @tab @code{SMSSS @var{a},@var{b}}
6498 @item @code{void __SMU (sw1, sw1)}
6499 @tab @code{__SMU (@var{a}, @var{b})}
6500 @tab @code{SMU @var{a},@var{b}}
6501 @item @code{sw2 __SMUL (sw1, sw1)}
6502 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6503 @tab @code{SMUL @var{a},@var{b},@var{c}}
6504 @item @code{sw1 __SUBSS (sw1, sw1)}
6505 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6506 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6507 @item @code{uw2 __UMUL (uw1, uw1)}
6508 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6509 @tab @code{UMUL @var{a},@var{b},@var{c}}
6512 @node Directly-mapped Media Functions
6513 @subsubsection Directly-mapped Media Functions
6515 The functions listed below map directly to FR-V M-type instructions.
6517 @multitable @columnfractions .45 .32 .23
6518 @item Function prototype @tab Example usage @tab Assembly output
6519 @item @code{uw1 __MABSHS (sw1)}
6520 @tab @code{@var{b} = __MABSHS (@var{a})}
6521 @tab @code{MABSHS @var{a},@var{b}}
6522 @item @code{void __MADDACCS (acc, acc)}
6523 @tab @code{__MADDACCS (@var{b}, @var{a})}
6524 @tab @code{MADDACCS @var{a},@var{b}}
6525 @item @code{sw1 __MADDHSS (sw1, sw1)}
6526 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6527 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6528 @item @code{uw1 __MADDHUS (uw1, uw1)}
6529 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6530 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6531 @item @code{uw1 __MAND (uw1, uw1)}
6532 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6533 @tab @code{MAND @var{a},@var{b},@var{c}}
6534 @item @code{void __MASACCS (acc, acc)}
6535 @tab @code{__MASACCS (@var{b}, @var{a})}
6536 @tab @code{MASACCS @var{a},@var{b}}
6537 @item @code{uw1 __MAVEH (uw1, uw1)}
6538 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6539 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6540 @item @code{uw2 __MBTOH (uw1)}
6541 @tab @code{@var{b} = __MBTOH (@var{a})}
6542 @tab @code{MBTOH @var{a},@var{b}}
6543 @item @code{void __MBTOHE (uw1 *, uw1)}
6544 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6545 @tab @code{MBTOHE @var{a},@var{b}}
6546 @item @code{void __MCLRACC (acc)}
6547 @tab @code{__MCLRACC (@var{a})}
6548 @tab @code{MCLRACC @var{a}}
6549 @item @code{void __MCLRACCA (void)}
6550 @tab @code{__MCLRACCA ()}
6551 @tab @code{MCLRACCA}
6552 @item @code{uw1 __Mcop1 (uw1, uw1)}
6553 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6554 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6555 @item @code{uw1 __Mcop2 (uw1, uw1)}
6556 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6557 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6558 @item @code{uw1 __MCPLHI (uw2, const)}
6559 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6560 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6561 @item @code{uw1 __MCPLI (uw2, const)}
6562 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6563 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6564 @item @code{void __MCPXIS (acc, sw1, sw1)}
6565 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6566 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6567 @item @code{void __MCPXIU (acc, uw1, uw1)}
6568 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6569 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6570 @item @code{void __MCPXRS (acc, sw1, sw1)}
6571 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6572 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6573 @item @code{void __MCPXRU (acc, uw1, uw1)}
6574 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6575 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6576 @item @code{uw1 __MCUT (acc, uw1)}
6577 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6578 @tab @code{MCUT @var{a},@var{b},@var{c}}
6579 @item @code{uw1 __MCUTSS (acc, sw1)}
6580 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6581 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6582 @item @code{void __MDADDACCS (acc, acc)}
6583 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6584 @tab @code{MDADDACCS @var{a},@var{b}}
6585 @item @code{void __MDASACCS (acc, acc)}
6586 @tab @code{__MDASACCS (@var{b}, @var{a})}
6587 @tab @code{MDASACCS @var{a},@var{b}}
6588 @item @code{uw2 __MDCUTSSI (acc, const)}
6589 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6590 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6591 @item @code{uw2 __MDPACKH (uw2, uw2)}
6592 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6593 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6594 @item @code{uw2 __MDROTLI (uw2, const)}
6595 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6596 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6597 @item @code{void __MDSUBACCS (acc, acc)}
6598 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6599 @tab @code{MDSUBACCS @var{a},@var{b}}
6600 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6601 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6602 @tab @code{MDUNPACKH @var{a},@var{b}}
6603 @item @code{uw2 __MEXPDHD (uw1, const)}
6604 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6605 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6606 @item @code{uw1 __MEXPDHW (uw1, const)}
6607 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6608 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6609 @item @code{uw1 __MHDSETH (uw1, const)}
6610 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6611 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6612 @item @code{sw1 __MHDSETS (const)}
6613 @tab @code{@var{b} = __MHDSETS (@var{a})}
6614 @tab @code{MHDSETS #@var{a},@var{b}}
6615 @item @code{uw1 __MHSETHIH (uw1, const)}
6616 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6617 @tab @code{MHSETHIH #@var{a},@var{b}}
6618 @item @code{sw1 __MHSETHIS (sw1, const)}
6619 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6620 @tab @code{MHSETHIS #@var{a},@var{b}}
6621 @item @code{uw1 __MHSETLOH (uw1, const)}
6622 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6623 @tab @code{MHSETLOH #@var{a},@var{b}}
6624 @item @code{sw1 __MHSETLOS (sw1, const)}
6625 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6626 @tab @code{MHSETLOS #@var{a},@var{b}}
6627 @item @code{uw1 __MHTOB (uw2)}
6628 @tab @code{@var{b} = __MHTOB (@var{a})}
6629 @tab @code{MHTOB @var{a},@var{b}}
6630 @item @code{void __MMACHS (acc, sw1, sw1)}
6631 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6632 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6633 @item @code{void __MMACHU (acc, uw1, uw1)}
6634 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6635 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6636 @item @code{void __MMRDHS (acc, sw1, sw1)}
6637 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6638 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6639 @item @code{void __MMRDHU (acc, uw1, uw1)}
6640 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6641 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6642 @item @code{void __MMULHS (acc, sw1, sw1)}
6643 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6644 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6645 @item @code{void __MMULHU (acc, uw1, uw1)}
6646 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6647 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6648 @item @code{void __MMULXHS (acc, sw1, sw1)}
6649 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6650 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6651 @item @code{void __MMULXHU (acc, uw1, uw1)}
6652 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6653 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6654 @item @code{uw1 __MNOT (uw1)}
6655 @tab @code{@var{b} = __MNOT (@var{a})}
6656 @tab @code{MNOT @var{a},@var{b}}
6657 @item @code{uw1 __MOR (uw1, uw1)}
6658 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6659 @tab @code{MOR @var{a},@var{b},@var{c}}
6660 @item @code{uw1 __MPACKH (uh, uh)}
6661 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6662 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6663 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6664 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6665 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6666 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6667 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6668 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6669 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6670 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6671 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6672 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6673 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6674 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6675 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6676 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6677 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6678 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6679 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6680 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6681 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6682 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6683 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6684 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6685 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6686 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6687 @item @code{void __MQMACHS (acc, sw2, sw2)}
6688 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6689 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6690 @item @code{void __MQMACHU (acc, uw2, uw2)}
6691 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6692 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6693 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6694 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6695 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6696 @item @code{void __MQMULHS (acc, sw2, sw2)}
6697 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6698 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6699 @item @code{void __MQMULHU (acc, uw2, uw2)}
6700 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6701 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6702 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6703 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6704 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6705 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6706 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6707 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6708 @item @code{sw2 __MQSATHS (sw2, sw2)}
6709 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6710 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6711 @item @code{uw2 __MQSLLHI (uw2, int)}
6712 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6713 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6714 @item @code{sw2 __MQSRAHI (sw2, int)}
6715 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6716 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6717 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6718 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6719 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6720 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6721 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6722 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6723 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6724 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6725 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6726 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6727 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6728 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6729 @item @code{uw1 __MRDACC (acc)}
6730 @tab @code{@var{b} = __MRDACC (@var{a})}
6731 @tab @code{MRDACC @var{a},@var{b}}
6732 @item @code{uw1 __MRDACCG (acc)}
6733 @tab @code{@var{b} = __MRDACCG (@var{a})}
6734 @tab @code{MRDACCG @var{a},@var{b}}
6735 @item @code{uw1 __MROTLI (uw1, const)}
6736 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6737 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6738 @item @code{uw1 __MROTRI (uw1, const)}
6739 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6740 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6741 @item @code{sw1 __MSATHS (sw1, sw1)}
6742 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6743 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6744 @item @code{uw1 __MSATHU (uw1, uw1)}
6745 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6746 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6747 @item @code{uw1 __MSLLHI (uw1, const)}
6748 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6749 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6750 @item @code{sw1 __MSRAHI (sw1, const)}
6751 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6752 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6753 @item @code{uw1 __MSRLHI (uw1, const)}
6754 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6755 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6756 @item @code{void __MSUBACCS (acc, acc)}
6757 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6758 @tab @code{MSUBACCS @var{a},@var{b}}
6759 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6760 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6761 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6762 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6763 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6764 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6765 @item @code{void __MTRAP (void)}
6766 @tab @code{__MTRAP ()}
6768 @item @code{uw2 __MUNPACKH (uw1)}
6769 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6770 @tab @code{MUNPACKH @var{a},@var{b}}
6771 @item @code{uw1 __MWCUT (uw2, uw1)}
6772 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6773 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6774 @item @code{void __MWTACC (acc, uw1)}
6775 @tab @code{__MWTACC (@var{b}, @var{a})}
6776 @tab @code{MWTACC @var{a},@var{b}}
6777 @item @code{void __MWTACCG (acc, uw1)}
6778 @tab @code{__MWTACCG (@var{b}, @var{a})}
6779 @tab @code{MWTACCG @var{a},@var{b}}
6780 @item @code{uw1 __MXOR (uw1, uw1)}
6781 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6782 @tab @code{MXOR @var{a},@var{b},@var{c}}
6785 @node Raw read/write Functions
6786 @subsubsection Raw read/write Functions
6788 This sections describes built-in functions related to read and write
6789 instructions to access memory. These functions generate
6790 @code{membar} instructions to flush the I/O load and stores where
6791 appropriate, as described in Fujitsu's manual described above.
6795 @item unsigned char __builtin_read8 (void *@var{data})
6796 @item unsigned short __builtin_read16 (void *@var{data})
6797 @item unsigned long __builtin_read32 (void *@var{data})
6798 @item unsigned long long __builtin_read64 (void *@var{data})
6800 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6801 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6802 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6803 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6806 @node Other Built-in Functions
6807 @subsubsection Other Built-in Functions
6809 This section describes built-in functions that are not named after
6810 a specific FR-V instruction.
6813 @item sw2 __IACCreadll (iacc @var{reg})
6814 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6815 for future expansion and must be 0.
6817 @item sw1 __IACCreadl (iacc @var{reg})
6818 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6819 Other values of @var{reg} are rejected as invalid.
6821 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6822 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6823 is reserved for future expansion and must be 0.
6825 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6826 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6827 is 1. Other values of @var{reg} are rejected as invalid.
6829 @item void __data_prefetch0 (const void *@var{x})
6830 Use the @code{dcpl} instruction to load the contents of address @var{x}
6831 into the data cache.
6833 @item void __data_prefetch (const void *@var{x})
6834 Use the @code{nldub} instruction to load the contents of address @var{x}
6835 into the data cache. The instruction will be issued in slot I1@.
6838 @node X86 Built-in Functions
6839 @subsection X86 Built-in Functions
6841 These built-in functions are available for the i386 and x86-64 family
6842 of computers, depending on the command-line switches used.
6844 Note that, if you specify command-line switches such as @option{-msse},
6845 the compiler could use the extended instruction sets even if the built-ins
6846 are not used explicitly in the program. For this reason, applications
6847 which perform runtime CPU detection must compile separate files for each
6848 supported architecture, using the appropriate flags. In particular,
6849 the file containing the CPU detection code should be compiled without
6852 The following machine modes are available for use with MMX built-in functions
6853 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6854 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6855 vector of eight 8-bit integers. Some of the built-in functions operate on
6856 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6858 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6859 of two 32-bit floating point values.
6861 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6862 floating point values. Some instructions use a vector of four 32-bit
6863 integers, these use @code{V4SI}. Finally, some instructions operate on an
6864 entire vector register, interpreting it as a 128-bit integer, these use mode
6867 The following built-in functions are made available by @option{-mmmx}.
6868 All of them generate the machine instruction that is part of the name.
6871 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6872 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6873 v2si __builtin_ia32_paddd (v2si, v2si)
6874 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6875 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6876 v2si __builtin_ia32_psubd (v2si, v2si)
6877 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6878 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6879 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6880 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6881 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6882 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6883 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6884 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6885 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6886 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6887 di __builtin_ia32_pand (di, di)
6888 di __builtin_ia32_pandn (di,di)
6889 di __builtin_ia32_por (di, di)
6890 di __builtin_ia32_pxor (di, di)
6891 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6892 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6893 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6894 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6895 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6896 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6897 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6898 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6899 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6900 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6901 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6902 v2si __builtin_ia32_punpckldq (v2si, v2si)
6903 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6904 v4hi __builtin_ia32_packssdw (v2si, v2si)
6905 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6908 The following built-in functions are made available either with
6909 @option{-msse}, or with a combination of @option{-m3dnow} and
6910 @option{-march=athlon}. All of them generate the machine
6911 instruction that is part of the name.
6914 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6915 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6916 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6917 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6918 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6919 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6920 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6921 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6922 int __builtin_ia32_pextrw (v4hi, int)
6923 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6924 int __builtin_ia32_pmovmskb (v8qi)
6925 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6926 void __builtin_ia32_movntq (di *, di)
6927 void __builtin_ia32_sfence (void)
6930 The following built-in functions are available when @option{-msse} is used.
6931 All of them generate the machine instruction that is part of the name.
6934 int __builtin_ia32_comieq (v4sf, v4sf)
6935 int __builtin_ia32_comineq (v4sf, v4sf)
6936 int __builtin_ia32_comilt (v4sf, v4sf)
6937 int __builtin_ia32_comile (v4sf, v4sf)
6938 int __builtin_ia32_comigt (v4sf, v4sf)
6939 int __builtin_ia32_comige (v4sf, v4sf)
6940 int __builtin_ia32_ucomieq (v4sf, v4sf)
6941 int __builtin_ia32_ucomineq (v4sf, v4sf)
6942 int __builtin_ia32_ucomilt (v4sf, v4sf)
6943 int __builtin_ia32_ucomile (v4sf, v4sf)
6944 int __builtin_ia32_ucomigt (v4sf, v4sf)
6945 int __builtin_ia32_ucomige (v4sf, v4sf)
6946 v4sf __builtin_ia32_addps (v4sf, v4sf)
6947 v4sf __builtin_ia32_subps (v4sf, v4sf)
6948 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6949 v4sf __builtin_ia32_divps (v4sf, v4sf)
6950 v4sf __builtin_ia32_addss (v4sf, v4sf)
6951 v4sf __builtin_ia32_subss (v4sf, v4sf)
6952 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6953 v4sf __builtin_ia32_divss (v4sf, v4sf)
6954 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6955 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6956 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6957 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6958 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6959 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6960 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6961 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6962 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6963 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6964 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6965 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6966 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6967 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6968 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6969 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6970 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6971 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6972 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6973 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6974 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6975 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6976 v4sf __builtin_ia32_minps (v4sf, v4sf)
6977 v4sf __builtin_ia32_minss (v4sf, v4sf)
6978 v4sf __builtin_ia32_andps (v4sf, v4sf)
6979 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6980 v4sf __builtin_ia32_orps (v4sf, v4sf)
6981 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6982 v4sf __builtin_ia32_movss (v4sf, v4sf)
6983 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6984 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6985 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6986 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6987 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6988 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6989 v2si __builtin_ia32_cvtps2pi (v4sf)
6990 int __builtin_ia32_cvtss2si (v4sf)
6991 v2si __builtin_ia32_cvttps2pi (v4sf)
6992 int __builtin_ia32_cvttss2si (v4sf)
6993 v4sf __builtin_ia32_rcpps (v4sf)
6994 v4sf __builtin_ia32_rsqrtps (v4sf)
6995 v4sf __builtin_ia32_sqrtps (v4sf)
6996 v4sf __builtin_ia32_rcpss (v4sf)
6997 v4sf __builtin_ia32_rsqrtss (v4sf)
6998 v4sf __builtin_ia32_sqrtss (v4sf)
6999 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7000 void __builtin_ia32_movntps (float *, v4sf)
7001 int __builtin_ia32_movmskps (v4sf)
7004 The following built-in functions are available when @option{-msse} is used.
7007 @item v4sf __builtin_ia32_loadaps (float *)
7008 Generates the @code{movaps} machine instruction as a load from memory.
7009 @item void __builtin_ia32_storeaps (float *, v4sf)
7010 Generates the @code{movaps} machine instruction as a store to memory.
7011 @item v4sf __builtin_ia32_loadups (float *)
7012 Generates the @code{movups} machine instruction as a load from memory.
7013 @item void __builtin_ia32_storeups (float *, v4sf)
7014 Generates the @code{movups} machine instruction as a store to memory.
7015 @item v4sf __builtin_ia32_loadsss (float *)
7016 Generates the @code{movss} machine instruction as a load from memory.
7017 @item void __builtin_ia32_storess (float *, v4sf)
7018 Generates the @code{movss} machine instruction as a store to memory.
7019 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7020 Generates the @code{movhps} machine instruction as a load from memory.
7021 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7022 Generates the @code{movlps} machine instruction as a load from memory
7023 @item void __builtin_ia32_storehps (v4sf, v2si *)
7024 Generates the @code{movhps} machine instruction as a store to memory.
7025 @item void __builtin_ia32_storelps (v4sf, v2si *)
7026 Generates the @code{movlps} machine instruction as a store to memory.
7029 The following built-in functions are available when @option{-msse2} is used.
7030 All of them generate the machine instruction that is part of the name.
7033 int __builtin_ia32_comisdeq (v2df, v2df)
7034 int __builtin_ia32_comisdlt (v2df, v2df)
7035 int __builtin_ia32_comisdle (v2df, v2df)
7036 int __builtin_ia32_comisdgt (v2df, v2df)
7037 int __builtin_ia32_comisdge (v2df, v2df)
7038 int __builtin_ia32_comisdneq (v2df, v2df)
7039 int __builtin_ia32_ucomisdeq (v2df, v2df)
7040 int __builtin_ia32_ucomisdlt (v2df, v2df)
7041 int __builtin_ia32_ucomisdle (v2df, v2df)
7042 int __builtin_ia32_ucomisdgt (v2df, v2df)
7043 int __builtin_ia32_ucomisdge (v2df, v2df)
7044 int __builtin_ia32_ucomisdneq (v2df, v2df)
7045 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7046 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7047 v2df __builtin_ia32_cmplepd (v2df, v2df)
7048 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7049 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7050 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7051 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7052 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7053 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7054 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7055 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7056 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7057 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7058 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7059 v2df __builtin_ia32_cmplesd (v2df, v2df)
7060 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7061 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7062 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7063 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7064 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7065 v2di __builtin_ia32_paddq (v2di, v2di)
7066 v2di __builtin_ia32_psubq (v2di, v2di)
7067 v2df __builtin_ia32_addpd (v2df, v2df)
7068 v2df __builtin_ia32_subpd (v2df, v2df)
7069 v2df __builtin_ia32_mulpd (v2df, v2df)
7070 v2df __builtin_ia32_divpd (v2df, v2df)
7071 v2df __builtin_ia32_addsd (v2df, v2df)
7072 v2df __builtin_ia32_subsd (v2df, v2df)
7073 v2df __builtin_ia32_mulsd (v2df, v2df)
7074 v2df __builtin_ia32_divsd (v2df, v2df)
7075 v2df __builtin_ia32_minpd (v2df, v2df)
7076 v2df __builtin_ia32_maxpd (v2df, v2df)
7077 v2df __builtin_ia32_minsd (v2df, v2df)
7078 v2df __builtin_ia32_maxsd (v2df, v2df)
7079 v2df __builtin_ia32_andpd (v2df, v2df)
7080 v2df __builtin_ia32_andnpd (v2df, v2df)
7081 v2df __builtin_ia32_orpd (v2df, v2df)
7082 v2df __builtin_ia32_xorpd (v2df, v2df)
7083 v2df __builtin_ia32_movsd (v2df, v2df)
7084 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7085 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7086 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7087 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7088 v4si __builtin_ia32_paddd128 (v4si, v4si)
7089 v2di __builtin_ia32_paddq128 (v2di, v2di)
7090 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7091 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7092 v4si __builtin_ia32_psubd128 (v4si, v4si)
7093 v2di __builtin_ia32_psubq128 (v2di, v2di)
7094 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7095 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7096 v2di __builtin_ia32_pand128 (v2di, v2di)
7097 v2di __builtin_ia32_pandn128 (v2di, v2di)
7098 v2di __builtin_ia32_por128 (v2di, v2di)
7099 v2di __builtin_ia32_pxor128 (v2di, v2di)
7100 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7101 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7102 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7103 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7104 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7105 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7106 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7107 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7108 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7109 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7110 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7111 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7112 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7113 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7114 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7115 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7116 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7117 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7118 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7119 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7120 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7121 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7122 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7123 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7124 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7125 v2df __builtin_ia32_loadupd (double *)
7126 void __builtin_ia32_storeupd (double *, v2df)
7127 v2df __builtin_ia32_loadhpd (v2df, double *)
7128 v2df __builtin_ia32_loadlpd (v2df, double *)
7129 int __builtin_ia32_movmskpd (v2df)
7130 int __builtin_ia32_pmovmskb128 (v16qi)
7131 void __builtin_ia32_movnti (int *, int)
7132 void __builtin_ia32_movntpd (double *, v2df)
7133 void __builtin_ia32_movntdq (v2df *, v2df)
7134 v4si __builtin_ia32_pshufd (v4si, int)
7135 v8hi __builtin_ia32_pshuflw (v8hi, int)
7136 v8hi __builtin_ia32_pshufhw (v8hi, int)
7137 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7138 v2df __builtin_ia32_sqrtpd (v2df)
7139 v2df __builtin_ia32_sqrtsd (v2df)
7140 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7141 v2df __builtin_ia32_cvtdq2pd (v4si)
7142 v4sf __builtin_ia32_cvtdq2ps (v4si)
7143 v4si __builtin_ia32_cvtpd2dq (v2df)
7144 v2si __builtin_ia32_cvtpd2pi (v2df)
7145 v4sf __builtin_ia32_cvtpd2ps (v2df)
7146 v4si __builtin_ia32_cvttpd2dq (v2df)
7147 v2si __builtin_ia32_cvttpd2pi (v2df)
7148 v2df __builtin_ia32_cvtpi2pd (v2si)
7149 int __builtin_ia32_cvtsd2si (v2df)
7150 int __builtin_ia32_cvttsd2si (v2df)
7151 long long __builtin_ia32_cvtsd2si64 (v2df)
7152 long long __builtin_ia32_cvttsd2si64 (v2df)
7153 v4si __builtin_ia32_cvtps2dq (v4sf)
7154 v2df __builtin_ia32_cvtps2pd (v4sf)
7155 v4si __builtin_ia32_cvttps2dq (v4sf)
7156 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7157 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7158 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7159 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7160 void __builtin_ia32_clflush (const void *)
7161 void __builtin_ia32_lfence (void)
7162 void __builtin_ia32_mfence (void)
7163 v16qi __builtin_ia32_loaddqu (const char *)
7164 void __builtin_ia32_storedqu (char *, v16qi)
7165 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7166 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7167 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7168 v4si __builtin_ia32_pslld128 (v4si, v2di)
7169 v2di __builtin_ia32_psllq128 (v4si, v2di)
7170 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7171 v4si __builtin_ia32_psrld128 (v4si, v2di)
7172 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7173 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7174 v4si __builtin_ia32_psrad128 (v4si, v2di)
7175 v2di __builtin_ia32_pslldqi128 (v2di, int)
7176 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7177 v4si __builtin_ia32_pslldi128 (v4si, int)
7178 v2di __builtin_ia32_psllqi128 (v2di, int)
7179 v2di __builtin_ia32_psrldqi128 (v2di, int)
7180 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7181 v4si __builtin_ia32_psrldi128 (v4si, int)
7182 v2di __builtin_ia32_psrlqi128 (v2di, int)
7183 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7184 v4si __builtin_ia32_psradi128 (v4si, int)
7185 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7188 The following built-in functions are available when @option{-msse3} is used.
7189 All of them generate the machine instruction that is part of the name.
7192 v2df __builtin_ia32_addsubpd (v2df, v2df)
7193 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7194 v2df __builtin_ia32_haddpd (v2df, v2df)
7195 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7196 v2df __builtin_ia32_hsubpd (v2df, v2df)
7197 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7198 v16qi __builtin_ia32_lddqu (char const *)
7199 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7200 v2df __builtin_ia32_movddup (v2df)
7201 v4sf __builtin_ia32_movshdup (v4sf)
7202 v4sf __builtin_ia32_movsldup (v4sf)
7203 void __builtin_ia32_mwait (unsigned int, unsigned int)
7206 The following built-in functions are available when @option{-msse3} is used.
7209 @item v2df __builtin_ia32_loadddup (double const *)
7210 Generates the @code{movddup} machine instruction as a load from memory.
7213 The following built-in functions are available when @option{-mssse3} is used.
7214 All of them generate the machine instruction that is part of the name
7218 v2si __builtin_ia32_phaddd (v2si, v2si)
7219 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7220 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7221 v2si __builtin_ia32_phsubd (v2si, v2si)
7222 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7223 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7224 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7225 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7226 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7227 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7228 v2si __builtin_ia32_psignd (v2si, v2si)
7229 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7230 long long __builtin_ia32_palignr (long long, long long, int)
7231 v8qi __builtin_ia32_pabsb (v8qi)
7232 v2si __builtin_ia32_pabsd (v2si)
7233 v4hi __builtin_ia32_pabsw (v4hi)
7236 The following built-in functions are available when @option{-mssse3} is used.
7237 All of them generate the machine instruction that is part of the name
7241 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7242 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7243 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7244 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7245 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7246 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7247 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7248 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7249 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7250 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7251 v4si __builtin_ia32_psignd128 (v4si, v4si)
7252 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7253 v2di __builtin_ia32_palignr (v2di, v2di, int)
7254 v16qi __builtin_ia32_pabsb128 (v16qi)
7255 v4si __builtin_ia32_pabsd128 (v4si)
7256 v8hi __builtin_ia32_pabsw128 (v8hi)
7259 The following built-in functions are available when @option{-msse4a} is used.
7262 void _mm_stream_sd (double*,__m128d);
7263 Generates the @code{movntsd} machine instruction.
7264 void _mm_stream_ss (float*,__m128);
7265 Generates the @code{movntss} machine instruction.
7266 __m128i _mm_extract_si64 (__m128i, __m128i);
7267 Generates the @code{extrq} machine instruction with only SSE register operands.
7268 __m128i _mm_extracti_si64 (__m128i, int, int);
7269 Generates the @code{extrq} machine instruction with SSE register and immediate operands.
7270 __m128i _mm_insert_si64 (__m128i, __m128i);
7271 Generates the @code{insertq} machine instruction with only SSE register operands.
7272 __m128i _mm_inserti_si64 (__m128i, __m128i, int, int);
7273 Generates the @code{insertq} machine instruction with SSE register and immediate operands.
7276 The following built-in functions are available when @option{-m3dnow} is used.
7277 All of them generate the machine instruction that is part of the name.
7280 void __builtin_ia32_femms (void)
7281 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7282 v2si __builtin_ia32_pf2id (v2sf)
7283 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7284 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7285 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7286 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7287 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7288 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7289 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7290 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7291 v2sf __builtin_ia32_pfrcp (v2sf)
7292 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7293 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7294 v2sf __builtin_ia32_pfrsqrt (v2sf)
7295 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7296 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7297 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7298 v2sf __builtin_ia32_pi2fd (v2si)
7299 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7302 The following built-in functions are available when both @option{-m3dnow}
7303 and @option{-march=athlon} are used. All of them generate the machine
7304 instruction that is part of the name.
7307 v2si __builtin_ia32_pf2iw (v2sf)
7308 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7309 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7310 v2sf __builtin_ia32_pi2fw (v2si)
7311 v2sf __builtin_ia32_pswapdsf (v2sf)
7312 v2si __builtin_ia32_pswapdsi (v2si)
7315 @node MIPS DSP Built-in Functions
7316 @subsection MIPS DSP Built-in Functions
7318 The MIPS DSP Application-Specific Extension (ASE) includes new
7319 instructions that are designed to improve the performance of DSP and
7320 media applications. It provides instructions that operate on packed
7321 8-bit integer data, Q15 fractional data and Q31 fractional data.
7323 GCC supports MIPS DSP operations using both the generic
7324 vector extensions (@pxref{Vector Extensions}) and a collection of
7325 MIPS-specific built-in functions. Both kinds of support are
7326 enabled by the @option{-mdsp} command-line option.
7328 At present, GCC only provides support for operations on 32-bit
7329 vectors. The vector type associated with 8-bit integer data is
7330 usually called @code{v4i8} and the vector type associated with Q15 is
7331 usually called @code{v2q15}. They can be defined in C as follows:
7334 typedef char v4i8 __attribute__ ((vector_size(4)));
7335 typedef short v2q15 __attribute__ ((vector_size(4)));
7338 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7339 aggregates. For example:
7342 v4i8 a = @{1, 2, 3, 4@};
7344 b = (v4i8) @{5, 6, 7, 8@};
7346 v2q15 c = @{0x0fcb, 0x3a75@};
7348 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7351 @emph{Note:} The CPU's endianness determines the order in which values
7352 are packed. On little-endian targets, the first value is the least
7353 significant and the last value is the most significant. The opposite
7354 order applies to big-endian targets. For example, the code above will
7355 set the lowest byte of @code{a} to @code{1} on little-endian targets
7356 and @code{4} on big-endian targets.
7358 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7359 representation. As shown in this example, the integer representation
7360 of a Q15 value can be obtained by multiplying the fractional value by
7361 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7364 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7365 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7366 and @code{c} and @code{d} are @code{v2q15} values.
7368 @multitable @columnfractions .50 .50
7369 @item C code @tab MIPS instruction
7370 @item @code{a + b} @tab @code{addu.qb}
7371 @item @code{c + d} @tab @code{addq.ph}
7372 @item @code{a - b} @tab @code{subu.qb}
7373 @item @code{c - d} @tab @code{subq.ph}
7376 It is easier to describe the DSP built-in functions if we first define
7377 the following types:
7382 typedef long long a64;
7385 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7386 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7387 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7388 @code{long long}, but we use @code{a64} to indicate values that will
7389 be placed in one of the four DSP accumulators (@code{$ac0},
7390 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7392 Also, some built-in functions prefer or require immediate numbers as
7393 parameters, because the corresponding DSP instructions accept both immediate
7394 numbers and register operands, or accept immediate numbers only. The
7395 immediate parameters are listed as follows.
7403 imm_n32_31: -32 to 31.
7404 imm_n512_511: -512 to 511.
7407 The following built-in functions map directly to a particular MIPS DSP
7408 instruction. Please refer to the architecture specification
7409 for details on what each instruction does.
7412 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7413 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7414 q31 __builtin_mips_addq_s_w (q31, q31)
7415 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7416 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7417 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7418 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7419 q31 __builtin_mips_subq_s_w (q31, q31)
7420 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7421 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7422 i32 __builtin_mips_addsc (i32, i32)
7423 i32 __builtin_mips_addwc (i32, i32)
7424 i32 __builtin_mips_modsub (i32, i32)
7425 i32 __builtin_mips_raddu_w_qb (v4i8)
7426 v2q15 __builtin_mips_absq_s_ph (v2q15)
7427 q31 __builtin_mips_absq_s_w (q31)
7428 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7429 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7430 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7431 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7432 q31 __builtin_mips_preceq_w_phl (v2q15)
7433 q31 __builtin_mips_preceq_w_phr (v2q15)
7434 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7435 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7436 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7437 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7438 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7439 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7440 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7441 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7442 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7443 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7444 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7445 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7446 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7447 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7448 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7449 q31 __builtin_mips_shll_s_w (q31, i32)
7450 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7451 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7452 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7453 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7454 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7455 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7456 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7457 q31 __builtin_mips_shra_r_w (q31, i32)
7458 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7459 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7460 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7461 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7462 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7463 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7464 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7465 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7466 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7467 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7468 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7469 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7470 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7471 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7472 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7473 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7474 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7475 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7476 i32 __builtin_mips_bitrev (i32)
7477 i32 __builtin_mips_insv (i32, i32)
7478 v4i8 __builtin_mips_repl_qb (imm0_255)
7479 v4i8 __builtin_mips_repl_qb (i32)
7480 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7481 v2q15 __builtin_mips_repl_ph (i32)
7482 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7483 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7484 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7485 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7486 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7487 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7488 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7489 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7490 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7491 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7492 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7493 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7494 i32 __builtin_mips_extr_w (a64, imm0_31)
7495 i32 __builtin_mips_extr_w (a64, i32)
7496 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7497 i32 __builtin_mips_extr_s_h (a64, i32)
7498 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7499 i32 __builtin_mips_extr_rs_w (a64, i32)
7500 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7501 i32 __builtin_mips_extr_r_w (a64, i32)
7502 i32 __builtin_mips_extp (a64, imm0_31)
7503 i32 __builtin_mips_extp (a64, i32)
7504 i32 __builtin_mips_extpdp (a64, imm0_31)
7505 i32 __builtin_mips_extpdp (a64, i32)
7506 a64 __builtin_mips_shilo (a64, imm_n32_31)
7507 a64 __builtin_mips_shilo (a64, i32)
7508 a64 __builtin_mips_mthlip (a64, i32)
7509 void __builtin_mips_wrdsp (i32, imm0_63)
7510 i32 __builtin_mips_rddsp (imm0_63)
7511 i32 __builtin_mips_lbux (void *, i32)
7512 i32 __builtin_mips_lhx (void *, i32)
7513 i32 __builtin_mips_lwx (void *, i32)
7514 i32 __builtin_mips_bposge32 (void)
7517 @node MIPS Paired-Single Support
7518 @subsection MIPS Paired-Single Support
7520 The MIPS64 architecture includes a number of instructions that
7521 operate on pairs of single-precision floating-point values.
7522 Each pair is packed into a 64-bit floating-point register,
7523 with one element being designated the ``upper half'' and
7524 the other being designated the ``lower half''.
7526 GCC supports paired-single operations using both the generic
7527 vector extensions (@pxref{Vector Extensions}) and a collection of
7528 MIPS-specific built-in functions. Both kinds of support are
7529 enabled by the @option{-mpaired-single} command-line option.
7531 The vector type associated with paired-single values is usually
7532 called @code{v2sf}. It can be defined in C as follows:
7535 typedef float v2sf __attribute__ ((vector_size (8)));
7538 @code{v2sf} values are initialized in the same way as aggregates.
7542 v2sf a = @{1.5, 9.1@};
7545 b = (v2sf) @{e, f@};
7548 @emph{Note:} The CPU's endianness determines which value is stored in
7549 the upper half of a register and which value is stored in the lower half.
7550 On little-endian targets, the first value is the lower one and the second
7551 value is the upper one. The opposite order applies to big-endian targets.
7552 For example, the code above will set the lower half of @code{a} to
7553 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7556 * Paired-Single Arithmetic::
7557 * Paired-Single Built-in Functions::
7558 * MIPS-3D Built-in Functions::
7561 @node Paired-Single Arithmetic
7562 @subsubsection Paired-Single Arithmetic
7564 The table below lists the @code{v2sf} operations for which hardware
7565 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7566 values and @code{x} is an integral value.
7568 @multitable @columnfractions .50 .50
7569 @item C code @tab MIPS instruction
7570 @item @code{a + b} @tab @code{add.ps}
7571 @item @code{a - b} @tab @code{sub.ps}
7572 @item @code{-a} @tab @code{neg.ps}
7573 @item @code{a * b} @tab @code{mul.ps}
7574 @item @code{a * b + c} @tab @code{madd.ps}
7575 @item @code{a * b - c} @tab @code{msub.ps}
7576 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7577 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7578 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7581 Note that the multiply-accumulate instructions can be disabled
7582 using the command-line option @code{-mno-fused-madd}.
7584 @node Paired-Single Built-in Functions
7585 @subsubsection Paired-Single Built-in Functions
7587 The following paired-single functions map directly to a particular
7588 MIPS instruction. Please refer to the architecture specification
7589 for details on what each instruction does.
7592 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7593 Pair lower lower (@code{pll.ps}).
7595 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7596 Pair upper lower (@code{pul.ps}).
7598 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7599 Pair lower upper (@code{plu.ps}).
7601 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7602 Pair upper upper (@code{puu.ps}).
7604 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7605 Convert pair to paired single (@code{cvt.ps.s}).
7607 @item float __builtin_mips_cvt_s_pl (v2sf)
7608 Convert pair lower to single (@code{cvt.s.pl}).
7610 @item float __builtin_mips_cvt_s_pu (v2sf)
7611 Convert pair upper to single (@code{cvt.s.pu}).
7613 @item v2sf __builtin_mips_abs_ps (v2sf)
7614 Absolute value (@code{abs.ps}).
7616 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7617 Align variable (@code{alnv.ps}).
7619 @emph{Note:} The value of the third parameter must be 0 or 4
7620 modulo 8, otherwise the result will be unpredictable. Please read the
7621 instruction description for details.
7624 The following multi-instruction functions are also available.
7625 In each case, @var{cond} can be any of the 16 floating-point conditions:
7626 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7627 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7628 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7631 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7632 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7633 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7634 @code{movt.ps}/@code{movf.ps}).
7636 The @code{movt} functions return the value @var{x} computed by:
7639 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7640 mov.ps @var{x},@var{c}
7641 movt.ps @var{x},@var{d},@var{cc}
7644 The @code{movf} functions are similar but use @code{movf.ps} instead
7647 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7648 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7649 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7650 @code{bc1t}/@code{bc1f}).
7652 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7653 and return either the upper or lower half of the result. For example:
7657 if (__builtin_mips_upper_c_eq_ps (a, b))
7658 upper_halves_are_equal ();
7660 upper_halves_are_unequal ();
7662 if (__builtin_mips_lower_c_eq_ps (a, b))
7663 lower_halves_are_equal ();
7665 lower_halves_are_unequal ();
7669 @node MIPS-3D Built-in Functions
7670 @subsubsection MIPS-3D Built-in Functions
7672 The MIPS-3D Application-Specific Extension (ASE) includes additional
7673 paired-single instructions that are designed to improve the performance
7674 of 3D graphics operations. Support for these instructions is controlled
7675 by the @option{-mips3d} command-line option.
7677 The functions listed below map directly to a particular MIPS-3D
7678 instruction. Please refer to the architecture specification for
7679 more details on what each instruction does.
7682 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7683 Reduction add (@code{addr.ps}).
7685 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7686 Reduction multiply (@code{mulr.ps}).
7688 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7689 Convert paired single to paired word (@code{cvt.pw.ps}).
7691 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7692 Convert paired word to paired single (@code{cvt.ps.pw}).
7694 @item float __builtin_mips_recip1_s (float)
7695 @itemx double __builtin_mips_recip1_d (double)
7696 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7697 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7699 @item float __builtin_mips_recip2_s (float, float)
7700 @itemx double __builtin_mips_recip2_d (double, double)
7701 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7702 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7704 @item float __builtin_mips_rsqrt1_s (float)
7705 @itemx double __builtin_mips_rsqrt1_d (double)
7706 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7707 Reduced precision reciprocal square root (sequence step 1)
7708 (@code{rsqrt1.@var{fmt}}).
7710 @item float __builtin_mips_rsqrt2_s (float, float)
7711 @itemx double __builtin_mips_rsqrt2_d (double, double)
7712 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7713 Reduced precision reciprocal square root (sequence step 2)
7714 (@code{rsqrt2.@var{fmt}}).
7717 The following multi-instruction functions are also available.
7718 In each case, @var{cond} can be any of the 16 floating-point conditions:
7719 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7720 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7721 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7724 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7725 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7726 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7727 @code{bc1t}/@code{bc1f}).
7729 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7730 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7735 if (__builtin_mips_cabs_eq_s (a, b))
7741 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7742 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7743 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7744 @code{bc1t}/@code{bc1f}).
7746 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7747 and return either the upper or lower half of the result. For example:
7751 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7752 upper_halves_are_equal ();
7754 upper_halves_are_unequal ();
7756 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7757 lower_halves_are_equal ();
7759 lower_halves_are_unequal ();
7762 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7763 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7764 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7765 @code{movt.ps}/@code{movf.ps}).
7767 The @code{movt} functions return the value @var{x} computed by:
7770 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7771 mov.ps @var{x},@var{c}
7772 movt.ps @var{x},@var{d},@var{cc}
7775 The @code{movf} functions are similar but use @code{movf.ps} instead
7778 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7779 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7780 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7781 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7782 Comparison of two paired-single values
7783 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7784 @code{bc1any2t}/@code{bc1any2f}).
7786 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7787 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7788 result is true and the @code{all} forms return true if both results are true.
7793 if (__builtin_mips_any_c_eq_ps (a, b))
7798 if (__builtin_mips_all_c_eq_ps (a, b))
7804 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7805 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7806 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7807 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7808 Comparison of four paired-single values
7809 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7810 @code{bc1any4t}/@code{bc1any4f}).
7812 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7813 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7814 The @code{any} forms return true if any of the four results are true
7815 and the @code{all} forms return true if all four results are true.
7820 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7825 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7832 @node PowerPC AltiVec Built-in Functions
7833 @subsection PowerPC AltiVec Built-in Functions
7835 GCC provides an interface for the PowerPC family of processors to access
7836 the AltiVec operations described in Motorola's AltiVec Programming
7837 Interface Manual. The interface is made available by including
7838 @code{<altivec.h>} and using @option{-maltivec} and
7839 @option{-mabi=altivec}. The interface supports the following vector
7843 vector unsigned char
7847 vector unsigned short
7858 GCC's implementation of the high-level language interface available from
7859 C and C++ code differs from Motorola's documentation in several ways.
7864 A vector constant is a list of constant expressions within curly braces.
7867 A vector initializer requires no cast if the vector constant is of the
7868 same type as the variable it is initializing.
7871 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7872 vector type is the default signedness of the base type. The default
7873 varies depending on the operating system, so a portable program should
7874 always specify the signedness.
7877 Compiling with @option{-maltivec} adds keywords @code{__vector},
7878 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7879 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7883 GCC allows using a @code{typedef} name as the type specifier for a
7887 For C, overloaded functions are implemented with macros so the following
7891 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7894 Since @code{vec_add} is a macro, the vector constant in the example
7895 is treated as four separate arguments. Wrap the entire argument in
7896 parentheses for this to work.
7899 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
7900 Internally, GCC uses built-in functions to achieve the functionality in
7901 the aforementioned header file, but they are not supported and are
7902 subject to change without notice.
7904 The following interfaces are supported for the generic and specific
7905 AltiVec operations and the AltiVec predicates. In cases where there
7906 is a direct mapping between generic and specific operations, only the
7907 generic names are shown here, although the specific operations can also
7910 Arguments that are documented as @code{const int} require literal
7911 integral values within the range required for that operation.
7914 vector signed char vec_abs (vector signed char);
7915 vector signed short vec_abs (vector signed short);
7916 vector signed int vec_abs (vector signed int);
7917 vector float vec_abs (vector float);
7919 vector signed char vec_abss (vector signed char);
7920 vector signed short vec_abss (vector signed short);
7921 vector signed int vec_abss (vector signed int);
7923 vector signed char vec_add (vector bool char, vector signed char);
7924 vector signed char vec_add (vector signed char, vector bool char);
7925 vector signed char vec_add (vector signed char, vector signed char);
7926 vector unsigned char vec_add (vector bool char, vector unsigned char);
7927 vector unsigned char vec_add (vector unsigned char, vector bool char);
7928 vector unsigned char vec_add (vector unsigned char,
7929 vector unsigned char);
7930 vector signed short vec_add (vector bool short, vector signed short);
7931 vector signed short vec_add (vector signed short, vector bool short);
7932 vector signed short vec_add (vector signed short, vector signed short);
7933 vector unsigned short vec_add (vector bool short,
7934 vector unsigned short);
7935 vector unsigned short vec_add (vector unsigned short,
7937 vector unsigned short vec_add (vector unsigned short,
7938 vector unsigned short);
7939 vector signed int vec_add (vector bool int, vector signed int);
7940 vector signed int vec_add (vector signed int, vector bool int);
7941 vector signed int vec_add (vector signed int, vector signed int);
7942 vector unsigned int vec_add (vector bool int, vector unsigned int);
7943 vector unsigned int vec_add (vector unsigned int, vector bool int);
7944 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
7945 vector float vec_add (vector float, vector float);
7947 vector float vec_vaddfp (vector float, vector float);
7949 vector signed int vec_vadduwm (vector bool int, vector signed int);
7950 vector signed int vec_vadduwm (vector signed int, vector bool int);
7951 vector signed int vec_vadduwm (vector signed int, vector signed int);
7952 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
7953 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
7954 vector unsigned int vec_vadduwm (vector unsigned int,
7955 vector unsigned int);
7957 vector signed short vec_vadduhm (vector bool short,
7958 vector signed short);
7959 vector signed short vec_vadduhm (vector signed short,
7961 vector signed short vec_vadduhm (vector signed short,
7962 vector signed short);
7963 vector unsigned short vec_vadduhm (vector bool short,
7964 vector unsigned short);
7965 vector unsigned short vec_vadduhm (vector unsigned short,
7967 vector unsigned short vec_vadduhm (vector unsigned short,
7968 vector unsigned short);
7970 vector signed char vec_vaddubm (vector bool char, vector signed char);
7971 vector signed char vec_vaddubm (vector signed char, vector bool char);
7972 vector signed char vec_vaddubm (vector signed char, vector signed char);
7973 vector unsigned char vec_vaddubm (vector bool char,
7974 vector unsigned char);
7975 vector unsigned char vec_vaddubm (vector unsigned char,
7977 vector unsigned char vec_vaddubm (vector unsigned char,
7978 vector unsigned char);
7980 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
7982 vector unsigned char vec_adds (vector bool char, vector unsigned char);
7983 vector unsigned char vec_adds (vector unsigned char, vector bool char);
7984 vector unsigned char vec_adds (vector unsigned char,
7985 vector unsigned char);
7986 vector signed char vec_adds (vector bool char, vector signed char);
7987 vector signed char vec_adds (vector signed char, vector bool char);
7988 vector signed char vec_adds (vector signed char, vector signed char);
7989 vector unsigned short vec_adds (vector bool short,
7990 vector unsigned short);
7991 vector unsigned short vec_adds (vector unsigned short,
7993 vector unsigned short vec_adds (vector unsigned short,
7994 vector unsigned short);
7995 vector signed short vec_adds (vector bool short, vector signed short);
7996 vector signed short vec_adds (vector signed short, vector bool short);
7997 vector signed short vec_adds (vector signed short, vector signed short);
7998 vector unsigned int vec_adds (vector bool int, vector unsigned int);
7999 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8000 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8001 vector signed int vec_adds (vector bool int, vector signed int);
8002 vector signed int vec_adds (vector signed int, vector bool int);
8003 vector signed int vec_adds (vector signed int, vector signed int);
8005 vector signed int vec_vaddsws (vector bool int, vector signed int);
8006 vector signed int vec_vaddsws (vector signed int, vector bool int);
8007 vector signed int vec_vaddsws (vector signed int, vector signed int);
8009 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8010 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8011 vector unsigned int vec_vadduws (vector unsigned int,
8012 vector unsigned int);
8014 vector signed short vec_vaddshs (vector bool short,
8015 vector signed short);
8016 vector signed short vec_vaddshs (vector signed short,
8018 vector signed short vec_vaddshs (vector signed short,
8019 vector signed short);
8021 vector unsigned short vec_vadduhs (vector bool short,
8022 vector unsigned short);
8023 vector unsigned short vec_vadduhs (vector unsigned short,
8025 vector unsigned short vec_vadduhs (vector unsigned short,
8026 vector unsigned short);
8028 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8029 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8030 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8032 vector unsigned char vec_vaddubs (vector bool char,
8033 vector unsigned char);
8034 vector unsigned char vec_vaddubs (vector unsigned char,
8036 vector unsigned char vec_vaddubs (vector unsigned char,
8037 vector unsigned char);
8039 vector float vec_and (vector float, vector float);
8040 vector float vec_and (vector float, vector bool int);
8041 vector float vec_and (vector bool int, vector float);
8042 vector bool int vec_and (vector bool int, vector bool int);
8043 vector signed int vec_and (vector bool int, vector signed int);
8044 vector signed int vec_and (vector signed int, vector bool int);
8045 vector signed int vec_and (vector signed int, vector signed int);
8046 vector unsigned int vec_and (vector bool int, vector unsigned int);
8047 vector unsigned int vec_and (vector unsigned int, vector bool int);
8048 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8049 vector bool short vec_and (vector bool short, vector bool short);
8050 vector signed short vec_and (vector bool short, vector signed short);
8051 vector signed short vec_and (vector signed short, vector bool short);
8052 vector signed short vec_and (vector signed short, vector signed short);
8053 vector unsigned short vec_and (vector bool short,
8054 vector unsigned short);
8055 vector unsigned short vec_and (vector unsigned short,
8057 vector unsigned short vec_and (vector unsigned short,
8058 vector unsigned short);
8059 vector signed char vec_and (vector bool char, vector signed char);
8060 vector bool char vec_and (vector bool char, vector bool char);
8061 vector signed char vec_and (vector signed char, vector bool char);
8062 vector signed char vec_and (vector signed char, vector signed char);
8063 vector unsigned char vec_and (vector bool char, vector unsigned char);
8064 vector unsigned char vec_and (vector unsigned char, vector bool char);
8065 vector unsigned char vec_and (vector unsigned char,
8066 vector unsigned char);
8068 vector float vec_andc (vector float, vector float);
8069 vector float vec_andc (vector float, vector bool int);
8070 vector float vec_andc (vector bool int, vector float);
8071 vector bool int vec_andc (vector bool int, vector bool int);
8072 vector signed int vec_andc (vector bool int, vector signed int);
8073 vector signed int vec_andc (vector signed int, vector bool int);
8074 vector signed int vec_andc (vector signed int, vector signed int);
8075 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8076 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8077 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8078 vector bool short vec_andc (vector bool short, vector bool short);
8079 vector signed short vec_andc (vector bool short, vector signed short);
8080 vector signed short vec_andc (vector signed short, vector bool short);
8081 vector signed short vec_andc (vector signed short, vector signed short);
8082 vector unsigned short vec_andc (vector bool short,
8083 vector unsigned short);
8084 vector unsigned short vec_andc (vector unsigned short,
8086 vector unsigned short vec_andc (vector unsigned short,
8087 vector unsigned short);
8088 vector signed char vec_andc (vector bool char, vector signed char);
8089 vector bool char vec_andc (vector bool char, vector bool char);
8090 vector signed char vec_andc (vector signed char, vector bool char);
8091 vector signed char vec_andc (vector signed char, vector signed char);
8092 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8093 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8094 vector unsigned char vec_andc (vector unsigned char,
8095 vector unsigned char);
8097 vector unsigned char vec_avg (vector unsigned char,
8098 vector unsigned char);
8099 vector signed char vec_avg (vector signed char, vector signed char);
8100 vector unsigned short vec_avg (vector unsigned short,
8101 vector unsigned short);
8102 vector signed short vec_avg (vector signed short, vector signed short);
8103 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8104 vector signed int vec_avg (vector signed int, vector signed int);
8106 vector signed int vec_vavgsw (vector signed int, vector signed int);
8108 vector unsigned int vec_vavguw (vector unsigned int,
8109 vector unsigned int);
8111 vector signed short vec_vavgsh (vector signed short,
8112 vector signed short);
8114 vector unsigned short vec_vavguh (vector unsigned short,
8115 vector unsigned short);
8117 vector signed char vec_vavgsb (vector signed char, vector signed char);
8119 vector unsigned char vec_vavgub (vector unsigned char,
8120 vector unsigned char);
8122 vector float vec_ceil (vector float);
8124 vector signed int vec_cmpb (vector float, vector float);
8126 vector bool char vec_cmpeq (vector signed char, vector signed char);
8127 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8128 vector bool short vec_cmpeq (vector signed short, vector signed short);
8129 vector bool short vec_cmpeq (vector unsigned short,
8130 vector unsigned short);
8131 vector bool int vec_cmpeq (vector signed int, vector signed int);
8132 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8133 vector bool int vec_cmpeq (vector float, vector float);
8135 vector bool int vec_vcmpeqfp (vector float, vector float);
8137 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8138 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8140 vector bool short vec_vcmpequh (vector signed short,
8141 vector signed short);
8142 vector bool short vec_vcmpequh (vector unsigned short,
8143 vector unsigned short);
8145 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8146 vector bool char vec_vcmpequb (vector unsigned char,
8147 vector unsigned char);
8149 vector bool int vec_cmpge (vector float, vector float);
8151 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8152 vector bool char vec_cmpgt (vector signed char, vector signed char);
8153 vector bool short vec_cmpgt (vector unsigned short,
8154 vector unsigned short);
8155 vector bool short vec_cmpgt (vector signed short, vector signed short);
8156 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8157 vector bool int vec_cmpgt (vector signed int, vector signed int);
8158 vector bool int vec_cmpgt (vector float, vector float);
8160 vector bool int vec_vcmpgtfp (vector float, vector float);
8162 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8164 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8166 vector bool short vec_vcmpgtsh (vector signed short,
8167 vector signed short);
8169 vector bool short vec_vcmpgtuh (vector unsigned short,
8170 vector unsigned short);
8172 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8174 vector bool char vec_vcmpgtub (vector unsigned char,
8175 vector unsigned char);
8177 vector bool int vec_cmple (vector float, vector float);
8179 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8180 vector bool char vec_cmplt (vector signed char, vector signed char);
8181 vector bool short vec_cmplt (vector unsigned short,
8182 vector unsigned short);
8183 vector bool short vec_cmplt (vector signed short, vector signed short);
8184 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8185 vector bool int vec_cmplt (vector signed int, vector signed int);
8186 vector bool int vec_cmplt (vector float, vector float);
8188 vector float vec_ctf (vector unsigned int, const int);
8189 vector float vec_ctf (vector signed int, const int);
8191 vector float vec_vcfsx (vector signed int, const int);
8193 vector float vec_vcfux (vector unsigned int, const int);
8195 vector signed int vec_cts (vector float, const int);
8197 vector unsigned int vec_ctu (vector float, const int);
8199 void vec_dss (const int);
8201 void vec_dssall (void);
8203 void vec_dst (const vector unsigned char *, int, const int);
8204 void vec_dst (const vector signed char *, int, const int);
8205 void vec_dst (const vector bool char *, int, const int);
8206 void vec_dst (const vector unsigned short *, int, const int);
8207 void vec_dst (const vector signed short *, int, const int);
8208 void vec_dst (const vector bool short *, int, const int);
8209 void vec_dst (const vector pixel *, int, const int);
8210 void vec_dst (const vector unsigned int *, int, const int);
8211 void vec_dst (const vector signed int *, int, const int);
8212 void vec_dst (const vector bool int *, int, const int);
8213 void vec_dst (const vector float *, int, const int);
8214 void vec_dst (const unsigned char *, int, const int);
8215 void vec_dst (const signed char *, int, const int);
8216 void vec_dst (const unsigned short *, int, const int);
8217 void vec_dst (const short *, int, const int);
8218 void vec_dst (const unsigned int *, int, const int);
8219 void vec_dst (const int *, int, const int);
8220 void vec_dst (const unsigned long *, int, const int);
8221 void vec_dst (const long *, int, const int);
8222 void vec_dst (const float *, int, const int);
8224 void vec_dstst (const vector unsigned char *, int, const int);
8225 void vec_dstst (const vector signed char *, int, const int);
8226 void vec_dstst (const vector bool char *, int, const int);
8227 void vec_dstst (const vector unsigned short *, int, const int);
8228 void vec_dstst (const vector signed short *, int, const int);
8229 void vec_dstst (const vector bool short *, int, const int);
8230 void vec_dstst (const vector pixel *, int, const int);
8231 void vec_dstst (const vector unsigned int *, int, const int);
8232 void vec_dstst (const vector signed int *, int, const int);
8233 void vec_dstst (const vector bool int *, int, const int);
8234 void vec_dstst (const vector float *, int, const int);
8235 void vec_dstst (const unsigned char *, int, const int);
8236 void vec_dstst (const signed char *, int, const int);
8237 void vec_dstst (const unsigned short *, int, const int);
8238 void vec_dstst (const short *, int, const int);
8239 void vec_dstst (const unsigned int *, int, const int);
8240 void vec_dstst (const int *, int, const int);
8241 void vec_dstst (const unsigned long *, int, const int);
8242 void vec_dstst (const long *, int, const int);
8243 void vec_dstst (const float *, int, const int);
8245 void vec_dststt (const vector unsigned char *, int, const int);
8246 void vec_dststt (const vector signed char *, int, const int);
8247 void vec_dststt (const vector bool char *, int, const int);
8248 void vec_dststt (const vector unsigned short *, int, const int);
8249 void vec_dststt (const vector signed short *, int, const int);
8250 void vec_dststt (const vector bool short *, int, const int);
8251 void vec_dststt (const vector pixel *, int, const int);
8252 void vec_dststt (const vector unsigned int *, int, const int);
8253 void vec_dststt (const vector signed int *, int, const int);
8254 void vec_dststt (const vector bool int *, int, const int);
8255 void vec_dststt (const vector float *, int, const int);
8256 void vec_dststt (const unsigned char *, int, const int);
8257 void vec_dststt (const signed char *, int, const int);
8258 void vec_dststt (const unsigned short *, int, const int);
8259 void vec_dststt (const short *, int, const int);
8260 void vec_dststt (const unsigned int *, int, const int);
8261 void vec_dststt (const int *, int, const int);
8262 void vec_dststt (const unsigned long *, int, const int);
8263 void vec_dststt (const long *, int, const int);
8264 void vec_dststt (const float *, int, const int);
8266 void vec_dstt (const vector unsigned char *, int, const int);
8267 void vec_dstt (const vector signed char *, int, const int);
8268 void vec_dstt (const vector bool char *, int, const int);
8269 void vec_dstt (const vector unsigned short *, int, const int);
8270 void vec_dstt (const vector signed short *, int, const int);
8271 void vec_dstt (const vector bool short *, int, const int);
8272 void vec_dstt (const vector pixel *, int, const int);
8273 void vec_dstt (const vector unsigned int *, int, const int);
8274 void vec_dstt (const vector signed int *, int, const int);
8275 void vec_dstt (const vector bool int *, int, const int);
8276 void vec_dstt (const vector float *, int, const int);
8277 void vec_dstt (const unsigned char *, int, const int);
8278 void vec_dstt (const signed char *, int, const int);
8279 void vec_dstt (const unsigned short *, int, const int);
8280 void vec_dstt (const short *, int, const int);
8281 void vec_dstt (const unsigned int *, int, const int);
8282 void vec_dstt (const int *, int, const int);
8283 void vec_dstt (const unsigned long *, int, const int);
8284 void vec_dstt (const long *, int, const int);
8285 void vec_dstt (const float *, int, const int);
8287 vector float vec_expte (vector float);
8289 vector float vec_floor (vector float);
8291 vector float vec_ld (int, const vector float *);
8292 vector float vec_ld (int, const float *);
8293 vector bool int vec_ld (int, const vector bool int *);
8294 vector signed int vec_ld (int, const vector signed int *);
8295 vector signed int vec_ld (int, const int *);
8296 vector signed int vec_ld (int, const long *);
8297 vector unsigned int vec_ld (int, const vector unsigned int *);
8298 vector unsigned int vec_ld (int, const unsigned int *);
8299 vector unsigned int vec_ld (int, const unsigned long *);
8300 vector bool short vec_ld (int, const vector bool short *);
8301 vector pixel vec_ld (int, const vector pixel *);
8302 vector signed short vec_ld (int, const vector signed short *);
8303 vector signed short vec_ld (int, const short *);
8304 vector unsigned short vec_ld (int, const vector unsigned short *);
8305 vector unsigned short vec_ld (int, const unsigned short *);
8306 vector bool char vec_ld (int, const vector bool char *);
8307 vector signed char vec_ld (int, const vector signed char *);
8308 vector signed char vec_ld (int, const signed char *);
8309 vector unsigned char vec_ld (int, const vector unsigned char *);
8310 vector unsigned char vec_ld (int, const unsigned char *);
8312 vector signed char vec_lde (int, const signed char *);
8313 vector unsigned char vec_lde (int, const unsigned char *);
8314 vector signed short vec_lde (int, const short *);
8315 vector unsigned short vec_lde (int, const unsigned short *);
8316 vector float vec_lde (int, const float *);
8317 vector signed int vec_lde (int, const int *);
8318 vector unsigned int vec_lde (int, const unsigned int *);
8319 vector signed int vec_lde (int, const long *);
8320 vector unsigned int vec_lde (int, const unsigned long *);
8322 vector float vec_lvewx (int, float *);
8323 vector signed int vec_lvewx (int, int *);
8324 vector unsigned int vec_lvewx (int, unsigned int *);
8325 vector signed int vec_lvewx (int, long *);
8326 vector unsigned int vec_lvewx (int, unsigned long *);
8328 vector signed short vec_lvehx (int, short *);
8329 vector unsigned short vec_lvehx (int, unsigned short *);
8331 vector signed char vec_lvebx (int, char *);
8332 vector unsigned char vec_lvebx (int, unsigned char *);
8334 vector float vec_ldl (int, const vector float *);
8335 vector float vec_ldl (int, const float *);
8336 vector bool int vec_ldl (int, const vector bool int *);
8337 vector signed int vec_ldl (int, const vector signed int *);
8338 vector signed int vec_ldl (int, const int *);
8339 vector signed int vec_ldl (int, const long *);
8340 vector unsigned int vec_ldl (int, const vector unsigned int *);
8341 vector unsigned int vec_ldl (int, const unsigned int *);
8342 vector unsigned int vec_ldl (int, const unsigned long *);
8343 vector bool short vec_ldl (int, const vector bool short *);
8344 vector pixel vec_ldl (int, const vector pixel *);
8345 vector signed short vec_ldl (int, const vector signed short *);
8346 vector signed short vec_ldl (int, const short *);
8347 vector unsigned short vec_ldl (int, const vector unsigned short *);
8348 vector unsigned short vec_ldl (int, const unsigned short *);
8349 vector bool char vec_ldl (int, const vector bool char *);
8350 vector signed char vec_ldl (int, const vector signed char *);
8351 vector signed char vec_ldl (int, const signed char *);
8352 vector unsigned char vec_ldl (int, const vector unsigned char *);
8353 vector unsigned char vec_ldl (int, const unsigned char *);
8355 vector float vec_loge (vector float);
8357 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8358 vector unsigned char vec_lvsl (int, const volatile signed char *);
8359 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8360 vector unsigned char vec_lvsl (int, const volatile short *);
8361 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8362 vector unsigned char vec_lvsl (int, const volatile int *);
8363 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8364 vector unsigned char vec_lvsl (int, const volatile long *);
8365 vector unsigned char vec_lvsl (int, const volatile float *);
8367 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8368 vector unsigned char vec_lvsr (int, const volatile signed char *);
8369 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8370 vector unsigned char vec_lvsr (int, const volatile short *);
8371 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8372 vector unsigned char vec_lvsr (int, const volatile int *);
8373 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8374 vector unsigned char vec_lvsr (int, const volatile long *);
8375 vector unsigned char vec_lvsr (int, const volatile float *);
8377 vector float vec_madd (vector float, vector float, vector float);
8379 vector signed short vec_madds (vector signed short,
8380 vector signed short,
8381 vector signed short);
8383 vector unsigned char vec_max (vector bool char, vector unsigned char);
8384 vector unsigned char vec_max (vector unsigned char, vector bool char);
8385 vector unsigned char vec_max (vector unsigned char,
8386 vector unsigned char);
8387 vector signed char vec_max (vector bool char, vector signed char);
8388 vector signed char vec_max (vector signed char, vector bool char);
8389 vector signed char vec_max (vector signed char, vector signed char);
8390 vector unsigned short vec_max (vector bool short,
8391 vector unsigned short);
8392 vector unsigned short vec_max (vector unsigned short,
8394 vector unsigned short vec_max (vector unsigned short,
8395 vector unsigned short);
8396 vector signed short vec_max (vector bool short, vector signed short);
8397 vector signed short vec_max (vector signed short, vector bool short);
8398 vector signed short vec_max (vector signed short, vector signed short);
8399 vector unsigned int vec_max (vector bool int, vector unsigned int);
8400 vector unsigned int vec_max (vector unsigned int, vector bool int);
8401 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8402 vector signed int vec_max (vector bool int, vector signed int);
8403 vector signed int vec_max (vector signed int, vector bool int);
8404 vector signed int vec_max (vector signed int, vector signed int);
8405 vector float vec_max (vector float, vector float);
8407 vector float vec_vmaxfp (vector float, vector float);
8409 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8410 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8411 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8413 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8414 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8415 vector unsigned int vec_vmaxuw (vector unsigned int,
8416 vector unsigned int);
8418 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8419 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8420 vector signed short vec_vmaxsh (vector signed short,
8421 vector signed short);
8423 vector unsigned short vec_vmaxuh (vector bool short,
8424 vector unsigned short);
8425 vector unsigned short vec_vmaxuh (vector unsigned short,
8427 vector unsigned short vec_vmaxuh (vector unsigned short,
8428 vector unsigned short);
8430 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8431 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8432 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8434 vector unsigned char vec_vmaxub (vector bool char,
8435 vector unsigned char);
8436 vector unsigned char vec_vmaxub (vector unsigned char,
8438 vector unsigned char vec_vmaxub (vector unsigned char,
8439 vector unsigned char);
8441 vector bool char vec_mergeh (vector bool char, vector bool char);
8442 vector signed char vec_mergeh (vector signed char, vector signed char);
8443 vector unsigned char vec_mergeh (vector unsigned char,
8444 vector unsigned char);
8445 vector bool short vec_mergeh (vector bool short, vector bool short);
8446 vector pixel vec_mergeh (vector pixel, vector pixel);
8447 vector signed short vec_mergeh (vector signed short,
8448 vector signed short);
8449 vector unsigned short vec_mergeh (vector unsigned short,
8450 vector unsigned short);
8451 vector float vec_mergeh (vector float, vector float);
8452 vector bool int vec_mergeh (vector bool int, vector bool int);
8453 vector signed int vec_mergeh (vector signed int, vector signed int);
8454 vector unsigned int vec_mergeh (vector unsigned int,
8455 vector unsigned int);
8457 vector float vec_vmrghw (vector float, vector float);
8458 vector bool int vec_vmrghw (vector bool int, vector bool int);
8459 vector signed int vec_vmrghw (vector signed int, vector signed int);
8460 vector unsigned int vec_vmrghw (vector unsigned int,
8461 vector unsigned int);
8463 vector bool short vec_vmrghh (vector bool short, vector bool short);
8464 vector signed short vec_vmrghh (vector signed short,
8465 vector signed short);
8466 vector unsigned short vec_vmrghh (vector unsigned short,
8467 vector unsigned short);
8468 vector pixel vec_vmrghh (vector pixel, vector pixel);
8470 vector bool char vec_vmrghb (vector bool char, vector bool char);
8471 vector signed char vec_vmrghb (vector signed char, vector signed char);
8472 vector unsigned char vec_vmrghb (vector unsigned char,
8473 vector unsigned char);
8475 vector bool char vec_mergel (vector bool char, vector bool char);
8476 vector signed char vec_mergel (vector signed char, vector signed char);
8477 vector unsigned char vec_mergel (vector unsigned char,
8478 vector unsigned char);
8479 vector bool short vec_mergel (vector bool short, vector bool short);
8480 vector pixel vec_mergel (vector pixel, vector pixel);
8481 vector signed short vec_mergel (vector signed short,
8482 vector signed short);
8483 vector unsigned short vec_mergel (vector unsigned short,
8484 vector unsigned short);
8485 vector float vec_mergel (vector float, vector float);
8486 vector bool int vec_mergel (vector bool int, vector bool int);
8487 vector signed int vec_mergel (vector signed int, vector signed int);
8488 vector unsigned int vec_mergel (vector unsigned int,
8489 vector unsigned int);
8491 vector float vec_vmrglw (vector float, vector float);
8492 vector signed int vec_vmrglw (vector signed int, vector signed int);
8493 vector unsigned int vec_vmrglw (vector unsigned int,
8494 vector unsigned int);
8495 vector bool int vec_vmrglw (vector bool int, vector bool int);
8497 vector bool short vec_vmrglh (vector bool short, vector bool short);
8498 vector signed short vec_vmrglh (vector signed short,
8499 vector signed short);
8500 vector unsigned short vec_vmrglh (vector unsigned short,
8501 vector unsigned short);
8502 vector pixel vec_vmrglh (vector pixel, vector pixel);
8504 vector bool char vec_vmrglb (vector bool char, vector bool char);
8505 vector signed char vec_vmrglb (vector signed char, vector signed char);
8506 vector unsigned char vec_vmrglb (vector unsigned char,
8507 vector unsigned char);
8509 vector unsigned short vec_mfvscr (void);
8511 vector unsigned char vec_min (vector bool char, vector unsigned char);
8512 vector unsigned char vec_min (vector unsigned char, vector bool char);
8513 vector unsigned char vec_min (vector unsigned char,
8514 vector unsigned char);
8515 vector signed char vec_min (vector bool char, vector signed char);
8516 vector signed char vec_min (vector signed char, vector bool char);
8517 vector signed char vec_min (vector signed char, vector signed char);
8518 vector unsigned short vec_min (vector bool short,
8519 vector unsigned short);
8520 vector unsigned short vec_min (vector unsigned short,
8522 vector unsigned short vec_min (vector unsigned short,
8523 vector unsigned short);
8524 vector signed short vec_min (vector bool short, vector signed short);
8525 vector signed short vec_min (vector signed short, vector bool short);
8526 vector signed short vec_min (vector signed short, vector signed short);
8527 vector unsigned int vec_min (vector bool int, vector unsigned int);
8528 vector unsigned int vec_min (vector unsigned int, vector bool int);
8529 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8530 vector signed int vec_min (vector bool int, vector signed int);
8531 vector signed int vec_min (vector signed int, vector bool int);
8532 vector signed int vec_min (vector signed int, vector signed int);
8533 vector float vec_min (vector float, vector float);
8535 vector float vec_vminfp (vector float, vector float);
8537 vector signed int vec_vminsw (vector bool int, vector signed int);
8538 vector signed int vec_vminsw (vector signed int, vector bool int);
8539 vector signed int vec_vminsw (vector signed int, vector signed int);
8541 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8542 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8543 vector unsigned int vec_vminuw (vector unsigned int,
8544 vector unsigned int);
8546 vector signed short vec_vminsh (vector bool short, vector signed short);
8547 vector signed short vec_vminsh (vector signed short, vector bool short);
8548 vector signed short vec_vminsh (vector signed short,
8549 vector signed short);
8551 vector unsigned short vec_vminuh (vector bool short,
8552 vector unsigned short);
8553 vector unsigned short vec_vminuh (vector unsigned short,
8555 vector unsigned short vec_vminuh (vector unsigned short,
8556 vector unsigned short);
8558 vector signed char vec_vminsb (vector bool char, vector signed char);
8559 vector signed char vec_vminsb (vector signed char, vector bool char);
8560 vector signed char vec_vminsb (vector signed char, vector signed char);
8562 vector unsigned char vec_vminub (vector bool char,
8563 vector unsigned char);
8564 vector unsigned char vec_vminub (vector unsigned char,
8566 vector unsigned char vec_vminub (vector unsigned char,
8567 vector unsigned char);
8569 vector signed short vec_mladd (vector signed short,
8570 vector signed short,
8571 vector signed short);
8572 vector signed short vec_mladd (vector signed short,
8573 vector unsigned short,
8574 vector unsigned short);
8575 vector signed short vec_mladd (vector unsigned short,
8576 vector signed short,
8577 vector signed short);
8578 vector unsigned short vec_mladd (vector unsigned short,
8579 vector unsigned short,
8580 vector unsigned short);
8582 vector signed short vec_mradds (vector signed short,
8583 vector signed short,
8584 vector signed short);
8586 vector unsigned int vec_msum (vector unsigned char,
8587 vector unsigned char,
8588 vector unsigned int);
8589 vector signed int vec_msum (vector signed char,
8590 vector unsigned char,
8592 vector unsigned int vec_msum (vector unsigned short,
8593 vector unsigned short,
8594 vector unsigned int);
8595 vector signed int vec_msum (vector signed short,
8596 vector signed short,
8599 vector signed int vec_vmsumshm (vector signed short,
8600 vector signed short,
8603 vector unsigned int vec_vmsumuhm (vector unsigned short,
8604 vector unsigned short,
8605 vector unsigned int);
8607 vector signed int vec_vmsummbm (vector signed char,
8608 vector unsigned char,
8611 vector unsigned int vec_vmsumubm (vector unsigned char,
8612 vector unsigned char,
8613 vector unsigned int);
8615 vector unsigned int vec_msums (vector unsigned short,
8616 vector unsigned short,
8617 vector unsigned int);
8618 vector signed int vec_msums (vector signed short,
8619 vector signed short,
8622 vector signed int vec_vmsumshs (vector signed short,
8623 vector signed short,
8626 vector unsigned int vec_vmsumuhs (vector unsigned short,
8627 vector unsigned short,
8628 vector unsigned int);
8630 void vec_mtvscr (vector signed int);
8631 void vec_mtvscr (vector unsigned int);
8632 void vec_mtvscr (vector bool int);
8633 void vec_mtvscr (vector signed short);
8634 void vec_mtvscr (vector unsigned short);
8635 void vec_mtvscr (vector bool short);
8636 void vec_mtvscr (vector pixel);
8637 void vec_mtvscr (vector signed char);
8638 void vec_mtvscr (vector unsigned char);
8639 void vec_mtvscr (vector bool char);
8641 vector unsigned short vec_mule (vector unsigned char,
8642 vector unsigned char);
8643 vector signed short vec_mule (vector signed char,
8644 vector signed char);
8645 vector unsigned int vec_mule (vector unsigned short,
8646 vector unsigned short);
8647 vector signed int vec_mule (vector signed short, vector signed short);
8649 vector signed int vec_vmulesh (vector signed short,
8650 vector signed short);
8652 vector unsigned int vec_vmuleuh (vector unsigned short,
8653 vector unsigned short);
8655 vector signed short vec_vmulesb (vector signed char,
8656 vector signed char);
8658 vector unsigned short vec_vmuleub (vector unsigned char,
8659 vector unsigned char);
8661 vector unsigned short vec_mulo (vector unsigned char,
8662 vector unsigned char);
8663 vector signed short vec_mulo (vector signed char, vector signed char);
8664 vector unsigned int vec_mulo (vector unsigned short,
8665 vector unsigned short);
8666 vector signed int vec_mulo (vector signed short, vector signed short);
8668 vector signed int vec_vmulosh (vector signed short,
8669 vector signed short);
8671 vector unsigned int vec_vmulouh (vector unsigned short,
8672 vector unsigned short);
8674 vector signed short vec_vmulosb (vector signed char,
8675 vector signed char);
8677 vector unsigned short vec_vmuloub (vector unsigned char,
8678 vector unsigned char);
8680 vector float vec_nmsub (vector float, vector float, vector float);
8682 vector float vec_nor (vector float, vector float);
8683 vector signed int vec_nor (vector signed int, vector signed int);
8684 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8685 vector bool int vec_nor (vector bool int, vector bool int);
8686 vector signed short vec_nor (vector signed short, vector signed short);
8687 vector unsigned short vec_nor (vector unsigned short,
8688 vector unsigned short);
8689 vector bool short vec_nor (vector bool short, vector bool short);
8690 vector signed char vec_nor (vector signed char, vector signed char);
8691 vector unsigned char vec_nor (vector unsigned char,
8692 vector unsigned char);
8693 vector bool char vec_nor (vector bool char, vector bool char);
8695 vector float vec_or (vector float, vector float);
8696 vector float vec_or (vector float, vector bool int);
8697 vector float vec_or (vector bool int, vector float);
8698 vector bool int vec_or (vector bool int, vector bool int);
8699 vector signed int vec_or (vector bool int, vector signed int);
8700 vector signed int vec_or (vector signed int, vector bool int);
8701 vector signed int vec_or (vector signed int, vector signed int);
8702 vector unsigned int vec_or (vector bool int, vector unsigned int);
8703 vector unsigned int vec_or (vector unsigned int, vector bool int);
8704 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8705 vector bool short vec_or (vector bool short, vector bool short);
8706 vector signed short vec_or (vector bool short, vector signed short);
8707 vector signed short vec_or (vector signed short, vector bool short);
8708 vector signed short vec_or (vector signed short, vector signed short);
8709 vector unsigned short vec_or (vector bool short, vector unsigned short);
8710 vector unsigned short vec_or (vector unsigned short, vector bool short);
8711 vector unsigned short vec_or (vector unsigned short,
8712 vector unsigned short);
8713 vector signed char vec_or (vector bool char, vector signed char);
8714 vector bool char vec_or (vector bool char, vector bool char);
8715 vector signed char vec_or (vector signed char, vector bool char);
8716 vector signed char vec_or (vector signed char, vector signed char);
8717 vector unsigned char vec_or (vector bool char, vector unsigned char);
8718 vector unsigned char vec_or (vector unsigned char, vector bool char);
8719 vector unsigned char vec_or (vector unsigned char,
8720 vector unsigned char);
8722 vector signed char vec_pack (vector signed short, vector signed short);
8723 vector unsigned char vec_pack (vector unsigned short,
8724 vector unsigned short);
8725 vector bool char vec_pack (vector bool short, vector bool short);
8726 vector signed short vec_pack (vector signed int, vector signed int);
8727 vector unsigned short vec_pack (vector unsigned int,
8728 vector unsigned int);
8729 vector bool short vec_pack (vector bool int, vector bool int);
8731 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8732 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8733 vector unsigned short vec_vpkuwum (vector unsigned int,
8734 vector unsigned int);
8736 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8737 vector signed char vec_vpkuhum (vector signed short,
8738 vector signed short);
8739 vector unsigned char vec_vpkuhum (vector unsigned short,
8740 vector unsigned short);
8742 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8744 vector unsigned char vec_packs (vector unsigned short,
8745 vector unsigned short);
8746 vector signed char vec_packs (vector signed short, vector signed short);
8747 vector unsigned short vec_packs (vector unsigned int,
8748 vector unsigned int);
8749 vector signed short vec_packs (vector signed int, vector signed int);
8751 vector signed short vec_vpkswss (vector signed int, vector signed int);
8753 vector unsigned short vec_vpkuwus (vector unsigned int,
8754 vector unsigned int);
8756 vector signed char vec_vpkshss (vector signed short,
8757 vector signed short);
8759 vector unsigned char vec_vpkuhus (vector unsigned short,
8760 vector unsigned short);
8762 vector unsigned char vec_packsu (vector unsigned short,
8763 vector unsigned short);
8764 vector unsigned char vec_packsu (vector signed short,
8765 vector signed short);
8766 vector unsigned short vec_packsu (vector unsigned int,
8767 vector unsigned int);
8768 vector unsigned short vec_packsu (vector signed int, vector signed int);
8770 vector unsigned short vec_vpkswus (vector signed int,
8773 vector unsigned char vec_vpkshus (vector signed short,
8774 vector signed short);
8776 vector float vec_perm (vector float,
8778 vector unsigned char);
8779 vector signed int vec_perm (vector signed int,
8781 vector unsigned char);
8782 vector unsigned int vec_perm (vector unsigned int,
8783 vector unsigned int,
8784 vector unsigned char);
8785 vector bool int vec_perm (vector bool int,
8787 vector unsigned char);
8788 vector signed short vec_perm (vector signed short,
8789 vector signed short,
8790 vector unsigned char);
8791 vector unsigned short vec_perm (vector unsigned short,
8792 vector unsigned short,
8793 vector unsigned char);
8794 vector bool short vec_perm (vector bool short,
8796 vector unsigned char);
8797 vector pixel vec_perm (vector pixel,
8799 vector unsigned char);
8800 vector signed char vec_perm (vector signed char,
8802 vector unsigned char);
8803 vector unsigned char vec_perm (vector unsigned char,
8804 vector unsigned char,
8805 vector unsigned char);
8806 vector bool char vec_perm (vector bool char,
8808 vector unsigned char);
8810 vector float vec_re (vector float);
8812 vector signed char vec_rl (vector signed char,
8813 vector unsigned char);
8814 vector unsigned char vec_rl (vector unsigned char,
8815 vector unsigned char);
8816 vector signed short vec_rl (vector signed short, vector unsigned short);
8817 vector unsigned short vec_rl (vector unsigned short,
8818 vector unsigned short);
8819 vector signed int vec_rl (vector signed int, vector unsigned int);
8820 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8822 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8823 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8825 vector signed short vec_vrlh (vector signed short,
8826 vector unsigned short);
8827 vector unsigned short vec_vrlh (vector unsigned short,
8828 vector unsigned short);
8830 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8831 vector unsigned char vec_vrlb (vector unsigned char,
8832 vector unsigned char);
8834 vector float vec_round (vector float);
8836 vector float vec_rsqrte (vector float);
8838 vector float vec_sel (vector float, vector float, vector bool int);
8839 vector float vec_sel (vector float, vector float, vector unsigned int);
8840 vector signed int vec_sel (vector signed int,
8843 vector signed int vec_sel (vector signed int,
8845 vector unsigned int);
8846 vector unsigned int vec_sel (vector unsigned int,
8847 vector unsigned int,
8849 vector unsigned int vec_sel (vector unsigned int,
8850 vector unsigned int,
8851 vector unsigned int);
8852 vector bool int vec_sel (vector bool int,
8855 vector bool int vec_sel (vector bool int,
8857 vector unsigned int);
8858 vector signed short vec_sel (vector signed short,
8859 vector signed short,
8861 vector signed short vec_sel (vector signed short,
8862 vector signed short,
8863 vector unsigned short);
8864 vector unsigned short vec_sel (vector unsigned short,
8865 vector unsigned short,
8867 vector unsigned short vec_sel (vector unsigned short,
8868 vector unsigned short,
8869 vector unsigned short);
8870 vector bool short vec_sel (vector bool short,
8873 vector bool short vec_sel (vector bool short,
8875 vector unsigned short);
8876 vector signed char vec_sel (vector signed char,
8879 vector signed char vec_sel (vector signed char,
8881 vector unsigned char);
8882 vector unsigned char vec_sel (vector unsigned char,
8883 vector unsigned char,
8885 vector unsigned char vec_sel (vector unsigned char,
8886 vector unsigned char,
8887 vector unsigned char);
8888 vector bool char vec_sel (vector bool char,
8891 vector bool char vec_sel (vector bool char,
8893 vector unsigned char);
8895 vector signed char vec_sl (vector signed char,
8896 vector unsigned char);
8897 vector unsigned char vec_sl (vector unsigned char,
8898 vector unsigned char);
8899 vector signed short vec_sl (vector signed short, vector unsigned short);
8900 vector unsigned short vec_sl (vector unsigned short,
8901 vector unsigned short);
8902 vector signed int vec_sl (vector signed int, vector unsigned int);
8903 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
8905 vector signed int vec_vslw (vector signed int, vector unsigned int);
8906 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
8908 vector signed short vec_vslh (vector signed short,
8909 vector unsigned short);
8910 vector unsigned short vec_vslh (vector unsigned short,
8911 vector unsigned short);
8913 vector signed char vec_vslb (vector signed char, vector unsigned char);
8914 vector unsigned char vec_vslb (vector unsigned char,
8915 vector unsigned char);
8917 vector float vec_sld (vector float, vector float, const int);
8918 vector signed int vec_sld (vector signed int,
8921 vector unsigned int vec_sld (vector unsigned int,
8922 vector unsigned int,
8924 vector bool int vec_sld (vector bool int,
8927 vector signed short vec_sld (vector signed short,
8928 vector signed short,
8930 vector unsigned short vec_sld (vector unsigned short,
8931 vector unsigned short,
8933 vector bool short vec_sld (vector bool short,
8936 vector pixel vec_sld (vector pixel,
8939 vector signed char vec_sld (vector signed char,
8942 vector unsigned char vec_sld (vector unsigned char,
8943 vector unsigned char,
8945 vector bool char vec_sld (vector bool char,
8949 vector signed int vec_sll (vector signed int,
8950 vector unsigned int);
8951 vector signed int vec_sll (vector signed int,
8952 vector unsigned short);
8953 vector signed int vec_sll (vector signed int,
8954 vector unsigned char);
8955 vector unsigned int vec_sll (vector unsigned int,
8956 vector unsigned int);
8957 vector unsigned int vec_sll (vector unsigned int,
8958 vector unsigned short);
8959 vector unsigned int vec_sll (vector unsigned int,
8960 vector unsigned char);
8961 vector bool int vec_sll (vector bool int,
8962 vector unsigned int);
8963 vector bool int vec_sll (vector bool int,
8964 vector unsigned short);
8965 vector bool int vec_sll (vector bool int,
8966 vector unsigned char);
8967 vector signed short vec_sll (vector signed short,
8968 vector unsigned int);
8969 vector signed short vec_sll (vector signed short,
8970 vector unsigned short);
8971 vector signed short vec_sll (vector signed short,
8972 vector unsigned char);
8973 vector unsigned short vec_sll (vector unsigned short,
8974 vector unsigned int);
8975 vector unsigned short vec_sll (vector unsigned short,
8976 vector unsigned short);
8977 vector unsigned short vec_sll (vector unsigned short,
8978 vector unsigned char);
8979 vector bool short vec_sll (vector bool short, vector unsigned int);
8980 vector bool short vec_sll (vector bool short, vector unsigned short);
8981 vector bool short vec_sll (vector bool short, vector unsigned char);
8982 vector pixel vec_sll (vector pixel, vector unsigned int);
8983 vector pixel vec_sll (vector pixel, vector unsigned short);
8984 vector pixel vec_sll (vector pixel, vector unsigned char);
8985 vector signed char vec_sll (vector signed char, vector unsigned int);
8986 vector signed char vec_sll (vector signed char, vector unsigned short);
8987 vector signed char vec_sll (vector signed char, vector unsigned char);
8988 vector unsigned char vec_sll (vector unsigned char,
8989 vector unsigned int);
8990 vector unsigned char vec_sll (vector unsigned char,
8991 vector unsigned short);
8992 vector unsigned char vec_sll (vector unsigned char,
8993 vector unsigned char);
8994 vector bool char vec_sll (vector bool char, vector unsigned int);
8995 vector bool char vec_sll (vector bool char, vector unsigned short);
8996 vector bool char vec_sll (vector bool char, vector unsigned char);
8998 vector float vec_slo (vector float, vector signed char);
8999 vector float vec_slo (vector float, vector unsigned char);
9000 vector signed int vec_slo (vector signed int, vector signed char);
9001 vector signed int vec_slo (vector signed int, vector unsigned char);
9002 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9003 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9004 vector signed short vec_slo (vector signed short, vector signed char);
9005 vector signed short vec_slo (vector signed short, vector unsigned char);
9006 vector unsigned short vec_slo (vector unsigned short,
9007 vector signed char);
9008 vector unsigned short vec_slo (vector unsigned short,
9009 vector unsigned char);
9010 vector pixel vec_slo (vector pixel, vector signed char);
9011 vector pixel vec_slo (vector pixel, vector unsigned char);
9012 vector signed char vec_slo (vector signed char, vector signed char);
9013 vector signed char vec_slo (vector signed char, vector unsigned char);
9014 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9015 vector unsigned char vec_slo (vector unsigned char,
9016 vector unsigned char);
9018 vector signed char vec_splat (vector signed char, const int);
9019 vector unsigned char vec_splat (vector unsigned char, const int);
9020 vector bool char vec_splat (vector bool char, const int);
9021 vector signed short vec_splat (vector signed short, const int);
9022 vector unsigned short vec_splat (vector unsigned short, const int);
9023 vector bool short vec_splat (vector bool short, const int);
9024 vector pixel vec_splat (vector pixel, const int);
9025 vector float vec_splat (vector float, const int);
9026 vector signed int vec_splat (vector signed int, const int);
9027 vector unsigned int vec_splat (vector unsigned int, const int);
9028 vector bool int vec_splat (vector bool int, const int);
9030 vector float vec_vspltw (vector float, const int);
9031 vector signed int vec_vspltw (vector signed int, const int);
9032 vector unsigned int vec_vspltw (vector unsigned int, const int);
9033 vector bool int vec_vspltw (vector bool int, const int);
9035 vector bool short vec_vsplth (vector bool short, const int);
9036 vector signed short vec_vsplth (vector signed short, const int);
9037 vector unsigned short vec_vsplth (vector unsigned short, const int);
9038 vector pixel vec_vsplth (vector pixel, const int);
9040 vector signed char vec_vspltb (vector signed char, const int);
9041 vector unsigned char vec_vspltb (vector unsigned char, const int);
9042 vector bool char vec_vspltb (vector bool char, const int);
9044 vector signed char vec_splat_s8 (const int);
9046 vector signed short vec_splat_s16 (const int);
9048 vector signed int vec_splat_s32 (const int);
9050 vector unsigned char vec_splat_u8 (const int);
9052 vector unsigned short vec_splat_u16 (const int);
9054 vector unsigned int vec_splat_u32 (const int);
9056 vector signed char vec_sr (vector signed char, vector unsigned char);
9057 vector unsigned char vec_sr (vector unsigned char,
9058 vector unsigned char);
9059 vector signed short vec_sr (vector signed short,
9060 vector unsigned short);
9061 vector unsigned short vec_sr (vector unsigned short,
9062 vector unsigned short);
9063 vector signed int vec_sr (vector signed int, vector unsigned int);
9064 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9066 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9067 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9069 vector signed short vec_vsrh (vector signed short,
9070 vector unsigned short);
9071 vector unsigned short vec_vsrh (vector unsigned short,
9072 vector unsigned short);
9074 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9075 vector unsigned char vec_vsrb (vector unsigned char,
9076 vector unsigned char);
9078 vector signed char vec_sra (vector signed char, vector unsigned char);
9079 vector unsigned char vec_sra (vector unsigned char,
9080 vector unsigned char);
9081 vector signed short vec_sra (vector signed short,
9082 vector unsigned short);
9083 vector unsigned short vec_sra (vector unsigned short,
9084 vector unsigned short);
9085 vector signed int vec_sra (vector signed int, vector unsigned int);
9086 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9088 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9089 vector unsigned int vec_vsraw (vector unsigned int,
9090 vector unsigned int);
9092 vector signed short vec_vsrah (vector signed short,
9093 vector unsigned short);
9094 vector unsigned short vec_vsrah (vector unsigned short,
9095 vector unsigned short);
9097 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9098 vector unsigned char vec_vsrab (vector unsigned char,
9099 vector unsigned char);
9101 vector signed int vec_srl (vector signed int, vector unsigned int);
9102 vector signed int vec_srl (vector signed int, vector unsigned short);
9103 vector signed int vec_srl (vector signed int, vector unsigned char);
9104 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9105 vector unsigned int vec_srl (vector unsigned int,
9106 vector unsigned short);
9107 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9108 vector bool int vec_srl (vector bool int, vector unsigned int);
9109 vector bool int vec_srl (vector bool int, vector unsigned short);
9110 vector bool int vec_srl (vector bool int, vector unsigned char);
9111 vector signed short vec_srl (vector signed short, vector unsigned int);
9112 vector signed short vec_srl (vector signed short,
9113 vector unsigned short);
9114 vector signed short vec_srl (vector signed short, vector unsigned char);
9115 vector unsigned short vec_srl (vector unsigned short,
9116 vector unsigned int);
9117 vector unsigned short vec_srl (vector unsigned short,
9118 vector unsigned short);
9119 vector unsigned short vec_srl (vector unsigned short,
9120 vector unsigned char);
9121 vector bool short vec_srl (vector bool short, vector unsigned int);
9122 vector bool short vec_srl (vector bool short, vector unsigned short);
9123 vector bool short vec_srl (vector bool short, vector unsigned char);
9124 vector pixel vec_srl (vector pixel, vector unsigned int);
9125 vector pixel vec_srl (vector pixel, vector unsigned short);
9126 vector pixel vec_srl (vector pixel, vector unsigned char);
9127 vector signed char vec_srl (vector signed char, vector unsigned int);
9128 vector signed char vec_srl (vector signed char, vector unsigned short);
9129 vector signed char vec_srl (vector signed char, vector unsigned char);
9130 vector unsigned char vec_srl (vector unsigned char,
9131 vector unsigned int);
9132 vector unsigned char vec_srl (vector unsigned char,
9133 vector unsigned short);
9134 vector unsigned char vec_srl (vector unsigned char,
9135 vector unsigned char);
9136 vector bool char vec_srl (vector bool char, vector unsigned int);
9137 vector bool char vec_srl (vector bool char, vector unsigned short);
9138 vector bool char vec_srl (vector bool char, vector unsigned char);
9140 vector float vec_sro (vector float, vector signed char);
9141 vector float vec_sro (vector float, vector unsigned char);
9142 vector signed int vec_sro (vector signed int, vector signed char);
9143 vector signed int vec_sro (vector signed int, vector unsigned char);
9144 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9145 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9146 vector signed short vec_sro (vector signed short, vector signed char);
9147 vector signed short vec_sro (vector signed short, vector unsigned char);
9148 vector unsigned short vec_sro (vector unsigned short,
9149 vector signed char);
9150 vector unsigned short vec_sro (vector unsigned short,
9151 vector unsigned char);
9152 vector pixel vec_sro (vector pixel, vector signed char);
9153 vector pixel vec_sro (vector pixel, vector unsigned char);
9154 vector signed char vec_sro (vector signed char, vector signed char);
9155 vector signed char vec_sro (vector signed char, vector unsigned char);
9156 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9157 vector unsigned char vec_sro (vector unsigned char,
9158 vector unsigned char);
9160 void vec_st (vector float, int, vector float *);
9161 void vec_st (vector float, int, float *);
9162 void vec_st (vector signed int, int, vector signed int *);
9163 void vec_st (vector signed int, int, int *);
9164 void vec_st (vector unsigned int, int, vector unsigned int *);
9165 void vec_st (vector unsigned int, int, unsigned int *);
9166 void vec_st (vector bool int, int, vector bool int *);
9167 void vec_st (vector bool int, int, unsigned int *);
9168 void vec_st (vector bool int, int, int *);
9169 void vec_st (vector signed short, int, vector signed short *);
9170 void vec_st (vector signed short, int, short *);
9171 void vec_st (vector unsigned short, int, vector unsigned short *);
9172 void vec_st (vector unsigned short, int, unsigned short *);
9173 void vec_st (vector bool short, int, vector bool short *);
9174 void vec_st (vector bool short, int, unsigned short *);
9175 void vec_st (vector pixel, int, vector pixel *);
9176 void vec_st (vector pixel, int, unsigned short *);
9177 void vec_st (vector pixel, int, short *);
9178 void vec_st (vector bool short, int, short *);
9179 void vec_st (vector signed char, int, vector signed char *);
9180 void vec_st (vector signed char, int, signed char *);
9181 void vec_st (vector unsigned char, int, vector unsigned char *);
9182 void vec_st (vector unsigned char, int, unsigned char *);
9183 void vec_st (vector bool char, int, vector bool char *);
9184 void vec_st (vector bool char, int, unsigned char *);
9185 void vec_st (vector bool char, int, signed char *);
9187 void vec_ste (vector signed char, int, signed char *);
9188 void vec_ste (vector unsigned char, int, unsigned char *);
9189 void vec_ste (vector bool char, int, signed char *);
9190 void vec_ste (vector bool char, int, unsigned char *);
9191 void vec_ste (vector signed short, int, short *);
9192 void vec_ste (vector unsigned short, int, unsigned short *);
9193 void vec_ste (vector bool short, int, short *);
9194 void vec_ste (vector bool short, int, unsigned short *);
9195 void vec_ste (vector pixel, int, short *);
9196 void vec_ste (vector pixel, int, unsigned short *);
9197 void vec_ste (vector float, int, float *);
9198 void vec_ste (vector signed int, int, int *);
9199 void vec_ste (vector unsigned int, int, unsigned int *);
9200 void vec_ste (vector bool int, int, int *);
9201 void vec_ste (vector bool int, int, unsigned int *);
9203 void vec_stvewx (vector float, int, float *);
9204 void vec_stvewx (vector signed int, int, int *);
9205 void vec_stvewx (vector unsigned int, int, unsigned int *);
9206 void vec_stvewx (vector bool int, int, int *);
9207 void vec_stvewx (vector bool int, int, unsigned int *);
9209 void vec_stvehx (vector signed short, int, short *);
9210 void vec_stvehx (vector unsigned short, int, unsigned short *);
9211 void vec_stvehx (vector bool short, int, short *);
9212 void vec_stvehx (vector bool short, int, unsigned short *);
9213 void vec_stvehx (vector pixel, int, short *);
9214 void vec_stvehx (vector pixel, int, unsigned short *);
9216 void vec_stvebx (vector signed char, int, signed char *);
9217 void vec_stvebx (vector unsigned char, int, unsigned char *);
9218 void vec_stvebx (vector bool char, int, signed char *);
9219 void vec_stvebx (vector bool char, int, unsigned char *);
9221 void vec_stl (vector float, int, vector float *);
9222 void vec_stl (vector float, int, float *);
9223 void vec_stl (vector signed int, int, vector signed int *);
9224 void vec_stl (vector signed int, int, int *);
9225 void vec_stl (vector unsigned int, int, vector unsigned int *);
9226 void vec_stl (vector unsigned int, int, unsigned int *);
9227 void vec_stl (vector bool int, int, vector bool int *);
9228 void vec_stl (vector bool int, int, unsigned int *);
9229 void vec_stl (vector bool int, int, int *);
9230 void vec_stl (vector signed short, int, vector signed short *);
9231 void vec_stl (vector signed short, int, short *);
9232 void vec_stl (vector unsigned short, int, vector unsigned short *);
9233 void vec_stl (vector unsigned short, int, unsigned short *);
9234 void vec_stl (vector bool short, int, vector bool short *);
9235 void vec_stl (vector bool short, int, unsigned short *);
9236 void vec_stl (vector bool short, int, short *);
9237 void vec_stl (vector pixel, int, vector pixel *);
9238 void vec_stl (vector pixel, int, unsigned short *);
9239 void vec_stl (vector pixel, int, short *);
9240 void vec_stl (vector signed char, int, vector signed char *);
9241 void vec_stl (vector signed char, int, signed char *);
9242 void vec_stl (vector unsigned char, int, vector unsigned char *);
9243 void vec_stl (vector unsigned char, int, unsigned char *);
9244 void vec_stl (vector bool char, int, vector bool char *);
9245 void vec_stl (vector bool char, int, unsigned char *);
9246 void vec_stl (vector bool char, int, signed char *);
9248 vector signed char vec_sub (vector bool char, vector signed char);
9249 vector signed char vec_sub (vector signed char, vector bool char);
9250 vector signed char vec_sub (vector signed char, vector signed char);
9251 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9252 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9253 vector unsigned char vec_sub (vector unsigned char,
9254 vector unsigned char);
9255 vector signed short vec_sub (vector bool short, vector signed short);
9256 vector signed short vec_sub (vector signed short, vector bool short);
9257 vector signed short vec_sub (vector signed short, vector signed short);
9258 vector unsigned short vec_sub (vector bool short,
9259 vector unsigned short);
9260 vector unsigned short vec_sub (vector unsigned short,
9262 vector unsigned short vec_sub (vector unsigned short,
9263 vector unsigned short);
9264 vector signed int vec_sub (vector bool int, vector signed int);
9265 vector signed int vec_sub (vector signed int, vector bool int);
9266 vector signed int vec_sub (vector signed int, vector signed int);
9267 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9268 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9269 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9270 vector float vec_sub (vector float, vector float);
9272 vector float vec_vsubfp (vector float, vector float);
9274 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9275 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9276 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9277 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9278 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9279 vector unsigned int vec_vsubuwm (vector unsigned int,
9280 vector unsigned int);
9282 vector signed short vec_vsubuhm (vector bool short,
9283 vector signed short);
9284 vector signed short vec_vsubuhm (vector signed short,
9286 vector signed short vec_vsubuhm (vector signed short,
9287 vector signed short);
9288 vector unsigned short vec_vsubuhm (vector bool short,
9289 vector unsigned short);
9290 vector unsigned short vec_vsubuhm (vector unsigned short,
9292 vector unsigned short vec_vsubuhm (vector unsigned short,
9293 vector unsigned short);
9295 vector signed char vec_vsububm (vector bool char, vector signed char);
9296 vector signed char vec_vsububm (vector signed char, vector bool char);
9297 vector signed char vec_vsububm (vector signed char, vector signed char);
9298 vector unsigned char vec_vsububm (vector bool char,
9299 vector unsigned char);
9300 vector unsigned char vec_vsububm (vector unsigned char,
9302 vector unsigned char vec_vsububm (vector unsigned char,
9303 vector unsigned char);
9305 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9307 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9308 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9309 vector unsigned char vec_subs (vector unsigned char,
9310 vector unsigned char);
9311 vector signed char vec_subs (vector bool char, vector signed char);
9312 vector signed char vec_subs (vector signed char, vector bool char);
9313 vector signed char vec_subs (vector signed char, vector signed char);
9314 vector unsigned short vec_subs (vector bool short,
9315 vector unsigned short);
9316 vector unsigned short vec_subs (vector unsigned short,
9318 vector unsigned short vec_subs (vector unsigned short,
9319 vector unsigned short);
9320 vector signed short vec_subs (vector bool short, vector signed short);
9321 vector signed short vec_subs (vector signed short, vector bool short);
9322 vector signed short vec_subs (vector signed short, vector signed short);
9323 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9324 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9325 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9326 vector signed int vec_subs (vector bool int, vector signed int);
9327 vector signed int vec_subs (vector signed int, vector bool int);
9328 vector signed int vec_subs (vector signed int, vector signed int);
9330 vector signed int vec_vsubsws (vector bool int, vector signed int);
9331 vector signed int vec_vsubsws (vector signed int, vector bool int);
9332 vector signed int vec_vsubsws (vector signed int, vector signed int);
9334 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9335 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9336 vector unsigned int vec_vsubuws (vector unsigned int,
9337 vector unsigned int);
9339 vector signed short vec_vsubshs (vector bool short,
9340 vector signed short);
9341 vector signed short vec_vsubshs (vector signed short,
9343 vector signed short vec_vsubshs (vector signed short,
9344 vector signed short);
9346 vector unsigned short vec_vsubuhs (vector bool short,
9347 vector unsigned short);
9348 vector unsigned short vec_vsubuhs (vector unsigned short,
9350 vector unsigned short vec_vsubuhs (vector unsigned short,
9351 vector unsigned short);
9353 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9354 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9355 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9357 vector unsigned char vec_vsububs (vector bool char,
9358 vector unsigned char);
9359 vector unsigned char vec_vsububs (vector unsigned char,
9361 vector unsigned char vec_vsububs (vector unsigned char,
9362 vector unsigned char);
9364 vector unsigned int vec_sum4s (vector unsigned char,
9365 vector unsigned int);
9366 vector signed int vec_sum4s (vector signed char, vector signed int);
9367 vector signed int vec_sum4s (vector signed short, vector signed int);
9369 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9371 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9373 vector unsigned int vec_vsum4ubs (vector unsigned char,
9374 vector unsigned int);
9376 vector signed int vec_sum2s (vector signed int, vector signed int);
9378 vector signed int vec_sums (vector signed int, vector signed int);
9380 vector float vec_trunc (vector float);
9382 vector signed short vec_unpackh (vector signed char);
9383 vector bool short vec_unpackh (vector bool char);
9384 vector signed int vec_unpackh (vector signed short);
9385 vector bool int vec_unpackh (vector bool short);
9386 vector unsigned int vec_unpackh (vector pixel);
9388 vector bool int vec_vupkhsh (vector bool short);
9389 vector signed int vec_vupkhsh (vector signed short);
9391 vector unsigned int vec_vupkhpx (vector pixel);
9393 vector bool short vec_vupkhsb (vector bool char);
9394 vector signed short vec_vupkhsb (vector signed char);
9396 vector signed short vec_unpackl (vector signed char);
9397 vector bool short vec_unpackl (vector bool char);
9398 vector unsigned int vec_unpackl (vector pixel);
9399 vector signed int vec_unpackl (vector signed short);
9400 vector bool int vec_unpackl (vector bool short);
9402 vector unsigned int vec_vupklpx (vector pixel);
9404 vector bool int vec_vupklsh (vector bool short);
9405 vector signed int vec_vupklsh (vector signed short);
9407 vector bool short vec_vupklsb (vector bool char);
9408 vector signed short vec_vupklsb (vector signed char);
9410 vector float vec_xor (vector float, vector float);
9411 vector float vec_xor (vector float, vector bool int);
9412 vector float vec_xor (vector bool int, vector float);
9413 vector bool int vec_xor (vector bool int, vector bool int);
9414 vector signed int vec_xor (vector bool int, vector signed int);
9415 vector signed int vec_xor (vector signed int, vector bool int);
9416 vector signed int vec_xor (vector signed int, vector signed int);
9417 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9418 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9419 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9420 vector bool short vec_xor (vector bool short, vector bool short);
9421 vector signed short vec_xor (vector bool short, vector signed short);
9422 vector signed short vec_xor (vector signed short, vector bool short);
9423 vector signed short vec_xor (vector signed short, vector signed short);
9424 vector unsigned short vec_xor (vector bool short,
9425 vector unsigned short);
9426 vector unsigned short vec_xor (vector unsigned short,
9428 vector unsigned short vec_xor (vector unsigned short,
9429 vector unsigned short);
9430 vector signed char vec_xor (vector bool char, vector signed char);
9431 vector bool char vec_xor (vector bool char, vector bool char);
9432 vector signed char vec_xor (vector signed char, vector bool char);
9433 vector signed char vec_xor (vector signed char, vector signed char);
9434 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9435 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9436 vector unsigned char vec_xor (vector unsigned char,
9437 vector unsigned char);
9439 int vec_all_eq (vector signed char, vector bool char);
9440 int vec_all_eq (vector signed char, vector signed char);
9441 int vec_all_eq (vector unsigned char, vector bool char);
9442 int vec_all_eq (vector unsigned char, vector unsigned char);
9443 int vec_all_eq (vector bool char, vector bool char);
9444 int vec_all_eq (vector bool char, vector unsigned char);
9445 int vec_all_eq (vector bool char, vector signed char);
9446 int vec_all_eq (vector signed short, vector bool short);
9447 int vec_all_eq (vector signed short, vector signed short);
9448 int vec_all_eq (vector unsigned short, vector bool short);
9449 int vec_all_eq (vector unsigned short, vector unsigned short);
9450 int vec_all_eq (vector bool short, vector bool short);
9451 int vec_all_eq (vector bool short, vector unsigned short);
9452 int vec_all_eq (vector bool short, vector signed short);
9453 int vec_all_eq (vector pixel, vector pixel);
9454 int vec_all_eq (vector signed int, vector bool int);
9455 int vec_all_eq (vector signed int, vector signed int);
9456 int vec_all_eq (vector unsigned int, vector bool int);
9457 int vec_all_eq (vector unsigned int, vector unsigned int);
9458 int vec_all_eq (vector bool int, vector bool int);
9459 int vec_all_eq (vector bool int, vector unsigned int);
9460 int vec_all_eq (vector bool int, vector signed int);
9461 int vec_all_eq (vector float, vector float);
9463 int vec_all_ge (vector bool char, vector unsigned char);
9464 int vec_all_ge (vector unsigned char, vector bool char);
9465 int vec_all_ge (vector unsigned char, vector unsigned char);
9466 int vec_all_ge (vector bool char, vector signed char);
9467 int vec_all_ge (vector signed char, vector bool char);
9468 int vec_all_ge (vector signed char, vector signed char);
9469 int vec_all_ge (vector bool short, vector unsigned short);
9470 int vec_all_ge (vector unsigned short, vector bool short);
9471 int vec_all_ge (vector unsigned short, vector unsigned short);
9472 int vec_all_ge (vector signed short, vector signed short);
9473 int vec_all_ge (vector bool short, vector signed short);
9474 int vec_all_ge (vector signed short, vector bool short);
9475 int vec_all_ge (vector bool int, vector unsigned int);
9476 int vec_all_ge (vector unsigned int, vector bool int);
9477 int vec_all_ge (vector unsigned int, vector unsigned int);
9478 int vec_all_ge (vector bool int, vector signed int);
9479 int vec_all_ge (vector signed int, vector bool int);
9480 int vec_all_ge (vector signed int, vector signed int);
9481 int vec_all_ge (vector float, vector float);
9483 int vec_all_gt (vector bool char, vector unsigned char);
9484 int vec_all_gt (vector unsigned char, vector bool char);
9485 int vec_all_gt (vector unsigned char, vector unsigned char);
9486 int vec_all_gt (vector bool char, vector signed char);
9487 int vec_all_gt (vector signed char, vector bool char);
9488 int vec_all_gt (vector signed char, vector signed char);
9489 int vec_all_gt (vector bool short, vector unsigned short);
9490 int vec_all_gt (vector unsigned short, vector bool short);
9491 int vec_all_gt (vector unsigned short, vector unsigned short);
9492 int vec_all_gt (vector bool short, vector signed short);
9493 int vec_all_gt (vector signed short, vector bool short);
9494 int vec_all_gt (vector signed short, vector signed short);
9495 int vec_all_gt (vector bool int, vector unsigned int);
9496 int vec_all_gt (vector unsigned int, vector bool int);
9497 int vec_all_gt (vector unsigned int, vector unsigned int);
9498 int vec_all_gt (vector bool int, vector signed int);
9499 int vec_all_gt (vector signed int, vector bool int);
9500 int vec_all_gt (vector signed int, vector signed int);
9501 int vec_all_gt (vector float, vector float);
9503 int vec_all_in (vector float, vector float);
9505 int vec_all_le (vector bool char, vector unsigned char);
9506 int vec_all_le (vector unsigned char, vector bool char);
9507 int vec_all_le (vector unsigned char, vector unsigned char);
9508 int vec_all_le (vector bool char, vector signed char);
9509 int vec_all_le (vector signed char, vector bool char);
9510 int vec_all_le (vector signed char, vector signed char);
9511 int vec_all_le (vector bool short, vector unsigned short);
9512 int vec_all_le (vector unsigned short, vector bool short);
9513 int vec_all_le (vector unsigned short, vector unsigned short);
9514 int vec_all_le (vector bool short, vector signed short);
9515 int vec_all_le (vector signed short, vector bool short);
9516 int vec_all_le (vector signed short, vector signed short);
9517 int vec_all_le (vector bool int, vector unsigned int);
9518 int vec_all_le (vector unsigned int, vector bool int);
9519 int vec_all_le (vector unsigned int, vector unsigned int);
9520 int vec_all_le (vector bool int, vector signed int);
9521 int vec_all_le (vector signed int, vector bool int);
9522 int vec_all_le (vector signed int, vector signed int);
9523 int vec_all_le (vector float, vector float);
9525 int vec_all_lt (vector bool char, vector unsigned char);
9526 int vec_all_lt (vector unsigned char, vector bool char);
9527 int vec_all_lt (vector unsigned char, vector unsigned char);
9528 int vec_all_lt (vector bool char, vector signed char);
9529 int vec_all_lt (vector signed char, vector bool char);
9530 int vec_all_lt (vector signed char, vector signed char);
9531 int vec_all_lt (vector bool short, vector unsigned short);
9532 int vec_all_lt (vector unsigned short, vector bool short);
9533 int vec_all_lt (vector unsigned short, vector unsigned short);
9534 int vec_all_lt (vector bool short, vector signed short);
9535 int vec_all_lt (vector signed short, vector bool short);
9536 int vec_all_lt (vector signed short, vector signed short);
9537 int vec_all_lt (vector bool int, vector unsigned int);
9538 int vec_all_lt (vector unsigned int, vector bool int);
9539 int vec_all_lt (vector unsigned int, vector unsigned int);
9540 int vec_all_lt (vector bool int, vector signed int);
9541 int vec_all_lt (vector signed int, vector bool int);
9542 int vec_all_lt (vector signed int, vector signed int);
9543 int vec_all_lt (vector float, vector float);
9545 int vec_all_nan (vector float);
9547 int vec_all_ne (vector signed char, vector bool char);
9548 int vec_all_ne (vector signed char, vector signed char);
9549 int vec_all_ne (vector unsigned char, vector bool char);
9550 int vec_all_ne (vector unsigned char, vector unsigned char);
9551 int vec_all_ne (vector bool char, vector bool char);
9552 int vec_all_ne (vector bool char, vector unsigned char);
9553 int vec_all_ne (vector bool char, vector signed char);
9554 int vec_all_ne (vector signed short, vector bool short);
9555 int vec_all_ne (vector signed short, vector signed short);
9556 int vec_all_ne (vector unsigned short, vector bool short);
9557 int vec_all_ne (vector unsigned short, vector unsigned short);
9558 int vec_all_ne (vector bool short, vector bool short);
9559 int vec_all_ne (vector bool short, vector unsigned short);
9560 int vec_all_ne (vector bool short, vector signed short);
9561 int vec_all_ne (vector pixel, vector pixel);
9562 int vec_all_ne (vector signed int, vector bool int);
9563 int vec_all_ne (vector signed int, vector signed int);
9564 int vec_all_ne (vector unsigned int, vector bool int);
9565 int vec_all_ne (vector unsigned int, vector unsigned int);
9566 int vec_all_ne (vector bool int, vector bool int);
9567 int vec_all_ne (vector bool int, vector unsigned int);
9568 int vec_all_ne (vector bool int, vector signed int);
9569 int vec_all_ne (vector float, vector float);
9571 int vec_all_nge (vector float, vector float);
9573 int vec_all_ngt (vector float, vector float);
9575 int vec_all_nle (vector float, vector float);
9577 int vec_all_nlt (vector float, vector float);
9579 int vec_all_numeric (vector float);
9581 int vec_any_eq (vector signed char, vector bool char);
9582 int vec_any_eq (vector signed char, vector signed char);
9583 int vec_any_eq (vector unsigned char, vector bool char);
9584 int vec_any_eq (vector unsigned char, vector unsigned char);
9585 int vec_any_eq (vector bool char, vector bool char);
9586 int vec_any_eq (vector bool char, vector unsigned char);
9587 int vec_any_eq (vector bool char, vector signed char);
9588 int vec_any_eq (vector signed short, vector bool short);
9589 int vec_any_eq (vector signed short, vector signed short);
9590 int vec_any_eq (vector unsigned short, vector bool short);
9591 int vec_any_eq (vector unsigned short, vector unsigned short);
9592 int vec_any_eq (vector bool short, vector bool short);
9593 int vec_any_eq (vector bool short, vector unsigned short);
9594 int vec_any_eq (vector bool short, vector signed short);
9595 int vec_any_eq (vector pixel, vector pixel);
9596 int vec_any_eq (vector signed int, vector bool int);
9597 int vec_any_eq (vector signed int, vector signed int);
9598 int vec_any_eq (vector unsigned int, vector bool int);
9599 int vec_any_eq (vector unsigned int, vector unsigned int);
9600 int vec_any_eq (vector bool int, vector bool int);
9601 int vec_any_eq (vector bool int, vector unsigned int);
9602 int vec_any_eq (vector bool int, vector signed int);
9603 int vec_any_eq (vector float, vector float);
9605 int vec_any_ge (vector signed char, vector bool char);
9606 int vec_any_ge (vector unsigned char, vector bool char);
9607 int vec_any_ge (vector unsigned char, vector unsigned char);
9608 int vec_any_ge (vector signed char, vector signed char);
9609 int vec_any_ge (vector bool char, vector unsigned char);
9610 int vec_any_ge (vector bool char, vector signed char);
9611 int vec_any_ge (vector unsigned short, vector bool short);
9612 int vec_any_ge (vector unsigned short, vector unsigned short);
9613 int vec_any_ge (vector signed short, vector signed short);
9614 int vec_any_ge (vector signed short, vector bool short);
9615 int vec_any_ge (vector bool short, vector unsigned short);
9616 int vec_any_ge (vector bool short, vector signed short);
9617 int vec_any_ge (vector signed int, vector bool int);
9618 int vec_any_ge (vector unsigned int, vector bool int);
9619 int vec_any_ge (vector unsigned int, vector unsigned int);
9620 int vec_any_ge (vector signed int, vector signed int);
9621 int vec_any_ge (vector bool int, vector unsigned int);
9622 int vec_any_ge (vector bool int, vector signed int);
9623 int vec_any_ge (vector float, vector float);
9625 int vec_any_gt (vector bool char, vector unsigned char);
9626 int vec_any_gt (vector unsigned char, vector bool char);
9627 int vec_any_gt (vector unsigned char, vector unsigned char);
9628 int vec_any_gt (vector bool char, vector signed char);
9629 int vec_any_gt (vector signed char, vector bool char);
9630 int vec_any_gt (vector signed char, vector signed char);
9631 int vec_any_gt (vector bool short, vector unsigned short);
9632 int vec_any_gt (vector unsigned short, vector bool short);
9633 int vec_any_gt (vector unsigned short, vector unsigned short);
9634 int vec_any_gt (vector bool short, vector signed short);
9635 int vec_any_gt (vector signed short, vector bool short);
9636 int vec_any_gt (vector signed short, vector signed short);
9637 int vec_any_gt (vector bool int, vector unsigned int);
9638 int vec_any_gt (vector unsigned int, vector bool int);
9639 int vec_any_gt (vector unsigned int, vector unsigned int);
9640 int vec_any_gt (vector bool int, vector signed int);
9641 int vec_any_gt (vector signed int, vector bool int);
9642 int vec_any_gt (vector signed int, vector signed int);
9643 int vec_any_gt (vector float, vector float);
9645 int vec_any_le (vector bool char, vector unsigned char);
9646 int vec_any_le (vector unsigned char, vector bool char);
9647 int vec_any_le (vector unsigned char, vector unsigned char);
9648 int vec_any_le (vector bool char, vector signed char);
9649 int vec_any_le (vector signed char, vector bool char);
9650 int vec_any_le (vector signed char, vector signed char);
9651 int vec_any_le (vector bool short, vector unsigned short);
9652 int vec_any_le (vector unsigned short, vector bool short);
9653 int vec_any_le (vector unsigned short, vector unsigned short);
9654 int vec_any_le (vector bool short, vector signed short);
9655 int vec_any_le (vector signed short, vector bool short);
9656 int vec_any_le (vector signed short, vector signed short);
9657 int vec_any_le (vector bool int, vector unsigned int);
9658 int vec_any_le (vector unsigned int, vector bool int);
9659 int vec_any_le (vector unsigned int, vector unsigned int);
9660 int vec_any_le (vector bool int, vector signed int);
9661 int vec_any_le (vector signed int, vector bool int);
9662 int vec_any_le (vector signed int, vector signed int);
9663 int vec_any_le (vector float, vector float);
9665 int vec_any_lt (vector bool char, vector unsigned char);
9666 int vec_any_lt (vector unsigned char, vector bool char);
9667 int vec_any_lt (vector unsigned char, vector unsigned char);
9668 int vec_any_lt (vector bool char, vector signed char);
9669 int vec_any_lt (vector signed char, vector bool char);
9670 int vec_any_lt (vector signed char, vector signed char);
9671 int vec_any_lt (vector bool short, vector unsigned short);
9672 int vec_any_lt (vector unsigned short, vector bool short);
9673 int vec_any_lt (vector unsigned short, vector unsigned short);
9674 int vec_any_lt (vector bool short, vector signed short);
9675 int vec_any_lt (vector signed short, vector bool short);
9676 int vec_any_lt (vector signed short, vector signed short);
9677 int vec_any_lt (vector bool int, vector unsigned int);
9678 int vec_any_lt (vector unsigned int, vector bool int);
9679 int vec_any_lt (vector unsigned int, vector unsigned int);
9680 int vec_any_lt (vector bool int, vector signed int);
9681 int vec_any_lt (vector signed int, vector bool int);
9682 int vec_any_lt (vector signed int, vector signed int);
9683 int vec_any_lt (vector float, vector float);
9685 int vec_any_nan (vector float);
9687 int vec_any_ne (vector signed char, vector bool char);
9688 int vec_any_ne (vector signed char, vector signed char);
9689 int vec_any_ne (vector unsigned char, vector bool char);
9690 int vec_any_ne (vector unsigned char, vector unsigned char);
9691 int vec_any_ne (vector bool char, vector bool char);
9692 int vec_any_ne (vector bool char, vector unsigned char);
9693 int vec_any_ne (vector bool char, vector signed char);
9694 int vec_any_ne (vector signed short, vector bool short);
9695 int vec_any_ne (vector signed short, vector signed short);
9696 int vec_any_ne (vector unsigned short, vector bool short);
9697 int vec_any_ne (vector unsigned short, vector unsigned short);
9698 int vec_any_ne (vector bool short, vector bool short);
9699 int vec_any_ne (vector bool short, vector unsigned short);
9700 int vec_any_ne (vector bool short, vector signed short);
9701 int vec_any_ne (vector pixel, vector pixel);
9702 int vec_any_ne (vector signed int, vector bool int);
9703 int vec_any_ne (vector signed int, vector signed int);
9704 int vec_any_ne (vector unsigned int, vector bool int);
9705 int vec_any_ne (vector unsigned int, vector unsigned int);
9706 int vec_any_ne (vector bool int, vector bool int);
9707 int vec_any_ne (vector bool int, vector unsigned int);
9708 int vec_any_ne (vector bool int, vector signed int);
9709 int vec_any_ne (vector float, vector float);
9711 int vec_any_nge (vector float, vector float);
9713 int vec_any_ngt (vector float, vector float);
9715 int vec_any_nle (vector float, vector float);
9717 int vec_any_nlt (vector float, vector float);
9719 int vec_any_numeric (vector float);
9721 int vec_any_out (vector float, vector float);
9724 @node SPARC VIS Built-in Functions
9725 @subsection SPARC VIS Built-in Functions
9727 GCC supports SIMD operations on the SPARC using both the generic vector
9728 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9729 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9730 switch, the VIS extension is exposed as the following built-in functions:
9733 typedef int v2si __attribute__ ((vector_size (8)));
9734 typedef short v4hi __attribute__ ((vector_size (8)));
9735 typedef short v2hi __attribute__ ((vector_size (4)));
9736 typedef char v8qi __attribute__ ((vector_size (8)));
9737 typedef char v4qi __attribute__ ((vector_size (4)));
9739 void * __builtin_vis_alignaddr (void *, long);
9740 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9741 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9742 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9743 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9745 v4hi __builtin_vis_fexpand (v4qi);
9747 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9748 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9749 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9750 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9751 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9752 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9753 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9755 v4qi __builtin_vis_fpack16 (v4hi);
9756 v8qi __builtin_vis_fpack32 (v2si, v2si);
9757 v2hi __builtin_vis_fpackfix (v2si);
9758 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9760 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9763 @node Target Format Checks
9764 @section Format Checks Specific to Particular Target Machines
9766 For some target machines, GCC supports additional options to the
9768 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9771 * Solaris Format Checks::
9774 @node Solaris Format Checks
9775 @subsection Solaris Format Checks
9777 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9778 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9779 conversions, and the two-argument @code{%b} conversion for displaying
9780 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9783 @section Pragmas Accepted by GCC
9787 GCC supports several types of pragmas, primarily in order to compile
9788 code originally written for other compilers. Note that in general
9789 we do not recommend the use of pragmas; @xref{Function Attributes},
9790 for further explanation.
9795 * RS/6000 and PowerPC Pragmas::
9798 * Symbol-Renaming Pragmas::
9799 * Structure-Packing Pragmas::
9801 * Diagnostic Pragmas::
9802 * Visibility Pragmas::
9806 @subsection ARM Pragmas
9808 The ARM target defines pragmas for controlling the default addition of
9809 @code{long_call} and @code{short_call} attributes to functions.
9810 @xref{Function Attributes}, for information about the effects of these
9815 @cindex pragma, long_calls
9816 Set all subsequent functions to have the @code{long_call} attribute.
9819 @cindex pragma, no_long_calls
9820 Set all subsequent functions to have the @code{short_call} attribute.
9822 @item long_calls_off
9823 @cindex pragma, long_calls_off
9824 Do not affect the @code{long_call} or @code{short_call} attributes of
9825 subsequent functions.
9829 @subsection M32C Pragmas
9832 @item memregs @var{number}
9833 @cindex pragma, memregs
9834 Overrides the command line option @code{-memregs=} for the current
9835 file. Use with care! This pragma must be before any function in the
9836 file, and mixing different memregs values in different objects may
9837 make them incompatible. This pragma is useful when a
9838 performance-critical function uses a memreg for temporary values,
9839 as it may allow you to reduce the number of memregs used.
9843 @node RS/6000 and PowerPC Pragmas
9844 @subsection RS/6000 and PowerPC Pragmas
9846 The RS/6000 and PowerPC targets define one pragma for controlling
9847 whether or not the @code{longcall} attribute is added to function
9848 declarations by default. This pragma overrides the @option{-mlongcall}
9849 option, but not the @code{longcall} and @code{shortcall} attributes.
9850 @xref{RS/6000 and PowerPC Options}, for more information about when long
9851 calls are and are not necessary.
9855 @cindex pragma, longcall
9856 Apply the @code{longcall} attribute to all subsequent function
9860 Do not apply the @code{longcall} attribute to subsequent function
9864 @c Describe c4x pragmas here.
9865 @c Describe h8300 pragmas here.
9866 @c Describe sh pragmas here.
9867 @c Describe v850 pragmas here.
9869 @node Darwin Pragmas
9870 @subsection Darwin Pragmas
9872 The following pragmas are available for all architectures running the
9873 Darwin operating system. These are useful for compatibility with other
9877 @item mark @var{tokens}@dots{}
9878 @cindex pragma, mark
9879 This pragma is accepted, but has no effect.
9881 @item options align=@var{alignment}
9882 @cindex pragma, options align
9883 This pragma sets the alignment of fields in structures. The values of
9884 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9885 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9886 properly; to restore the previous setting, use @code{reset} for the
9889 @item segment @var{tokens}@dots{}
9890 @cindex pragma, segment
9891 This pragma is accepted, but has no effect.
9893 @item unused (@var{var} [, @var{var}]@dots{})
9894 @cindex pragma, unused
9895 This pragma declares variables to be possibly unused. GCC will not
9896 produce warnings for the listed variables. The effect is similar to
9897 that of the @code{unused} attribute, except that this pragma may appear
9898 anywhere within the variables' scopes.
9901 @node Solaris Pragmas
9902 @subsection Solaris Pragmas
9904 The Solaris target supports @code{#pragma redefine_extname}
9905 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
9906 @code{#pragma} directives for compatibility with the system compiler.
9909 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
9910 @cindex pragma, align
9912 Increase the minimum alignment of each @var{variable} to @var{alignment}.
9913 This is the same as GCC's @code{aligned} attribute @pxref{Variable
9914 Attributes}). Macro expansion occurs on the arguments to this pragma
9915 when compiling C. It does not currently occur when compiling C++, but
9916 this is a bug which may be fixed in a future release.
9918 @item fini (@var{function} [, @var{function}]...)
9919 @cindex pragma, fini
9921 This pragma causes each listed @var{function} to be called after
9922 main, or during shared module unloading, by adding a call to the
9923 @code{.fini} section.
9925 @item init (@var{function} [, @var{function}]...)
9926 @cindex pragma, init
9928 This pragma causes each listed @var{function} to be called during
9929 initialization (before @code{main}) or during shared module loading, by
9930 adding a call to the @code{.init} section.
9934 @node Symbol-Renaming Pragmas
9935 @subsection Symbol-Renaming Pragmas
9937 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
9938 supports two @code{#pragma} directives which change the name used in
9939 assembly for a given declaration. These pragmas are only available on
9940 platforms whose system headers need them. To get this effect on all
9941 platforms supported by GCC, use the asm labels extension (@pxref{Asm
9945 @item redefine_extname @var{oldname} @var{newname}
9946 @cindex pragma, redefine_extname
9948 This pragma gives the C function @var{oldname} the assembly symbol
9949 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
9950 will be defined if this pragma is available (currently only on
9953 @item extern_prefix @var{string}
9954 @cindex pragma, extern_prefix
9956 This pragma causes all subsequent external function and variable
9957 declarations to have @var{string} prepended to their assembly symbols.
9958 This effect may be terminated with another @code{extern_prefix} pragma
9959 whose argument is an empty string. The preprocessor macro
9960 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
9961 available (currently only on Tru64 UNIX)@.
9964 These pragmas and the asm labels extension interact in a complicated
9965 manner. Here are some corner cases you may want to be aware of.
9968 @item Both pragmas silently apply only to declarations with external
9969 linkage. Asm labels do not have this restriction.
9971 @item In C++, both pragmas silently apply only to declarations with
9972 ``C'' linkage. Again, asm labels do not have this restriction.
9974 @item If any of the three ways of changing the assembly name of a
9975 declaration is applied to a declaration whose assembly name has
9976 already been determined (either by a previous use of one of these
9977 features, or because the compiler needed the assembly name in order to
9978 generate code), and the new name is different, a warning issues and
9979 the name does not change.
9981 @item The @var{oldname} used by @code{#pragma redefine_extname} is
9982 always the C-language name.
9984 @item If @code{#pragma extern_prefix} is in effect, and a declaration
9985 occurs with an asm label attached, the prefix is silently ignored for
9988 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
9989 apply to the same declaration, whichever triggered first wins, and a
9990 warning issues if they contradict each other. (We would like to have
9991 @code{#pragma redefine_extname} always win, for consistency with asm
9992 labels, but if @code{#pragma extern_prefix} triggers first we have no
9993 way of knowing that that happened.)
9996 @node Structure-Packing Pragmas
9997 @subsection Structure-Packing Pragmas
9999 For compatibility with Win32, GCC supports a set of @code{#pragma}
10000 directives which change the maximum alignment of members of structures
10001 (other than zero-width bitfields), unions, and classes subsequently
10002 defined. The @var{n} value below always is required to be a small power
10003 of two and specifies the new alignment in bytes.
10006 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10007 @item @code{#pragma pack()} sets the alignment to the one that was in
10008 effect when compilation started (see also command line option
10009 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10010 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10011 setting on an internal stack and then optionally sets the new alignment.
10012 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10013 saved at the top of the internal stack (and removes that stack entry).
10014 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10015 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10016 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10017 @code{#pragma pack(pop)}.
10020 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10021 @code{#pragma} which lays out a structure as the documented
10022 @code{__attribute__ ((ms_struct))}.
10024 @item @code{#pragma ms_struct on} turns on the layout for structures
10026 @item @code{#pragma ms_struct off} turns off the layout for structures
10028 @item @code{#pragma ms_struct reset} goes back to the default layout.
10032 @subsection Weak Pragmas
10034 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10035 directives for declaring symbols to be weak, and defining weak
10039 @item #pragma weak @var{symbol}
10040 @cindex pragma, weak
10041 This pragma declares @var{symbol} to be weak, as if the declaration
10042 had the attribute of the same name. The pragma may appear before
10043 or after the declaration of @var{symbol}, but must appear before
10044 either its first use or its definition. It is not an error for
10045 @var{symbol} to never be defined at all.
10047 @item #pragma weak @var{symbol1} = @var{symbol2}
10048 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10049 It is an error if @var{symbol2} is not defined in the current
10053 @node Diagnostic Pragmas
10054 @subsection Diagnostic Pragmas
10056 GCC allows the user to selectively enable or disable certain types of
10057 diagnostics, and change the kind of the diagnostic. For example, a
10058 project's policy might require that all sources compile with
10059 @option{-Werror} but certain files might have exceptions allowing
10060 specific types of warnings. Or, a project might selectively enable
10061 diagnostics and treat them as errors depending on which preprocessor
10062 macros are defined.
10065 @item #pragma GCC diagnostic @var{kind} @var{option}
10066 @cindex pragma, diagnostic
10068 Modifies the disposition of a diagnostic. Note that not all
10069 diagnostics are modifiable; at the moment only warnings (normally
10070 controlled by @samp{-W...}) can be controlled, and not all of them.
10071 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10072 are controllable and which option controls them.
10074 @var{kind} is @samp{error} to treat this diagnostic as an error,
10075 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10076 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10077 @var{option} is a double quoted string which matches the command line
10081 #pragma GCC diagnostic warning "-Wformat"
10082 #pragma GCC diagnostic error "-Wformat"
10083 #pragma GCC diagnostic ignored "-Wformat"
10086 Note that these pragmas override any command line options. Also,
10087 while it is syntactically valid to put these pragmas anywhere in your
10088 sources, the only supported location for them is before any data or
10089 functions are defined. Doing otherwise may result in unpredictable
10090 results depending on how the optimizer manages your sources. If the
10091 same option is listed multiple times, the last one specified is the
10092 one that is in effect. This pragma is not intended to be a general
10093 purpose replacement for command line options, but for implementing
10094 strict control over project policies.
10098 @node Visibility Pragmas
10099 @subsection Visibility Pragmas
10102 @item #pragma GCC visibility push(@var{visibility})
10103 @itemx #pragma GCC visibility pop
10104 @cindex pragma, visibility
10106 This pragma allows the user to set the visibility for multiple
10107 declarations without having to give each a visibility attribute
10108 @xref{Function Attributes}, for more information about visibility and
10109 the attribute syntax.
10111 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10112 declarations. Class members and template specializations are not
10113 affected; if you want to override the visibility for a particular
10114 member or instantiation, you must use an attribute.
10118 @node Unnamed Fields
10119 @section Unnamed struct/union fields within structs/unions
10123 For compatibility with other compilers, GCC allows you to define
10124 a structure or union that contains, as fields, structures and unions
10125 without names. For example:
10138 In this example, the user would be able to access members of the unnamed
10139 union with code like @samp{foo.b}. Note that only unnamed structs and
10140 unions are allowed, you may not have, for example, an unnamed
10143 You must never create such structures that cause ambiguous field definitions.
10144 For example, this structure:
10155 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10156 Such constructs are not supported and must be avoided. In the future,
10157 such constructs may be detected and treated as compilation errors.
10159 @opindex fms-extensions
10160 Unless @option{-fms-extensions} is used, the unnamed field must be a
10161 structure or union definition without a tag (for example, @samp{struct
10162 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10163 also be a definition with a tag such as @samp{struct foo @{ int a;
10164 @};}, a reference to a previously defined structure or union such as
10165 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10166 previously defined structure or union type.
10169 @section Thread-Local Storage
10170 @cindex Thread-Local Storage
10171 @cindex @acronym{TLS}
10174 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10175 are allocated such that there is one instance of the variable per extant
10176 thread. The run-time model GCC uses to implement this originates
10177 in the IA-64 processor-specific ABI, but has since been migrated
10178 to other processors as well. It requires significant support from
10179 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10180 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10181 is not available everywhere.
10183 At the user level, the extension is visible with a new storage
10184 class keyword: @code{__thread}. For example:
10188 extern __thread struct state s;
10189 static __thread char *p;
10192 The @code{__thread} specifier may be used alone, with the @code{extern}
10193 or @code{static} specifiers, but with no other storage class specifier.
10194 When used with @code{extern} or @code{static}, @code{__thread} must appear
10195 immediately after the other storage class specifier.
10197 The @code{__thread} specifier may be applied to any global, file-scoped
10198 static, function-scoped static, or static data member of a class. It may
10199 not be applied to block-scoped automatic or non-static data member.
10201 When the address-of operator is applied to a thread-local variable, it is
10202 evaluated at run-time and returns the address of the current thread's
10203 instance of that variable. An address so obtained may be used by any
10204 thread. When a thread terminates, any pointers to thread-local variables
10205 in that thread become invalid.
10207 No static initialization may refer to the address of a thread-local variable.
10209 In C++, if an initializer is present for a thread-local variable, it must
10210 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10213 See @uref{http://people.redhat.com/drepper/tls.pdf,
10214 ELF Handling For Thread-Local Storage} for a detailed explanation of
10215 the four thread-local storage addressing models, and how the run-time
10216 is expected to function.
10219 * C99 Thread-Local Edits::
10220 * C++98 Thread-Local Edits::
10223 @node C99 Thread-Local Edits
10224 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10226 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10227 that document the exact semantics of the language extension.
10231 @cite{5.1.2 Execution environments}
10233 Add new text after paragraph 1
10236 Within either execution environment, a @dfn{thread} is a flow of
10237 control within a program. It is implementation defined whether
10238 or not there may be more than one thread associated with a program.
10239 It is implementation defined how threads beyond the first are
10240 created, the name and type of the function called at thread
10241 startup, and how threads may be terminated. However, objects
10242 with thread storage duration shall be initialized before thread
10247 @cite{6.2.4 Storage durations of objects}
10249 Add new text before paragraph 3
10252 An object whose identifier is declared with the storage-class
10253 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10254 Its lifetime is the entire execution of the thread, and its
10255 stored value is initialized only once, prior to thread startup.
10259 @cite{6.4.1 Keywords}
10261 Add @code{__thread}.
10264 @cite{6.7.1 Storage-class specifiers}
10266 Add @code{__thread} to the list of storage class specifiers in
10269 Change paragraph 2 to
10272 With the exception of @code{__thread}, at most one storage-class
10273 specifier may be given [@dots{}]. The @code{__thread} specifier may
10274 be used alone, or immediately following @code{extern} or
10278 Add new text after paragraph 6
10281 The declaration of an identifier for a variable that has
10282 block scope that specifies @code{__thread} shall also
10283 specify either @code{extern} or @code{static}.
10285 The @code{__thread} specifier shall be used only with
10290 @node C++98 Thread-Local Edits
10291 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10293 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10294 that document the exact semantics of the language extension.
10298 @b{[intro.execution]}
10300 New text after paragraph 4
10303 A @dfn{thread} is a flow of control within the abstract machine.
10304 It is implementation defined whether or not there may be more than
10308 New text after paragraph 7
10311 It is unspecified whether additional action must be taken to
10312 ensure when and whether side effects are visible to other threads.
10318 Add @code{__thread}.
10321 @b{[basic.start.main]}
10323 Add after paragraph 5
10326 The thread that begins execution at the @code{main} function is called
10327 the @dfn{main thread}. It is implementation defined how functions
10328 beginning threads other than the main thread are designated or typed.
10329 A function so designated, as well as the @code{main} function, is called
10330 a @dfn{thread startup function}. It is implementation defined what
10331 happens if a thread startup function returns. It is implementation
10332 defined what happens to other threads when any thread calls @code{exit}.
10336 @b{[basic.start.init]}
10338 Add after paragraph 4
10341 The storage for an object of thread storage duration shall be
10342 statically initialized before the first statement of the thread startup
10343 function. An object of thread storage duration shall not require
10344 dynamic initialization.
10348 @b{[basic.start.term]}
10350 Add after paragraph 3
10353 The type of an object with thread storage duration shall not have a
10354 non-trivial destructor, nor shall it be an array type whose elements
10355 (directly or indirectly) have non-trivial destructors.
10361 Add ``thread storage duration'' to the list in paragraph 1.
10366 Thread, static, and automatic storage durations are associated with
10367 objects introduced by declarations [@dots{}].
10370 Add @code{__thread} to the list of specifiers in paragraph 3.
10373 @b{[basic.stc.thread]}
10375 New section before @b{[basic.stc.static]}
10378 The keyword @code{__thread} applied to a non-local object gives the
10379 object thread storage duration.
10381 A local variable or class data member declared both @code{static}
10382 and @code{__thread} gives the variable or member thread storage
10387 @b{[basic.stc.static]}
10392 All objects which have neither thread storage duration, dynamic
10393 storage duration nor are local [@dots{}].
10399 Add @code{__thread} to the list in paragraph 1.
10404 With the exception of @code{__thread}, at most one
10405 @var{storage-class-specifier} shall appear in a given
10406 @var{decl-specifier-seq}. The @code{__thread} specifier may
10407 be used alone, or immediately following the @code{extern} or
10408 @code{static} specifiers. [@dots{}]
10411 Add after paragraph 5
10414 The @code{__thread} specifier can be applied only to the names of objects
10415 and to anonymous unions.
10421 Add after paragraph 6
10424 Non-@code{static} members shall not be @code{__thread}.
10428 @node Binary constants
10429 @section Binary constants using the @samp{0b} prefix
10430 @cindex Binary constants using the @samp{0b} prefix
10432 Integer constants can be written as binary constants, consisting of a
10433 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
10434 @samp{0B}. This is particularly useful in environments that operate a
10435 lot on the bit-level (like microcontrollers).
10437 The following statements are identical:
10446 The type of these constants follows the same rules as for octal or
10447 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
10450 @node C++ Extensions
10451 @chapter Extensions to the C++ Language
10452 @cindex extensions, C++ language
10453 @cindex C++ language extensions
10455 The GNU compiler provides these extensions to the C++ language (and you
10456 can also use most of the C language extensions in your C++ programs). If you
10457 want to write code that checks whether these features are available, you can
10458 test for the GNU compiler the same way as for C programs: check for a
10459 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10460 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10461 Predefined Macros,cpp,The GNU C Preprocessor}).
10464 * Volatiles:: What constitutes an access to a volatile object.
10465 * Restricted Pointers:: C99 restricted pointers and references.
10466 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10467 * C++ Interface:: You can use a single C++ header file for both
10468 declarations and definitions.
10469 * Template Instantiation:: Methods for ensuring that exactly one copy of
10470 each needed template instantiation is emitted.
10471 * Bound member functions:: You can extract a function pointer to the
10472 method denoted by a @samp{->*} or @samp{.*} expression.
10473 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10474 * Namespace Association:: Strong using-directives for namespace association.
10475 * Java Exceptions:: Tweaking exception handling to work with Java.
10476 * Deprecated Features:: Things will disappear from g++.
10477 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10481 @section When is a Volatile Object Accessed?
10482 @cindex accessing volatiles
10483 @cindex volatile read
10484 @cindex volatile write
10485 @cindex volatile access
10487 Both the C and C++ standard have the concept of volatile objects. These
10488 are normally accessed by pointers and used for accessing hardware. The
10489 standards encourage compilers to refrain from optimizations concerning
10490 accesses to volatile objects. The C standard leaves it implementation
10491 defined as to what constitutes a volatile access. The C++ standard omits
10492 to specify this, except to say that C++ should behave in a similar manner
10493 to C with respect to volatiles, where possible. The minimum either
10494 standard specifies is that at a sequence point all previous accesses to
10495 volatile objects have stabilized and no subsequent accesses have
10496 occurred. Thus an implementation is free to reorder and combine
10497 volatile accesses which occur between sequence points, but cannot do so
10498 for accesses across a sequence point. The use of volatiles does not
10499 allow you to violate the restriction on updating objects multiple times
10500 within a sequence point.
10502 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10504 The behavior differs slightly between C and C++ in the non-obvious cases:
10507 volatile int *src = @var{somevalue};
10511 With C, such expressions are rvalues, and GCC interprets this either as a
10512 read of the volatile object being pointed to or only as request to evaluate
10513 the side-effects. The C++ standard specifies that such expressions do not
10514 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10515 object may be incomplete. The C++ standard does not specify explicitly
10516 that it is this lvalue to rvalue conversion which may be responsible for
10517 causing an access. However, there is reason to believe that it is,
10518 because otherwise certain simple expressions become undefined. However,
10519 because it would surprise most programmers, G++ treats dereferencing a
10520 pointer to volatile object of complete type when the value is unused as
10521 GCC would do for an equivalent type in C. When the object has incomplete
10522 type, G++ issues a warning; if you wish to force an error, you must
10523 force a conversion to rvalue with, for instance, a static cast.
10525 When using a reference to volatile, G++ does not treat equivalent
10526 expressions as accesses to volatiles, but instead issues a warning that
10527 no volatile is accessed. The rationale for this is that otherwise it
10528 becomes difficult to determine where volatile access occur, and not
10529 possible to ignore the return value from functions returning volatile
10530 references. Again, if you wish to force a read, cast the reference to
10533 @node Restricted Pointers
10534 @section Restricting Pointer Aliasing
10535 @cindex restricted pointers
10536 @cindex restricted references
10537 @cindex restricted this pointer
10539 As with the C front end, G++ understands the C99 feature of restricted pointers,
10540 specified with the @code{__restrict__}, or @code{__restrict} type
10541 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10542 language flag, @code{restrict} is not a keyword in C++.
10544 In addition to allowing restricted pointers, you can specify restricted
10545 references, which indicate that the reference is not aliased in the local
10549 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10556 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10557 @var{rref} refers to a (different) unaliased integer.
10559 You may also specify whether a member function's @var{this} pointer is
10560 unaliased by using @code{__restrict__} as a member function qualifier.
10563 void T::fn () __restrict__
10570 Within the body of @code{T::fn}, @var{this} will have the effective
10571 definition @code{T *__restrict__ const this}. Notice that the
10572 interpretation of a @code{__restrict__} member function qualifier is
10573 different to that of @code{const} or @code{volatile} qualifier, in that it
10574 is applied to the pointer rather than the object. This is consistent with
10575 other compilers which implement restricted pointers.
10577 As with all outermost parameter qualifiers, @code{__restrict__} is
10578 ignored in function definition matching. This means you only need to
10579 specify @code{__restrict__} in a function definition, rather than
10580 in a function prototype as well.
10582 @node Vague Linkage
10583 @section Vague Linkage
10584 @cindex vague linkage
10586 There are several constructs in C++ which require space in the object
10587 file but are not clearly tied to a single translation unit. We say that
10588 these constructs have ``vague linkage''. Typically such constructs are
10589 emitted wherever they are needed, though sometimes we can be more
10593 @item Inline Functions
10594 Inline functions are typically defined in a header file which can be
10595 included in many different compilations. Hopefully they can usually be
10596 inlined, but sometimes an out-of-line copy is necessary, if the address
10597 of the function is taken or if inlining fails. In general, we emit an
10598 out-of-line copy in all translation units where one is needed. As an
10599 exception, we only emit inline virtual functions with the vtable, since
10600 it will always require a copy.
10602 Local static variables and string constants used in an inline function
10603 are also considered to have vague linkage, since they must be shared
10604 between all inlined and out-of-line instances of the function.
10608 C++ virtual functions are implemented in most compilers using a lookup
10609 table, known as a vtable. The vtable contains pointers to the virtual
10610 functions provided by a class, and each object of the class contains a
10611 pointer to its vtable (or vtables, in some multiple-inheritance
10612 situations). If the class declares any non-inline, non-pure virtual
10613 functions, the first one is chosen as the ``key method'' for the class,
10614 and the vtable is only emitted in the translation unit where the key
10617 @emph{Note:} If the chosen key method is later defined as inline, the
10618 vtable will still be emitted in every translation unit which defines it.
10619 Make sure that any inline virtuals are declared inline in the class
10620 body, even if they are not defined there.
10622 @item type_info objects
10625 C++ requires information about types to be written out in order to
10626 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10627 For polymorphic classes (classes with virtual functions), the type_info
10628 object is written out along with the vtable so that @samp{dynamic_cast}
10629 can determine the dynamic type of a class object at runtime. For all
10630 other types, we write out the type_info object when it is used: when
10631 applying @samp{typeid} to an expression, throwing an object, or
10632 referring to a type in a catch clause or exception specification.
10634 @item Template Instantiations
10635 Most everything in this section also applies to template instantiations,
10636 but there are other options as well.
10637 @xref{Template Instantiation,,Where's the Template?}.
10641 When used with GNU ld version 2.8 or later on an ELF system such as
10642 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10643 these constructs will be discarded at link time. This is known as
10646 On targets that don't support COMDAT, but do support weak symbols, GCC
10647 will use them. This way one copy will override all the others, but
10648 the unused copies will still take up space in the executable.
10650 For targets which do not support either COMDAT or weak symbols,
10651 most entities with vague linkage will be emitted as local symbols to
10652 avoid duplicate definition errors from the linker. This will not happen
10653 for local statics in inlines, however, as having multiple copies will
10654 almost certainly break things.
10656 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10657 another way to control placement of these constructs.
10659 @node C++ Interface
10660 @section #pragma interface and implementation
10662 @cindex interface and implementation headers, C++
10663 @cindex C++ interface and implementation headers
10664 @cindex pragmas, interface and implementation
10666 @code{#pragma interface} and @code{#pragma implementation} provide the
10667 user with a way of explicitly directing the compiler to emit entities
10668 with vague linkage (and debugging information) in a particular
10671 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10672 most cases, because of COMDAT support and the ``key method'' heuristic
10673 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10674 program to grow due to unnecessary out-of-line copies of inline
10675 functions. Currently (3.4) the only benefit of these
10676 @code{#pragma}s is reduced duplication of debugging information, and
10677 that should be addressed soon on DWARF 2 targets with the use of
10681 @item #pragma interface
10682 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10683 @kindex #pragma interface
10684 Use this directive in @emph{header files} that define object classes, to save
10685 space in most of the object files that use those classes. Normally,
10686 local copies of certain information (backup copies of inline member
10687 functions, debugging information, and the internal tables that implement
10688 virtual functions) must be kept in each object file that includes class
10689 definitions. You can use this pragma to avoid such duplication. When a
10690 header file containing @samp{#pragma interface} is included in a
10691 compilation, this auxiliary information will not be generated (unless
10692 the main input source file itself uses @samp{#pragma implementation}).
10693 Instead, the object files will contain references to be resolved at link
10696 The second form of this directive is useful for the case where you have
10697 multiple headers with the same name in different directories. If you
10698 use this form, you must specify the same string to @samp{#pragma
10701 @item #pragma implementation
10702 @itemx #pragma implementation "@var{objects}.h"
10703 @kindex #pragma implementation
10704 Use this pragma in a @emph{main input file}, when you want full output from
10705 included header files to be generated (and made globally visible). The
10706 included header file, in turn, should use @samp{#pragma interface}.
10707 Backup copies of inline member functions, debugging information, and the
10708 internal tables used to implement virtual functions are all generated in
10709 implementation files.
10711 @cindex implied @code{#pragma implementation}
10712 @cindex @code{#pragma implementation}, implied
10713 @cindex naming convention, implementation headers
10714 If you use @samp{#pragma implementation} with no argument, it applies to
10715 an include file with the same basename@footnote{A file's @dfn{basename}
10716 was the name stripped of all leading path information and of trailing
10717 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10718 file. For example, in @file{allclass.cc}, giving just
10719 @samp{#pragma implementation}
10720 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10722 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10723 an implementation file whenever you would include it from
10724 @file{allclass.cc} even if you never specified @samp{#pragma
10725 implementation}. This was deemed to be more trouble than it was worth,
10726 however, and disabled.
10728 Use the string argument if you want a single implementation file to
10729 include code from multiple header files. (You must also use
10730 @samp{#include} to include the header file; @samp{#pragma
10731 implementation} only specifies how to use the file---it doesn't actually
10734 There is no way to split up the contents of a single header file into
10735 multiple implementation files.
10738 @cindex inlining and C++ pragmas
10739 @cindex C++ pragmas, effect on inlining
10740 @cindex pragmas in C++, effect on inlining
10741 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10742 effect on function inlining.
10744 If you define a class in a header file marked with @samp{#pragma
10745 interface}, the effect on an inline function defined in that class is
10746 similar to an explicit @code{extern} declaration---the compiler emits
10747 no code at all to define an independent version of the function. Its
10748 definition is used only for inlining with its callers.
10750 @opindex fno-implement-inlines
10751 Conversely, when you include the same header file in a main source file
10752 that declares it as @samp{#pragma implementation}, the compiler emits
10753 code for the function itself; this defines a version of the function
10754 that can be found via pointers (or by callers compiled without
10755 inlining). If all calls to the function can be inlined, you can avoid
10756 emitting the function by compiling with @option{-fno-implement-inlines}.
10757 If any calls were not inlined, you will get linker errors.
10759 @node Template Instantiation
10760 @section Where's the Template?
10761 @cindex template instantiation
10763 C++ templates are the first language feature to require more
10764 intelligence from the environment than one usually finds on a UNIX
10765 system. Somehow the compiler and linker have to make sure that each
10766 template instance occurs exactly once in the executable if it is needed,
10767 and not at all otherwise. There are two basic approaches to this
10768 problem, which are referred to as the Borland model and the Cfront model.
10771 @item Borland model
10772 Borland C++ solved the template instantiation problem by adding the code
10773 equivalent of common blocks to their linker; the compiler emits template
10774 instances in each translation unit that uses them, and the linker
10775 collapses them together. The advantage of this model is that the linker
10776 only has to consider the object files themselves; there is no external
10777 complexity to worry about. This disadvantage is that compilation time
10778 is increased because the template code is being compiled repeatedly.
10779 Code written for this model tends to include definitions of all
10780 templates in the header file, since they must be seen to be
10784 The AT&T C++ translator, Cfront, solved the template instantiation
10785 problem by creating the notion of a template repository, an
10786 automatically maintained place where template instances are stored. A
10787 more modern version of the repository works as follows: As individual
10788 object files are built, the compiler places any template definitions and
10789 instantiations encountered in the repository. At link time, the link
10790 wrapper adds in the objects in the repository and compiles any needed
10791 instances that were not previously emitted. The advantages of this
10792 model are more optimal compilation speed and the ability to use the
10793 system linker; to implement the Borland model a compiler vendor also
10794 needs to replace the linker. The disadvantages are vastly increased
10795 complexity, and thus potential for error; for some code this can be
10796 just as transparent, but in practice it can been very difficult to build
10797 multiple programs in one directory and one program in multiple
10798 directories. Code written for this model tends to separate definitions
10799 of non-inline member templates into a separate file, which should be
10800 compiled separately.
10803 When used with GNU ld version 2.8 or later on an ELF system such as
10804 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10805 Borland model. On other systems, G++ implements neither automatic
10808 A future version of G++ will support a hybrid model whereby the compiler
10809 will emit any instantiations for which the template definition is
10810 included in the compile, and store template definitions and
10811 instantiation context information into the object file for the rest.
10812 The link wrapper will extract that information as necessary and invoke
10813 the compiler to produce the remaining instantiations. The linker will
10814 then combine duplicate instantiations.
10816 In the mean time, you have the following options for dealing with
10817 template instantiations:
10822 Compile your template-using code with @option{-frepo}. The compiler will
10823 generate files with the extension @samp{.rpo} listing all of the
10824 template instantiations used in the corresponding object files which
10825 could be instantiated there; the link wrapper, @samp{collect2}, will
10826 then update the @samp{.rpo} files to tell the compiler where to place
10827 those instantiations and rebuild any affected object files. The
10828 link-time overhead is negligible after the first pass, as the compiler
10829 will continue to place the instantiations in the same files.
10831 This is your best option for application code written for the Borland
10832 model, as it will just work. Code written for the Cfront model will
10833 need to be modified so that the template definitions are available at
10834 one or more points of instantiation; usually this is as simple as adding
10835 @code{#include <tmethods.cc>} to the end of each template header.
10837 For library code, if you want the library to provide all of the template
10838 instantiations it needs, just try to link all of its object files
10839 together; the link will fail, but cause the instantiations to be
10840 generated as a side effect. Be warned, however, that this may cause
10841 conflicts if multiple libraries try to provide the same instantiations.
10842 For greater control, use explicit instantiation as described in the next
10846 @opindex fno-implicit-templates
10847 Compile your code with @option{-fno-implicit-templates} to disable the
10848 implicit generation of template instances, and explicitly instantiate
10849 all the ones you use. This approach requires more knowledge of exactly
10850 which instances you need than do the others, but it's less
10851 mysterious and allows greater control. You can scatter the explicit
10852 instantiations throughout your program, perhaps putting them in the
10853 translation units where the instances are used or the translation units
10854 that define the templates themselves; you can put all of the explicit
10855 instantiations you need into one big file; or you can create small files
10862 template class Foo<int>;
10863 template ostream& operator <<
10864 (ostream&, const Foo<int>&);
10867 for each of the instances you need, and create a template instantiation
10868 library from those.
10870 If you are using Cfront-model code, you can probably get away with not
10871 using @option{-fno-implicit-templates} when compiling files that don't
10872 @samp{#include} the member template definitions.
10874 If you use one big file to do the instantiations, you may want to
10875 compile it without @option{-fno-implicit-templates} so you get all of the
10876 instances required by your explicit instantiations (but not by any
10877 other files) without having to specify them as well.
10879 G++ has extended the template instantiation syntax given in the ISO
10880 standard to allow forward declaration of explicit instantiations
10881 (with @code{extern}), instantiation of the compiler support data for a
10882 template class (i.e.@: the vtable) without instantiating any of its
10883 members (with @code{inline}), and instantiation of only the static data
10884 members of a template class, without the support data or member
10885 functions (with (@code{static}):
10888 extern template int max (int, int);
10889 inline template class Foo<int>;
10890 static template class Foo<int>;
10894 Do nothing. Pretend G++ does implement automatic instantiation
10895 management. Code written for the Borland model will work fine, but
10896 each translation unit will contain instances of each of the templates it
10897 uses. In a large program, this can lead to an unacceptable amount of code
10901 @node Bound member functions
10902 @section Extracting the function pointer from a bound pointer to member function
10904 @cindex pointer to member function
10905 @cindex bound pointer to member function
10907 In C++, pointer to member functions (PMFs) are implemented using a wide
10908 pointer of sorts to handle all the possible call mechanisms; the PMF
10909 needs to store information about how to adjust the @samp{this} pointer,
10910 and if the function pointed to is virtual, where to find the vtable, and
10911 where in the vtable to look for the member function. If you are using
10912 PMFs in an inner loop, you should really reconsider that decision. If
10913 that is not an option, you can extract the pointer to the function that
10914 would be called for a given object/PMF pair and call it directly inside
10915 the inner loop, to save a bit of time.
10917 Note that you will still be paying the penalty for the call through a
10918 function pointer; on most modern architectures, such a call defeats the
10919 branch prediction features of the CPU@. This is also true of normal
10920 virtual function calls.
10922 The syntax for this extension is
10926 extern int (A::*fp)();
10927 typedef int (*fptr)(A *);
10929 fptr p = (fptr)(a.*fp);
10932 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
10933 no object is needed to obtain the address of the function. They can be
10934 converted to function pointers directly:
10937 fptr p1 = (fptr)(&A::foo);
10940 @opindex Wno-pmf-conversions
10941 You must specify @option{-Wno-pmf-conversions} to use this extension.
10943 @node C++ Attributes
10944 @section C++-Specific Variable, Function, and Type Attributes
10946 Some attributes only make sense for C++ programs.
10949 @item init_priority (@var{priority})
10950 @cindex init_priority attribute
10953 In Standard C++, objects defined at namespace scope are guaranteed to be
10954 initialized in an order in strict accordance with that of their definitions
10955 @emph{in a given translation unit}. No guarantee is made for initializations
10956 across translation units. However, GNU C++ allows users to control the
10957 order of initialization of objects defined at namespace scope with the
10958 @code{init_priority} attribute by specifying a relative @var{priority},
10959 a constant integral expression currently bounded between 101 and 65535
10960 inclusive. Lower numbers indicate a higher priority.
10962 In the following example, @code{A} would normally be created before
10963 @code{B}, but the @code{init_priority} attribute has reversed that order:
10966 Some_Class A __attribute__ ((init_priority (2000)));
10967 Some_Class B __attribute__ ((init_priority (543)));
10971 Note that the particular values of @var{priority} do not matter; only their
10974 @item java_interface
10975 @cindex java_interface attribute
10977 This type attribute informs C++ that the class is a Java interface. It may
10978 only be applied to classes declared within an @code{extern "Java"} block.
10979 Calls to methods declared in this interface will be dispatched using GCJ's
10980 interface table mechanism, instead of regular virtual table dispatch.
10984 See also @xref{Namespace Association}.
10986 @node Namespace Association
10987 @section Namespace Association
10989 @strong{Caution:} The semantics of this extension are not fully
10990 defined. Users should refrain from using this extension as its
10991 semantics may change subtly over time. It is possible that this
10992 extension will be removed in future versions of G++.
10994 A using-directive with @code{__attribute ((strong))} is stronger
10995 than a normal using-directive in two ways:
10999 Templates from the used namespace can be specialized and explicitly
11000 instantiated as though they were members of the using namespace.
11003 The using namespace is considered an associated namespace of all
11004 templates in the used namespace for purposes of argument-dependent
11008 The used namespace must be nested within the using namespace so that
11009 normal unqualified lookup works properly.
11011 This is useful for composing a namespace transparently from
11012 implementation namespaces. For example:
11017 template <class T> struct A @{ @};
11019 using namespace debug __attribute ((__strong__));
11020 template <> struct A<int> @{ @}; // @r{ok to specialize}
11022 template <class T> void f (A<T>);
11027 f (std::A<float>()); // @r{lookup finds} std::f
11032 @node Java Exceptions
11033 @section Java Exceptions
11035 The Java language uses a slightly different exception handling model
11036 from C++. Normally, GNU C++ will automatically detect when you are
11037 writing C++ code that uses Java exceptions, and handle them
11038 appropriately. However, if C++ code only needs to execute destructors
11039 when Java exceptions are thrown through it, GCC will guess incorrectly.
11040 Sample problematic code is:
11043 struct S @{ ~S(); @};
11044 extern void bar(); // @r{is written in Java, and may throw exceptions}
11053 The usual effect of an incorrect guess is a link failure, complaining of
11054 a missing routine called @samp{__gxx_personality_v0}.
11056 You can inform the compiler that Java exceptions are to be used in a
11057 translation unit, irrespective of what it might think, by writing
11058 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11059 @samp{#pragma} must appear before any functions that throw or catch
11060 exceptions, or run destructors when exceptions are thrown through them.
11062 You cannot mix Java and C++ exceptions in the same translation unit. It
11063 is believed to be safe to throw a C++ exception from one file through
11064 another file compiled for the Java exception model, or vice versa, but
11065 there may be bugs in this area.
11067 @node Deprecated Features
11068 @section Deprecated Features
11070 In the past, the GNU C++ compiler was extended to experiment with new
11071 features, at a time when the C++ language was still evolving. Now that
11072 the C++ standard is complete, some of those features are superseded by
11073 superior alternatives. Using the old features might cause a warning in
11074 some cases that the feature will be dropped in the future. In other
11075 cases, the feature might be gone already.
11077 While the list below is not exhaustive, it documents some of the options
11078 that are now deprecated:
11081 @item -fexternal-templates
11082 @itemx -falt-external-templates
11083 These are two of the many ways for G++ to implement template
11084 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11085 defines how template definitions have to be organized across
11086 implementation units. G++ has an implicit instantiation mechanism that
11087 should work just fine for standard-conforming code.
11089 @item -fstrict-prototype
11090 @itemx -fno-strict-prototype
11091 Previously it was possible to use an empty prototype parameter list to
11092 indicate an unspecified number of parameters (like C), rather than no
11093 parameters, as C++ demands. This feature has been removed, except where
11094 it is required for backwards compatibility @xref{Backwards Compatibility}.
11097 G++ allows a virtual function returning @samp{void *} to be overridden
11098 by one returning a different pointer type. This extension to the
11099 covariant return type rules is now deprecated and will be removed from a
11102 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11103 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11104 and will be removed in a future version. Code using these operators
11105 should be modified to use @code{std::min} and @code{std::max} instead.
11107 The named return value extension has been deprecated, and is now
11110 The use of initializer lists with new expressions has been deprecated,
11111 and is now removed from G++.
11113 Floating and complex non-type template parameters have been deprecated,
11114 and are now removed from G++.
11116 The implicit typename extension has been deprecated and is now
11119 The use of default arguments in function pointers, function typedefs
11120 and other places where they are not permitted by the standard is
11121 deprecated and will be removed from a future version of G++.
11123 G++ allows floating-point literals to appear in integral constant expressions,
11124 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11125 This extension is deprecated and will be removed from a future version.
11127 G++ allows static data members of const floating-point type to be declared
11128 with an initializer in a class definition. The standard only allows
11129 initializers for static members of const integral types and const
11130 enumeration types so this extension has been deprecated and will be removed
11131 from a future version.
11133 @node Backwards Compatibility
11134 @section Backwards Compatibility
11135 @cindex Backwards Compatibility
11136 @cindex ARM [Annotated C++ Reference Manual]
11138 Now that there is a definitive ISO standard C++, G++ has a specification
11139 to adhere to. The C++ language evolved over time, and features that
11140 used to be acceptable in previous drafts of the standard, such as the ARM
11141 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11142 compilation of C++ written to such drafts, G++ contains some backwards
11143 compatibilities. @emph{All such backwards compatibility features are
11144 liable to disappear in future versions of G++.} They should be considered
11145 deprecated @xref{Deprecated Features}.
11149 If a variable is declared at for scope, it used to remain in scope until
11150 the end of the scope which contained the for statement (rather than just
11151 within the for scope). G++ retains this, but issues a warning, if such a
11152 variable is accessed outside the for scope.
11154 @item Implicit C language
11155 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11156 scope to set the language. On such systems, all header files are
11157 implicitly scoped inside a C language scope. Also, an empty prototype
11158 @code{()} will be treated as an unspecified number of arguments, rather
11159 than no arguments, as C++ demands.