1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C and Objective-C@. Most of them are
20 also available in C++. @xref{C++ Extensions,,Extensions to the
21 C++ Language}, for extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * 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 When the value of a decimal floating type cannot be represented in the
860 integer type to which it is being converted, the result is undefined
861 rather than the result value specified by the draft technical report.
864 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
865 are supported by the DWARF2 debug information format.
871 ISO C99 supports floating-point numbers written not only in the usual
872 decimal notation, such as @code{1.55e1}, but also numbers such as
873 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
874 supports this in C89 mode (except in some cases when strictly
875 conforming) and in C++. In that format the
876 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
877 mandatory. The exponent is a decimal number that indicates the power of
878 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
885 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
886 is the same as @code{1.55e1}.
888 Unlike for floating-point numbers in the decimal notation the exponent
889 is always required in the hexadecimal notation. Otherwise the compiler
890 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
891 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
892 extension for floating-point constants of type @code{float}.
895 @section Arrays of Length Zero
896 @cindex arrays of length zero
897 @cindex zero-length arrays
898 @cindex length-zero arrays
899 @cindex flexible array members
901 Zero-length arrays are allowed in GNU C@. They are very useful as the
902 last element of a structure which is really a header for a variable-length
911 struct line *thisline = (struct line *)
912 malloc (sizeof (struct line) + this_length);
913 thisline->length = this_length;
916 In ISO C90, you would have to give @code{contents} a length of 1, which
917 means either you waste space or complicate the argument to @code{malloc}.
919 In ISO C99, you would use a @dfn{flexible array member}, which is
920 slightly different in syntax and semantics:
924 Flexible array members are written as @code{contents[]} without
928 Flexible array members have incomplete type, and so the @code{sizeof}
929 operator may not be applied. As a quirk of the original implementation
930 of zero-length arrays, @code{sizeof} evaluates to zero.
933 Flexible array members may only appear as the last member of a
934 @code{struct} that is otherwise non-empty.
937 A structure containing a flexible array member, or a union containing
938 such a structure (possibly recursively), may not be a member of a
939 structure or an element of an array. (However, these uses are
940 permitted by GCC as extensions.)
943 GCC versions before 3.0 allowed zero-length arrays to be statically
944 initialized, as if they were flexible arrays. In addition to those
945 cases that were useful, it also allowed initializations in situations
946 that would corrupt later data. Non-empty initialization of zero-length
947 arrays is now treated like any case where there are more initializer
948 elements than the array holds, in that a suitable warning about "excess
949 elements in array" is given, and the excess elements (all of them, in
950 this case) are ignored.
952 Instead GCC allows static initialization of flexible array members.
953 This is equivalent to defining a new structure containing the original
954 structure followed by an array of sufficient size to contain the data.
955 I.e.@: in the following, @code{f1} is constructed as if it were declared
961 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
964 struct f1 f1; int data[3];
965 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
969 The convenience of this extension is that @code{f1} has the desired
970 type, eliminating the need to consistently refer to @code{f2.f1}.
972 This has symmetry with normal static arrays, in that an array of
973 unknown size is also written with @code{[]}.
975 Of course, this extension only makes sense if the extra data comes at
976 the end of a top-level object, as otherwise we would be overwriting
977 data at subsequent offsets. To avoid undue complication and confusion
978 with initialization of deeply nested arrays, we simply disallow any
979 non-empty initialization except when the structure is the top-level
983 struct foo @{ int x; int y[]; @};
984 struct bar @{ struct foo z; @};
986 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
987 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
988 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
989 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
992 @node Empty Structures
993 @section Structures With No Members
994 @cindex empty structures
995 @cindex zero-size structures
997 GCC permits a C structure to have no members:
1004 The structure will have size zero. In C++, empty structures are part
1005 of the language. G++ treats empty structures as if they had a single
1006 member of type @code{char}.
1008 @node Variable Length
1009 @section Arrays of Variable Length
1010 @cindex variable-length arrays
1011 @cindex arrays of variable length
1014 Variable-length automatic arrays are allowed in ISO C99, and as an
1015 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1016 implementation of variable-length arrays does not yet conform in detail
1017 to the ISO C99 standard.) These arrays are
1018 declared like any other automatic arrays, but with a length that is not
1019 a constant expression. The storage is allocated at the point of
1020 declaration and deallocated when the brace-level is exited. For
1025 concat_fopen (char *s1, char *s2, char *mode)
1027 char str[strlen (s1) + strlen (s2) + 1];
1030 return fopen (str, mode);
1034 @cindex scope of a variable length array
1035 @cindex variable-length array scope
1036 @cindex deallocating variable length arrays
1037 Jumping or breaking out of the scope of the array name deallocates the
1038 storage. Jumping into the scope is not allowed; you get an error
1041 @cindex @code{alloca} vs variable-length arrays
1042 You can use the function @code{alloca} to get an effect much like
1043 variable-length arrays. The function @code{alloca} is available in
1044 many other C implementations (but not in all). On the other hand,
1045 variable-length arrays are more elegant.
1047 There are other differences between these two methods. Space allocated
1048 with @code{alloca} exists until the containing @emph{function} returns.
1049 The space for a variable-length array is deallocated as soon as the array
1050 name's scope ends. (If you use both variable-length arrays and
1051 @code{alloca} in the same function, deallocation of a variable-length array
1052 will also deallocate anything more recently allocated with @code{alloca}.)
1054 You can also use variable-length arrays as arguments to functions:
1058 tester (int len, char data[len][len])
1064 The length of an array is computed once when the storage is allocated
1065 and is remembered for the scope of the array in case you access it with
1068 If you want to pass the array first and the length afterward, you can
1069 use a forward declaration in the parameter list---another GNU extension.
1073 tester (int len; char data[len][len], int len)
1079 @cindex parameter forward declaration
1080 The @samp{int len} before the semicolon is a @dfn{parameter forward
1081 declaration}, and it serves the purpose of making the name @code{len}
1082 known when the declaration of @code{data} is parsed.
1084 You can write any number of such parameter forward declarations in the
1085 parameter list. They can be separated by commas or semicolons, but the
1086 last one must end with a semicolon, which is followed by the ``real''
1087 parameter declarations. Each forward declaration must match a ``real''
1088 declaration in parameter name and data type. ISO C99 does not support
1089 parameter forward declarations.
1091 @node Variadic Macros
1092 @section Macros with a Variable Number of Arguments.
1093 @cindex variable number of arguments
1094 @cindex macro with variable arguments
1095 @cindex rest argument (in macro)
1096 @cindex variadic macros
1098 In the ISO C standard of 1999, a macro can be declared to accept a
1099 variable number of arguments much as a function can. The syntax for
1100 defining the macro is similar to that of a function. Here is an
1104 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1107 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1108 such a macro, it represents the zero or more tokens until the closing
1109 parenthesis that ends the invocation, including any commas. This set of
1110 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1111 wherever it appears. See the CPP manual for more information.
1113 GCC has long supported variadic macros, and used a different syntax that
1114 allowed you to give a name to the variable arguments just like any other
1115 argument. Here is an example:
1118 #define debug(format, args...) fprintf (stderr, format, args)
1121 This is in all ways equivalent to the ISO C example above, but arguably
1122 more readable and descriptive.
1124 GNU CPP has two further variadic macro extensions, and permits them to
1125 be used with either of the above forms of macro definition.
1127 In standard C, you are not allowed to leave the variable argument out
1128 entirely; but you are allowed to pass an empty argument. For example,
1129 this invocation is invalid in ISO C, because there is no comma after
1136 GNU CPP permits you to completely omit the variable arguments in this
1137 way. In the above examples, the compiler would complain, though since
1138 the expansion of the macro still has the extra comma after the format
1141 To help solve this problem, CPP behaves specially for variable arguments
1142 used with the token paste operator, @samp{##}. If instead you write
1145 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1148 and if the variable arguments are omitted or empty, the @samp{##}
1149 operator causes the preprocessor to remove the comma before it. If you
1150 do provide some variable arguments in your macro invocation, GNU CPP
1151 does not complain about the paste operation and instead places the
1152 variable arguments after the comma. Just like any other pasted macro
1153 argument, these arguments are not macro expanded.
1155 @node Escaped Newlines
1156 @section Slightly Looser Rules for Escaped Newlines
1157 @cindex escaped newlines
1158 @cindex newlines (escaped)
1160 Recently, the preprocessor has relaxed its treatment of escaped
1161 newlines. Previously, the newline had to immediately follow a
1162 backslash. The current implementation allows whitespace in the form
1163 of spaces, horizontal and vertical tabs, and form feeds between the
1164 backslash and the subsequent newline. The preprocessor issues a
1165 warning, but treats it as a valid escaped newline and combines the two
1166 lines to form a single logical line. This works within comments and
1167 tokens, as well as between tokens. Comments are @emph{not} treated as
1168 whitespace for the purposes of this relaxation, since they have not
1169 yet been replaced with spaces.
1172 @section Non-Lvalue Arrays May Have Subscripts
1173 @cindex subscripting
1174 @cindex arrays, non-lvalue
1176 @cindex subscripting and function values
1177 In ISO C99, arrays that are not lvalues still decay to pointers, and
1178 may be subscripted, although they may not be modified or used after
1179 the next sequence point and the unary @samp{&} operator may not be
1180 applied to them. As an extension, GCC allows such arrays to be
1181 subscripted in C89 mode, though otherwise they do not decay to
1182 pointers outside C99 mode. For example,
1183 this is valid in GNU C though not valid in C89:
1187 struct foo @{int a[4];@};
1193 return f().a[index];
1199 @section Arithmetic on @code{void}- and Function-Pointers
1200 @cindex void pointers, arithmetic
1201 @cindex void, size of pointer to
1202 @cindex function pointers, arithmetic
1203 @cindex function, size of pointer to
1205 In GNU C, addition and subtraction operations are supported on pointers to
1206 @code{void} and on pointers to functions. This is done by treating the
1207 size of a @code{void} or of a function as 1.
1209 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1210 and on function types, and returns 1.
1212 @opindex Wpointer-arith
1213 The option @option{-Wpointer-arith} requests a warning if these extensions
1217 @section Non-Constant Initializers
1218 @cindex initializers, non-constant
1219 @cindex non-constant initializers
1221 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1222 automatic variable are not required to be constant expressions in GNU C@.
1223 Here is an example of an initializer with run-time varying elements:
1226 foo (float f, float g)
1228 float beat_freqs[2] = @{ f-g, f+g @};
1233 @node Compound Literals
1234 @section Compound Literals
1235 @cindex constructor expressions
1236 @cindex initializations in expressions
1237 @cindex structures, constructor expression
1238 @cindex expressions, constructor
1239 @cindex compound literals
1240 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1242 ISO C99 supports compound literals. A compound literal looks like
1243 a cast containing an initializer. Its value is an object of the
1244 type specified in the cast, containing the elements specified in
1245 the initializer; it is an lvalue. As an extension, GCC supports
1246 compound literals in C89 mode and in C++.
1248 Usually, the specified type is a structure. Assume that
1249 @code{struct foo} and @code{structure} are declared as shown:
1252 struct foo @{int a; char b[2];@} structure;
1256 Here is an example of constructing a @code{struct foo} with a compound literal:
1259 structure = ((struct foo) @{x + y, 'a', 0@});
1263 This is equivalent to writing the following:
1267 struct foo temp = @{x + y, 'a', 0@};
1272 You can also construct an array. If all the elements of the compound literal
1273 are (made up of) simple constant expressions, suitable for use in
1274 initializers of objects of static storage duration, then the compound
1275 literal can be coerced to a pointer to its first element and used in
1276 such an initializer, as shown here:
1279 char **foo = (char *[]) @{ "x", "y", "z" @};
1282 Compound literals for scalar types and union types are is
1283 also allowed, but then the compound literal is equivalent
1286 As a GNU extension, GCC allows initialization of objects with static storage
1287 duration by compound literals (which is not possible in ISO C99, because
1288 the initializer is not a constant).
1289 It is handled as if the object was initialized only with the bracket
1290 enclosed list if the types of the compound literal and the object match.
1291 The initializer list of the compound literal must be constant.
1292 If the object being initialized has array type of unknown size, the size is
1293 determined by compound literal size.
1296 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1297 static int y[] = (int []) @{1, 2, 3@};
1298 static int z[] = (int [3]) @{1@};
1302 The above lines are equivalent to the following:
1304 static struct foo x = @{1, 'a', 'b'@};
1305 static int y[] = @{1, 2, 3@};
1306 static int z[] = @{1, 0, 0@};
1309 @node Designated Inits
1310 @section Designated Initializers
1311 @cindex initializers with labeled elements
1312 @cindex labeled elements in initializers
1313 @cindex case labels in initializers
1314 @cindex designated initializers
1316 Standard C89 requires the elements of an initializer to appear in a fixed
1317 order, the same as the order of the elements in the array or structure
1320 In ISO C99 you can give the elements in any order, specifying the array
1321 indices or structure field names they apply to, and GNU C allows this as
1322 an extension in C89 mode as well. This extension is not
1323 implemented in GNU C++.
1325 To specify an array index, write
1326 @samp{[@var{index}] =} before the element value. For example,
1329 int a[6] = @{ [4] = 29, [2] = 15 @};
1336 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1340 The index values must be constant expressions, even if the array being
1341 initialized is automatic.
1343 An alternative syntax for this which has been obsolete since GCC 2.5 but
1344 GCC still accepts is to write @samp{[@var{index}]} before the element
1345 value, with no @samp{=}.
1347 To initialize a range of elements to the same value, write
1348 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1349 extension. For example,
1352 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1356 If the value in it has side-effects, the side-effects will happen only once,
1357 not for each initialized field by the range initializer.
1360 Note that the length of the array is the highest value specified
1363 In a structure initializer, specify the name of a field to initialize
1364 with @samp{.@var{fieldname} =} before the element value. For example,
1365 given the following structure,
1368 struct point @{ int x, y; @};
1372 the following initialization
1375 struct point p = @{ .y = yvalue, .x = xvalue @};
1382 struct point p = @{ xvalue, yvalue @};
1385 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1386 @samp{@var{fieldname}:}, as shown here:
1389 struct point p = @{ y: yvalue, x: xvalue @};
1393 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1394 @dfn{designator}. You can also use a designator (or the obsolete colon
1395 syntax) when initializing a union, to specify which element of the union
1396 should be used. For example,
1399 union foo @{ int i; double d; @};
1401 union foo f = @{ .d = 4 @};
1405 will convert 4 to a @code{double} to store it in the union using
1406 the second element. By contrast, casting 4 to type @code{union foo}
1407 would store it into the union as the integer @code{i}, since it is
1408 an integer. (@xref{Cast to Union}.)
1410 You can combine this technique of naming elements with ordinary C
1411 initialization of successive elements. Each initializer element that
1412 does not have a designator applies to the next consecutive element of the
1413 array or structure. For example,
1416 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1423 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1426 Labeling the elements of an array initializer is especially useful
1427 when the indices are characters or belong to an @code{enum} type.
1432 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1433 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1436 @cindex designator lists
1437 You can also write a series of @samp{.@var{fieldname}} and
1438 @samp{[@var{index}]} designators before an @samp{=} to specify a
1439 nested subobject to initialize; the list is taken relative to the
1440 subobject corresponding to the closest surrounding brace pair. For
1441 example, with the @samp{struct point} declaration above:
1444 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1448 If the same field is initialized multiple times, it will have value from
1449 the last initialization. If any such overridden initialization has
1450 side-effect, it is unspecified whether the side-effect happens or not.
1451 Currently, GCC will discard them and issue a warning.
1454 @section Case Ranges
1456 @cindex ranges in case statements
1458 You can specify a range of consecutive values in a single @code{case} label,
1462 case @var{low} ... @var{high}:
1466 This has the same effect as the proper number of individual @code{case}
1467 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1469 This feature is especially useful for ranges of ASCII character codes:
1475 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1476 it may be parsed wrong when you use it with integer values. For example,
1491 @section Cast to a Union Type
1492 @cindex cast to a union
1493 @cindex union, casting to a
1495 A cast to union type is similar to other casts, except that the type
1496 specified is a union type. You can specify the type either with
1497 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1498 a constructor though, not a cast, and hence does not yield an lvalue like
1499 normal casts. (@xref{Compound Literals}.)
1501 The types that may be cast to the union type are those of the members
1502 of the union. Thus, given the following union and variables:
1505 union foo @{ int i; double d; @};
1511 both @code{x} and @code{y} can be cast to type @code{union foo}.
1513 Using the cast as the right-hand side of an assignment to a variable of
1514 union type is equivalent to storing in a member of the union:
1519 u = (union foo) x @equiv{} u.i = x
1520 u = (union foo) y @equiv{} u.d = y
1523 You can also use the union cast as a function argument:
1526 void hack (union foo);
1528 hack ((union foo) x);
1531 @node Mixed Declarations
1532 @section Mixed Declarations and Code
1533 @cindex mixed declarations and code
1534 @cindex declarations, mixed with code
1535 @cindex code, mixed with declarations
1537 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1538 within compound statements. As an extension, GCC also allows this in
1539 C89 mode. For example, you could do:
1548 Each identifier is visible from where it is declared until the end of
1549 the enclosing block.
1551 @node Function Attributes
1552 @section Declaring Attributes of Functions
1553 @cindex function attributes
1554 @cindex declaring attributes of functions
1555 @cindex functions that never return
1556 @cindex functions that return more than once
1557 @cindex functions that have no side effects
1558 @cindex functions in arbitrary sections
1559 @cindex functions that behave like malloc
1560 @cindex @code{volatile} applied to function
1561 @cindex @code{const} applied to function
1562 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1563 @cindex functions with non-null pointer arguments
1564 @cindex functions that are passed arguments in registers on the 386
1565 @cindex functions that pop the argument stack on the 386
1566 @cindex functions that do not pop the argument stack on the 386
1568 In GNU C, you declare certain things about functions called in your program
1569 which help the compiler optimize function calls and check your code more
1572 The keyword @code{__attribute__} allows you to specify special
1573 attributes when making a declaration. This keyword is followed by an
1574 attribute specification inside double parentheses. The following
1575 attributes are currently defined for functions on all targets:
1576 @code{alloc_size}, @code{noreturn}, @code{returns_twice}, @code{noinline},
1577 @code{always_inline}, @code{flatten}, @code{pure}, @code{const},
1578 @code{nothrow}, @code{sentinel}, @code{format}, @code{format_arg},
1579 @code{no_instrument_function}, @code{section}, @code{constructor},
1580 @code{destructor}, @code{used}, @code{unused}, @code{deprecated},
1581 @code{weak}, @code{malloc}, @code{alias}, @code{warn_unused_result},
1582 @code{nonnull}, @code{gnu_inline} and @code{externally_visible},
1583 @code{hot}, @code{cold}.
1584 Several other attributes are defined for functions on particular target
1585 systems. Other attributes, including @code{section} are supported for
1586 variables declarations (@pxref{Variable Attributes}) and for types (@pxref{Type
1589 You may also specify attributes with @samp{__} preceding and following
1590 each keyword. This allows you to use them in header files without
1591 being concerned about a possible macro of the same name. For example,
1592 you may use @code{__noreturn__} instead of @code{noreturn}.
1594 @xref{Attribute Syntax}, for details of the exact syntax for using
1598 @c Keep this table alphabetized by attribute name. Treat _ as space.
1600 @item alias ("@var{target}")
1601 @cindex @code{alias} attribute
1602 The @code{alias} attribute causes the declaration to be emitted as an
1603 alias for another symbol, which must be specified. For instance,
1606 void __f () @{ /* @r{Do something.} */; @}
1607 void f () __attribute__ ((weak, alias ("__f")));
1610 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1611 mangled name for the target must be used. It is an error if @samp{__f}
1612 is not defined in the same translation unit.
1614 Not all target machines support this attribute.
1617 @cindex @code{alloc_size} attribute
1618 The @code{alloc_size} attribute is used to tell the compiler that the
1619 function return value points to memory, where the size is given by
1620 one or two of the functions parameters. GCC uses this
1621 information to improve the correctness of @code{__builtin_object_size}.
1623 The function parameter(s) denoting the allocated size are specified by
1624 one or two integer arguments supplied to the attribute. The allocated size
1625 is either the value of the single function argument specified or the product
1626 of the two function arguments specified. Argument numbering starts at
1632 void* my_calloc(size_t, size_t) __attribute__((alloc_size(1,2)))
1633 void my_realloc(void* size_t) __attribute__((alloc_size(2)))
1636 declares that my_calloc will return memory of the size given by
1637 the product of parameter 1 and 2 and that my_realloc will return memory
1638 of the size given by parameter 2.
1641 @cindex @code{always_inline} function attribute
1642 Generally, functions are not inlined unless optimization is specified.
1643 For functions declared inline, this attribute inlines the function even
1644 if no optimization level was specified.
1647 @cindex @code{gnu_inline} function attribute
1648 This attribute should be used with a function which is also declared
1649 with the @code{inline} keyword. It directs GCC to treat the function
1650 as if it were defined in gnu89 mode even when compiling in C99 or
1653 If the function is declared @code{extern}, then this definition of the
1654 function is used only for inlining. In no case is the function
1655 compiled as a standalone function, not even if you take its address
1656 explicitly. Such an address becomes an external reference, as if you
1657 had only declared the function, and had not defined it. This has
1658 almost the effect of a macro. The way to use this is to put a
1659 function definition in a header file with this attribute, and put
1660 another copy of the function, without @code{extern}, in a library
1661 file. The definition in the header file will cause most calls to the
1662 function to be inlined. If any uses of the function remain, they will
1663 refer to the single copy in the library. Note that the two
1664 definitions of the functions need not be precisely the same, although
1665 if they do not have the same effect your program may behave oddly.
1667 If the function is neither @code{extern} nor @code{static}, then the
1668 function is compiled as a standalone function, as well as being
1669 inlined where possible.
1671 This is how GCC traditionally handled functions declared
1672 @code{inline}. Since ISO C99 specifies a different semantics for
1673 @code{inline}, this function attribute is provided as a transition
1674 measure and as a useful feature in its own right. This attribute is
1675 available in GCC 4.1.3 and later. It is available if either of the
1676 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1677 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1678 Function is As Fast As a Macro}.
1680 @cindex @code{flatten} function attribute
1682 Generally, inlining into a function is limited. For a function marked with
1683 this attribute, every call inside this function will be inlined, if possible.
1684 Whether the function itself is considered for inlining depends on its size and
1685 the current inlining parameters. The @code{flatten} attribute only works
1686 reliably in unit-at-a-time mode.
1689 @cindex functions that do pop the argument stack on the 386
1691 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1692 assume that the calling function will pop off the stack space used to
1693 pass arguments. This is
1694 useful to override the effects of the @option{-mrtd} switch.
1697 @cindex @code{const} function attribute
1698 Many functions do not examine any values except their arguments, and
1699 have no effects except the return value. Basically this is just slightly
1700 more strict class than the @code{pure} attribute below, since function is not
1701 allowed to read global memory.
1703 @cindex pointer arguments
1704 Note that a function that has pointer arguments and examines the data
1705 pointed to must @emph{not} be declared @code{const}. Likewise, a
1706 function that calls a non-@code{const} function usually must not be
1707 @code{const}. It does not make sense for a @code{const} function to
1710 The attribute @code{const} is not implemented in GCC versions earlier
1711 than 2.5. An alternative way to declare that a function has no side
1712 effects, which works in the current version and in some older versions,
1716 typedef int intfn ();
1718 extern const intfn square;
1721 This approach does not work in GNU C++ from 2.6.0 on, since the language
1722 specifies that the @samp{const} must be attached to the return value.
1726 @itemx constructor (@var{priority})
1727 @itemx destructor (@var{priority})
1728 @cindex @code{constructor} function attribute
1729 @cindex @code{destructor} function attribute
1730 The @code{constructor} attribute causes the function to be called
1731 automatically before execution enters @code{main ()}. Similarly, the
1732 @code{destructor} attribute causes the function to be called
1733 automatically after @code{main ()} has completed or @code{exit ()} has
1734 been called. Functions with these attributes are useful for
1735 initializing data that will be used implicitly during the execution of
1738 You may provide an optional integer priority to control the order in
1739 which constructor and destructor functions are run. A constructor
1740 with a smaller priority number runs before a constructor with a larger
1741 priority number; the opposite relationship holds for destructors. So,
1742 if you have a constructor that allocates a resource and a destructor
1743 that deallocates the same resource, both functions typically have the
1744 same priority. The priorities for constructor and destructor
1745 functions are the same as those specified for namespace-scope C++
1746 objects (@pxref{C++ Attributes}).
1748 These attributes are not currently implemented for Objective-C@.
1751 @cindex @code{deprecated} attribute.
1752 The @code{deprecated} attribute results in a warning if the function
1753 is used anywhere in the source file. This is useful when identifying
1754 functions that are expected to be removed in a future version of a
1755 program. The warning also includes the location of the declaration
1756 of the deprecated function, to enable users to easily find further
1757 information about why the function is deprecated, or what they should
1758 do instead. Note that the warnings only occurs for uses:
1761 int old_fn () __attribute__ ((deprecated));
1763 int (*fn_ptr)() = old_fn;
1766 results in a warning on line 3 but not line 2.
1768 The @code{deprecated} attribute can also be used for variables and
1769 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1772 @cindex @code{__declspec(dllexport)}
1773 On Microsoft Windows targets and Symbian OS targets the
1774 @code{dllexport} attribute causes the compiler to provide a global
1775 pointer to a pointer in a DLL, so that it can be referenced with the
1776 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1777 name is formed by combining @code{_imp__} and the function or variable
1780 You can use @code{__declspec(dllexport)} as a synonym for
1781 @code{__attribute__ ((dllexport))} for compatibility with other
1784 On systems that support the @code{visibility} attribute, this
1785 attribute also implies ``default'' visibility. It is an error to
1786 explicitly specify any other visibility.
1788 Currently, the @code{dllexport} attribute is ignored for inlined
1789 functions, unless the @option{-fkeep-inline-functions} flag has been
1790 used. The attribute is also ignored for undefined symbols.
1792 When applied to C++ classes, the attribute marks defined non-inlined
1793 member functions and static data members as exports. Static consts
1794 initialized in-class are not marked unless they are also defined
1797 For Microsoft Windows targets there are alternative methods for
1798 including the symbol in the DLL's export table such as using a
1799 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1800 the @option{--export-all} linker flag.
1803 @cindex @code{__declspec(dllimport)}
1804 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1805 attribute causes the compiler to reference a function or variable via
1806 a global pointer to a pointer that is set up by the DLL exporting the
1807 symbol. The attribute implies @code{extern}. On Microsoft Windows
1808 targets, the pointer name is formed by combining @code{_imp__} and the
1809 function or variable name.
1811 You can use @code{__declspec(dllimport)} as a synonym for
1812 @code{__attribute__ ((dllimport))} for compatibility with other
1815 On systems that support the @code{visibility} attribute, this
1816 attribute also implies ``default'' visibility. It is an error to
1817 explicitly specify any other visibility.
1819 Currently, the attribute is ignored for inlined functions. If the
1820 attribute is applied to a symbol @emph{definition}, an error is reported.
1821 If a symbol previously declared @code{dllimport} is later defined, the
1822 attribute is ignored in subsequent references, and a warning is emitted.
1823 The attribute is also overridden by a subsequent declaration as
1826 When applied to C++ classes, the attribute marks non-inlined
1827 member functions and static data members as imports. However, the
1828 attribute is ignored for virtual methods to allow creation of vtables
1831 On the SH Symbian OS target the @code{dllimport} attribute also has
1832 another affect---it can cause the vtable and run-time type information
1833 for a class to be exported. This happens when the class has a
1834 dllimport'ed constructor or a non-inline, non-pure virtual function
1835 and, for either of those two conditions, the class also has a inline
1836 constructor or destructor and has a key function that is defined in
1837 the current translation unit.
1839 For Microsoft Windows based targets the use of the @code{dllimport}
1840 attribute on functions is not necessary, but provides a small
1841 performance benefit by eliminating a thunk in the DLL@. The use of the
1842 @code{dllimport} attribute on imported variables was required on older
1843 versions of the GNU linker, but can now be avoided by passing the
1844 @option{--enable-auto-import} switch to the GNU linker. As with
1845 functions, using the attribute for a variable eliminates a thunk in
1848 One drawback to using this attribute is that a pointer to a function
1849 or variable marked as @code{dllimport} cannot be used as a constant
1850 address. On Microsoft Windows targets, the attribute can be disabled
1851 for functions by setting the @option{-mnop-fun-dllimport} flag.
1854 @cindex eight bit data on the H8/300, H8/300H, and H8S
1855 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1856 variable should be placed into the eight bit data section.
1857 The compiler will generate more efficient code for certain operations
1858 on data in the eight bit data area. Note the eight bit data area is limited to
1861 You must use GAS and GLD from GNU binutils version 2.7 or later for
1862 this attribute to work correctly.
1864 @item exception_handler
1865 @cindex exception handler functions on the Blackfin processor
1866 Use this attribute on the Blackfin to indicate that the specified function
1867 is an exception handler. The compiler will generate function entry and
1868 exit sequences suitable for use in an exception handler when this
1869 attribute is present.
1872 @cindex functions which handle memory bank switching
1873 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1874 use a calling convention that takes care of switching memory banks when
1875 entering and leaving a function. This calling convention is also the
1876 default when using the @option{-mlong-calls} option.
1878 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1879 to call and return from a function.
1881 On 68HC11 the compiler will generate a sequence of instructions
1882 to invoke a board-specific routine to switch the memory bank and call the
1883 real function. The board-specific routine simulates a @code{call}.
1884 At the end of a function, it will jump to a board-specific routine
1885 instead of using @code{rts}. The board-specific return routine simulates
1889 @cindex functions that pop the argument stack on the 386
1890 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1891 pass the first argument (if of integral type) in the register ECX and
1892 the second argument (if of integral type) in the register EDX@. Subsequent
1893 and other typed arguments are passed on the stack. The called function will
1894 pop the arguments off the stack. If the number of arguments is variable all
1895 arguments are pushed on the stack.
1897 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1898 @cindex @code{format} function attribute
1900 The @code{format} attribute specifies that a function takes @code{printf},
1901 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1902 should be type-checked against a format string. For example, the
1907 my_printf (void *my_object, const char *my_format, ...)
1908 __attribute__ ((format (printf, 2, 3)));
1912 causes the compiler to check the arguments in calls to @code{my_printf}
1913 for consistency with the @code{printf} style format string argument
1916 The parameter @var{archetype} determines how the format string is
1917 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1918 or @code{strfmon}. (You can also use @code{__printf__},
1919 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1920 parameter @var{string-index} specifies which argument is the format
1921 string argument (starting from 1), while @var{first-to-check} is the
1922 number of the first argument to check against the format string. For
1923 functions where the arguments are not available to be checked (such as
1924 @code{vprintf}), specify the third parameter as zero. In this case the
1925 compiler only checks the format string for consistency. For
1926 @code{strftime} formats, the third parameter is required to be zero.
1927 Since non-static C++ methods have an implicit @code{this} argument, the
1928 arguments of such methods should be counted from two, not one, when
1929 giving values for @var{string-index} and @var{first-to-check}.
1931 In the example above, the format string (@code{my_format}) is the second
1932 argument of the function @code{my_print}, and the arguments to check
1933 start with the third argument, so the correct parameters for the format
1934 attribute are 2 and 3.
1936 @opindex ffreestanding
1937 @opindex fno-builtin
1938 The @code{format} attribute allows you to identify your own functions
1939 which take format strings as arguments, so that GCC can check the
1940 calls to these functions for errors. The compiler always (unless
1941 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1942 for the standard library functions @code{printf}, @code{fprintf},
1943 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1944 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1945 warnings are requested (using @option{-Wformat}), so there is no need to
1946 modify the header file @file{stdio.h}. In C99 mode, the functions
1947 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1948 @code{vsscanf} are also checked. Except in strictly conforming C
1949 standard modes, the X/Open function @code{strfmon} is also checked as
1950 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1951 @xref{C Dialect Options,,Options Controlling C Dialect}.
1953 The target may provide additional types of format checks.
1954 @xref{Target Format Checks,,Format Checks Specific to Particular
1957 @item format_arg (@var{string-index})
1958 @cindex @code{format_arg} function attribute
1959 @opindex Wformat-nonliteral
1960 The @code{format_arg} attribute specifies that a function takes a format
1961 string for a @code{printf}, @code{scanf}, @code{strftime} or
1962 @code{strfmon} style function and modifies it (for example, to translate
1963 it into another language), so the result can be passed to a
1964 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1965 function (with the remaining arguments to the format function the same
1966 as they would have been for the unmodified string). For example, the
1971 my_dgettext (char *my_domain, const char *my_format)
1972 __attribute__ ((format_arg (2)));
1976 causes the compiler to check the arguments in calls to a @code{printf},
1977 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1978 format string argument is a call to the @code{my_dgettext} function, for
1979 consistency with the format string argument @code{my_format}. If the
1980 @code{format_arg} attribute had not been specified, all the compiler
1981 could tell in such calls to format functions would be that the format
1982 string argument is not constant; this would generate a warning when
1983 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1984 without the attribute.
1986 The parameter @var{string-index} specifies which argument is the format
1987 string argument (starting from one). Since non-static C++ methods have
1988 an implicit @code{this} argument, the arguments of such methods should
1989 be counted from two.
1991 The @code{format-arg} attribute allows you to identify your own
1992 functions which modify format strings, so that GCC can check the
1993 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1994 type function whose operands are a call to one of your own function.
1995 The compiler always treats @code{gettext}, @code{dgettext}, and
1996 @code{dcgettext} in this manner except when strict ISO C support is
1997 requested by @option{-ansi} or an appropriate @option{-std} option, or
1998 @option{-ffreestanding} or @option{-fno-builtin}
1999 is used. @xref{C Dialect Options,,Options
2000 Controlling C Dialect}.
2002 @item function_vector
2003 @cindex calling functions through the function vector on H8/300, M16C, and M32C processors
2004 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2005 function should be called through the function vector. Calling a
2006 function through the function vector will reduce code size, however;
2007 the function vector has a limited size (maximum 128 entries on the H8/300
2008 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2010 You must use GAS and GLD from GNU binutils version 2.7 or later for
2011 this attribute to work correctly.
2013 On M16C/M32C targets, the @code{function_vector} attribute declares a
2014 special page subroutine call function. Use of this attribute reduces
2015 the code size by 2 bytes for each call generated to the
2016 subroutine. The argument to the attribute is the vector number entry
2017 from the special page vector table which contains the 16 low-order
2018 bits of the subroutine's entry address. Each vector table has special
2019 page number (18 to 255) which are used in @code{jsrs} instruction.
2020 Jump addresses of the routines are generated by adding 0x0F0000 (in
2021 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
2022 byte addresses set in the vector table. Therefore you need to ensure
2023 that all the special page vector routines should get mapped within the
2024 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
2027 In the following example 2 bytes will be saved for each call to
2028 function @code{foo}.
2031 void foo (void) __attribute__((function_vector(0x18)));
2042 If functions are defined in one file and are called in another file,
2043 then be sure to write this declaration in both files.
2045 This attribute is ignored for R8C target.
2048 @cindex interrupt handler functions
2049 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, m68k, MS1,
2050 and Xstormy16 ports to indicate that the specified function is an
2051 interrupt handler. The compiler will generate function entry and exit
2052 sequences suitable for use in an interrupt handler when this attribute
2055 Note, interrupt handlers for the Blackfin, H8/300, H8/300H, H8S, and
2056 SH processors can be specified via the @code{interrupt_handler} attribute.
2058 Note, on the AVR, interrupts will be enabled inside the function.
2060 Note, for the ARM, you can specify the kind of interrupt to be handled by
2061 adding an optional parameter to the interrupt attribute like this:
2064 void f () __attribute__ ((interrupt ("IRQ")));
2067 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2069 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2070 may be called with a word aligned stack pointer.
2072 @item interrupt_handler
2073 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2074 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2075 indicate that the specified function is an interrupt handler. The compiler
2076 will generate function entry and exit sequences suitable for use in an
2077 interrupt handler when this attribute is present.
2079 @item interrupt_thread
2080 @cindex interrupt thread functions on fido
2081 Use this attribute on fido, a subarchitecture of the m68k, to indicate
2082 that the specified function is an interrupt handler that is designed
2083 to run as a thread. The compiler omits generate prologue/epilogue
2084 sequences and replaces the return instruction with a @code{sleep}
2085 instruction. This attribute is available only on fido.
2088 @cindex User stack pointer in interrupts on the Blackfin
2089 When used together with @code{interrupt_handler}, @code{exception_handler}
2090 or @code{nmi_handler}, code will be generated to load the stack pointer
2091 from the USP register in the function prologue.
2093 @item long_call/short_call
2094 @cindex indirect calls on ARM
2095 This attribute specifies how a particular function is called on
2096 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2097 command line switch and @code{#pragma long_calls} settings. The
2098 @code{long_call} attribute indicates that the function might be far
2099 away from the call site and require a different (more expensive)
2100 calling sequence. The @code{short_call} attribute always places
2101 the offset to the function from the call site into the @samp{BL}
2102 instruction directly.
2104 @item longcall/shortcall
2105 @cindex functions called via pointer on the RS/6000 and PowerPC
2106 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2107 indicates that the function might be far away from the call site and
2108 require a different (more expensive) calling sequence. The
2109 @code{shortcall} attribute indicates that the function is always close
2110 enough for the shorter calling sequence to be used. These attributes
2111 override both the @option{-mlongcall} switch and, on the RS/6000 and
2112 PowerPC, the @code{#pragma longcall} setting.
2114 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2115 calls are necessary.
2117 @item long_call/near/far
2118 @cindex indirect calls on MIPS
2119 These attributes specify how a particular function is called on MIPS@.
2120 The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
2121 command-line switch. The @code{long_call} and @code{far} attributes are
2122 synonyms, and cause the compiler to always call
2123 the function by first loading its address into a register, and then using
2124 the contents of that register. The @code{near} attribute has the opposite
2125 effect; it specifies that non-PIC calls should be made using the more
2126 efficient @code{jal} instruction.
2129 @cindex @code{malloc} attribute
2130 The @code{malloc} attribute is used to tell the compiler that a function
2131 may be treated as if any non-@code{NULL} pointer it returns cannot
2132 alias any other pointer valid when the function returns.
2133 This will often improve optimization.
2134 Standard functions with this property include @code{malloc} and
2135 @code{calloc}. @code{realloc}-like functions have this property as
2136 long as the old pointer is never referred to (including comparing it
2137 to the new pointer) after the function returns a non-@code{NULL}
2140 @item model (@var{model-name})
2141 @cindex function addressability on the M32R/D
2142 @cindex variable addressability on the IA-64
2144 On the M32R/D, use this attribute to set the addressability of an
2145 object, and of the code generated for a function. The identifier
2146 @var{model-name} is one of @code{small}, @code{medium}, or
2147 @code{large}, representing each of the code models.
2149 Small model objects live in the lower 16MB of memory (so that their
2150 addresses can be loaded with the @code{ld24} instruction), and are
2151 callable with the @code{bl} instruction.
2153 Medium model objects may live anywhere in the 32-bit address space (the
2154 compiler will generate @code{seth/add3} instructions to load their addresses),
2155 and are callable with the @code{bl} instruction.
2157 Large model objects may live anywhere in the 32-bit address space (the
2158 compiler will generate @code{seth/add3} instructions to load their addresses),
2159 and may not be reachable with the @code{bl} instruction (the compiler will
2160 generate the much slower @code{seth/add3/jl} instruction sequence).
2162 On IA-64, use this attribute to set the addressability of an object.
2163 At present, the only supported identifier for @var{model-name} is
2164 @code{small}, indicating addressability via ``small'' (22-bit)
2165 addresses (so that their addresses can be loaded with the @code{addl}
2166 instruction). Caveat: such addressing is by definition not position
2167 independent and hence this attribute must not be used for objects
2168 defined by shared libraries.
2171 @cindex function without a prologue/epilogue code
2172 Use this attribute on the ARM, AVR, C4x, IP2K and SPU ports to indicate that
2173 the specified function does not need prologue/epilogue sequences generated by
2174 the compiler. It is up to the programmer to provide these sequences.
2177 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2178 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2179 use the normal calling convention based on @code{jsr} and @code{rts}.
2180 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2184 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2185 Use this attribute together with @code{interrupt_handler},
2186 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2187 entry code should enable nested interrupts or exceptions.
2190 @cindex NMI handler functions on the Blackfin processor
2191 Use this attribute on the Blackfin to indicate that the specified function
2192 is an NMI handler. The compiler will generate function entry and
2193 exit sequences suitable for use in an NMI handler when this
2194 attribute is present.
2196 @item no_instrument_function
2197 @cindex @code{no_instrument_function} function attribute
2198 @opindex finstrument-functions
2199 If @option{-finstrument-functions} is given, profiling function calls will
2200 be generated at entry and exit of most user-compiled functions.
2201 Functions with this attribute will not be so instrumented.
2204 @cindex @code{noinline} function attribute
2205 This function attribute prevents a function from being considered for
2208 @item nonnull (@var{arg-index}, @dots{})
2209 @cindex @code{nonnull} function attribute
2210 The @code{nonnull} attribute specifies that some function parameters should
2211 be non-null pointers. For instance, the declaration:
2215 my_memcpy (void *dest, const void *src, size_t len)
2216 __attribute__((nonnull (1, 2)));
2220 causes the compiler to check that, in calls to @code{my_memcpy},
2221 arguments @var{dest} and @var{src} are non-null. If the compiler
2222 determines that a null pointer is passed in an argument slot marked
2223 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2224 is issued. The compiler may also choose to make optimizations based
2225 on the knowledge that certain function arguments will not be null.
2227 If no argument index list is given to the @code{nonnull} attribute,
2228 all pointer arguments are marked as non-null. To illustrate, the
2229 following declaration is equivalent to the previous example:
2233 my_memcpy (void *dest, const void *src, size_t len)
2234 __attribute__((nonnull));
2238 @cindex @code{noreturn} function attribute
2239 A few standard library functions, such as @code{abort} and @code{exit},
2240 cannot return. GCC knows this automatically. Some programs define
2241 their own functions that never return. You can declare them
2242 @code{noreturn} to tell the compiler this fact. For example,
2246 void fatal () __attribute__ ((noreturn));
2249 fatal (/* @r{@dots{}} */)
2251 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2257 The @code{noreturn} keyword tells the compiler to assume that
2258 @code{fatal} cannot return. It can then optimize without regard to what
2259 would happen if @code{fatal} ever did return. This makes slightly
2260 better code. More importantly, it helps avoid spurious warnings of
2261 uninitialized variables.
2263 The @code{noreturn} keyword does not affect the exceptional path when that
2264 applies: a @code{noreturn}-marked function may still return to the caller
2265 by throwing an exception or calling @code{longjmp}.
2267 Do not assume that registers saved by the calling function are
2268 restored before calling the @code{noreturn} function.
2270 It does not make sense for a @code{noreturn} function to have a return
2271 type other than @code{void}.
2273 The attribute @code{noreturn} is not implemented in GCC versions
2274 earlier than 2.5. An alternative way to declare that a function does
2275 not return, which works in the current version and in some older
2276 versions, is as follows:
2279 typedef void voidfn ();
2281 volatile voidfn fatal;
2284 This approach does not work in GNU C++.
2287 @cindex @code{nothrow} function attribute
2288 The @code{nothrow} attribute is used to inform the compiler that a
2289 function cannot throw an exception. For example, most functions in
2290 the standard C library can be guaranteed not to throw an exception
2291 with the notable exceptions of @code{qsort} and @code{bsearch} that
2292 take function pointer arguments. The @code{nothrow} attribute is not
2293 implemented in GCC versions earlier than 3.3.
2296 @cindex @code{pure} function attribute
2297 Many functions have no effects except the return value and their
2298 return value depends only on the parameters and/or global variables.
2299 Such a function can be subject
2300 to common subexpression elimination and loop optimization just as an
2301 arithmetic operator would be. These functions should be declared
2302 with the attribute @code{pure}. For example,
2305 int square (int) __attribute__ ((pure));
2309 says that the hypothetical function @code{square} is safe to call
2310 fewer times than the program says.
2312 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2313 Interesting non-pure functions are functions with infinite loops or those
2314 depending on volatile memory or other system resource, that may change between
2315 two consecutive calls (such as @code{feof} in a multithreading environment).
2317 The attribute @code{pure} is not implemented in GCC versions earlier
2321 @cindex @code{hot} function attribute
2322 The @code{hot} attribute is used to inform the compiler that a function is a
2323 hot spot of the compiled program. The function is optimized more aggressively
2324 and on many target it is placed into special subsection of the text section so
2325 all hot functions appears close together improving locality.
2327 When profile feedback is available, via @option{-fprofile-use}, hot functions
2328 are automatically detected and this attribute is ignored.
2330 The @code{hot} attribute is not implemented in GCC versions earlier than 4.3.
2333 @cindex @code{cold} function attribute
2334 The @code{cold} attribute is used to inform the compiler that a function is
2335 unlikely executed. The function is optimized for size rather than speed and on
2336 many targets it is placed into special subsection of the text section so all
2337 cold functions appears close together improving code locality of non-cold parts
2338 of program. The paths leading to call of cold functions within code are marked
2339 as unlikely by the branch prediction mechanism. It is thus useful to mark
2340 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2341 improve optimization of hot functions that do call marked functions in rare
2344 When profile feedback is available, via @option{-fprofile-use}, hot functions
2345 are automatically detected and this attribute is ignored.
2347 The @code{hot} attribute is not implemented in GCC versions earlier than 4.3.
2349 @item regparm (@var{number})
2350 @cindex @code{regparm} attribute
2351 @cindex functions that are passed arguments in registers on the 386
2352 On the Intel 386, the @code{regparm} attribute causes the compiler to
2353 pass arguments number one to @var{number} if they are of integral type
2354 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2355 take a variable number of arguments will continue to be passed all of their
2356 arguments on the stack.
2358 Beware that on some ELF systems this attribute is unsuitable for
2359 global functions in shared libraries with lazy binding (which is the
2360 default). Lazy binding will send the first call via resolving code in
2361 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2362 per the standard calling conventions. Solaris 8 is affected by this.
2363 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2364 safe since the loaders there save all registers. (Lazy binding can be
2365 disabled with the linker or the loader if desired, to avoid the
2369 @cindex @code{sseregparm} attribute
2370 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2371 causes the compiler to pass up to 3 floating point arguments in
2372 SSE registers instead of on the stack. Functions that take a
2373 variable number of arguments will continue to pass all of their
2374 floating point arguments on the stack.
2376 @item force_align_arg_pointer
2377 @cindex @code{force_align_arg_pointer} attribute
2378 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2379 applied to individual function definitions, generating an alternate
2380 prologue and epilogue that realigns the runtime stack. This supports
2381 mixing legacy codes that run with a 4-byte aligned stack with modern
2382 codes that keep a 16-byte stack for SSE compatibility. The alternate
2383 prologue and epilogue are slower and bigger than the regular ones, and
2384 the alternate prologue requires a scratch register; this lowers the
2385 number of registers available if used in conjunction with the
2386 @code{regparm} attribute. The @code{force_align_arg_pointer}
2387 attribute is incompatible with nested functions; this is considered a
2391 @cindex @code{returns_twice} attribute
2392 The @code{returns_twice} attribute tells the compiler that a function may
2393 return more than one time. The compiler will ensure that all registers
2394 are dead before calling such a function and will emit a warning about
2395 the variables that may be clobbered after the second return from the
2396 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2397 The @code{longjmp}-like counterpart of such function, if any, might need
2398 to be marked with the @code{noreturn} attribute.
2401 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2402 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2403 all registers except the stack pointer should be saved in the prologue
2404 regardless of whether they are used or not.
2406 @item section ("@var{section-name}")
2407 @cindex @code{section} function attribute
2408 Normally, the compiler places the code it generates in the @code{text} section.
2409 Sometimes, however, you need additional sections, or you need certain
2410 particular functions to appear in special sections. The @code{section}
2411 attribute specifies that a function lives in a particular section.
2412 For example, the declaration:
2415 extern void foobar (void) __attribute__ ((section ("bar")));
2419 puts the function @code{foobar} in the @code{bar} section.
2421 Some file formats do not support arbitrary sections so the @code{section}
2422 attribute is not available on all platforms.
2423 If you need to map the entire contents of a module to a particular
2424 section, consider using the facilities of the linker instead.
2427 @cindex @code{sentinel} function attribute
2428 This function attribute ensures that a parameter in a function call is
2429 an explicit @code{NULL}. The attribute is only valid on variadic
2430 functions. By default, the sentinel is located at position zero, the
2431 last parameter of the function call. If an optional integer position
2432 argument P is supplied to the attribute, the sentinel must be located at
2433 position P counting backwards from the end of the argument list.
2436 __attribute__ ((sentinel))
2438 __attribute__ ((sentinel(0)))
2441 The attribute is automatically set with a position of 0 for the built-in
2442 functions @code{execl} and @code{execlp}. The built-in function
2443 @code{execle} has the attribute set with a position of 1.
2445 A valid @code{NULL} in this context is defined as zero with any pointer
2446 type. If your system defines the @code{NULL} macro with an integer type
2447 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2448 with a copy that redefines NULL appropriately.
2450 The warnings for missing or incorrect sentinels are enabled with
2454 See long_call/short_call.
2457 See longcall/shortcall.
2460 @cindex signal handler functions on the AVR processors
2461 Use this attribute on the AVR to indicate that the specified
2462 function is a signal handler. The compiler will generate function
2463 entry and exit sequences suitable for use in a signal handler when this
2464 attribute is present. Interrupts will be disabled inside the function.
2467 Use this attribute on the SH to indicate an @code{interrupt_handler}
2468 function should switch to an alternate stack. It expects a string
2469 argument that names a global variable holding the address of the
2474 void f () __attribute__ ((interrupt_handler,
2475 sp_switch ("alt_stack")));
2479 @cindex functions that pop the argument stack on the 386
2480 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2481 assume that the called function will pop off the stack space used to
2482 pass arguments, unless it takes a variable number of arguments.
2485 @cindex tiny data section on the H8/300H and H8S
2486 Use this attribute on the H8/300H and H8S to indicate that the specified
2487 variable should be placed into the tiny data section.
2488 The compiler will generate more efficient code for loads and stores
2489 on data in the tiny data section. Note the tiny data area is limited to
2490 slightly under 32kbytes of data.
2493 Use this attribute on the SH for an @code{interrupt_handler} to return using
2494 @code{trapa} instead of @code{rte}. This attribute expects an integer
2495 argument specifying the trap number to be used.
2498 @cindex @code{unused} attribute.
2499 This attribute, attached to a function, means that the function is meant
2500 to be possibly unused. GCC will not produce a warning for this
2504 @cindex @code{used} attribute.
2505 This attribute, attached to a function, means that code must be emitted
2506 for the function even if it appears that the function is not referenced.
2507 This is useful, for example, when the function is referenced only in
2511 @cindex @code{version_id} attribute on IA64 HP-UX
2512 This attribute, attached to a global variable or function, renames a
2513 symbol to contain a version string, thus allowing for function level
2514 versioning. HP-UX system header files may use version level functioning
2515 for some system calls.
2518 extern int foo () __attribute__((version_id ("20040821")));
2521 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
2523 @item visibility ("@var{visibility_type}")
2524 @cindex @code{visibility} attribute
2525 This attribute affects the linkage of the declaration to which it is attached.
2526 There are four supported @var{visibility_type} values: default,
2527 hidden, protected or internal visibility.
2530 void __attribute__ ((visibility ("protected")))
2531 f () @{ /* @r{Do something.} */; @}
2532 int i __attribute__ ((visibility ("hidden")));
2535 The possible values of @var{visibility_type} correspond to the
2536 visibility settings in the ELF gABI.
2539 @c keep this list of visibilities in alphabetical order.
2542 Default visibility is the normal case for the object file format.
2543 This value is available for the visibility attribute to override other
2544 options that may change the assumed visibility of entities.
2546 On ELF, default visibility means that the declaration is visible to other
2547 modules and, in shared libraries, means that the declared entity may be
2550 On Darwin, default visibility means that the declaration is visible to
2553 Default visibility corresponds to ``external linkage'' in the language.
2556 Hidden visibility indicates that the entity declared will have a new
2557 form of linkage, which we'll call ``hidden linkage''. Two
2558 declarations of an object with hidden linkage refer to the same object
2559 if they are in the same shared object.
2562 Internal visibility is like hidden visibility, but with additional
2563 processor specific semantics. Unless otherwise specified by the
2564 psABI, GCC defines internal visibility to mean that a function is
2565 @emph{never} called from another module. Compare this with hidden
2566 functions which, while they cannot be referenced directly by other
2567 modules, can be referenced indirectly via function pointers. By
2568 indicating that a function cannot be called from outside the module,
2569 GCC may for instance omit the load of a PIC register since it is known
2570 that the calling function loaded the correct value.
2573 Protected visibility is like default visibility except that it
2574 indicates that references within the defining module will bind to the
2575 definition in that module. That is, the declared entity cannot be
2576 overridden by another module.
2580 All visibilities are supported on many, but not all, ELF targets
2581 (supported when the assembler supports the @samp{.visibility}
2582 pseudo-op). Default visibility is supported everywhere. Hidden
2583 visibility is supported on Darwin targets.
2585 The visibility attribute should be applied only to declarations which
2586 would otherwise have external linkage. The attribute should be applied
2587 consistently, so that the same entity should not be declared with
2588 different settings of the attribute.
2590 In C++, the visibility attribute applies to types as well as functions
2591 and objects, because in C++ types have linkage. A class must not have
2592 greater visibility than its non-static data member types and bases,
2593 and class members default to the visibility of their class. Also, a
2594 declaration without explicit visibility is limited to the visibility
2597 In C++, you can mark member functions and static member variables of a
2598 class with the visibility attribute. This is useful if if you know a
2599 particular method or static member variable should only be used from
2600 one shared object; then you can mark it hidden while the rest of the
2601 class has default visibility. Care must be taken to avoid breaking
2602 the One Definition Rule; for example, it is usually not useful to mark
2603 an inline method as hidden without marking the whole class as hidden.
2605 A C++ namespace declaration can also have the visibility attribute.
2606 This attribute applies only to the particular namespace body, not to
2607 other definitions of the same namespace; it is equivalent to using
2608 @samp{#pragma GCC visibility} before and after the namespace
2609 definition (@pxref{Visibility Pragmas}).
2611 In C++, if a template argument has limited visibility, this
2612 restriction is implicitly propagated to the template instantiation.
2613 Otherwise, template instantiations and specializations default to the
2614 visibility of their template.
2616 If both the template and enclosing class have explicit visibility, the
2617 visibility from the template is used.
2619 @item warn_unused_result
2620 @cindex @code{warn_unused_result} attribute
2621 The @code{warn_unused_result} attribute causes a warning to be emitted
2622 if a caller of the function with this attribute does not use its
2623 return value. This is useful for functions where not checking
2624 the result is either a security problem or always a bug, such as
2628 int fn () __attribute__ ((warn_unused_result));
2631 if (fn () < 0) return -1;
2637 results in warning on line 5.
2640 @cindex @code{weak} attribute
2641 The @code{weak} attribute causes the declaration to be emitted as a weak
2642 symbol rather than a global. This is primarily useful in defining
2643 library functions which can be overridden in user code, though it can
2644 also be used with non-function declarations. Weak symbols are supported
2645 for ELF targets, and also for a.out targets when using the GNU assembler
2649 @itemx weakref ("@var{target}")
2650 @cindex @code{weakref} attribute
2651 The @code{weakref} attribute marks a declaration as a weak reference.
2652 Without arguments, it should be accompanied by an @code{alias} attribute
2653 naming the target symbol. Optionally, the @var{target} may be given as
2654 an argument to @code{weakref} itself. In either case, @code{weakref}
2655 implicitly marks the declaration as @code{weak}. Without a
2656 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2657 @code{weakref} is equivalent to @code{weak}.
2660 static int x() __attribute__ ((weakref ("y")));
2661 /* is equivalent to... */
2662 static int x() __attribute__ ((weak, weakref, alias ("y")));
2664 static int x() __attribute__ ((weakref));
2665 static int x() __attribute__ ((alias ("y")));
2668 A weak reference is an alias that does not by itself require a
2669 definition to be given for the target symbol. If the target symbol is
2670 only referenced through weak references, then the becomes a @code{weak}
2671 undefined symbol. If it is directly referenced, however, then such
2672 strong references prevail, and a definition will be required for the
2673 symbol, not necessarily in the same translation unit.
2675 The effect is equivalent to moving all references to the alias to a
2676 separate translation unit, renaming the alias to the aliased symbol,
2677 declaring it as weak, compiling the two separate translation units and
2678 performing a reloadable link on them.
2680 At present, a declaration to which @code{weakref} is attached can
2681 only be @code{static}.
2683 @item externally_visible
2684 @cindex @code{externally_visible} attribute.
2685 This attribute, attached to a global variable or function nullify
2686 effect of @option{-fwhole-program} command line option, so the object
2687 remain visible outside the current compilation unit
2691 You can specify multiple attributes in a declaration by separating them
2692 by commas within the double parentheses or by immediately following an
2693 attribute declaration with another attribute declaration.
2695 @cindex @code{#pragma}, reason for not using
2696 @cindex pragma, reason for not using
2697 Some people object to the @code{__attribute__} feature, suggesting that
2698 ISO C's @code{#pragma} should be used instead. At the time
2699 @code{__attribute__} was designed, there were two reasons for not doing
2704 It is impossible to generate @code{#pragma} commands from a macro.
2707 There is no telling what the same @code{#pragma} might mean in another
2711 These two reasons applied to almost any application that might have been
2712 proposed for @code{#pragma}. It was basically a mistake to use
2713 @code{#pragma} for @emph{anything}.
2715 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2716 to be generated from macros. In addition, a @code{#pragma GCC}
2717 namespace is now in use for GCC-specific pragmas. However, it has been
2718 found convenient to use @code{__attribute__} to achieve a natural
2719 attachment of attributes to their corresponding declarations, whereas
2720 @code{#pragma GCC} is of use for constructs that do not naturally form
2721 part of the grammar. @xref{Other Directives,,Miscellaneous
2722 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2724 @node Attribute Syntax
2725 @section Attribute Syntax
2726 @cindex attribute syntax
2728 This section describes the syntax with which @code{__attribute__} may be
2729 used, and the constructs to which attribute specifiers bind, for the C
2730 language. Some details may vary for C++ and Objective-C@. Because of
2731 infelicities in the grammar for attributes, some forms described here
2732 may not be successfully parsed in all cases.
2734 There are some problems with the semantics of attributes in C++. For
2735 example, there are no manglings for attributes, although they may affect
2736 code generation, so problems may arise when attributed types are used in
2737 conjunction with templates or overloading. Similarly, @code{typeid}
2738 does not distinguish between types with different attributes. Support
2739 for attributes in C++ may be restricted in future to attributes on
2740 declarations only, but not on nested declarators.
2742 @xref{Function Attributes}, for details of the semantics of attributes
2743 applying to functions. @xref{Variable Attributes}, for details of the
2744 semantics of attributes applying to variables. @xref{Type Attributes},
2745 for details of the semantics of attributes applying to structure, union
2746 and enumerated types.
2748 An @dfn{attribute specifier} is of the form
2749 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2750 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2751 each attribute is one of the following:
2755 Empty. Empty attributes are ignored.
2758 A word (which may be an identifier such as @code{unused}, or a reserved
2759 word such as @code{const}).
2762 A word, followed by, in parentheses, parameters for the attribute.
2763 These parameters take one of the following forms:
2767 An identifier. For example, @code{mode} attributes use this form.
2770 An identifier followed by a comma and a non-empty comma-separated list
2771 of expressions. For example, @code{format} attributes use this form.
2774 A possibly empty comma-separated list of expressions. For example,
2775 @code{format_arg} attributes use this form with the list being a single
2776 integer constant expression, and @code{alias} attributes use this form
2777 with the list being a single string constant.
2781 An @dfn{attribute specifier list} is a sequence of one or more attribute
2782 specifiers, not separated by any other tokens.
2784 In GNU C, an attribute specifier list may appear after the colon following a
2785 label, other than a @code{case} or @code{default} label. The only
2786 attribute it makes sense to use after a label is @code{unused}. This
2787 feature is intended for code generated by programs which contains labels
2788 that may be unused but which is compiled with @option{-Wall}. It would
2789 not normally be appropriate to use in it human-written code, though it
2790 could be useful in cases where the code that jumps to the label is
2791 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2792 such placement of attribute lists, as it is permissible for a
2793 declaration, which could begin with an attribute list, to be labelled in
2794 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2795 does not arise there.
2797 An attribute specifier list may appear as part of a @code{struct},
2798 @code{union} or @code{enum} specifier. It may go either immediately
2799 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2800 the closing brace. The former syntax is preferred.
2801 Where attribute specifiers follow the closing brace, they are considered
2802 to relate to the structure, union or enumerated type defined, not to any
2803 enclosing declaration the type specifier appears in, and the type
2804 defined is not complete until after the attribute specifiers.
2805 @c Otherwise, there would be the following problems: a shift/reduce
2806 @c conflict between attributes binding the struct/union/enum and
2807 @c binding to the list of specifiers/qualifiers; and "aligned"
2808 @c attributes could use sizeof for the structure, but the size could be
2809 @c changed later by "packed" attributes.
2811 Otherwise, an attribute specifier appears as part of a declaration,
2812 counting declarations of unnamed parameters and type names, and relates
2813 to that declaration (which may be nested in another declaration, for
2814 example in the case of a parameter declaration), or to a particular declarator
2815 within a declaration. Where an
2816 attribute specifier is applied to a parameter declared as a function or
2817 an array, it should apply to the function or array rather than the
2818 pointer to which the parameter is implicitly converted, but this is not
2819 yet correctly implemented.
2821 Any list of specifiers and qualifiers at the start of a declaration may
2822 contain attribute specifiers, whether or not such a list may in that
2823 context contain storage class specifiers. (Some attributes, however,
2824 are essentially in the nature of storage class specifiers, and only make
2825 sense where storage class specifiers may be used; for example,
2826 @code{section}.) There is one necessary limitation to this syntax: the
2827 first old-style parameter declaration in a function definition cannot
2828 begin with an attribute specifier, because such an attribute applies to
2829 the function instead by syntax described below (which, however, is not
2830 yet implemented in this case). In some other cases, attribute
2831 specifiers are permitted by this grammar but not yet supported by the
2832 compiler. All attribute specifiers in this place relate to the
2833 declaration as a whole. In the obsolescent usage where a type of
2834 @code{int} is implied by the absence of type specifiers, such a list of
2835 specifiers and qualifiers may be an attribute specifier list with no
2836 other specifiers or qualifiers.
2838 At present, the first parameter in a function prototype must have some
2839 type specifier which is not an attribute specifier; this resolves an
2840 ambiguity in the interpretation of @code{void f(int
2841 (__attribute__((foo)) x))}, but is subject to change. At present, if
2842 the parentheses of a function declarator contain only attributes then
2843 those attributes are ignored, rather than yielding an error or warning
2844 or implying a single parameter of type int, but this is subject to
2847 An attribute specifier list may appear immediately before a declarator
2848 (other than the first) in a comma-separated list of declarators in a
2849 declaration of more than one identifier using a single list of
2850 specifiers and qualifiers. Such attribute specifiers apply
2851 only to the identifier before whose declarator they appear. For
2855 __attribute__((noreturn)) void d0 (void),
2856 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2861 the @code{noreturn} attribute applies to all the functions
2862 declared; the @code{format} attribute only applies to @code{d1}.
2864 An attribute specifier list may appear immediately before the comma,
2865 @code{=} or semicolon terminating the declaration of an identifier other
2866 than a function definition. At present, such attribute specifiers apply
2867 to the declared object or function, but in future they may attach to the
2868 outermost adjacent declarator. In simple cases there is no difference,
2869 but, for example, in
2872 void (****f)(void) __attribute__((noreturn));
2876 at present the @code{noreturn} attribute applies to @code{f}, which
2877 causes a warning since @code{f} is not a function, but in future it may
2878 apply to the function @code{****f}. The precise semantics of what
2879 attributes in such cases will apply to are not yet specified. Where an
2880 assembler name for an object or function is specified (@pxref{Asm
2881 Labels}), at present the attribute must follow the @code{asm}
2882 specification; in future, attributes before the @code{asm} specification
2883 may apply to the adjacent declarator, and those after it to the declared
2886 An attribute specifier list may, in future, be permitted to appear after
2887 the declarator in a function definition (before any old-style parameter
2888 declarations or the function body).
2890 Attribute specifiers may be mixed with type qualifiers appearing inside
2891 the @code{[]} of a parameter array declarator, in the C99 construct by
2892 which such qualifiers are applied to the pointer to which the array is
2893 implicitly converted. Such attribute specifiers apply to the pointer,
2894 not to the array, but at present this is not implemented and they are
2897 An attribute specifier list may appear at the start of a nested
2898 declarator. At present, there are some limitations in this usage: the
2899 attributes correctly apply to the declarator, but for most individual
2900 attributes the semantics this implies are not implemented.
2901 When attribute specifiers follow the @code{*} of a pointer
2902 declarator, they may be mixed with any type qualifiers present.
2903 The following describes the formal semantics of this syntax. It will make the
2904 most sense if you are familiar with the formal specification of
2905 declarators in the ISO C standard.
2907 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2908 D1}, where @code{T} contains declaration specifiers that specify a type
2909 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2910 contains an identifier @var{ident}. The type specified for @var{ident}
2911 for derived declarators whose type does not include an attribute
2912 specifier is as in the ISO C standard.
2914 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2915 and the declaration @code{T D} specifies the type
2916 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2917 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2918 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2920 If @code{D1} has the form @code{*
2921 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2922 declaration @code{T D} specifies the type
2923 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2924 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2925 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2931 void (__attribute__((noreturn)) ****f) (void);
2935 specifies the type ``pointer to pointer to pointer to pointer to
2936 non-returning function returning @code{void}''. As another example,
2939 char *__attribute__((aligned(8))) *f;
2943 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2944 Note again that this does not work with most attributes; for example,
2945 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2946 is not yet supported.
2948 For compatibility with existing code written for compiler versions that
2949 did not implement attributes on nested declarators, some laxity is
2950 allowed in the placing of attributes. If an attribute that only applies
2951 to types is applied to a declaration, it will be treated as applying to
2952 the type of that declaration. If an attribute that only applies to
2953 declarations is applied to the type of a declaration, it will be treated
2954 as applying to that declaration; and, for compatibility with code
2955 placing the attributes immediately before the identifier declared, such
2956 an attribute applied to a function return type will be treated as
2957 applying to the function type, and such an attribute applied to an array
2958 element type will be treated as applying to the array type. If an
2959 attribute that only applies to function types is applied to a
2960 pointer-to-function type, it will be treated as applying to the pointer
2961 target type; if such an attribute is applied to a function return type
2962 that is not a pointer-to-function type, it will be treated as applying
2963 to the function type.
2965 @node Function Prototypes
2966 @section Prototypes and Old-Style Function Definitions
2967 @cindex function prototype declarations
2968 @cindex old-style function definitions
2969 @cindex promotion of formal parameters
2971 GNU C extends ISO C to allow a function prototype to override a later
2972 old-style non-prototype definition. Consider the following example:
2975 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2982 /* @r{Prototype function declaration.} */
2983 int isroot P((uid_t));
2985 /* @r{Old-style function definition.} */
2987 isroot (x) /* @r{??? lossage here ???} */
2994 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2995 not allow this example, because subword arguments in old-style
2996 non-prototype definitions are promoted. Therefore in this example the
2997 function definition's argument is really an @code{int}, which does not
2998 match the prototype argument type of @code{short}.
3000 This restriction of ISO C makes it hard to write code that is portable
3001 to traditional C compilers, because the programmer does not know
3002 whether the @code{uid_t} type is @code{short}, @code{int}, or
3003 @code{long}. Therefore, in cases like these GNU C allows a prototype
3004 to override a later old-style definition. More precisely, in GNU C, a
3005 function prototype argument type overrides the argument type specified
3006 by a later old-style definition if the former type is the same as the
3007 latter type before promotion. Thus in GNU C the above example is
3008 equivalent to the following:
3021 GNU C++ does not support old-style function definitions, so this
3022 extension is irrelevant.
3025 @section C++ Style Comments
3027 @cindex C++ comments
3028 @cindex comments, C++ style
3030 In GNU C, you may use C++ style comments, which start with @samp{//} and
3031 continue until the end of the line. Many other C implementations allow
3032 such comments, and they are included in the 1999 C standard. However,
3033 C++ style comments are not recognized if you specify an @option{-std}
3034 option specifying a version of ISO C before C99, or @option{-ansi}
3035 (equivalent to @option{-std=c89}).
3038 @section Dollar Signs in Identifier Names
3040 @cindex dollar signs in identifier names
3041 @cindex identifier names, dollar signs in
3043 In GNU C, you may normally use dollar signs in identifier names.
3044 This is because many traditional C implementations allow such identifiers.
3045 However, dollar signs in identifiers are not supported on a few target
3046 machines, typically because the target assembler does not allow them.
3048 @node Character Escapes
3049 @section The Character @key{ESC} in Constants
3051 You can use the sequence @samp{\e} in a string or character constant to
3052 stand for the ASCII character @key{ESC}.
3055 @section Inquiring on Alignment of Types or Variables
3057 @cindex type alignment
3058 @cindex variable alignment
3060 The keyword @code{__alignof__} allows you to inquire about how an object
3061 is aligned, or the minimum alignment usually required by a type. Its
3062 syntax is just like @code{sizeof}.
3064 For example, if the target machine requires a @code{double} value to be
3065 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3066 This is true on many RISC machines. On more traditional machine
3067 designs, @code{__alignof__ (double)} is 4 or even 2.
3069 Some machines never actually require alignment; they allow reference to any
3070 data type even at an odd address. For these machines, @code{__alignof__}
3071 reports the @emph{recommended} alignment of a type.
3073 If the operand of @code{__alignof__} is an lvalue rather than a type,
3074 its value is the required alignment for its type, taking into account
3075 any minimum alignment specified with GCC's @code{__attribute__}
3076 extension (@pxref{Variable Attributes}). For example, after this
3080 struct foo @{ int x; char y; @} foo1;
3084 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3085 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3087 It is an error to ask for the alignment of an incomplete type.
3089 @node Variable Attributes
3090 @section Specifying Attributes of Variables
3091 @cindex attribute of variables
3092 @cindex variable attributes
3094 The keyword @code{__attribute__} allows you to specify special
3095 attributes of variables or structure fields. This keyword is followed
3096 by an attribute specification inside double parentheses. Some
3097 attributes are currently defined generically for variables.
3098 Other attributes are defined for variables on particular target
3099 systems. Other attributes are available for functions
3100 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3101 Other front ends might define more attributes
3102 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3104 You may also specify attributes with @samp{__} preceding and following
3105 each keyword. This allows you to use them in header files without
3106 being concerned about a possible macro of the same name. For example,
3107 you may use @code{__aligned__} instead of @code{aligned}.
3109 @xref{Attribute Syntax}, for details of the exact syntax for using
3113 @cindex @code{aligned} attribute
3114 @item aligned (@var{alignment})
3115 This attribute specifies a minimum alignment for the variable or
3116 structure field, measured in bytes. For example, the declaration:
3119 int x __attribute__ ((aligned (16))) = 0;
3123 causes the compiler to allocate the global variable @code{x} on a
3124 16-byte boundary. On a 68040, this could be used in conjunction with
3125 an @code{asm} expression to access the @code{move16} instruction which
3126 requires 16-byte aligned operands.
3128 You can also specify the alignment of structure fields. For example, to
3129 create a double-word aligned @code{int} pair, you could write:
3132 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3136 This is an alternative to creating a union with a @code{double} member
3137 that forces the union to be double-word aligned.
3139 As in the preceding examples, you can explicitly specify the alignment
3140 (in bytes) that you wish the compiler to use for a given variable or
3141 structure field. Alternatively, you can leave out the alignment factor
3142 and just ask the compiler to align a variable or field to the maximum
3143 useful alignment for the target machine you are compiling for. For
3144 example, you could write:
3147 short array[3] __attribute__ ((aligned));
3150 Whenever you leave out the alignment factor in an @code{aligned} attribute
3151 specification, the compiler automatically sets the alignment for the declared
3152 variable or field to the largest alignment which is ever used for any data
3153 type on the target machine you are compiling for. Doing this can often make
3154 copy operations more efficient, because the compiler can use whatever
3155 instructions copy the biggest chunks of memory when performing copies to
3156 or from the variables or fields that you have aligned this way.
3158 When used on a struct, or struct member, the @code{aligned} attribute can
3159 only increase the alignment; in order to decrease it, the @code{packed}
3160 attribute must be specified as well. When used as part of a typedef, the
3161 @code{aligned} attribute can both increase and decrease alignment, and
3162 specifying the @code{packed} attribute will generate a warning.
3164 Note that the effectiveness of @code{aligned} attributes may be limited
3165 by inherent limitations in your linker. On many systems, the linker is
3166 only able to arrange for variables to be aligned up to a certain maximum
3167 alignment. (For some linkers, the maximum supported alignment may
3168 be very very small.) If your linker is only able to align variables
3169 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3170 in an @code{__attribute__} will still only provide you with 8 byte
3171 alignment. See your linker documentation for further information.
3173 @item cleanup (@var{cleanup_function})
3174 @cindex @code{cleanup} attribute
3175 The @code{cleanup} attribute runs a function when the variable goes
3176 out of scope. This attribute can only be applied to auto function
3177 scope variables; it may not be applied to parameters or variables
3178 with static storage duration. The function must take one parameter,
3179 a pointer to a type compatible with the variable. The return value
3180 of the function (if any) is ignored.
3182 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3183 will be run during the stack unwinding that happens during the
3184 processing of the exception. Note that the @code{cleanup} attribute
3185 does not allow the exception to be caught, only to perform an action.
3186 It is undefined what happens if @var{cleanup_function} does not
3191 @cindex @code{common} attribute
3192 @cindex @code{nocommon} attribute
3195 The @code{common} attribute requests GCC to place a variable in
3196 ``common'' storage. The @code{nocommon} attribute requests the
3197 opposite---to allocate space for it directly.
3199 These attributes override the default chosen by the
3200 @option{-fno-common} and @option{-fcommon} flags respectively.
3203 @cindex @code{deprecated} attribute
3204 The @code{deprecated} attribute results in a warning if the variable
3205 is used anywhere in the source file. This is useful when identifying
3206 variables that are expected to be removed in a future version of a
3207 program. The warning also includes the location of the declaration
3208 of the deprecated variable, to enable users to easily find further
3209 information about why the variable is deprecated, or what they should
3210 do instead. Note that the warning only occurs for uses:
3213 extern int old_var __attribute__ ((deprecated));
3215 int new_fn () @{ return old_var; @}
3218 results in a warning on line 3 but not line 2.
3220 The @code{deprecated} attribute can also be used for functions and
3221 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3223 @item mode (@var{mode})
3224 @cindex @code{mode} attribute
3225 This attribute specifies the data type for the declaration---whichever
3226 type corresponds to the mode @var{mode}. This in effect lets you
3227 request an integer or floating point type according to its width.
3229 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3230 indicate the mode corresponding to a one-byte integer, @samp{word} or
3231 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3232 or @samp{__pointer__} for the mode used to represent pointers.
3235 @cindex @code{packed} attribute
3236 The @code{packed} attribute specifies that a variable or structure field
3237 should have the smallest possible alignment---one byte for a variable,
3238 and one bit for a field, unless you specify a larger value with the
3239 @code{aligned} attribute.
3241 Here is a structure in which the field @code{x} is packed, so that it
3242 immediately follows @code{a}:
3248 int x[2] __attribute__ ((packed));
3252 @item section ("@var{section-name}")
3253 @cindex @code{section} variable attribute
3254 Normally, the compiler places the objects it generates in sections like
3255 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3256 or you need certain particular variables to appear in special sections,
3257 for example to map to special hardware. The @code{section}
3258 attribute specifies that a variable (or function) lives in a particular
3259 section. For example, this small program uses several specific section names:
3262 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3263 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3264 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3265 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3269 /* @r{Initialize stack pointer} */
3270 init_sp (stack + sizeof (stack));
3272 /* @r{Initialize initialized data} */
3273 memcpy (&init_data, &data, &edata - &data);
3275 /* @r{Turn on the serial ports} */
3282 Use the @code{section} attribute with an @emph{initialized} definition
3283 of a @emph{global} variable, as shown in the example. GCC issues
3284 a warning and otherwise ignores the @code{section} attribute in
3285 uninitialized variable declarations.
3287 You may only use the @code{section} attribute with a fully initialized
3288 global definition because of the way linkers work. The linker requires
3289 each object be defined once, with the exception that uninitialized
3290 variables tentatively go in the @code{common} (or @code{bss}) section
3291 and can be multiply ``defined''. You can force a variable to be
3292 initialized with the @option{-fno-common} flag or the @code{nocommon}
3295 Some file formats do not support arbitrary sections so the @code{section}
3296 attribute is not available on all platforms.
3297 If you need to map the entire contents of a module to a particular
3298 section, consider using the facilities of the linker instead.
3301 @cindex @code{shared} variable attribute
3302 On Microsoft Windows, in addition to putting variable definitions in a named
3303 section, the section can also be shared among all running copies of an
3304 executable or DLL@. For example, this small program defines shared data
3305 by putting it in a named section @code{shared} and marking the section
3309 int foo __attribute__((section ("shared"), shared)) = 0;
3314 /* @r{Read and write foo. All running
3315 copies see the same value.} */
3321 You may only use the @code{shared} attribute along with @code{section}
3322 attribute with a fully initialized global definition because of the way
3323 linkers work. See @code{section} attribute for more information.
3325 The @code{shared} attribute is only available on Microsoft Windows@.
3327 @item tls_model ("@var{tls_model}")
3328 @cindex @code{tls_model} attribute
3329 The @code{tls_model} attribute sets thread-local storage model
3330 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3331 overriding @option{-ftls-model=} command line switch on a per-variable
3333 The @var{tls_model} argument should be one of @code{global-dynamic},
3334 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3336 Not all targets support this attribute.
3339 This attribute, attached to a variable, means that the variable is meant
3340 to be possibly unused. GCC will not produce a warning for this
3344 This attribute, attached to a variable, means that the variable must be
3345 emitted even if it appears that the variable is not referenced.
3347 @item vector_size (@var{bytes})
3348 This attribute specifies the vector size for the variable, measured in
3349 bytes. For example, the declaration:
3352 int foo __attribute__ ((vector_size (16)));
3356 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3357 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3358 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3360 This attribute is only applicable to integral and float scalars,
3361 although arrays, pointers, and function return values are allowed in
3362 conjunction with this construct.
3364 Aggregates with this attribute are invalid, even if they are of the same
3365 size as a corresponding scalar. For example, the declaration:
3368 struct S @{ int a; @};
3369 struct S __attribute__ ((vector_size (16))) foo;
3373 is invalid even if the size of the structure is the same as the size of
3377 The @code{selectany} attribute causes an initialized global variable to
3378 have link-once semantics. When multiple definitions of the variable are
3379 encountered by the linker, the first is selected and the remainder are
3380 discarded. Following usage by the Microsoft compiler, the linker is told
3381 @emph{not} to warn about size or content differences of the multiple
3384 Although the primary usage of this attribute is for POD types, the
3385 attribute can also be applied to global C++ objects that are initialized
3386 by a constructor. In this case, the static initialization and destruction
3387 code for the object is emitted in each translation defining the object,
3388 but the calls to the constructor and destructor are protected by a
3389 link-once guard variable.
3391 The @code{selectany} attribute is only available on Microsoft Windows
3392 targets. You can use @code{__declspec (selectany)} as a synonym for
3393 @code{__attribute__ ((selectany))} for compatibility with other
3397 The @code{weak} attribute is described in @xref{Function Attributes}.
3400 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3403 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3407 @subsection M32R/D Variable Attributes
3409 One attribute is currently defined for the M32R/D@.
3412 @item model (@var{model-name})
3413 @cindex variable addressability on the M32R/D
3414 Use this attribute on the M32R/D to set the addressability of an object.
3415 The identifier @var{model-name} is one of @code{small}, @code{medium},
3416 or @code{large}, representing each of the code models.
3418 Small model objects live in the lower 16MB of memory (so that their
3419 addresses can be loaded with the @code{ld24} instruction).
3421 Medium and large model objects may live anywhere in the 32-bit address space
3422 (the compiler will generate @code{seth/add3} instructions to load their
3426 @anchor{i386 Variable Attributes}
3427 @subsection i386 Variable Attributes
3429 Two attributes are currently defined for i386 configurations:
3430 @code{ms_struct} and @code{gcc_struct}
3435 @cindex @code{ms_struct} attribute
3436 @cindex @code{gcc_struct} attribute
3438 If @code{packed} is used on a structure, or if bit-fields are used
3439 it may be that the Microsoft ABI packs them differently
3440 than GCC would normally pack them. Particularly when moving packed
3441 data between functions compiled with GCC and the native Microsoft compiler
3442 (either via function call or as data in a file), it may be necessary to access
3445 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3446 compilers to match the native Microsoft compiler.
3448 The Microsoft structure layout algorithm is fairly simple with the exception
3449 of the bitfield packing:
3451 The padding and alignment of members of structures and whether a bit field
3452 can straddle a storage-unit boundary
3455 @item Structure members are stored sequentially in the order in which they are
3456 declared: the first member has the lowest memory address and the last member
3459 @item Every data object has an alignment-requirement. The alignment-requirement
3460 for all data except structures, unions, and arrays is either the size of the
3461 object or the current packing size (specified with either the aligned attribute
3462 or the pack pragma), whichever is less. For structures, unions, and arrays,
3463 the alignment-requirement is the largest alignment-requirement of its members.
3464 Every object is allocated an offset so that:
3466 offset % alignment-requirement == 0
3468 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3469 unit if the integral types are the same size and if the next bit field fits
3470 into the current allocation unit without crossing the boundary imposed by the
3471 common alignment requirements of the bit fields.
3474 Handling of zero-length bitfields:
3476 MSVC interprets zero-length bitfields in the following ways:
3479 @item If a zero-length bitfield is inserted between two bitfields that would
3480 normally be coalesced, the bitfields will not be coalesced.
3487 unsigned long bf_1 : 12;
3489 unsigned long bf_2 : 12;
3493 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3494 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3496 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3497 alignment of the zero-length bitfield is greater than the member that follows it,
3498 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3518 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3519 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3520 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3523 Taking this into account, it is important to note the following:
3526 @item If a zero-length bitfield follows a normal bitfield, the type of the
3527 zero-length bitfield may affect the alignment of the structure as whole. For
3528 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3529 normal bitfield, and is of type short.
3531 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3532 still affect the alignment of the structure:
3542 Here, @code{t4} will take up 4 bytes.
3545 @item Zero-length bitfields following non-bitfield members are ignored:
3556 Here, @code{t5} will take up 2 bytes.
3560 @subsection PowerPC Variable Attributes
3562 Three attributes currently are defined for PowerPC configurations:
3563 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3565 For full documentation of the struct attributes please see the
3566 documentation in the @xref{i386 Variable Attributes}, section.
3568 For documentation of @code{altivec} attribute please see the
3569 documentation in the @xref{PowerPC Type Attributes}, section.
3571 @subsection SPU Variable Attributes
3573 The SPU supports the @code{spu_vector} attribute for variables. For
3574 documentation of this attribute please see the documentation in the
3575 @xref{SPU Type Attributes}, section.
3577 @subsection Xstormy16 Variable Attributes
3579 One attribute is currently defined for xstormy16 configurations:
3584 @cindex @code{below100} attribute
3586 If a variable has the @code{below100} attribute (@code{BELOW100} is
3587 allowed also), GCC will place the variable in the first 0x100 bytes of
3588 memory and use special opcodes to access it. Such variables will be
3589 placed in either the @code{.bss_below100} section or the
3590 @code{.data_below100} section.
3594 @node Type Attributes
3595 @section Specifying Attributes of Types
3596 @cindex attribute of types
3597 @cindex type attributes
3599 The keyword @code{__attribute__} allows you to specify special
3600 attributes of @code{struct} and @code{union} types when you define
3601 such types. This keyword is followed by an attribute specification
3602 inside double parentheses. Seven attributes are currently defined for
3603 types: @code{aligned}, @code{packed}, @code{transparent_union},
3604 @code{unused}, @code{deprecated}, @code{visibility}, and
3605 @code{may_alias}. Other attributes are defined for functions
3606 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3609 You may also specify any one of these attributes with @samp{__}
3610 preceding and following its keyword. This allows you to use these
3611 attributes in header files without being concerned about a possible
3612 macro of the same name. For example, you may use @code{__aligned__}
3613 instead of @code{aligned}.
3615 You may specify type attributes either in a @code{typedef} declaration
3616 or in an enum, struct or union type declaration or definition.
3618 For an enum, struct or union type, you may specify attributes either
3619 between the enum, struct or union tag and the name of the type, or
3620 just past the closing curly brace of the @emph{definition}. The
3621 former syntax is preferred.
3623 @xref{Attribute Syntax}, for details of the exact syntax for using
3627 @cindex @code{aligned} attribute
3628 @item aligned (@var{alignment})
3629 This attribute specifies a minimum alignment (in bytes) for variables
3630 of the specified type. For example, the declarations:
3633 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3634 typedef int more_aligned_int __attribute__ ((aligned (8)));
3638 force the compiler to insure (as far as it can) that each variable whose
3639 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3640 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3641 variables of type @code{struct S} aligned to 8-byte boundaries allows
3642 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3643 store) instructions when copying one variable of type @code{struct S} to
3644 another, thus improving run-time efficiency.
3646 Note that the alignment of any given @code{struct} or @code{union} type
3647 is required by the ISO C standard to be at least a perfect multiple of
3648 the lowest common multiple of the alignments of all of the members of
3649 the @code{struct} or @code{union} in question. This means that you @emph{can}
3650 effectively adjust the alignment of a @code{struct} or @code{union}
3651 type by attaching an @code{aligned} attribute to any one of the members
3652 of such a type, but the notation illustrated in the example above is a
3653 more obvious, intuitive, and readable way to request the compiler to
3654 adjust the alignment of an entire @code{struct} or @code{union} type.
3656 As in the preceding example, you can explicitly specify the alignment
3657 (in bytes) that you wish the compiler to use for a given @code{struct}
3658 or @code{union} type. Alternatively, you can leave out the alignment factor
3659 and just ask the compiler to align a type to the maximum
3660 useful alignment for the target machine you are compiling for. For
3661 example, you could write:
3664 struct S @{ short f[3]; @} __attribute__ ((aligned));
3667 Whenever you leave out the alignment factor in an @code{aligned}
3668 attribute specification, the compiler automatically sets the alignment
3669 for the type to the largest alignment which is ever used for any data
3670 type on the target machine you are compiling for. Doing this can often
3671 make copy operations more efficient, because the compiler can use
3672 whatever instructions copy the biggest chunks of memory when performing
3673 copies to or from the variables which have types that you have aligned
3676 In the example above, if the size of each @code{short} is 2 bytes, then
3677 the size of the entire @code{struct S} type is 6 bytes. The smallest
3678 power of two which is greater than or equal to that is 8, so the
3679 compiler sets the alignment for the entire @code{struct S} type to 8
3682 Note that although you can ask the compiler to select a time-efficient
3683 alignment for a given type and then declare only individual stand-alone
3684 objects of that type, the compiler's ability to select a time-efficient
3685 alignment is primarily useful only when you plan to create arrays of
3686 variables having the relevant (efficiently aligned) type. If you
3687 declare or use arrays of variables of an efficiently-aligned type, then
3688 it is likely that your program will also be doing pointer arithmetic (or
3689 subscripting, which amounts to the same thing) on pointers to the
3690 relevant type, and the code that the compiler generates for these
3691 pointer arithmetic operations will often be more efficient for
3692 efficiently-aligned types than for other types.
3694 The @code{aligned} attribute can only increase the alignment; but you
3695 can decrease it by specifying @code{packed} as well. See below.
3697 Note that the effectiveness of @code{aligned} attributes may be limited
3698 by inherent limitations in your linker. On many systems, the linker is
3699 only able to arrange for variables to be aligned up to a certain maximum
3700 alignment. (For some linkers, the maximum supported alignment may
3701 be very very small.) If your linker is only able to align variables
3702 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3703 in an @code{__attribute__} will still only provide you with 8 byte
3704 alignment. See your linker documentation for further information.
3707 This attribute, attached to @code{struct} or @code{union} type
3708 definition, specifies that each member (other than zero-width bitfields)
3709 of the structure or union is placed to minimize the memory required. When
3710 attached to an @code{enum} definition, it indicates that the smallest
3711 integral type should be used.
3713 @opindex fshort-enums
3714 Specifying this attribute for @code{struct} and @code{union} types is
3715 equivalent to specifying the @code{packed} attribute on each of the
3716 structure or union members. Specifying the @option{-fshort-enums}
3717 flag on the line is equivalent to specifying the @code{packed}
3718 attribute on all @code{enum} definitions.
3720 In the following example @code{struct my_packed_struct}'s members are
3721 packed closely together, but the internal layout of its @code{s} member
3722 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3726 struct my_unpacked_struct
3732 struct __attribute__ ((__packed__)) my_packed_struct
3736 struct my_unpacked_struct s;
3740 You may only specify this attribute on the definition of a @code{enum},
3741 @code{struct} or @code{union}, not on a @code{typedef} which does not
3742 also define the enumerated type, structure or union.
3744 @item transparent_union
3745 This attribute, attached to a @code{union} type definition, indicates
3746 that any function parameter having that union type causes calls to that
3747 function to be treated in a special way.
3749 First, the argument corresponding to a transparent union type can be of
3750 any type in the union; no cast is required. Also, if the union contains
3751 a pointer type, the corresponding argument can be a null pointer
3752 constant or a void pointer expression; and if the union contains a void
3753 pointer type, the corresponding argument can be any pointer expression.
3754 If the union member type is a pointer, qualifiers like @code{const} on
3755 the referenced type must be respected, just as with normal pointer
3758 Second, the argument is passed to the function using the calling
3759 conventions of the first member of the transparent union, not the calling
3760 conventions of the union itself. All members of the union must have the
3761 same machine representation; this is necessary for this argument passing
3764 Transparent unions are designed for library functions that have multiple
3765 interfaces for compatibility reasons. For example, suppose the
3766 @code{wait} function must accept either a value of type @code{int *} to
3767 comply with Posix, or a value of type @code{union wait *} to comply with
3768 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3769 @code{wait} would accept both kinds of arguments, but it would also
3770 accept any other pointer type and this would make argument type checking
3771 less useful. Instead, @code{<sys/wait.h>} might define the interface
3779 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3781 pid_t wait (wait_status_ptr_t);
3784 This interface allows either @code{int *} or @code{union wait *}
3785 arguments to be passed, using the @code{int *} calling convention.
3786 The program can call @code{wait} with arguments of either type:
3789 int w1 () @{ int w; return wait (&w); @}
3790 int w2 () @{ union wait w; return wait (&w); @}
3793 With this interface, @code{wait}'s implementation might look like this:
3796 pid_t wait (wait_status_ptr_t p)
3798 return waitpid (-1, p.__ip, 0);
3803 When attached to a type (including a @code{union} or a @code{struct}),
3804 this attribute means that variables of that type are meant to appear
3805 possibly unused. GCC will not produce a warning for any variables of
3806 that type, even if the variable appears to do nothing. This is often
3807 the case with lock or thread classes, which are usually defined and then
3808 not referenced, but contain constructors and destructors that have
3809 nontrivial bookkeeping functions.
3812 The @code{deprecated} attribute results in a warning if the type
3813 is used anywhere in the source file. This is useful when identifying
3814 types that are expected to be removed in a future version of a program.
3815 If possible, the warning also includes the location of the declaration
3816 of the deprecated type, to enable users to easily find further
3817 information about why the type is deprecated, or what they should do
3818 instead. Note that the warnings only occur for uses and then only
3819 if the type is being applied to an identifier that itself is not being
3820 declared as deprecated.
3823 typedef int T1 __attribute__ ((deprecated));
3827 typedef T1 T3 __attribute__ ((deprecated));
3828 T3 z __attribute__ ((deprecated));
3831 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3832 warning is issued for line 4 because T2 is not explicitly
3833 deprecated. Line 5 has no warning because T3 is explicitly
3834 deprecated. Similarly for line 6.
3836 The @code{deprecated} attribute can also be used for functions and
3837 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3840 Accesses to objects with types with this attribute are not subjected to
3841 type-based alias analysis, but are instead assumed to be able to alias
3842 any other type of objects, just like the @code{char} type. See
3843 @option{-fstrict-aliasing} for more information on aliasing issues.
3848 typedef short __attribute__((__may_alias__)) short_a;
3854 short_a *b = (short_a *) &a;
3858 if (a == 0x12345678)
3865 If you replaced @code{short_a} with @code{short} in the variable
3866 declaration, the above program would abort when compiled with
3867 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3868 above in recent GCC versions.
3871 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3872 applied to class, struct, union and enum types. Unlike other type
3873 attributes, the attribute must appear between the initial keyword and
3874 the name of the type; it cannot appear after the body of the type.
3876 Note that the type visibility is applied to vague linkage entities
3877 associated with the class (vtable, typeinfo node, etc.). In
3878 particular, if a class is thrown as an exception in one shared object
3879 and caught in another, the class must have default visibility.
3880 Otherwise the two shared objects will be unable to use the same
3881 typeinfo node and exception handling will break.
3883 @subsection ARM Type Attributes
3885 On those ARM targets that support @code{dllimport} (such as Symbian
3886 OS), you can use the @code{notshared} attribute to indicate that the
3887 virtual table and other similar data for a class should not be
3888 exported from a DLL@. For example:
3891 class __declspec(notshared) C @{
3893 __declspec(dllimport) C();
3897 __declspec(dllexport)
3901 In this code, @code{C::C} is exported from the current DLL, but the
3902 virtual table for @code{C} is not exported. (You can use
3903 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3904 most Symbian OS code uses @code{__declspec}.)
3906 @anchor{i386 Type Attributes}
3907 @subsection i386 Type Attributes
3909 Two attributes are currently defined for i386 configurations:
3910 @code{ms_struct} and @code{gcc_struct}
3914 @cindex @code{ms_struct}
3915 @cindex @code{gcc_struct}
3917 If @code{packed} is used on a structure, or if bit-fields are used
3918 it may be that the Microsoft ABI packs them differently
3919 than GCC would normally pack them. Particularly when moving packed
3920 data between functions compiled with GCC and the native Microsoft compiler
3921 (either via function call or as data in a file), it may be necessary to access
3924 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3925 compilers to match the native Microsoft compiler.
3928 To specify multiple attributes, separate them by commas within the
3929 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3932 @anchor{PowerPC Type Attributes}
3933 @subsection PowerPC Type Attributes
3935 Three attributes currently are defined for PowerPC configurations:
3936 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3938 For full documentation of the struct attributes please see the
3939 documentation in the @xref{i386 Type Attributes}, section.
3941 The @code{altivec} attribute allows one to declare AltiVec vector data
3942 types supported by the AltiVec Programming Interface Manual. The
3943 attribute requires an argument to specify one of three vector types:
3944 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3945 and @code{bool__} (always followed by unsigned).
3948 __attribute__((altivec(vector__)))
3949 __attribute__((altivec(pixel__))) unsigned short
3950 __attribute__((altivec(bool__))) unsigned
3953 These attributes mainly are intended to support the @code{__vector},
3954 @code{__pixel}, and @code{__bool} AltiVec keywords.
3956 @anchor{SPU Type Attributes}
3957 @subsection SPU Type Attributes
3959 The SPU supports the @code{spu_vector} attribute for types. This attribute
3960 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
3961 Language Extensions Specification. It is intended to support the
3962 @code{__vector} keyword.
3966 @section An Inline Function is As Fast As a Macro
3967 @cindex inline functions
3968 @cindex integrating function code
3970 @cindex macros, inline alternative
3972 By declaring a function inline, you can direct GCC to make
3973 calls to that function faster. One way GCC can achieve this is to
3974 integrate that function's code into the code for its callers. This
3975 makes execution faster by eliminating the function-call overhead; in
3976 addition, if any of the actual argument values are constant, their
3977 known values may permit simplifications at compile time so that not
3978 all of the inline function's code needs to be included. The effect on
3979 code size is less predictable; object code may be larger or smaller
3980 with function inlining, depending on the particular case. You can
3981 also direct GCC to try to integrate all ``simple enough'' functions
3982 into their callers with the option @option{-finline-functions}.
3984 GCC implements three different semantics of declaring a function
3985 inline. One is available with @option{-std=gnu89} or
3986 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
3987 on all inline declarations, another when @option{-std=c99} or
3988 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
3989 is used when compiling C++.
3991 To declare a function inline, use the @code{inline} keyword in its
3992 declaration, like this:
4002 If you are writing a header file to be included in ISO C89 programs, write
4003 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
4005 The three types of inlining behave similarly in two important cases:
4006 when the @code{inline} keyword is used on a @code{static} function,
4007 like the example above, and when a function is first declared without
4008 using the @code{inline} keyword and then is defined with
4009 @code{inline}, like this:
4012 extern int inc (int *a);
4020 In both of these common cases, the program behaves the same as if you
4021 had not used the @code{inline} keyword, except for its speed.
4023 @cindex inline functions, omission of
4024 @opindex fkeep-inline-functions
4025 When a function is both inline and @code{static}, if all calls to the
4026 function are integrated into the caller, and the function's address is
4027 never used, then the function's own assembler code is never referenced.
4028 In this case, GCC does not actually output assembler code for the
4029 function, unless you specify the option @option{-fkeep-inline-functions}.
4030 Some calls cannot be integrated for various reasons (in particular,
4031 calls that precede the function's definition cannot be integrated, and
4032 neither can recursive calls within the definition). If there is a
4033 nonintegrated call, then the function is compiled to assembler code as
4034 usual. The function must also be compiled as usual if the program
4035 refers to its address, because that can't be inlined.
4038 Note that certain usages in a function definition can make it unsuitable
4039 for inline substitution. Among these usages are: use of varargs, use of
4040 alloca, use of variable sized data types (@pxref{Variable Length}),
4041 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
4042 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4043 will warn when a function marked @code{inline} could not be substituted,
4044 and will give the reason for the failure.
4046 @cindex automatic @code{inline} for C++ member fns
4047 @cindex @code{inline} automatic for C++ member fns
4048 @cindex member fns, automatically @code{inline}
4049 @cindex C++ member fns, automatically @code{inline}
4050 @opindex fno-default-inline
4051 As required by ISO C++, GCC considers member functions defined within
4052 the body of a class to be marked inline even if they are
4053 not explicitly declared with the @code{inline} keyword. You can
4054 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4055 Options,,Options Controlling C++ Dialect}.
4057 GCC does not inline any functions when not optimizing unless you specify
4058 the @samp{always_inline} attribute for the function, like this:
4061 /* @r{Prototype.} */
4062 inline void foo (const char) __attribute__((always_inline));
4065 The remainder of this section is specific to GNU C89 inlining.
4067 @cindex non-static inline function
4068 When an inline function is not @code{static}, then the compiler must assume
4069 that there may be calls from other source files; since a global symbol can
4070 be defined only once in any program, the function must not be defined in
4071 the other source files, so the calls therein cannot be integrated.
4072 Therefore, a non-@code{static} inline function is always compiled on its
4073 own in the usual fashion.
4075 If you specify both @code{inline} and @code{extern} in the function
4076 definition, then the definition is used only for inlining. In no case
4077 is the function compiled on its own, not even if you refer to its
4078 address explicitly. Such an address becomes an external reference, as
4079 if you had only declared the function, and had not defined it.
4081 This combination of @code{inline} and @code{extern} has almost the
4082 effect of a macro. The way to use it is to put a function definition in
4083 a header file with these keywords, and put another copy of the
4084 definition (lacking @code{inline} and @code{extern}) in a library file.
4085 The definition in the header file will cause most calls to the function
4086 to be inlined. If any uses of the function remain, they will refer to
4087 the single copy in the library.
4090 @section Assembler Instructions with C Expression Operands
4091 @cindex extended @code{asm}
4092 @cindex @code{asm} expressions
4093 @cindex assembler instructions
4096 In an assembler instruction using @code{asm}, you can specify the
4097 operands of the instruction using C expressions. This means you need not
4098 guess which registers or memory locations will contain the data you want
4101 You must specify an assembler instruction template much like what
4102 appears in a machine description, plus an operand constraint string for
4105 For example, here is how to use the 68881's @code{fsinx} instruction:
4108 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4112 Here @code{angle} is the C expression for the input operand while
4113 @code{result} is that of the output operand. Each has @samp{"f"} as its
4114 operand constraint, saying that a floating point register is required.
4115 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4116 output operands' constraints must use @samp{=}. The constraints use the
4117 same language used in the machine description (@pxref{Constraints}).
4119 Each operand is described by an operand-constraint string followed by
4120 the C expression in parentheses. A colon separates the assembler
4121 template from the first output operand and another separates the last
4122 output operand from the first input, if any. Commas separate the
4123 operands within each group. The total number of operands is currently
4124 limited to 30; this limitation may be lifted in some future version of
4127 If there are no output operands but there are input operands, you must
4128 place two consecutive colons surrounding the place where the output
4131 As of GCC version 3.1, it is also possible to specify input and output
4132 operands using symbolic names which can be referenced within the
4133 assembler code. These names are specified inside square brackets
4134 preceding the constraint string, and can be referenced inside the
4135 assembler code using @code{%[@var{name}]} instead of a percentage sign
4136 followed by the operand number. Using named operands the above example
4140 asm ("fsinx %[angle],%[output]"
4141 : [output] "=f" (result)
4142 : [angle] "f" (angle));
4146 Note that the symbolic operand names have no relation whatsoever to
4147 other C identifiers. You may use any name you like, even those of
4148 existing C symbols, but you must ensure that no two operands within the same
4149 assembler construct use the same symbolic name.
4151 Output operand expressions must be lvalues; the compiler can check this.
4152 The input operands need not be lvalues. The compiler cannot check
4153 whether the operands have data types that are reasonable for the
4154 instruction being executed. It does not parse the assembler instruction
4155 template and does not know what it means or even whether it is valid
4156 assembler input. The extended @code{asm} feature is most often used for
4157 machine instructions the compiler itself does not know exist. If
4158 the output expression cannot be directly addressed (for example, it is a
4159 bit-field), your constraint must allow a register. In that case, GCC
4160 will use the register as the output of the @code{asm}, and then store
4161 that register into the output.
4163 The ordinary output operands must be write-only; GCC will assume that
4164 the values in these operands before the instruction are dead and need
4165 not be generated. Extended asm supports input-output or read-write
4166 operands. Use the constraint character @samp{+} to indicate such an
4167 operand and list it with the output operands. You should only use
4168 read-write operands when the constraints for the operand (or the
4169 operand in which only some of the bits are to be changed) allow a
4172 You may, as an alternative, logically split its function into two
4173 separate operands, one input operand and one write-only output
4174 operand. The connection between them is expressed by constraints
4175 which say they need to be in the same location when the instruction
4176 executes. You can use the same C expression for both operands, or
4177 different expressions. For example, here we write the (fictitious)
4178 @samp{combine} instruction with @code{bar} as its read-only source
4179 operand and @code{foo} as its read-write destination:
4182 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4186 The constraint @samp{"0"} for operand 1 says that it must occupy the
4187 same location as operand 0. A number in constraint is allowed only in
4188 an input operand and it must refer to an output operand.
4190 Only a number in the constraint can guarantee that one operand will be in
4191 the same place as another. The mere fact that @code{foo} is the value
4192 of both operands is not enough to guarantee that they will be in the
4193 same place in the generated assembler code. The following would not
4197 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4200 Various optimizations or reloading could cause operands 0 and 1 to be in
4201 different registers; GCC knows no reason not to do so. For example, the
4202 compiler might find a copy of the value of @code{foo} in one register and
4203 use it for operand 1, but generate the output operand 0 in a different
4204 register (copying it afterward to @code{foo}'s own address). Of course,
4205 since the register for operand 1 is not even mentioned in the assembler
4206 code, the result will not work, but GCC can't tell that.
4208 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4209 the operand number for a matching constraint. For example:
4212 asm ("cmoveq %1,%2,%[result]"
4213 : [result] "=r"(result)
4214 : "r" (test), "r"(new), "[result]"(old));
4217 Sometimes you need to make an @code{asm} operand be a specific register,
4218 but there's no matching constraint letter for that register @emph{by
4219 itself}. To force the operand into that register, use a local variable
4220 for the operand and specify the register in the variable declaration.
4221 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4222 register constraint letter that matches the register:
4225 register int *p1 asm ("r0") = @dots{};
4226 register int *p2 asm ("r1") = @dots{};
4227 register int *result asm ("r0");
4228 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4231 @anchor{Example of asm with clobbered asm reg}
4232 In the above example, beware that a register that is call-clobbered by
4233 the target ABI will be overwritten by any function call in the
4234 assignment, including library calls for arithmetic operators.
4235 Assuming it is a call-clobbered register, this may happen to @code{r0}
4236 above by the assignment to @code{p2}. If you have to use such a
4237 register, use temporary variables for expressions between the register
4242 register int *p1 asm ("r0") = @dots{};
4243 register int *p2 asm ("r1") = t1;
4244 register int *result asm ("r0");
4245 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4248 Some instructions clobber specific hard registers. To describe this,
4249 write a third colon after the input operands, followed by the names of
4250 the clobbered hard registers (given as strings). Here is a realistic
4251 example for the VAX:
4254 asm volatile ("movc3 %0,%1,%2"
4255 : /* @r{no outputs} */
4256 : "g" (from), "g" (to), "g" (count)
4257 : "r0", "r1", "r2", "r3", "r4", "r5");
4260 You may not write a clobber description in a way that overlaps with an
4261 input or output operand. For example, you may not have an operand
4262 describing a register class with one member if you mention that register
4263 in the clobber list. Variables declared to live in specific registers
4264 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4265 have no part mentioned in the clobber description.
4266 There is no way for you to specify that an input
4267 operand is modified without also specifying it as an output
4268 operand. Note that if all the output operands you specify are for this
4269 purpose (and hence unused), you will then also need to specify
4270 @code{volatile} for the @code{asm} construct, as described below, to
4271 prevent GCC from deleting the @code{asm} statement as unused.
4273 If you refer to a particular hardware register from the assembler code,
4274 you will probably have to list the register after the third colon to
4275 tell the compiler the register's value is modified. In some assemblers,
4276 the register names begin with @samp{%}; to produce one @samp{%} in the
4277 assembler code, you must write @samp{%%} in the input.
4279 If your assembler instruction can alter the condition code register, add
4280 @samp{cc} to the list of clobbered registers. GCC on some machines
4281 represents the condition codes as a specific hardware register;
4282 @samp{cc} serves to name this register. On other machines, the
4283 condition code is handled differently, and specifying @samp{cc} has no
4284 effect. But it is valid no matter what the machine.
4286 If your assembler instructions access memory in an unpredictable
4287 fashion, add @samp{memory} to the list of clobbered registers. This
4288 will cause GCC to not keep memory values cached in registers across the
4289 assembler instruction and not optimize stores or loads to that memory.
4290 You will also want to add the @code{volatile} keyword if the memory
4291 affected is not listed in the inputs or outputs of the @code{asm}, as
4292 the @samp{memory} clobber does not count as a side-effect of the
4293 @code{asm}. If you know how large the accessed memory is, you can add
4294 it as input or output but if this is not known, you should add
4295 @samp{memory}. As an example, if you access ten bytes of a string, you
4296 can use a memory input like:
4299 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4302 Note that in the following example the memory input is necessary,
4303 otherwise GCC might optimize the store to @code{x} away:
4310 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4311 "=&d" (r) : "a" (y), "m" (*y));
4316 You can put multiple assembler instructions together in a single
4317 @code{asm} template, separated by the characters normally used in assembly
4318 code for the system. A combination that works in most places is a newline
4319 to break the line, plus a tab character to move to the instruction field
4320 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4321 assembler allows semicolons as a line-breaking character. Note that some
4322 assembler dialects use semicolons to start a comment.
4323 The input operands are guaranteed not to use any of the clobbered
4324 registers, and neither will the output operands' addresses, so you can
4325 read and write the clobbered registers as many times as you like. Here
4326 is an example of multiple instructions in a template; it assumes the
4327 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4330 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4332 : "g" (from), "g" (to)
4336 Unless an output operand has the @samp{&} constraint modifier, GCC
4337 may allocate it in the same register as an unrelated input operand, on
4338 the assumption the inputs are consumed before the outputs are produced.
4339 This assumption may be false if the assembler code actually consists of
4340 more than one instruction. In such a case, use @samp{&} for each output
4341 operand that may not overlap an input. @xref{Modifiers}.
4343 If you want to test the condition code produced by an assembler
4344 instruction, you must include a branch and a label in the @code{asm}
4345 construct, as follows:
4348 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4354 This assumes your assembler supports local labels, as the GNU assembler
4355 and most Unix assemblers do.
4357 Speaking of labels, jumps from one @code{asm} to another are not
4358 supported. The compiler's optimizers do not know about these jumps, and
4359 therefore they cannot take account of them when deciding how to
4362 @cindex macros containing @code{asm}
4363 Usually the most convenient way to use these @code{asm} instructions is to
4364 encapsulate them in macros that look like functions. For example,
4368 (@{ double __value, __arg = (x); \
4369 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4374 Here the variable @code{__arg} is used to make sure that the instruction
4375 operates on a proper @code{double} value, and to accept only those
4376 arguments @code{x} which can convert automatically to a @code{double}.
4378 Another way to make sure the instruction operates on the correct data
4379 type is to use a cast in the @code{asm}. This is different from using a
4380 variable @code{__arg} in that it converts more different types. For
4381 example, if the desired type were @code{int}, casting the argument to
4382 @code{int} would accept a pointer with no complaint, while assigning the
4383 argument to an @code{int} variable named @code{__arg} would warn about
4384 using a pointer unless the caller explicitly casts it.
4386 If an @code{asm} has output operands, GCC assumes for optimization
4387 purposes the instruction has no side effects except to change the output
4388 operands. This does not mean instructions with a side effect cannot be
4389 used, but you must be careful, because the compiler may eliminate them
4390 if the output operands aren't used, or move them out of loops, or
4391 replace two with one if they constitute a common subexpression. Also,
4392 if your instruction does have a side effect on a variable that otherwise
4393 appears not to change, the old value of the variable may be reused later
4394 if it happens to be found in a register.
4396 You can prevent an @code{asm} instruction from being deleted
4397 by writing the keyword @code{volatile} after
4398 the @code{asm}. For example:
4401 #define get_and_set_priority(new) \
4403 asm volatile ("get_and_set_priority %0, %1" \
4404 : "=g" (__old) : "g" (new)); \
4409 The @code{volatile} keyword indicates that the instruction has
4410 important side-effects. GCC will not delete a volatile @code{asm} if
4411 it is reachable. (The instruction can still be deleted if GCC can
4412 prove that control-flow will never reach the location of the
4413 instruction.) Note that even a volatile @code{asm} instruction
4414 can be moved relative to other code, including across jump
4415 instructions. For example, on many targets there is a system
4416 register which can be set to control the rounding mode of
4417 floating point operations. You might try
4418 setting it with a volatile @code{asm}, like this PowerPC example:
4421 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4426 This will not work reliably, as the compiler may move the addition back
4427 before the volatile @code{asm}. To make it work you need to add an
4428 artificial dependency to the @code{asm} referencing a variable in the code
4429 you don't want moved, for example:
4432 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4436 Similarly, you can't expect a
4437 sequence of volatile @code{asm} instructions to remain perfectly
4438 consecutive. If you want consecutive output, use a single @code{asm}.
4439 Also, GCC will perform some optimizations across a volatile @code{asm}
4440 instruction; GCC does not ``forget everything'' when it encounters
4441 a volatile @code{asm} instruction the way some other compilers do.
4443 An @code{asm} instruction without any output operands will be treated
4444 identically to a volatile @code{asm} instruction.
4446 It is a natural idea to look for a way to give access to the condition
4447 code left by the assembler instruction. However, when we attempted to
4448 implement this, we found no way to make it work reliably. The problem
4449 is that output operands might need reloading, which would result in
4450 additional following ``store'' instructions. On most machines, these
4451 instructions would alter the condition code before there was time to
4452 test it. This problem doesn't arise for ordinary ``test'' and
4453 ``compare'' instructions because they don't have any output operands.
4455 For reasons similar to those described above, it is not possible to give
4456 an assembler instruction access to the condition code left by previous
4459 If you are writing a header file that should be includable in ISO C
4460 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4463 @subsection Size of an @code{asm}
4465 Some targets require that GCC track the size of each instruction used in
4466 order to generate correct code. Because the final length of an
4467 @code{asm} is only known by the assembler, GCC must make an estimate as
4468 to how big it will be. The estimate is formed by counting the number of
4469 statements in the pattern of the @code{asm} and multiplying that by the
4470 length of the longest instruction on that processor. Statements in the
4471 @code{asm} are identified by newline characters and whatever statement
4472 separator characters are supported by the assembler; on most processors
4473 this is the `@code{;}' character.
4475 Normally, GCC's estimate is perfectly adequate to ensure that correct
4476 code is generated, but it is possible to confuse the compiler if you use
4477 pseudo instructions or assembler macros that expand into multiple real
4478 instructions or if you use assembler directives that expand to more
4479 space in the object file than would be needed for a single instruction.
4480 If this happens then the assembler will produce a diagnostic saying that
4481 a label is unreachable.
4483 @subsection i386 floating point asm operands
4485 There are several rules on the usage of stack-like regs in
4486 asm_operands insns. These rules apply only to the operands that are
4491 Given a set of input regs that die in an asm_operands, it is
4492 necessary to know which are implicitly popped by the asm, and
4493 which must be explicitly popped by gcc.
4495 An input reg that is implicitly popped by the asm must be
4496 explicitly clobbered, unless it is constrained to match an
4500 For any input reg that is implicitly popped by an asm, it is
4501 necessary to know how to adjust the stack to compensate for the pop.
4502 If any non-popped input is closer to the top of the reg-stack than
4503 the implicitly popped reg, it would not be possible to know what the
4504 stack looked like---it's not clear how the rest of the stack ``slides
4507 All implicitly popped input regs must be closer to the top of
4508 the reg-stack than any input that is not implicitly popped.
4510 It is possible that if an input dies in an insn, reload might
4511 use the input reg for an output reload. Consider this example:
4514 asm ("foo" : "=t" (a) : "f" (b));
4517 This asm says that input B is not popped by the asm, and that
4518 the asm pushes a result onto the reg-stack, i.e., the stack is one
4519 deeper after the asm than it was before. But, it is possible that
4520 reload will think that it can use the same reg for both the input and
4521 the output, if input B dies in this insn.
4523 If any input operand uses the @code{f} constraint, all output reg
4524 constraints must use the @code{&} earlyclobber.
4526 The asm above would be written as
4529 asm ("foo" : "=&t" (a) : "f" (b));
4533 Some operands need to be in particular places on the stack. All
4534 output operands fall in this category---there is no other way to
4535 know which regs the outputs appear in unless the user indicates
4536 this in the constraints.
4538 Output operands must specifically indicate which reg an output
4539 appears in after an asm. @code{=f} is not allowed: the operand
4540 constraints must select a class with a single reg.
4543 Output operands may not be ``inserted'' between existing stack regs.
4544 Since no 387 opcode uses a read/write operand, all output operands
4545 are dead before the asm_operands, and are pushed by the asm_operands.
4546 It makes no sense to push anywhere but the top of the reg-stack.
4548 Output operands must start at the top of the reg-stack: output
4549 operands may not ``skip'' a reg.
4552 Some asm statements may need extra stack space for internal
4553 calculations. This can be guaranteed by clobbering stack registers
4554 unrelated to the inputs and outputs.
4558 Here are a couple of reasonable asms to want to write. This asm
4559 takes one input, which is internally popped, and produces two outputs.
4562 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4565 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4566 and replaces them with one output. The user must code the @code{st(1)}
4567 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4570 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4576 @section Controlling Names Used in Assembler Code
4577 @cindex assembler names for identifiers
4578 @cindex names used in assembler code
4579 @cindex identifiers, names in assembler code
4581 You can specify the name to be used in the assembler code for a C
4582 function or variable by writing the @code{asm} (or @code{__asm__})
4583 keyword after the declarator as follows:
4586 int foo asm ("myfoo") = 2;
4590 This specifies that the name to be used for the variable @code{foo} in
4591 the assembler code should be @samp{myfoo} rather than the usual
4594 On systems where an underscore is normally prepended to the name of a C
4595 function or variable, this feature allows you to define names for the
4596 linker that do not start with an underscore.
4598 It does not make sense to use this feature with a non-static local
4599 variable since such variables do not have assembler names. If you are
4600 trying to put the variable in a particular register, see @ref{Explicit
4601 Reg Vars}. GCC presently accepts such code with a warning, but will
4602 probably be changed to issue an error, rather than a warning, in the
4605 You cannot use @code{asm} in this way in a function @emph{definition}; but
4606 you can get the same effect by writing a declaration for the function
4607 before its definition and putting @code{asm} there, like this:
4610 extern func () asm ("FUNC");
4617 It is up to you to make sure that the assembler names you choose do not
4618 conflict with any other assembler symbols. Also, you must not use a
4619 register name; that would produce completely invalid assembler code. GCC
4620 does not as yet have the ability to store static variables in registers.
4621 Perhaps that will be added.
4623 @node Explicit Reg Vars
4624 @section Variables in Specified Registers
4625 @cindex explicit register variables
4626 @cindex variables in specified registers
4627 @cindex specified registers
4628 @cindex registers, global allocation
4630 GNU C allows you to put a few global variables into specified hardware
4631 registers. You can also specify the register in which an ordinary
4632 register variable should be allocated.
4636 Global register variables reserve registers throughout the program.
4637 This may be useful in programs such as programming language
4638 interpreters which have a couple of global variables that are accessed
4642 Local register variables in specific registers do not reserve the
4643 registers, except at the point where they are used as input or output
4644 operands in an @code{asm} statement and the @code{asm} statement itself is
4645 not deleted. The compiler's data flow analysis is capable of determining
4646 where the specified registers contain live values, and where they are
4647 available for other uses. Stores into local register variables may be deleted
4648 when they appear to be dead according to dataflow analysis. References
4649 to local register variables may be deleted or moved or simplified.
4651 These local variables are sometimes convenient for use with the extended
4652 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4653 output of the assembler instruction directly into a particular register.
4654 (This will work provided the register you specify fits the constraints
4655 specified for that operand in the @code{asm}.)
4663 @node Global Reg Vars
4664 @subsection Defining Global Register Variables
4665 @cindex global register variables
4666 @cindex registers, global variables in
4668 You can define a global register variable in GNU C like this:
4671 register int *foo asm ("a5");
4675 Here @code{a5} is the name of the register which should be used. Choose a
4676 register which is normally saved and restored by function calls on your
4677 machine, so that library routines will not clobber it.
4679 Naturally the register name is cpu-dependent, so you would need to
4680 conditionalize your program according to cpu type. The register
4681 @code{a5} would be a good choice on a 68000 for a variable of pointer
4682 type. On machines with register windows, be sure to choose a ``global''
4683 register that is not affected magically by the function call mechanism.
4685 In addition, operating systems on one type of cpu may differ in how they
4686 name the registers; then you would need additional conditionals. For
4687 example, some 68000 operating systems call this register @code{%a5}.
4689 Eventually there may be a way of asking the compiler to choose a register
4690 automatically, but first we need to figure out how it should choose and
4691 how to enable you to guide the choice. No solution is evident.
4693 Defining a global register variable in a certain register reserves that
4694 register entirely for this use, at least within the current compilation.
4695 The register will not be allocated for any other purpose in the functions
4696 in the current compilation. The register will not be saved and restored by
4697 these functions. Stores into this register are never deleted even if they
4698 would appear to be dead, but references may be deleted or moved or
4701 It is not safe to access the global register variables from signal
4702 handlers, or from more than one thread of control, because the system
4703 library routines may temporarily use the register for other things (unless
4704 you recompile them specially for the task at hand).
4706 @cindex @code{qsort}, and global register variables
4707 It is not safe for one function that uses a global register variable to
4708 call another such function @code{foo} by way of a third function
4709 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4710 different source file in which the variable wasn't declared). This is
4711 because @code{lose} might save the register and put some other value there.
4712 For example, you can't expect a global register variable to be available in
4713 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4714 might have put something else in that register. (If you are prepared to
4715 recompile @code{qsort} with the same global register variable, you can
4716 solve this problem.)
4718 If you want to recompile @code{qsort} or other source files which do not
4719 actually use your global register variable, so that they will not use that
4720 register for any other purpose, then it suffices to specify the compiler
4721 option @option{-ffixed-@var{reg}}. You need not actually add a global
4722 register declaration to their source code.
4724 A function which can alter the value of a global register variable cannot
4725 safely be called from a function compiled without this variable, because it
4726 could clobber the value the caller expects to find there on return.
4727 Therefore, the function which is the entry point into the part of the
4728 program that uses the global register variable must explicitly save and
4729 restore the value which belongs to its caller.
4731 @cindex register variable after @code{longjmp}
4732 @cindex global register after @code{longjmp}
4733 @cindex value after @code{longjmp}
4736 On most machines, @code{longjmp} will restore to each global register
4737 variable the value it had at the time of the @code{setjmp}. On some
4738 machines, however, @code{longjmp} will not change the value of global
4739 register variables. To be portable, the function that called @code{setjmp}
4740 should make other arrangements to save the values of the global register
4741 variables, and to restore them in a @code{longjmp}. This way, the same
4742 thing will happen regardless of what @code{longjmp} does.
4744 All global register variable declarations must precede all function
4745 definitions. If such a declaration could appear after function
4746 definitions, the declaration would be too late to prevent the register from
4747 being used for other purposes in the preceding functions.
4749 Global register variables may not have initial values, because an
4750 executable file has no means to supply initial contents for a register.
4752 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4753 registers, but certain library functions, such as @code{getwd}, as well
4754 as the subroutines for division and remainder, modify g3 and g4. g1 and
4755 g2 are local temporaries.
4757 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4758 Of course, it will not do to use more than a few of those.
4760 @node Local Reg Vars
4761 @subsection Specifying Registers for Local Variables
4762 @cindex local variables, specifying registers
4763 @cindex specifying registers for local variables
4764 @cindex registers for local variables
4766 You can define a local register variable with a specified register
4770 register int *foo asm ("a5");
4774 Here @code{a5} is the name of the register which should be used. Note
4775 that this is the same syntax used for defining global register
4776 variables, but for a local variable it would appear within a function.
4778 Naturally the register name is cpu-dependent, but this is not a
4779 problem, since specific registers are most often useful with explicit
4780 assembler instructions (@pxref{Extended Asm}). Both of these things
4781 generally require that you conditionalize your program according to
4784 In addition, operating systems on one type of cpu may differ in how they
4785 name the registers; then you would need additional conditionals. For
4786 example, some 68000 operating systems call this register @code{%a5}.
4788 Defining such a register variable does not reserve the register; it
4789 remains available for other uses in places where flow control determines
4790 the variable's value is not live.
4792 This option does not guarantee that GCC will generate code that has
4793 this variable in the register you specify at all times. You may not
4794 code an explicit reference to this register in the @emph{assembler
4795 instruction template} part of an @code{asm} statement and assume it will
4796 always refer to this variable. However, using the variable as an
4797 @code{asm} @emph{operand} guarantees that the specified register is used
4800 Stores into local register variables may be deleted when they appear to be dead
4801 according to dataflow analysis. References to local register variables may
4802 be deleted or moved or simplified.
4804 As for global register variables, it's recommended that you choose a
4805 register which is normally saved and restored by function calls on
4806 your machine, so that library routines will not clobber it. A common
4807 pitfall is to initialize multiple call-clobbered registers with
4808 arbitrary expressions, where a function call or library call for an
4809 arithmetic operator will overwrite a register value from a previous
4810 assignment, for example @code{r0} below:
4812 register int *p1 asm ("r0") = @dots{};
4813 register int *p2 asm ("r1") = @dots{};
4815 In those cases, a solution is to use a temporary variable for
4816 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4818 @node Alternate Keywords
4819 @section Alternate Keywords
4820 @cindex alternate keywords
4821 @cindex keywords, alternate
4823 @option{-ansi} and the various @option{-std} options disable certain
4824 keywords. This causes trouble when you want to use GNU C extensions, or
4825 a general-purpose header file that should be usable by all programs,
4826 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4827 @code{inline} are not available in programs compiled with
4828 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4829 program compiled with @option{-std=c99}). The ISO C99 keyword
4830 @code{restrict} is only available when @option{-std=gnu99} (which will
4831 eventually be the default) or @option{-std=c99} (or the equivalent
4832 @option{-std=iso9899:1999}) is used.
4834 The way to solve these problems is to put @samp{__} at the beginning and
4835 end of each problematical keyword. For example, use @code{__asm__}
4836 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4838 Other C compilers won't accept these alternative keywords; if you want to
4839 compile with another compiler, you can define the alternate keywords as
4840 macros to replace them with the customary keywords. It looks like this:
4848 @findex __extension__
4850 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4852 prevent such warnings within one expression by writing
4853 @code{__extension__} before the expression. @code{__extension__} has no
4854 effect aside from this.
4856 @node Incomplete Enums
4857 @section Incomplete @code{enum} Types
4859 You can define an @code{enum} tag without specifying its possible values.
4860 This results in an incomplete type, much like what you get if you write
4861 @code{struct foo} without describing the elements. A later declaration
4862 which does specify the possible values completes the type.
4864 You can't allocate variables or storage using the type while it is
4865 incomplete. However, you can work with pointers to that type.
4867 This extension may not be very useful, but it makes the handling of
4868 @code{enum} more consistent with the way @code{struct} and @code{union}
4871 This extension is not supported by GNU C++.
4873 @node Function Names
4874 @section Function Names as Strings
4875 @cindex @code{__func__} identifier
4876 @cindex @code{__FUNCTION__} identifier
4877 @cindex @code{__PRETTY_FUNCTION__} identifier
4879 GCC provides three magic variables which hold the name of the current
4880 function, as a string. The first of these is @code{__func__}, which
4881 is part of the C99 standard:
4884 The identifier @code{__func__} is implicitly declared by the translator
4885 as if, immediately following the opening brace of each function
4886 definition, the declaration
4889 static const char __func__[] = "function-name";
4892 appeared, where function-name is the name of the lexically-enclosing
4893 function. This name is the unadorned name of the function.
4896 @code{__FUNCTION__} is another name for @code{__func__}. Older
4897 versions of GCC recognize only this name. However, it is not
4898 standardized. For maximum portability, we recommend you use
4899 @code{__func__}, but provide a fallback definition with the
4903 #if __STDC_VERSION__ < 199901L
4905 # define __func__ __FUNCTION__
4907 # define __func__ "<unknown>"
4912 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4913 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4914 the type signature of the function as well as its bare name. For
4915 example, this program:
4919 extern int printf (char *, ...);
4926 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4927 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4945 __PRETTY_FUNCTION__ = void a::sub(int)
4948 These identifiers are not preprocessor macros. In GCC 3.3 and
4949 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4950 were treated as string literals; they could be used to initialize
4951 @code{char} arrays, and they could be concatenated with other string
4952 literals. GCC 3.4 and later treat them as variables, like
4953 @code{__func__}. In C++, @code{__FUNCTION__} and
4954 @code{__PRETTY_FUNCTION__} have always been variables.
4956 @node Return Address
4957 @section Getting the Return or Frame Address of a Function
4959 These functions may be used to get information about the callers of a
4962 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4963 This function returns the return address of the current function, or of
4964 one of its callers. The @var{level} argument is number of frames to
4965 scan up the call stack. A value of @code{0} yields the return address
4966 of the current function, a value of @code{1} yields the return address
4967 of the caller of the current function, and so forth. When inlining
4968 the expected behavior is that the function will return the address of
4969 the function that will be returned to. To work around this behavior use
4970 the @code{noinline} function attribute.
4972 The @var{level} argument must be a constant integer.
4974 On some machines it may be impossible to determine the return address of
4975 any function other than the current one; in such cases, or when the top
4976 of the stack has been reached, this function will return @code{0} or a
4977 random value. In addition, @code{__builtin_frame_address} may be used
4978 to determine if the top of the stack has been reached.
4980 This function should only be used with a nonzero argument for debugging
4984 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4985 This function is similar to @code{__builtin_return_address}, but it
4986 returns the address of the function frame rather than the return address
4987 of the function. Calling @code{__builtin_frame_address} with a value of
4988 @code{0} yields the frame address of the current function, a value of
4989 @code{1} yields the frame address of the caller of the current function,
4992 The frame is the area on the stack which holds local variables and saved
4993 registers. The frame address is normally the address of the first word
4994 pushed on to the stack by the function. However, the exact definition
4995 depends upon the processor and the calling convention. If the processor
4996 has a dedicated frame pointer register, and the function has a frame,
4997 then @code{__builtin_frame_address} will return the value of the frame
5000 On some machines it may be impossible to determine the frame address of
5001 any function other than the current one; in such cases, or when the top
5002 of the stack has been reached, this function will return @code{0} if
5003 the first frame pointer is properly initialized by the startup code.
5005 This function should only be used with a nonzero argument for debugging
5009 @node Vector Extensions
5010 @section Using vector instructions through built-in functions
5012 On some targets, the instruction set contains SIMD vector instructions that
5013 operate on multiple values contained in one large register at the same time.
5014 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
5017 The first step in using these extensions is to provide the necessary data
5018 types. This should be done using an appropriate @code{typedef}:
5021 typedef int v4si __attribute__ ((vector_size (16)));
5024 The @code{int} type specifies the base type, while the attribute specifies
5025 the vector size for the variable, measured in bytes. For example, the
5026 declaration above causes the compiler to set the mode for the @code{v4si}
5027 type to be 16 bytes wide and divided into @code{int} sized units. For
5028 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
5029 corresponding mode of @code{foo} will be @acronym{V4SI}.
5031 The @code{vector_size} attribute is only applicable to integral and
5032 float scalars, although arrays, pointers, and function return values
5033 are allowed in conjunction with this construct.
5035 All the basic integer types can be used as base types, both as signed
5036 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
5037 @code{long long}. In addition, @code{float} and @code{double} can be
5038 used to build floating-point vector types.
5040 Specifying a combination that is not valid for the current architecture
5041 will cause GCC to synthesize the instructions using a narrower mode.
5042 For example, if you specify a variable of type @code{V4SI} and your
5043 architecture does not allow for this specific SIMD type, GCC will
5044 produce code that uses 4 @code{SIs}.
5046 The types defined in this manner can be used with a subset of normal C
5047 operations. Currently, GCC will allow using the following operators
5048 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5050 The operations behave like C++ @code{valarrays}. Addition is defined as
5051 the addition of the corresponding elements of the operands. For
5052 example, in the code below, each of the 4 elements in @var{a} will be
5053 added to the corresponding 4 elements in @var{b} and the resulting
5054 vector will be stored in @var{c}.
5057 typedef int v4si __attribute__ ((vector_size (16)));
5064 Subtraction, multiplication, division, and the logical operations
5065 operate in a similar manner. Likewise, the result of using the unary
5066 minus or complement operators on a vector type is a vector whose
5067 elements are the negative or complemented values of the corresponding
5068 elements in the operand.
5070 You can declare variables and use them in function calls and returns, as
5071 well as in assignments and some casts. You can specify a vector type as
5072 a return type for a function. Vector types can also be used as function
5073 arguments. It is possible to cast from one vector type to another,
5074 provided they are of the same size (in fact, you can also cast vectors
5075 to and from other datatypes of the same size).
5077 You cannot operate between vectors of different lengths or different
5078 signedness without a cast.
5080 A port that supports hardware vector operations, usually provides a set
5081 of built-in functions that can be used to operate on vectors. For
5082 example, a function to add two vectors and multiply the result by a
5083 third could look like this:
5086 v4si f (v4si a, v4si b, v4si c)
5088 v4si tmp = __builtin_addv4si (a, b);
5089 return __builtin_mulv4si (tmp, c);
5096 @findex __builtin_offsetof
5098 GCC implements for both C and C++ a syntactic extension to implement
5099 the @code{offsetof} macro.
5103 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5105 offsetof_member_designator:
5107 | offsetof_member_designator "." @code{identifier}
5108 | offsetof_member_designator "[" @code{expr} "]"
5111 This extension is sufficient such that
5114 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5117 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5118 may be dependent. In either case, @var{member} may consist of a single
5119 identifier, or a sequence of member accesses and array references.
5121 @node Atomic Builtins
5122 @section Built-in functions for atomic memory access
5124 The following builtins are intended to be compatible with those described
5125 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5126 section 7.4. As such, they depart from the normal GCC practice of using
5127 the ``__builtin_'' prefix, and further that they are overloaded such that
5128 they work on multiple types.
5130 The definition given in the Intel documentation allows only for the use of
5131 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5132 counterparts. GCC will allow any integral scalar or pointer type that is
5133 1, 2, 4 or 8 bytes in length.
5135 Not all operations are supported by all target processors. If a particular
5136 operation cannot be implemented on the target processor, a warning will be
5137 generated and a call an external function will be generated. The external
5138 function will carry the same name as the builtin, with an additional suffix
5139 @samp{_@var{n}} where @var{n} is the size of the data type.
5141 @c ??? Should we have a mechanism to suppress this warning? This is almost
5142 @c useful for implementing the operation under the control of an external
5145 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5146 no memory operand will be moved across the operation, either forward or
5147 backward. Further, instructions will be issued as necessary to prevent the
5148 processor from speculating loads across the operation and from queuing stores
5149 after the operation.
5151 All of the routines are are described in the Intel documentation to take
5152 ``an optional list of variables protected by the memory barrier''. It's
5153 not clear what is meant by that; it could mean that @emph{only} the
5154 following variables are protected, or it could mean that these variables
5155 should in addition be protected. At present GCC ignores this list and
5156 protects all variables which are globally accessible. If in the future
5157 we make some use of this list, an empty list will continue to mean all
5158 globally accessible variables.
5161 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5162 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5163 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5164 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5165 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5166 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5167 @findex __sync_fetch_and_add
5168 @findex __sync_fetch_and_sub
5169 @findex __sync_fetch_and_or
5170 @findex __sync_fetch_and_and
5171 @findex __sync_fetch_and_xor
5172 @findex __sync_fetch_and_nand
5173 These builtins perform the operation suggested by the name, and
5174 returns the value that had previously been in memory. That is,
5177 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5178 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5181 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5182 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5183 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5184 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5185 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5186 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5187 @findex __sync_add_and_fetch
5188 @findex __sync_sub_and_fetch
5189 @findex __sync_or_and_fetch
5190 @findex __sync_and_and_fetch
5191 @findex __sync_xor_and_fetch
5192 @findex __sync_nand_and_fetch
5193 These builtins perform the operation suggested by the name, and
5194 return the new value. That is,
5197 @{ *ptr @var{op}= value; return *ptr; @}
5198 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5201 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5202 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5203 @findex __sync_bool_compare_and_swap
5204 @findex __sync_val_compare_and_swap
5205 These builtins perform an atomic compare and swap. That is, if the current
5206 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5209 The ``bool'' version returns true if the comparison is successful and
5210 @var{newval} was written. The ``val'' version returns the contents
5211 of @code{*@var{ptr}} before the operation.
5213 @item __sync_synchronize (...)
5214 @findex __sync_synchronize
5215 This builtin issues a full memory barrier.
5217 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5218 @findex __sync_lock_test_and_set
5219 This builtin, as described by Intel, is not a traditional test-and-set
5220 operation, but rather an atomic exchange operation. It writes @var{value}
5221 into @code{*@var{ptr}}, and returns the previous contents of
5224 Many targets have only minimal support for such locks, and do not support
5225 a full exchange operation. In this case, a target may support reduced
5226 functionality here by which the @emph{only} valid value to store is the
5227 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5228 is implementation defined.
5230 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5231 This means that references after the builtin cannot move to (or be
5232 speculated to) before the builtin, but previous memory stores may not
5233 be globally visible yet, and previous memory loads may not yet be
5236 @item void __sync_lock_release (@var{type} *ptr, ...)
5237 @findex __sync_lock_release
5238 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5239 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5241 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5242 This means that all previous memory stores are globally visible, and all
5243 previous memory loads have been satisfied, but following memory reads
5244 are not prevented from being speculated to before the barrier.
5247 @node Object Size Checking
5248 @section Object Size Checking Builtins
5249 @findex __builtin_object_size
5250 @findex __builtin___memcpy_chk
5251 @findex __builtin___mempcpy_chk
5252 @findex __builtin___memmove_chk
5253 @findex __builtin___memset_chk
5254 @findex __builtin___strcpy_chk
5255 @findex __builtin___stpcpy_chk
5256 @findex __builtin___strncpy_chk
5257 @findex __builtin___strcat_chk
5258 @findex __builtin___strncat_chk
5259 @findex __builtin___sprintf_chk
5260 @findex __builtin___snprintf_chk
5261 @findex __builtin___vsprintf_chk
5262 @findex __builtin___vsnprintf_chk
5263 @findex __builtin___printf_chk
5264 @findex __builtin___vprintf_chk
5265 @findex __builtin___fprintf_chk
5266 @findex __builtin___vfprintf_chk
5268 GCC implements a limited buffer overflow protection mechanism
5269 that can prevent some buffer overflow attacks.
5271 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5272 is a built-in construct that returns a constant number of bytes from
5273 @var{ptr} to the end of the object @var{ptr} pointer points to
5274 (if known at compile time). @code{__builtin_object_size} never evaluates
5275 its arguments for side-effects. If there are any side-effects in them, it
5276 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5277 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5278 point to and all of them are known at compile time, the returned number
5279 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5280 0 and minimum if nonzero. If it is not possible to determine which objects
5281 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5282 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5283 for @var{type} 2 or 3.
5285 @var{type} is an integer constant from 0 to 3. If the least significant
5286 bit is clear, objects are whole variables, if it is set, a closest
5287 surrounding subobject is considered the object a pointer points to.
5288 The second bit determines if maximum or minimum of remaining bytes
5292 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5293 char *p = &var.buf1[1], *q = &var.b;
5295 /* Here the object p points to is var. */
5296 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5297 /* The subobject p points to is var.buf1. */
5298 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5299 /* The object q points to is var. */
5300 assert (__builtin_object_size (q, 0)
5301 == (char *) (&var + 1) - (char *) &var.b);
5302 /* The subobject q points to is var.b. */
5303 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5307 There are built-in functions added for many common string operation
5308 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5309 built-in is provided. This built-in has an additional last argument,
5310 which is the number of bytes remaining in object the @var{dest}
5311 argument points to or @code{(size_t) -1} if the size is not known.
5313 The built-in functions are optimized into the normal string functions
5314 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5315 it is known at compile time that the destination object will not
5316 be overflown. If the compiler can determine at compile time the
5317 object will be always overflown, it issues a warning.
5319 The intended use can be e.g.
5323 #define bos0(dest) __builtin_object_size (dest, 0)
5324 #define memcpy(dest, src, n) \
5325 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5329 /* It is unknown what object p points to, so this is optimized
5330 into plain memcpy - no checking is possible. */
5331 memcpy (p, "abcde", n);
5332 /* Destination is known and length too. It is known at compile
5333 time there will be no overflow. */
5334 memcpy (&buf[5], "abcde", 5);
5335 /* Destination is known, but the length is not known at compile time.
5336 This will result in __memcpy_chk call that can check for overflow
5338 memcpy (&buf[5], "abcde", n);
5339 /* Destination is known and it is known at compile time there will
5340 be overflow. There will be a warning and __memcpy_chk call that
5341 will abort the program at runtime. */
5342 memcpy (&buf[6], "abcde", 5);
5345 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5346 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5347 @code{strcat} and @code{strncat}.
5349 There are also checking built-in functions for formatted output functions.
5351 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5352 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5353 const char *fmt, ...);
5354 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5356 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5357 const char *fmt, va_list ap);
5360 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5361 etc. functions and can contain implementation specific flags on what
5362 additional security measures the checking function might take, such as
5363 handling @code{%n} differently.
5365 The @var{os} argument is the object size @var{s} points to, like in the
5366 other built-in functions. There is a small difference in the behavior
5367 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5368 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5369 the checking function is called with @var{os} argument set to
5372 In addition to this, there are checking built-in functions
5373 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5374 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5375 These have just one additional argument, @var{flag}, right before
5376 format string @var{fmt}. If the compiler is able to optimize them to
5377 @code{fputc} etc. functions, it will, otherwise the checking function
5378 should be called and the @var{flag} argument passed to it.
5380 @node Other Builtins
5381 @section Other built-in functions provided by GCC
5382 @cindex built-in functions
5383 @findex __builtin_isgreater
5384 @findex __builtin_isgreaterequal
5385 @findex __builtin_isless
5386 @findex __builtin_islessequal
5387 @findex __builtin_islessgreater
5388 @findex __builtin_isunordered
5389 @findex __builtin_powi
5390 @findex __builtin_powif
5391 @findex __builtin_powil
5549 @findex fprintf_unlocked
5551 @findex fputs_unlocked
5668 @findex printf_unlocked
5700 @findex significandf
5701 @findex significandl
5772 GCC provides a large number of built-in functions other than the ones
5773 mentioned above. Some of these are for internal use in the processing
5774 of exceptions or variable-length argument lists and will not be
5775 documented here because they may change from time to time; we do not
5776 recommend general use of these functions.
5778 The remaining functions are provided for optimization purposes.
5780 @opindex fno-builtin
5781 GCC includes built-in versions of many of the functions in the standard
5782 C library. The versions prefixed with @code{__builtin_} will always be
5783 treated as having the same meaning as the C library function even if you
5784 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5785 Many of these functions are only optimized in certain cases; if they are
5786 not optimized in a particular case, a call to the library function will
5791 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5792 @option{-std=c99}), the functions
5793 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5794 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5795 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5796 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked},
5797 @code{fputs_unlocked}, @code{gammaf}, @code{gammal}, @code{gamma},
5798 @code{gammaf_r}, @code{gammal_r}, @code{gamma_r}, @code{gettext},
5799 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5800 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5801 @code{lgammaf_r}, @code{lgammal_r}, @code{lgamma_r}, @code{mempcpy},
5802 @code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
5803 @code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
5804 @code{signbit}, @code{signbitf}, @code{signbitl}, @code{signbitd32},
5805 @code{signbitd64}, @code{signbitd128}, @code{significandf},
5806 @code{significandl}, @code{significand}, @code{sincosf},
5807 @code{sincosl}, @code{sincos}, @code{stpcpy}, @code{stpncpy},
5808 @code{strcasecmp}, @code{strdup}, @code{strfmon}, @code{strncasecmp},
5809 @code{strndup}, @code{toascii}, @code{y0f}, @code{y0l}, @code{y0},
5810 @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
5812 may be handled as built-in functions.
5813 All these functions have corresponding versions
5814 prefixed with @code{__builtin_}, which may be used even in strict C89
5817 The ISO C99 functions
5818 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5819 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5820 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5821 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5822 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5823 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5824 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5825 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5826 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5827 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5828 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5829 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5830 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5831 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5832 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5833 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5834 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5835 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5836 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5837 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5838 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5839 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5840 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5841 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5842 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5843 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5844 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5845 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5846 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5847 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5848 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5849 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5850 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5851 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5852 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5853 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5854 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5855 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5856 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5857 are handled as built-in functions
5858 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5860 There are also built-in versions of the ISO C99 functions
5861 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5862 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5863 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5864 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5865 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5866 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5867 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5868 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5869 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5870 that are recognized in any mode since ISO C90 reserves these names for
5871 the purpose to which ISO C99 puts them. All these functions have
5872 corresponding versions prefixed with @code{__builtin_}.
5874 The ISO C94 functions
5875 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5876 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5877 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5879 are handled as built-in functions
5880 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5882 The ISO C90 functions
5883 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5884 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5885 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5886 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5887 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5888 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5889 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5890 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5891 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
5892 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
5893 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
5894 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
5895 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
5896 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
5897 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
5898 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
5899 are all recognized as built-in functions unless
5900 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5901 is specified for an individual function). All of these functions have
5902 corresponding versions prefixed with @code{__builtin_}.
5904 GCC provides built-in versions of the ISO C99 floating point comparison
5905 macros that avoid raising exceptions for unordered operands. They have
5906 the same names as the standard macros ( @code{isgreater},
5907 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5908 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5909 prefixed. We intend for a library implementor to be able to simply
5910 @code{#define} each standard macro to its built-in equivalent.
5912 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5914 You can use the built-in function @code{__builtin_types_compatible_p} to
5915 determine whether two types are the same.
5917 This built-in function returns 1 if the unqualified versions of the
5918 types @var{type1} and @var{type2} (which are types, not expressions) are
5919 compatible, 0 otherwise. The result of this built-in function can be
5920 used in integer constant expressions.
5922 This built-in function ignores top level qualifiers (e.g., @code{const},
5923 @code{volatile}). For example, @code{int} is equivalent to @code{const
5926 The type @code{int[]} and @code{int[5]} are compatible. On the other
5927 hand, @code{int} and @code{char *} are not compatible, even if the size
5928 of their types, on the particular architecture are the same. Also, the
5929 amount of pointer indirection is taken into account when determining
5930 similarity. Consequently, @code{short *} is not similar to
5931 @code{short **}. Furthermore, two types that are typedefed are
5932 considered compatible if their underlying types are compatible.
5934 An @code{enum} type is not considered to be compatible with another
5935 @code{enum} type even if both are compatible with the same integer
5936 type; this is what the C standard specifies.
5937 For example, @code{enum @{foo, bar@}} is not similar to
5938 @code{enum @{hot, dog@}}.
5940 You would typically use this function in code whose execution varies
5941 depending on the arguments' types. For example:
5946 typeof (x) tmp = (x); \
5947 if (__builtin_types_compatible_p (typeof (x), long double)) \
5948 tmp = foo_long_double (tmp); \
5949 else if (__builtin_types_compatible_p (typeof (x), double)) \
5950 tmp = foo_double (tmp); \
5951 else if (__builtin_types_compatible_p (typeof (x), float)) \
5952 tmp = foo_float (tmp); \
5959 @emph{Note:} This construct is only available for C@.
5963 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5965 You can use the built-in function @code{__builtin_choose_expr} to
5966 evaluate code depending on the value of a constant expression. This
5967 built-in function returns @var{exp1} if @var{const_exp}, which is a
5968 constant expression that must be able to be determined at compile time,
5969 is nonzero. Otherwise it returns 0.
5971 This built-in function is analogous to the @samp{? :} operator in C,
5972 except that the expression returned has its type unaltered by promotion
5973 rules. Also, the built-in function does not evaluate the expression
5974 that was not chosen. For example, if @var{const_exp} evaluates to true,
5975 @var{exp2} is not evaluated even if it has side-effects.
5977 This built-in function can return an lvalue if the chosen argument is an
5980 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5981 type. Similarly, if @var{exp2} is returned, its return type is the same
5988 __builtin_choose_expr ( \
5989 __builtin_types_compatible_p (typeof (x), double), \
5991 __builtin_choose_expr ( \
5992 __builtin_types_compatible_p (typeof (x), float), \
5994 /* @r{The void expression results in a compile-time error} \
5995 @r{when assigning the result to something.} */ \
5999 @emph{Note:} This construct is only available for C@. Furthermore, the
6000 unused expression (@var{exp1} or @var{exp2} depending on the value of
6001 @var{const_exp}) may still generate syntax errors. This may change in
6006 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
6007 You can use the built-in function @code{__builtin_constant_p} to
6008 determine if a value is known to be constant at compile-time and hence
6009 that GCC can perform constant-folding on expressions involving that
6010 value. The argument of the function is the value to test. The function
6011 returns the integer 1 if the argument is known to be a compile-time
6012 constant and 0 if it is not known to be a compile-time constant. A
6013 return of 0 does not indicate that the value is @emph{not} a constant,
6014 but merely that GCC cannot prove it is a constant with the specified
6015 value of the @option{-O} option.
6017 You would typically use this function in an embedded application where
6018 memory was a critical resource. If you have some complex calculation,
6019 you may want it to be folded if it involves constants, but need to call
6020 a function if it does not. For example:
6023 #define Scale_Value(X) \
6024 (__builtin_constant_p (X) \
6025 ? ((X) * SCALE + OFFSET) : Scale (X))
6028 You may use this built-in function in either a macro or an inline
6029 function. However, if you use it in an inlined function and pass an
6030 argument of the function as the argument to the built-in, GCC will
6031 never return 1 when you call the inline function with a string constant
6032 or compound literal (@pxref{Compound Literals}) and will not return 1
6033 when you pass a constant numeric value to the inline function unless you
6034 specify the @option{-O} option.
6036 You may also use @code{__builtin_constant_p} in initializers for static
6037 data. For instance, you can write
6040 static const int table[] = @{
6041 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
6047 This is an acceptable initializer even if @var{EXPRESSION} is not a
6048 constant expression. GCC must be more conservative about evaluating the
6049 built-in in this case, because it has no opportunity to perform
6052 Previous versions of GCC did not accept this built-in in data
6053 initializers. The earliest version where it is completely safe is
6057 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6058 @opindex fprofile-arcs
6059 You may use @code{__builtin_expect} to provide the compiler with
6060 branch prediction information. In general, you should prefer to
6061 use actual profile feedback for this (@option{-fprofile-arcs}), as
6062 programmers are notoriously bad at predicting how their programs
6063 actually perform. However, there are applications in which this
6064 data is hard to collect.
6066 The return value is the value of @var{exp}, which should be an integral
6067 expression. The semantics of the built-in are that it is expected that
6068 @var{exp} == @var{c}. For example:
6071 if (__builtin_expect (x, 0))
6076 would indicate that we do not expect to call @code{foo}, since
6077 we expect @code{x} to be zero. Since you are limited to integral
6078 expressions for @var{exp}, you should use constructions such as
6081 if (__builtin_expect (ptr != NULL, 1))
6086 when testing pointer or floating-point values.
6089 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6090 This function is used to minimize cache-miss latency by moving data into
6091 a cache before it is accessed.
6092 You can insert calls to @code{__builtin_prefetch} into code for which
6093 you know addresses of data in memory that is likely to be accessed soon.
6094 If the target supports them, data prefetch instructions will be generated.
6095 If the prefetch is done early enough before the access then the data will
6096 be in the cache by the time it is accessed.
6098 The value of @var{addr} is the address of the memory to prefetch.
6099 There are two optional arguments, @var{rw} and @var{locality}.
6100 The value of @var{rw} is a compile-time constant one or zero; one
6101 means that the prefetch is preparing for a write to the memory address
6102 and zero, the default, means that the prefetch is preparing for a read.
6103 The value @var{locality} must be a compile-time constant integer between
6104 zero and three. A value of zero means that the data has no temporal
6105 locality, so it need not be left in the cache after the access. A value
6106 of three means that the data has a high degree of temporal locality and
6107 should be left in all levels of cache possible. Values of one and two
6108 mean, respectively, a low or moderate degree of temporal locality. The
6112 for (i = 0; i < n; i++)
6115 __builtin_prefetch (&a[i+j], 1, 1);
6116 __builtin_prefetch (&b[i+j], 0, 1);
6121 Data prefetch does not generate faults if @var{addr} is invalid, but
6122 the address expression itself must be valid. For example, a prefetch
6123 of @code{p->next} will not fault if @code{p->next} is not a valid
6124 address, but evaluation will fault if @code{p} is not a valid address.
6126 If the target does not support data prefetch, the address expression
6127 is evaluated if it includes side effects but no other code is generated
6128 and GCC does not issue a warning.
6131 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6132 Returns a positive infinity, if supported by the floating-point format,
6133 else @code{DBL_MAX}. This function is suitable for implementing the
6134 ISO C macro @code{HUGE_VAL}.
6137 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6138 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6141 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6142 Similar to @code{__builtin_huge_val}, except the return
6143 type is @code{long double}.
6146 @deftypefn {Built-in Function} double __builtin_inf (void)
6147 Similar to @code{__builtin_huge_val}, except a warning is generated
6148 if the target floating-point format does not support infinities.
6151 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6152 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6155 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6156 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6159 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6160 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6163 @deftypefn {Built-in Function} float __builtin_inff (void)
6164 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6165 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6168 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6169 Similar to @code{__builtin_inf}, except the return
6170 type is @code{long double}.
6173 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6174 This is an implementation of the ISO C99 function @code{nan}.
6176 Since ISO C99 defines this function in terms of @code{strtod}, which we
6177 do not implement, a description of the parsing is in order. The string
6178 is parsed as by @code{strtol}; that is, the base is recognized by
6179 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6180 in the significand such that the least significant bit of the number
6181 is at the least significant bit of the significand. The number is
6182 truncated to fit the significand field provided. The significand is
6183 forced to be a quiet NaN@.
6185 This function, if given a string literal all of which would have been
6186 consumed by strtol, is evaluated early enough that it is considered a
6187 compile-time constant.
6190 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6191 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6194 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6195 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6198 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6199 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6202 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6203 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6206 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6207 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6210 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6211 Similar to @code{__builtin_nan}, except the significand is forced
6212 to be a signaling NaN@. The @code{nans} function is proposed by
6213 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6216 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6217 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6220 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6221 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6224 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6225 Returns one plus the index of the least significant 1-bit of @var{x}, or
6226 if @var{x} is zero, returns zero.
6229 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6230 Returns the number of leading 0-bits in @var{x}, starting at the most
6231 significant bit position. If @var{x} is 0, the result is undefined.
6234 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6235 Returns the number of trailing 0-bits in @var{x}, starting at the least
6236 significant bit position. If @var{x} is 0, the result is undefined.
6239 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6240 Returns the number of 1-bits in @var{x}.
6243 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6244 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6248 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6249 Similar to @code{__builtin_ffs}, except the argument type is
6250 @code{unsigned long}.
6253 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6254 Similar to @code{__builtin_clz}, except the argument type is
6255 @code{unsigned long}.
6258 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6259 Similar to @code{__builtin_ctz}, except the argument type is
6260 @code{unsigned long}.
6263 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6264 Similar to @code{__builtin_popcount}, except the argument type is
6265 @code{unsigned long}.
6268 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6269 Similar to @code{__builtin_parity}, except the argument type is
6270 @code{unsigned long}.
6273 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6274 Similar to @code{__builtin_ffs}, except the argument type is
6275 @code{unsigned long long}.
6278 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6279 Similar to @code{__builtin_clz}, except the argument type is
6280 @code{unsigned long long}.
6283 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6284 Similar to @code{__builtin_ctz}, except the argument type is
6285 @code{unsigned long long}.
6288 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6289 Similar to @code{__builtin_popcount}, except the argument type is
6290 @code{unsigned long long}.
6293 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6294 Similar to @code{__builtin_parity}, except the argument type is
6295 @code{unsigned long long}.
6298 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6299 Returns the first argument raised to the power of the second. Unlike the
6300 @code{pow} function no guarantees about precision and rounding are made.
6303 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6304 Similar to @code{__builtin_powi}, except the argument and return types
6308 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6309 Similar to @code{__builtin_powi}, except the argument and return types
6310 are @code{long double}.
6313 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6314 Returns @var{x} with the order of the bytes reversed; for example,
6315 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6319 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6320 Similar to @code{__builtin_bswap32}, except the argument and return types
6324 @node Target Builtins
6325 @section Built-in Functions Specific to Particular Target Machines
6327 On some target machines, GCC supports many built-in functions specific
6328 to those machines. Generally these generate calls to specific machine
6329 instructions, but allow the compiler to schedule those calls.
6332 * Alpha Built-in Functions::
6333 * ARM Built-in Functions::
6334 * Blackfin Built-in Functions::
6335 * FR-V Built-in Functions::
6336 * X86 Built-in Functions::
6337 * MIPS DSP Built-in Functions::
6338 * MIPS Paired-Single Support::
6339 * PowerPC AltiVec Built-in Functions::
6340 * SPARC VIS Built-in Functions::
6341 * SPU Built-in Functions::
6344 @node Alpha Built-in Functions
6345 @subsection Alpha Built-in Functions
6347 These built-in functions are available for the Alpha family of
6348 processors, depending on the command-line switches used.
6350 The following built-in functions are always available. They
6351 all generate the machine instruction that is part of the name.
6354 long __builtin_alpha_implver (void)
6355 long __builtin_alpha_rpcc (void)
6356 long __builtin_alpha_amask (long)
6357 long __builtin_alpha_cmpbge (long, long)
6358 long __builtin_alpha_extbl (long, long)
6359 long __builtin_alpha_extwl (long, long)
6360 long __builtin_alpha_extll (long, long)
6361 long __builtin_alpha_extql (long, long)
6362 long __builtin_alpha_extwh (long, long)
6363 long __builtin_alpha_extlh (long, long)
6364 long __builtin_alpha_extqh (long, long)
6365 long __builtin_alpha_insbl (long, long)
6366 long __builtin_alpha_inswl (long, long)
6367 long __builtin_alpha_insll (long, long)
6368 long __builtin_alpha_insql (long, long)
6369 long __builtin_alpha_inswh (long, long)
6370 long __builtin_alpha_inslh (long, long)
6371 long __builtin_alpha_insqh (long, long)
6372 long __builtin_alpha_mskbl (long, long)
6373 long __builtin_alpha_mskwl (long, long)
6374 long __builtin_alpha_mskll (long, long)
6375 long __builtin_alpha_mskql (long, long)
6376 long __builtin_alpha_mskwh (long, long)
6377 long __builtin_alpha_msklh (long, long)
6378 long __builtin_alpha_mskqh (long, long)
6379 long __builtin_alpha_umulh (long, long)
6380 long __builtin_alpha_zap (long, long)
6381 long __builtin_alpha_zapnot (long, long)
6384 The following built-in functions are always with @option{-mmax}
6385 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6386 later. They all generate the machine instruction that is part
6390 long __builtin_alpha_pklb (long)
6391 long __builtin_alpha_pkwb (long)
6392 long __builtin_alpha_unpkbl (long)
6393 long __builtin_alpha_unpkbw (long)
6394 long __builtin_alpha_minub8 (long, long)
6395 long __builtin_alpha_minsb8 (long, long)
6396 long __builtin_alpha_minuw4 (long, long)
6397 long __builtin_alpha_minsw4 (long, long)
6398 long __builtin_alpha_maxub8 (long, long)
6399 long __builtin_alpha_maxsb8 (long, long)
6400 long __builtin_alpha_maxuw4 (long, long)
6401 long __builtin_alpha_maxsw4 (long, long)
6402 long __builtin_alpha_perr (long, long)
6405 The following built-in functions are always with @option{-mcix}
6406 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6407 later. They all generate the machine instruction that is part
6411 long __builtin_alpha_cttz (long)
6412 long __builtin_alpha_ctlz (long)
6413 long __builtin_alpha_ctpop (long)
6416 The following builtins are available on systems that use the OSF/1
6417 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6418 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6419 @code{rdval} and @code{wrval}.
6422 void *__builtin_thread_pointer (void)
6423 void __builtin_set_thread_pointer (void *)
6426 @node ARM Built-in Functions
6427 @subsection ARM Built-in Functions
6429 These built-in functions are available for the ARM family of
6430 processors, when the @option{-mcpu=iwmmxt} switch is used:
6433 typedef int v2si __attribute__ ((vector_size (8)));
6434 typedef short v4hi __attribute__ ((vector_size (8)));
6435 typedef char v8qi __attribute__ ((vector_size (8)));
6437 int __builtin_arm_getwcx (int)
6438 void __builtin_arm_setwcx (int, int)
6439 int __builtin_arm_textrmsb (v8qi, int)
6440 int __builtin_arm_textrmsh (v4hi, int)
6441 int __builtin_arm_textrmsw (v2si, int)
6442 int __builtin_arm_textrmub (v8qi, int)
6443 int __builtin_arm_textrmuh (v4hi, int)
6444 int __builtin_arm_textrmuw (v2si, int)
6445 v8qi __builtin_arm_tinsrb (v8qi, int)
6446 v4hi __builtin_arm_tinsrh (v4hi, int)
6447 v2si __builtin_arm_tinsrw (v2si, int)
6448 long long __builtin_arm_tmia (long long, int, int)
6449 long long __builtin_arm_tmiabb (long long, int, int)
6450 long long __builtin_arm_tmiabt (long long, int, int)
6451 long long __builtin_arm_tmiaph (long long, int, int)
6452 long long __builtin_arm_tmiatb (long long, int, int)
6453 long long __builtin_arm_tmiatt (long long, int, int)
6454 int __builtin_arm_tmovmskb (v8qi)
6455 int __builtin_arm_tmovmskh (v4hi)
6456 int __builtin_arm_tmovmskw (v2si)
6457 long long __builtin_arm_waccb (v8qi)
6458 long long __builtin_arm_wacch (v4hi)
6459 long long __builtin_arm_waccw (v2si)
6460 v8qi __builtin_arm_waddb (v8qi, v8qi)
6461 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6462 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6463 v4hi __builtin_arm_waddh (v4hi, v4hi)
6464 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6465 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6466 v2si __builtin_arm_waddw (v2si, v2si)
6467 v2si __builtin_arm_waddwss (v2si, v2si)
6468 v2si __builtin_arm_waddwus (v2si, v2si)
6469 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6470 long long __builtin_arm_wand(long long, long long)
6471 long long __builtin_arm_wandn (long long, long long)
6472 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6473 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6474 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6475 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6476 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6477 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6478 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6479 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6480 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6481 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6482 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6483 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6484 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6485 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6486 long long __builtin_arm_wmacsz (v4hi, v4hi)
6487 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6488 long long __builtin_arm_wmacuz (v4hi, v4hi)
6489 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6490 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6491 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6492 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6493 v2si __builtin_arm_wmaxsw (v2si, v2si)
6494 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6495 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6496 v2si __builtin_arm_wmaxuw (v2si, v2si)
6497 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6498 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6499 v2si __builtin_arm_wminsw (v2si, v2si)
6500 v8qi __builtin_arm_wminub (v8qi, v8qi)
6501 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6502 v2si __builtin_arm_wminuw (v2si, v2si)
6503 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6504 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6505 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6506 long long __builtin_arm_wor (long long, long long)
6507 v2si __builtin_arm_wpackdss (long long, long long)
6508 v2si __builtin_arm_wpackdus (long long, long long)
6509 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6510 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6511 v4hi __builtin_arm_wpackwss (v2si, v2si)
6512 v4hi __builtin_arm_wpackwus (v2si, v2si)
6513 long long __builtin_arm_wrord (long long, long long)
6514 long long __builtin_arm_wrordi (long long, int)
6515 v4hi __builtin_arm_wrorh (v4hi, long long)
6516 v4hi __builtin_arm_wrorhi (v4hi, int)
6517 v2si __builtin_arm_wrorw (v2si, long long)
6518 v2si __builtin_arm_wrorwi (v2si, int)
6519 v2si __builtin_arm_wsadb (v8qi, v8qi)
6520 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6521 v2si __builtin_arm_wsadh (v4hi, v4hi)
6522 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6523 v4hi __builtin_arm_wshufh (v4hi, int)
6524 long long __builtin_arm_wslld (long long, long long)
6525 long long __builtin_arm_wslldi (long long, int)
6526 v4hi __builtin_arm_wsllh (v4hi, long long)
6527 v4hi __builtin_arm_wsllhi (v4hi, int)
6528 v2si __builtin_arm_wsllw (v2si, long long)
6529 v2si __builtin_arm_wsllwi (v2si, int)
6530 long long __builtin_arm_wsrad (long long, long long)
6531 long long __builtin_arm_wsradi (long long, int)
6532 v4hi __builtin_arm_wsrah (v4hi, long long)
6533 v4hi __builtin_arm_wsrahi (v4hi, int)
6534 v2si __builtin_arm_wsraw (v2si, long long)
6535 v2si __builtin_arm_wsrawi (v2si, int)
6536 long long __builtin_arm_wsrld (long long, long long)
6537 long long __builtin_arm_wsrldi (long long, int)
6538 v4hi __builtin_arm_wsrlh (v4hi, long long)
6539 v4hi __builtin_arm_wsrlhi (v4hi, int)
6540 v2si __builtin_arm_wsrlw (v2si, long long)
6541 v2si __builtin_arm_wsrlwi (v2si, int)
6542 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6543 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6544 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6545 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6546 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6547 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6548 v2si __builtin_arm_wsubw (v2si, v2si)
6549 v2si __builtin_arm_wsubwss (v2si, v2si)
6550 v2si __builtin_arm_wsubwus (v2si, v2si)
6551 v4hi __builtin_arm_wunpckehsb (v8qi)
6552 v2si __builtin_arm_wunpckehsh (v4hi)
6553 long long __builtin_arm_wunpckehsw (v2si)
6554 v4hi __builtin_arm_wunpckehub (v8qi)
6555 v2si __builtin_arm_wunpckehuh (v4hi)
6556 long long __builtin_arm_wunpckehuw (v2si)
6557 v4hi __builtin_arm_wunpckelsb (v8qi)
6558 v2si __builtin_arm_wunpckelsh (v4hi)
6559 long long __builtin_arm_wunpckelsw (v2si)
6560 v4hi __builtin_arm_wunpckelub (v8qi)
6561 v2si __builtin_arm_wunpckeluh (v4hi)
6562 long long __builtin_arm_wunpckeluw (v2si)
6563 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6564 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6565 v2si __builtin_arm_wunpckihw (v2si, v2si)
6566 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6567 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6568 v2si __builtin_arm_wunpckilw (v2si, v2si)
6569 long long __builtin_arm_wxor (long long, long long)
6570 long long __builtin_arm_wzero ()
6573 @node Blackfin Built-in Functions
6574 @subsection Blackfin Built-in Functions
6576 Currently, there are two Blackfin-specific built-in functions. These are
6577 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6578 using inline assembly; by using these built-in functions the compiler can
6579 automatically add workarounds for hardware errata involving these
6580 instructions. These functions are named as follows:
6583 void __builtin_bfin_csync (void)
6584 void __builtin_bfin_ssync (void)
6587 @node FR-V Built-in Functions
6588 @subsection FR-V Built-in Functions
6590 GCC provides many FR-V-specific built-in functions. In general,
6591 these functions are intended to be compatible with those described
6592 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6593 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6594 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6595 pointer rather than by value.
6597 Most of the functions are named after specific FR-V instructions.
6598 Such functions are said to be ``directly mapped'' and are summarized
6599 here in tabular form.
6603 * Directly-mapped Integer Functions::
6604 * Directly-mapped Media Functions::
6605 * Raw read/write Functions::
6606 * Other Built-in Functions::
6609 @node Argument Types
6610 @subsubsection Argument Types
6612 The arguments to the built-in functions can be divided into three groups:
6613 register numbers, compile-time constants and run-time values. In order
6614 to make this classification clear at a glance, the arguments and return
6615 values are given the following pseudo types:
6617 @multitable @columnfractions .20 .30 .15 .35
6618 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6619 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6620 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6621 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6622 @item @code{uw2} @tab @code{unsigned long long} @tab No
6623 @tab an unsigned doubleword
6624 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6625 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6626 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6627 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6630 These pseudo types are not defined by GCC, they are simply a notational
6631 convenience used in this manual.
6633 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6634 and @code{sw2} are evaluated at run time. They correspond to
6635 register operands in the underlying FR-V instructions.
6637 @code{const} arguments represent immediate operands in the underlying
6638 FR-V instructions. They must be compile-time constants.
6640 @code{acc} arguments are evaluated at compile time and specify the number
6641 of an accumulator register. For example, an @code{acc} argument of 2
6642 will select the ACC2 register.
6644 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6645 number of an IACC register. See @pxref{Other Built-in Functions}
6648 @node Directly-mapped Integer Functions
6649 @subsubsection Directly-mapped Integer Functions
6651 The functions listed below map directly to FR-V I-type instructions.
6653 @multitable @columnfractions .45 .32 .23
6654 @item Function prototype @tab Example usage @tab Assembly output
6655 @item @code{sw1 __ADDSS (sw1, sw1)}
6656 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6657 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6658 @item @code{sw1 __SCAN (sw1, sw1)}
6659 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6660 @tab @code{SCAN @var{a},@var{b},@var{c}}
6661 @item @code{sw1 __SCUTSS (sw1)}
6662 @tab @code{@var{b} = __SCUTSS (@var{a})}
6663 @tab @code{SCUTSS @var{a},@var{b}}
6664 @item @code{sw1 __SLASS (sw1, sw1)}
6665 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6666 @tab @code{SLASS @var{a},@var{b},@var{c}}
6667 @item @code{void __SMASS (sw1, sw1)}
6668 @tab @code{__SMASS (@var{a}, @var{b})}
6669 @tab @code{SMASS @var{a},@var{b}}
6670 @item @code{void __SMSSS (sw1, sw1)}
6671 @tab @code{__SMSSS (@var{a}, @var{b})}
6672 @tab @code{SMSSS @var{a},@var{b}}
6673 @item @code{void __SMU (sw1, sw1)}
6674 @tab @code{__SMU (@var{a}, @var{b})}
6675 @tab @code{SMU @var{a},@var{b}}
6676 @item @code{sw2 __SMUL (sw1, sw1)}
6677 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6678 @tab @code{SMUL @var{a},@var{b},@var{c}}
6679 @item @code{sw1 __SUBSS (sw1, sw1)}
6680 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6681 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6682 @item @code{uw2 __UMUL (uw1, uw1)}
6683 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6684 @tab @code{UMUL @var{a},@var{b},@var{c}}
6687 @node Directly-mapped Media Functions
6688 @subsubsection Directly-mapped Media Functions
6690 The functions listed below map directly to FR-V M-type instructions.
6692 @multitable @columnfractions .45 .32 .23
6693 @item Function prototype @tab Example usage @tab Assembly output
6694 @item @code{uw1 __MABSHS (sw1)}
6695 @tab @code{@var{b} = __MABSHS (@var{a})}
6696 @tab @code{MABSHS @var{a},@var{b}}
6697 @item @code{void __MADDACCS (acc, acc)}
6698 @tab @code{__MADDACCS (@var{b}, @var{a})}
6699 @tab @code{MADDACCS @var{a},@var{b}}
6700 @item @code{sw1 __MADDHSS (sw1, sw1)}
6701 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6702 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6703 @item @code{uw1 __MADDHUS (uw1, uw1)}
6704 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6705 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6706 @item @code{uw1 __MAND (uw1, uw1)}
6707 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6708 @tab @code{MAND @var{a},@var{b},@var{c}}
6709 @item @code{void __MASACCS (acc, acc)}
6710 @tab @code{__MASACCS (@var{b}, @var{a})}
6711 @tab @code{MASACCS @var{a},@var{b}}
6712 @item @code{uw1 __MAVEH (uw1, uw1)}
6713 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6714 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6715 @item @code{uw2 __MBTOH (uw1)}
6716 @tab @code{@var{b} = __MBTOH (@var{a})}
6717 @tab @code{MBTOH @var{a},@var{b}}
6718 @item @code{void __MBTOHE (uw1 *, uw1)}
6719 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6720 @tab @code{MBTOHE @var{a},@var{b}}
6721 @item @code{void __MCLRACC (acc)}
6722 @tab @code{__MCLRACC (@var{a})}
6723 @tab @code{MCLRACC @var{a}}
6724 @item @code{void __MCLRACCA (void)}
6725 @tab @code{__MCLRACCA ()}
6726 @tab @code{MCLRACCA}
6727 @item @code{uw1 __Mcop1 (uw1, uw1)}
6728 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6729 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6730 @item @code{uw1 __Mcop2 (uw1, uw1)}
6731 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6732 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6733 @item @code{uw1 __MCPLHI (uw2, const)}
6734 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6735 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6736 @item @code{uw1 __MCPLI (uw2, const)}
6737 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6738 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6739 @item @code{void __MCPXIS (acc, sw1, sw1)}
6740 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6741 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6742 @item @code{void __MCPXIU (acc, uw1, uw1)}
6743 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6744 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6745 @item @code{void __MCPXRS (acc, sw1, sw1)}
6746 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6747 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6748 @item @code{void __MCPXRU (acc, uw1, uw1)}
6749 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6750 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6751 @item @code{uw1 __MCUT (acc, uw1)}
6752 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6753 @tab @code{MCUT @var{a},@var{b},@var{c}}
6754 @item @code{uw1 __MCUTSS (acc, sw1)}
6755 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6756 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6757 @item @code{void __MDADDACCS (acc, acc)}
6758 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6759 @tab @code{MDADDACCS @var{a},@var{b}}
6760 @item @code{void __MDASACCS (acc, acc)}
6761 @tab @code{__MDASACCS (@var{b}, @var{a})}
6762 @tab @code{MDASACCS @var{a},@var{b}}
6763 @item @code{uw2 __MDCUTSSI (acc, const)}
6764 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6765 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6766 @item @code{uw2 __MDPACKH (uw2, uw2)}
6767 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6768 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6769 @item @code{uw2 __MDROTLI (uw2, const)}
6770 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6771 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6772 @item @code{void __MDSUBACCS (acc, acc)}
6773 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6774 @tab @code{MDSUBACCS @var{a},@var{b}}
6775 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6776 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6777 @tab @code{MDUNPACKH @var{a},@var{b}}
6778 @item @code{uw2 __MEXPDHD (uw1, const)}
6779 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6780 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6781 @item @code{uw1 __MEXPDHW (uw1, const)}
6782 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6783 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6784 @item @code{uw1 __MHDSETH (uw1, const)}
6785 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6786 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6787 @item @code{sw1 __MHDSETS (const)}
6788 @tab @code{@var{b} = __MHDSETS (@var{a})}
6789 @tab @code{MHDSETS #@var{a},@var{b}}
6790 @item @code{uw1 __MHSETHIH (uw1, const)}
6791 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6792 @tab @code{MHSETHIH #@var{a},@var{b}}
6793 @item @code{sw1 __MHSETHIS (sw1, const)}
6794 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6795 @tab @code{MHSETHIS #@var{a},@var{b}}
6796 @item @code{uw1 __MHSETLOH (uw1, const)}
6797 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6798 @tab @code{MHSETLOH #@var{a},@var{b}}
6799 @item @code{sw1 __MHSETLOS (sw1, const)}
6800 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6801 @tab @code{MHSETLOS #@var{a},@var{b}}
6802 @item @code{uw1 __MHTOB (uw2)}
6803 @tab @code{@var{b} = __MHTOB (@var{a})}
6804 @tab @code{MHTOB @var{a},@var{b}}
6805 @item @code{void __MMACHS (acc, sw1, sw1)}
6806 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6807 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6808 @item @code{void __MMACHU (acc, uw1, uw1)}
6809 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6810 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6811 @item @code{void __MMRDHS (acc, sw1, sw1)}
6812 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6813 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6814 @item @code{void __MMRDHU (acc, uw1, uw1)}
6815 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6816 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6817 @item @code{void __MMULHS (acc, sw1, sw1)}
6818 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6819 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6820 @item @code{void __MMULHU (acc, uw1, uw1)}
6821 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6822 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6823 @item @code{void __MMULXHS (acc, sw1, sw1)}
6824 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6825 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6826 @item @code{void __MMULXHU (acc, uw1, uw1)}
6827 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6828 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6829 @item @code{uw1 __MNOT (uw1)}
6830 @tab @code{@var{b} = __MNOT (@var{a})}
6831 @tab @code{MNOT @var{a},@var{b}}
6832 @item @code{uw1 __MOR (uw1, uw1)}
6833 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6834 @tab @code{MOR @var{a},@var{b},@var{c}}
6835 @item @code{uw1 __MPACKH (uh, uh)}
6836 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6837 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6838 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6839 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6840 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6841 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6842 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6843 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6844 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6845 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6846 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6847 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6848 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6849 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6850 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6851 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6852 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6853 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6854 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6855 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6856 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6857 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6858 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6859 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6860 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6861 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6862 @item @code{void __MQMACHS (acc, sw2, sw2)}
6863 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6864 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6865 @item @code{void __MQMACHU (acc, uw2, uw2)}
6866 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6867 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6868 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6869 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6870 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6871 @item @code{void __MQMULHS (acc, sw2, sw2)}
6872 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6873 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6874 @item @code{void __MQMULHU (acc, uw2, uw2)}
6875 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6876 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6877 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6878 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6879 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6880 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6881 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6882 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6883 @item @code{sw2 __MQSATHS (sw2, sw2)}
6884 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6885 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6886 @item @code{uw2 __MQSLLHI (uw2, int)}
6887 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6888 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6889 @item @code{sw2 __MQSRAHI (sw2, int)}
6890 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6891 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6892 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6893 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6894 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6895 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6896 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6897 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6898 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6899 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6900 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6901 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6902 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6903 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6904 @item @code{uw1 __MRDACC (acc)}
6905 @tab @code{@var{b} = __MRDACC (@var{a})}
6906 @tab @code{MRDACC @var{a},@var{b}}
6907 @item @code{uw1 __MRDACCG (acc)}
6908 @tab @code{@var{b} = __MRDACCG (@var{a})}
6909 @tab @code{MRDACCG @var{a},@var{b}}
6910 @item @code{uw1 __MROTLI (uw1, const)}
6911 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6912 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6913 @item @code{uw1 __MROTRI (uw1, const)}
6914 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6915 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6916 @item @code{sw1 __MSATHS (sw1, sw1)}
6917 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6918 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6919 @item @code{uw1 __MSATHU (uw1, uw1)}
6920 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6921 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6922 @item @code{uw1 __MSLLHI (uw1, const)}
6923 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6924 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6925 @item @code{sw1 __MSRAHI (sw1, const)}
6926 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6927 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6928 @item @code{uw1 __MSRLHI (uw1, const)}
6929 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6930 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6931 @item @code{void __MSUBACCS (acc, acc)}
6932 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6933 @tab @code{MSUBACCS @var{a},@var{b}}
6934 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6935 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6936 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6937 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6938 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6939 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6940 @item @code{void __MTRAP (void)}
6941 @tab @code{__MTRAP ()}
6943 @item @code{uw2 __MUNPACKH (uw1)}
6944 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6945 @tab @code{MUNPACKH @var{a},@var{b}}
6946 @item @code{uw1 __MWCUT (uw2, uw1)}
6947 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6948 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6949 @item @code{void __MWTACC (acc, uw1)}
6950 @tab @code{__MWTACC (@var{b}, @var{a})}
6951 @tab @code{MWTACC @var{a},@var{b}}
6952 @item @code{void __MWTACCG (acc, uw1)}
6953 @tab @code{__MWTACCG (@var{b}, @var{a})}
6954 @tab @code{MWTACCG @var{a},@var{b}}
6955 @item @code{uw1 __MXOR (uw1, uw1)}
6956 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6957 @tab @code{MXOR @var{a},@var{b},@var{c}}
6960 @node Raw read/write Functions
6961 @subsubsection Raw read/write Functions
6963 This sections describes built-in functions related to read and write
6964 instructions to access memory. These functions generate
6965 @code{membar} instructions to flush the I/O load and stores where
6966 appropriate, as described in Fujitsu's manual described above.
6970 @item unsigned char __builtin_read8 (void *@var{data})
6971 @item unsigned short __builtin_read16 (void *@var{data})
6972 @item unsigned long __builtin_read32 (void *@var{data})
6973 @item unsigned long long __builtin_read64 (void *@var{data})
6975 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6976 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6977 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6978 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6981 @node Other Built-in Functions
6982 @subsubsection Other Built-in Functions
6984 This section describes built-in functions that are not named after
6985 a specific FR-V instruction.
6988 @item sw2 __IACCreadll (iacc @var{reg})
6989 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6990 for future expansion and must be 0.
6992 @item sw1 __IACCreadl (iacc @var{reg})
6993 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6994 Other values of @var{reg} are rejected as invalid.
6996 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6997 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6998 is reserved for future expansion and must be 0.
7000 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
7001 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
7002 is 1. Other values of @var{reg} are rejected as invalid.
7004 @item void __data_prefetch0 (const void *@var{x})
7005 Use the @code{dcpl} instruction to load the contents of address @var{x}
7006 into the data cache.
7008 @item void __data_prefetch (const void *@var{x})
7009 Use the @code{nldub} instruction to load the contents of address @var{x}
7010 into the data cache. The instruction will be issued in slot I1@.
7013 @node X86 Built-in Functions
7014 @subsection X86 Built-in Functions
7016 These built-in functions are available for the i386 and x86-64 family
7017 of computers, depending on the command-line switches used.
7019 Note that, if you specify command-line switches such as @option{-msse},
7020 the compiler could use the extended instruction sets even if the built-ins
7021 are not used explicitly in the program. For this reason, applications
7022 which perform runtime CPU detection must compile separate files for each
7023 supported architecture, using the appropriate flags. In particular,
7024 the file containing the CPU detection code should be compiled without
7027 The following machine modes are available for use with MMX built-in functions
7028 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
7029 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
7030 vector of eight 8-bit integers. Some of the built-in functions operate on
7031 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
7033 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
7034 of two 32-bit floating point values.
7036 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
7037 floating point values. Some instructions use a vector of four 32-bit
7038 integers, these use @code{V4SI}. Finally, some instructions operate on an
7039 entire vector register, interpreting it as a 128-bit integer, these use mode
7042 In the 64-bit mode, x86-64 family of processors uses additional built-in
7043 functions for efficient use of @code{TF} (@code{__float128}) 128-bit
7044 floating point and @code{TC} 128-bit complex floating point values.
7046 The following floating point built-in functions are made available in the
7047 64-bit mode. All of them implement the function that is part of the name.
7050 __float128 __builtin_fabsq (__float128)
7051 __float128 __builtin_copysignq (__float128, __float128)
7054 The following floating point built-in functions are made available in the
7058 @item __float128 __builtin_infq (void)
7059 Similar to @code{__builtin_inf}, except the return type is @code{__float128}.
7062 The following built-in functions are made available by @option{-mmmx}.
7063 All of them generate the machine instruction that is part of the name.
7066 v8qi __builtin_ia32_paddb (v8qi, v8qi)
7067 v4hi __builtin_ia32_paddw (v4hi, v4hi)
7068 v2si __builtin_ia32_paddd (v2si, v2si)
7069 v8qi __builtin_ia32_psubb (v8qi, v8qi)
7070 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7071 v2si __builtin_ia32_psubd (v2si, v2si)
7072 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7073 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7074 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7075 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7076 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7077 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7078 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7079 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7080 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7081 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7082 di __builtin_ia32_pand (di, di)
7083 di __builtin_ia32_pandn (di,di)
7084 di __builtin_ia32_por (di, di)
7085 di __builtin_ia32_pxor (di, di)
7086 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7087 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7088 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7089 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7090 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7091 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7092 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7093 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7094 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7095 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7096 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7097 v2si __builtin_ia32_punpckldq (v2si, v2si)
7098 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7099 v4hi __builtin_ia32_packssdw (v2si, v2si)
7100 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7103 The following built-in functions are made available either with
7104 @option{-msse}, or with a combination of @option{-m3dnow} and
7105 @option{-march=athlon}. All of them generate the machine
7106 instruction that is part of the name.
7109 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7110 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7111 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7112 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
7113 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7114 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7115 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7116 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7117 int __builtin_ia32_pextrw (v4hi, int)
7118 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7119 int __builtin_ia32_pmovmskb (v8qi)
7120 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7121 void __builtin_ia32_movntq (di *, di)
7122 void __builtin_ia32_sfence (void)
7125 The following built-in functions are available when @option{-msse} is used.
7126 All of them generate the machine instruction that is part of the name.
7129 int __builtin_ia32_comieq (v4sf, v4sf)
7130 int __builtin_ia32_comineq (v4sf, v4sf)
7131 int __builtin_ia32_comilt (v4sf, v4sf)
7132 int __builtin_ia32_comile (v4sf, v4sf)
7133 int __builtin_ia32_comigt (v4sf, v4sf)
7134 int __builtin_ia32_comige (v4sf, v4sf)
7135 int __builtin_ia32_ucomieq (v4sf, v4sf)
7136 int __builtin_ia32_ucomineq (v4sf, v4sf)
7137 int __builtin_ia32_ucomilt (v4sf, v4sf)
7138 int __builtin_ia32_ucomile (v4sf, v4sf)
7139 int __builtin_ia32_ucomigt (v4sf, v4sf)
7140 int __builtin_ia32_ucomige (v4sf, v4sf)
7141 v4sf __builtin_ia32_addps (v4sf, v4sf)
7142 v4sf __builtin_ia32_subps (v4sf, v4sf)
7143 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7144 v4sf __builtin_ia32_divps (v4sf, v4sf)
7145 v4sf __builtin_ia32_addss (v4sf, v4sf)
7146 v4sf __builtin_ia32_subss (v4sf, v4sf)
7147 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7148 v4sf __builtin_ia32_divss (v4sf, v4sf)
7149 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7150 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7151 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7152 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7153 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7154 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7155 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7156 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7157 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7158 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7159 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7160 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7161 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7162 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7163 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7164 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7165 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7166 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7167 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7168 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7169 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7170 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7171 v4sf __builtin_ia32_minps (v4sf, v4sf)
7172 v4sf __builtin_ia32_minss (v4sf, v4sf)
7173 v4sf __builtin_ia32_andps (v4sf, v4sf)
7174 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7175 v4sf __builtin_ia32_orps (v4sf, v4sf)
7176 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7177 v4sf __builtin_ia32_movss (v4sf, v4sf)
7178 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7179 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7180 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7181 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7182 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7183 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7184 v2si __builtin_ia32_cvtps2pi (v4sf)
7185 int __builtin_ia32_cvtss2si (v4sf)
7186 v2si __builtin_ia32_cvttps2pi (v4sf)
7187 int __builtin_ia32_cvttss2si (v4sf)
7188 v4sf __builtin_ia32_rcpps (v4sf)
7189 v4sf __builtin_ia32_rsqrtps (v4sf)
7190 v4sf __builtin_ia32_sqrtps (v4sf)
7191 v4sf __builtin_ia32_rcpss (v4sf)
7192 v4sf __builtin_ia32_rsqrtss (v4sf)
7193 v4sf __builtin_ia32_sqrtss (v4sf)
7194 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7195 void __builtin_ia32_movntps (float *, v4sf)
7196 int __builtin_ia32_movmskps (v4sf)
7199 The following built-in functions are available when @option{-msse} is used.
7202 @item v4sf __builtin_ia32_loadaps (float *)
7203 Generates the @code{movaps} machine instruction as a load from memory.
7204 @item void __builtin_ia32_storeaps (float *, v4sf)
7205 Generates the @code{movaps} machine instruction as a store to memory.
7206 @item v4sf __builtin_ia32_loadups (float *)
7207 Generates the @code{movups} machine instruction as a load from memory.
7208 @item void __builtin_ia32_storeups (float *, v4sf)
7209 Generates the @code{movups} machine instruction as a store to memory.
7210 @item v4sf __builtin_ia32_loadsss (float *)
7211 Generates the @code{movss} machine instruction as a load from memory.
7212 @item void __builtin_ia32_storess (float *, v4sf)
7213 Generates the @code{movss} machine instruction as a store to memory.
7214 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7215 Generates the @code{movhps} machine instruction as a load from memory.
7216 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7217 Generates the @code{movlps} machine instruction as a load from memory
7218 @item void __builtin_ia32_storehps (v4sf, v2si *)
7219 Generates the @code{movhps} machine instruction as a store to memory.
7220 @item void __builtin_ia32_storelps (v4sf, v2si *)
7221 Generates the @code{movlps} machine instruction as a store to memory.
7224 The following built-in functions are available when @option{-msse2} is used.
7225 All of them generate the machine instruction that is part of the name.
7228 int __builtin_ia32_comisdeq (v2df, v2df)
7229 int __builtin_ia32_comisdlt (v2df, v2df)
7230 int __builtin_ia32_comisdle (v2df, v2df)
7231 int __builtin_ia32_comisdgt (v2df, v2df)
7232 int __builtin_ia32_comisdge (v2df, v2df)
7233 int __builtin_ia32_comisdneq (v2df, v2df)
7234 int __builtin_ia32_ucomisdeq (v2df, v2df)
7235 int __builtin_ia32_ucomisdlt (v2df, v2df)
7236 int __builtin_ia32_ucomisdle (v2df, v2df)
7237 int __builtin_ia32_ucomisdgt (v2df, v2df)
7238 int __builtin_ia32_ucomisdge (v2df, v2df)
7239 int __builtin_ia32_ucomisdneq (v2df, v2df)
7240 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7241 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7242 v2df __builtin_ia32_cmplepd (v2df, v2df)
7243 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7244 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7245 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7246 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7247 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7248 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7249 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7250 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7251 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7252 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7253 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7254 v2df __builtin_ia32_cmplesd (v2df, v2df)
7255 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7256 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7257 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7258 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7259 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7260 v2di __builtin_ia32_paddq (v2di, v2di)
7261 v2di __builtin_ia32_psubq (v2di, v2di)
7262 v2df __builtin_ia32_addpd (v2df, v2df)
7263 v2df __builtin_ia32_subpd (v2df, v2df)
7264 v2df __builtin_ia32_mulpd (v2df, v2df)
7265 v2df __builtin_ia32_divpd (v2df, v2df)
7266 v2df __builtin_ia32_addsd (v2df, v2df)
7267 v2df __builtin_ia32_subsd (v2df, v2df)
7268 v2df __builtin_ia32_mulsd (v2df, v2df)
7269 v2df __builtin_ia32_divsd (v2df, v2df)
7270 v2df __builtin_ia32_minpd (v2df, v2df)
7271 v2df __builtin_ia32_maxpd (v2df, v2df)
7272 v2df __builtin_ia32_minsd (v2df, v2df)
7273 v2df __builtin_ia32_maxsd (v2df, v2df)
7274 v2df __builtin_ia32_andpd (v2df, v2df)
7275 v2df __builtin_ia32_andnpd (v2df, v2df)
7276 v2df __builtin_ia32_orpd (v2df, v2df)
7277 v2df __builtin_ia32_xorpd (v2df, v2df)
7278 v2df __builtin_ia32_movsd (v2df, v2df)
7279 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7280 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7281 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7282 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7283 v4si __builtin_ia32_paddd128 (v4si, v4si)
7284 v2di __builtin_ia32_paddq128 (v2di, v2di)
7285 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7286 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7287 v4si __builtin_ia32_psubd128 (v4si, v4si)
7288 v2di __builtin_ia32_psubq128 (v2di, v2di)
7289 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7290 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7291 v2di __builtin_ia32_pand128 (v2di, v2di)
7292 v2di __builtin_ia32_pandn128 (v2di, v2di)
7293 v2di __builtin_ia32_por128 (v2di, v2di)
7294 v2di __builtin_ia32_pxor128 (v2di, v2di)
7295 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7296 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7297 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7298 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7299 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7300 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7301 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7302 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7303 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7304 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7305 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7306 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7307 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7308 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7309 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7310 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7311 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7312 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7313 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7314 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7315 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7316 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7317 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7318 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7319 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7320 v2df __builtin_ia32_loadupd (double *)
7321 void __builtin_ia32_storeupd (double *, v2df)
7322 v2df __builtin_ia32_loadhpd (v2df, double *)
7323 v2df __builtin_ia32_loadlpd (v2df, double *)
7324 int __builtin_ia32_movmskpd (v2df)
7325 int __builtin_ia32_pmovmskb128 (v16qi)
7326 void __builtin_ia32_movnti (int *, int)
7327 void __builtin_ia32_movntpd (double *, v2df)
7328 void __builtin_ia32_movntdq (v2df *, v2df)
7329 v4si __builtin_ia32_pshufd (v4si, int)
7330 v8hi __builtin_ia32_pshuflw (v8hi, int)
7331 v8hi __builtin_ia32_pshufhw (v8hi, int)
7332 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7333 v2df __builtin_ia32_sqrtpd (v2df)
7334 v2df __builtin_ia32_sqrtsd (v2df)
7335 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7336 v2df __builtin_ia32_cvtdq2pd (v4si)
7337 v4sf __builtin_ia32_cvtdq2ps (v4si)
7338 v4si __builtin_ia32_cvtpd2dq (v2df)
7339 v2si __builtin_ia32_cvtpd2pi (v2df)
7340 v4sf __builtin_ia32_cvtpd2ps (v2df)
7341 v4si __builtin_ia32_cvttpd2dq (v2df)
7342 v2si __builtin_ia32_cvttpd2pi (v2df)
7343 v2df __builtin_ia32_cvtpi2pd (v2si)
7344 int __builtin_ia32_cvtsd2si (v2df)
7345 int __builtin_ia32_cvttsd2si (v2df)
7346 long long __builtin_ia32_cvtsd2si64 (v2df)
7347 long long __builtin_ia32_cvttsd2si64 (v2df)
7348 v4si __builtin_ia32_cvtps2dq (v4sf)
7349 v2df __builtin_ia32_cvtps2pd (v4sf)
7350 v4si __builtin_ia32_cvttps2dq (v4sf)
7351 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7352 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7353 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7354 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7355 void __builtin_ia32_clflush (const void *)
7356 void __builtin_ia32_lfence (void)
7357 void __builtin_ia32_mfence (void)
7358 v16qi __builtin_ia32_loaddqu (const char *)
7359 void __builtin_ia32_storedqu (char *, v16qi)
7360 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7361 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7362 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7363 v4si __builtin_ia32_pslld128 (v4si, v2di)
7364 v2di __builtin_ia32_psllq128 (v4si, v2di)
7365 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7366 v4si __builtin_ia32_psrld128 (v4si, v2di)
7367 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7368 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7369 v4si __builtin_ia32_psrad128 (v4si, v2di)
7370 v2di __builtin_ia32_pslldqi128 (v2di, int)
7371 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7372 v4si __builtin_ia32_pslldi128 (v4si, int)
7373 v2di __builtin_ia32_psllqi128 (v2di, int)
7374 v2di __builtin_ia32_psrldqi128 (v2di, int)
7375 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7376 v4si __builtin_ia32_psrldi128 (v4si, int)
7377 v2di __builtin_ia32_psrlqi128 (v2di, int)
7378 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7379 v4si __builtin_ia32_psradi128 (v4si, int)
7380 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7383 The following built-in functions are available when @option{-msse3} is used.
7384 All of them generate the machine instruction that is part of the name.
7387 v2df __builtin_ia32_addsubpd (v2df, v2df)
7388 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7389 v2df __builtin_ia32_haddpd (v2df, v2df)
7390 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7391 v2df __builtin_ia32_hsubpd (v2df, v2df)
7392 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7393 v16qi __builtin_ia32_lddqu (char const *)
7394 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7395 v2df __builtin_ia32_movddup (v2df)
7396 v4sf __builtin_ia32_movshdup (v4sf)
7397 v4sf __builtin_ia32_movsldup (v4sf)
7398 void __builtin_ia32_mwait (unsigned int, unsigned int)
7401 The following built-in functions are available when @option{-msse3} is used.
7404 @item v2df __builtin_ia32_loadddup (double const *)
7405 Generates the @code{movddup} machine instruction as a load from memory.
7408 The following built-in functions are available when @option{-mssse3} is used.
7409 All of them generate the machine instruction that is part of the name
7413 v2si __builtin_ia32_phaddd (v2si, v2si)
7414 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7415 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7416 v2si __builtin_ia32_phsubd (v2si, v2si)
7417 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7418 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7419 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7420 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7421 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7422 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7423 v2si __builtin_ia32_psignd (v2si, v2si)
7424 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7425 long long __builtin_ia32_palignr (long long, long long, int)
7426 v8qi __builtin_ia32_pabsb (v8qi)
7427 v2si __builtin_ia32_pabsd (v2si)
7428 v4hi __builtin_ia32_pabsw (v4hi)
7431 The following built-in functions are available when @option{-mssse3} is used.
7432 All of them generate the machine instruction that is part of the name
7436 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7437 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7438 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7439 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7440 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7441 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7442 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7443 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7444 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7445 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7446 v4si __builtin_ia32_psignd128 (v4si, v4si)
7447 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7448 v2di __builtin_ia32_palignr (v2di, v2di, int)
7449 v16qi __builtin_ia32_pabsb128 (v16qi)
7450 v4si __builtin_ia32_pabsd128 (v4si)
7451 v8hi __builtin_ia32_pabsw128 (v8hi)
7454 The following built-in functions are available when @option{-msse4.1} is
7455 used. All of them generate the machine instruction that is part of the
7459 v2df __builtin_ia32_blendpd (v2df, v2df, const int)
7460 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int)
7461 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df)
7462 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf)
7463 v2df __builtin_ia32_dppd (v2df, v2df, const int)
7464 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int)
7465 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int)
7466 v2di __builtin_ia32_movntdqa (v2di *);
7467 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int)
7468 v8hi __builtin_ia32_packusdw128 (v4si, v4si)
7469 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi)
7470 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int)
7471 v2di __builtin_ia32_pcmpeqq (v2di, v2di)
7472 v8hi __builtin_ia32_phminposuw128 (v8hi)
7473 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi)
7474 v4si __builtin_ia32_pmaxsd128 (v4si, v4si)
7475 v4si __builtin_ia32_pmaxud128 (v4si, v4si)
7476 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi)
7477 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi)
7478 v4si __builtin_ia32_pminsd128 (v4si, v4si)
7479 v4si __builtin_ia32_pminud128 (v4si, v4si)
7480 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi)
7481 v4si __builtin_ia32_pmovsxbd128 (v16qi)
7482 v2di __builtin_ia32_pmovsxbq128 (v16qi)
7483 v8hi __builtin_ia32_pmovsxbw128 (v16qi)
7484 v2di __builtin_ia32_pmovsxdq128 (v4si)
7485 v4si __builtin_ia32_pmovsxwd128 (v8hi)
7486 v2di __builtin_ia32_pmovsxwq128 (v8hi)
7487 v4si __builtin_ia32_pmovzxbd128 (v16qi)
7488 v2di __builtin_ia32_pmovzxbq128 (v16qi)
7489 v8hi __builtin_ia32_pmovzxbw128 (v16qi)
7490 v2di __builtin_ia32_pmovzxdq128 (v4si)
7491 v4si __builtin_ia32_pmovzxwd128 (v8hi)
7492 v2di __builtin_ia32_pmovzxwq128 (v8hi)
7493 v2di __builtin_ia32_pmuldq128 (v4si, v4si)
7494 v4si __builtin_ia32_pmulld128 (v4si, v4si)
7495 int __builtin_ia32_ptestc128 (v2di, v2di)
7496 int __builtin_ia32_ptestnzc128 (v2di, v2di)
7497 int __builtin_ia32_ptestz128 (v2di, v2di)
7498 v2df __builtin_ia32_roundpd (v2df, const int)
7499 v4sf __builtin_ia32_roundps (v4sf, const int)
7500 v2df __builtin_ia32_roundsd (v2df, v2df, const int)
7501 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int)
7504 The following built-in functions are available when @option{-msse4.1} is
7508 @item v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)
7509 Generates the @code{insertps} machine instruction.
7510 @item int __builtin_ia32_vec_ext_v16qi (v16qi, const int)
7511 Generates the @code{pextrb} machine instruction.
7512 @item v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)
7513 Generates the @code{pinsrb} machine instruction.
7514 @item v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)
7515 Generates the @code{pinsrd} machine instruction.
7516 @item v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)
7517 Generates the @code{pinsrq} machine instruction in 64bit mode.
7520 The following built-in functions are changed to generate new SSE4.1
7521 instructions when @option{-msse4.1} is used.
7524 @item float __builtin_ia32_vec_ext_v4sf (v4sf, const int)
7525 Generates the @code{extractps} machine instruction.
7526 @item int __builtin_ia32_vec_ext_v4si (v4si, const int)
7527 Generates the @code{pextrd} machine instruction.
7528 @item long long __builtin_ia32_vec_ext_v2di (v2di, const int)
7529 Generates the @code{pextrq} machine instruction in 64bit mode.
7532 The following built-in functions are available when @option{-msse4.2} is
7533 used. All of them generate the machine instruction that is part of the
7537 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int)
7538 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int)
7539 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int)
7540 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int)
7541 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int)
7542 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int)
7543 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int)
7544 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int)
7545 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int)
7546 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int)
7547 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int)
7548 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int)
7549 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int)
7550 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int)
7551 v2di __builtin_ia32_pcmpgtq (v2di, v2di)
7554 The following built-in functions are available when @option{-msse4.2} is
7558 @item unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)
7559 Generates the @code{crc32b} machine instruction.
7560 @item unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)
7561 Generates the @code{crc32w} machine instruction.
7562 @item unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)
7563 Generates the @code{crc32l} machine instruction.
7564 @item unsigned long long __builtin_ia32_crc32di (unsigned int, unsigned long long)
7567 The following built-in functions are changed to generate new SSE4.2
7568 instructions when @option{-msse4.2} is used.
7571 @item int __builtin_popcount (unsigned int)
7572 Generates the @code{popcntl} machine instruction.
7573 @item int __builtin_popcountl (unsigned long)
7574 Generates the @code{popcntl} or @code{popcntq} machine instruction,
7575 depending on the size of @code{unsigned long}.
7576 @item int __builtin_popcountll (unsigned long long)
7577 Generates the @code{popcntq} machine instruction.
7580 The following built-in functions are available when @option{-msse4a} is used.
7581 All of them generate the machine instruction that is part of the name.
7584 void __builtin_ia32_movntsd (double *, v2df)
7585 void __builtin_ia32_movntss (float *, v4sf)
7586 v2di __builtin_ia32_extrq (v2di, v16qi)
7587 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int)
7588 v2di __builtin_ia32_insertq (v2di, v2di)
7589 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int)
7592 The following built-in functions are available when @option{-m3dnow} is used.
7593 All of them generate the machine instruction that is part of the name.
7596 void __builtin_ia32_femms (void)
7597 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7598 v2si __builtin_ia32_pf2id (v2sf)
7599 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7600 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7601 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7602 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7603 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7604 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7605 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7606 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7607 v2sf __builtin_ia32_pfrcp (v2sf)
7608 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7609 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7610 v2sf __builtin_ia32_pfrsqrt (v2sf)
7611 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7612 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7613 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7614 v2sf __builtin_ia32_pi2fd (v2si)
7615 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7618 The following built-in functions are available when both @option{-m3dnow}
7619 and @option{-march=athlon} are used. All of them generate the machine
7620 instruction that is part of the name.
7623 v2si __builtin_ia32_pf2iw (v2sf)
7624 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7625 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7626 v2sf __builtin_ia32_pi2fw (v2si)
7627 v2sf __builtin_ia32_pswapdsf (v2sf)
7628 v2si __builtin_ia32_pswapdsi (v2si)
7631 @node MIPS DSP Built-in Functions
7632 @subsection MIPS DSP Built-in Functions
7634 The MIPS DSP Application-Specific Extension (ASE) includes new
7635 instructions that are designed to improve the performance of DSP and
7636 media applications. It provides instructions that operate on packed
7637 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
7639 GCC supports MIPS DSP operations using both the generic
7640 vector extensions (@pxref{Vector Extensions}) and a collection of
7641 MIPS-specific built-in functions. Both kinds of support are
7642 enabled by the @option{-mdsp} command-line option.
7644 Revision 2 of the ASE was introduced in the second half of 2006.
7645 This revision adds extra instructions to the original ASE, but is
7646 otherwise backwards-compatible with it. You can select revision 2
7647 using the command-line option @option{-mdspr2}; this option implies
7650 At present, GCC only provides support for operations on 32-bit
7651 vectors. The vector type associated with 8-bit integer data is
7652 usually called @code{v4i8}, the vector type associated with Q7
7653 is usually called @code{v4q7}, the vector type associated with 16-bit
7654 integer data is usually called @code{v2i16}, and the vector type
7655 associated with Q15 is usually called @code{v2q15}. They can be
7656 defined in C as follows:
7659 typedef signed char v4i8 __attribute__ ((vector_size(4)));
7660 typedef signed char v4q7 __attribute__ ((vector_size(4)));
7661 typedef short v2i16 __attribute__ ((vector_size(4)));
7662 typedef short v2q15 __attribute__ ((vector_size(4)));
7665 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
7666 initialized in the same way as aggregates. For example:
7669 v4i8 a = @{1, 2, 3, 4@};
7671 b = (v4i8) @{5, 6, 7, 8@};
7673 v2q15 c = @{0x0fcb, 0x3a75@};
7675 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7678 @emph{Note:} The CPU's endianness determines the order in which values
7679 are packed. On little-endian targets, the first value is the least
7680 significant and the last value is the most significant. The opposite
7681 order applies to big-endian targets. For example, the code above will
7682 set the lowest byte of @code{a} to @code{1} on little-endian targets
7683 and @code{4} on big-endian targets.
7685 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
7686 representation. As shown in this example, the integer representation
7687 of a Q7 value can be obtained by multiplying the fractional value by
7688 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
7689 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7692 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7693 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7694 and @code{c} and @code{d} are @code{v2q15} values.
7696 @multitable @columnfractions .50 .50
7697 @item C code @tab MIPS instruction
7698 @item @code{a + b} @tab @code{addu.qb}
7699 @item @code{c + d} @tab @code{addq.ph}
7700 @item @code{a - b} @tab @code{subu.qb}
7701 @item @code{c - d} @tab @code{subq.ph}
7704 The table below lists the @code{v2i16} operation for which
7705 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
7706 @code{v2i16} values.
7708 @multitable @columnfractions .50 .50
7709 @item C code @tab MIPS instruction
7710 @item @code{e * f} @tab @code{mul.ph}
7713 It is easier to describe the DSP built-in functions if we first define
7714 the following types:
7719 typedef unsigned int ui32;
7720 typedef long long a64;
7723 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7724 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7725 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7726 @code{long long}, but we use @code{a64} to indicate values that will
7727 be placed in one of the four DSP accumulators (@code{$ac0},
7728 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7730 Also, some built-in functions prefer or require immediate numbers as
7731 parameters, because the corresponding DSP instructions accept both immediate
7732 numbers and register operands, or accept immediate numbers only. The
7733 immediate parameters are listed as follows.
7742 imm_n32_31: -32 to 31.
7743 imm_n512_511: -512 to 511.
7746 The following built-in functions map directly to a particular MIPS DSP
7747 instruction. Please refer to the architecture specification
7748 for details on what each instruction does.
7751 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7752 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7753 q31 __builtin_mips_addq_s_w (q31, q31)
7754 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7755 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7756 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7757 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7758 q31 __builtin_mips_subq_s_w (q31, q31)
7759 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7760 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7761 i32 __builtin_mips_addsc (i32, i32)
7762 i32 __builtin_mips_addwc (i32, i32)
7763 i32 __builtin_mips_modsub (i32, i32)
7764 i32 __builtin_mips_raddu_w_qb (v4i8)
7765 v2q15 __builtin_mips_absq_s_ph (v2q15)
7766 q31 __builtin_mips_absq_s_w (q31)
7767 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7768 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7769 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7770 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7771 q31 __builtin_mips_preceq_w_phl (v2q15)
7772 q31 __builtin_mips_preceq_w_phr (v2q15)
7773 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7774 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7775 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7776 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7777 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7778 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7779 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7780 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7781 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7782 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7783 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7784 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7785 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7786 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7787 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7788 q31 __builtin_mips_shll_s_w (q31, i32)
7789 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7790 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7791 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7792 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7793 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7794 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7795 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7796 q31 __builtin_mips_shra_r_w (q31, i32)
7797 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7798 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7799 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7800 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7801 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7802 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7803 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7804 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7805 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7806 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7807 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7808 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7809 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7810 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7811 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7812 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7813 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7814 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7815 i32 __builtin_mips_bitrev (i32)
7816 i32 __builtin_mips_insv (i32, i32)
7817 v4i8 __builtin_mips_repl_qb (imm0_255)
7818 v4i8 __builtin_mips_repl_qb (i32)
7819 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7820 v2q15 __builtin_mips_repl_ph (i32)
7821 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7822 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7823 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7824 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7825 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7826 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7827 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7828 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7829 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7830 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7831 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7832 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7833 i32 __builtin_mips_extr_w (a64, imm0_31)
7834 i32 __builtin_mips_extr_w (a64, i32)
7835 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7836 i32 __builtin_mips_extr_s_h (a64, i32)
7837 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7838 i32 __builtin_mips_extr_rs_w (a64, i32)
7839 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7840 i32 __builtin_mips_extr_r_w (a64, i32)
7841 i32 __builtin_mips_extp (a64, imm0_31)
7842 i32 __builtin_mips_extp (a64, i32)
7843 i32 __builtin_mips_extpdp (a64, imm0_31)
7844 i32 __builtin_mips_extpdp (a64, i32)
7845 a64 __builtin_mips_shilo (a64, imm_n32_31)
7846 a64 __builtin_mips_shilo (a64, i32)
7847 a64 __builtin_mips_mthlip (a64, i32)
7848 void __builtin_mips_wrdsp (i32, imm0_63)
7849 i32 __builtin_mips_rddsp (imm0_63)
7850 i32 __builtin_mips_lbux (void *, i32)
7851 i32 __builtin_mips_lhx (void *, i32)
7852 i32 __builtin_mips_lwx (void *, i32)
7853 i32 __builtin_mips_bposge32 (void)
7856 The following built-in functions map directly to a particular MIPS DSP REV 2
7857 instruction. Please refer to the architecture specification
7858 for details on what each instruction does.
7861 v4q7 __builtin_mips_absq_s_qb (v4q7);
7862 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
7863 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
7864 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
7865 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
7866 i32 __builtin_mips_append (i32, i32, imm0_31);
7867 i32 __builtin_mips_balign (i32, i32, imm0_3);
7868 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
7869 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
7870 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
7871 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
7872 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
7873 a64 __builtin_mips_madd (a64, i32, i32);
7874 a64 __builtin_mips_maddu (a64, ui32, ui32);
7875 a64 __builtin_mips_msub (a64, i32, i32);
7876 a64 __builtin_mips_msubu (a64, ui32, ui32);
7877 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
7878 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
7879 q31 __builtin_mips_mulq_rs_w (q31, q31);
7880 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
7881 q31 __builtin_mips_mulq_s_w (q31, q31);
7882 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
7883 a64 __builtin_mips_mult (i32, i32);
7884 a64 __builtin_mips_multu (ui32, ui32);
7885 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
7886 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
7887 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
7888 i32 __builtin_mips_prepend (i32, i32, imm0_31);
7889 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
7890 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
7891 v4i8 __builtin_mips_shra_qb (v4i8, i32);
7892 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
7893 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
7894 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
7895 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
7896 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
7897 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
7898 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
7899 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
7900 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
7901 q31 __builtin_mips_addqh_w (q31, q31);
7902 q31 __builtin_mips_addqh_r_w (q31, q31);
7903 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
7904 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
7905 q31 __builtin_mips_subqh_w (q31, q31);
7906 q31 __builtin_mips_subqh_r_w (q31, q31);
7907 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
7908 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
7909 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
7910 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
7911 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
7912 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
7916 @node MIPS Paired-Single Support
7917 @subsection MIPS Paired-Single Support
7919 The MIPS64 architecture includes a number of instructions that
7920 operate on pairs of single-precision floating-point values.
7921 Each pair is packed into a 64-bit floating-point register,
7922 with one element being designated the ``upper half'' and
7923 the other being designated the ``lower half''.
7925 GCC supports paired-single operations using both the generic
7926 vector extensions (@pxref{Vector Extensions}) and a collection of
7927 MIPS-specific built-in functions. Both kinds of support are
7928 enabled by the @option{-mpaired-single} command-line option.
7930 The vector type associated with paired-single values is usually
7931 called @code{v2sf}. It can be defined in C as follows:
7934 typedef float v2sf __attribute__ ((vector_size (8)));
7937 @code{v2sf} values are initialized in the same way as aggregates.
7941 v2sf a = @{1.5, 9.1@};
7944 b = (v2sf) @{e, f@};
7947 @emph{Note:} The CPU's endianness determines which value is stored in
7948 the upper half of a register and which value is stored in the lower half.
7949 On little-endian targets, the first value is the lower one and the second
7950 value is the upper one. The opposite order applies to big-endian targets.
7951 For example, the code above will set the lower half of @code{a} to
7952 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7955 * Paired-Single Arithmetic::
7956 * Paired-Single Built-in Functions::
7957 * MIPS-3D Built-in Functions::
7960 @node Paired-Single Arithmetic
7961 @subsubsection Paired-Single Arithmetic
7963 The table below lists the @code{v2sf} operations for which hardware
7964 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7965 values and @code{x} is an integral value.
7967 @multitable @columnfractions .50 .50
7968 @item C code @tab MIPS instruction
7969 @item @code{a + b} @tab @code{add.ps}
7970 @item @code{a - b} @tab @code{sub.ps}
7971 @item @code{-a} @tab @code{neg.ps}
7972 @item @code{a * b} @tab @code{mul.ps}
7973 @item @code{a * b + c} @tab @code{madd.ps}
7974 @item @code{a * b - c} @tab @code{msub.ps}
7975 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7976 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7977 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7980 Note that the multiply-accumulate instructions can be disabled
7981 using the command-line option @code{-mno-fused-madd}.
7983 @node Paired-Single Built-in Functions
7984 @subsubsection Paired-Single Built-in Functions
7986 The following paired-single functions map directly to a particular
7987 MIPS instruction. Please refer to the architecture specification
7988 for details on what each instruction does.
7991 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7992 Pair lower lower (@code{pll.ps}).
7994 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7995 Pair upper lower (@code{pul.ps}).
7997 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7998 Pair lower upper (@code{plu.ps}).
8000 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
8001 Pair upper upper (@code{puu.ps}).
8003 @item v2sf __builtin_mips_cvt_ps_s (float, float)
8004 Convert pair to paired single (@code{cvt.ps.s}).
8006 @item float __builtin_mips_cvt_s_pl (v2sf)
8007 Convert pair lower to single (@code{cvt.s.pl}).
8009 @item float __builtin_mips_cvt_s_pu (v2sf)
8010 Convert pair upper to single (@code{cvt.s.pu}).
8012 @item v2sf __builtin_mips_abs_ps (v2sf)
8013 Absolute value (@code{abs.ps}).
8015 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
8016 Align variable (@code{alnv.ps}).
8018 @emph{Note:} The value of the third parameter must be 0 or 4
8019 modulo 8, otherwise the result will be unpredictable. Please read the
8020 instruction description for details.
8023 The following multi-instruction functions are also available.
8024 In each case, @var{cond} can be any of the 16 floating-point conditions:
8025 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
8026 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
8027 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
8030 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8031 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8032 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
8033 @code{movt.ps}/@code{movf.ps}).
8035 The @code{movt} functions return the value @var{x} computed by:
8038 c.@var{cond}.ps @var{cc},@var{a},@var{b}
8039 mov.ps @var{x},@var{c}
8040 movt.ps @var{x},@var{d},@var{cc}
8043 The @code{movf} functions are similar but use @code{movf.ps} instead
8046 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8047 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8048 Comparison of two paired-single values (@code{c.@var{cond}.ps},
8049 @code{bc1t}/@code{bc1f}).
8051 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
8052 and return either the upper or lower half of the result. For example:
8056 if (__builtin_mips_upper_c_eq_ps (a, b))
8057 upper_halves_are_equal ();
8059 upper_halves_are_unequal ();
8061 if (__builtin_mips_lower_c_eq_ps (a, b))
8062 lower_halves_are_equal ();
8064 lower_halves_are_unequal ();
8068 @node MIPS-3D Built-in Functions
8069 @subsubsection MIPS-3D Built-in Functions
8071 The MIPS-3D Application-Specific Extension (ASE) includes additional
8072 paired-single instructions that are designed to improve the performance
8073 of 3D graphics operations. Support for these instructions is controlled
8074 by the @option{-mips3d} command-line option.
8076 The functions listed below map directly to a particular MIPS-3D
8077 instruction. Please refer to the architecture specification for
8078 more details on what each instruction does.
8081 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
8082 Reduction add (@code{addr.ps}).
8084 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
8085 Reduction multiply (@code{mulr.ps}).
8087 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
8088 Convert paired single to paired word (@code{cvt.pw.ps}).
8090 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
8091 Convert paired word to paired single (@code{cvt.ps.pw}).
8093 @item float __builtin_mips_recip1_s (float)
8094 @itemx double __builtin_mips_recip1_d (double)
8095 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
8096 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
8098 @item float __builtin_mips_recip2_s (float, float)
8099 @itemx double __builtin_mips_recip2_d (double, double)
8100 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
8101 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
8103 @item float __builtin_mips_rsqrt1_s (float)
8104 @itemx double __builtin_mips_rsqrt1_d (double)
8105 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
8106 Reduced precision reciprocal square root (sequence step 1)
8107 (@code{rsqrt1.@var{fmt}}).
8109 @item float __builtin_mips_rsqrt2_s (float, float)
8110 @itemx double __builtin_mips_rsqrt2_d (double, double)
8111 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
8112 Reduced precision reciprocal square root (sequence step 2)
8113 (@code{rsqrt2.@var{fmt}}).
8116 The following multi-instruction functions are also available.
8117 In each case, @var{cond} can be any of the 16 floating-point conditions:
8118 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
8119 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
8120 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
8123 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
8124 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
8125 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
8126 @code{bc1t}/@code{bc1f}).
8128 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
8129 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
8134 if (__builtin_mips_cabs_eq_s (a, b))
8140 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8141 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8142 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
8143 @code{bc1t}/@code{bc1f}).
8145 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
8146 and return either the upper or lower half of the result. For example:
8150 if (__builtin_mips_upper_cabs_eq_ps (a, b))
8151 upper_halves_are_equal ();
8153 upper_halves_are_unequal ();
8155 if (__builtin_mips_lower_cabs_eq_ps (a, b))
8156 lower_halves_are_equal ();
8158 lower_halves_are_unequal ();
8161 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8162 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8163 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
8164 @code{movt.ps}/@code{movf.ps}).
8166 The @code{movt} functions return the value @var{x} computed by:
8169 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
8170 mov.ps @var{x},@var{c}
8171 movt.ps @var{x},@var{d},@var{cc}
8174 The @code{movf} functions are similar but use @code{movf.ps} instead
8177 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8178 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8179 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8180 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
8181 Comparison of two paired-single values
8182 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
8183 @code{bc1any2t}/@code{bc1any2f}).
8185 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
8186 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
8187 result is true and the @code{all} forms return true if both results are true.
8192 if (__builtin_mips_any_c_eq_ps (a, b))
8197 if (__builtin_mips_all_c_eq_ps (a, b))
8203 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8204 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8205 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8206 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8207 Comparison of four paired-single values
8208 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
8209 @code{bc1any4t}/@code{bc1any4f}).
8211 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
8212 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
8213 The @code{any} forms return true if any of the four results are true
8214 and the @code{all} forms return true if all four results are true.
8219 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
8224 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
8231 @node PowerPC AltiVec Built-in Functions
8232 @subsection PowerPC AltiVec Built-in Functions
8234 GCC provides an interface for the PowerPC family of processors to access
8235 the AltiVec operations described in Motorola's AltiVec Programming
8236 Interface Manual. The interface is made available by including
8237 @code{<altivec.h>} and using @option{-maltivec} and
8238 @option{-mabi=altivec}. The interface supports the following vector
8242 vector unsigned char
8246 vector unsigned short
8257 GCC's implementation of the high-level language interface available from
8258 C and C++ code differs from Motorola's documentation in several ways.
8263 A vector constant is a list of constant expressions within curly braces.
8266 A vector initializer requires no cast if the vector constant is of the
8267 same type as the variable it is initializing.
8270 If @code{signed} or @code{unsigned} is omitted, the signedness of the
8271 vector type is the default signedness of the base type. The default
8272 varies depending on the operating system, so a portable program should
8273 always specify the signedness.
8276 Compiling with @option{-maltivec} adds keywords @code{__vector},
8277 @code{__pixel}, and @code{__bool}. Macros @option{vector},
8278 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
8282 GCC allows using a @code{typedef} name as the type specifier for a
8286 For C, overloaded functions are implemented with macros so the following
8290 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
8293 Since @code{vec_add} is a macro, the vector constant in the example
8294 is treated as four separate arguments. Wrap the entire argument in
8295 parentheses for this to work.
8298 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
8299 Internally, GCC uses built-in functions to achieve the functionality in
8300 the aforementioned header file, but they are not supported and are
8301 subject to change without notice.
8303 The following interfaces are supported for the generic and specific
8304 AltiVec operations and the AltiVec predicates. In cases where there
8305 is a direct mapping between generic and specific operations, only the
8306 generic names are shown here, although the specific operations can also
8309 Arguments that are documented as @code{const int} require literal
8310 integral values within the range required for that operation.
8313 vector signed char vec_abs (vector signed char);
8314 vector signed short vec_abs (vector signed short);
8315 vector signed int vec_abs (vector signed int);
8316 vector float vec_abs (vector float);
8318 vector signed char vec_abss (vector signed char);
8319 vector signed short vec_abss (vector signed short);
8320 vector signed int vec_abss (vector signed int);
8322 vector signed char vec_add (vector bool char, vector signed char);
8323 vector signed char vec_add (vector signed char, vector bool char);
8324 vector signed char vec_add (vector signed char, vector signed char);
8325 vector unsigned char vec_add (vector bool char, vector unsigned char);
8326 vector unsigned char vec_add (vector unsigned char, vector bool char);
8327 vector unsigned char vec_add (vector unsigned char,
8328 vector unsigned char);
8329 vector signed short vec_add (vector bool short, vector signed short);
8330 vector signed short vec_add (vector signed short, vector bool short);
8331 vector signed short vec_add (vector signed short, vector signed short);
8332 vector unsigned short vec_add (vector bool short,
8333 vector unsigned short);
8334 vector unsigned short vec_add (vector unsigned short,
8336 vector unsigned short vec_add (vector unsigned short,
8337 vector unsigned short);
8338 vector signed int vec_add (vector bool int, vector signed int);
8339 vector signed int vec_add (vector signed int, vector bool int);
8340 vector signed int vec_add (vector signed int, vector signed int);
8341 vector unsigned int vec_add (vector bool int, vector unsigned int);
8342 vector unsigned int vec_add (vector unsigned int, vector bool int);
8343 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
8344 vector float vec_add (vector float, vector float);
8346 vector float vec_vaddfp (vector float, vector float);
8348 vector signed int vec_vadduwm (vector bool int, vector signed int);
8349 vector signed int vec_vadduwm (vector signed int, vector bool int);
8350 vector signed int vec_vadduwm (vector signed int, vector signed int);
8351 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
8352 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
8353 vector unsigned int vec_vadduwm (vector unsigned int,
8354 vector unsigned int);
8356 vector signed short vec_vadduhm (vector bool short,
8357 vector signed short);
8358 vector signed short vec_vadduhm (vector signed short,
8360 vector signed short vec_vadduhm (vector signed short,
8361 vector signed short);
8362 vector unsigned short vec_vadduhm (vector bool short,
8363 vector unsigned short);
8364 vector unsigned short vec_vadduhm (vector unsigned short,
8366 vector unsigned short vec_vadduhm (vector unsigned short,
8367 vector unsigned short);
8369 vector signed char vec_vaddubm (vector bool char, vector signed char);
8370 vector signed char vec_vaddubm (vector signed char, vector bool char);
8371 vector signed char vec_vaddubm (vector signed char, vector signed char);
8372 vector unsigned char vec_vaddubm (vector bool char,
8373 vector unsigned char);
8374 vector unsigned char vec_vaddubm (vector unsigned char,
8376 vector unsigned char vec_vaddubm (vector unsigned char,
8377 vector unsigned char);
8379 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
8381 vector unsigned char vec_adds (vector bool char, vector unsigned char);
8382 vector unsigned char vec_adds (vector unsigned char, vector bool char);
8383 vector unsigned char vec_adds (vector unsigned char,
8384 vector unsigned char);
8385 vector signed char vec_adds (vector bool char, vector signed char);
8386 vector signed char vec_adds (vector signed char, vector bool char);
8387 vector signed char vec_adds (vector signed char, vector signed char);
8388 vector unsigned short vec_adds (vector bool short,
8389 vector unsigned short);
8390 vector unsigned short vec_adds (vector unsigned short,
8392 vector unsigned short vec_adds (vector unsigned short,
8393 vector unsigned short);
8394 vector signed short vec_adds (vector bool short, vector signed short);
8395 vector signed short vec_adds (vector signed short, vector bool short);
8396 vector signed short vec_adds (vector signed short, vector signed short);
8397 vector unsigned int vec_adds (vector bool int, vector unsigned int);
8398 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8399 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8400 vector signed int vec_adds (vector bool int, vector signed int);
8401 vector signed int vec_adds (vector signed int, vector bool int);
8402 vector signed int vec_adds (vector signed int, vector signed int);
8404 vector signed int vec_vaddsws (vector bool int, vector signed int);
8405 vector signed int vec_vaddsws (vector signed int, vector bool int);
8406 vector signed int vec_vaddsws (vector signed int, vector signed int);
8408 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8409 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8410 vector unsigned int vec_vadduws (vector unsigned int,
8411 vector unsigned int);
8413 vector signed short vec_vaddshs (vector bool short,
8414 vector signed short);
8415 vector signed short vec_vaddshs (vector signed short,
8417 vector signed short vec_vaddshs (vector signed short,
8418 vector signed short);
8420 vector unsigned short vec_vadduhs (vector bool short,
8421 vector unsigned short);
8422 vector unsigned short vec_vadduhs (vector unsigned short,
8424 vector unsigned short vec_vadduhs (vector unsigned short,
8425 vector unsigned short);
8427 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8428 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8429 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8431 vector unsigned char vec_vaddubs (vector bool char,
8432 vector unsigned char);
8433 vector unsigned char vec_vaddubs (vector unsigned char,
8435 vector unsigned char vec_vaddubs (vector unsigned char,
8436 vector unsigned char);
8438 vector float vec_and (vector float, vector float);
8439 vector float vec_and (vector float, vector bool int);
8440 vector float vec_and (vector bool int, vector float);
8441 vector bool int vec_and (vector bool int, vector bool int);
8442 vector signed int vec_and (vector bool int, vector signed int);
8443 vector signed int vec_and (vector signed int, vector bool int);
8444 vector signed int vec_and (vector signed int, vector signed int);
8445 vector unsigned int vec_and (vector bool int, vector unsigned int);
8446 vector unsigned int vec_and (vector unsigned int, vector bool int);
8447 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8448 vector bool short vec_and (vector bool short, vector bool short);
8449 vector signed short vec_and (vector bool short, vector signed short);
8450 vector signed short vec_and (vector signed short, vector bool short);
8451 vector signed short vec_and (vector signed short, vector signed short);
8452 vector unsigned short vec_and (vector bool short,
8453 vector unsigned short);
8454 vector unsigned short vec_and (vector unsigned short,
8456 vector unsigned short vec_and (vector unsigned short,
8457 vector unsigned short);
8458 vector signed char vec_and (vector bool char, vector signed char);
8459 vector bool char vec_and (vector bool char, vector bool char);
8460 vector signed char vec_and (vector signed char, vector bool char);
8461 vector signed char vec_and (vector signed char, vector signed char);
8462 vector unsigned char vec_and (vector bool char, vector unsigned char);
8463 vector unsigned char vec_and (vector unsigned char, vector bool char);
8464 vector unsigned char vec_and (vector unsigned char,
8465 vector unsigned char);
8467 vector float vec_andc (vector float, vector float);
8468 vector float vec_andc (vector float, vector bool int);
8469 vector float vec_andc (vector bool int, vector float);
8470 vector bool int vec_andc (vector bool int, vector bool int);
8471 vector signed int vec_andc (vector bool int, vector signed int);
8472 vector signed int vec_andc (vector signed int, vector bool int);
8473 vector signed int vec_andc (vector signed int, vector signed int);
8474 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8475 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8476 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8477 vector bool short vec_andc (vector bool short, vector bool short);
8478 vector signed short vec_andc (vector bool short, vector signed short);
8479 vector signed short vec_andc (vector signed short, vector bool short);
8480 vector signed short vec_andc (vector signed short, vector signed short);
8481 vector unsigned short vec_andc (vector bool short,
8482 vector unsigned short);
8483 vector unsigned short vec_andc (vector unsigned short,
8485 vector unsigned short vec_andc (vector unsigned short,
8486 vector unsigned short);
8487 vector signed char vec_andc (vector bool char, vector signed char);
8488 vector bool char vec_andc (vector bool char, vector bool char);
8489 vector signed char vec_andc (vector signed char, vector bool char);
8490 vector signed char vec_andc (vector signed char, vector signed char);
8491 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8492 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8493 vector unsigned char vec_andc (vector unsigned char,
8494 vector unsigned char);
8496 vector unsigned char vec_avg (vector unsigned char,
8497 vector unsigned char);
8498 vector signed char vec_avg (vector signed char, vector signed char);
8499 vector unsigned short vec_avg (vector unsigned short,
8500 vector unsigned short);
8501 vector signed short vec_avg (vector signed short, vector signed short);
8502 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8503 vector signed int vec_avg (vector signed int, vector signed int);
8505 vector signed int vec_vavgsw (vector signed int, vector signed int);
8507 vector unsigned int vec_vavguw (vector unsigned int,
8508 vector unsigned int);
8510 vector signed short vec_vavgsh (vector signed short,
8511 vector signed short);
8513 vector unsigned short vec_vavguh (vector unsigned short,
8514 vector unsigned short);
8516 vector signed char vec_vavgsb (vector signed char, vector signed char);
8518 vector unsigned char vec_vavgub (vector unsigned char,
8519 vector unsigned char);
8521 vector float vec_ceil (vector float);
8523 vector signed int vec_cmpb (vector float, vector float);
8525 vector bool char vec_cmpeq (vector signed char, vector signed char);
8526 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8527 vector bool short vec_cmpeq (vector signed short, vector signed short);
8528 vector bool short vec_cmpeq (vector unsigned short,
8529 vector unsigned short);
8530 vector bool int vec_cmpeq (vector signed int, vector signed int);
8531 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8532 vector bool int vec_cmpeq (vector float, vector float);
8534 vector bool int vec_vcmpeqfp (vector float, vector float);
8536 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8537 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8539 vector bool short vec_vcmpequh (vector signed short,
8540 vector signed short);
8541 vector bool short vec_vcmpequh (vector unsigned short,
8542 vector unsigned short);
8544 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8545 vector bool char vec_vcmpequb (vector unsigned char,
8546 vector unsigned char);
8548 vector bool int vec_cmpge (vector float, vector float);
8550 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8551 vector bool char vec_cmpgt (vector signed char, vector signed char);
8552 vector bool short vec_cmpgt (vector unsigned short,
8553 vector unsigned short);
8554 vector bool short vec_cmpgt (vector signed short, vector signed short);
8555 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8556 vector bool int vec_cmpgt (vector signed int, vector signed int);
8557 vector bool int vec_cmpgt (vector float, vector float);
8559 vector bool int vec_vcmpgtfp (vector float, vector float);
8561 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8563 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8565 vector bool short vec_vcmpgtsh (vector signed short,
8566 vector signed short);
8568 vector bool short vec_vcmpgtuh (vector unsigned short,
8569 vector unsigned short);
8571 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8573 vector bool char vec_vcmpgtub (vector unsigned char,
8574 vector unsigned char);
8576 vector bool int vec_cmple (vector float, vector float);
8578 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8579 vector bool char vec_cmplt (vector signed char, vector signed char);
8580 vector bool short vec_cmplt (vector unsigned short,
8581 vector unsigned short);
8582 vector bool short vec_cmplt (vector signed short, vector signed short);
8583 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8584 vector bool int vec_cmplt (vector signed int, vector signed int);
8585 vector bool int vec_cmplt (vector float, vector float);
8587 vector float vec_ctf (vector unsigned int, const int);
8588 vector float vec_ctf (vector signed int, const int);
8590 vector float vec_vcfsx (vector signed int, const int);
8592 vector float vec_vcfux (vector unsigned int, const int);
8594 vector signed int vec_cts (vector float, const int);
8596 vector unsigned int vec_ctu (vector float, const int);
8598 void vec_dss (const int);
8600 void vec_dssall (void);
8602 void vec_dst (const vector unsigned char *, int, const int);
8603 void vec_dst (const vector signed char *, int, const int);
8604 void vec_dst (const vector bool char *, int, const int);
8605 void vec_dst (const vector unsigned short *, int, const int);
8606 void vec_dst (const vector signed short *, int, const int);
8607 void vec_dst (const vector bool short *, int, const int);
8608 void vec_dst (const vector pixel *, int, const int);
8609 void vec_dst (const vector unsigned int *, int, const int);
8610 void vec_dst (const vector signed int *, int, const int);
8611 void vec_dst (const vector bool int *, int, const int);
8612 void vec_dst (const vector float *, int, const int);
8613 void vec_dst (const unsigned char *, int, const int);
8614 void vec_dst (const signed char *, int, const int);
8615 void vec_dst (const unsigned short *, int, const int);
8616 void vec_dst (const short *, int, const int);
8617 void vec_dst (const unsigned int *, int, const int);
8618 void vec_dst (const int *, int, const int);
8619 void vec_dst (const unsigned long *, int, const int);
8620 void vec_dst (const long *, int, const int);
8621 void vec_dst (const float *, int, const int);
8623 void vec_dstst (const vector unsigned char *, int, const int);
8624 void vec_dstst (const vector signed char *, int, const int);
8625 void vec_dstst (const vector bool char *, int, const int);
8626 void vec_dstst (const vector unsigned short *, int, const int);
8627 void vec_dstst (const vector signed short *, int, const int);
8628 void vec_dstst (const vector bool short *, int, const int);
8629 void vec_dstst (const vector pixel *, int, const int);
8630 void vec_dstst (const vector unsigned int *, int, const int);
8631 void vec_dstst (const vector signed int *, int, const int);
8632 void vec_dstst (const vector bool int *, int, const int);
8633 void vec_dstst (const vector float *, int, const int);
8634 void vec_dstst (const unsigned char *, int, const int);
8635 void vec_dstst (const signed char *, int, const int);
8636 void vec_dstst (const unsigned short *, int, const int);
8637 void vec_dstst (const short *, int, const int);
8638 void vec_dstst (const unsigned int *, int, const int);
8639 void vec_dstst (const int *, int, const int);
8640 void vec_dstst (const unsigned long *, int, const int);
8641 void vec_dstst (const long *, int, const int);
8642 void vec_dstst (const float *, int, const int);
8644 void vec_dststt (const vector unsigned char *, int, const int);
8645 void vec_dststt (const vector signed char *, int, const int);
8646 void vec_dststt (const vector bool char *, int, const int);
8647 void vec_dststt (const vector unsigned short *, int, const int);
8648 void vec_dststt (const vector signed short *, int, const int);
8649 void vec_dststt (const vector bool short *, int, const int);
8650 void vec_dststt (const vector pixel *, int, const int);
8651 void vec_dststt (const vector unsigned int *, int, const int);
8652 void vec_dststt (const vector signed int *, int, const int);
8653 void vec_dststt (const vector bool int *, int, const int);
8654 void vec_dststt (const vector float *, int, const int);
8655 void vec_dststt (const unsigned char *, int, const int);
8656 void vec_dststt (const signed char *, int, const int);
8657 void vec_dststt (const unsigned short *, int, const int);
8658 void vec_dststt (const short *, int, const int);
8659 void vec_dststt (const unsigned int *, int, const int);
8660 void vec_dststt (const int *, int, const int);
8661 void vec_dststt (const unsigned long *, int, const int);
8662 void vec_dststt (const long *, int, const int);
8663 void vec_dststt (const float *, int, const int);
8665 void vec_dstt (const vector unsigned char *, int, const int);
8666 void vec_dstt (const vector signed char *, int, const int);
8667 void vec_dstt (const vector bool char *, int, const int);
8668 void vec_dstt (const vector unsigned short *, int, const int);
8669 void vec_dstt (const vector signed short *, int, const int);
8670 void vec_dstt (const vector bool short *, int, const int);
8671 void vec_dstt (const vector pixel *, int, const int);
8672 void vec_dstt (const vector unsigned int *, int, const int);
8673 void vec_dstt (const vector signed int *, int, const int);
8674 void vec_dstt (const vector bool int *, int, const int);
8675 void vec_dstt (const vector float *, int, const int);
8676 void vec_dstt (const unsigned char *, int, const int);
8677 void vec_dstt (const signed char *, int, const int);
8678 void vec_dstt (const unsigned short *, int, const int);
8679 void vec_dstt (const short *, int, const int);
8680 void vec_dstt (const unsigned int *, int, const int);
8681 void vec_dstt (const int *, int, const int);
8682 void vec_dstt (const unsigned long *, int, const int);
8683 void vec_dstt (const long *, int, const int);
8684 void vec_dstt (const float *, int, const int);
8686 vector float vec_expte (vector float);
8688 vector float vec_floor (vector float);
8690 vector float vec_ld (int, const vector float *);
8691 vector float vec_ld (int, const float *);
8692 vector bool int vec_ld (int, const vector bool int *);
8693 vector signed int vec_ld (int, const vector signed int *);
8694 vector signed int vec_ld (int, const int *);
8695 vector signed int vec_ld (int, const long *);
8696 vector unsigned int vec_ld (int, const vector unsigned int *);
8697 vector unsigned int vec_ld (int, const unsigned int *);
8698 vector unsigned int vec_ld (int, const unsigned long *);
8699 vector bool short vec_ld (int, const vector bool short *);
8700 vector pixel vec_ld (int, const vector pixel *);
8701 vector signed short vec_ld (int, const vector signed short *);
8702 vector signed short vec_ld (int, const short *);
8703 vector unsigned short vec_ld (int, const vector unsigned short *);
8704 vector unsigned short vec_ld (int, const unsigned short *);
8705 vector bool char vec_ld (int, const vector bool char *);
8706 vector signed char vec_ld (int, const vector signed char *);
8707 vector signed char vec_ld (int, const signed char *);
8708 vector unsigned char vec_ld (int, const vector unsigned char *);
8709 vector unsigned char vec_ld (int, const unsigned char *);
8711 vector signed char vec_lde (int, const signed char *);
8712 vector unsigned char vec_lde (int, const unsigned char *);
8713 vector signed short vec_lde (int, const short *);
8714 vector unsigned short vec_lde (int, const unsigned short *);
8715 vector float vec_lde (int, const float *);
8716 vector signed int vec_lde (int, const int *);
8717 vector unsigned int vec_lde (int, const unsigned int *);
8718 vector signed int vec_lde (int, const long *);
8719 vector unsigned int vec_lde (int, const unsigned long *);
8721 vector float vec_lvewx (int, float *);
8722 vector signed int vec_lvewx (int, int *);
8723 vector unsigned int vec_lvewx (int, unsigned int *);
8724 vector signed int vec_lvewx (int, long *);
8725 vector unsigned int vec_lvewx (int, unsigned long *);
8727 vector signed short vec_lvehx (int, short *);
8728 vector unsigned short vec_lvehx (int, unsigned short *);
8730 vector signed char vec_lvebx (int, char *);
8731 vector unsigned char vec_lvebx (int, unsigned char *);
8733 vector float vec_ldl (int, const vector float *);
8734 vector float vec_ldl (int, const float *);
8735 vector bool int vec_ldl (int, const vector bool int *);
8736 vector signed int vec_ldl (int, const vector signed int *);
8737 vector signed int vec_ldl (int, const int *);
8738 vector signed int vec_ldl (int, const long *);
8739 vector unsigned int vec_ldl (int, const vector unsigned int *);
8740 vector unsigned int vec_ldl (int, const unsigned int *);
8741 vector unsigned int vec_ldl (int, const unsigned long *);
8742 vector bool short vec_ldl (int, const vector bool short *);
8743 vector pixel vec_ldl (int, const vector pixel *);
8744 vector signed short vec_ldl (int, const vector signed short *);
8745 vector signed short vec_ldl (int, const short *);
8746 vector unsigned short vec_ldl (int, const vector unsigned short *);
8747 vector unsigned short vec_ldl (int, const unsigned short *);
8748 vector bool char vec_ldl (int, const vector bool char *);
8749 vector signed char vec_ldl (int, const vector signed char *);
8750 vector signed char vec_ldl (int, const signed char *);
8751 vector unsigned char vec_ldl (int, const vector unsigned char *);
8752 vector unsigned char vec_ldl (int, const unsigned char *);
8754 vector float vec_loge (vector float);
8756 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8757 vector unsigned char vec_lvsl (int, const volatile signed char *);
8758 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8759 vector unsigned char vec_lvsl (int, const volatile short *);
8760 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8761 vector unsigned char vec_lvsl (int, const volatile int *);
8762 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8763 vector unsigned char vec_lvsl (int, const volatile long *);
8764 vector unsigned char vec_lvsl (int, const volatile float *);
8766 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8767 vector unsigned char vec_lvsr (int, const volatile signed char *);
8768 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8769 vector unsigned char vec_lvsr (int, const volatile short *);
8770 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8771 vector unsigned char vec_lvsr (int, const volatile int *);
8772 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8773 vector unsigned char vec_lvsr (int, const volatile long *);
8774 vector unsigned char vec_lvsr (int, const volatile float *);
8776 vector float vec_madd (vector float, vector float, vector float);
8778 vector signed short vec_madds (vector signed short,
8779 vector signed short,
8780 vector signed short);
8782 vector unsigned char vec_max (vector bool char, vector unsigned char);
8783 vector unsigned char vec_max (vector unsigned char, vector bool char);
8784 vector unsigned char vec_max (vector unsigned char,
8785 vector unsigned char);
8786 vector signed char vec_max (vector bool char, vector signed char);
8787 vector signed char vec_max (vector signed char, vector bool char);
8788 vector signed char vec_max (vector signed char, vector signed char);
8789 vector unsigned short vec_max (vector bool short,
8790 vector unsigned short);
8791 vector unsigned short vec_max (vector unsigned short,
8793 vector unsigned short vec_max (vector unsigned short,
8794 vector unsigned short);
8795 vector signed short vec_max (vector bool short, vector signed short);
8796 vector signed short vec_max (vector signed short, vector bool short);
8797 vector signed short vec_max (vector signed short, vector signed short);
8798 vector unsigned int vec_max (vector bool int, vector unsigned int);
8799 vector unsigned int vec_max (vector unsigned int, vector bool int);
8800 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8801 vector signed int vec_max (vector bool int, vector signed int);
8802 vector signed int vec_max (vector signed int, vector bool int);
8803 vector signed int vec_max (vector signed int, vector signed int);
8804 vector float vec_max (vector float, vector float);
8806 vector float vec_vmaxfp (vector float, vector float);
8808 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8809 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8810 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8812 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8813 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8814 vector unsigned int vec_vmaxuw (vector unsigned int,
8815 vector unsigned int);
8817 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8818 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8819 vector signed short vec_vmaxsh (vector signed short,
8820 vector signed short);
8822 vector unsigned short vec_vmaxuh (vector bool short,
8823 vector unsigned short);
8824 vector unsigned short vec_vmaxuh (vector unsigned short,
8826 vector unsigned short vec_vmaxuh (vector unsigned short,
8827 vector unsigned short);
8829 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8830 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8831 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8833 vector unsigned char vec_vmaxub (vector bool char,
8834 vector unsigned char);
8835 vector unsigned char vec_vmaxub (vector unsigned char,
8837 vector unsigned char vec_vmaxub (vector unsigned char,
8838 vector unsigned char);
8840 vector bool char vec_mergeh (vector bool char, vector bool char);
8841 vector signed char vec_mergeh (vector signed char, vector signed char);
8842 vector unsigned char vec_mergeh (vector unsigned char,
8843 vector unsigned char);
8844 vector bool short vec_mergeh (vector bool short, vector bool short);
8845 vector pixel vec_mergeh (vector pixel, vector pixel);
8846 vector signed short vec_mergeh (vector signed short,
8847 vector signed short);
8848 vector unsigned short vec_mergeh (vector unsigned short,
8849 vector unsigned short);
8850 vector float vec_mergeh (vector float, vector float);
8851 vector bool int vec_mergeh (vector bool int, vector bool int);
8852 vector signed int vec_mergeh (vector signed int, vector signed int);
8853 vector unsigned int vec_mergeh (vector unsigned int,
8854 vector unsigned int);
8856 vector float vec_vmrghw (vector float, vector float);
8857 vector bool int vec_vmrghw (vector bool int, vector bool int);
8858 vector signed int vec_vmrghw (vector signed int, vector signed int);
8859 vector unsigned int vec_vmrghw (vector unsigned int,
8860 vector unsigned int);
8862 vector bool short vec_vmrghh (vector bool short, vector bool short);
8863 vector signed short vec_vmrghh (vector signed short,
8864 vector signed short);
8865 vector unsigned short vec_vmrghh (vector unsigned short,
8866 vector unsigned short);
8867 vector pixel vec_vmrghh (vector pixel, vector pixel);
8869 vector bool char vec_vmrghb (vector bool char, vector bool char);
8870 vector signed char vec_vmrghb (vector signed char, vector signed char);
8871 vector unsigned char vec_vmrghb (vector unsigned char,
8872 vector unsigned char);
8874 vector bool char vec_mergel (vector bool char, vector bool char);
8875 vector signed char vec_mergel (vector signed char, vector signed char);
8876 vector unsigned char vec_mergel (vector unsigned char,
8877 vector unsigned char);
8878 vector bool short vec_mergel (vector bool short, vector bool short);
8879 vector pixel vec_mergel (vector pixel, vector pixel);
8880 vector signed short vec_mergel (vector signed short,
8881 vector signed short);
8882 vector unsigned short vec_mergel (vector unsigned short,
8883 vector unsigned short);
8884 vector float vec_mergel (vector float, vector float);
8885 vector bool int vec_mergel (vector bool int, vector bool int);
8886 vector signed int vec_mergel (vector signed int, vector signed int);
8887 vector unsigned int vec_mergel (vector unsigned int,
8888 vector unsigned int);
8890 vector float vec_vmrglw (vector float, vector float);
8891 vector signed int vec_vmrglw (vector signed int, vector signed int);
8892 vector unsigned int vec_vmrglw (vector unsigned int,
8893 vector unsigned int);
8894 vector bool int vec_vmrglw (vector bool int, vector bool int);
8896 vector bool short vec_vmrglh (vector bool short, vector bool short);
8897 vector signed short vec_vmrglh (vector signed short,
8898 vector signed short);
8899 vector unsigned short vec_vmrglh (vector unsigned short,
8900 vector unsigned short);
8901 vector pixel vec_vmrglh (vector pixel, vector pixel);
8903 vector bool char vec_vmrglb (vector bool char, vector bool char);
8904 vector signed char vec_vmrglb (vector signed char, vector signed char);
8905 vector unsigned char vec_vmrglb (vector unsigned char,
8906 vector unsigned char);
8908 vector unsigned short vec_mfvscr (void);
8910 vector unsigned char vec_min (vector bool char, vector unsigned char);
8911 vector unsigned char vec_min (vector unsigned char, vector bool char);
8912 vector unsigned char vec_min (vector unsigned char,
8913 vector unsigned char);
8914 vector signed char vec_min (vector bool char, vector signed char);
8915 vector signed char vec_min (vector signed char, vector bool char);
8916 vector signed char vec_min (vector signed char, vector signed char);
8917 vector unsigned short vec_min (vector bool short,
8918 vector unsigned short);
8919 vector unsigned short vec_min (vector unsigned short,
8921 vector unsigned short vec_min (vector unsigned short,
8922 vector unsigned short);
8923 vector signed short vec_min (vector bool short, vector signed short);
8924 vector signed short vec_min (vector signed short, vector bool short);
8925 vector signed short vec_min (vector signed short, vector signed short);
8926 vector unsigned int vec_min (vector bool int, vector unsigned int);
8927 vector unsigned int vec_min (vector unsigned int, vector bool int);
8928 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8929 vector signed int vec_min (vector bool int, vector signed int);
8930 vector signed int vec_min (vector signed int, vector bool int);
8931 vector signed int vec_min (vector signed int, vector signed int);
8932 vector float vec_min (vector float, vector float);
8934 vector float vec_vminfp (vector float, vector float);
8936 vector signed int vec_vminsw (vector bool int, vector signed int);
8937 vector signed int vec_vminsw (vector signed int, vector bool int);
8938 vector signed int vec_vminsw (vector signed int, vector signed int);
8940 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8941 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8942 vector unsigned int vec_vminuw (vector unsigned int,
8943 vector unsigned int);
8945 vector signed short vec_vminsh (vector bool short, vector signed short);
8946 vector signed short vec_vminsh (vector signed short, vector bool short);
8947 vector signed short vec_vminsh (vector signed short,
8948 vector signed short);
8950 vector unsigned short vec_vminuh (vector bool short,
8951 vector unsigned short);
8952 vector unsigned short vec_vminuh (vector unsigned short,
8954 vector unsigned short vec_vminuh (vector unsigned short,
8955 vector unsigned short);
8957 vector signed char vec_vminsb (vector bool char, vector signed char);
8958 vector signed char vec_vminsb (vector signed char, vector bool char);
8959 vector signed char vec_vminsb (vector signed char, vector signed char);
8961 vector unsigned char vec_vminub (vector bool char,
8962 vector unsigned char);
8963 vector unsigned char vec_vminub (vector unsigned char,
8965 vector unsigned char vec_vminub (vector unsigned char,
8966 vector unsigned char);
8968 vector signed short vec_mladd (vector signed short,
8969 vector signed short,
8970 vector signed short);
8971 vector signed short vec_mladd (vector signed short,
8972 vector unsigned short,
8973 vector unsigned short);
8974 vector signed short vec_mladd (vector unsigned short,
8975 vector signed short,
8976 vector signed short);
8977 vector unsigned short vec_mladd (vector unsigned short,
8978 vector unsigned short,
8979 vector unsigned short);
8981 vector signed short vec_mradds (vector signed short,
8982 vector signed short,
8983 vector signed short);
8985 vector unsigned int vec_msum (vector unsigned char,
8986 vector unsigned char,
8987 vector unsigned int);
8988 vector signed int vec_msum (vector signed char,
8989 vector unsigned char,
8991 vector unsigned int vec_msum (vector unsigned short,
8992 vector unsigned short,
8993 vector unsigned int);
8994 vector signed int vec_msum (vector signed short,
8995 vector signed short,
8998 vector signed int vec_vmsumshm (vector signed short,
8999 vector signed short,
9002 vector unsigned int vec_vmsumuhm (vector unsigned short,
9003 vector unsigned short,
9004 vector unsigned int);
9006 vector signed int vec_vmsummbm (vector signed char,
9007 vector unsigned char,
9010 vector unsigned int vec_vmsumubm (vector unsigned char,
9011 vector unsigned char,
9012 vector unsigned int);
9014 vector unsigned int vec_msums (vector unsigned short,
9015 vector unsigned short,
9016 vector unsigned int);
9017 vector signed int vec_msums (vector signed short,
9018 vector signed short,
9021 vector signed int vec_vmsumshs (vector signed short,
9022 vector signed short,
9025 vector unsigned int vec_vmsumuhs (vector unsigned short,
9026 vector unsigned short,
9027 vector unsigned int);
9029 void vec_mtvscr (vector signed int);
9030 void vec_mtvscr (vector unsigned int);
9031 void vec_mtvscr (vector bool int);
9032 void vec_mtvscr (vector signed short);
9033 void vec_mtvscr (vector unsigned short);
9034 void vec_mtvscr (vector bool short);
9035 void vec_mtvscr (vector pixel);
9036 void vec_mtvscr (vector signed char);
9037 void vec_mtvscr (vector unsigned char);
9038 void vec_mtvscr (vector bool char);
9040 vector unsigned short vec_mule (vector unsigned char,
9041 vector unsigned char);
9042 vector signed short vec_mule (vector signed char,
9043 vector signed char);
9044 vector unsigned int vec_mule (vector unsigned short,
9045 vector unsigned short);
9046 vector signed int vec_mule (vector signed short, vector signed short);
9048 vector signed int vec_vmulesh (vector signed short,
9049 vector signed short);
9051 vector unsigned int vec_vmuleuh (vector unsigned short,
9052 vector unsigned short);
9054 vector signed short vec_vmulesb (vector signed char,
9055 vector signed char);
9057 vector unsigned short vec_vmuleub (vector unsigned char,
9058 vector unsigned char);
9060 vector unsigned short vec_mulo (vector unsigned char,
9061 vector unsigned char);
9062 vector signed short vec_mulo (vector signed char, vector signed char);
9063 vector unsigned int vec_mulo (vector unsigned short,
9064 vector unsigned short);
9065 vector signed int vec_mulo (vector signed short, vector signed short);
9067 vector signed int vec_vmulosh (vector signed short,
9068 vector signed short);
9070 vector unsigned int vec_vmulouh (vector unsigned short,
9071 vector unsigned short);
9073 vector signed short vec_vmulosb (vector signed char,
9074 vector signed char);
9076 vector unsigned short vec_vmuloub (vector unsigned char,
9077 vector unsigned char);
9079 vector float vec_nmsub (vector float, vector float, vector float);
9081 vector float vec_nor (vector float, vector float);
9082 vector signed int vec_nor (vector signed int, vector signed int);
9083 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
9084 vector bool int vec_nor (vector bool int, vector bool int);
9085 vector signed short vec_nor (vector signed short, vector signed short);
9086 vector unsigned short vec_nor (vector unsigned short,
9087 vector unsigned short);
9088 vector bool short vec_nor (vector bool short, vector bool short);
9089 vector signed char vec_nor (vector signed char, vector signed char);
9090 vector unsigned char vec_nor (vector unsigned char,
9091 vector unsigned char);
9092 vector bool char vec_nor (vector bool char, vector bool char);
9094 vector float vec_or (vector float, vector float);
9095 vector float vec_or (vector float, vector bool int);
9096 vector float vec_or (vector bool int, vector float);
9097 vector bool int vec_or (vector bool int, vector bool int);
9098 vector signed int vec_or (vector bool int, vector signed int);
9099 vector signed int vec_or (vector signed int, vector bool int);
9100 vector signed int vec_or (vector signed int, vector signed int);
9101 vector unsigned int vec_or (vector bool int, vector unsigned int);
9102 vector unsigned int vec_or (vector unsigned int, vector bool int);
9103 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
9104 vector bool short vec_or (vector bool short, vector bool short);
9105 vector signed short vec_or (vector bool short, vector signed short);
9106 vector signed short vec_or (vector signed short, vector bool short);
9107 vector signed short vec_or (vector signed short, vector signed short);
9108 vector unsigned short vec_or (vector bool short, vector unsigned short);
9109 vector unsigned short vec_or (vector unsigned short, vector bool short);
9110 vector unsigned short vec_or (vector unsigned short,
9111 vector unsigned short);
9112 vector signed char vec_or (vector bool char, vector signed char);
9113 vector bool char vec_or (vector bool char, vector bool char);
9114 vector signed char vec_or (vector signed char, vector bool char);
9115 vector signed char vec_or (vector signed char, vector signed char);
9116 vector unsigned char vec_or (vector bool char, vector unsigned char);
9117 vector unsigned char vec_or (vector unsigned char, vector bool char);
9118 vector unsigned char vec_or (vector unsigned char,
9119 vector unsigned char);
9121 vector signed char vec_pack (vector signed short, vector signed short);
9122 vector unsigned char vec_pack (vector unsigned short,
9123 vector unsigned short);
9124 vector bool char vec_pack (vector bool short, vector bool short);
9125 vector signed short vec_pack (vector signed int, vector signed int);
9126 vector unsigned short vec_pack (vector unsigned int,
9127 vector unsigned int);
9128 vector bool short vec_pack (vector bool int, vector bool int);
9130 vector bool short vec_vpkuwum (vector bool int, vector bool int);
9131 vector signed short vec_vpkuwum (vector signed int, vector signed int);
9132 vector unsigned short vec_vpkuwum (vector unsigned int,
9133 vector unsigned int);
9135 vector bool char vec_vpkuhum (vector bool short, vector bool short);
9136 vector signed char vec_vpkuhum (vector signed short,
9137 vector signed short);
9138 vector unsigned char vec_vpkuhum (vector unsigned short,
9139 vector unsigned short);
9141 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
9143 vector unsigned char vec_packs (vector unsigned short,
9144 vector unsigned short);
9145 vector signed char vec_packs (vector signed short, vector signed short);
9146 vector unsigned short vec_packs (vector unsigned int,
9147 vector unsigned int);
9148 vector signed short vec_packs (vector signed int, vector signed int);
9150 vector signed short vec_vpkswss (vector signed int, vector signed int);
9152 vector unsigned short vec_vpkuwus (vector unsigned int,
9153 vector unsigned int);
9155 vector signed char vec_vpkshss (vector signed short,
9156 vector signed short);
9158 vector unsigned char vec_vpkuhus (vector unsigned short,
9159 vector unsigned short);
9161 vector unsigned char vec_packsu (vector unsigned short,
9162 vector unsigned short);
9163 vector unsigned char vec_packsu (vector signed short,
9164 vector signed short);
9165 vector unsigned short vec_packsu (vector unsigned int,
9166 vector unsigned int);
9167 vector unsigned short vec_packsu (vector signed int, vector signed int);
9169 vector unsigned short vec_vpkswus (vector signed int,
9172 vector unsigned char vec_vpkshus (vector signed short,
9173 vector signed short);
9175 vector float vec_perm (vector float,
9177 vector unsigned char);
9178 vector signed int vec_perm (vector signed int,
9180 vector unsigned char);
9181 vector unsigned int vec_perm (vector unsigned int,
9182 vector unsigned int,
9183 vector unsigned char);
9184 vector bool int vec_perm (vector bool int,
9186 vector unsigned char);
9187 vector signed short vec_perm (vector signed short,
9188 vector signed short,
9189 vector unsigned char);
9190 vector unsigned short vec_perm (vector unsigned short,
9191 vector unsigned short,
9192 vector unsigned char);
9193 vector bool short vec_perm (vector bool short,
9195 vector unsigned char);
9196 vector pixel vec_perm (vector pixel,
9198 vector unsigned char);
9199 vector signed char vec_perm (vector signed char,
9201 vector unsigned char);
9202 vector unsigned char vec_perm (vector unsigned char,
9203 vector unsigned char,
9204 vector unsigned char);
9205 vector bool char vec_perm (vector bool char,
9207 vector unsigned char);
9209 vector float vec_re (vector float);
9211 vector signed char vec_rl (vector signed char,
9212 vector unsigned char);
9213 vector unsigned char vec_rl (vector unsigned char,
9214 vector unsigned char);
9215 vector signed short vec_rl (vector signed short, vector unsigned short);
9216 vector unsigned short vec_rl (vector unsigned short,
9217 vector unsigned short);
9218 vector signed int vec_rl (vector signed int, vector unsigned int);
9219 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
9221 vector signed int vec_vrlw (vector signed int, vector unsigned int);
9222 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
9224 vector signed short vec_vrlh (vector signed short,
9225 vector unsigned short);
9226 vector unsigned short vec_vrlh (vector unsigned short,
9227 vector unsigned short);
9229 vector signed char vec_vrlb (vector signed char, vector unsigned char);
9230 vector unsigned char vec_vrlb (vector unsigned char,
9231 vector unsigned char);
9233 vector float vec_round (vector float);
9235 vector float vec_rsqrte (vector float);
9237 vector float vec_sel (vector float, vector float, vector bool int);
9238 vector float vec_sel (vector float, vector float, vector unsigned int);
9239 vector signed int vec_sel (vector signed int,
9242 vector signed int vec_sel (vector signed int,
9244 vector unsigned int);
9245 vector unsigned int vec_sel (vector unsigned int,
9246 vector unsigned int,
9248 vector unsigned int vec_sel (vector unsigned int,
9249 vector unsigned int,
9250 vector unsigned int);
9251 vector bool int vec_sel (vector bool int,
9254 vector bool int vec_sel (vector bool int,
9256 vector unsigned int);
9257 vector signed short vec_sel (vector signed short,
9258 vector signed short,
9260 vector signed short vec_sel (vector signed short,
9261 vector signed short,
9262 vector unsigned short);
9263 vector unsigned short vec_sel (vector unsigned short,
9264 vector unsigned short,
9266 vector unsigned short vec_sel (vector unsigned short,
9267 vector unsigned short,
9268 vector unsigned short);
9269 vector bool short vec_sel (vector bool short,
9272 vector bool short vec_sel (vector bool short,
9274 vector unsigned short);
9275 vector signed char vec_sel (vector signed char,
9278 vector signed char vec_sel (vector signed char,
9280 vector unsigned char);
9281 vector unsigned char vec_sel (vector unsigned char,
9282 vector unsigned char,
9284 vector unsigned char vec_sel (vector unsigned char,
9285 vector unsigned char,
9286 vector unsigned char);
9287 vector bool char vec_sel (vector bool char,
9290 vector bool char vec_sel (vector bool char,
9292 vector unsigned char);
9294 vector signed char vec_sl (vector signed char,
9295 vector unsigned char);
9296 vector unsigned char vec_sl (vector unsigned char,
9297 vector unsigned char);
9298 vector signed short vec_sl (vector signed short, vector unsigned short);
9299 vector unsigned short vec_sl (vector unsigned short,
9300 vector unsigned short);
9301 vector signed int vec_sl (vector signed int, vector unsigned int);
9302 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
9304 vector signed int vec_vslw (vector signed int, vector unsigned int);
9305 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
9307 vector signed short vec_vslh (vector signed short,
9308 vector unsigned short);
9309 vector unsigned short vec_vslh (vector unsigned short,
9310 vector unsigned short);
9312 vector signed char vec_vslb (vector signed char, vector unsigned char);
9313 vector unsigned char vec_vslb (vector unsigned char,
9314 vector unsigned char);
9316 vector float vec_sld (vector float, vector float, const int);
9317 vector signed int vec_sld (vector signed int,
9320 vector unsigned int vec_sld (vector unsigned int,
9321 vector unsigned int,
9323 vector bool int vec_sld (vector bool int,
9326 vector signed short vec_sld (vector signed short,
9327 vector signed short,
9329 vector unsigned short vec_sld (vector unsigned short,
9330 vector unsigned short,
9332 vector bool short vec_sld (vector bool short,
9335 vector pixel vec_sld (vector pixel,
9338 vector signed char vec_sld (vector signed char,
9341 vector unsigned char vec_sld (vector unsigned char,
9342 vector unsigned char,
9344 vector bool char vec_sld (vector bool char,
9348 vector signed int vec_sll (vector signed int,
9349 vector unsigned int);
9350 vector signed int vec_sll (vector signed int,
9351 vector unsigned short);
9352 vector signed int vec_sll (vector signed int,
9353 vector unsigned char);
9354 vector unsigned int vec_sll (vector unsigned int,
9355 vector unsigned int);
9356 vector unsigned int vec_sll (vector unsigned int,
9357 vector unsigned short);
9358 vector unsigned int vec_sll (vector unsigned int,
9359 vector unsigned char);
9360 vector bool int vec_sll (vector bool int,
9361 vector unsigned int);
9362 vector bool int vec_sll (vector bool int,
9363 vector unsigned short);
9364 vector bool int vec_sll (vector bool int,
9365 vector unsigned char);
9366 vector signed short vec_sll (vector signed short,
9367 vector unsigned int);
9368 vector signed short vec_sll (vector signed short,
9369 vector unsigned short);
9370 vector signed short vec_sll (vector signed short,
9371 vector unsigned char);
9372 vector unsigned short vec_sll (vector unsigned short,
9373 vector unsigned int);
9374 vector unsigned short vec_sll (vector unsigned short,
9375 vector unsigned short);
9376 vector unsigned short vec_sll (vector unsigned short,
9377 vector unsigned char);
9378 vector bool short vec_sll (vector bool short, vector unsigned int);
9379 vector bool short vec_sll (vector bool short, vector unsigned short);
9380 vector bool short vec_sll (vector bool short, vector unsigned char);
9381 vector pixel vec_sll (vector pixel, vector unsigned int);
9382 vector pixel vec_sll (vector pixel, vector unsigned short);
9383 vector pixel vec_sll (vector pixel, vector unsigned char);
9384 vector signed char vec_sll (vector signed char, vector unsigned int);
9385 vector signed char vec_sll (vector signed char, vector unsigned short);
9386 vector signed char vec_sll (vector signed char, vector unsigned char);
9387 vector unsigned char vec_sll (vector unsigned char,
9388 vector unsigned int);
9389 vector unsigned char vec_sll (vector unsigned char,
9390 vector unsigned short);
9391 vector unsigned char vec_sll (vector unsigned char,
9392 vector unsigned char);
9393 vector bool char vec_sll (vector bool char, vector unsigned int);
9394 vector bool char vec_sll (vector bool char, vector unsigned short);
9395 vector bool char vec_sll (vector bool char, vector unsigned char);
9397 vector float vec_slo (vector float, vector signed char);
9398 vector float vec_slo (vector float, vector unsigned char);
9399 vector signed int vec_slo (vector signed int, vector signed char);
9400 vector signed int vec_slo (vector signed int, vector unsigned char);
9401 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9402 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9403 vector signed short vec_slo (vector signed short, vector signed char);
9404 vector signed short vec_slo (vector signed short, vector unsigned char);
9405 vector unsigned short vec_slo (vector unsigned short,
9406 vector signed char);
9407 vector unsigned short vec_slo (vector unsigned short,
9408 vector unsigned char);
9409 vector pixel vec_slo (vector pixel, vector signed char);
9410 vector pixel vec_slo (vector pixel, vector unsigned char);
9411 vector signed char vec_slo (vector signed char, vector signed char);
9412 vector signed char vec_slo (vector signed char, vector unsigned char);
9413 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9414 vector unsigned char vec_slo (vector unsigned char,
9415 vector unsigned char);
9417 vector signed char vec_splat (vector signed char, const int);
9418 vector unsigned char vec_splat (vector unsigned char, const int);
9419 vector bool char vec_splat (vector bool char, const int);
9420 vector signed short vec_splat (vector signed short, const int);
9421 vector unsigned short vec_splat (vector unsigned short, const int);
9422 vector bool short vec_splat (vector bool short, const int);
9423 vector pixel vec_splat (vector pixel, const int);
9424 vector float vec_splat (vector float, const int);
9425 vector signed int vec_splat (vector signed int, const int);
9426 vector unsigned int vec_splat (vector unsigned int, const int);
9427 vector bool int vec_splat (vector bool int, const int);
9429 vector float vec_vspltw (vector float, const int);
9430 vector signed int vec_vspltw (vector signed int, const int);
9431 vector unsigned int vec_vspltw (vector unsigned int, const int);
9432 vector bool int vec_vspltw (vector bool int, const int);
9434 vector bool short vec_vsplth (vector bool short, const int);
9435 vector signed short vec_vsplth (vector signed short, const int);
9436 vector unsigned short vec_vsplth (vector unsigned short, const int);
9437 vector pixel vec_vsplth (vector pixel, const int);
9439 vector signed char vec_vspltb (vector signed char, const int);
9440 vector unsigned char vec_vspltb (vector unsigned char, const int);
9441 vector bool char vec_vspltb (vector bool char, const int);
9443 vector signed char vec_splat_s8 (const int);
9445 vector signed short vec_splat_s16 (const int);
9447 vector signed int vec_splat_s32 (const int);
9449 vector unsigned char vec_splat_u8 (const int);
9451 vector unsigned short vec_splat_u16 (const int);
9453 vector unsigned int vec_splat_u32 (const int);
9455 vector signed char vec_sr (vector signed char, vector unsigned char);
9456 vector unsigned char vec_sr (vector unsigned char,
9457 vector unsigned char);
9458 vector signed short vec_sr (vector signed short,
9459 vector unsigned short);
9460 vector unsigned short vec_sr (vector unsigned short,
9461 vector unsigned short);
9462 vector signed int vec_sr (vector signed int, vector unsigned int);
9463 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9465 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9466 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9468 vector signed short vec_vsrh (vector signed short,
9469 vector unsigned short);
9470 vector unsigned short vec_vsrh (vector unsigned short,
9471 vector unsigned short);
9473 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9474 vector unsigned char vec_vsrb (vector unsigned char,
9475 vector unsigned char);
9477 vector signed char vec_sra (vector signed char, vector unsigned char);
9478 vector unsigned char vec_sra (vector unsigned char,
9479 vector unsigned char);
9480 vector signed short vec_sra (vector signed short,
9481 vector unsigned short);
9482 vector unsigned short vec_sra (vector unsigned short,
9483 vector unsigned short);
9484 vector signed int vec_sra (vector signed int, vector unsigned int);
9485 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9487 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9488 vector unsigned int vec_vsraw (vector unsigned int,
9489 vector unsigned int);
9491 vector signed short vec_vsrah (vector signed short,
9492 vector unsigned short);
9493 vector unsigned short vec_vsrah (vector unsigned short,
9494 vector unsigned short);
9496 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9497 vector unsigned char vec_vsrab (vector unsigned char,
9498 vector unsigned char);
9500 vector signed int vec_srl (vector signed int, vector unsigned int);
9501 vector signed int vec_srl (vector signed int, vector unsigned short);
9502 vector signed int vec_srl (vector signed int, vector unsigned char);
9503 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9504 vector unsigned int vec_srl (vector unsigned int,
9505 vector unsigned short);
9506 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9507 vector bool int vec_srl (vector bool int, vector unsigned int);
9508 vector bool int vec_srl (vector bool int, vector unsigned short);
9509 vector bool int vec_srl (vector bool int, vector unsigned char);
9510 vector signed short vec_srl (vector signed short, vector unsigned int);
9511 vector signed short vec_srl (vector signed short,
9512 vector unsigned short);
9513 vector signed short vec_srl (vector signed short, vector unsigned char);
9514 vector unsigned short vec_srl (vector unsigned short,
9515 vector unsigned int);
9516 vector unsigned short vec_srl (vector unsigned short,
9517 vector unsigned short);
9518 vector unsigned short vec_srl (vector unsigned short,
9519 vector unsigned char);
9520 vector bool short vec_srl (vector bool short, vector unsigned int);
9521 vector bool short vec_srl (vector bool short, vector unsigned short);
9522 vector bool short vec_srl (vector bool short, vector unsigned char);
9523 vector pixel vec_srl (vector pixel, vector unsigned int);
9524 vector pixel vec_srl (vector pixel, vector unsigned short);
9525 vector pixel vec_srl (vector pixel, vector unsigned char);
9526 vector signed char vec_srl (vector signed char, vector unsigned int);
9527 vector signed char vec_srl (vector signed char, vector unsigned short);
9528 vector signed char vec_srl (vector signed char, vector unsigned char);
9529 vector unsigned char vec_srl (vector unsigned char,
9530 vector unsigned int);
9531 vector unsigned char vec_srl (vector unsigned char,
9532 vector unsigned short);
9533 vector unsigned char vec_srl (vector unsigned char,
9534 vector unsigned char);
9535 vector bool char vec_srl (vector bool char, vector unsigned int);
9536 vector bool char vec_srl (vector bool char, vector unsigned short);
9537 vector bool char vec_srl (vector bool char, vector unsigned char);
9539 vector float vec_sro (vector float, vector signed char);
9540 vector float vec_sro (vector float, vector unsigned char);
9541 vector signed int vec_sro (vector signed int, vector signed char);
9542 vector signed int vec_sro (vector signed int, vector unsigned char);
9543 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9544 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9545 vector signed short vec_sro (vector signed short, vector signed char);
9546 vector signed short vec_sro (vector signed short, vector unsigned char);
9547 vector unsigned short vec_sro (vector unsigned short,
9548 vector signed char);
9549 vector unsigned short vec_sro (vector unsigned short,
9550 vector unsigned char);
9551 vector pixel vec_sro (vector pixel, vector signed char);
9552 vector pixel vec_sro (vector pixel, vector unsigned char);
9553 vector signed char vec_sro (vector signed char, vector signed char);
9554 vector signed char vec_sro (vector signed char, vector unsigned char);
9555 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9556 vector unsigned char vec_sro (vector unsigned char,
9557 vector unsigned char);
9559 void vec_st (vector float, int, vector float *);
9560 void vec_st (vector float, int, float *);
9561 void vec_st (vector signed int, int, vector signed int *);
9562 void vec_st (vector signed int, int, int *);
9563 void vec_st (vector unsigned int, int, vector unsigned int *);
9564 void vec_st (vector unsigned int, int, unsigned int *);
9565 void vec_st (vector bool int, int, vector bool int *);
9566 void vec_st (vector bool int, int, unsigned int *);
9567 void vec_st (vector bool int, int, int *);
9568 void vec_st (vector signed short, int, vector signed short *);
9569 void vec_st (vector signed short, int, short *);
9570 void vec_st (vector unsigned short, int, vector unsigned short *);
9571 void vec_st (vector unsigned short, int, unsigned short *);
9572 void vec_st (vector bool short, int, vector bool short *);
9573 void vec_st (vector bool short, int, unsigned short *);
9574 void vec_st (vector pixel, int, vector pixel *);
9575 void vec_st (vector pixel, int, unsigned short *);
9576 void vec_st (vector pixel, int, short *);
9577 void vec_st (vector bool short, int, short *);
9578 void vec_st (vector signed char, int, vector signed char *);
9579 void vec_st (vector signed char, int, signed char *);
9580 void vec_st (vector unsigned char, int, vector unsigned char *);
9581 void vec_st (vector unsigned char, int, unsigned char *);
9582 void vec_st (vector bool char, int, vector bool char *);
9583 void vec_st (vector bool char, int, unsigned char *);
9584 void vec_st (vector bool char, int, signed char *);
9586 void vec_ste (vector signed char, int, signed char *);
9587 void vec_ste (vector unsigned char, int, unsigned char *);
9588 void vec_ste (vector bool char, int, signed char *);
9589 void vec_ste (vector bool char, int, unsigned char *);
9590 void vec_ste (vector signed short, int, short *);
9591 void vec_ste (vector unsigned short, int, unsigned short *);
9592 void vec_ste (vector bool short, int, short *);
9593 void vec_ste (vector bool short, int, unsigned short *);
9594 void vec_ste (vector pixel, int, short *);
9595 void vec_ste (vector pixel, int, unsigned short *);
9596 void vec_ste (vector float, int, float *);
9597 void vec_ste (vector signed int, int, int *);
9598 void vec_ste (vector unsigned int, int, unsigned int *);
9599 void vec_ste (vector bool int, int, int *);
9600 void vec_ste (vector bool int, int, unsigned int *);
9602 void vec_stvewx (vector float, int, float *);
9603 void vec_stvewx (vector signed int, int, int *);
9604 void vec_stvewx (vector unsigned int, int, unsigned int *);
9605 void vec_stvewx (vector bool int, int, int *);
9606 void vec_stvewx (vector bool int, int, unsigned int *);
9608 void vec_stvehx (vector signed short, int, short *);
9609 void vec_stvehx (vector unsigned short, int, unsigned short *);
9610 void vec_stvehx (vector bool short, int, short *);
9611 void vec_stvehx (vector bool short, int, unsigned short *);
9612 void vec_stvehx (vector pixel, int, short *);
9613 void vec_stvehx (vector pixel, int, unsigned short *);
9615 void vec_stvebx (vector signed char, int, signed char *);
9616 void vec_stvebx (vector unsigned char, int, unsigned char *);
9617 void vec_stvebx (vector bool char, int, signed char *);
9618 void vec_stvebx (vector bool char, int, unsigned char *);
9620 void vec_stl (vector float, int, vector float *);
9621 void vec_stl (vector float, int, float *);
9622 void vec_stl (vector signed int, int, vector signed int *);
9623 void vec_stl (vector signed int, int, int *);
9624 void vec_stl (vector unsigned int, int, vector unsigned int *);
9625 void vec_stl (vector unsigned int, int, unsigned int *);
9626 void vec_stl (vector bool int, int, vector bool int *);
9627 void vec_stl (vector bool int, int, unsigned int *);
9628 void vec_stl (vector bool int, int, int *);
9629 void vec_stl (vector signed short, int, vector signed short *);
9630 void vec_stl (vector signed short, int, short *);
9631 void vec_stl (vector unsigned short, int, vector unsigned short *);
9632 void vec_stl (vector unsigned short, int, unsigned short *);
9633 void vec_stl (vector bool short, int, vector bool short *);
9634 void vec_stl (vector bool short, int, unsigned short *);
9635 void vec_stl (vector bool short, int, short *);
9636 void vec_stl (vector pixel, int, vector pixel *);
9637 void vec_stl (vector pixel, int, unsigned short *);
9638 void vec_stl (vector pixel, int, short *);
9639 void vec_stl (vector signed char, int, vector signed char *);
9640 void vec_stl (vector signed char, int, signed char *);
9641 void vec_stl (vector unsigned char, int, vector unsigned char *);
9642 void vec_stl (vector unsigned char, int, unsigned char *);
9643 void vec_stl (vector bool char, int, vector bool char *);
9644 void vec_stl (vector bool char, int, unsigned char *);
9645 void vec_stl (vector bool char, int, signed char *);
9647 vector signed char vec_sub (vector bool char, vector signed char);
9648 vector signed char vec_sub (vector signed char, vector bool char);
9649 vector signed char vec_sub (vector signed char, vector signed char);
9650 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9651 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9652 vector unsigned char vec_sub (vector unsigned char,
9653 vector unsigned char);
9654 vector signed short vec_sub (vector bool short, vector signed short);
9655 vector signed short vec_sub (vector signed short, vector bool short);
9656 vector signed short vec_sub (vector signed short, vector signed short);
9657 vector unsigned short vec_sub (vector bool short,
9658 vector unsigned short);
9659 vector unsigned short vec_sub (vector unsigned short,
9661 vector unsigned short vec_sub (vector unsigned short,
9662 vector unsigned short);
9663 vector signed int vec_sub (vector bool int, vector signed int);
9664 vector signed int vec_sub (vector signed int, vector bool int);
9665 vector signed int vec_sub (vector signed int, vector signed int);
9666 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9667 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9668 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9669 vector float vec_sub (vector float, vector float);
9671 vector float vec_vsubfp (vector float, vector float);
9673 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9674 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9675 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9676 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9677 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9678 vector unsigned int vec_vsubuwm (vector unsigned int,
9679 vector unsigned int);
9681 vector signed short vec_vsubuhm (vector bool short,
9682 vector signed short);
9683 vector signed short vec_vsubuhm (vector signed short,
9685 vector signed short vec_vsubuhm (vector signed short,
9686 vector signed short);
9687 vector unsigned short vec_vsubuhm (vector bool short,
9688 vector unsigned short);
9689 vector unsigned short vec_vsubuhm (vector unsigned short,
9691 vector unsigned short vec_vsubuhm (vector unsigned short,
9692 vector unsigned short);
9694 vector signed char vec_vsububm (vector bool char, vector signed char);
9695 vector signed char vec_vsububm (vector signed char, vector bool char);
9696 vector signed char vec_vsububm (vector signed char, vector signed char);
9697 vector unsigned char vec_vsububm (vector bool char,
9698 vector unsigned char);
9699 vector unsigned char vec_vsububm (vector unsigned char,
9701 vector unsigned char vec_vsububm (vector unsigned char,
9702 vector unsigned char);
9704 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9706 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9707 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9708 vector unsigned char vec_subs (vector unsigned char,
9709 vector unsigned char);
9710 vector signed char vec_subs (vector bool char, vector signed char);
9711 vector signed char vec_subs (vector signed char, vector bool char);
9712 vector signed char vec_subs (vector signed char, vector signed char);
9713 vector unsigned short vec_subs (vector bool short,
9714 vector unsigned short);
9715 vector unsigned short vec_subs (vector unsigned short,
9717 vector unsigned short vec_subs (vector unsigned short,
9718 vector unsigned short);
9719 vector signed short vec_subs (vector bool short, vector signed short);
9720 vector signed short vec_subs (vector signed short, vector bool short);
9721 vector signed short vec_subs (vector signed short, vector signed short);
9722 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9723 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9724 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9725 vector signed int vec_subs (vector bool int, vector signed int);
9726 vector signed int vec_subs (vector signed int, vector bool int);
9727 vector signed int vec_subs (vector signed int, vector signed int);
9729 vector signed int vec_vsubsws (vector bool int, vector signed int);
9730 vector signed int vec_vsubsws (vector signed int, vector bool int);
9731 vector signed int vec_vsubsws (vector signed int, vector signed int);
9733 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9734 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9735 vector unsigned int vec_vsubuws (vector unsigned int,
9736 vector unsigned int);
9738 vector signed short vec_vsubshs (vector bool short,
9739 vector signed short);
9740 vector signed short vec_vsubshs (vector signed short,
9742 vector signed short vec_vsubshs (vector signed short,
9743 vector signed short);
9745 vector unsigned short vec_vsubuhs (vector bool short,
9746 vector unsigned short);
9747 vector unsigned short vec_vsubuhs (vector unsigned short,
9749 vector unsigned short vec_vsubuhs (vector unsigned short,
9750 vector unsigned short);
9752 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9753 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9754 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9756 vector unsigned char vec_vsububs (vector bool char,
9757 vector unsigned char);
9758 vector unsigned char vec_vsububs (vector unsigned char,
9760 vector unsigned char vec_vsububs (vector unsigned char,
9761 vector unsigned char);
9763 vector unsigned int vec_sum4s (vector unsigned char,
9764 vector unsigned int);
9765 vector signed int vec_sum4s (vector signed char, vector signed int);
9766 vector signed int vec_sum4s (vector signed short, vector signed int);
9768 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9770 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9772 vector unsigned int vec_vsum4ubs (vector unsigned char,
9773 vector unsigned int);
9775 vector signed int vec_sum2s (vector signed int, vector signed int);
9777 vector signed int vec_sums (vector signed int, vector signed int);
9779 vector float vec_trunc (vector float);
9781 vector signed short vec_unpackh (vector signed char);
9782 vector bool short vec_unpackh (vector bool char);
9783 vector signed int vec_unpackh (vector signed short);
9784 vector bool int vec_unpackh (vector bool short);
9785 vector unsigned int vec_unpackh (vector pixel);
9787 vector bool int vec_vupkhsh (vector bool short);
9788 vector signed int vec_vupkhsh (vector signed short);
9790 vector unsigned int vec_vupkhpx (vector pixel);
9792 vector bool short vec_vupkhsb (vector bool char);
9793 vector signed short vec_vupkhsb (vector signed char);
9795 vector signed short vec_unpackl (vector signed char);
9796 vector bool short vec_unpackl (vector bool char);
9797 vector unsigned int vec_unpackl (vector pixel);
9798 vector signed int vec_unpackl (vector signed short);
9799 vector bool int vec_unpackl (vector bool short);
9801 vector unsigned int vec_vupklpx (vector pixel);
9803 vector bool int vec_vupklsh (vector bool short);
9804 vector signed int vec_vupklsh (vector signed short);
9806 vector bool short vec_vupklsb (vector bool char);
9807 vector signed short vec_vupklsb (vector signed char);
9809 vector float vec_xor (vector float, vector float);
9810 vector float vec_xor (vector float, vector bool int);
9811 vector float vec_xor (vector bool int, vector float);
9812 vector bool int vec_xor (vector bool int, vector bool int);
9813 vector signed int vec_xor (vector bool int, vector signed int);
9814 vector signed int vec_xor (vector signed int, vector bool int);
9815 vector signed int vec_xor (vector signed int, vector signed int);
9816 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9817 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9818 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9819 vector bool short vec_xor (vector bool short, vector bool short);
9820 vector signed short vec_xor (vector bool short, vector signed short);
9821 vector signed short vec_xor (vector signed short, vector bool short);
9822 vector signed short vec_xor (vector signed short, vector signed short);
9823 vector unsigned short vec_xor (vector bool short,
9824 vector unsigned short);
9825 vector unsigned short vec_xor (vector unsigned short,
9827 vector unsigned short vec_xor (vector unsigned short,
9828 vector unsigned short);
9829 vector signed char vec_xor (vector bool char, vector signed char);
9830 vector bool char vec_xor (vector bool char, vector bool char);
9831 vector signed char vec_xor (vector signed char, vector bool char);
9832 vector signed char vec_xor (vector signed char, vector signed char);
9833 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9834 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9835 vector unsigned char vec_xor (vector unsigned char,
9836 vector unsigned char);
9838 int vec_all_eq (vector signed char, vector bool char);
9839 int vec_all_eq (vector signed char, vector signed char);
9840 int vec_all_eq (vector unsigned char, vector bool char);
9841 int vec_all_eq (vector unsigned char, vector unsigned char);
9842 int vec_all_eq (vector bool char, vector bool char);
9843 int vec_all_eq (vector bool char, vector unsigned char);
9844 int vec_all_eq (vector bool char, vector signed char);
9845 int vec_all_eq (vector signed short, vector bool short);
9846 int vec_all_eq (vector signed short, vector signed short);
9847 int vec_all_eq (vector unsigned short, vector bool short);
9848 int vec_all_eq (vector unsigned short, vector unsigned short);
9849 int vec_all_eq (vector bool short, vector bool short);
9850 int vec_all_eq (vector bool short, vector unsigned short);
9851 int vec_all_eq (vector bool short, vector signed short);
9852 int vec_all_eq (vector pixel, vector pixel);
9853 int vec_all_eq (vector signed int, vector bool int);
9854 int vec_all_eq (vector signed int, vector signed int);
9855 int vec_all_eq (vector unsigned int, vector bool int);
9856 int vec_all_eq (vector unsigned int, vector unsigned int);
9857 int vec_all_eq (vector bool int, vector bool int);
9858 int vec_all_eq (vector bool int, vector unsigned int);
9859 int vec_all_eq (vector bool int, vector signed int);
9860 int vec_all_eq (vector float, vector float);
9862 int vec_all_ge (vector bool char, vector unsigned char);
9863 int vec_all_ge (vector unsigned char, vector bool char);
9864 int vec_all_ge (vector unsigned char, vector unsigned char);
9865 int vec_all_ge (vector bool char, vector signed char);
9866 int vec_all_ge (vector signed char, vector bool char);
9867 int vec_all_ge (vector signed char, vector signed char);
9868 int vec_all_ge (vector bool short, vector unsigned short);
9869 int vec_all_ge (vector unsigned short, vector bool short);
9870 int vec_all_ge (vector unsigned short, vector unsigned short);
9871 int vec_all_ge (vector signed short, vector signed short);
9872 int vec_all_ge (vector bool short, vector signed short);
9873 int vec_all_ge (vector signed short, vector bool short);
9874 int vec_all_ge (vector bool int, vector unsigned int);
9875 int vec_all_ge (vector unsigned int, vector bool int);
9876 int vec_all_ge (vector unsigned int, vector unsigned int);
9877 int vec_all_ge (vector bool int, vector signed int);
9878 int vec_all_ge (vector signed int, vector bool int);
9879 int vec_all_ge (vector signed int, vector signed int);
9880 int vec_all_ge (vector float, vector float);
9882 int vec_all_gt (vector bool char, vector unsigned char);
9883 int vec_all_gt (vector unsigned char, vector bool char);
9884 int vec_all_gt (vector unsigned char, vector unsigned char);
9885 int vec_all_gt (vector bool char, vector signed char);
9886 int vec_all_gt (vector signed char, vector bool char);
9887 int vec_all_gt (vector signed char, vector signed char);
9888 int vec_all_gt (vector bool short, vector unsigned short);
9889 int vec_all_gt (vector unsigned short, vector bool short);
9890 int vec_all_gt (vector unsigned short, vector unsigned short);
9891 int vec_all_gt (vector bool short, vector signed short);
9892 int vec_all_gt (vector signed short, vector bool short);
9893 int vec_all_gt (vector signed short, vector signed short);
9894 int vec_all_gt (vector bool int, vector unsigned int);
9895 int vec_all_gt (vector unsigned int, vector bool int);
9896 int vec_all_gt (vector unsigned int, vector unsigned int);
9897 int vec_all_gt (vector bool int, vector signed int);
9898 int vec_all_gt (vector signed int, vector bool int);
9899 int vec_all_gt (vector signed int, vector signed int);
9900 int vec_all_gt (vector float, vector float);
9902 int vec_all_in (vector float, vector float);
9904 int vec_all_le (vector bool char, vector unsigned char);
9905 int vec_all_le (vector unsigned char, vector bool char);
9906 int vec_all_le (vector unsigned char, vector unsigned char);
9907 int vec_all_le (vector bool char, vector signed char);
9908 int vec_all_le (vector signed char, vector bool char);
9909 int vec_all_le (vector signed char, vector signed char);
9910 int vec_all_le (vector bool short, vector unsigned short);
9911 int vec_all_le (vector unsigned short, vector bool short);
9912 int vec_all_le (vector unsigned short, vector unsigned short);
9913 int vec_all_le (vector bool short, vector signed short);
9914 int vec_all_le (vector signed short, vector bool short);
9915 int vec_all_le (vector signed short, vector signed short);
9916 int vec_all_le (vector bool int, vector unsigned int);
9917 int vec_all_le (vector unsigned int, vector bool int);
9918 int vec_all_le (vector unsigned int, vector unsigned int);
9919 int vec_all_le (vector bool int, vector signed int);
9920 int vec_all_le (vector signed int, vector bool int);
9921 int vec_all_le (vector signed int, vector signed int);
9922 int vec_all_le (vector float, vector float);
9924 int vec_all_lt (vector bool char, vector unsigned char);
9925 int vec_all_lt (vector unsigned char, vector bool char);
9926 int vec_all_lt (vector unsigned char, vector unsigned char);
9927 int vec_all_lt (vector bool char, vector signed char);
9928 int vec_all_lt (vector signed char, vector bool char);
9929 int vec_all_lt (vector signed char, vector signed char);
9930 int vec_all_lt (vector bool short, vector unsigned short);
9931 int vec_all_lt (vector unsigned short, vector bool short);
9932 int vec_all_lt (vector unsigned short, vector unsigned short);
9933 int vec_all_lt (vector bool short, vector signed short);
9934 int vec_all_lt (vector signed short, vector bool short);
9935 int vec_all_lt (vector signed short, vector signed short);
9936 int vec_all_lt (vector bool int, vector unsigned int);
9937 int vec_all_lt (vector unsigned int, vector bool int);
9938 int vec_all_lt (vector unsigned int, vector unsigned int);
9939 int vec_all_lt (vector bool int, vector signed int);
9940 int vec_all_lt (vector signed int, vector bool int);
9941 int vec_all_lt (vector signed int, vector signed int);
9942 int vec_all_lt (vector float, vector float);
9944 int vec_all_nan (vector float);
9946 int vec_all_ne (vector signed char, vector bool char);
9947 int vec_all_ne (vector signed char, vector signed char);
9948 int vec_all_ne (vector unsigned char, vector bool char);
9949 int vec_all_ne (vector unsigned char, vector unsigned char);
9950 int vec_all_ne (vector bool char, vector bool char);
9951 int vec_all_ne (vector bool char, vector unsigned char);
9952 int vec_all_ne (vector bool char, vector signed char);
9953 int vec_all_ne (vector signed short, vector bool short);
9954 int vec_all_ne (vector signed short, vector signed short);
9955 int vec_all_ne (vector unsigned short, vector bool short);
9956 int vec_all_ne (vector unsigned short, vector unsigned short);
9957 int vec_all_ne (vector bool short, vector bool short);
9958 int vec_all_ne (vector bool short, vector unsigned short);
9959 int vec_all_ne (vector bool short, vector signed short);
9960 int vec_all_ne (vector pixel, vector pixel);
9961 int vec_all_ne (vector signed int, vector bool int);
9962 int vec_all_ne (vector signed int, vector signed int);
9963 int vec_all_ne (vector unsigned int, vector bool int);
9964 int vec_all_ne (vector unsigned int, vector unsigned int);
9965 int vec_all_ne (vector bool int, vector bool int);
9966 int vec_all_ne (vector bool int, vector unsigned int);
9967 int vec_all_ne (vector bool int, vector signed int);
9968 int vec_all_ne (vector float, vector float);
9970 int vec_all_nge (vector float, vector float);
9972 int vec_all_ngt (vector float, vector float);
9974 int vec_all_nle (vector float, vector float);
9976 int vec_all_nlt (vector float, vector float);
9978 int vec_all_numeric (vector float);
9980 int vec_any_eq (vector signed char, vector bool char);
9981 int vec_any_eq (vector signed char, vector signed char);
9982 int vec_any_eq (vector unsigned char, vector bool char);
9983 int vec_any_eq (vector unsigned char, vector unsigned char);
9984 int vec_any_eq (vector bool char, vector bool char);
9985 int vec_any_eq (vector bool char, vector unsigned char);
9986 int vec_any_eq (vector bool char, vector signed char);
9987 int vec_any_eq (vector signed short, vector bool short);
9988 int vec_any_eq (vector signed short, vector signed short);
9989 int vec_any_eq (vector unsigned short, vector bool short);
9990 int vec_any_eq (vector unsigned short, vector unsigned short);
9991 int vec_any_eq (vector bool short, vector bool short);
9992 int vec_any_eq (vector bool short, vector unsigned short);
9993 int vec_any_eq (vector bool short, vector signed short);
9994 int vec_any_eq (vector pixel, vector pixel);
9995 int vec_any_eq (vector signed int, vector bool int);
9996 int vec_any_eq (vector signed int, vector signed int);
9997 int vec_any_eq (vector unsigned int, vector bool int);
9998 int vec_any_eq (vector unsigned int, vector unsigned int);
9999 int vec_any_eq (vector bool int, vector bool int);
10000 int vec_any_eq (vector bool int, vector unsigned int);
10001 int vec_any_eq (vector bool int, vector signed int);
10002 int vec_any_eq (vector float, vector float);
10004 int vec_any_ge (vector signed char, vector bool char);
10005 int vec_any_ge (vector unsigned char, vector bool char);
10006 int vec_any_ge (vector unsigned char, vector unsigned char);
10007 int vec_any_ge (vector signed char, vector signed char);
10008 int vec_any_ge (vector bool char, vector unsigned char);
10009 int vec_any_ge (vector bool char, vector signed char);
10010 int vec_any_ge (vector unsigned short, vector bool short);
10011 int vec_any_ge (vector unsigned short, vector unsigned short);
10012 int vec_any_ge (vector signed short, vector signed short);
10013 int vec_any_ge (vector signed short, vector bool short);
10014 int vec_any_ge (vector bool short, vector unsigned short);
10015 int vec_any_ge (vector bool short, vector signed short);
10016 int vec_any_ge (vector signed int, vector bool int);
10017 int vec_any_ge (vector unsigned int, vector bool int);
10018 int vec_any_ge (vector unsigned int, vector unsigned int);
10019 int vec_any_ge (vector signed int, vector signed int);
10020 int vec_any_ge (vector bool int, vector unsigned int);
10021 int vec_any_ge (vector bool int, vector signed int);
10022 int vec_any_ge (vector float, vector float);
10024 int vec_any_gt (vector bool char, vector unsigned char);
10025 int vec_any_gt (vector unsigned char, vector bool char);
10026 int vec_any_gt (vector unsigned char, vector unsigned char);
10027 int vec_any_gt (vector bool char, vector signed char);
10028 int vec_any_gt (vector signed char, vector bool char);
10029 int vec_any_gt (vector signed char, vector signed char);
10030 int vec_any_gt (vector bool short, vector unsigned short);
10031 int vec_any_gt (vector unsigned short, vector bool short);
10032 int vec_any_gt (vector unsigned short, vector unsigned short);
10033 int vec_any_gt (vector bool short, vector signed short);
10034 int vec_any_gt (vector signed short, vector bool short);
10035 int vec_any_gt (vector signed short, vector signed short);
10036 int vec_any_gt (vector bool int, vector unsigned int);
10037 int vec_any_gt (vector unsigned int, vector bool int);
10038 int vec_any_gt (vector unsigned int, vector unsigned int);
10039 int vec_any_gt (vector bool int, vector signed int);
10040 int vec_any_gt (vector signed int, vector bool int);
10041 int vec_any_gt (vector signed int, vector signed int);
10042 int vec_any_gt (vector float, vector float);
10044 int vec_any_le (vector bool char, vector unsigned char);
10045 int vec_any_le (vector unsigned char, vector bool char);
10046 int vec_any_le (vector unsigned char, vector unsigned char);
10047 int vec_any_le (vector bool char, vector signed char);
10048 int vec_any_le (vector signed char, vector bool char);
10049 int vec_any_le (vector signed char, vector signed char);
10050 int vec_any_le (vector bool short, vector unsigned short);
10051 int vec_any_le (vector unsigned short, vector bool short);
10052 int vec_any_le (vector unsigned short, vector unsigned short);
10053 int vec_any_le (vector bool short, vector signed short);
10054 int vec_any_le (vector signed short, vector bool short);
10055 int vec_any_le (vector signed short, vector signed short);
10056 int vec_any_le (vector bool int, vector unsigned int);
10057 int vec_any_le (vector unsigned int, vector bool int);
10058 int vec_any_le (vector unsigned int, vector unsigned int);
10059 int vec_any_le (vector bool int, vector signed int);
10060 int vec_any_le (vector signed int, vector bool int);
10061 int vec_any_le (vector signed int, vector signed int);
10062 int vec_any_le (vector float, vector float);
10064 int vec_any_lt (vector bool char, vector unsigned char);
10065 int vec_any_lt (vector unsigned char, vector bool char);
10066 int vec_any_lt (vector unsigned char, vector unsigned char);
10067 int vec_any_lt (vector bool char, vector signed char);
10068 int vec_any_lt (vector signed char, vector bool char);
10069 int vec_any_lt (vector signed char, vector signed char);
10070 int vec_any_lt (vector bool short, vector unsigned short);
10071 int vec_any_lt (vector unsigned short, vector bool short);
10072 int vec_any_lt (vector unsigned short, vector unsigned short);
10073 int vec_any_lt (vector bool short, vector signed short);
10074 int vec_any_lt (vector signed short, vector bool short);
10075 int vec_any_lt (vector signed short, vector signed short);
10076 int vec_any_lt (vector bool int, vector unsigned int);
10077 int vec_any_lt (vector unsigned int, vector bool int);
10078 int vec_any_lt (vector unsigned int, vector unsigned int);
10079 int vec_any_lt (vector bool int, vector signed int);
10080 int vec_any_lt (vector signed int, vector bool int);
10081 int vec_any_lt (vector signed int, vector signed int);
10082 int vec_any_lt (vector float, vector float);
10084 int vec_any_nan (vector float);
10086 int vec_any_ne (vector signed char, vector bool char);
10087 int vec_any_ne (vector signed char, vector signed char);
10088 int vec_any_ne (vector unsigned char, vector bool char);
10089 int vec_any_ne (vector unsigned char, vector unsigned char);
10090 int vec_any_ne (vector bool char, vector bool char);
10091 int vec_any_ne (vector bool char, vector unsigned char);
10092 int vec_any_ne (vector bool char, vector signed char);
10093 int vec_any_ne (vector signed short, vector bool short);
10094 int vec_any_ne (vector signed short, vector signed short);
10095 int vec_any_ne (vector unsigned short, vector bool short);
10096 int vec_any_ne (vector unsigned short, vector unsigned short);
10097 int vec_any_ne (vector bool short, vector bool short);
10098 int vec_any_ne (vector bool short, vector unsigned short);
10099 int vec_any_ne (vector bool short, vector signed short);
10100 int vec_any_ne (vector pixel, vector pixel);
10101 int vec_any_ne (vector signed int, vector bool int);
10102 int vec_any_ne (vector signed int, vector signed int);
10103 int vec_any_ne (vector unsigned int, vector bool int);
10104 int vec_any_ne (vector unsigned int, vector unsigned int);
10105 int vec_any_ne (vector bool int, vector bool int);
10106 int vec_any_ne (vector bool int, vector unsigned int);
10107 int vec_any_ne (vector bool int, vector signed int);
10108 int vec_any_ne (vector float, vector float);
10110 int vec_any_nge (vector float, vector float);
10112 int vec_any_ngt (vector float, vector float);
10114 int vec_any_nle (vector float, vector float);
10116 int vec_any_nlt (vector float, vector float);
10118 int vec_any_numeric (vector float);
10120 int vec_any_out (vector float, vector float);
10123 @node SPARC VIS Built-in Functions
10124 @subsection SPARC VIS Built-in Functions
10126 GCC supports SIMD operations on the SPARC using both the generic vector
10127 extensions (@pxref{Vector Extensions}) as well as built-in functions for
10128 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
10129 switch, the VIS extension is exposed as the following built-in functions:
10132 typedef int v2si __attribute__ ((vector_size (8)));
10133 typedef short v4hi __attribute__ ((vector_size (8)));
10134 typedef short v2hi __attribute__ ((vector_size (4)));
10135 typedef char v8qi __attribute__ ((vector_size (8)));
10136 typedef char v4qi __attribute__ ((vector_size (4)));
10138 void * __builtin_vis_alignaddr (void *, long);
10139 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
10140 v2si __builtin_vis_faligndatav2si (v2si, v2si);
10141 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
10142 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
10144 v4hi __builtin_vis_fexpand (v4qi);
10146 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
10147 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
10148 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
10149 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
10150 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
10151 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
10152 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
10154 v4qi __builtin_vis_fpack16 (v4hi);
10155 v8qi __builtin_vis_fpack32 (v2si, v2si);
10156 v2hi __builtin_vis_fpackfix (v2si);
10157 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
10159 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
10162 @node SPU Built-in Functions
10163 @subsection SPU Built-in Functions
10165 GCC provides extensions for the SPU processor as described in the
10166 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
10167 found at @uref{http://cell.scei.co.jp/} or
10168 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
10169 implementation differs in several ways.
10174 The optional extension of specifying vector constants in parentheses is
10178 A vector initializer requires no cast if the vector constant is of the
10179 same type as the variable it is initializing.
10182 If @code{signed} or @code{unsigned} is omitted, the signedness of the
10183 vector type is the default signedness of the base type. The default
10184 varies depending on the operating system, so a portable program should
10185 always specify the signedness.
10188 By default, the keyword @code{__vector} is added. The macro
10189 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
10193 GCC allows using a @code{typedef} name as the type specifier for a
10197 For C, overloaded functions are implemented with macros so the following
10201 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
10204 Since @code{spu_add} is a macro, the vector constant in the example
10205 is treated as four separate arguments. Wrap the entire argument in
10206 parentheses for this to work.
10209 The extended version of @code{__builtin_expect} is not supported.
10213 @emph{Note:} Only the interface described in the aforementioned
10214 specification is supported. Internally, GCC uses built-in functions to
10215 implement the required functionality, but these are not supported and
10216 are subject to change without notice.
10218 @node Target Format Checks
10219 @section Format Checks Specific to Particular Target Machines
10221 For some target machines, GCC supports additional options to the
10223 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
10226 * Solaris Format Checks::
10229 @node Solaris Format Checks
10230 @subsection Solaris Format Checks
10232 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
10233 check. @code{cmn_err} accepts a subset of the standard @code{printf}
10234 conversions, and the two-argument @code{%b} conversion for displaying
10235 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
10238 @section Pragmas Accepted by GCC
10242 GCC supports several types of pragmas, primarily in order to compile
10243 code originally written for other compilers. Note that in general
10244 we do not recommend the use of pragmas; @xref{Function Attributes},
10245 for further explanation.
10250 * RS/6000 and PowerPC Pragmas::
10252 * Solaris Pragmas::
10253 * Symbol-Renaming Pragmas::
10254 * Structure-Packing Pragmas::
10256 * Diagnostic Pragmas::
10257 * Visibility Pragmas::
10261 @subsection ARM Pragmas
10263 The ARM target defines pragmas for controlling the default addition of
10264 @code{long_call} and @code{short_call} attributes to functions.
10265 @xref{Function Attributes}, for information about the effects of these
10270 @cindex pragma, long_calls
10271 Set all subsequent functions to have the @code{long_call} attribute.
10273 @item no_long_calls
10274 @cindex pragma, no_long_calls
10275 Set all subsequent functions to have the @code{short_call} attribute.
10277 @item long_calls_off
10278 @cindex pragma, long_calls_off
10279 Do not affect the @code{long_call} or @code{short_call} attributes of
10280 subsequent functions.
10284 @subsection M32C Pragmas
10287 @item memregs @var{number}
10288 @cindex pragma, memregs
10289 Overrides the command line option @code{-memregs=} for the current
10290 file. Use with care! This pragma must be before any function in the
10291 file, and mixing different memregs values in different objects may
10292 make them incompatible. This pragma is useful when a
10293 performance-critical function uses a memreg for temporary values,
10294 as it may allow you to reduce the number of memregs used.
10298 @node RS/6000 and PowerPC Pragmas
10299 @subsection RS/6000 and PowerPC Pragmas
10301 The RS/6000 and PowerPC targets define one pragma for controlling
10302 whether or not the @code{longcall} attribute is added to function
10303 declarations by default. This pragma overrides the @option{-mlongcall}
10304 option, but not the @code{longcall} and @code{shortcall} attributes.
10305 @xref{RS/6000 and PowerPC Options}, for more information about when long
10306 calls are and are not necessary.
10310 @cindex pragma, longcall
10311 Apply the @code{longcall} attribute to all subsequent function
10315 Do not apply the @code{longcall} attribute to subsequent function
10319 @c Describe c4x pragmas here.
10320 @c Describe h8300 pragmas here.
10321 @c Describe sh pragmas here.
10322 @c Describe v850 pragmas here.
10324 @node Darwin Pragmas
10325 @subsection Darwin Pragmas
10327 The following pragmas are available for all architectures running the
10328 Darwin operating system. These are useful for compatibility with other
10332 @item mark @var{tokens}@dots{}
10333 @cindex pragma, mark
10334 This pragma is accepted, but has no effect.
10336 @item options align=@var{alignment}
10337 @cindex pragma, options align
10338 This pragma sets the alignment of fields in structures. The values of
10339 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
10340 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
10341 properly; to restore the previous setting, use @code{reset} for the
10344 @item segment @var{tokens}@dots{}
10345 @cindex pragma, segment
10346 This pragma is accepted, but has no effect.
10348 @item unused (@var{var} [, @var{var}]@dots{})
10349 @cindex pragma, unused
10350 This pragma declares variables to be possibly unused. GCC will not
10351 produce warnings for the listed variables. The effect is similar to
10352 that of the @code{unused} attribute, except that this pragma may appear
10353 anywhere within the variables' scopes.
10356 @node Solaris Pragmas
10357 @subsection Solaris Pragmas
10359 The Solaris target supports @code{#pragma redefine_extname}
10360 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
10361 @code{#pragma} directives for compatibility with the system compiler.
10364 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
10365 @cindex pragma, align
10367 Increase the minimum alignment of each @var{variable} to @var{alignment}.
10368 This is the same as GCC's @code{aligned} attribute @pxref{Variable
10369 Attributes}). Macro expansion occurs on the arguments to this pragma
10370 when compiling C and Objective-C. It does not currently occur when
10371 compiling C++, but this is a bug which may be fixed in a future
10374 @item fini (@var{function} [, @var{function}]...)
10375 @cindex pragma, fini
10377 This pragma causes each listed @var{function} to be called after
10378 main, or during shared module unloading, by adding a call to the
10379 @code{.fini} section.
10381 @item init (@var{function} [, @var{function}]...)
10382 @cindex pragma, init
10384 This pragma causes each listed @var{function} to be called during
10385 initialization (before @code{main}) or during shared module loading, by
10386 adding a call to the @code{.init} section.
10390 @node Symbol-Renaming Pragmas
10391 @subsection Symbol-Renaming Pragmas
10393 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
10394 supports two @code{#pragma} directives which change the name used in
10395 assembly for a given declaration. These pragmas are only available on
10396 platforms whose system headers need them. To get this effect on all
10397 platforms supported by GCC, use the asm labels extension (@pxref{Asm
10401 @item redefine_extname @var{oldname} @var{newname}
10402 @cindex pragma, redefine_extname
10404 This pragma gives the C function @var{oldname} the assembly symbol
10405 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
10406 will be defined if this pragma is available (currently only on
10409 @item extern_prefix @var{string}
10410 @cindex pragma, extern_prefix
10412 This pragma causes all subsequent external function and variable
10413 declarations to have @var{string} prepended to their assembly symbols.
10414 This effect may be terminated with another @code{extern_prefix} pragma
10415 whose argument is an empty string. The preprocessor macro
10416 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10417 available (currently only on Tru64 UNIX)@.
10420 These pragmas and the asm labels extension interact in a complicated
10421 manner. Here are some corner cases you may want to be aware of.
10424 @item Both pragmas silently apply only to declarations with external
10425 linkage. Asm labels do not have this restriction.
10427 @item In C++, both pragmas silently apply only to declarations with
10428 ``C'' linkage. Again, asm labels do not have this restriction.
10430 @item If any of the three ways of changing the assembly name of a
10431 declaration is applied to a declaration whose assembly name has
10432 already been determined (either by a previous use of one of these
10433 features, or because the compiler needed the assembly name in order to
10434 generate code), and the new name is different, a warning issues and
10435 the name does not change.
10437 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10438 always the C-language name.
10440 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10441 occurs with an asm label attached, the prefix is silently ignored for
10444 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10445 apply to the same declaration, whichever triggered first wins, and a
10446 warning issues if they contradict each other. (We would like to have
10447 @code{#pragma redefine_extname} always win, for consistency with asm
10448 labels, but if @code{#pragma extern_prefix} triggers first we have no
10449 way of knowing that that happened.)
10452 @node Structure-Packing Pragmas
10453 @subsection Structure-Packing Pragmas
10455 For compatibility with Win32, GCC supports a set of @code{#pragma}
10456 directives which change the maximum alignment of members of structures
10457 (other than zero-width bitfields), unions, and classes subsequently
10458 defined. The @var{n} value below always is required to be a small power
10459 of two and specifies the new alignment in bytes.
10462 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10463 @item @code{#pragma pack()} sets the alignment to the one that was in
10464 effect when compilation started (see also command line option
10465 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10466 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10467 setting on an internal stack and then optionally sets the new alignment.
10468 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10469 saved at the top of the internal stack (and removes that stack entry).
10470 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10471 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10472 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10473 @code{#pragma pack(pop)}.
10476 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10477 @code{#pragma} which lays out a structure as the documented
10478 @code{__attribute__ ((ms_struct))}.
10480 @item @code{#pragma ms_struct on} turns on the layout for structures
10482 @item @code{#pragma ms_struct off} turns off the layout for structures
10484 @item @code{#pragma ms_struct reset} goes back to the default layout.
10488 @subsection Weak Pragmas
10490 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10491 directives for declaring symbols to be weak, and defining weak
10495 @item #pragma weak @var{symbol}
10496 @cindex pragma, weak
10497 This pragma declares @var{symbol} to be weak, as if the declaration
10498 had the attribute of the same name. The pragma may appear before
10499 or after the declaration of @var{symbol}, but must appear before
10500 either its first use or its definition. It is not an error for
10501 @var{symbol} to never be defined at all.
10503 @item #pragma weak @var{symbol1} = @var{symbol2}
10504 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10505 It is an error if @var{symbol2} is not defined in the current
10509 @node Diagnostic Pragmas
10510 @subsection Diagnostic Pragmas
10512 GCC allows the user to selectively enable or disable certain types of
10513 diagnostics, and change the kind of the diagnostic. For example, a
10514 project's policy might require that all sources compile with
10515 @option{-Werror} but certain files might have exceptions allowing
10516 specific types of warnings. Or, a project might selectively enable
10517 diagnostics and treat them as errors depending on which preprocessor
10518 macros are defined.
10521 @item #pragma GCC diagnostic @var{kind} @var{option}
10522 @cindex pragma, diagnostic
10524 Modifies the disposition of a diagnostic. Note that not all
10525 diagnostics are modifiable; at the moment only warnings (normally
10526 controlled by @samp{-W...}) can be controlled, and not all of them.
10527 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10528 are controllable and which option controls them.
10530 @var{kind} is @samp{error} to treat this diagnostic as an error,
10531 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10532 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10533 @var{option} is a double quoted string which matches the command line
10537 #pragma GCC diagnostic warning "-Wformat"
10538 #pragma GCC diagnostic error "-Wformat"
10539 #pragma GCC diagnostic ignored "-Wformat"
10542 Note that these pragmas override any command line options. Also,
10543 while it is syntactically valid to put these pragmas anywhere in your
10544 sources, the only supported location for them is before any data or
10545 functions are defined. Doing otherwise may result in unpredictable
10546 results depending on how the optimizer manages your sources. If the
10547 same option is listed multiple times, the last one specified is the
10548 one that is in effect. This pragma is not intended to be a general
10549 purpose replacement for command line options, but for implementing
10550 strict control over project policies.
10554 @node Visibility Pragmas
10555 @subsection Visibility Pragmas
10558 @item #pragma GCC visibility push(@var{visibility})
10559 @itemx #pragma GCC visibility pop
10560 @cindex pragma, visibility
10562 This pragma allows the user to set the visibility for multiple
10563 declarations without having to give each a visibility attribute
10564 @xref{Function Attributes}, for more information about visibility and
10565 the attribute syntax.
10567 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10568 declarations. Class members and template specializations are not
10569 affected; if you want to override the visibility for a particular
10570 member or instantiation, you must use an attribute.
10574 @node Unnamed Fields
10575 @section Unnamed struct/union fields within structs/unions
10579 For compatibility with other compilers, GCC allows you to define
10580 a structure or union that contains, as fields, structures and unions
10581 without names. For example:
10594 In this example, the user would be able to access members of the unnamed
10595 union with code like @samp{foo.b}. Note that only unnamed structs and
10596 unions are allowed, you may not have, for example, an unnamed
10599 You must never create such structures that cause ambiguous field definitions.
10600 For example, this structure:
10611 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10612 Such constructs are not supported and must be avoided. In the future,
10613 such constructs may be detected and treated as compilation errors.
10615 @opindex fms-extensions
10616 Unless @option{-fms-extensions} is used, the unnamed field must be a
10617 structure or union definition without a tag (for example, @samp{struct
10618 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10619 also be a definition with a tag such as @samp{struct foo @{ int a;
10620 @};}, a reference to a previously defined structure or union such as
10621 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10622 previously defined structure or union type.
10625 @section Thread-Local Storage
10626 @cindex Thread-Local Storage
10627 @cindex @acronym{TLS}
10630 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10631 are allocated such that there is one instance of the variable per extant
10632 thread. The run-time model GCC uses to implement this originates
10633 in the IA-64 processor-specific ABI, but has since been migrated
10634 to other processors as well. It requires significant support from
10635 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10636 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10637 is not available everywhere.
10639 At the user level, the extension is visible with a new storage
10640 class keyword: @code{__thread}. For example:
10644 extern __thread struct state s;
10645 static __thread char *p;
10648 The @code{__thread} specifier may be used alone, with the @code{extern}
10649 or @code{static} specifiers, but with no other storage class specifier.
10650 When used with @code{extern} or @code{static}, @code{__thread} must appear
10651 immediately after the other storage class specifier.
10653 The @code{__thread} specifier may be applied to any global, file-scoped
10654 static, function-scoped static, or static data member of a class. It may
10655 not be applied to block-scoped automatic or non-static data member.
10657 When the address-of operator is applied to a thread-local variable, it is
10658 evaluated at run-time and returns the address of the current thread's
10659 instance of that variable. An address so obtained may be used by any
10660 thread. When a thread terminates, any pointers to thread-local variables
10661 in that thread become invalid.
10663 No static initialization may refer to the address of a thread-local variable.
10665 In C++, if an initializer is present for a thread-local variable, it must
10666 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10669 See @uref{http://people.redhat.com/drepper/tls.pdf,
10670 ELF Handling For Thread-Local Storage} for a detailed explanation of
10671 the four thread-local storage addressing models, and how the run-time
10672 is expected to function.
10675 * C99 Thread-Local Edits::
10676 * C++98 Thread-Local Edits::
10679 @node C99 Thread-Local Edits
10680 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10682 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10683 that document the exact semantics of the language extension.
10687 @cite{5.1.2 Execution environments}
10689 Add new text after paragraph 1
10692 Within either execution environment, a @dfn{thread} is a flow of
10693 control within a program. It is implementation defined whether
10694 or not there may be more than one thread associated with a program.
10695 It is implementation defined how threads beyond the first are
10696 created, the name and type of the function called at thread
10697 startup, and how threads may be terminated. However, objects
10698 with thread storage duration shall be initialized before thread
10703 @cite{6.2.4 Storage durations of objects}
10705 Add new text before paragraph 3
10708 An object whose identifier is declared with the storage-class
10709 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10710 Its lifetime is the entire execution of the thread, and its
10711 stored value is initialized only once, prior to thread startup.
10715 @cite{6.4.1 Keywords}
10717 Add @code{__thread}.
10720 @cite{6.7.1 Storage-class specifiers}
10722 Add @code{__thread} to the list of storage class specifiers in
10725 Change paragraph 2 to
10728 With the exception of @code{__thread}, at most one storage-class
10729 specifier may be given [@dots{}]. The @code{__thread} specifier may
10730 be used alone, or immediately following @code{extern} or
10734 Add new text after paragraph 6
10737 The declaration of an identifier for a variable that has
10738 block scope that specifies @code{__thread} shall also
10739 specify either @code{extern} or @code{static}.
10741 The @code{__thread} specifier shall be used only with
10746 @node C++98 Thread-Local Edits
10747 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10749 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10750 that document the exact semantics of the language extension.
10754 @b{[intro.execution]}
10756 New text after paragraph 4
10759 A @dfn{thread} is a flow of control within the abstract machine.
10760 It is implementation defined whether or not there may be more than
10764 New text after paragraph 7
10767 It is unspecified whether additional action must be taken to
10768 ensure when and whether side effects are visible to other threads.
10774 Add @code{__thread}.
10777 @b{[basic.start.main]}
10779 Add after paragraph 5
10782 The thread that begins execution at the @code{main} function is called
10783 the @dfn{main thread}. It is implementation defined how functions
10784 beginning threads other than the main thread are designated or typed.
10785 A function so designated, as well as the @code{main} function, is called
10786 a @dfn{thread startup function}. It is implementation defined what
10787 happens if a thread startup function returns. It is implementation
10788 defined what happens to other threads when any thread calls @code{exit}.
10792 @b{[basic.start.init]}
10794 Add after paragraph 4
10797 The storage for an object of thread storage duration shall be
10798 statically initialized before the first statement of the thread startup
10799 function. An object of thread storage duration shall not require
10800 dynamic initialization.
10804 @b{[basic.start.term]}
10806 Add after paragraph 3
10809 The type of an object with thread storage duration shall not have a
10810 non-trivial destructor, nor shall it be an array type whose elements
10811 (directly or indirectly) have non-trivial destructors.
10817 Add ``thread storage duration'' to the list in paragraph 1.
10822 Thread, static, and automatic storage durations are associated with
10823 objects introduced by declarations [@dots{}].
10826 Add @code{__thread} to the list of specifiers in paragraph 3.
10829 @b{[basic.stc.thread]}
10831 New section before @b{[basic.stc.static]}
10834 The keyword @code{__thread} applied to a non-local object gives the
10835 object thread storage duration.
10837 A local variable or class data member declared both @code{static}
10838 and @code{__thread} gives the variable or member thread storage
10843 @b{[basic.stc.static]}
10848 All objects which have neither thread storage duration, dynamic
10849 storage duration nor are local [@dots{}].
10855 Add @code{__thread} to the list in paragraph 1.
10860 With the exception of @code{__thread}, at most one
10861 @var{storage-class-specifier} shall appear in a given
10862 @var{decl-specifier-seq}. The @code{__thread} specifier may
10863 be used alone, or immediately following the @code{extern} or
10864 @code{static} specifiers. [@dots{}]
10867 Add after paragraph 5
10870 The @code{__thread} specifier can be applied only to the names of objects
10871 and to anonymous unions.
10877 Add after paragraph 6
10880 Non-@code{static} members shall not be @code{__thread}.
10884 @node Binary constants
10885 @section Binary constants using the @samp{0b} prefix
10886 @cindex Binary constants using the @samp{0b} prefix
10888 Integer constants can be written as binary constants, consisting of a
10889 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
10890 @samp{0B}. This is particularly useful in environments that operate a
10891 lot on the bit-level (like microcontrollers).
10893 The following statements are identical:
10902 The type of these constants follows the same rules as for octal or
10903 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
10906 @node C++ Extensions
10907 @chapter Extensions to the C++ Language
10908 @cindex extensions, C++ language
10909 @cindex C++ language extensions
10911 The GNU compiler provides these extensions to the C++ language (and you
10912 can also use most of the C language extensions in your C++ programs). If you
10913 want to write code that checks whether these features are available, you can
10914 test for the GNU compiler the same way as for C programs: check for a
10915 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10916 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10917 Predefined Macros,cpp,The GNU C Preprocessor}).
10920 * Volatiles:: What constitutes an access to a volatile object.
10921 * Restricted Pointers:: C99 restricted pointers and references.
10922 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10923 * C++ Interface:: You can use a single C++ header file for both
10924 declarations and definitions.
10925 * Template Instantiation:: Methods for ensuring that exactly one copy of
10926 each needed template instantiation is emitted.
10927 * Bound member functions:: You can extract a function pointer to the
10928 method denoted by a @samp{->*} or @samp{.*} expression.
10929 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10930 * Namespace Association:: Strong using-directives for namespace association.
10931 * Type Traits:: Compiler support for type traits
10932 * Java Exceptions:: Tweaking exception handling to work with Java.
10933 * Deprecated Features:: Things will disappear from g++.
10934 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10938 @section When is a Volatile Object Accessed?
10939 @cindex accessing volatiles
10940 @cindex volatile read
10941 @cindex volatile write
10942 @cindex volatile access
10944 Both the C and C++ standard have the concept of volatile objects. These
10945 are normally accessed by pointers and used for accessing hardware. The
10946 standards encourage compilers to refrain from optimizations concerning
10947 accesses to volatile objects. The C standard leaves it implementation
10948 defined as to what constitutes a volatile access. The C++ standard omits
10949 to specify this, except to say that C++ should behave in a similar manner
10950 to C with respect to volatiles, where possible. The minimum either
10951 standard specifies is that at a sequence point all previous accesses to
10952 volatile objects have stabilized and no subsequent accesses have
10953 occurred. Thus an implementation is free to reorder and combine
10954 volatile accesses which occur between sequence points, but cannot do so
10955 for accesses across a sequence point. The use of volatiles does not
10956 allow you to violate the restriction on updating objects multiple times
10957 within a sequence point.
10959 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10961 The behavior differs slightly between C and C++ in the non-obvious cases:
10964 volatile int *src = @var{somevalue};
10968 With C, such expressions are rvalues, and GCC interprets this either as a
10969 read of the volatile object being pointed to or only as request to evaluate
10970 the side-effects. The C++ standard specifies that such expressions do not
10971 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10972 object may be incomplete. The C++ standard does not specify explicitly
10973 that it is this lvalue to rvalue conversion which may be responsible for
10974 causing an access. However, there is reason to believe that it is,
10975 because otherwise certain simple expressions become undefined. However,
10976 because it would surprise most programmers, G++ treats dereferencing a
10977 pointer to volatile object of complete type when the value is unused as
10978 GCC would do for an equivalent type in C. When the object has incomplete
10979 type, G++ issues a warning; if you wish to force an error, you must
10980 force a conversion to rvalue with, for instance, a static cast.
10982 When using a reference to volatile, G++ does not treat equivalent
10983 expressions as accesses to volatiles, but instead issues a warning that
10984 no volatile is accessed. The rationale for this is that otherwise it
10985 becomes difficult to determine where volatile access occur, and not
10986 possible to ignore the return value from functions returning volatile
10987 references. Again, if you wish to force a read, cast the reference to
10990 @node Restricted Pointers
10991 @section Restricting Pointer Aliasing
10992 @cindex restricted pointers
10993 @cindex restricted references
10994 @cindex restricted this pointer
10996 As with the C front end, G++ understands the C99 feature of restricted pointers,
10997 specified with the @code{__restrict__}, or @code{__restrict} type
10998 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10999 language flag, @code{restrict} is not a keyword in C++.
11001 In addition to allowing restricted pointers, you can specify restricted
11002 references, which indicate that the reference is not aliased in the local
11006 void fn (int *__restrict__ rptr, int &__restrict__ rref)
11013 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
11014 @var{rref} refers to a (different) unaliased integer.
11016 You may also specify whether a member function's @var{this} pointer is
11017 unaliased by using @code{__restrict__} as a member function qualifier.
11020 void T::fn () __restrict__
11027 Within the body of @code{T::fn}, @var{this} will have the effective
11028 definition @code{T *__restrict__ const this}. Notice that the
11029 interpretation of a @code{__restrict__} member function qualifier is
11030 different to that of @code{const} or @code{volatile} qualifier, in that it
11031 is applied to the pointer rather than the object. This is consistent with
11032 other compilers which implement restricted pointers.
11034 As with all outermost parameter qualifiers, @code{__restrict__} is
11035 ignored in function definition matching. This means you only need to
11036 specify @code{__restrict__} in a function definition, rather than
11037 in a function prototype as well.
11039 @node Vague Linkage
11040 @section Vague Linkage
11041 @cindex vague linkage
11043 There are several constructs in C++ which require space in the object
11044 file but are not clearly tied to a single translation unit. We say that
11045 these constructs have ``vague linkage''. Typically such constructs are
11046 emitted wherever they are needed, though sometimes we can be more
11050 @item Inline Functions
11051 Inline functions are typically defined in a header file which can be
11052 included in many different compilations. Hopefully they can usually be
11053 inlined, but sometimes an out-of-line copy is necessary, if the address
11054 of the function is taken or if inlining fails. In general, we emit an
11055 out-of-line copy in all translation units where one is needed. As an
11056 exception, we only emit inline virtual functions with the vtable, since
11057 it will always require a copy.
11059 Local static variables and string constants used in an inline function
11060 are also considered to have vague linkage, since they must be shared
11061 between all inlined and out-of-line instances of the function.
11065 C++ virtual functions are implemented in most compilers using a lookup
11066 table, known as a vtable. The vtable contains pointers to the virtual
11067 functions provided by a class, and each object of the class contains a
11068 pointer to its vtable (or vtables, in some multiple-inheritance
11069 situations). If the class declares any non-inline, non-pure virtual
11070 functions, the first one is chosen as the ``key method'' for the class,
11071 and the vtable is only emitted in the translation unit where the key
11074 @emph{Note:} If the chosen key method is later defined as inline, the
11075 vtable will still be emitted in every translation unit which defines it.
11076 Make sure that any inline virtuals are declared inline in the class
11077 body, even if they are not defined there.
11079 @item type_info objects
11082 C++ requires information about types to be written out in order to
11083 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
11084 For polymorphic classes (classes with virtual functions), the type_info
11085 object is written out along with the vtable so that @samp{dynamic_cast}
11086 can determine the dynamic type of a class object at runtime. For all
11087 other types, we write out the type_info object when it is used: when
11088 applying @samp{typeid} to an expression, throwing an object, or
11089 referring to a type in a catch clause or exception specification.
11091 @item Template Instantiations
11092 Most everything in this section also applies to template instantiations,
11093 but there are other options as well.
11094 @xref{Template Instantiation,,Where's the Template?}.
11098 When used with GNU ld version 2.8 or later on an ELF system such as
11099 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
11100 these constructs will be discarded at link time. This is known as
11103 On targets that don't support COMDAT, but do support weak symbols, GCC
11104 will use them. This way one copy will override all the others, but
11105 the unused copies will still take up space in the executable.
11107 For targets which do not support either COMDAT or weak symbols,
11108 most entities with vague linkage will be emitted as local symbols to
11109 avoid duplicate definition errors from the linker. This will not happen
11110 for local statics in inlines, however, as having multiple copies will
11111 almost certainly break things.
11113 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
11114 another way to control placement of these constructs.
11116 @node C++ Interface
11117 @section #pragma interface and implementation
11119 @cindex interface and implementation headers, C++
11120 @cindex C++ interface and implementation headers
11121 @cindex pragmas, interface and implementation
11123 @code{#pragma interface} and @code{#pragma implementation} provide the
11124 user with a way of explicitly directing the compiler to emit entities
11125 with vague linkage (and debugging information) in a particular
11128 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
11129 most cases, because of COMDAT support and the ``key method'' heuristic
11130 mentioned in @ref{Vague Linkage}. Using them can actually cause your
11131 program to grow due to unnecessary out-of-line copies of inline
11132 functions. Currently (3.4) the only benefit of these
11133 @code{#pragma}s is reduced duplication of debugging information, and
11134 that should be addressed soon on DWARF 2 targets with the use of
11138 @item #pragma interface
11139 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
11140 @kindex #pragma interface
11141 Use this directive in @emph{header files} that define object classes, to save
11142 space in most of the object files that use those classes. Normally,
11143 local copies of certain information (backup copies of inline member
11144 functions, debugging information, and the internal tables that implement
11145 virtual functions) must be kept in each object file that includes class
11146 definitions. You can use this pragma to avoid such duplication. When a
11147 header file containing @samp{#pragma interface} is included in a
11148 compilation, this auxiliary information will not be generated (unless
11149 the main input source file itself uses @samp{#pragma implementation}).
11150 Instead, the object files will contain references to be resolved at link
11153 The second form of this directive is useful for the case where you have
11154 multiple headers with the same name in different directories. If you
11155 use this form, you must specify the same string to @samp{#pragma
11158 @item #pragma implementation
11159 @itemx #pragma implementation "@var{objects}.h"
11160 @kindex #pragma implementation
11161 Use this pragma in a @emph{main input file}, when you want full output from
11162 included header files to be generated (and made globally visible). The
11163 included header file, in turn, should use @samp{#pragma interface}.
11164 Backup copies of inline member functions, debugging information, and the
11165 internal tables used to implement virtual functions are all generated in
11166 implementation files.
11168 @cindex implied @code{#pragma implementation}
11169 @cindex @code{#pragma implementation}, implied
11170 @cindex naming convention, implementation headers
11171 If you use @samp{#pragma implementation} with no argument, it applies to
11172 an include file with the same basename@footnote{A file's @dfn{basename}
11173 was the name stripped of all leading path information and of trailing
11174 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
11175 file. For example, in @file{allclass.cc}, giving just
11176 @samp{#pragma implementation}
11177 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
11179 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
11180 an implementation file whenever you would include it from
11181 @file{allclass.cc} even if you never specified @samp{#pragma
11182 implementation}. This was deemed to be more trouble than it was worth,
11183 however, and disabled.
11185 Use the string argument if you want a single implementation file to
11186 include code from multiple header files. (You must also use
11187 @samp{#include} to include the header file; @samp{#pragma
11188 implementation} only specifies how to use the file---it doesn't actually
11191 There is no way to split up the contents of a single header file into
11192 multiple implementation files.
11195 @cindex inlining and C++ pragmas
11196 @cindex C++ pragmas, effect on inlining
11197 @cindex pragmas in C++, effect on inlining
11198 @samp{#pragma implementation} and @samp{#pragma interface} also have an
11199 effect on function inlining.
11201 If you define a class in a header file marked with @samp{#pragma
11202 interface}, the effect on an inline function defined in that class is
11203 similar to an explicit @code{extern} declaration---the compiler emits
11204 no code at all to define an independent version of the function. Its
11205 definition is used only for inlining with its callers.
11207 @opindex fno-implement-inlines
11208 Conversely, when you include the same header file in a main source file
11209 that declares it as @samp{#pragma implementation}, the compiler emits
11210 code for the function itself; this defines a version of the function
11211 that can be found via pointers (or by callers compiled without
11212 inlining). If all calls to the function can be inlined, you can avoid
11213 emitting the function by compiling with @option{-fno-implement-inlines}.
11214 If any calls were not inlined, you will get linker errors.
11216 @node Template Instantiation
11217 @section Where's the Template?
11218 @cindex template instantiation
11220 C++ templates are the first language feature to require more
11221 intelligence from the environment than one usually finds on a UNIX
11222 system. Somehow the compiler and linker have to make sure that each
11223 template instance occurs exactly once in the executable if it is needed,
11224 and not at all otherwise. There are two basic approaches to this
11225 problem, which are referred to as the Borland model and the Cfront model.
11228 @item Borland model
11229 Borland C++ solved the template instantiation problem by adding the code
11230 equivalent of common blocks to their linker; the compiler emits template
11231 instances in each translation unit that uses them, and the linker
11232 collapses them together. The advantage of this model is that the linker
11233 only has to consider the object files themselves; there is no external
11234 complexity to worry about. This disadvantage is that compilation time
11235 is increased because the template code is being compiled repeatedly.
11236 Code written for this model tends to include definitions of all
11237 templates in the header file, since they must be seen to be
11241 The AT&T C++ translator, Cfront, solved the template instantiation
11242 problem by creating the notion of a template repository, an
11243 automatically maintained place where template instances are stored. A
11244 more modern version of the repository works as follows: As individual
11245 object files are built, the compiler places any template definitions and
11246 instantiations encountered in the repository. At link time, the link
11247 wrapper adds in the objects in the repository and compiles any needed
11248 instances that were not previously emitted. The advantages of this
11249 model are more optimal compilation speed and the ability to use the
11250 system linker; to implement the Borland model a compiler vendor also
11251 needs to replace the linker. The disadvantages are vastly increased
11252 complexity, and thus potential for error; for some code this can be
11253 just as transparent, but in practice it can been very difficult to build
11254 multiple programs in one directory and one program in multiple
11255 directories. Code written for this model tends to separate definitions
11256 of non-inline member templates into a separate file, which should be
11257 compiled separately.
11260 When used with GNU ld version 2.8 or later on an ELF system such as
11261 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
11262 Borland model. On other systems, G++ implements neither automatic
11265 A future version of G++ will support a hybrid model whereby the compiler
11266 will emit any instantiations for which the template definition is
11267 included in the compile, and store template definitions and
11268 instantiation context information into the object file for the rest.
11269 The link wrapper will extract that information as necessary and invoke
11270 the compiler to produce the remaining instantiations. The linker will
11271 then combine duplicate instantiations.
11273 In the mean time, you have the following options for dealing with
11274 template instantiations:
11279 Compile your template-using code with @option{-frepo}. The compiler will
11280 generate files with the extension @samp{.rpo} listing all of the
11281 template instantiations used in the corresponding object files which
11282 could be instantiated there; the link wrapper, @samp{collect2}, will
11283 then update the @samp{.rpo} files to tell the compiler where to place
11284 those instantiations and rebuild any affected object files. The
11285 link-time overhead is negligible after the first pass, as the compiler
11286 will continue to place the instantiations in the same files.
11288 This is your best option for application code written for the Borland
11289 model, as it will just work. Code written for the Cfront model will
11290 need to be modified so that the template definitions are available at
11291 one or more points of instantiation; usually this is as simple as adding
11292 @code{#include <tmethods.cc>} to the end of each template header.
11294 For library code, if you want the library to provide all of the template
11295 instantiations it needs, just try to link all of its object files
11296 together; the link will fail, but cause the instantiations to be
11297 generated as a side effect. Be warned, however, that this may cause
11298 conflicts if multiple libraries try to provide the same instantiations.
11299 For greater control, use explicit instantiation as described in the next
11303 @opindex fno-implicit-templates
11304 Compile your code with @option{-fno-implicit-templates} to disable the
11305 implicit generation of template instances, and explicitly instantiate
11306 all the ones you use. This approach requires more knowledge of exactly
11307 which instances you need than do the others, but it's less
11308 mysterious and allows greater control. You can scatter the explicit
11309 instantiations throughout your program, perhaps putting them in the
11310 translation units where the instances are used or the translation units
11311 that define the templates themselves; you can put all of the explicit
11312 instantiations you need into one big file; or you can create small files
11319 template class Foo<int>;
11320 template ostream& operator <<
11321 (ostream&, const Foo<int>&);
11324 for each of the instances you need, and create a template instantiation
11325 library from those.
11327 If you are using Cfront-model code, you can probably get away with not
11328 using @option{-fno-implicit-templates} when compiling files that don't
11329 @samp{#include} the member template definitions.
11331 If you use one big file to do the instantiations, you may want to
11332 compile it without @option{-fno-implicit-templates} so you get all of the
11333 instances required by your explicit instantiations (but not by any
11334 other files) without having to specify them as well.
11336 G++ has extended the template instantiation syntax given in the ISO
11337 standard to allow forward declaration of explicit instantiations
11338 (with @code{extern}), instantiation of the compiler support data for a
11339 template class (i.e.@: the vtable) without instantiating any of its
11340 members (with @code{inline}), and instantiation of only the static data
11341 members of a template class, without the support data or member
11342 functions (with (@code{static}):
11345 extern template int max (int, int);
11346 inline template class Foo<int>;
11347 static template class Foo<int>;
11351 Do nothing. Pretend G++ does implement automatic instantiation
11352 management. Code written for the Borland model will work fine, but
11353 each translation unit will contain instances of each of the templates it
11354 uses. In a large program, this can lead to an unacceptable amount of code
11358 @node Bound member functions
11359 @section Extracting the function pointer from a bound pointer to member function
11361 @cindex pointer to member function
11362 @cindex bound pointer to member function
11364 In C++, pointer to member functions (PMFs) are implemented using a wide
11365 pointer of sorts to handle all the possible call mechanisms; the PMF
11366 needs to store information about how to adjust the @samp{this} pointer,
11367 and if the function pointed to is virtual, where to find the vtable, and
11368 where in the vtable to look for the member function. If you are using
11369 PMFs in an inner loop, you should really reconsider that decision. If
11370 that is not an option, you can extract the pointer to the function that
11371 would be called for a given object/PMF pair and call it directly inside
11372 the inner loop, to save a bit of time.
11374 Note that you will still be paying the penalty for the call through a
11375 function pointer; on most modern architectures, such a call defeats the
11376 branch prediction features of the CPU@. This is also true of normal
11377 virtual function calls.
11379 The syntax for this extension is
11383 extern int (A::*fp)();
11384 typedef int (*fptr)(A *);
11386 fptr p = (fptr)(a.*fp);
11389 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
11390 no object is needed to obtain the address of the function. They can be
11391 converted to function pointers directly:
11394 fptr p1 = (fptr)(&A::foo);
11397 @opindex Wno-pmf-conversions
11398 You must specify @option{-Wno-pmf-conversions} to use this extension.
11400 @node C++ Attributes
11401 @section C++-Specific Variable, Function, and Type Attributes
11403 Some attributes only make sense for C++ programs.
11406 @item init_priority (@var{priority})
11407 @cindex init_priority attribute
11410 In Standard C++, objects defined at namespace scope are guaranteed to be
11411 initialized in an order in strict accordance with that of their definitions
11412 @emph{in a given translation unit}. No guarantee is made for initializations
11413 across translation units. However, GNU C++ allows users to control the
11414 order of initialization of objects defined at namespace scope with the
11415 @code{init_priority} attribute by specifying a relative @var{priority},
11416 a constant integral expression currently bounded between 101 and 65535
11417 inclusive. Lower numbers indicate a higher priority.
11419 In the following example, @code{A} would normally be created before
11420 @code{B}, but the @code{init_priority} attribute has reversed that order:
11423 Some_Class A __attribute__ ((init_priority (2000)));
11424 Some_Class B __attribute__ ((init_priority (543)));
11428 Note that the particular values of @var{priority} do not matter; only their
11431 @item java_interface
11432 @cindex java_interface attribute
11434 This type attribute informs C++ that the class is a Java interface. It may
11435 only be applied to classes declared within an @code{extern "Java"} block.
11436 Calls to methods declared in this interface will be dispatched using GCJ's
11437 interface table mechanism, instead of regular virtual table dispatch.
11441 See also @xref{Namespace Association}.
11443 @node Namespace Association
11444 @section Namespace Association
11446 @strong{Caution:} The semantics of this extension are not fully
11447 defined. Users should refrain from using this extension as its
11448 semantics may change subtly over time. It is possible that this
11449 extension will be removed in future versions of G++.
11451 A using-directive with @code{__attribute ((strong))} is stronger
11452 than a normal using-directive in two ways:
11456 Templates from the used namespace can be specialized and explicitly
11457 instantiated as though they were members of the using namespace.
11460 The using namespace is considered an associated namespace of all
11461 templates in the used namespace for purposes of argument-dependent
11465 The used namespace must be nested within the using namespace so that
11466 normal unqualified lookup works properly.
11468 This is useful for composing a namespace transparently from
11469 implementation namespaces. For example:
11474 template <class T> struct A @{ @};
11476 using namespace debug __attribute ((__strong__));
11477 template <> struct A<int> @{ @}; // @r{ok to specialize}
11479 template <class T> void f (A<T>);
11484 f (std::A<float>()); // @r{lookup finds} std::f
11490 @section Type Traits
11492 The C++ front-end implements syntactic extensions that allow to
11493 determine at compile time various characteristics of a type (or of a
11497 @item __has_nothrow_assign (type)
11498 If @code{type} is const qualified or is a reference type then the trait is
11499 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
11500 is true, else if @code{type} is a cv class or union type with copy assignment
11501 operators that are known not to throw an exception then the trait is true,
11502 else it is false. Requires: @code{type} shall be a complete type, an array
11503 type of unknown bound, or is a @code{void} type.
11505 @item __has_nothrow_copy (type)
11506 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
11507 @code{type} is a cv class or union type with copy constructors that
11508 are known not to throw an exception then the trait is true, else it is false.
11509 Requires: @code{type} shall be a complete type, an array type of
11510 unknown bound, or is a @code{void} type.
11512 @item __has_nothrow_constructor (type)
11513 If @code{__has_trivial_constructor (type)} is true then the trait is
11514 true, else if @code{type} is a cv class or union type (or array
11515 thereof) with a default constructor that is known not to throw an
11516 exception then the trait is true, else it is false. Requires:
11517 @code{type} shall be a complete type, an array type of unknown bound,
11518 or is a @code{void} type.
11520 @item __has_trivial_assign (type)
11521 If @code{type} is const qualified or is a reference type then the trait is
11522 false. Otherwise if @code{__is_pod (type)} is true then the trait is
11523 true, else if @code{type} is a cv class or union type with a trivial
11524 copy assignment ([class.copy]) then the trait is true, else it is
11525 false. Requires: @code{type} shall be a complete type, an array type
11526 of unknown bound, or is a @code{void} type.
11528 @item __has_trivial_copy (type)
11529 If @code{__is_pod (type)} is true or @code{type} is a reference type
11530 then the trait is true, else if @code{type} is a cv class or union type
11531 with a trivial copy constructor ([class.copy]) then the trait
11532 is true, else it is false. Requires: @code{type} shall be a complete
11533 type, an array type of unknown bound, or is a @code{void} type.
11535 @item __has_trivial_constructor (type)
11536 If @code{__is_pod (type)} is true then the trait is true, else if
11537 @code{type} is a cv class or union type (or array thereof) with a
11538 trivial default constructor ([class.ctor]) then the trait is true,
11539 else it is false. Requires: @code{type} shall be a complete type, an
11540 array type of unknown bound, or is a @code{void} type.
11542 @item __has_trivial_destructor (type)
11543 If @code{__is_pod (type)} is true or @code{type} is a reference type then
11544 the trait is true, else if @code{type} is a cv class or union type (or
11545 array thereof) with a trivial destructor ([class.dtor]) then the trait
11546 is true, else it is false. Requires: @code{type} shall be a complete
11547 type, an array type of unknown bound, or is a @code{void} type.
11549 @item __has_virtual_destructor (type)
11550 If @code{type} is a class type with a virtual destructor
11551 ([class.dtor]) then the trait is true, else it is false. Requires:
11552 @code{type} shall be a complete type, an array type of unknown bound,
11553 or is a @code{void} type.
11555 @item __is_abstract (type)
11556 If @code{type} is an abstract class ([class.abstract]) then the trait
11557 is true, else it is false. Requires: @code{type} shall be a complete
11558 type, an array type of unknown bound, or is a @code{void} type.
11560 @item __is_base_of (base_type, derived_type)
11561 If @code{base_type} is a base class of @code{derived_type}
11562 ([class.derived]) then the trait is true, otherwise it is false.
11563 Top-level cv qualifications of @code{base_type} and
11564 @code{derived_type} are ignored. For the purposes of this trait, a
11565 class type is considered is own base. Requires: if @code{__is_class
11566 (base_type)} and @code{__is_class (derived_type)} are true and
11567 @code{base_type} and @code{derived_type} are not the same type
11568 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
11569 type. Diagnostic is produced if this requirement is not met.
11571 @item __is_class (type)
11572 If @code{type} is a cv class type, and not a union type
11573 ([basic.compound]) the the trait is true, else it is false.
11575 @item __is_empty (type)
11576 If @code{__is_class (type)} is false then the trait is false.
11577 Otherwise @code{type} is considered empty if and only if: @code{type}
11578 has no non-static data members, or all non-static data members, if
11579 any, are bit-fields of lenght 0, and @code{type} has no virtual
11580 members, and @code{type} has no virtual base classes, and @code{type}
11581 has no base classes @code{base_type} for which
11582 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
11583 be a complete type, an array type of unknown bound, or is a
11586 @item __is_enum (type)
11587 If @code{type} is a cv enumeration type ([basic.compound]) the the trait is
11588 true, else it is false.
11590 @item __is_pod (type)
11591 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
11592 else it is false. Requires: @code{type} shall be a complete type,
11593 an array type of unknown bound, or is a @code{void} type.
11595 @item __is_polymorphic (type)
11596 If @code{type} is a polymorphic class ([class.virtual]) then the trait
11597 is true, else it is false. Requires: @code{type} shall be a complete
11598 type, an array type of unknown bound, or is a @code{void} type.
11600 @item __is_union (type)
11601 If @code{type} is a cv union type ([basic.compound]) the the trait is
11602 true, else it is false.
11606 @node Java Exceptions
11607 @section Java Exceptions
11609 The Java language uses a slightly different exception handling model
11610 from C++. Normally, GNU C++ will automatically detect when you are
11611 writing C++ code that uses Java exceptions, and handle them
11612 appropriately. However, if C++ code only needs to execute destructors
11613 when Java exceptions are thrown through it, GCC will guess incorrectly.
11614 Sample problematic code is:
11617 struct S @{ ~S(); @};
11618 extern void bar(); // @r{is written in Java, and may throw exceptions}
11627 The usual effect of an incorrect guess is a link failure, complaining of
11628 a missing routine called @samp{__gxx_personality_v0}.
11630 You can inform the compiler that Java exceptions are to be used in a
11631 translation unit, irrespective of what it might think, by writing
11632 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11633 @samp{#pragma} must appear before any functions that throw or catch
11634 exceptions, or run destructors when exceptions are thrown through them.
11636 You cannot mix Java and C++ exceptions in the same translation unit. It
11637 is believed to be safe to throw a C++ exception from one file through
11638 another file compiled for the Java exception model, or vice versa, but
11639 there may be bugs in this area.
11641 @node Deprecated Features
11642 @section Deprecated Features
11644 In the past, the GNU C++ compiler was extended to experiment with new
11645 features, at a time when the C++ language was still evolving. Now that
11646 the C++ standard is complete, some of those features are superseded by
11647 superior alternatives. Using the old features might cause a warning in
11648 some cases that the feature will be dropped in the future. In other
11649 cases, the feature might be gone already.
11651 While the list below is not exhaustive, it documents some of the options
11652 that are now deprecated:
11655 @item -fexternal-templates
11656 @itemx -falt-external-templates
11657 These are two of the many ways for G++ to implement template
11658 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11659 defines how template definitions have to be organized across
11660 implementation units. G++ has an implicit instantiation mechanism that
11661 should work just fine for standard-conforming code.
11663 @item -fstrict-prototype
11664 @itemx -fno-strict-prototype
11665 Previously it was possible to use an empty prototype parameter list to
11666 indicate an unspecified number of parameters (like C), rather than no
11667 parameters, as C++ demands. This feature has been removed, except where
11668 it is required for backwards compatibility @xref{Backwards Compatibility}.
11671 G++ allows a virtual function returning @samp{void *} to be overridden
11672 by one returning a different pointer type. This extension to the
11673 covariant return type rules is now deprecated and will be removed from a
11676 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11677 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11678 and will be removed in a future version. Code using these operators
11679 should be modified to use @code{std::min} and @code{std::max} instead.
11681 The named return value extension has been deprecated, and is now
11684 The use of initializer lists with new expressions has been deprecated,
11685 and is now removed from G++.
11687 Floating and complex non-type template parameters have been deprecated,
11688 and are now removed from G++.
11690 The implicit typename extension has been deprecated and is now
11693 The use of default arguments in function pointers, function typedefs
11694 and other places where they are not permitted by the standard is
11695 deprecated and will be removed from a future version of G++.
11697 G++ allows floating-point literals to appear in integral constant expressions,
11698 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11699 This extension is deprecated and will be removed from a future version.
11701 G++ allows static data members of const floating-point type to be declared
11702 with an initializer in a class definition. The standard only allows
11703 initializers for static members of const integral types and const
11704 enumeration types so this extension has been deprecated and will be removed
11705 from a future version.
11707 @node Backwards Compatibility
11708 @section Backwards Compatibility
11709 @cindex Backwards Compatibility
11710 @cindex ARM [Annotated C++ Reference Manual]
11712 Now that there is a definitive ISO standard C++, G++ has a specification
11713 to adhere to. The C++ language evolved over time, and features that
11714 used to be acceptable in previous drafts of the standard, such as the ARM
11715 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11716 compilation of C++ written to such drafts, G++ contains some backwards
11717 compatibilities. @emph{All such backwards compatibility features are
11718 liable to disappear in future versions of G++.} They should be considered
11719 deprecated @xref{Deprecated Features}.
11723 If a variable is declared at for scope, it used to remain in scope until
11724 the end of the scope which contained the for statement (rather than just
11725 within the for scope). G++ retains this, but issues a warning, if such a
11726 variable is accessed outside the for scope.
11728 @item Implicit C language
11729 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11730 scope to set the language. On such systems, all header files are
11731 implicitly scoped inside a C language scope. Also, an empty prototype
11732 @code{()} will be treated as an unspecified number of arguments, rather
11733 than no arguments, as C++ demands.