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.
87 @section Statements and Declarations in Expressions
88 @cindex statements inside expressions
89 @cindex declarations inside expressions
90 @cindex expressions containing statements
91 @cindex macros, statements in expressions
93 @c the above section title wrapped and causes an underfull hbox.. i
94 @c changed it from "within" to "in". --mew 4feb93
95 A compound statement enclosed in parentheses may appear as an expression
96 in GNU C@. This allows you to use loops, switches, and local variables
99 Recall that a compound statement is a sequence of statements surrounded
100 by braces; in this construct, parentheses go around the braces. For
104 (@{ int y = foo (); int z;
111 is a valid (though slightly more complex than necessary) expression
112 for the absolute value of @code{foo ()}.
114 The last thing in the compound statement should be an expression
115 followed by a semicolon; the value of this subexpression serves as the
116 value of the entire construct. (If you use some other kind of statement
117 last within the braces, the construct has type @code{void}, and thus
118 effectively no value.)
120 This feature is especially useful in making macro definitions ``safe'' (so
121 that they evaluate each operand exactly once). For example, the
122 ``maximum'' function is commonly defined as a macro in standard C as
126 #define max(a,b) ((a) > (b) ? (a) : (b))
130 @cindex side effects, macro argument
131 But this definition computes either @var{a} or @var{b} twice, with bad
132 results if the operand has side effects. In GNU C, if you know the
133 type of the operands (here taken as @code{int}), you can define
134 the macro safely as follows:
137 #define maxint(a,b) \
138 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
141 Embedded statements are not allowed in constant expressions, such as
142 the value of an enumeration constant, the width of a bit-field, or
143 the initial value of a static variable.
145 If you don't know the type of the operand, you can still do this, but you
146 must use @code{typeof} (@pxref{Typeof}).
148 In G++, the result value of a statement expression undergoes array and
149 function pointer decay, and is returned by value to the enclosing
150 expression. For instance, if @code{A} is a class, then
159 will construct a temporary @code{A} object to hold the result of the
160 statement expression, and that will be used to invoke @code{Foo}.
161 Therefore the @code{this} pointer observed by @code{Foo} will not be the
164 Any temporaries created within a statement within a statement expression
165 will be destroyed at the statement's end. This makes statement
166 expressions inside macros slightly different from function calls. In
167 the latter case temporaries introduced during argument evaluation will
168 be destroyed at the end of the statement that includes the function
169 call. In the statement expression case they will be destroyed during
170 the statement expression. For instance,
173 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
174 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
184 will have different places where temporaries are destroyed. For the
185 @code{macro} case, the temporary @code{X} will be destroyed just after
186 the initialization of @code{b}. In the @code{function} case that
187 temporary will be destroyed when the function returns.
189 These considerations mean that it is probably a bad idea to use
190 statement-expressions of this form in header files that are designed to
191 work with C++. (Note that some versions of the GNU C Library contained
192 header files using statement-expression that lead to precisely this
195 Jumping into a statement expression with @code{goto} or using a
196 @code{switch} statement outside the statement expression with a
197 @code{case} or @code{default} label inside the statement expression is
198 not permitted. Jumping into a statement expression with a computed
199 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
200 Jumping out of a statement expression is permitted, but if the
201 statement expression is part of a larger expression then it is
202 unspecified which other subexpressions of that expression have been
203 evaluated except where the language definition requires certain
204 subexpressions to be evaluated before or after the statement
205 expression. In any case, as with a function call the evaluation of a
206 statement expression is not interleaved with the evaluation of other
207 parts of the containing expression. For example,
210 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
214 will call @code{foo} and @code{bar1} and will not call @code{baz} but
215 may or may not call @code{bar2}. If @code{bar2} is called, it will be
216 called after @code{foo} and before @code{bar1}
219 @section Locally Declared Labels
221 @cindex macros, local labels
223 GCC allows you to declare @dfn{local labels} in any nested block
224 scope. A local label is just like an ordinary label, but you can
225 only reference it (with a @code{goto} statement, or by taking its
226 address) within the block in which it was declared.
228 A local label declaration looks like this:
231 __label__ @var{label};
238 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
241 Local label declarations must come at the beginning of the block,
242 before any ordinary declarations or statements.
244 The label declaration defines the label @emph{name}, but does not define
245 the label itself. You must do this in the usual way, with
246 @code{@var{label}:}, within the statements of the statement expression.
248 The local label feature is useful for complex macros. If a macro
249 contains nested loops, a @code{goto} can be useful for breaking out of
250 them. However, an ordinary label whose scope is the whole function
251 cannot be used: if the macro can be expanded several times in one
252 function, the label will be multiply defined in that function. A
253 local label avoids this problem. For example:
256 #define SEARCH(value, array, target) \
259 typeof (target) _SEARCH_target = (target); \
260 typeof (*(array)) *_SEARCH_array = (array); \
263 for (i = 0; i < max; i++) \
264 for (j = 0; j < max; j++) \
265 if (_SEARCH_array[i][j] == _SEARCH_target) \
266 @{ (value) = i; goto found; @} \
272 This could also be written using a statement-expression:
275 #define SEARCH(array, target) \
278 typeof (target) _SEARCH_target = (target); \
279 typeof (*(array)) *_SEARCH_array = (array); \
282 for (i = 0; i < max; i++) \
283 for (j = 0; j < max; j++) \
284 if (_SEARCH_array[i][j] == _SEARCH_target) \
285 @{ value = i; goto found; @} \
292 Local label declarations also make the labels they declare visible to
293 nested functions, if there are any. @xref{Nested Functions}, for details.
295 @node Labels as Values
296 @section Labels as Values
297 @cindex labels as values
298 @cindex computed gotos
299 @cindex goto with computed label
300 @cindex address of a label
302 You can get the address of a label defined in the current function
303 (or a containing function) with the unary operator @samp{&&}. The
304 value has type @code{void *}. This value is a constant and can be used
305 wherever a constant of that type is valid. For example:
313 To use these values, you need to be able to jump to one. This is done
314 with the computed goto statement@footnote{The analogous feature in
315 Fortran is called an assigned goto, but that name seems inappropriate in
316 C, where one can do more than simply store label addresses in label
317 variables.}, @code{goto *@var{exp};}. For example,
324 Any expression of type @code{void *} is allowed.
326 One way of using these constants is in initializing a static array that
327 will serve as a jump table:
330 static void *array[] = @{ &&foo, &&bar, &&hack @};
333 Then you can select a label with indexing, like this:
340 Note that this does not check whether the subscript is in bounds---array
341 indexing in C never does that.
343 Such an array of label values serves a purpose much like that of the
344 @code{switch} statement. The @code{switch} statement is cleaner, so
345 use that rather than an array unless the problem does not fit a
346 @code{switch} statement very well.
348 Another use of label values is in an interpreter for threaded code.
349 The labels within the interpreter function can be stored in the
350 threaded code for super-fast dispatching.
352 You may not use this mechanism to jump to code in a different function.
353 If you do that, totally unpredictable things will happen. The best way to
354 avoid this is to store the label address only in automatic variables and
355 never pass it as an argument.
357 An alternate way to write the above example is
360 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
362 goto *(&&foo + array[i]);
366 This is more friendly to code living in shared libraries, as it reduces
367 the number of dynamic relocations that are needed, and by consequence,
368 allows the data to be read-only.
370 @node Nested Functions
371 @section Nested Functions
372 @cindex nested functions
373 @cindex downward funargs
376 A @dfn{nested function} is a function defined inside another function.
377 (Nested functions are not supported for GNU C++.) The nested function's
378 name is local to the block where it is defined. For example, here we
379 define a nested function named @code{square}, and call it twice:
383 foo (double a, double b)
385 double square (double z) @{ return z * z; @}
387 return square (a) + square (b);
392 The nested function can access all the variables of the containing
393 function that are visible at the point of its definition. This is
394 called @dfn{lexical scoping}. For example, here we show a nested
395 function which uses an inherited variable named @code{offset}:
399 bar (int *array, int offset, int size)
401 int access (int *array, int index)
402 @{ return array[index + offset]; @}
405 for (i = 0; i < size; i++)
406 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
411 Nested function definitions are permitted within functions in the places
412 where variable definitions are allowed; that is, in any block, mixed
413 with the other declarations and statements in the block.
415 It is possible to call the nested function from outside the scope of its
416 name by storing its address or passing the address to another function:
419 hack (int *array, int size)
421 void store (int index, int value)
422 @{ array[index] = value; @}
424 intermediate (store, size);
428 Here, the function @code{intermediate} receives the address of
429 @code{store} as an argument. If @code{intermediate} calls @code{store},
430 the arguments given to @code{store} are used to store into @code{array}.
431 But this technique works only so long as the containing function
432 (@code{hack}, in this example) does not exit.
434 If you try to call the nested function through its address after the
435 containing function has exited, all hell will break loose. If you try
436 to call it after a containing scope level has exited, and if it refers
437 to some of the variables that are no longer in scope, you may be lucky,
438 but it's not wise to take the risk. If, however, the nested function
439 does not refer to anything that has gone out of scope, you should be
442 GCC implements taking the address of a nested function using a technique
443 called @dfn{trampolines}. A paper describing them is available as
446 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
448 A nested function can jump to a label inherited from a containing
449 function, provided the label was explicitly declared in the containing
450 function (@pxref{Local Labels}). Such a jump returns instantly to the
451 containing function, exiting the nested function which did the
452 @code{goto} and any intermediate functions as well. Here is an example:
456 bar (int *array, int offset, int size)
459 int access (int *array, int index)
463 return array[index + offset];
467 for (i = 0; i < size; i++)
468 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
472 /* @r{Control comes here from @code{access}
473 if it detects an error.} */
480 A nested function always has no linkage. Declaring one with
481 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
482 before its definition, use @code{auto} (which is otherwise meaningless
483 for function declarations).
486 bar (int *array, int offset, int size)
489 auto int access (int *, int);
491 int access (int *array, int index)
495 return array[index + offset];
501 @node Constructing Calls
502 @section Constructing Function Calls
503 @cindex constructing calls
504 @cindex forwarding calls
506 Using the built-in functions described below, you can record
507 the arguments a function received, and call another function
508 with the same arguments, without knowing the number or types
511 You can also record the return value of that function call,
512 and later return that value, without knowing what data type
513 the function tried to return (as long as your caller expects
516 However, these built-in functions may interact badly with some
517 sophisticated features or other extensions of the language. It
518 is, therefore, not recommended to use them outside very simple
519 functions acting as mere forwarders for their arguments.
521 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
522 This built-in function returns a pointer to data
523 describing how to perform a call with the same arguments as were passed
524 to the current function.
526 The function saves the arg pointer register, structure value address,
527 and all registers that might be used to pass arguments to a function
528 into a block of memory allocated on the stack. Then it returns the
529 address of that block.
532 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
533 This built-in function invokes @var{function}
534 with a copy of the parameters described by @var{arguments}
537 The value of @var{arguments} should be the value returned by
538 @code{__builtin_apply_args}. The argument @var{size} specifies the size
539 of the stack argument data, in bytes.
541 This function returns a pointer to data describing
542 how to return whatever value was returned by @var{function}. The data
543 is saved in a block of memory allocated on the stack.
545 It is not always simple to compute the proper value for @var{size}. The
546 value is used by @code{__builtin_apply} to compute the amount of data
547 that should be pushed on the stack and copied from the incoming argument
551 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
552 This built-in function returns the value described by @var{result} from
553 the containing function. You should specify, for @var{result}, a value
554 returned by @code{__builtin_apply}.
558 @section Referring to a Type with @code{typeof}
561 @cindex macros, types of arguments
563 Another way to refer to the type of an expression is with @code{typeof}.
564 The syntax of using of this keyword looks like @code{sizeof}, but the
565 construct acts semantically like a type name defined with @code{typedef}.
567 There are two ways of writing the argument to @code{typeof}: with an
568 expression or with a type. Here is an example with an expression:
575 This assumes that @code{x} is an array of pointers to functions;
576 the type described is that of the values of the functions.
578 Here is an example with a typename as the argument:
585 Here the type described is that of pointers to @code{int}.
587 If you are writing a header file that must work when included in ISO C
588 programs, write @code{__typeof__} instead of @code{typeof}.
589 @xref{Alternate Keywords}.
591 A @code{typeof}-construct can be used anywhere a typedef name could be
592 used. For example, you can use it in a declaration, in a cast, or inside
593 of @code{sizeof} or @code{typeof}.
595 @code{typeof} is often useful in conjunction with the
596 statements-within-expressions feature. Here is how the two together can
597 be used to define a safe ``maximum'' macro that operates on any
598 arithmetic type and evaluates each of its arguments exactly once:
602 (@{ typeof (a) _a = (a); \
603 typeof (b) _b = (b); \
604 _a > _b ? _a : _b; @})
607 @cindex underscores in variables in macros
608 @cindex @samp{_} in variables in macros
609 @cindex local variables in macros
610 @cindex variables, local, in macros
611 @cindex macros, local variables in
613 The reason for using names that start with underscores for the local
614 variables is to avoid conflicts with variable names that occur within the
615 expressions that are substituted for @code{a} and @code{b}. Eventually we
616 hope to design a new form of declaration syntax that allows you to declare
617 variables whose scopes start only after their initializers; this will be a
618 more reliable way to prevent such conflicts.
621 Some more examples of the use of @code{typeof}:
625 This declares @code{y} with the type of what @code{x} points to.
632 This declares @code{y} as an array of such values.
639 This declares @code{y} as an array of pointers to characters:
642 typeof (typeof (char *)[4]) y;
646 It is equivalent to the following traditional C declaration:
652 To see the meaning of the declaration using @code{typeof}, and why it
653 might be a useful way to write, rewrite it with these macros:
656 #define pointer(T) typeof(T *)
657 #define array(T, N) typeof(T [N])
661 Now the declaration can be rewritten this way:
664 array (pointer (char), 4) y;
668 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
669 pointers to @code{char}.
672 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
673 a more limited extension which permitted one to write
676 typedef @var{T} = @var{expr};
680 with the effect of declaring @var{T} to have the type of the expression
681 @var{expr}. This extension does not work with GCC 3 (versions between
682 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
683 relies on it should be rewritten to use @code{typeof}:
686 typedef typeof(@var{expr}) @var{T};
690 This will work with all versions of GCC@.
693 @section Conditionals with Omitted Operands
694 @cindex conditional expressions, extensions
695 @cindex omitted middle-operands
696 @cindex middle-operands, omitted
697 @cindex extensions, @code{?:}
698 @cindex @code{?:} extensions
700 The middle operand in a conditional expression may be omitted. Then
701 if the first operand is nonzero, its value is the value of the conditional
704 Therefore, the expression
711 has the value of @code{x} if that is nonzero; otherwise, the value of
714 This example is perfectly equivalent to
720 @cindex side effect in ?:
721 @cindex ?: side effect
723 In this simple case, the ability to omit the middle operand is not
724 especially useful. When it becomes useful is when the first operand does,
725 or may (if it is a macro argument), contain a side effect. Then repeating
726 the operand in the middle would perform the side effect twice. Omitting
727 the middle operand uses the value already computed without the undesirable
728 effects of recomputing it.
731 @section Double-Word Integers
732 @cindex @code{long long} data types
733 @cindex double-word arithmetic
734 @cindex multiprecision arithmetic
735 @cindex @code{LL} integer suffix
736 @cindex @code{ULL} integer suffix
738 ISO C99 supports data types for integers that are at least 64 bits wide,
739 and as an extension GCC supports them in C89 mode and in C++.
740 Simply write @code{long long int} for a signed integer, or
741 @code{unsigned long long int} for an unsigned integer. To make an
742 integer constant of type @code{long long int}, add the suffix @samp{LL}
743 to the integer. To make an integer constant of type @code{unsigned long
744 long int}, add the suffix @samp{ULL} to the integer.
746 You can use these types in arithmetic like any other integer types.
747 Addition, subtraction, and bitwise boolean operations on these types
748 are open-coded on all types of machines. Multiplication is open-coded
749 if the machine supports fullword-to-doubleword a widening multiply
750 instruction. Division and shifts are open-coded only on machines that
751 provide special support. The operations that are not open-coded use
752 special library routines that come with GCC@.
754 There may be pitfalls when you use @code{long long} types for function
755 arguments, unless you declare function prototypes. If a function
756 expects type @code{int} for its argument, and you pass a value of type
757 @code{long long int}, confusion will result because the caller and the
758 subroutine will disagree about the number of bytes for the argument.
759 Likewise, if the function expects @code{long long int} and you pass
760 @code{int}. The best way to avoid such problems is to use prototypes.
763 @section Complex Numbers
764 @cindex complex numbers
765 @cindex @code{_Complex} keyword
766 @cindex @code{__complex__} keyword
768 ISO C99 supports complex floating data types, and as an extension GCC
769 supports them in C89 mode and in C++, and supports complex integer data
770 types which are not part of ISO C99. You can declare complex types
771 using the keyword @code{_Complex}. As an extension, the older GNU
772 keyword @code{__complex__} is also supported.
774 For example, @samp{_Complex double x;} declares @code{x} as a
775 variable whose real part and imaginary part are both of type
776 @code{double}. @samp{_Complex short int y;} declares @code{y} to
777 have real and imaginary parts of type @code{short int}; this is not
778 likely to be useful, but it shows that the set of complex types is
781 To write a constant with a complex data type, use the suffix @samp{i} or
782 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
783 has type @code{_Complex float} and @code{3i} has type
784 @code{_Complex int}. Such a constant always has a pure imaginary
785 value, but you can form any complex value you like by adding one to a
786 real constant. This is a GNU extension; if you have an ISO C99
787 conforming C library (such as GNU libc), and want to construct complex
788 constants of floating type, you should include @code{<complex.h>} and
789 use the macros @code{I} or @code{_Complex_I} instead.
791 @cindex @code{__real__} keyword
792 @cindex @code{__imag__} keyword
793 To extract the real part of a complex-valued expression @var{exp}, write
794 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
795 extract the imaginary part. This is a GNU extension; for values of
796 floating type, you should use the ISO C99 functions @code{crealf},
797 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
798 @code{cimagl}, declared in @code{<complex.h>} and also provided as
799 built-in functions by GCC@.
801 @cindex complex conjugation
802 The operator @samp{~} performs complex conjugation when used on a value
803 with a complex type. This is a GNU extension; for values of
804 floating type, you should use the ISO C99 functions @code{conjf},
805 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
806 provided as built-in functions by GCC@.
808 GCC can allocate complex automatic variables in a noncontiguous
809 fashion; it's even possible for the real part to be in a register while
810 the imaginary part is on the stack (or vice-versa). Only the DWARF2
811 debug info format can represent this, so use of DWARF2 is recommended.
812 If you are using the stabs debug info format, GCC describes a noncontiguous
813 complex variable as if it were two separate variables of noncomplex type.
814 If the variable's actual name is @code{foo}, the two fictitious
815 variables are named @code{foo$real} and @code{foo$imag}. You can
816 examine and set these two fictitious variables with your debugger.
819 @section Decimal Floating Types
820 @cindex decimal floating types
821 @cindex @code{_Decimal32} data type
822 @cindex @code{_Decimal64} data type
823 @cindex @code{_Decimal128} data type
824 @cindex @code{df} integer suffix
825 @cindex @code{dd} integer suffix
826 @cindex @code{dl} integer suffix
827 @cindex @code{DF} integer suffix
828 @cindex @code{DD} integer suffix
829 @cindex @code{DL} integer suffix
831 As an extension, the GNU C compiler supports decimal floating types as
832 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
833 floating types in GCC will evolve as the draft technical report changes.
834 Calling conventions for any target might also change. Not all targets
835 support decimal floating types.
837 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
838 @code{_Decimal128}. They use a radix of ten, unlike the floating types
839 @code{float}, @code{double}, and @code{long double} whose radix is not
840 specified by the C standard but is usually two.
842 Support for decimal floating types includes the arithmetic operators
843 add, subtract, multiply, divide; unary arithmetic operators;
844 relational operators; equality operators; and conversions to and from
845 integer and other floating types. Use a suffix @samp{df} or
846 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
847 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
850 GCC support of decimal float as specified by the draft technical report
855 Translation time data type (TTDT) is not supported.
858 When the value of a decimal floating type cannot be represented in the
859 integer type to which it is being converted, the result is undefined
860 rather than the result value specified by the draft technical report.
863 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
864 are supported by the DWARF2 debug information format.
870 ISO C99 supports floating-point numbers written not only in the usual
871 decimal notation, such as @code{1.55e1}, but also numbers such as
872 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
873 supports this in C89 mode (except in some cases when strictly
874 conforming) and in C++. In that format the
875 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
876 mandatory. The exponent is a decimal number that indicates the power of
877 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
884 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
885 is the same as @code{1.55e1}.
887 Unlike for floating-point numbers in the decimal notation the exponent
888 is always required in the hexadecimal notation. Otherwise the compiler
889 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
890 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
891 extension for floating-point constants of type @code{float}.
894 @section Arrays of Length Zero
895 @cindex arrays of length zero
896 @cindex zero-length arrays
897 @cindex length-zero arrays
898 @cindex flexible array members
900 Zero-length arrays are allowed in GNU C@. They are very useful as the
901 last element of a structure which is really a header for a variable-length
910 struct line *thisline = (struct line *)
911 malloc (sizeof (struct line) + this_length);
912 thisline->length = this_length;
915 In ISO C90, you would have to give @code{contents} a length of 1, which
916 means either you waste space or complicate the argument to @code{malloc}.
918 In ISO C99, you would use a @dfn{flexible array member}, which is
919 slightly different in syntax and semantics:
923 Flexible array members are written as @code{contents[]} without
927 Flexible array members have incomplete type, and so the @code{sizeof}
928 operator may not be applied. As a quirk of the original implementation
929 of zero-length arrays, @code{sizeof} evaluates to zero.
932 Flexible array members may only appear as the last member of a
933 @code{struct} that is otherwise non-empty.
936 A structure containing a flexible array member, or a union containing
937 such a structure (possibly recursively), may not be a member of a
938 structure or an element of an array. (However, these uses are
939 permitted by GCC as extensions.)
942 GCC versions before 3.0 allowed zero-length arrays to be statically
943 initialized, as if they were flexible arrays. In addition to those
944 cases that were useful, it also allowed initializations in situations
945 that would corrupt later data. Non-empty initialization of zero-length
946 arrays is now treated like any case where there are more initializer
947 elements than the array holds, in that a suitable warning about "excess
948 elements in array" is given, and the excess elements (all of them, in
949 this case) are ignored.
951 Instead GCC allows static initialization of flexible array members.
952 This is equivalent to defining a new structure containing the original
953 structure followed by an array of sufficient size to contain the data.
954 I.e.@: in the following, @code{f1} is constructed as if it were declared
960 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
963 struct f1 f1; int data[3];
964 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
968 The convenience of this extension is that @code{f1} has the desired
969 type, eliminating the need to consistently refer to @code{f2.f1}.
971 This has symmetry with normal static arrays, in that an array of
972 unknown size is also written with @code{[]}.
974 Of course, this extension only makes sense if the extra data comes at
975 the end of a top-level object, as otherwise we would be overwriting
976 data at subsequent offsets. To avoid undue complication and confusion
977 with initialization of deeply nested arrays, we simply disallow any
978 non-empty initialization except when the structure is the top-level
982 struct foo @{ int x; int y[]; @};
983 struct bar @{ struct foo z; @};
985 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
986 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
987 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
988 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
991 @node Empty Structures
992 @section Structures With No Members
993 @cindex empty structures
994 @cindex zero-size structures
996 GCC permits a C structure to have no members:
1003 The structure will have size zero. In C++, empty structures are part
1004 of the language. G++ treats empty structures as if they had a single
1005 member of type @code{char}.
1007 @node Variable Length
1008 @section Arrays of Variable Length
1009 @cindex variable-length arrays
1010 @cindex arrays of variable length
1013 Variable-length automatic arrays are allowed in ISO C99, and as an
1014 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1015 implementation of variable-length arrays does not yet conform in detail
1016 to the ISO C99 standard.) These arrays are
1017 declared like any other automatic arrays, but with a length that is not
1018 a constant expression. The storage is allocated at the point of
1019 declaration and deallocated when the brace-level is exited. For
1024 concat_fopen (char *s1, char *s2, char *mode)
1026 char str[strlen (s1) + strlen (s2) + 1];
1029 return fopen (str, mode);
1033 @cindex scope of a variable length array
1034 @cindex variable-length array scope
1035 @cindex deallocating variable length arrays
1036 Jumping or breaking out of the scope of the array name deallocates the
1037 storage. Jumping into the scope is not allowed; you get an error
1040 @cindex @code{alloca} vs variable-length arrays
1041 You can use the function @code{alloca} to get an effect much like
1042 variable-length arrays. The function @code{alloca} is available in
1043 many other C implementations (but not in all). On the other hand,
1044 variable-length arrays are more elegant.
1046 There are other differences between these two methods. Space allocated
1047 with @code{alloca} exists until the containing @emph{function} returns.
1048 The space for a variable-length array is deallocated as soon as the array
1049 name's scope ends. (If you use both variable-length arrays and
1050 @code{alloca} in the same function, deallocation of a variable-length array
1051 will also deallocate anything more recently allocated with @code{alloca}.)
1053 You can also use variable-length arrays as arguments to functions:
1057 tester (int len, char data[len][len])
1063 The length of an array is computed once when the storage is allocated
1064 and is remembered for the scope of the array in case you access it with
1067 If you want to pass the array first and the length afterward, you can
1068 use a forward declaration in the parameter list---another GNU extension.
1072 tester (int len; char data[len][len], int len)
1078 @cindex parameter forward declaration
1079 The @samp{int len} before the semicolon is a @dfn{parameter forward
1080 declaration}, and it serves the purpose of making the name @code{len}
1081 known when the declaration of @code{data} is parsed.
1083 You can write any number of such parameter forward declarations in the
1084 parameter list. They can be separated by commas or semicolons, but the
1085 last one must end with a semicolon, which is followed by the ``real''
1086 parameter declarations. Each forward declaration must match a ``real''
1087 declaration in parameter name and data type. ISO C99 does not support
1088 parameter forward declarations.
1090 @node Variadic Macros
1091 @section Macros with a Variable Number of Arguments.
1092 @cindex variable number of arguments
1093 @cindex macro with variable arguments
1094 @cindex rest argument (in macro)
1095 @cindex variadic macros
1097 In the ISO C standard of 1999, a macro can be declared to accept a
1098 variable number of arguments much as a function can. The syntax for
1099 defining the macro is similar to that of a function. Here is an
1103 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1106 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1107 such a macro, it represents the zero or more tokens until the closing
1108 parenthesis that ends the invocation, including any commas. This set of
1109 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1110 wherever it appears. See the CPP manual for more information.
1112 GCC has long supported variadic macros, and used a different syntax that
1113 allowed you to give a name to the variable arguments just like any other
1114 argument. Here is an example:
1117 #define debug(format, args...) fprintf (stderr, format, args)
1120 This is in all ways equivalent to the ISO C example above, but arguably
1121 more readable and descriptive.
1123 GNU CPP has two further variadic macro extensions, and permits them to
1124 be used with either of the above forms of macro definition.
1126 In standard C, you are not allowed to leave the variable argument out
1127 entirely; but you are allowed to pass an empty argument. For example,
1128 this invocation is invalid in ISO C, because there is no comma after
1135 GNU CPP permits you to completely omit the variable arguments in this
1136 way. In the above examples, the compiler would complain, though since
1137 the expansion of the macro still has the extra comma after the format
1140 To help solve this problem, CPP behaves specially for variable arguments
1141 used with the token paste operator, @samp{##}. If instead you write
1144 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1147 and if the variable arguments are omitted or empty, the @samp{##}
1148 operator causes the preprocessor to remove the comma before it. If you
1149 do provide some variable arguments in your macro invocation, GNU CPP
1150 does not complain about the paste operation and instead places the
1151 variable arguments after the comma. Just like any other pasted macro
1152 argument, these arguments are not macro expanded.
1154 @node Escaped Newlines
1155 @section Slightly Looser Rules for Escaped Newlines
1156 @cindex escaped newlines
1157 @cindex newlines (escaped)
1159 Recently, the preprocessor has relaxed its treatment of escaped
1160 newlines. Previously, the newline had to immediately follow a
1161 backslash. The current implementation allows whitespace in the form
1162 of spaces, horizontal and vertical tabs, and form feeds between the
1163 backslash and the subsequent newline. The preprocessor issues a
1164 warning, but treats it as a valid escaped newline and combines the two
1165 lines to form a single logical line. This works within comments and
1166 tokens, as well as between tokens. Comments are @emph{not} treated as
1167 whitespace for the purposes of this relaxation, since they have not
1168 yet been replaced with spaces.
1171 @section Non-Lvalue Arrays May Have Subscripts
1172 @cindex subscripting
1173 @cindex arrays, non-lvalue
1175 @cindex subscripting and function values
1176 In ISO C99, arrays that are not lvalues still decay to pointers, and
1177 may be subscripted, although they may not be modified or used after
1178 the next sequence point and the unary @samp{&} operator may not be
1179 applied to them. As an extension, GCC allows such arrays to be
1180 subscripted in C89 mode, though otherwise they do not decay to
1181 pointers outside C99 mode. For example,
1182 this is valid in GNU C though not valid in C89:
1186 struct foo @{int a[4];@};
1192 return f().a[index];
1198 @section Arithmetic on @code{void}- and Function-Pointers
1199 @cindex void pointers, arithmetic
1200 @cindex void, size of pointer to
1201 @cindex function pointers, arithmetic
1202 @cindex function, size of pointer to
1204 In GNU C, addition and subtraction operations are supported on pointers to
1205 @code{void} and on pointers to functions. This is done by treating the
1206 size of a @code{void} or of a function as 1.
1208 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1209 and on function types, and returns 1.
1211 @opindex Wpointer-arith
1212 The option @option{-Wpointer-arith} requests a warning if these extensions
1216 @section Non-Constant Initializers
1217 @cindex initializers, non-constant
1218 @cindex non-constant initializers
1220 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1221 automatic variable are not required to be constant expressions in GNU C@.
1222 Here is an example of an initializer with run-time varying elements:
1225 foo (float f, float g)
1227 float beat_freqs[2] = @{ f-g, f+g @};
1232 @node Compound Literals
1233 @section Compound Literals
1234 @cindex constructor expressions
1235 @cindex initializations in expressions
1236 @cindex structures, constructor expression
1237 @cindex expressions, constructor
1238 @cindex compound literals
1239 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1241 ISO C99 supports compound literals. A compound literal looks like
1242 a cast containing an initializer. Its value is an object of the
1243 type specified in the cast, containing the elements specified in
1244 the initializer; it is an lvalue. As an extension, GCC supports
1245 compound literals in C89 mode and in C++.
1247 Usually, the specified type is a structure. Assume that
1248 @code{struct foo} and @code{structure} are declared as shown:
1251 struct foo @{int a; char b[2];@} structure;
1255 Here is an example of constructing a @code{struct foo} with a compound literal:
1258 structure = ((struct foo) @{x + y, 'a', 0@});
1262 This is equivalent to writing the following:
1266 struct foo temp = @{x + y, 'a', 0@};
1271 You can also construct an array. If all the elements of the compound literal
1272 are (made up of) simple constant expressions, suitable for use in
1273 initializers of objects of static storage duration, then the compound
1274 literal can be coerced to a pointer to its first element and used in
1275 such an initializer, as shown here:
1278 char **foo = (char *[]) @{ "x", "y", "z" @};
1281 Compound literals for scalar types and union types are is
1282 also allowed, but then the compound literal is equivalent
1285 As a GNU extension, GCC allows initialization of objects with static storage
1286 duration by compound literals (which is not possible in ISO C99, because
1287 the initializer is not a constant).
1288 It is handled as if the object was initialized only with the bracket
1289 enclosed list if the types of the compound literal and the object match.
1290 The initializer list of the compound literal must be constant.
1291 If the object being initialized has array type of unknown size, the size is
1292 determined by compound literal size.
1295 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1296 static int y[] = (int []) @{1, 2, 3@};
1297 static int z[] = (int [3]) @{1@};
1301 The above lines are equivalent to the following:
1303 static struct foo x = @{1, 'a', 'b'@};
1304 static int y[] = @{1, 2, 3@};
1305 static int z[] = @{1, 0, 0@};
1308 @node Designated Inits
1309 @section Designated Initializers
1310 @cindex initializers with labeled elements
1311 @cindex labeled elements in initializers
1312 @cindex case labels in initializers
1313 @cindex designated initializers
1315 Standard C89 requires the elements of an initializer to appear in a fixed
1316 order, the same as the order of the elements in the array or structure
1319 In ISO C99 you can give the elements in any order, specifying the array
1320 indices or structure field names they apply to, and GNU C allows this as
1321 an extension in C89 mode as well. This extension is not
1322 implemented in GNU C++.
1324 To specify an array index, write
1325 @samp{[@var{index}] =} before the element value. For example,
1328 int a[6] = @{ [4] = 29, [2] = 15 @};
1335 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1339 The index values must be constant expressions, even if the array being
1340 initialized is automatic.
1342 An alternative syntax for this which has been obsolete since GCC 2.5 but
1343 GCC still accepts is to write @samp{[@var{index}]} before the element
1344 value, with no @samp{=}.
1346 To initialize a range of elements to the same value, write
1347 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1348 extension. For example,
1351 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1355 If the value in it has side-effects, the side-effects will happen only once,
1356 not for each initialized field by the range initializer.
1359 Note that the length of the array is the highest value specified
1362 In a structure initializer, specify the name of a field to initialize
1363 with @samp{.@var{fieldname} =} before the element value. For example,
1364 given the following structure,
1367 struct point @{ int x, y; @};
1371 the following initialization
1374 struct point p = @{ .y = yvalue, .x = xvalue @};
1381 struct point p = @{ xvalue, yvalue @};
1384 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1385 @samp{@var{fieldname}:}, as shown here:
1388 struct point p = @{ y: yvalue, x: xvalue @};
1392 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1393 @dfn{designator}. You can also use a designator (or the obsolete colon
1394 syntax) when initializing a union, to specify which element of the union
1395 should be used. For example,
1398 union foo @{ int i; double d; @};
1400 union foo f = @{ .d = 4 @};
1404 will convert 4 to a @code{double} to store it in the union using
1405 the second element. By contrast, casting 4 to type @code{union foo}
1406 would store it into the union as the integer @code{i}, since it is
1407 an integer. (@xref{Cast to Union}.)
1409 You can combine this technique of naming elements with ordinary C
1410 initialization of successive elements. Each initializer element that
1411 does not have a designator applies to the next consecutive element of the
1412 array or structure. For example,
1415 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1422 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1425 Labeling the elements of an array initializer is especially useful
1426 when the indices are characters or belong to an @code{enum} type.
1431 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1432 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1435 @cindex designator lists
1436 You can also write a series of @samp{.@var{fieldname}} and
1437 @samp{[@var{index}]} designators before an @samp{=} to specify a
1438 nested subobject to initialize; the list is taken relative to the
1439 subobject corresponding to the closest surrounding brace pair. For
1440 example, with the @samp{struct point} declaration above:
1443 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1447 If the same field is initialized multiple times, it will have value from
1448 the last initialization. If any such overridden initialization has
1449 side-effect, it is unspecified whether the side-effect happens or not.
1450 Currently, GCC will discard them and issue a warning.
1453 @section Case Ranges
1455 @cindex ranges in case statements
1457 You can specify a range of consecutive values in a single @code{case} label,
1461 case @var{low} ... @var{high}:
1465 This has the same effect as the proper number of individual @code{case}
1466 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1468 This feature is especially useful for ranges of ASCII character codes:
1474 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1475 it may be parsed wrong when you use it with integer values. For example,
1490 @section Cast to a Union Type
1491 @cindex cast to a union
1492 @cindex union, casting to a
1494 A cast to union type is similar to other casts, except that the type
1495 specified is a union type. You can specify the type either with
1496 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1497 a constructor though, not a cast, and hence does not yield an lvalue like
1498 normal casts. (@xref{Compound Literals}.)
1500 The types that may be cast to the union type are those of the members
1501 of the union. Thus, given the following union and variables:
1504 union foo @{ int i; double d; @};
1510 both @code{x} and @code{y} can be cast to type @code{union foo}.
1512 Using the cast as the right-hand side of an assignment to a variable of
1513 union type is equivalent to storing in a member of the union:
1518 u = (union foo) x @equiv{} u.i = x
1519 u = (union foo) y @equiv{} u.d = y
1522 You can also use the union cast as a function argument:
1525 void hack (union foo);
1527 hack ((union foo) x);
1530 @node Mixed Declarations
1531 @section Mixed Declarations and Code
1532 @cindex mixed declarations and code
1533 @cindex declarations, mixed with code
1534 @cindex code, mixed with declarations
1536 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1537 within compound statements. As an extension, GCC also allows this in
1538 C89 mode. For example, you could do:
1547 Each identifier is visible from where it is declared until the end of
1548 the enclosing block.
1550 @node Function Attributes
1551 @section Declaring Attributes of Functions
1552 @cindex function attributes
1553 @cindex declaring attributes of functions
1554 @cindex functions that never return
1555 @cindex functions that return more than once
1556 @cindex functions that have no side effects
1557 @cindex functions in arbitrary sections
1558 @cindex functions that behave like malloc
1559 @cindex @code{volatile} applied to function
1560 @cindex @code{const} applied to function
1561 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1562 @cindex functions with non-null pointer arguments
1563 @cindex functions that are passed arguments in registers on the 386
1564 @cindex functions that pop the argument stack on the 386
1565 @cindex functions that do not pop the argument stack on the 386
1567 In GNU C, you declare certain things about functions called in your program
1568 which help the compiler optimize function calls and check your code more
1571 The keyword @code{__attribute__} allows you to specify special
1572 attributes when making a declaration. This keyword is followed by an
1573 attribute specification inside double parentheses. The following
1574 attributes are currently defined for functions on all targets:
1575 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1576 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1577 @code{format}, @code{format_arg}, @code{no_instrument_function},
1578 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1579 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1580 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1581 @code{gnu_inline} and @code{externally_visible}, @code{hot}, @code{cold}.
1582 Several other attributes are defined for functions on particular target
1583 systems. Other attributes, including @code{section} are supported for
1584 variables declarations (@pxref{Variable Attributes}) and for types (@pxref{Type
1587 You may also specify attributes with @samp{__} preceding and following
1588 each keyword. This allows you to use them in header files without
1589 being concerned about a possible macro of the same name. For example,
1590 you may use @code{__noreturn__} instead of @code{noreturn}.
1592 @xref{Attribute Syntax}, for details of the exact syntax for using
1596 @c Keep this table alphabetized by attribute name. Treat _ as space.
1598 @item alias ("@var{target}")
1599 @cindex @code{alias} attribute
1600 The @code{alias} attribute causes the declaration to be emitted as an
1601 alias for another symbol, which must be specified. For instance,
1604 void __f () @{ /* @r{Do something.} */; @}
1605 void f () __attribute__ ((weak, alias ("__f")));
1608 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1609 mangled name for the target must be used. It is an error if @samp{__f}
1610 is not defined in the same translation unit.
1612 Not all target machines support this attribute.
1615 @cindex @code{always_inline} function attribute
1616 Generally, functions are not inlined unless optimization is specified.
1617 For functions declared inline, this attribute inlines the function even
1618 if no optimization level was specified.
1621 @cindex @code{gnu_inline} function attribute
1622 This attribute should be used with a function which is also declared
1623 with the @code{inline} keyword. It directs GCC to treat the function
1624 as if it were defined in gnu89 mode even when compiling in C99 or
1627 If the function is declared @code{extern}, then this definition of the
1628 function is used only for inlining. In no case is the function
1629 compiled as a standalone function, not even if you take its address
1630 explicitly. Such an address becomes an external reference, as if you
1631 had only declared the function, and had not defined it. This has
1632 almost the effect of a macro. The way to use this is to put a
1633 function definition in a header file with this attribute, and put
1634 another copy of the function, without @code{extern}, in a library
1635 file. The definition in the header file will cause most calls to the
1636 function to be inlined. If any uses of the function remain, they will
1637 refer to the single copy in the library. Note that the two
1638 definitions of the functions need not be precisely the same, although
1639 if they do not have the same effect your program may behave oddly.
1641 If the function is neither @code{extern} nor @code{static}, then the
1642 function is compiled as a standalone function, as well as being
1643 inlined where possible.
1645 This is how GCC traditionally handled functions declared
1646 @code{inline}. Since ISO C99 specifies a different semantics for
1647 @code{inline}, this function attribute is provided as a transition
1648 measure and as a useful feature in its own right. This attribute is
1649 available in GCC 4.1.3 and later. It is available if either of the
1650 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1651 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1652 Function is As Fast As a Macro}.
1654 @cindex @code{flatten} function attribute
1656 Generally, inlining into a function is limited. For a function marked with
1657 this attribute, every call inside this function will be inlined, if possible.
1658 Whether the function itself is considered for inlining depends on its size and
1659 the current inlining parameters. The @code{flatten} attribute only works
1660 reliably in unit-at-a-time mode.
1663 @cindex functions that do pop the argument stack on the 386
1665 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1666 assume that the calling function will pop off the stack space used to
1667 pass arguments. This is
1668 useful to override the effects of the @option{-mrtd} switch.
1671 @cindex @code{const} function attribute
1672 Many functions do not examine any values except their arguments, and
1673 have no effects except the return value. Basically this is just slightly
1674 more strict class than the @code{pure} attribute below, since function is not
1675 allowed to read global memory.
1677 @cindex pointer arguments
1678 Note that a function that has pointer arguments and examines the data
1679 pointed to must @emph{not} be declared @code{const}. Likewise, a
1680 function that calls a non-@code{const} function usually must not be
1681 @code{const}. It does not make sense for a @code{const} function to
1684 The attribute @code{const} is not implemented in GCC versions earlier
1685 than 2.5. An alternative way to declare that a function has no side
1686 effects, which works in the current version and in some older versions,
1690 typedef int intfn ();
1692 extern const intfn square;
1695 This approach does not work in GNU C++ from 2.6.0 on, since the language
1696 specifies that the @samp{const} must be attached to the return value.
1700 @itemx constructor (@var{priority})
1701 @itemx destructor (@var{priority})
1702 @cindex @code{constructor} function attribute
1703 @cindex @code{destructor} function attribute
1704 The @code{constructor} attribute causes the function to be called
1705 automatically before execution enters @code{main ()}. Similarly, the
1706 @code{destructor} attribute causes the function to be called
1707 automatically after @code{main ()} has completed or @code{exit ()} has
1708 been called. Functions with these attributes are useful for
1709 initializing data that will be used implicitly during the execution of
1712 You may provide an optional integer priority to control the order in
1713 which constructor and destructor functions are run. A constructor
1714 with a smaller priority number runs before a constructor with a larger
1715 priority number; the opposite relationship holds for destructors. So,
1716 if you have a constructor that allocates a resource and a destructor
1717 that deallocates the same resource, both functions typically have the
1718 same priority. The priorities for constructor and destructor
1719 functions are the same as those specified for namespace-scope C++
1720 objects (@pxref{C++ Attributes}).
1722 These attributes are not currently implemented for Objective-C@.
1725 @cindex @code{deprecated} attribute.
1726 The @code{deprecated} attribute results in a warning if the function
1727 is used anywhere in the source file. This is useful when identifying
1728 functions that are expected to be removed in a future version of a
1729 program. The warning also includes the location of the declaration
1730 of the deprecated function, to enable users to easily find further
1731 information about why the function is deprecated, or what they should
1732 do instead. Note that the warnings only occurs for uses:
1735 int old_fn () __attribute__ ((deprecated));
1737 int (*fn_ptr)() = old_fn;
1740 results in a warning on line 3 but not line 2.
1742 The @code{deprecated} attribute can also be used for variables and
1743 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1746 @cindex @code{__declspec(dllexport)}
1747 On Microsoft Windows targets and Symbian OS targets the
1748 @code{dllexport} attribute causes the compiler to provide a global
1749 pointer to a pointer in a DLL, so that it can be referenced with the
1750 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1751 name is formed by combining @code{_imp__} and the function or variable
1754 You can use @code{__declspec(dllexport)} as a synonym for
1755 @code{__attribute__ ((dllexport))} for compatibility with other
1758 On systems that support the @code{visibility} attribute, this
1759 attribute also implies ``default'' visibility, unless a
1760 @code{visibility} attribute is explicitly specified. You should avoid
1761 the use of @code{dllexport} with ``hidden'' or ``internal''
1762 visibility; in the future GCC may issue an error for those cases.
1764 Currently, the @code{dllexport} attribute is ignored for inlined
1765 functions, unless the @option{-fkeep-inline-functions} flag has been
1766 used. The attribute is also ignored for undefined symbols.
1768 When applied to C++ classes, the attribute marks defined non-inlined
1769 member functions and static data members as exports. Static consts
1770 initialized in-class are not marked unless they are also defined
1773 For Microsoft Windows targets there are alternative methods for
1774 including the symbol in the DLL's export table such as using a
1775 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1776 the @option{--export-all} linker flag.
1779 @cindex @code{__declspec(dllimport)}
1780 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1781 attribute causes the compiler to reference a function or variable via
1782 a global pointer to a pointer that is set up by the DLL exporting the
1783 symbol. The attribute implies @code{extern} storage. On Microsoft
1784 Windows targets, the pointer name is formed by combining @code{_imp__}
1785 and the function or variable name.
1787 You can use @code{__declspec(dllimport)} as a synonym for
1788 @code{__attribute__ ((dllimport))} for compatibility with other
1791 Currently, the attribute is ignored for inlined functions. If the
1792 attribute is applied to a symbol @emph{definition}, an error is reported.
1793 If a symbol previously declared @code{dllimport} is later defined, the
1794 attribute is ignored in subsequent references, and a warning is emitted.
1795 The attribute is also overridden by a subsequent declaration as
1798 When applied to C++ classes, the attribute marks non-inlined
1799 member functions and static data members as imports. However, the
1800 attribute is ignored for virtual methods to allow creation of vtables
1803 On the SH Symbian OS target the @code{dllimport} attribute also has
1804 another affect---it can cause the vtable and run-time type information
1805 for a class to be exported. This happens when the class has a
1806 dllimport'ed constructor or a non-inline, non-pure virtual function
1807 and, for either of those two conditions, the class also has a inline
1808 constructor or destructor and has a key function that is defined in
1809 the current translation unit.
1811 For Microsoft Windows based targets the use of the @code{dllimport}
1812 attribute on functions is not necessary, but provides a small
1813 performance benefit by eliminating a thunk in the DLL@. The use of the
1814 @code{dllimport} attribute on imported variables was required on older
1815 versions of the GNU linker, but can now be avoided by passing the
1816 @option{--enable-auto-import} switch to the GNU linker. As with
1817 functions, using the attribute for a variable eliminates a thunk in
1820 One drawback to using this attribute is that a pointer to a function
1821 or variable marked as @code{dllimport} cannot be used as a constant
1822 address. On Microsoft Windows targets, the attribute can be disabled
1823 for functions by setting the @option{-mnop-fun-dllimport} flag.
1826 @cindex eight bit data on the H8/300, H8/300H, and H8S
1827 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1828 variable should be placed into the eight bit data section.
1829 The compiler will generate more efficient code for certain operations
1830 on data in the eight bit data area. Note the eight bit data area is limited to
1833 You must use GAS and GLD from GNU binutils version 2.7 or later for
1834 this attribute to work correctly.
1836 @item exception_handler
1837 @cindex exception handler functions on the Blackfin processor
1838 Use this attribute on the Blackfin to indicate that the specified function
1839 is an exception handler. The compiler will generate function entry and
1840 exit sequences suitable for use in an exception handler when this
1841 attribute is present.
1844 @cindex functions which handle memory bank switching
1845 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1846 use a calling convention that takes care of switching memory banks when
1847 entering and leaving a function. This calling convention is also the
1848 default when using the @option{-mlong-calls} option.
1850 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1851 to call and return from a function.
1853 On 68HC11 the compiler will generate a sequence of instructions
1854 to invoke a board-specific routine to switch the memory bank and call the
1855 real function. The board-specific routine simulates a @code{call}.
1856 At the end of a function, it will jump to a board-specific routine
1857 instead of using @code{rts}. The board-specific return routine simulates
1861 @cindex functions that pop the argument stack on the 386
1862 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1863 pass the first argument (if of integral type) in the register ECX and
1864 the second argument (if of integral type) in the register EDX@. Subsequent
1865 and other typed arguments are passed on the stack. The called function will
1866 pop the arguments off the stack. If the number of arguments is variable all
1867 arguments are pushed on the stack.
1869 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1870 @cindex @code{format} function attribute
1872 The @code{format} attribute specifies that a function takes @code{printf},
1873 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1874 should be type-checked against a format string. For example, the
1879 my_printf (void *my_object, const char *my_format, ...)
1880 __attribute__ ((format (printf, 2, 3)));
1884 causes the compiler to check the arguments in calls to @code{my_printf}
1885 for consistency with the @code{printf} style format string argument
1888 The parameter @var{archetype} determines how the format string is
1889 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1890 or @code{strfmon}. (You can also use @code{__printf__},
1891 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1892 parameter @var{string-index} specifies which argument is the format
1893 string argument (starting from 1), while @var{first-to-check} is the
1894 number of the first argument to check against the format string. For
1895 functions where the arguments are not available to be checked (such as
1896 @code{vprintf}), specify the third parameter as zero. In this case the
1897 compiler only checks the format string for consistency. For
1898 @code{strftime} formats, the third parameter is required to be zero.
1899 Since non-static C++ methods have an implicit @code{this} argument, the
1900 arguments of such methods should be counted from two, not one, when
1901 giving values for @var{string-index} and @var{first-to-check}.
1903 In the example above, the format string (@code{my_format}) is the second
1904 argument of the function @code{my_print}, and the arguments to check
1905 start with the third argument, so the correct parameters for the format
1906 attribute are 2 and 3.
1908 @opindex ffreestanding
1909 @opindex fno-builtin
1910 The @code{format} attribute allows you to identify your own functions
1911 which take format strings as arguments, so that GCC can check the
1912 calls to these functions for errors. The compiler always (unless
1913 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1914 for the standard library functions @code{printf}, @code{fprintf},
1915 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1916 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1917 warnings are requested (using @option{-Wformat}), so there is no need to
1918 modify the header file @file{stdio.h}. In C99 mode, the functions
1919 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1920 @code{vsscanf} are also checked. Except in strictly conforming C
1921 standard modes, the X/Open function @code{strfmon} is also checked as
1922 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1923 @xref{C Dialect Options,,Options Controlling C Dialect}.
1925 The target may provide additional types of format checks.
1926 @xref{Target Format Checks,,Format Checks Specific to Particular
1929 @item format_arg (@var{string-index})
1930 @cindex @code{format_arg} function attribute
1931 @opindex Wformat-nonliteral
1932 The @code{format_arg} attribute specifies that a function takes a format
1933 string for a @code{printf}, @code{scanf}, @code{strftime} or
1934 @code{strfmon} style function and modifies it (for example, to translate
1935 it into another language), so the result can be passed to a
1936 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1937 function (with the remaining arguments to the format function the same
1938 as they would have been for the unmodified string). For example, the
1943 my_dgettext (char *my_domain, const char *my_format)
1944 __attribute__ ((format_arg (2)));
1948 causes the compiler to check the arguments in calls to a @code{printf},
1949 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1950 format string argument is a call to the @code{my_dgettext} function, for
1951 consistency with the format string argument @code{my_format}. If the
1952 @code{format_arg} attribute had not been specified, all the compiler
1953 could tell in such calls to format functions would be that the format
1954 string argument is not constant; this would generate a warning when
1955 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1956 without the attribute.
1958 The parameter @var{string-index} specifies which argument is the format
1959 string argument (starting from one). Since non-static C++ methods have
1960 an implicit @code{this} argument, the arguments of such methods should
1961 be counted from two.
1963 The @code{format-arg} attribute allows you to identify your own
1964 functions which modify format strings, so that GCC can check the
1965 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1966 type function whose operands are a call to one of your own function.
1967 The compiler always treats @code{gettext}, @code{dgettext}, and
1968 @code{dcgettext} in this manner except when strict ISO C support is
1969 requested by @option{-ansi} or an appropriate @option{-std} option, or
1970 @option{-ffreestanding} or @option{-fno-builtin}
1971 is used. @xref{C Dialect Options,,Options
1972 Controlling C Dialect}.
1974 @item function_vector
1975 @cindex calling functions through the function vector on H8/300, M16C, and M32C processors
1976 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1977 function should be called through the function vector. Calling a
1978 function through the function vector will reduce code size, however;
1979 the function vector has a limited size (maximum 128 entries on the H8/300
1980 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1982 You must use GAS and GLD from GNU binutils version 2.7 or later for
1983 this attribute to work correctly.
1985 On M16C/M32C targets, the @code{function_vector} attribute declares a
1986 special page subroutine call function. Use of this attribute reduces
1987 the code size by 2 bytes for each call generated to the
1988 subroutine. The argument to the attribute is the vector number entry
1989 from the special page vector table which contains the 16 low-order
1990 bits of the subroutine's entry address. Each vector table has special
1991 page number (18 to 255) which are used in @code{jsrs} instruction.
1992 Jump addresses of the routines are generated by adding 0x0F0000 (in
1993 case of M16C targets) or 0xFF0000 (in case of M32C targets), to the 2
1994 byte addresses set in the vector table. Therefore you need to ensure
1995 that all the special page vector routines should get mapped within the
1996 address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
1999 In the following example 2 bytes will be saved for each call to
2000 function @code{foo}.
2003 void foo (void) __attribute__((function_vector(0x18)));
2014 If functions are defined in one file and are called in another file,
2015 then be sure to write this declaration in both files.
2017 This attribute is ignored for R8C target.
2020 @cindex interrupt handler functions
2021 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
2022 ports to indicate that the specified function is an interrupt handler.
2023 The compiler will generate function entry and exit sequences suitable
2024 for use in an interrupt handler when this attribute is present.
2026 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
2027 SH processors can be specified via the @code{interrupt_handler} attribute.
2029 Note, on the AVR, interrupts will be enabled inside the function.
2031 Note, for the ARM, you can specify the kind of interrupt to be handled by
2032 adding an optional parameter to the interrupt attribute like this:
2035 void f () __attribute__ ((interrupt ("IRQ")));
2038 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2040 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2041 may be called with a word aligned stack pointer.
2043 @item interrupt_handler
2044 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2045 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2046 indicate that the specified function is an interrupt handler. The compiler
2047 will generate function entry and exit sequences suitable for use in an
2048 interrupt handler when this attribute is present.
2051 @cindex User stack pointer in interrupts on the Blackfin
2052 When used together with @code{interrupt_handler}, @code{exception_handler}
2053 or @code{nmi_handler}, code will be generated to load the stack pointer
2054 from the USP register in the function prologue.
2056 @item long_call/short_call
2057 @cindex indirect calls on ARM
2058 This attribute specifies how a particular function is called on
2059 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2060 command line switch and @code{#pragma long_calls} settings. The
2061 @code{long_call} attribute indicates that the function might be far
2062 away from the call site and require a different (more expensive)
2063 calling sequence. The @code{short_call} attribute always places
2064 the offset to the function from the call site into the @samp{BL}
2065 instruction directly.
2067 @item longcall/shortcall
2068 @cindex functions called via pointer on the RS/6000 and PowerPC
2069 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2070 indicates that the function might be far away from the call site and
2071 require a different (more expensive) calling sequence. The
2072 @code{shortcall} attribute indicates that the function is always close
2073 enough for the shorter calling sequence to be used. These attributes
2074 override both the @option{-mlongcall} switch and, on the RS/6000 and
2075 PowerPC, the @code{#pragma longcall} setting.
2077 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2078 calls are necessary.
2081 @cindex indirect calls on MIPS
2082 This attribute specifies how a particular function is called on MIPS@.
2083 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2084 command line switch. This attribute causes the compiler to always call
2085 the function by first loading its address into a register, and then using
2086 the contents of that register.
2089 @cindex @code{malloc} attribute
2090 The @code{malloc} attribute is used to tell the compiler that a function
2091 may be treated as if any non-@code{NULL} pointer it returns cannot
2092 alias any other pointer valid when the function returns.
2093 This will often improve optimization.
2094 Standard functions with this property include @code{malloc} and
2095 @code{calloc}. @code{realloc}-like functions have this property as
2096 long as the old pointer is never referred to (including comparing it
2097 to the new pointer) after the function returns a non-@code{NULL}
2100 @item model (@var{model-name})
2101 @cindex function addressability on the M32R/D
2102 @cindex variable addressability on the IA-64
2104 On the M32R/D, use this attribute to set the addressability of an
2105 object, and of the code generated for a function. The identifier
2106 @var{model-name} is one of @code{small}, @code{medium}, or
2107 @code{large}, representing each of the code models.
2109 Small model objects live in the lower 16MB of memory (so that their
2110 addresses can be loaded with the @code{ld24} instruction), and are
2111 callable with the @code{bl} instruction.
2113 Medium model objects may live anywhere in the 32-bit address space (the
2114 compiler will generate @code{seth/add3} instructions to load their addresses),
2115 and are callable with the @code{bl} instruction.
2117 Large model objects may live anywhere in the 32-bit address space (the
2118 compiler will generate @code{seth/add3} instructions to load their addresses),
2119 and may not be reachable with the @code{bl} instruction (the compiler will
2120 generate the much slower @code{seth/add3/jl} instruction sequence).
2122 On IA-64, use this attribute to set the addressability of an object.
2123 At present, the only supported identifier for @var{model-name} is
2124 @code{small}, indicating addressability via ``small'' (22-bit)
2125 addresses (so that their addresses can be loaded with the @code{addl}
2126 instruction). Caveat: such addressing is by definition not position
2127 independent and hence this attribute must not be used for objects
2128 defined by shared libraries.
2131 @cindex function without a prologue/epilogue code
2132 Use this attribute on the ARM, AVR, C4x, IP2K and SPU ports to indicate that
2133 the specified function does not need prologue/epilogue sequences generated by
2134 the compiler. It is up to the programmer to provide these sequences.
2137 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2138 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2139 use the normal calling convention based on @code{jsr} and @code{rts}.
2140 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2144 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2145 Use this attribute together with @code{interrupt_handler},
2146 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2147 entry code should enable nested interrupts or exceptions.
2150 @cindex NMI handler functions on the Blackfin processor
2151 Use this attribute on the Blackfin to indicate that the specified function
2152 is an NMI handler. The compiler will generate function entry and
2153 exit sequences suitable for use in an NMI handler when this
2154 attribute is present.
2156 @item no_instrument_function
2157 @cindex @code{no_instrument_function} function attribute
2158 @opindex finstrument-functions
2159 If @option{-finstrument-functions} is given, profiling function calls will
2160 be generated at entry and exit of most user-compiled functions.
2161 Functions with this attribute will not be so instrumented.
2164 @cindex @code{noinline} function attribute
2165 This function attribute prevents a function from being considered for
2168 @item nonnull (@var{arg-index}, @dots{})
2169 @cindex @code{nonnull} function attribute
2170 The @code{nonnull} attribute specifies that some function parameters should
2171 be non-null pointers. For instance, the declaration:
2175 my_memcpy (void *dest, const void *src, size_t len)
2176 __attribute__((nonnull (1, 2)));
2180 causes the compiler to check that, in calls to @code{my_memcpy},
2181 arguments @var{dest} and @var{src} are non-null. If the compiler
2182 determines that a null pointer is passed in an argument slot marked
2183 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2184 is issued. The compiler may also choose to make optimizations based
2185 on the knowledge that certain function arguments will not be null.
2187 If no argument index list is given to the @code{nonnull} attribute,
2188 all pointer arguments are marked as non-null. To illustrate, the
2189 following declaration is equivalent to the previous example:
2193 my_memcpy (void *dest, const void *src, size_t len)
2194 __attribute__((nonnull));
2198 @cindex @code{noreturn} function attribute
2199 A few standard library functions, such as @code{abort} and @code{exit},
2200 cannot return. GCC knows this automatically. Some programs define
2201 their own functions that never return. You can declare them
2202 @code{noreturn} to tell the compiler this fact. For example,
2206 void fatal () __attribute__ ((noreturn));
2209 fatal (/* @r{@dots{}} */)
2211 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2217 The @code{noreturn} keyword tells the compiler to assume that
2218 @code{fatal} cannot return. It can then optimize without regard to what
2219 would happen if @code{fatal} ever did return. This makes slightly
2220 better code. More importantly, it helps avoid spurious warnings of
2221 uninitialized variables.
2223 The @code{noreturn} keyword does not affect the exceptional path when that
2224 applies: a @code{noreturn}-marked function may still return to the caller
2225 by throwing an exception or calling @code{longjmp}.
2227 Do not assume that registers saved by the calling function are
2228 restored before calling the @code{noreturn} function.
2230 It does not make sense for a @code{noreturn} function to have a return
2231 type other than @code{void}.
2233 The attribute @code{noreturn} is not implemented in GCC versions
2234 earlier than 2.5. An alternative way to declare that a function does
2235 not return, which works in the current version and in some older
2236 versions, is as follows:
2239 typedef void voidfn ();
2241 volatile voidfn fatal;
2244 This approach does not work in GNU C++.
2247 @cindex @code{nothrow} function attribute
2248 The @code{nothrow} attribute is used to inform the compiler that a
2249 function cannot throw an exception. For example, most functions in
2250 the standard C library can be guaranteed not to throw an exception
2251 with the notable exceptions of @code{qsort} and @code{bsearch} that
2252 take function pointer arguments. The @code{nothrow} attribute is not
2253 implemented in GCC versions earlier than 3.3.
2256 @cindex @code{pure} function attribute
2257 Many functions have no effects except the return value and their
2258 return value depends only on the parameters and/or global variables.
2259 Such a function can be subject
2260 to common subexpression elimination and loop optimization just as an
2261 arithmetic operator would be. These functions should be declared
2262 with the attribute @code{pure}. For example,
2265 int square (int) __attribute__ ((pure));
2269 says that the hypothetical function @code{square} is safe to call
2270 fewer times than the program says.
2272 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2273 Interesting non-pure functions are functions with infinite loops or those
2274 depending on volatile memory or other system resource, that may change between
2275 two consecutive calls (such as @code{feof} in a multithreading environment).
2277 The attribute @code{pure} is not implemented in GCC versions earlier
2281 @cindex @code{hot} function attribute
2282 The @code{hot} attribute is used to inform the compiler that a function is a
2283 hot spot of the compiled program. The function is optimized more aggressively
2284 and on many target it is placed into special subsection of the text section so
2285 all hot functions appears close together improving locality.
2287 When profile feedback is available, via @option{-fprofile-use}, hot functions
2288 are automatically detected and this attribute is ignored.
2290 The @code{hot} attribute is not implemented in GCC versions earlier than 4.3.
2293 @cindex @code{cold} function attribute
2294 The @code{cold} attribute is used to inform the compiler that a function is
2295 unlikely executed. The function is optimized for size rather than speed and on
2296 many targets it is placed into special subsection of the text section so all
2297 cold functions appears close together improving code locality of non-cold parts
2298 of program. The paths leading to call of cold functions within code are marked
2299 as unlikely by the branch prediction mechanism. It is thus useful to mark
2300 functions used to handle unlikely conditions, such as @code{perror}, as cold to
2301 improve optimization of hot functions that do call marked functions in rare
2304 When profile feedback is available, via @option{-fprofile-use}, hot functions
2305 are automatically detected and this attribute is ignored.
2307 The @code{hot} attribute is not implemented in GCC versions earlier than 4.3.
2309 @item regparm (@var{number})
2310 @cindex @code{regparm} attribute
2311 @cindex functions that are passed arguments in registers on the 386
2312 On the Intel 386, the @code{regparm} attribute causes the compiler to
2313 pass arguments number one to @var{number} if they are of integral type
2314 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2315 take a variable number of arguments will continue to be passed all of their
2316 arguments on the stack.
2318 Beware that on some ELF systems this attribute is unsuitable for
2319 global functions in shared libraries with lazy binding (which is the
2320 default). Lazy binding will send the first call via resolving code in
2321 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2322 per the standard calling conventions. Solaris 8 is affected by this.
2323 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2324 safe since the loaders there save all registers. (Lazy binding can be
2325 disabled with the linker or the loader if desired, to avoid the
2329 @cindex @code{sseregparm} attribute
2330 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2331 causes the compiler to pass up to 3 floating point arguments in
2332 SSE registers instead of on the stack. Functions that take a
2333 variable number of arguments will continue to pass all of their
2334 floating point arguments on the stack.
2336 @item force_align_arg_pointer
2337 @cindex @code{force_align_arg_pointer} attribute
2338 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2339 applied to individual function definitions, generating an alternate
2340 prologue and epilogue that realigns the runtime stack. This supports
2341 mixing legacy codes that run with a 4-byte aligned stack with modern
2342 codes that keep a 16-byte stack for SSE compatibility. The alternate
2343 prologue and epilogue are slower and bigger than the regular ones, and
2344 the alternate prologue requires a scratch register; this lowers the
2345 number of registers available if used in conjunction with the
2346 @code{regparm} attribute. The @code{force_align_arg_pointer}
2347 attribute is incompatible with nested functions; this is considered a
2351 @cindex @code{returns_twice} attribute
2352 The @code{returns_twice} attribute tells the compiler that a function may
2353 return more than one time. The compiler will ensure that all registers
2354 are dead before calling such a function and will emit a warning about
2355 the variables that may be clobbered after the second return from the
2356 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2357 The @code{longjmp}-like counterpart of such function, if any, might need
2358 to be marked with the @code{noreturn} attribute.
2361 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2362 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2363 all registers except the stack pointer should be saved in the prologue
2364 regardless of whether they are used or not.
2366 @item section ("@var{section-name}")
2367 @cindex @code{section} function attribute
2368 Normally, the compiler places the code it generates in the @code{text} section.
2369 Sometimes, however, you need additional sections, or you need certain
2370 particular functions to appear in special sections. The @code{section}
2371 attribute specifies that a function lives in a particular section.
2372 For example, the declaration:
2375 extern void foobar (void) __attribute__ ((section ("bar")));
2379 puts the function @code{foobar} in the @code{bar} section.
2381 Some file formats do not support arbitrary sections so the @code{section}
2382 attribute is not available on all platforms.
2383 If you need to map the entire contents of a module to a particular
2384 section, consider using the facilities of the linker instead.
2387 @cindex @code{sentinel} function attribute
2388 This function attribute ensures that a parameter in a function call is
2389 an explicit @code{NULL}. The attribute is only valid on variadic
2390 functions. By default, the sentinel is located at position zero, the
2391 last parameter of the function call. If an optional integer position
2392 argument P is supplied to the attribute, the sentinel must be located at
2393 position P counting backwards from the end of the argument list.
2396 __attribute__ ((sentinel))
2398 __attribute__ ((sentinel(0)))
2401 The attribute is automatically set with a position of 0 for the built-in
2402 functions @code{execl} and @code{execlp}. The built-in function
2403 @code{execle} has the attribute set with a position of 1.
2405 A valid @code{NULL} in this context is defined as zero with any pointer
2406 type. If your system defines the @code{NULL} macro with an integer type
2407 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2408 with a copy that redefines NULL appropriately.
2410 The warnings for missing or incorrect sentinels are enabled with
2414 See long_call/short_call.
2417 See longcall/shortcall.
2420 @cindex signal handler functions on the AVR processors
2421 Use this attribute on the AVR to indicate that the specified
2422 function is a signal handler. The compiler will generate function
2423 entry and exit sequences suitable for use in a signal handler when this
2424 attribute is present. Interrupts will be disabled inside the function.
2427 Use this attribute on the SH to indicate an @code{interrupt_handler}
2428 function should switch to an alternate stack. It expects a string
2429 argument that names a global variable holding the address of the
2434 void f () __attribute__ ((interrupt_handler,
2435 sp_switch ("alt_stack")));
2439 @cindex functions that pop the argument stack on the 386
2440 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2441 assume that the called function will pop off the stack space used to
2442 pass arguments, unless it takes a variable number of arguments.
2445 @cindex tiny data section on the H8/300H and H8S
2446 Use this attribute on the H8/300H and H8S to indicate that the specified
2447 variable should be placed into the tiny data section.
2448 The compiler will generate more efficient code for loads and stores
2449 on data in the tiny data section. Note the tiny data area is limited to
2450 slightly under 32kbytes of data.
2453 Use this attribute on the SH for an @code{interrupt_handler} to return using
2454 @code{trapa} instead of @code{rte}. This attribute expects an integer
2455 argument specifying the trap number to be used.
2458 @cindex @code{unused} attribute.
2459 This attribute, attached to a function, means that the function is meant
2460 to be possibly unused. GCC will not produce a warning for this
2464 @cindex @code{used} attribute.
2465 This attribute, attached to a function, means that code must be emitted
2466 for the function even if it appears that the function is not referenced.
2467 This is useful, for example, when the function is referenced only in
2471 @cindex @code{version_id} attribute on IA64 HP-UX
2472 This attribute, attached to a global variable or function, renames a
2473 symbol to contain a version string, thus allowing for function level
2474 versioning. HP-UX system header files may use version level functioning
2475 for some system calls.
2478 extern int foo () __attribute__((version_id ("20040821")));
2481 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
2483 @item visibility ("@var{visibility_type}")
2484 @cindex @code{visibility} attribute
2485 This attribute affects the linkage of the declaration to which it is attached.
2486 There are four supported @var{visibility_type} values: default,
2487 hidden, protected or internal visibility.
2490 void __attribute__ ((visibility ("protected")))
2491 f () @{ /* @r{Do something.} */; @}
2492 int i __attribute__ ((visibility ("hidden")));
2495 The possible values of @var{visibility_type} correspond to the
2496 visibility settings in the ELF gABI.
2499 @c keep this list of visibilities in alphabetical order.
2502 Default visibility is the normal case for the object file format.
2503 This value is available for the visibility attribute to override other
2504 options that may change the assumed visibility of entities.
2506 On ELF, default visibility means that the declaration is visible to other
2507 modules and, in shared libraries, means that the declared entity may be
2510 On Darwin, default visibility means that the declaration is visible to
2513 Default visibility corresponds to ``external linkage'' in the language.
2516 Hidden visibility indicates that the entity declared will have a new
2517 form of linkage, which we'll call ``hidden linkage''. Two
2518 declarations of an object with hidden linkage refer to the same object
2519 if they are in the same shared object.
2522 Internal visibility is like hidden visibility, but with additional
2523 processor specific semantics. Unless otherwise specified by the
2524 psABI, GCC defines internal visibility to mean that a function is
2525 @emph{never} called from another module. Compare this with hidden
2526 functions which, while they cannot be referenced directly by other
2527 modules, can be referenced indirectly via function pointers. By
2528 indicating that a function cannot be called from outside the module,
2529 GCC may for instance omit the load of a PIC register since it is known
2530 that the calling function loaded the correct value.
2533 Protected visibility is like default visibility except that it
2534 indicates that references within the defining module will bind to the
2535 definition in that module. That is, the declared entity cannot be
2536 overridden by another module.
2540 All visibilities are supported on many, but not all, ELF targets
2541 (supported when the assembler supports the @samp{.visibility}
2542 pseudo-op). Default visibility is supported everywhere. Hidden
2543 visibility is supported on Darwin targets.
2545 The visibility attribute should be applied only to declarations which
2546 would otherwise have external linkage. The attribute should be applied
2547 consistently, so that the same entity should not be declared with
2548 different settings of the attribute.
2550 In C++, the visibility attribute applies to types as well as functions
2551 and objects, because in C++ types have linkage. A class must not have
2552 greater visibility than its non-static data member types and bases,
2553 and class members default to the visibility of their class. Also, a
2554 declaration without explicit visibility is limited to the visibility
2557 In C++, you can mark member functions and static member variables of a
2558 class with the visibility attribute. This is useful if if you know a
2559 particular method or static member variable should only be used from
2560 one shared object; then you can mark it hidden while the rest of the
2561 class has default visibility. Care must be taken to avoid breaking
2562 the One Definition Rule; for example, it is usually not useful to mark
2563 an inline method as hidden without marking the whole class as hidden.
2565 A C++ namespace declaration can also have the visibility attribute.
2566 This attribute applies only to the particular namespace body, not to
2567 other definitions of the same namespace; it is equivalent to using
2568 @samp{#pragma GCC visibility} before and after the namespace
2569 definition (@pxref{Visibility Pragmas}).
2571 In C++, if a template argument has limited visibility, this
2572 restriction is implicitly propagated to the template instantiation.
2573 Otherwise, template instantiations and specializations default to the
2574 visibility of their template.
2576 If both the template and enclosing class have explicit visibility, the
2577 visibility from the template is used.
2579 @item warn_unused_result
2580 @cindex @code{warn_unused_result} attribute
2581 The @code{warn_unused_result} attribute causes a warning to be emitted
2582 if a caller of the function with this attribute does not use its
2583 return value. This is useful for functions where not checking
2584 the result is either a security problem or always a bug, such as
2588 int fn () __attribute__ ((warn_unused_result));
2591 if (fn () < 0) return -1;
2597 results in warning on line 5.
2600 @cindex @code{weak} attribute
2601 The @code{weak} attribute causes the declaration to be emitted as a weak
2602 symbol rather than a global. This is primarily useful in defining
2603 library functions which can be overridden in user code, though it can
2604 also be used with non-function declarations. Weak symbols are supported
2605 for ELF targets, and also for a.out targets when using the GNU assembler
2609 @itemx weakref ("@var{target}")
2610 @cindex @code{weakref} attribute
2611 The @code{weakref} attribute marks a declaration as a weak reference.
2612 Without arguments, it should be accompanied by an @code{alias} attribute
2613 naming the target symbol. Optionally, the @var{target} may be given as
2614 an argument to @code{weakref} itself. In either case, @code{weakref}
2615 implicitly marks the declaration as @code{weak}. Without a
2616 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2617 @code{weakref} is equivalent to @code{weak}.
2620 static int x() __attribute__ ((weakref ("y")));
2621 /* is equivalent to... */
2622 static int x() __attribute__ ((weak, weakref, alias ("y")));
2624 static int x() __attribute__ ((weakref));
2625 static int x() __attribute__ ((alias ("y")));
2628 A weak reference is an alias that does not by itself require a
2629 definition to be given for the target symbol. If the target symbol is
2630 only referenced through weak references, then the becomes a @code{weak}
2631 undefined symbol. If it is directly referenced, however, then such
2632 strong references prevail, and a definition will be required for the
2633 symbol, not necessarily in the same translation unit.
2635 The effect is equivalent to moving all references to the alias to a
2636 separate translation unit, renaming the alias to the aliased symbol,
2637 declaring it as weak, compiling the two separate translation units and
2638 performing a reloadable link on them.
2640 At present, a declaration to which @code{weakref} is attached can
2641 only be @code{static}.
2643 @item externally_visible
2644 @cindex @code{externally_visible} attribute.
2645 This attribute, attached to a global variable or function nullify
2646 effect of @option{-fwhole-program} command line option, so the object
2647 remain visible outside the current compilation unit
2651 You can specify multiple attributes in a declaration by separating them
2652 by commas within the double parentheses or by immediately following an
2653 attribute declaration with another attribute declaration.
2655 @cindex @code{#pragma}, reason for not using
2656 @cindex pragma, reason for not using
2657 Some people object to the @code{__attribute__} feature, suggesting that
2658 ISO C's @code{#pragma} should be used instead. At the time
2659 @code{__attribute__} was designed, there were two reasons for not doing
2664 It is impossible to generate @code{#pragma} commands from a macro.
2667 There is no telling what the same @code{#pragma} might mean in another
2671 These two reasons applied to almost any application that might have been
2672 proposed for @code{#pragma}. It was basically a mistake to use
2673 @code{#pragma} for @emph{anything}.
2675 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2676 to be generated from macros. In addition, a @code{#pragma GCC}
2677 namespace is now in use for GCC-specific pragmas. However, it has been
2678 found convenient to use @code{__attribute__} to achieve a natural
2679 attachment of attributes to their corresponding declarations, whereas
2680 @code{#pragma GCC} is of use for constructs that do not naturally form
2681 part of the grammar. @xref{Other Directives,,Miscellaneous
2682 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2684 @node Attribute Syntax
2685 @section Attribute Syntax
2686 @cindex attribute syntax
2688 This section describes the syntax with which @code{__attribute__} may be
2689 used, and the constructs to which attribute specifiers bind, for the C
2690 language. Some details may vary for C++ and Objective-C@. Because of
2691 infelicities in the grammar for attributes, some forms described here
2692 may not be successfully parsed in all cases.
2694 There are some problems with the semantics of attributes in C++. For
2695 example, there are no manglings for attributes, although they may affect
2696 code generation, so problems may arise when attributed types are used in
2697 conjunction with templates or overloading. Similarly, @code{typeid}
2698 does not distinguish between types with different attributes. Support
2699 for attributes in C++ may be restricted in future to attributes on
2700 declarations only, but not on nested declarators.
2702 @xref{Function Attributes}, for details of the semantics of attributes
2703 applying to functions. @xref{Variable Attributes}, for details of the
2704 semantics of attributes applying to variables. @xref{Type Attributes},
2705 for details of the semantics of attributes applying to structure, union
2706 and enumerated types.
2708 An @dfn{attribute specifier} is of the form
2709 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2710 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2711 each attribute is one of the following:
2715 Empty. Empty attributes are ignored.
2718 A word (which may be an identifier such as @code{unused}, or a reserved
2719 word such as @code{const}).
2722 A word, followed by, in parentheses, parameters for the attribute.
2723 These parameters take one of the following forms:
2727 An identifier. For example, @code{mode} attributes use this form.
2730 An identifier followed by a comma and a non-empty comma-separated list
2731 of expressions. For example, @code{format} attributes use this form.
2734 A possibly empty comma-separated list of expressions. For example,
2735 @code{format_arg} attributes use this form with the list being a single
2736 integer constant expression, and @code{alias} attributes use this form
2737 with the list being a single string constant.
2741 An @dfn{attribute specifier list} is a sequence of one or more attribute
2742 specifiers, not separated by any other tokens.
2744 In GNU C, an attribute specifier list may appear after the colon following a
2745 label, other than a @code{case} or @code{default} label. The only
2746 attribute it makes sense to use after a label is @code{unused}. This
2747 feature is intended for code generated by programs which contains labels
2748 that may be unused but which is compiled with @option{-Wall}. It would
2749 not normally be appropriate to use in it human-written code, though it
2750 could be useful in cases where the code that jumps to the label is
2751 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2752 such placement of attribute lists, as it is permissible for a
2753 declaration, which could begin with an attribute list, to be labelled in
2754 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2755 does not arise there.
2757 An attribute specifier list may appear as part of a @code{struct},
2758 @code{union} or @code{enum} specifier. It may go either immediately
2759 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2760 the closing brace. The former syntax is preferred.
2761 Where attribute specifiers follow the closing brace, they are considered
2762 to relate to the structure, union or enumerated type defined, not to any
2763 enclosing declaration the type specifier appears in, and the type
2764 defined is not complete until after the attribute specifiers.
2765 @c Otherwise, there would be the following problems: a shift/reduce
2766 @c conflict between attributes binding the struct/union/enum and
2767 @c binding to the list of specifiers/qualifiers; and "aligned"
2768 @c attributes could use sizeof for the structure, but the size could be
2769 @c changed later by "packed" attributes.
2771 Otherwise, an attribute specifier appears as part of a declaration,
2772 counting declarations of unnamed parameters and type names, and relates
2773 to that declaration (which may be nested in another declaration, for
2774 example in the case of a parameter declaration), or to a particular declarator
2775 within a declaration. Where an
2776 attribute specifier is applied to a parameter declared as a function or
2777 an array, it should apply to the function or array rather than the
2778 pointer to which the parameter is implicitly converted, but this is not
2779 yet correctly implemented.
2781 Any list of specifiers and qualifiers at the start of a declaration may
2782 contain attribute specifiers, whether or not such a list may in that
2783 context contain storage class specifiers. (Some attributes, however,
2784 are essentially in the nature of storage class specifiers, and only make
2785 sense where storage class specifiers may be used; for example,
2786 @code{section}.) There is one necessary limitation to this syntax: the
2787 first old-style parameter declaration in a function definition cannot
2788 begin with an attribute specifier, because such an attribute applies to
2789 the function instead by syntax described below (which, however, is not
2790 yet implemented in this case). In some other cases, attribute
2791 specifiers are permitted by this grammar but not yet supported by the
2792 compiler. All attribute specifiers in this place relate to the
2793 declaration as a whole. In the obsolescent usage where a type of
2794 @code{int} is implied by the absence of type specifiers, such a list of
2795 specifiers and qualifiers may be an attribute specifier list with no
2796 other specifiers or qualifiers.
2798 At present, the first parameter in a function prototype must have some
2799 type specifier which is not an attribute specifier; this resolves an
2800 ambiguity in the interpretation of @code{void f(int
2801 (__attribute__((foo)) x))}, but is subject to change. At present, if
2802 the parentheses of a function declarator contain only attributes then
2803 those attributes are ignored, rather than yielding an error or warning
2804 or implying a single parameter of type int, but this is subject to
2807 An attribute specifier list may appear immediately before a declarator
2808 (other than the first) in a comma-separated list of declarators in a
2809 declaration of more than one identifier using a single list of
2810 specifiers and qualifiers. Such attribute specifiers apply
2811 only to the identifier before whose declarator they appear. For
2815 __attribute__((noreturn)) void d0 (void),
2816 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2821 the @code{noreturn} attribute applies to all the functions
2822 declared; the @code{format} attribute only applies to @code{d1}.
2824 An attribute specifier list may appear immediately before the comma,
2825 @code{=} or semicolon terminating the declaration of an identifier other
2826 than a function definition. At present, such attribute specifiers apply
2827 to the declared object or function, but in future they may attach to the
2828 outermost adjacent declarator. In simple cases there is no difference,
2829 but, for example, in
2832 void (****f)(void) __attribute__((noreturn));
2836 at present the @code{noreturn} attribute applies to @code{f}, which
2837 causes a warning since @code{f} is not a function, but in future it may
2838 apply to the function @code{****f}. The precise semantics of what
2839 attributes in such cases will apply to are not yet specified. Where an
2840 assembler name for an object or function is specified (@pxref{Asm
2841 Labels}), at present the attribute must follow the @code{asm}
2842 specification; in future, attributes before the @code{asm} specification
2843 may apply to the adjacent declarator, and those after it to the declared
2846 An attribute specifier list may, in future, be permitted to appear after
2847 the declarator in a function definition (before any old-style parameter
2848 declarations or the function body).
2850 Attribute specifiers may be mixed with type qualifiers appearing inside
2851 the @code{[]} of a parameter array declarator, in the C99 construct by
2852 which such qualifiers are applied to the pointer to which the array is
2853 implicitly converted. Such attribute specifiers apply to the pointer,
2854 not to the array, but at present this is not implemented and they are
2857 An attribute specifier list may appear at the start of a nested
2858 declarator. At present, there are some limitations in this usage: the
2859 attributes correctly apply to the declarator, but for most individual
2860 attributes the semantics this implies are not implemented.
2861 When attribute specifiers follow the @code{*} of a pointer
2862 declarator, they may be mixed with any type qualifiers present.
2863 The following describes the formal semantics of this syntax. It will make the
2864 most sense if you are familiar with the formal specification of
2865 declarators in the ISO C standard.
2867 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2868 D1}, where @code{T} contains declaration specifiers that specify a type
2869 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2870 contains an identifier @var{ident}. The type specified for @var{ident}
2871 for derived declarators whose type does not include an attribute
2872 specifier is as in the ISO C standard.
2874 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2875 and the declaration @code{T D} specifies the type
2876 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2877 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2878 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2880 If @code{D1} has the form @code{*
2881 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2882 declaration @code{T D} specifies the type
2883 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2884 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2885 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2891 void (__attribute__((noreturn)) ****f) (void);
2895 specifies the type ``pointer to pointer to pointer to pointer to
2896 non-returning function returning @code{void}''. As another example,
2899 char *__attribute__((aligned(8))) *f;
2903 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2904 Note again that this does not work with most attributes; for example,
2905 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2906 is not yet supported.
2908 For compatibility with existing code written for compiler versions that
2909 did not implement attributes on nested declarators, some laxity is
2910 allowed in the placing of attributes. If an attribute that only applies
2911 to types is applied to a declaration, it will be treated as applying to
2912 the type of that declaration. If an attribute that only applies to
2913 declarations is applied to the type of a declaration, it will be treated
2914 as applying to that declaration; and, for compatibility with code
2915 placing the attributes immediately before the identifier declared, such
2916 an attribute applied to a function return type will be treated as
2917 applying to the function type, and such an attribute applied to an array
2918 element type will be treated as applying to the array type. If an
2919 attribute that only applies to function types is applied to a
2920 pointer-to-function type, it will be treated as applying to the pointer
2921 target type; if such an attribute is applied to a function return type
2922 that is not a pointer-to-function type, it will be treated as applying
2923 to the function type.
2925 @node Function Prototypes
2926 @section Prototypes and Old-Style Function Definitions
2927 @cindex function prototype declarations
2928 @cindex old-style function definitions
2929 @cindex promotion of formal parameters
2931 GNU C extends ISO C to allow a function prototype to override a later
2932 old-style non-prototype definition. Consider the following example:
2935 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2942 /* @r{Prototype function declaration.} */
2943 int isroot P((uid_t));
2945 /* @r{Old-style function definition.} */
2947 isroot (x) /* @r{??? lossage here ???} */
2954 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2955 not allow this example, because subword arguments in old-style
2956 non-prototype definitions are promoted. Therefore in this example the
2957 function definition's argument is really an @code{int}, which does not
2958 match the prototype argument type of @code{short}.
2960 This restriction of ISO C makes it hard to write code that is portable
2961 to traditional C compilers, because the programmer does not know
2962 whether the @code{uid_t} type is @code{short}, @code{int}, or
2963 @code{long}. Therefore, in cases like these GNU C allows a prototype
2964 to override a later old-style definition. More precisely, in GNU C, a
2965 function prototype argument type overrides the argument type specified
2966 by a later old-style definition if the former type is the same as the
2967 latter type before promotion. Thus in GNU C the above example is
2968 equivalent to the following:
2981 GNU C++ does not support old-style function definitions, so this
2982 extension is irrelevant.
2985 @section C++ Style Comments
2987 @cindex C++ comments
2988 @cindex comments, C++ style
2990 In GNU C, you may use C++ style comments, which start with @samp{//} and
2991 continue until the end of the line. Many other C implementations allow
2992 such comments, and they are included in the 1999 C standard. However,
2993 C++ style comments are not recognized if you specify an @option{-std}
2994 option specifying a version of ISO C before C99, or @option{-ansi}
2995 (equivalent to @option{-std=c89}).
2998 @section Dollar Signs in Identifier Names
3000 @cindex dollar signs in identifier names
3001 @cindex identifier names, dollar signs in
3003 In GNU C, you may normally use dollar signs in identifier names.
3004 This is because many traditional C implementations allow such identifiers.
3005 However, dollar signs in identifiers are not supported on a few target
3006 machines, typically because the target assembler does not allow them.
3008 @node Character Escapes
3009 @section The Character @key{ESC} in Constants
3011 You can use the sequence @samp{\e} in a string or character constant to
3012 stand for the ASCII character @key{ESC}.
3015 @section Inquiring on Alignment of Types or Variables
3017 @cindex type alignment
3018 @cindex variable alignment
3020 The keyword @code{__alignof__} allows you to inquire about how an object
3021 is aligned, or the minimum alignment usually required by a type. Its
3022 syntax is just like @code{sizeof}.
3024 For example, if the target machine requires a @code{double} value to be
3025 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3026 This is true on many RISC machines. On more traditional machine
3027 designs, @code{__alignof__ (double)} is 4 or even 2.
3029 Some machines never actually require alignment; they allow reference to any
3030 data type even at an odd address. For these machines, @code{__alignof__}
3031 reports the @emph{recommended} alignment of a type.
3033 If the operand of @code{__alignof__} is an lvalue rather than a type,
3034 its value is the required alignment for its type, taking into account
3035 any minimum alignment specified with GCC's @code{__attribute__}
3036 extension (@pxref{Variable Attributes}). For example, after this
3040 struct foo @{ int x; char y; @} foo1;
3044 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3045 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3047 It is an error to ask for the alignment of an incomplete type.
3049 @node Variable Attributes
3050 @section Specifying Attributes of Variables
3051 @cindex attribute of variables
3052 @cindex variable attributes
3054 The keyword @code{__attribute__} allows you to specify special
3055 attributes of variables or structure fields. This keyword is followed
3056 by an attribute specification inside double parentheses. Some
3057 attributes are currently defined generically for variables.
3058 Other attributes are defined for variables on particular target
3059 systems. Other attributes are available for functions
3060 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3061 Other front ends might define more attributes
3062 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3064 You may also specify attributes with @samp{__} preceding and following
3065 each keyword. This allows you to use them in header files without
3066 being concerned about a possible macro of the same name. For example,
3067 you may use @code{__aligned__} instead of @code{aligned}.
3069 @xref{Attribute Syntax}, for details of the exact syntax for using
3073 @cindex @code{aligned} attribute
3074 @item aligned (@var{alignment})
3075 This attribute specifies a minimum alignment for the variable or
3076 structure field, measured in bytes. For example, the declaration:
3079 int x __attribute__ ((aligned (16))) = 0;
3083 causes the compiler to allocate the global variable @code{x} on a
3084 16-byte boundary. On a 68040, this could be used in conjunction with
3085 an @code{asm} expression to access the @code{move16} instruction which
3086 requires 16-byte aligned operands.
3088 You can also specify the alignment of structure fields. For example, to
3089 create a double-word aligned @code{int} pair, you could write:
3092 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3096 This is an alternative to creating a union with a @code{double} member
3097 that forces the union to be double-word aligned.
3099 As in the preceding examples, you can explicitly specify the alignment
3100 (in bytes) that you wish the compiler to use for a given variable or
3101 structure field. Alternatively, you can leave out the alignment factor
3102 and just ask the compiler to align a variable or field to the maximum
3103 useful alignment for the target machine you are compiling for. For
3104 example, you could write:
3107 short array[3] __attribute__ ((aligned));
3110 Whenever you leave out the alignment factor in an @code{aligned} attribute
3111 specification, the compiler automatically sets the alignment for the declared
3112 variable or field to the largest alignment which is ever used for any data
3113 type on the target machine you are compiling for. Doing this can often make
3114 copy operations more efficient, because the compiler can use whatever
3115 instructions copy the biggest chunks of memory when performing copies to
3116 or from the variables or fields that you have aligned this way.
3118 The @code{aligned} attribute can only increase the alignment; but you
3119 can decrease it by specifying @code{packed} as well. See below.
3121 Note that the effectiveness of @code{aligned} attributes may be limited
3122 by inherent limitations in your linker. On many systems, the linker is
3123 only able to arrange for variables to be aligned up to a certain maximum
3124 alignment. (For some linkers, the maximum supported alignment may
3125 be very very small.) If your linker is only able to align variables
3126 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3127 in an @code{__attribute__} will still only provide you with 8 byte
3128 alignment. See your linker documentation for further information.
3130 @item cleanup (@var{cleanup_function})
3131 @cindex @code{cleanup} attribute
3132 The @code{cleanup} attribute runs a function when the variable goes
3133 out of scope. This attribute can only be applied to auto function
3134 scope variables; it may not be applied to parameters or variables
3135 with static storage duration. The function must take one parameter,
3136 a pointer to a type compatible with the variable. The return value
3137 of the function (if any) is ignored.
3139 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3140 will be run during the stack unwinding that happens during the
3141 processing of the exception. Note that the @code{cleanup} attribute
3142 does not allow the exception to be caught, only to perform an action.
3143 It is undefined what happens if @var{cleanup_function} does not
3148 @cindex @code{common} attribute
3149 @cindex @code{nocommon} attribute
3152 The @code{common} attribute requests GCC to place a variable in
3153 ``common'' storage. The @code{nocommon} attribute requests the
3154 opposite---to allocate space for it directly.
3156 These attributes override the default chosen by the
3157 @option{-fno-common} and @option{-fcommon} flags respectively.
3160 @cindex @code{deprecated} attribute
3161 The @code{deprecated} attribute results in a warning if the variable
3162 is used anywhere in the source file. This is useful when identifying
3163 variables that are expected to be removed in a future version of a
3164 program. The warning also includes the location of the declaration
3165 of the deprecated variable, to enable users to easily find further
3166 information about why the variable is deprecated, or what they should
3167 do instead. Note that the warning only occurs for uses:
3170 extern int old_var __attribute__ ((deprecated));
3172 int new_fn () @{ return old_var; @}
3175 results in a warning on line 3 but not line 2.
3177 The @code{deprecated} attribute can also be used for functions and
3178 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3180 @item mode (@var{mode})
3181 @cindex @code{mode} attribute
3182 This attribute specifies the data type for the declaration---whichever
3183 type corresponds to the mode @var{mode}. This in effect lets you
3184 request an integer or floating point type according to its width.
3186 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3187 indicate the mode corresponding to a one-byte integer, @samp{word} or
3188 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3189 or @samp{__pointer__} for the mode used to represent pointers.
3192 @cindex @code{packed} attribute
3193 The @code{packed} attribute specifies that a variable or structure field
3194 should have the smallest possible alignment---one byte for a variable,
3195 and one bit for a field, unless you specify a larger value with the
3196 @code{aligned} attribute.
3198 Here is a structure in which the field @code{x} is packed, so that it
3199 immediately follows @code{a}:
3205 int x[2] __attribute__ ((packed));
3209 @item section ("@var{section-name}")
3210 @cindex @code{section} variable attribute
3211 Normally, the compiler places the objects it generates in sections like
3212 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3213 or you need certain particular variables to appear in special sections,
3214 for example to map to special hardware. The @code{section}
3215 attribute specifies that a variable (or function) lives in a particular
3216 section. For example, this small program uses several specific section names:
3219 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3220 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3221 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3222 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3226 /* @r{Initialize stack pointer} */
3227 init_sp (stack + sizeof (stack));
3229 /* @r{Initialize initialized data} */
3230 memcpy (&init_data, &data, &edata - &data);
3232 /* @r{Turn on the serial ports} */
3239 Use the @code{section} attribute with an @emph{initialized} definition
3240 of a @emph{global} variable, as shown in the example. GCC issues
3241 a warning and otherwise ignores the @code{section} attribute in
3242 uninitialized variable declarations.
3244 You may only use the @code{section} attribute with a fully initialized
3245 global definition because of the way linkers work. The linker requires
3246 each object be defined once, with the exception that uninitialized
3247 variables tentatively go in the @code{common} (or @code{bss}) section
3248 and can be multiply ``defined''. You can force a variable to be
3249 initialized with the @option{-fno-common} flag or the @code{nocommon}
3252 Some file formats do not support arbitrary sections so the @code{section}
3253 attribute is not available on all platforms.
3254 If you need to map the entire contents of a module to a particular
3255 section, consider using the facilities of the linker instead.
3258 @cindex @code{shared} variable attribute
3259 On Microsoft Windows, in addition to putting variable definitions in a named
3260 section, the section can also be shared among all running copies of an
3261 executable or DLL@. For example, this small program defines shared data
3262 by putting it in a named section @code{shared} and marking the section
3266 int foo __attribute__((section ("shared"), shared)) = 0;
3271 /* @r{Read and write foo. All running
3272 copies see the same value.} */
3278 You may only use the @code{shared} attribute along with @code{section}
3279 attribute with a fully initialized global definition because of the way
3280 linkers work. See @code{section} attribute for more information.
3282 The @code{shared} attribute is only available on Microsoft Windows@.
3284 @item tls_model ("@var{tls_model}")
3285 @cindex @code{tls_model} attribute
3286 The @code{tls_model} attribute sets thread-local storage model
3287 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3288 overriding @option{-ftls-model=} command line switch on a per-variable
3290 The @var{tls_model} argument should be one of @code{global-dynamic},
3291 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3293 Not all targets support this attribute.
3296 This attribute, attached to a variable, means that the variable is meant
3297 to be possibly unused. GCC will not produce a warning for this
3301 This attribute, attached to a variable, means that the variable must be
3302 emitted even if it appears that the variable is not referenced.
3304 @item vector_size (@var{bytes})
3305 This attribute specifies the vector size for the variable, measured in
3306 bytes. For example, the declaration:
3309 int foo __attribute__ ((vector_size (16)));
3313 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3314 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3315 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3317 This attribute is only applicable to integral and float scalars,
3318 although arrays, pointers, and function return values are allowed in
3319 conjunction with this construct.
3321 Aggregates with this attribute are invalid, even if they are of the same
3322 size as a corresponding scalar. For example, the declaration:
3325 struct S @{ int a; @};
3326 struct S __attribute__ ((vector_size (16))) foo;
3330 is invalid even if the size of the structure is the same as the size of
3334 The @code{selectany} attribute causes an initialized global variable to
3335 have link-once semantics. When multiple definitions of the variable are
3336 encountered by the linker, the first is selected and the remainder are
3337 discarded. Following usage by the Microsoft compiler, the linker is told
3338 @emph{not} to warn about size or content differences of the multiple
3341 Although the primary usage of this attribute is for POD types, the
3342 attribute can also be applied to global C++ objects that are initialized
3343 by a constructor. In this case, the static initialization and destruction
3344 code for the object is emitted in each translation defining the object,
3345 but the calls to the constructor and destructor are protected by a
3346 link-once guard variable.
3348 The @code{selectany} attribute is only available on Microsoft Windows
3349 targets. You can use @code{__declspec (selectany)} as a synonym for
3350 @code{__attribute__ ((selectany))} for compatibility with other
3354 The @code{weak} attribute is described in @xref{Function Attributes}.
3357 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3360 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3364 @subsection M32R/D Variable Attributes
3366 One attribute is currently defined for the M32R/D@.
3369 @item model (@var{model-name})
3370 @cindex variable addressability on the M32R/D
3371 Use this attribute on the M32R/D to set the addressability of an object.
3372 The identifier @var{model-name} is one of @code{small}, @code{medium},
3373 or @code{large}, representing each of the code models.
3375 Small model objects live in the lower 16MB of memory (so that their
3376 addresses can be loaded with the @code{ld24} instruction).
3378 Medium and large model objects may live anywhere in the 32-bit address space
3379 (the compiler will generate @code{seth/add3} instructions to load their
3383 @anchor{i386 Variable Attributes}
3384 @subsection i386 Variable Attributes
3386 Two attributes are currently defined for i386 configurations:
3387 @code{ms_struct} and @code{gcc_struct}
3392 @cindex @code{ms_struct} attribute
3393 @cindex @code{gcc_struct} attribute
3395 If @code{packed} is used on a structure, or if bit-fields are used
3396 it may be that the Microsoft ABI packs them differently
3397 than GCC would normally pack them. Particularly when moving packed
3398 data between functions compiled with GCC and the native Microsoft compiler
3399 (either via function call or as data in a file), it may be necessary to access
3402 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3403 compilers to match the native Microsoft compiler.
3405 The Microsoft structure layout algorithm is fairly simple with the exception
3406 of the bitfield packing:
3408 The padding and alignment of members of structures and whether a bit field
3409 can straddle a storage-unit boundary
3412 @item Structure members are stored sequentially in the order in which they are
3413 declared: the first member has the lowest memory address and the last member
3416 @item Every data object has an alignment-requirement. The alignment-requirement
3417 for all data except structures, unions, and arrays is either the size of the
3418 object or the current packing size (specified with either the aligned attribute
3419 or the pack pragma), whichever is less. For structures, unions, and arrays,
3420 the alignment-requirement is the largest alignment-requirement of its members.
3421 Every object is allocated an offset so that:
3423 offset % alignment-requirement == 0
3425 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3426 unit if the integral types are the same size and if the next bit field fits
3427 into the current allocation unit without crossing the boundary imposed by the
3428 common alignment requirements of the bit fields.
3431 Handling of zero-length bitfields:
3433 MSVC interprets zero-length bitfields in the following ways:
3436 @item If a zero-length bitfield is inserted between two bitfields that would
3437 normally be coalesced, the bitfields will not be coalesced.
3444 unsigned long bf_1 : 12;
3446 unsigned long bf_2 : 12;
3450 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3451 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3453 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3454 alignment of the zero-length bitfield is greater than the member that follows it,
3455 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3475 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3476 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3477 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3480 Taking this into account, it is important to note the following:
3483 @item If a zero-length bitfield follows a normal bitfield, the type of the
3484 zero-length bitfield may affect the alignment of the structure as whole. For
3485 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3486 normal bitfield, and is of type short.
3488 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3489 still affect the alignment of the structure:
3499 Here, @code{t4} will take up 4 bytes.
3502 @item Zero-length bitfields following non-bitfield members are ignored:
3513 Here, @code{t5} will take up 2 bytes.
3517 @subsection PowerPC Variable Attributes
3519 Three attributes currently are defined for PowerPC configurations:
3520 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3522 For full documentation of the struct attributes please see the
3523 documentation in the @xref{i386 Variable Attributes}, section.
3525 For documentation of @code{altivec} attribute please see the
3526 documentation in the @xref{PowerPC Type Attributes}, section.
3528 @subsection SPU Variable Attributes
3530 The SPU supports the @code{spu_vector} attribute for variables. For
3531 documentation of this attribute please see the documentation in the
3532 @xref{SPU Type Attributes}, section.
3534 @subsection Xstormy16 Variable Attributes
3536 One attribute is currently defined for xstormy16 configurations:
3541 @cindex @code{below100} attribute
3543 If a variable has the @code{below100} attribute (@code{BELOW100} is
3544 allowed also), GCC will place the variable in the first 0x100 bytes of
3545 memory and use special opcodes to access it. Such variables will be
3546 placed in either the @code{.bss_below100} section or the
3547 @code{.data_below100} section.
3551 @node Type Attributes
3552 @section Specifying Attributes of Types
3553 @cindex attribute of types
3554 @cindex type attributes
3556 The keyword @code{__attribute__} allows you to specify special
3557 attributes of @code{struct} and @code{union} types when you define
3558 such types. This keyword is followed by an attribute specification
3559 inside double parentheses. Seven attributes are currently defined for
3560 types: @code{aligned}, @code{packed}, @code{transparent_union},
3561 @code{unused}, @code{deprecated}, @code{visibility}, and
3562 @code{may_alias}. Other attributes are defined for functions
3563 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3566 You may also specify any one of these attributes with @samp{__}
3567 preceding and following its keyword. This allows you to use these
3568 attributes in header files without being concerned about a possible
3569 macro of the same name. For example, you may use @code{__aligned__}
3570 instead of @code{aligned}.
3572 You may specify type attributes either in a @code{typedef} declaration
3573 or in an enum, struct or union type declaration or definition.
3575 For an enum, struct or union type, you may specify attributes either
3576 between the enum, struct or union tag and the name of the type, or
3577 just past the closing curly brace of the @emph{definition}. The
3578 former syntax is preferred.
3580 @xref{Attribute Syntax}, for details of the exact syntax for using
3584 @cindex @code{aligned} attribute
3585 @item aligned (@var{alignment})
3586 This attribute specifies a minimum alignment (in bytes) for variables
3587 of the specified type. For example, the declarations:
3590 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3591 typedef int more_aligned_int __attribute__ ((aligned (8)));
3595 force the compiler to insure (as far as it can) that each variable whose
3596 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3597 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3598 variables of type @code{struct S} aligned to 8-byte boundaries allows
3599 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3600 store) instructions when copying one variable of type @code{struct S} to
3601 another, thus improving run-time efficiency.
3603 Note that the alignment of any given @code{struct} or @code{union} type
3604 is required by the ISO C standard to be at least a perfect multiple of
3605 the lowest common multiple of the alignments of all of the members of
3606 the @code{struct} or @code{union} in question. This means that you @emph{can}
3607 effectively adjust the alignment of a @code{struct} or @code{union}
3608 type by attaching an @code{aligned} attribute to any one of the members
3609 of such a type, but the notation illustrated in the example above is a
3610 more obvious, intuitive, and readable way to request the compiler to
3611 adjust the alignment of an entire @code{struct} or @code{union} type.
3613 As in the preceding example, you can explicitly specify the alignment
3614 (in bytes) that you wish the compiler to use for a given @code{struct}
3615 or @code{union} type. Alternatively, you can leave out the alignment factor
3616 and just ask the compiler to align a type to the maximum
3617 useful alignment for the target machine you are compiling for. For
3618 example, you could write:
3621 struct S @{ short f[3]; @} __attribute__ ((aligned));
3624 Whenever you leave out the alignment factor in an @code{aligned}
3625 attribute specification, the compiler automatically sets the alignment
3626 for the type to the largest alignment which is ever used for any data
3627 type on the target machine you are compiling for. Doing this can often
3628 make copy operations more efficient, because the compiler can use
3629 whatever instructions copy the biggest chunks of memory when performing
3630 copies to or from the variables which have types that you have aligned
3633 In the example above, if the size of each @code{short} is 2 bytes, then
3634 the size of the entire @code{struct S} type is 6 bytes. The smallest
3635 power of two which is greater than or equal to that is 8, so the
3636 compiler sets the alignment for the entire @code{struct S} type to 8
3639 Note that although you can ask the compiler to select a time-efficient
3640 alignment for a given type and then declare only individual stand-alone
3641 objects of that type, the compiler's ability to select a time-efficient
3642 alignment is primarily useful only when you plan to create arrays of
3643 variables having the relevant (efficiently aligned) type. If you
3644 declare or use arrays of variables of an efficiently-aligned type, then
3645 it is likely that your program will also be doing pointer arithmetic (or
3646 subscripting, which amounts to the same thing) on pointers to the
3647 relevant type, and the code that the compiler generates for these
3648 pointer arithmetic operations will often be more efficient for
3649 efficiently-aligned types than for other types.
3651 The @code{aligned} attribute can only increase the alignment; but you
3652 can decrease it by specifying @code{packed} as well. See below.
3654 Note that the effectiveness of @code{aligned} attributes may be limited
3655 by inherent limitations in your linker. On many systems, the linker is
3656 only able to arrange for variables to be aligned up to a certain maximum
3657 alignment. (For some linkers, the maximum supported alignment may
3658 be very very small.) If your linker is only able to align variables
3659 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3660 in an @code{__attribute__} will still only provide you with 8 byte
3661 alignment. See your linker documentation for further information.
3664 This attribute, attached to @code{struct} or @code{union} type
3665 definition, specifies that each member (other than zero-width bitfields)
3666 of the structure or union is placed to minimize the memory required. When
3667 attached to an @code{enum} definition, it indicates that the smallest
3668 integral type should be used.
3670 @opindex fshort-enums
3671 Specifying this attribute for @code{struct} and @code{union} types is
3672 equivalent to specifying the @code{packed} attribute on each of the
3673 structure or union members. Specifying the @option{-fshort-enums}
3674 flag on the line is equivalent to specifying the @code{packed}
3675 attribute on all @code{enum} definitions.
3677 In the following example @code{struct my_packed_struct}'s members are
3678 packed closely together, but the internal layout of its @code{s} member
3679 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3683 struct my_unpacked_struct
3689 struct __attribute__ ((__packed__)) my_packed_struct
3693 struct my_unpacked_struct s;
3697 You may only specify this attribute on the definition of a @code{enum},
3698 @code{struct} or @code{union}, not on a @code{typedef} which does not
3699 also define the enumerated type, structure or union.
3701 @item transparent_union
3702 This attribute, attached to a @code{union} type definition, indicates
3703 that any function parameter having that union type causes calls to that
3704 function to be treated in a special way.
3706 First, the argument corresponding to a transparent union type can be of
3707 any type in the union; no cast is required. Also, if the union contains
3708 a pointer type, the corresponding argument can be a null pointer
3709 constant or a void pointer expression; and if the union contains a void
3710 pointer type, the corresponding argument can be any pointer expression.
3711 If the union member type is a pointer, qualifiers like @code{const} on
3712 the referenced type must be respected, just as with normal pointer
3715 Second, the argument is passed to the function using the calling
3716 conventions of the first member of the transparent union, not the calling
3717 conventions of the union itself. All members of the union must have the
3718 same machine representation; this is necessary for this argument passing
3721 Transparent unions are designed for library functions that have multiple
3722 interfaces for compatibility reasons. For example, suppose the
3723 @code{wait} function must accept either a value of type @code{int *} to
3724 comply with Posix, or a value of type @code{union wait *} to comply with
3725 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3726 @code{wait} would accept both kinds of arguments, but it would also
3727 accept any other pointer type and this would make argument type checking
3728 less useful. Instead, @code{<sys/wait.h>} might define the interface
3736 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3738 pid_t wait (wait_status_ptr_t);
3741 This interface allows either @code{int *} or @code{union wait *}
3742 arguments to be passed, using the @code{int *} calling convention.
3743 The program can call @code{wait} with arguments of either type:
3746 int w1 () @{ int w; return wait (&w); @}
3747 int w2 () @{ union wait w; return wait (&w); @}
3750 With this interface, @code{wait}'s implementation might look like this:
3753 pid_t wait (wait_status_ptr_t p)
3755 return waitpid (-1, p.__ip, 0);
3760 When attached to a type (including a @code{union} or a @code{struct}),
3761 this attribute means that variables of that type are meant to appear
3762 possibly unused. GCC will not produce a warning for any variables of
3763 that type, even if the variable appears to do nothing. This is often
3764 the case with lock or thread classes, which are usually defined and then
3765 not referenced, but contain constructors and destructors that have
3766 nontrivial bookkeeping functions.
3769 The @code{deprecated} attribute results in a warning if the type
3770 is used anywhere in the source file. This is useful when identifying
3771 types that are expected to be removed in a future version of a program.
3772 If possible, the warning also includes the location of the declaration
3773 of the deprecated type, to enable users to easily find further
3774 information about why the type is deprecated, or what they should do
3775 instead. Note that the warnings only occur for uses and then only
3776 if the type is being applied to an identifier that itself is not being
3777 declared as deprecated.
3780 typedef int T1 __attribute__ ((deprecated));
3784 typedef T1 T3 __attribute__ ((deprecated));
3785 T3 z __attribute__ ((deprecated));
3788 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3789 warning is issued for line 4 because T2 is not explicitly
3790 deprecated. Line 5 has no warning because T3 is explicitly
3791 deprecated. Similarly for line 6.
3793 The @code{deprecated} attribute can also be used for functions and
3794 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3797 Accesses to objects with types with this attribute are not subjected to
3798 type-based alias analysis, but are instead assumed to be able to alias
3799 any other type of objects, just like the @code{char} type. See
3800 @option{-fstrict-aliasing} for more information on aliasing issues.
3805 typedef short __attribute__((__may_alias__)) short_a;
3811 short_a *b = (short_a *) &a;
3815 if (a == 0x12345678)
3822 If you replaced @code{short_a} with @code{short} in the variable
3823 declaration, the above program would abort when compiled with
3824 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3825 above in recent GCC versions.
3828 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3829 applied to class, struct, union and enum types. Unlike other type
3830 attributes, the attribute must appear between the initial keyword and
3831 the name of the type; it cannot appear after the body of the type.
3833 Note that the type visibility is applied to vague linkage entities
3834 associated with the class (vtable, typeinfo node, etc.). In
3835 particular, if a class is thrown as an exception in one shared object
3836 and caught in another, the class must have default visibility.
3837 Otherwise the two shared objects will be unable to use the same
3838 typeinfo node and exception handling will break.
3840 @subsection ARM Type Attributes
3842 On those ARM targets that support @code{dllimport} (such as Symbian
3843 OS), you can use the @code{notshared} attribute to indicate that the
3844 virtual table and other similar data for a class should not be
3845 exported from a DLL@. For example:
3848 class __declspec(notshared) C @{
3850 __declspec(dllimport) C();
3854 __declspec(dllexport)
3858 In this code, @code{C::C} is exported from the current DLL, but the
3859 virtual table for @code{C} is not exported. (You can use
3860 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3861 most Symbian OS code uses @code{__declspec}.)
3863 @anchor{i386 Type Attributes}
3864 @subsection i386 Type Attributes
3866 Two attributes are currently defined for i386 configurations:
3867 @code{ms_struct} and @code{gcc_struct}
3871 @cindex @code{ms_struct}
3872 @cindex @code{gcc_struct}
3874 If @code{packed} is used on a structure, or if bit-fields are used
3875 it may be that the Microsoft ABI packs them differently
3876 than GCC would normally pack them. Particularly when moving packed
3877 data between functions compiled with GCC and the native Microsoft compiler
3878 (either via function call or as data in a file), it may be necessary to access
3881 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3882 compilers to match the native Microsoft compiler.
3885 To specify multiple attributes, separate them by commas within the
3886 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3889 @anchor{PowerPC Type Attributes}
3890 @subsection PowerPC Type Attributes
3892 Three attributes currently are defined for PowerPC configurations:
3893 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3895 For full documentation of the struct attributes please see the
3896 documentation in the @xref{i386 Type Attributes}, section.
3898 The @code{altivec} attribute allows one to declare AltiVec vector data
3899 types supported by the AltiVec Programming Interface Manual. The
3900 attribute requires an argument to specify one of three vector types:
3901 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3902 and @code{bool__} (always followed by unsigned).
3905 __attribute__((altivec(vector__)))
3906 __attribute__((altivec(pixel__))) unsigned short
3907 __attribute__((altivec(bool__))) unsigned
3910 These attributes mainly are intended to support the @code{__vector},
3911 @code{__pixel}, and @code{__bool} AltiVec keywords.
3913 @anchor{SPU Type Attributes}
3914 @subsection SPU Type Attributes
3916 The SPU supports the @code{spu_vector} attribute for types. This attribute
3917 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
3918 Language Extensions Specification. It is intended to support the
3919 @code{__vector} keyword.
3923 @section An Inline Function is As Fast As a Macro
3924 @cindex inline functions
3925 @cindex integrating function code
3927 @cindex macros, inline alternative
3929 By declaring a function inline, you can direct GCC to make
3930 calls to that function faster. One way GCC can achieve this is to
3931 integrate that function's code into the code for its callers. This
3932 makes execution faster by eliminating the function-call overhead; in
3933 addition, if any of the actual argument values are constant, their
3934 known values may permit simplifications at compile time so that not
3935 all of the inline function's code needs to be included. The effect on
3936 code size is less predictable; object code may be larger or smaller
3937 with function inlining, depending on the particular case. You can
3938 also direct GCC to try to integrate all ``simple enough'' functions
3939 into their callers with the option @option{-finline-functions}.
3941 GCC implements three different semantics of declaring a function
3942 inline. One is available with @option{-std=gnu89} or
3943 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
3944 on all inline declarations, another when @option{-std=c99} or
3945 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
3946 is used when compiling C++.
3948 To declare a function inline, use the @code{inline} keyword in its
3949 declaration, like this:
3959 If you are writing a header file to be included in ISO C89 programs, write
3960 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3962 The three types of inlining behave similarly in two important cases:
3963 when the @code{inline} keyword is used on a @code{static} function,
3964 like the example above, and when a function is first declared without
3965 using the @code{inline} keyword and then is defined with
3966 @code{inline}, like this:
3969 extern int inc (int *a);
3977 In both of these common cases, the program behaves the same as if you
3978 had not used the @code{inline} keyword, except for its speed.
3980 @cindex inline functions, omission of
3981 @opindex fkeep-inline-functions
3982 When a function is both inline and @code{static}, if all calls to the
3983 function are integrated into the caller, and the function's address is
3984 never used, then the function's own assembler code is never referenced.
3985 In this case, GCC does not actually output assembler code for the
3986 function, unless you specify the option @option{-fkeep-inline-functions}.
3987 Some calls cannot be integrated for various reasons (in particular,
3988 calls that precede the function's definition cannot be integrated, and
3989 neither can recursive calls within the definition). If there is a
3990 nonintegrated call, then the function is compiled to assembler code as
3991 usual. The function must also be compiled as usual if the program
3992 refers to its address, because that can't be inlined.
3995 Note that certain usages in a function definition can make it unsuitable
3996 for inline substitution. Among these usages are: use of varargs, use of
3997 alloca, use of variable sized data types (@pxref{Variable Length}),
3998 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3999 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
4000 will warn when a function marked @code{inline} could not be substituted,
4001 and will give the reason for the failure.
4003 @cindex automatic @code{inline} for C++ member fns
4004 @cindex @code{inline} automatic for C++ member fns
4005 @cindex member fns, automatically @code{inline}
4006 @cindex C++ member fns, automatically @code{inline}
4007 @opindex fno-default-inline
4008 As required by ISO C++, GCC considers member functions defined within
4009 the body of a class to be marked inline even if they are
4010 not explicitly declared with the @code{inline} keyword. You can
4011 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4012 Options,,Options Controlling C++ Dialect}.
4014 GCC does not inline any functions when not optimizing unless you specify
4015 the @samp{always_inline} attribute for the function, like this:
4018 /* @r{Prototype.} */
4019 inline void foo (const char) __attribute__((always_inline));
4022 The remainder of this section is specific to GNU C89 inlining.
4024 @cindex non-static inline function
4025 When an inline function is not @code{static}, then the compiler must assume
4026 that there may be calls from other source files; since a global symbol can
4027 be defined only once in any program, the function must not be defined in
4028 the other source files, so the calls therein cannot be integrated.
4029 Therefore, a non-@code{static} inline function is always compiled on its
4030 own in the usual fashion.
4032 If you specify both @code{inline} and @code{extern} in the function
4033 definition, then the definition is used only for inlining. In no case
4034 is the function compiled on its own, not even if you refer to its
4035 address explicitly. Such an address becomes an external reference, as
4036 if you had only declared the function, and had not defined it.
4038 This combination of @code{inline} and @code{extern} has almost the
4039 effect of a macro. The way to use it is to put a function definition in
4040 a header file with these keywords, and put another copy of the
4041 definition (lacking @code{inline} and @code{extern}) in a library file.
4042 The definition in the header file will cause most calls to the function
4043 to be inlined. If any uses of the function remain, they will refer to
4044 the single copy in the library.
4047 @section Assembler Instructions with C Expression Operands
4048 @cindex extended @code{asm}
4049 @cindex @code{asm} expressions
4050 @cindex assembler instructions
4053 In an assembler instruction using @code{asm}, you can specify the
4054 operands of the instruction using C expressions. This means you need not
4055 guess which registers or memory locations will contain the data you want
4058 You must specify an assembler instruction template much like what
4059 appears in a machine description, plus an operand constraint string for
4062 For example, here is how to use the 68881's @code{fsinx} instruction:
4065 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4069 Here @code{angle} is the C expression for the input operand while
4070 @code{result} is that of the output operand. Each has @samp{"f"} as its
4071 operand constraint, saying that a floating point register is required.
4072 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4073 output operands' constraints must use @samp{=}. The constraints use the
4074 same language used in the machine description (@pxref{Constraints}).
4076 Each operand is described by an operand-constraint string followed by
4077 the C expression in parentheses. A colon separates the assembler
4078 template from the first output operand and another separates the last
4079 output operand from the first input, if any. Commas separate the
4080 operands within each group. The total number of operands is currently
4081 limited to 30; this limitation may be lifted in some future version of
4084 If there are no output operands but there are input operands, you must
4085 place two consecutive colons surrounding the place where the output
4088 As of GCC version 3.1, it is also possible to specify input and output
4089 operands using symbolic names which can be referenced within the
4090 assembler code. These names are specified inside square brackets
4091 preceding the constraint string, and can be referenced inside the
4092 assembler code using @code{%[@var{name}]} instead of a percentage sign
4093 followed by the operand number. Using named operands the above example
4097 asm ("fsinx %[angle],%[output]"
4098 : [output] "=f" (result)
4099 : [angle] "f" (angle));
4103 Note that the symbolic operand names have no relation whatsoever to
4104 other C identifiers. You may use any name you like, even those of
4105 existing C symbols, but you must ensure that no two operands within the same
4106 assembler construct use the same symbolic name.
4108 Output operand expressions must be lvalues; the compiler can check this.
4109 The input operands need not be lvalues. The compiler cannot check
4110 whether the operands have data types that are reasonable for the
4111 instruction being executed. It does not parse the assembler instruction
4112 template and does not know what it means or even whether it is valid
4113 assembler input. The extended @code{asm} feature is most often used for
4114 machine instructions the compiler itself does not know exist. If
4115 the output expression cannot be directly addressed (for example, it is a
4116 bit-field), your constraint must allow a register. In that case, GCC
4117 will use the register as the output of the @code{asm}, and then store
4118 that register into the output.
4120 The ordinary output operands must be write-only; GCC will assume that
4121 the values in these operands before the instruction are dead and need
4122 not be generated. Extended asm supports input-output or read-write
4123 operands. Use the constraint character @samp{+} to indicate such an
4124 operand and list it with the output operands. You should only use
4125 read-write operands when the constraints for the operand (or the
4126 operand in which only some of the bits are to be changed) allow a
4129 You may, as an alternative, logically split its function into two
4130 separate operands, one input operand and one write-only output
4131 operand. The connection between them is expressed by constraints
4132 which say they need to be in the same location when the instruction
4133 executes. You can use the same C expression for both operands, or
4134 different expressions. For example, here we write the (fictitious)
4135 @samp{combine} instruction with @code{bar} as its read-only source
4136 operand and @code{foo} as its read-write destination:
4139 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4143 The constraint @samp{"0"} for operand 1 says that it must occupy the
4144 same location as operand 0. A number in constraint is allowed only in
4145 an input operand and it must refer to an output operand.
4147 Only a number in the constraint can guarantee that one operand will be in
4148 the same place as another. The mere fact that @code{foo} is the value
4149 of both operands is not enough to guarantee that they will be in the
4150 same place in the generated assembler code. The following would not
4154 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4157 Various optimizations or reloading could cause operands 0 and 1 to be in
4158 different registers; GCC knows no reason not to do so. For example, the
4159 compiler might find a copy of the value of @code{foo} in one register and
4160 use it for operand 1, but generate the output operand 0 in a different
4161 register (copying it afterward to @code{foo}'s own address). Of course,
4162 since the register for operand 1 is not even mentioned in the assembler
4163 code, the result will not work, but GCC can't tell that.
4165 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4166 the operand number for a matching constraint. For example:
4169 asm ("cmoveq %1,%2,%[result]"
4170 : [result] "=r"(result)
4171 : "r" (test), "r"(new), "[result]"(old));
4174 Sometimes you need to make an @code{asm} operand be a specific register,
4175 but there's no matching constraint letter for that register @emph{by
4176 itself}. To force the operand into that register, use a local variable
4177 for the operand and specify the register in the variable declaration.
4178 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4179 register constraint letter that matches the register:
4182 register int *p1 asm ("r0") = @dots{};
4183 register int *p2 asm ("r1") = @dots{};
4184 register int *result asm ("r0");
4185 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4188 @anchor{Example of asm with clobbered asm reg}
4189 In the above example, beware that a register that is call-clobbered by
4190 the target ABI will be overwritten by any function call in the
4191 assignment, including library calls for arithmetic operators.
4192 Assuming it is a call-clobbered register, this may happen to @code{r0}
4193 above by the assignment to @code{p2}. If you have to use such a
4194 register, use temporary variables for expressions between the register
4199 register int *p1 asm ("r0") = @dots{};
4200 register int *p2 asm ("r1") = t1;
4201 register int *result asm ("r0");
4202 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4205 Some instructions clobber specific hard registers. To describe this,
4206 write a third colon after the input operands, followed by the names of
4207 the clobbered hard registers (given as strings). Here is a realistic
4208 example for the VAX:
4211 asm volatile ("movc3 %0,%1,%2"
4212 : /* @r{no outputs} */
4213 : "g" (from), "g" (to), "g" (count)
4214 : "r0", "r1", "r2", "r3", "r4", "r5");
4217 You may not write a clobber description in a way that overlaps with an
4218 input or output operand. For example, you may not have an operand
4219 describing a register class with one member if you mention that register
4220 in the clobber list. Variables declared to live in specific registers
4221 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4222 have no part mentioned in the clobber description.
4223 There is no way for you to specify that an input
4224 operand is modified without also specifying it as an output
4225 operand. Note that if all the output operands you specify are for this
4226 purpose (and hence unused), you will then also need to specify
4227 @code{volatile} for the @code{asm} construct, as described below, to
4228 prevent GCC from deleting the @code{asm} statement as unused.
4230 If you refer to a particular hardware register from the assembler code,
4231 you will probably have to list the register after the third colon to
4232 tell the compiler the register's value is modified. In some assemblers,
4233 the register names begin with @samp{%}; to produce one @samp{%} in the
4234 assembler code, you must write @samp{%%} in the input.
4236 If your assembler instruction can alter the condition code register, add
4237 @samp{cc} to the list of clobbered registers. GCC on some machines
4238 represents the condition codes as a specific hardware register;
4239 @samp{cc} serves to name this register. On other machines, the
4240 condition code is handled differently, and specifying @samp{cc} has no
4241 effect. But it is valid no matter what the machine.
4243 If your assembler instructions access memory in an unpredictable
4244 fashion, add @samp{memory} to the list of clobbered registers. This
4245 will cause GCC to not keep memory values cached in registers across the
4246 assembler instruction and not optimize stores or loads to that memory.
4247 You will also want to add the @code{volatile} keyword if the memory
4248 affected is not listed in the inputs or outputs of the @code{asm}, as
4249 the @samp{memory} clobber does not count as a side-effect of the
4250 @code{asm}. If you know how large the accessed memory is, you can add
4251 it as input or output but if this is not known, you should add
4252 @samp{memory}. As an example, if you access ten bytes of a string, you
4253 can use a memory input like:
4256 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4259 Note that in the following example the memory input is necessary,
4260 otherwise GCC might optimize the store to @code{x} away:
4267 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4268 "=&d" (r) : "a" (y), "m" (*y));
4273 You can put multiple assembler instructions together in a single
4274 @code{asm} template, separated by the characters normally used in assembly
4275 code for the system. A combination that works in most places is a newline
4276 to break the line, plus a tab character to move to the instruction field
4277 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4278 assembler allows semicolons as a line-breaking character. Note that some
4279 assembler dialects use semicolons to start a comment.
4280 The input operands are guaranteed not to use any of the clobbered
4281 registers, and neither will the output operands' addresses, so you can
4282 read and write the clobbered registers as many times as you like. Here
4283 is an example of multiple instructions in a template; it assumes the
4284 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4287 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4289 : "g" (from), "g" (to)
4293 Unless an output operand has the @samp{&} constraint modifier, GCC
4294 may allocate it in the same register as an unrelated input operand, on
4295 the assumption the inputs are consumed before the outputs are produced.
4296 This assumption may be false if the assembler code actually consists of
4297 more than one instruction. In such a case, use @samp{&} for each output
4298 operand that may not overlap an input. @xref{Modifiers}.
4300 If you want to test the condition code produced by an assembler
4301 instruction, you must include a branch and a label in the @code{asm}
4302 construct, as follows:
4305 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4311 This assumes your assembler supports local labels, as the GNU assembler
4312 and most Unix assemblers do.
4314 Speaking of labels, jumps from one @code{asm} to another are not
4315 supported. The compiler's optimizers do not know about these jumps, and
4316 therefore they cannot take account of them when deciding how to
4319 @cindex macros containing @code{asm}
4320 Usually the most convenient way to use these @code{asm} instructions is to
4321 encapsulate them in macros that look like functions. For example,
4325 (@{ double __value, __arg = (x); \
4326 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4331 Here the variable @code{__arg} is used to make sure that the instruction
4332 operates on a proper @code{double} value, and to accept only those
4333 arguments @code{x} which can convert automatically to a @code{double}.
4335 Another way to make sure the instruction operates on the correct data
4336 type is to use a cast in the @code{asm}. This is different from using a
4337 variable @code{__arg} in that it converts more different types. For
4338 example, if the desired type were @code{int}, casting the argument to
4339 @code{int} would accept a pointer with no complaint, while assigning the
4340 argument to an @code{int} variable named @code{__arg} would warn about
4341 using a pointer unless the caller explicitly casts it.
4343 If an @code{asm} has output operands, GCC assumes for optimization
4344 purposes the instruction has no side effects except to change the output
4345 operands. This does not mean instructions with a side effect cannot be
4346 used, but you must be careful, because the compiler may eliminate them
4347 if the output operands aren't used, or move them out of loops, or
4348 replace two with one if they constitute a common subexpression. Also,
4349 if your instruction does have a side effect on a variable that otherwise
4350 appears not to change, the old value of the variable may be reused later
4351 if it happens to be found in a register.
4353 You can prevent an @code{asm} instruction from being deleted
4354 by writing the keyword @code{volatile} after
4355 the @code{asm}. For example:
4358 #define get_and_set_priority(new) \
4360 asm volatile ("get_and_set_priority %0, %1" \
4361 : "=g" (__old) : "g" (new)); \
4366 The @code{volatile} keyword indicates that the instruction has
4367 important side-effects. GCC will not delete a volatile @code{asm} if
4368 it is reachable. (The instruction can still be deleted if GCC can
4369 prove that control-flow will never reach the location of the
4370 instruction.) Note that even a volatile @code{asm} instruction
4371 can be moved relative to other code, including across jump
4372 instructions. For example, on many targets there is a system
4373 register which can be set to control the rounding mode of
4374 floating point operations. You might try
4375 setting it with a volatile @code{asm}, like this PowerPC example:
4378 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4383 This will not work reliably, as the compiler may move the addition back
4384 before the volatile @code{asm}. To make it work you need to add an
4385 artificial dependency to the @code{asm} referencing a variable in the code
4386 you don't want moved, for example:
4389 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4393 Similarly, you can't expect a
4394 sequence of volatile @code{asm} instructions to remain perfectly
4395 consecutive. If you want consecutive output, use a single @code{asm}.
4396 Also, GCC will perform some optimizations across a volatile @code{asm}
4397 instruction; GCC does not ``forget everything'' when it encounters
4398 a volatile @code{asm} instruction the way some other compilers do.
4400 An @code{asm} instruction without any output operands will be treated
4401 identically to a volatile @code{asm} instruction.
4403 It is a natural idea to look for a way to give access to the condition
4404 code left by the assembler instruction. However, when we attempted to
4405 implement this, we found no way to make it work reliably. The problem
4406 is that output operands might need reloading, which would result in
4407 additional following ``store'' instructions. On most machines, these
4408 instructions would alter the condition code before there was time to
4409 test it. This problem doesn't arise for ordinary ``test'' and
4410 ``compare'' instructions because they don't have any output operands.
4412 For reasons similar to those described above, it is not possible to give
4413 an assembler instruction access to the condition code left by previous
4416 If you are writing a header file that should be includable in ISO C
4417 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4420 @subsection Size of an @code{asm}
4422 Some targets require that GCC track the size of each instruction used in
4423 order to generate correct code. Because the final length of an
4424 @code{asm} is only known by the assembler, GCC must make an estimate as
4425 to how big it will be. The estimate is formed by counting the number of
4426 statements in the pattern of the @code{asm} and multiplying that by the
4427 length of the longest instruction on that processor. Statements in the
4428 @code{asm} are identified by newline characters and whatever statement
4429 separator characters are supported by the assembler; on most processors
4430 this is the `@code{;}' character.
4432 Normally, GCC's estimate is perfectly adequate to ensure that correct
4433 code is generated, but it is possible to confuse the compiler if you use
4434 pseudo instructions or assembler macros that expand into multiple real
4435 instructions or if you use assembler directives that expand to more
4436 space in the object file than would be needed for a single instruction.
4437 If this happens then the assembler will produce a diagnostic saying that
4438 a label is unreachable.
4440 @subsection i386 floating point asm operands
4442 There are several rules on the usage of stack-like regs in
4443 asm_operands insns. These rules apply only to the operands that are
4448 Given a set of input regs that die in an asm_operands, it is
4449 necessary to know which are implicitly popped by the asm, and
4450 which must be explicitly popped by gcc.
4452 An input reg that is implicitly popped by the asm must be
4453 explicitly clobbered, unless it is constrained to match an
4457 For any input reg that is implicitly popped by an asm, it is
4458 necessary to know how to adjust the stack to compensate for the pop.
4459 If any non-popped input is closer to the top of the reg-stack than
4460 the implicitly popped reg, it would not be possible to know what the
4461 stack looked like---it's not clear how the rest of the stack ``slides
4464 All implicitly popped input regs must be closer to the top of
4465 the reg-stack than any input that is not implicitly popped.
4467 It is possible that if an input dies in an insn, reload might
4468 use the input reg for an output reload. Consider this example:
4471 asm ("foo" : "=t" (a) : "f" (b));
4474 This asm says that input B is not popped by the asm, and that
4475 the asm pushes a result onto the reg-stack, i.e., the stack is one
4476 deeper after the asm than it was before. But, it is possible that
4477 reload will think that it can use the same reg for both the input and
4478 the output, if input B dies in this insn.
4480 If any input operand uses the @code{f} constraint, all output reg
4481 constraints must use the @code{&} earlyclobber.
4483 The asm above would be written as
4486 asm ("foo" : "=&t" (a) : "f" (b));
4490 Some operands need to be in particular places on the stack. All
4491 output operands fall in this category---there is no other way to
4492 know which regs the outputs appear in unless the user indicates
4493 this in the constraints.
4495 Output operands must specifically indicate which reg an output
4496 appears in after an asm. @code{=f} is not allowed: the operand
4497 constraints must select a class with a single reg.
4500 Output operands may not be ``inserted'' between existing stack regs.
4501 Since no 387 opcode uses a read/write operand, all output operands
4502 are dead before the asm_operands, and are pushed by the asm_operands.
4503 It makes no sense to push anywhere but the top of the reg-stack.
4505 Output operands must start at the top of the reg-stack: output
4506 operands may not ``skip'' a reg.
4509 Some asm statements may need extra stack space for internal
4510 calculations. This can be guaranteed by clobbering stack registers
4511 unrelated to the inputs and outputs.
4515 Here are a couple of reasonable asms to want to write. This asm
4516 takes one input, which is internally popped, and produces two outputs.
4519 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4522 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4523 and replaces them with one output. The user must code the @code{st(1)}
4524 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4527 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4533 @section Controlling Names Used in Assembler Code
4534 @cindex assembler names for identifiers
4535 @cindex names used in assembler code
4536 @cindex identifiers, names in assembler code
4538 You can specify the name to be used in the assembler code for a C
4539 function or variable by writing the @code{asm} (or @code{__asm__})
4540 keyword after the declarator as follows:
4543 int foo asm ("myfoo") = 2;
4547 This specifies that the name to be used for the variable @code{foo} in
4548 the assembler code should be @samp{myfoo} rather than the usual
4551 On systems where an underscore is normally prepended to the name of a C
4552 function or variable, this feature allows you to define names for the
4553 linker that do not start with an underscore.
4555 It does not make sense to use this feature with a non-static local
4556 variable since such variables do not have assembler names. If you are
4557 trying to put the variable in a particular register, see @ref{Explicit
4558 Reg Vars}. GCC presently accepts such code with a warning, but will
4559 probably be changed to issue an error, rather than a warning, in the
4562 You cannot use @code{asm} in this way in a function @emph{definition}; but
4563 you can get the same effect by writing a declaration for the function
4564 before its definition and putting @code{asm} there, like this:
4567 extern func () asm ("FUNC");
4574 It is up to you to make sure that the assembler names you choose do not
4575 conflict with any other assembler symbols. Also, you must not use a
4576 register name; that would produce completely invalid assembler code. GCC
4577 does not as yet have the ability to store static variables in registers.
4578 Perhaps that will be added.
4580 @node Explicit Reg Vars
4581 @section Variables in Specified Registers
4582 @cindex explicit register variables
4583 @cindex variables in specified registers
4584 @cindex specified registers
4585 @cindex registers, global allocation
4587 GNU C allows you to put a few global variables into specified hardware
4588 registers. You can also specify the register in which an ordinary
4589 register variable should be allocated.
4593 Global register variables reserve registers throughout the program.
4594 This may be useful in programs such as programming language
4595 interpreters which have a couple of global variables that are accessed
4599 Local register variables in specific registers do not reserve the
4600 registers, except at the point where they are used as input or output
4601 operands in an @code{asm} statement and the @code{asm} statement itself is
4602 not deleted. The compiler's data flow analysis is capable of determining
4603 where the specified registers contain live values, and where they are
4604 available for other uses. Stores into local register variables may be deleted
4605 when they appear to be dead according to dataflow analysis. References
4606 to local register variables may be deleted or moved or simplified.
4608 These local variables are sometimes convenient for use with the extended
4609 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4610 output of the assembler instruction directly into a particular register.
4611 (This will work provided the register you specify fits the constraints
4612 specified for that operand in the @code{asm}.)
4620 @node Global Reg Vars
4621 @subsection Defining Global Register Variables
4622 @cindex global register variables
4623 @cindex registers, global variables in
4625 You can define a global register variable in GNU C like this:
4628 register int *foo asm ("a5");
4632 Here @code{a5} is the name of the register which should be used. Choose a
4633 register which is normally saved and restored by function calls on your
4634 machine, so that library routines will not clobber it.
4636 Naturally the register name is cpu-dependent, so you would need to
4637 conditionalize your program according to cpu type. The register
4638 @code{a5} would be a good choice on a 68000 for a variable of pointer
4639 type. On machines with register windows, be sure to choose a ``global''
4640 register that is not affected magically by the function call mechanism.
4642 In addition, operating systems on one type of cpu may differ in how they
4643 name the registers; then you would need additional conditionals. For
4644 example, some 68000 operating systems call this register @code{%a5}.
4646 Eventually there may be a way of asking the compiler to choose a register
4647 automatically, but first we need to figure out how it should choose and
4648 how to enable you to guide the choice. No solution is evident.
4650 Defining a global register variable in a certain register reserves that
4651 register entirely for this use, at least within the current compilation.
4652 The register will not be allocated for any other purpose in the functions
4653 in the current compilation. The register will not be saved and restored by
4654 these functions. Stores into this register are never deleted even if they
4655 would appear to be dead, but references may be deleted or moved or
4658 It is not safe to access the global register variables from signal
4659 handlers, or from more than one thread of control, because the system
4660 library routines may temporarily use the register for other things (unless
4661 you recompile them specially for the task at hand).
4663 @cindex @code{qsort}, and global register variables
4664 It is not safe for one function that uses a global register variable to
4665 call another such function @code{foo} by way of a third function
4666 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4667 different source file in which the variable wasn't declared). This is
4668 because @code{lose} might save the register and put some other value there.
4669 For example, you can't expect a global register variable to be available in
4670 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4671 might have put something else in that register. (If you are prepared to
4672 recompile @code{qsort} with the same global register variable, you can
4673 solve this problem.)
4675 If you want to recompile @code{qsort} or other source files which do not
4676 actually use your global register variable, so that they will not use that
4677 register for any other purpose, then it suffices to specify the compiler
4678 option @option{-ffixed-@var{reg}}. You need not actually add a global
4679 register declaration to their source code.
4681 A function which can alter the value of a global register variable cannot
4682 safely be called from a function compiled without this variable, because it
4683 could clobber the value the caller expects to find there on return.
4684 Therefore, the function which is the entry point into the part of the
4685 program that uses the global register variable must explicitly save and
4686 restore the value which belongs to its caller.
4688 @cindex register variable after @code{longjmp}
4689 @cindex global register after @code{longjmp}
4690 @cindex value after @code{longjmp}
4693 On most machines, @code{longjmp} will restore to each global register
4694 variable the value it had at the time of the @code{setjmp}. On some
4695 machines, however, @code{longjmp} will not change the value of global
4696 register variables. To be portable, the function that called @code{setjmp}
4697 should make other arrangements to save the values of the global register
4698 variables, and to restore them in a @code{longjmp}. This way, the same
4699 thing will happen regardless of what @code{longjmp} does.
4701 All global register variable declarations must precede all function
4702 definitions. If such a declaration could appear after function
4703 definitions, the declaration would be too late to prevent the register from
4704 being used for other purposes in the preceding functions.
4706 Global register variables may not have initial values, because an
4707 executable file has no means to supply initial contents for a register.
4709 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4710 registers, but certain library functions, such as @code{getwd}, as well
4711 as the subroutines for division and remainder, modify g3 and g4. g1 and
4712 g2 are local temporaries.
4714 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4715 Of course, it will not do to use more than a few of those.
4717 @node Local Reg Vars
4718 @subsection Specifying Registers for Local Variables
4719 @cindex local variables, specifying registers
4720 @cindex specifying registers for local variables
4721 @cindex registers for local variables
4723 You can define a local register variable with a specified register
4727 register int *foo asm ("a5");
4731 Here @code{a5} is the name of the register which should be used. Note
4732 that this is the same syntax used for defining global register
4733 variables, but for a local variable it would appear within a function.
4735 Naturally the register name is cpu-dependent, but this is not a
4736 problem, since specific registers are most often useful with explicit
4737 assembler instructions (@pxref{Extended Asm}). Both of these things
4738 generally require that you conditionalize your program according to
4741 In addition, operating systems on one type of cpu may differ in how they
4742 name the registers; then you would need additional conditionals. For
4743 example, some 68000 operating systems call this register @code{%a5}.
4745 Defining such a register variable does not reserve the register; it
4746 remains available for other uses in places where flow control determines
4747 the variable's value is not live.
4749 This option does not guarantee that GCC will generate code that has
4750 this variable in the register you specify at all times. You may not
4751 code an explicit reference to this register in the @emph{assembler
4752 instruction template} part of an @code{asm} statement and assume it will
4753 always refer to this variable. However, using the variable as an
4754 @code{asm} @emph{operand} guarantees that the specified register is used
4757 Stores into local register variables may be deleted when they appear to be dead
4758 according to dataflow analysis. References to local register variables may
4759 be deleted or moved or simplified.
4761 As for global register variables, it's recommended that you choose a
4762 register which is normally saved and restored by function calls on
4763 your machine, so that library routines will not clobber it. A common
4764 pitfall is to initialize multiple call-clobbered registers with
4765 arbitrary expressions, where a function call or library call for an
4766 arithmetic operator will overwrite a register value from a previous
4767 assignment, for example @code{r0} below:
4769 register int *p1 asm ("r0") = @dots{};
4770 register int *p2 asm ("r1") = @dots{};
4772 In those cases, a solution is to use a temporary variable for
4773 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4775 @node Alternate Keywords
4776 @section Alternate Keywords
4777 @cindex alternate keywords
4778 @cindex keywords, alternate
4780 @option{-ansi} and the various @option{-std} options disable certain
4781 keywords. This causes trouble when you want to use GNU C extensions, or
4782 a general-purpose header file that should be usable by all programs,
4783 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4784 @code{inline} are not available in programs compiled with
4785 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4786 program compiled with @option{-std=c99}). The ISO C99 keyword
4787 @code{restrict} is only available when @option{-std=gnu99} (which will
4788 eventually be the default) or @option{-std=c99} (or the equivalent
4789 @option{-std=iso9899:1999}) is used.
4791 The way to solve these problems is to put @samp{__} at the beginning and
4792 end of each problematical keyword. For example, use @code{__asm__}
4793 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4795 Other C compilers won't accept these alternative keywords; if you want to
4796 compile with another compiler, you can define the alternate keywords as
4797 macros to replace them with the customary keywords. It looks like this:
4805 @findex __extension__
4807 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4809 prevent such warnings within one expression by writing
4810 @code{__extension__} before the expression. @code{__extension__} has no
4811 effect aside from this.
4813 @node Incomplete Enums
4814 @section Incomplete @code{enum} Types
4816 You can define an @code{enum} tag without specifying its possible values.
4817 This results in an incomplete type, much like what you get if you write
4818 @code{struct foo} without describing the elements. A later declaration
4819 which does specify the possible values completes the type.
4821 You can't allocate variables or storage using the type while it is
4822 incomplete. However, you can work with pointers to that type.
4824 This extension may not be very useful, but it makes the handling of
4825 @code{enum} more consistent with the way @code{struct} and @code{union}
4828 This extension is not supported by GNU C++.
4830 @node Function Names
4831 @section Function Names as Strings
4832 @cindex @code{__func__} identifier
4833 @cindex @code{__FUNCTION__} identifier
4834 @cindex @code{__PRETTY_FUNCTION__} identifier
4836 GCC provides three magic variables which hold the name of the current
4837 function, as a string. The first of these is @code{__func__}, which
4838 is part of the C99 standard:
4841 The identifier @code{__func__} is implicitly declared by the translator
4842 as if, immediately following the opening brace of each function
4843 definition, the declaration
4846 static const char __func__[] = "function-name";
4849 appeared, where function-name is the name of the lexically-enclosing
4850 function. This name is the unadorned name of the function.
4853 @code{__FUNCTION__} is another name for @code{__func__}. Older
4854 versions of GCC recognize only this name. However, it is not
4855 standardized. For maximum portability, we recommend you use
4856 @code{__func__}, but provide a fallback definition with the
4860 #if __STDC_VERSION__ < 199901L
4862 # define __func__ __FUNCTION__
4864 # define __func__ "<unknown>"
4869 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4870 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4871 the type signature of the function as well as its bare name. For
4872 example, this program:
4876 extern int printf (char *, ...);
4883 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4884 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4902 __PRETTY_FUNCTION__ = void a::sub(int)
4905 These identifiers are not preprocessor macros. In GCC 3.3 and
4906 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4907 were treated as string literals; they could be used to initialize
4908 @code{char} arrays, and they could be concatenated with other string
4909 literals. GCC 3.4 and later treat them as variables, like
4910 @code{__func__}. In C++, @code{__FUNCTION__} and
4911 @code{__PRETTY_FUNCTION__} have always been variables.
4913 @node Return Address
4914 @section Getting the Return or Frame Address of a Function
4916 These functions may be used to get information about the callers of a
4919 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4920 This function returns the return address of the current function, or of
4921 one of its callers. The @var{level} argument is number of frames to
4922 scan up the call stack. A value of @code{0} yields the return address
4923 of the current function, a value of @code{1} yields the return address
4924 of the caller of the current function, and so forth. When inlining
4925 the expected behavior is that the function will return the address of
4926 the function that will be returned to. To work around this behavior use
4927 the @code{noinline} function attribute.
4929 The @var{level} argument must be a constant integer.
4931 On some machines it may be impossible to determine the return address of
4932 any function other than the current one; in such cases, or when the top
4933 of the stack has been reached, this function will return @code{0} or a
4934 random value. In addition, @code{__builtin_frame_address} may be used
4935 to determine if the top of the stack has been reached.
4937 This function should only be used with a nonzero argument for debugging
4941 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4942 This function is similar to @code{__builtin_return_address}, but it
4943 returns the address of the function frame rather than the return address
4944 of the function. Calling @code{__builtin_frame_address} with a value of
4945 @code{0} yields the frame address of the current function, a value of
4946 @code{1} yields the frame address of the caller of the current function,
4949 The frame is the area on the stack which holds local variables and saved
4950 registers. The frame address is normally the address of the first word
4951 pushed on to the stack by the function. However, the exact definition
4952 depends upon the processor and the calling convention. If the processor
4953 has a dedicated frame pointer register, and the function has a frame,
4954 then @code{__builtin_frame_address} will return the value of the frame
4957 On some machines it may be impossible to determine the frame address of
4958 any function other than the current one; in such cases, or when the top
4959 of the stack has been reached, this function will return @code{0} if
4960 the first frame pointer is properly initialized by the startup code.
4962 This function should only be used with a nonzero argument for debugging
4966 @node Vector Extensions
4967 @section Using vector instructions through built-in functions
4969 On some targets, the instruction set contains SIMD vector instructions that
4970 operate on multiple values contained in one large register at the same time.
4971 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4974 The first step in using these extensions is to provide the necessary data
4975 types. This should be done using an appropriate @code{typedef}:
4978 typedef int v4si __attribute__ ((vector_size (16)));
4981 The @code{int} type specifies the base type, while the attribute specifies
4982 the vector size for the variable, measured in bytes. For example, the
4983 declaration above causes the compiler to set the mode for the @code{v4si}
4984 type to be 16 bytes wide and divided into @code{int} sized units. For
4985 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4986 corresponding mode of @code{foo} will be @acronym{V4SI}.
4988 The @code{vector_size} attribute is only applicable to integral and
4989 float scalars, although arrays, pointers, and function return values
4990 are allowed in conjunction with this construct.
4992 All the basic integer types can be used as base types, both as signed
4993 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4994 @code{long long}. In addition, @code{float} and @code{double} can be
4995 used to build floating-point vector types.
4997 Specifying a combination that is not valid for the current architecture
4998 will cause GCC to synthesize the instructions using a narrower mode.
4999 For example, if you specify a variable of type @code{V4SI} and your
5000 architecture does not allow for this specific SIMD type, GCC will
5001 produce code that uses 4 @code{SIs}.
5003 The types defined in this manner can be used with a subset of normal C
5004 operations. Currently, GCC will allow using the following operators
5005 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
5007 The operations behave like C++ @code{valarrays}. Addition is defined as
5008 the addition of the corresponding elements of the operands. For
5009 example, in the code below, each of the 4 elements in @var{a} will be
5010 added to the corresponding 4 elements in @var{b} and the resulting
5011 vector will be stored in @var{c}.
5014 typedef int v4si __attribute__ ((vector_size (16)));
5021 Subtraction, multiplication, division, and the logical operations
5022 operate in a similar manner. Likewise, the result of using the unary
5023 minus or complement operators on a vector type is a vector whose
5024 elements are the negative or complemented values of the corresponding
5025 elements in the operand.
5027 You can declare variables and use them in function calls and returns, as
5028 well as in assignments and some casts. You can specify a vector type as
5029 a return type for a function. Vector types can also be used as function
5030 arguments. It is possible to cast from one vector type to another,
5031 provided they are of the same size (in fact, you can also cast vectors
5032 to and from other datatypes of the same size).
5034 You cannot operate between vectors of different lengths or different
5035 signedness without a cast.
5037 A port that supports hardware vector operations, usually provides a set
5038 of built-in functions that can be used to operate on vectors. For
5039 example, a function to add two vectors and multiply the result by a
5040 third could look like this:
5043 v4si f (v4si a, v4si b, v4si c)
5045 v4si tmp = __builtin_addv4si (a, b);
5046 return __builtin_mulv4si (tmp, c);
5053 @findex __builtin_offsetof
5055 GCC implements for both C and C++ a syntactic extension to implement
5056 the @code{offsetof} macro.
5060 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5062 offsetof_member_designator:
5064 | offsetof_member_designator "." @code{identifier}
5065 | offsetof_member_designator "[" @code{expr} "]"
5068 This extension is sufficient such that
5071 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5074 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5075 may be dependent. In either case, @var{member} may consist of a single
5076 identifier, or a sequence of member accesses and array references.
5078 @node Atomic Builtins
5079 @section Built-in functions for atomic memory access
5081 The following builtins are intended to be compatible with those described
5082 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5083 section 7.4. As such, they depart from the normal GCC practice of using
5084 the ``__builtin_'' prefix, and further that they are overloaded such that
5085 they work on multiple types.
5087 The definition given in the Intel documentation allows only for the use of
5088 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5089 counterparts. GCC will allow any integral scalar or pointer type that is
5090 1, 2, 4 or 8 bytes in length.
5092 Not all operations are supported by all target processors. If a particular
5093 operation cannot be implemented on the target processor, a warning will be
5094 generated and a call an external function will be generated. The external
5095 function will carry the same name as the builtin, with an additional suffix
5096 @samp{_@var{n}} where @var{n} is the size of the data type.
5098 @c ??? Should we have a mechanism to suppress this warning? This is almost
5099 @c useful for implementing the operation under the control of an external
5102 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5103 no memory operand will be moved across the operation, either forward or
5104 backward. Further, instructions will be issued as necessary to prevent the
5105 processor from speculating loads across the operation and from queuing stores
5106 after the operation.
5108 All of the routines are are described in the Intel documentation to take
5109 ``an optional list of variables protected by the memory barrier''. It's
5110 not clear what is meant by that; it could mean that @emph{only} the
5111 following variables are protected, or it could mean that these variables
5112 should in addition be protected. At present GCC ignores this list and
5113 protects all variables which are globally accessible. If in the future
5114 we make some use of this list, an empty list will continue to mean all
5115 globally accessible variables.
5118 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5119 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5120 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5121 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5122 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5123 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5124 @findex __sync_fetch_and_add
5125 @findex __sync_fetch_and_sub
5126 @findex __sync_fetch_and_or
5127 @findex __sync_fetch_and_and
5128 @findex __sync_fetch_and_xor
5129 @findex __sync_fetch_and_nand
5130 These builtins perform the operation suggested by the name, and
5131 returns the value that had previously been in memory. That is,
5134 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5135 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5138 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5139 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5140 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5141 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5142 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5143 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5144 @findex __sync_add_and_fetch
5145 @findex __sync_sub_and_fetch
5146 @findex __sync_or_and_fetch
5147 @findex __sync_and_and_fetch
5148 @findex __sync_xor_and_fetch
5149 @findex __sync_nand_and_fetch
5150 These builtins perform the operation suggested by the name, and
5151 return the new value. That is,
5154 @{ *ptr @var{op}= value; return *ptr; @}
5155 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5158 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5159 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5160 @findex __sync_bool_compare_and_swap
5161 @findex __sync_val_compare_and_swap
5162 These builtins perform an atomic compare and swap. That is, if the current
5163 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5166 The ``bool'' version returns true if the comparison is successful and
5167 @var{newval} was written. The ``val'' version returns the contents
5168 of @code{*@var{ptr}} before the operation.
5170 @item __sync_synchronize (...)
5171 @findex __sync_synchronize
5172 This builtin issues a full memory barrier.
5174 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5175 @findex __sync_lock_test_and_set
5176 This builtin, as described by Intel, is not a traditional test-and-set
5177 operation, but rather an atomic exchange operation. It writes @var{value}
5178 into @code{*@var{ptr}}, and returns the previous contents of
5181 Many targets have only minimal support for such locks, and do not support
5182 a full exchange operation. In this case, a target may support reduced
5183 functionality here by which the @emph{only} valid value to store is the
5184 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5185 is implementation defined.
5187 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5188 This means that references after the builtin cannot move to (or be
5189 speculated to) before the builtin, but previous memory stores may not
5190 be globally visible yet, and previous memory loads may not yet be
5193 @item void __sync_lock_release (@var{type} *ptr, ...)
5194 @findex __sync_lock_release
5195 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5196 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5198 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5199 This means that all previous memory stores are globally visible, and all
5200 previous memory loads have been satisfied, but following memory reads
5201 are not prevented from being speculated to before the barrier.
5204 @node Object Size Checking
5205 @section Object Size Checking Builtins
5206 @findex __builtin_object_size
5207 @findex __builtin___memcpy_chk
5208 @findex __builtin___mempcpy_chk
5209 @findex __builtin___memmove_chk
5210 @findex __builtin___memset_chk
5211 @findex __builtin___strcpy_chk
5212 @findex __builtin___stpcpy_chk
5213 @findex __builtin___strncpy_chk
5214 @findex __builtin___strcat_chk
5215 @findex __builtin___strncat_chk
5216 @findex __builtin___sprintf_chk
5217 @findex __builtin___snprintf_chk
5218 @findex __builtin___vsprintf_chk
5219 @findex __builtin___vsnprintf_chk
5220 @findex __builtin___printf_chk
5221 @findex __builtin___vprintf_chk
5222 @findex __builtin___fprintf_chk
5223 @findex __builtin___vfprintf_chk
5225 GCC implements a limited buffer overflow protection mechanism
5226 that can prevent some buffer overflow attacks.
5228 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5229 is a built-in construct that returns a constant number of bytes from
5230 @var{ptr} to the end of the object @var{ptr} pointer points to
5231 (if known at compile time). @code{__builtin_object_size} never evaluates
5232 its arguments for side-effects. If there are any side-effects in them, it
5233 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5234 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5235 point to and all of them are known at compile time, the returned number
5236 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5237 0 and minimum if nonzero. If it is not possible to determine which objects
5238 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5239 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5240 for @var{type} 2 or 3.
5242 @var{type} is an integer constant from 0 to 3. If the least significant
5243 bit is clear, objects are whole variables, if it is set, a closest
5244 surrounding subobject is considered the object a pointer points to.
5245 The second bit determines if maximum or minimum of remaining bytes
5249 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5250 char *p = &var.buf1[1], *q = &var.b;
5252 /* Here the object p points to is var. */
5253 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5254 /* The subobject p points to is var.buf1. */
5255 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5256 /* The object q points to is var. */
5257 assert (__builtin_object_size (q, 0)
5258 == (char *) (&var + 1) - (char *) &var.b);
5259 /* The subobject q points to is var.b. */
5260 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5264 There are built-in functions added for many common string operation
5265 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5266 built-in is provided. This built-in has an additional last argument,
5267 which is the number of bytes remaining in object the @var{dest}
5268 argument points to or @code{(size_t) -1} if the size is not known.
5270 The built-in functions are optimized into the normal string functions
5271 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5272 it is known at compile time that the destination object will not
5273 be overflown. If the compiler can determine at compile time the
5274 object will be always overflown, it issues a warning.
5276 The intended use can be e.g.
5280 #define bos0(dest) __builtin_object_size (dest, 0)
5281 #define memcpy(dest, src, n) \
5282 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5286 /* It is unknown what object p points to, so this is optimized
5287 into plain memcpy - no checking is possible. */
5288 memcpy (p, "abcde", n);
5289 /* Destination is known and length too. It is known at compile
5290 time there will be no overflow. */
5291 memcpy (&buf[5], "abcde", 5);
5292 /* Destination is known, but the length is not known at compile time.
5293 This will result in __memcpy_chk call that can check for overflow
5295 memcpy (&buf[5], "abcde", n);
5296 /* Destination is known and it is known at compile time there will
5297 be overflow. There will be a warning and __memcpy_chk call that
5298 will abort the program at runtime. */
5299 memcpy (&buf[6], "abcde", 5);
5302 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5303 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5304 @code{strcat} and @code{strncat}.
5306 There are also checking built-in functions for formatted output functions.
5308 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5309 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5310 const char *fmt, ...);
5311 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5313 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5314 const char *fmt, va_list ap);
5317 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5318 etc. functions and can contain implementation specific flags on what
5319 additional security measures the checking function might take, such as
5320 handling @code{%n} differently.
5322 The @var{os} argument is the object size @var{s} points to, like in the
5323 other built-in functions. There is a small difference in the behavior
5324 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5325 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5326 the checking function is called with @var{os} argument set to
5329 In addition to this, there are checking built-in functions
5330 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5331 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5332 These have just one additional argument, @var{flag}, right before
5333 format string @var{fmt}. If the compiler is able to optimize them to
5334 @code{fputc} etc. functions, it will, otherwise the checking function
5335 should be called and the @var{flag} argument passed to it.
5337 @node Other Builtins
5338 @section Other built-in functions provided by GCC
5339 @cindex built-in functions
5340 @findex __builtin_isgreater
5341 @findex __builtin_isgreaterequal
5342 @findex __builtin_isless
5343 @findex __builtin_islessequal
5344 @findex __builtin_islessgreater
5345 @findex __builtin_isunordered
5346 @findex __builtin_powi
5347 @findex __builtin_powif
5348 @findex __builtin_powil
5506 @findex fprintf_unlocked
5508 @findex fputs_unlocked
5619 @findex printf_unlocked
5651 @findex significandf
5652 @findex significandl
5723 GCC provides a large number of built-in functions other than the ones
5724 mentioned above. Some of these are for internal use in the processing
5725 of exceptions or variable-length argument lists and will not be
5726 documented here because they may change from time to time; we do not
5727 recommend general use of these functions.
5729 The remaining functions are provided for optimization purposes.
5731 @opindex fno-builtin
5732 GCC includes built-in versions of many of the functions in the standard
5733 C library. The versions prefixed with @code{__builtin_} will always be
5734 treated as having the same meaning as the C library function even if you
5735 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5736 Many of these functions are only optimized in certain cases; if they are
5737 not optimized in a particular case, a call to the library function will
5742 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5743 @option{-std=c99}), the functions
5744 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5745 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5746 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5747 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5748 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5749 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5750 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5751 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5752 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5753 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5754 @code{signbitd32}, @code{signbitd64}, @code{signbitd128},
5755 @code{significandf}, @code{significandl}, @code{significand},
5756 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5757 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5758 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5759 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5760 @code{ynl} and @code{yn}
5761 may be handled as built-in functions.
5762 All these functions have corresponding versions
5763 prefixed with @code{__builtin_}, which may be used even in strict C89
5766 The ISO C99 functions
5767 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5768 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5769 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5770 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5771 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5772 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5773 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5774 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5775 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5776 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5777 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5778 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5779 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5780 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5781 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5782 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5783 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5784 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5785 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5786 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5787 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5788 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5789 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5790 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5791 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5792 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5793 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5794 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5795 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5796 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5797 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5798 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5799 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5800 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5801 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5802 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5803 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5804 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5805 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5806 are handled as built-in functions
5807 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5809 There are also built-in versions of the ISO C99 functions
5810 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5811 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5812 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5813 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5814 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5815 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5816 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5817 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5818 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5819 that are recognized in any mode since ISO C90 reserves these names for
5820 the purpose to which ISO C99 puts them. All these functions have
5821 corresponding versions prefixed with @code{__builtin_}.
5823 The ISO C94 functions
5824 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5825 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5826 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5828 are handled as built-in functions
5829 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5831 The ISO C90 functions
5832 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5833 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5834 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5835 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5836 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5837 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5838 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5839 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5840 @code{malloc}, @code{memchr}, @code{memcmp}, @code{memcpy},
5841 @code{memset}, @code{modf}, @code{pow}, @code{printf}, @code{putchar},
5842 @code{puts}, @code{scanf}, @code{sinh}, @code{sin}, @code{snprintf},
5843 @code{sprintf}, @code{sqrt}, @code{sscanf}, @code{strcat},
5844 @code{strchr}, @code{strcmp}, @code{strcpy}, @code{strcspn},
5845 @code{strlen}, @code{strncat}, @code{strncmp}, @code{strncpy},
5846 @code{strpbrk}, @code{strrchr}, @code{strspn}, @code{strstr},
5847 @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf} and @code{vsprintf}
5848 are all recognized as built-in functions unless
5849 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5850 is specified for an individual function). All of these functions have
5851 corresponding versions prefixed with @code{__builtin_}.
5853 GCC provides built-in versions of the ISO C99 floating point comparison
5854 macros that avoid raising exceptions for unordered operands. They have
5855 the same names as the standard macros ( @code{isgreater},
5856 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5857 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5858 prefixed. We intend for a library implementor to be able to simply
5859 @code{#define} each standard macro to its built-in equivalent.
5861 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5863 You can use the built-in function @code{__builtin_types_compatible_p} to
5864 determine whether two types are the same.
5866 This built-in function returns 1 if the unqualified versions of the
5867 types @var{type1} and @var{type2} (which are types, not expressions) are
5868 compatible, 0 otherwise. The result of this built-in function can be
5869 used in integer constant expressions.
5871 This built-in function ignores top level qualifiers (e.g., @code{const},
5872 @code{volatile}). For example, @code{int} is equivalent to @code{const
5875 The type @code{int[]} and @code{int[5]} are compatible. On the other
5876 hand, @code{int} and @code{char *} are not compatible, even if the size
5877 of their types, on the particular architecture are the same. Also, the
5878 amount of pointer indirection is taken into account when determining
5879 similarity. Consequently, @code{short *} is not similar to
5880 @code{short **}. Furthermore, two types that are typedefed are
5881 considered compatible if their underlying types are compatible.
5883 An @code{enum} type is not considered to be compatible with another
5884 @code{enum} type even if both are compatible with the same integer
5885 type; this is what the C standard specifies.
5886 For example, @code{enum @{foo, bar@}} is not similar to
5887 @code{enum @{hot, dog@}}.
5889 You would typically use this function in code whose execution varies
5890 depending on the arguments' types. For example:
5895 typeof (x) tmp = (x); \
5896 if (__builtin_types_compatible_p (typeof (x), long double)) \
5897 tmp = foo_long_double (tmp); \
5898 else if (__builtin_types_compatible_p (typeof (x), double)) \
5899 tmp = foo_double (tmp); \
5900 else if (__builtin_types_compatible_p (typeof (x), float)) \
5901 tmp = foo_float (tmp); \
5908 @emph{Note:} This construct is only available for C@.
5912 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5914 You can use the built-in function @code{__builtin_choose_expr} to
5915 evaluate code depending on the value of a constant expression. This
5916 built-in function returns @var{exp1} if @var{const_exp}, which is a
5917 constant expression that must be able to be determined at compile time,
5918 is nonzero. Otherwise it returns 0.
5920 This built-in function is analogous to the @samp{? :} operator in C,
5921 except that the expression returned has its type unaltered by promotion
5922 rules. Also, the built-in function does not evaluate the expression
5923 that was not chosen. For example, if @var{const_exp} evaluates to true,
5924 @var{exp2} is not evaluated even if it has side-effects.
5926 This built-in function can return an lvalue if the chosen argument is an
5929 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5930 type. Similarly, if @var{exp2} is returned, its return type is the same
5937 __builtin_choose_expr ( \
5938 __builtin_types_compatible_p (typeof (x), double), \
5940 __builtin_choose_expr ( \
5941 __builtin_types_compatible_p (typeof (x), float), \
5943 /* @r{The void expression results in a compile-time error} \
5944 @r{when assigning the result to something.} */ \
5948 @emph{Note:} This construct is only available for C@. Furthermore, the
5949 unused expression (@var{exp1} or @var{exp2} depending on the value of
5950 @var{const_exp}) may still generate syntax errors. This may change in
5955 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5956 You can use the built-in function @code{__builtin_constant_p} to
5957 determine if a value is known to be constant at compile-time and hence
5958 that GCC can perform constant-folding on expressions involving that
5959 value. The argument of the function is the value to test. The function
5960 returns the integer 1 if the argument is known to be a compile-time
5961 constant and 0 if it is not known to be a compile-time constant. A
5962 return of 0 does not indicate that the value is @emph{not} a constant,
5963 but merely that GCC cannot prove it is a constant with the specified
5964 value of the @option{-O} option.
5966 You would typically use this function in an embedded application where
5967 memory was a critical resource. If you have some complex calculation,
5968 you may want it to be folded if it involves constants, but need to call
5969 a function if it does not. For example:
5972 #define Scale_Value(X) \
5973 (__builtin_constant_p (X) \
5974 ? ((X) * SCALE + OFFSET) : Scale (X))
5977 You may use this built-in function in either a macro or an inline
5978 function. However, if you use it in an inlined function and pass an
5979 argument of the function as the argument to the built-in, GCC will
5980 never return 1 when you call the inline function with a string constant
5981 or compound literal (@pxref{Compound Literals}) and will not return 1
5982 when you pass a constant numeric value to the inline function unless you
5983 specify the @option{-O} option.
5985 You may also use @code{__builtin_constant_p} in initializers for static
5986 data. For instance, you can write
5989 static const int table[] = @{
5990 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5996 This is an acceptable initializer even if @var{EXPRESSION} is not a
5997 constant expression. GCC must be more conservative about evaluating the
5998 built-in in this case, because it has no opportunity to perform
6001 Previous versions of GCC did not accept this built-in in data
6002 initializers. The earliest version where it is completely safe is
6006 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
6007 @opindex fprofile-arcs
6008 You may use @code{__builtin_expect} to provide the compiler with
6009 branch prediction information. In general, you should prefer to
6010 use actual profile feedback for this (@option{-fprofile-arcs}), as
6011 programmers are notoriously bad at predicting how their programs
6012 actually perform. However, there are applications in which this
6013 data is hard to collect.
6015 The return value is the value of @var{exp}, which should be an integral
6016 expression. The semantics of the built-in are that it is expected that
6017 @var{exp} == @var{c}. For example:
6020 if (__builtin_expect (x, 0))
6025 would indicate that we do not expect to call @code{foo}, since
6026 we expect @code{x} to be zero. Since you are limited to integral
6027 expressions for @var{exp}, you should use constructions such as
6030 if (__builtin_expect (ptr != NULL, 1))
6035 when testing pointer or floating-point values.
6038 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6039 This function is used to minimize cache-miss latency by moving data into
6040 a cache before it is accessed.
6041 You can insert calls to @code{__builtin_prefetch} into code for which
6042 you know addresses of data in memory that is likely to be accessed soon.
6043 If the target supports them, data prefetch instructions will be generated.
6044 If the prefetch is done early enough before the access then the data will
6045 be in the cache by the time it is accessed.
6047 The value of @var{addr} is the address of the memory to prefetch.
6048 There are two optional arguments, @var{rw} and @var{locality}.
6049 The value of @var{rw} is a compile-time constant one or zero; one
6050 means that the prefetch is preparing for a write to the memory address
6051 and zero, the default, means that the prefetch is preparing for a read.
6052 The value @var{locality} must be a compile-time constant integer between
6053 zero and three. A value of zero means that the data has no temporal
6054 locality, so it need not be left in the cache after the access. A value
6055 of three means that the data has a high degree of temporal locality and
6056 should be left in all levels of cache possible. Values of one and two
6057 mean, respectively, a low or moderate degree of temporal locality. The
6061 for (i = 0; i < n; i++)
6064 __builtin_prefetch (&a[i+j], 1, 1);
6065 __builtin_prefetch (&b[i+j], 0, 1);
6070 Data prefetch does not generate faults if @var{addr} is invalid, but
6071 the address expression itself must be valid. For example, a prefetch
6072 of @code{p->next} will not fault if @code{p->next} is not a valid
6073 address, but evaluation will fault if @code{p} is not a valid address.
6075 If the target does not support data prefetch, the address expression
6076 is evaluated if it includes side effects but no other code is generated
6077 and GCC does not issue a warning.
6080 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6081 Returns a positive infinity, if supported by the floating-point format,
6082 else @code{DBL_MAX}. This function is suitable for implementing the
6083 ISO C macro @code{HUGE_VAL}.
6086 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6087 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6090 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6091 Similar to @code{__builtin_huge_val}, except the return
6092 type is @code{long double}.
6095 @deftypefn {Built-in Function} double __builtin_inf (void)
6096 Similar to @code{__builtin_huge_val}, except a warning is generated
6097 if the target floating-point format does not support infinities.
6100 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6101 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6104 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6105 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6108 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6109 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6112 @deftypefn {Built-in Function} float __builtin_inff (void)
6113 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6114 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6117 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6118 Similar to @code{__builtin_inf}, except the return
6119 type is @code{long double}.
6122 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6123 This is an implementation of the ISO C99 function @code{nan}.
6125 Since ISO C99 defines this function in terms of @code{strtod}, which we
6126 do not implement, a description of the parsing is in order. The string
6127 is parsed as by @code{strtol}; that is, the base is recognized by
6128 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6129 in the significand such that the least significant bit of the number
6130 is at the least significant bit of the significand. The number is
6131 truncated to fit the significand field provided. The significand is
6132 forced to be a quiet NaN@.
6134 This function, if given a string literal all of which would have been
6135 consumed by strtol, is evaluated early enough that it is considered a
6136 compile-time constant.
6139 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6140 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6143 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6144 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6147 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6148 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6151 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6152 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6155 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6156 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6159 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6160 Similar to @code{__builtin_nan}, except the significand is forced
6161 to be a signaling NaN@. The @code{nans} function is proposed by
6162 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6165 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6166 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6169 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6170 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6173 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6174 Returns one plus the index of the least significant 1-bit of @var{x}, or
6175 if @var{x} is zero, returns zero.
6178 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6179 Returns the number of leading 0-bits in @var{x}, starting at the most
6180 significant bit position. If @var{x} is 0, the result is undefined.
6183 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6184 Returns the number of trailing 0-bits in @var{x}, starting at the least
6185 significant bit position. If @var{x} is 0, the result is undefined.
6188 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6189 Returns the number of 1-bits in @var{x}.
6192 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6193 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6197 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6198 Similar to @code{__builtin_ffs}, except the argument type is
6199 @code{unsigned long}.
6202 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6203 Similar to @code{__builtin_clz}, except the argument type is
6204 @code{unsigned long}.
6207 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6208 Similar to @code{__builtin_ctz}, except the argument type is
6209 @code{unsigned long}.
6212 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6213 Similar to @code{__builtin_popcount}, except the argument type is
6214 @code{unsigned long}.
6217 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6218 Similar to @code{__builtin_parity}, except the argument type is
6219 @code{unsigned long}.
6222 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6223 Similar to @code{__builtin_ffs}, except the argument type is
6224 @code{unsigned long long}.
6227 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6228 Similar to @code{__builtin_clz}, except the argument type is
6229 @code{unsigned long long}.
6232 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6233 Similar to @code{__builtin_ctz}, except the argument type is
6234 @code{unsigned long long}.
6237 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6238 Similar to @code{__builtin_popcount}, except the argument type is
6239 @code{unsigned long long}.
6242 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6243 Similar to @code{__builtin_parity}, except the argument type is
6244 @code{unsigned long long}.
6247 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6248 Returns the first argument raised to the power of the second. Unlike the
6249 @code{pow} function no guarantees about precision and rounding are made.
6252 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6253 Similar to @code{__builtin_powi}, except the argument and return types
6257 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6258 Similar to @code{__builtin_powi}, except the argument and return types
6259 are @code{long double}.
6262 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6263 Returns @var{x} with the order of the bytes reversed; for example,
6264 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6268 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6269 Similar to @code{__builtin_bswap32}, except the argument and return types
6273 @node Target Builtins
6274 @section Built-in Functions Specific to Particular Target Machines
6276 On some target machines, GCC supports many built-in functions specific
6277 to those machines. Generally these generate calls to specific machine
6278 instructions, but allow the compiler to schedule those calls.
6281 * Alpha Built-in Functions::
6282 * ARM Built-in Functions::
6283 * Blackfin Built-in Functions::
6284 * FR-V Built-in Functions::
6285 * X86 Built-in Functions::
6286 * MIPS DSP Built-in Functions::
6287 * MIPS Paired-Single Support::
6288 * PowerPC AltiVec Built-in Functions::
6289 * SPARC VIS Built-in Functions::
6290 * SPU Built-in Functions::
6293 @node Alpha Built-in Functions
6294 @subsection Alpha Built-in Functions
6296 These built-in functions are available for the Alpha family of
6297 processors, depending on the command-line switches used.
6299 The following built-in functions are always available. They
6300 all generate the machine instruction that is part of the name.
6303 long __builtin_alpha_implver (void)
6304 long __builtin_alpha_rpcc (void)
6305 long __builtin_alpha_amask (long)
6306 long __builtin_alpha_cmpbge (long, long)
6307 long __builtin_alpha_extbl (long, long)
6308 long __builtin_alpha_extwl (long, long)
6309 long __builtin_alpha_extll (long, long)
6310 long __builtin_alpha_extql (long, long)
6311 long __builtin_alpha_extwh (long, long)
6312 long __builtin_alpha_extlh (long, long)
6313 long __builtin_alpha_extqh (long, long)
6314 long __builtin_alpha_insbl (long, long)
6315 long __builtin_alpha_inswl (long, long)
6316 long __builtin_alpha_insll (long, long)
6317 long __builtin_alpha_insql (long, long)
6318 long __builtin_alpha_inswh (long, long)
6319 long __builtin_alpha_inslh (long, long)
6320 long __builtin_alpha_insqh (long, long)
6321 long __builtin_alpha_mskbl (long, long)
6322 long __builtin_alpha_mskwl (long, long)
6323 long __builtin_alpha_mskll (long, long)
6324 long __builtin_alpha_mskql (long, long)
6325 long __builtin_alpha_mskwh (long, long)
6326 long __builtin_alpha_msklh (long, long)
6327 long __builtin_alpha_mskqh (long, long)
6328 long __builtin_alpha_umulh (long, long)
6329 long __builtin_alpha_zap (long, long)
6330 long __builtin_alpha_zapnot (long, long)
6333 The following built-in functions are always with @option{-mmax}
6334 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6335 later. They all generate the machine instruction that is part
6339 long __builtin_alpha_pklb (long)
6340 long __builtin_alpha_pkwb (long)
6341 long __builtin_alpha_unpkbl (long)
6342 long __builtin_alpha_unpkbw (long)
6343 long __builtin_alpha_minub8 (long, long)
6344 long __builtin_alpha_minsb8 (long, long)
6345 long __builtin_alpha_minuw4 (long, long)
6346 long __builtin_alpha_minsw4 (long, long)
6347 long __builtin_alpha_maxub8 (long, long)
6348 long __builtin_alpha_maxsb8 (long, long)
6349 long __builtin_alpha_maxuw4 (long, long)
6350 long __builtin_alpha_maxsw4 (long, long)
6351 long __builtin_alpha_perr (long, long)
6354 The following built-in functions are always with @option{-mcix}
6355 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6356 later. They all generate the machine instruction that is part
6360 long __builtin_alpha_cttz (long)
6361 long __builtin_alpha_ctlz (long)
6362 long __builtin_alpha_ctpop (long)
6365 The following builtins are available on systems that use the OSF/1
6366 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6367 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6368 @code{rdval} and @code{wrval}.
6371 void *__builtin_thread_pointer (void)
6372 void __builtin_set_thread_pointer (void *)
6375 @node ARM Built-in Functions
6376 @subsection ARM Built-in Functions
6378 These built-in functions are available for the ARM family of
6379 processors, when the @option{-mcpu=iwmmxt} switch is used:
6382 typedef int v2si __attribute__ ((vector_size (8)));
6383 typedef short v4hi __attribute__ ((vector_size (8)));
6384 typedef char v8qi __attribute__ ((vector_size (8)));
6386 int __builtin_arm_getwcx (int)
6387 void __builtin_arm_setwcx (int, int)
6388 int __builtin_arm_textrmsb (v8qi, int)
6389 int __builtin_arm_textrmsh (v4hi, int)
6390 int __builtin_arm_textrmsw (v2si, int)
6391 int __builtin_arm_textrmub (v8qi, int)
6392 int __builtin_arm_textrmuh (v4hi, int)
6393 int __builtin_arm_textrmuw (v2si, int)
6394 v8qi __builtin_arm_tinsrb (v8qi, int)
6395 v4hi __builtin_arm_tinsrh (v4hi, int)
6396 v2si __builtin_arm_tinsrw (v2si, int)
6397 long long __builtin_arm_tmia (long long, int, int)
6398 long long __builtin_arm_tmiabb (long long, int, int)
6399 long long __builtin_arm_tmiabt (long long, int, int)
6400 long long __builtin_arm_tmiaph (long long, int, int)
6401 long long __builtin_arm_tmiatb (long long, int, int)
6402 long long __builtin_arm_tmiatt (long long, int, int)
6403 int __builtin_arm_tmovmskb (v8qi)
6404 int __builtin_arm_tmovmskh (v4hi)
6405 int __builtin_arm_tmovmskw (v2si)
6406 long long __builtin_arm_waccb (v8qi)
6407 long long __builtin_arm_wacch (v4hi)
6408 long long __builtin_arm_waccw (v2si)
6409 v8qi __builtin_arm_waddb (v8qi, v8qi)
6410 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6411 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6412 v4hi __builtin_arm_waddh (v4hi, v4hi)
6413 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6414 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6415 v2si __builtin_arm_waddw (v2si, v2si)
6416 v2si __builtin_arm_waddwss (v2si, v2si)
6417 v2si __builtin_arm_waddwus (v2si, v2si)
6418 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6419 long long __builtin_arm_wand(long long, long long)
6420 long long __builtin_arm_wandn (long long, long long)
6421 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6422 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6423 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6424 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6425 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6426 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6427 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6428 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6429 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6430 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6431 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6432 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6433 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6434 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6435 long long __builtin_arm_wmacsz (v4hi, v4hi)
6436 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6437 long long __builtin_arm_wmacuz (v4hi, v4hi)
6438 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6439 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6440 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6441 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6442 v2si __builtin_arm_wmaxsw (v2si, v2si)
6443 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6444 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6445 v2si __builtin_arm_wmaxuw (v2si, v2si)
6446 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6447 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6448 v2si __builtin_arm_wminsw (v2si, v2si)
6449 v8qi __builtin_arm_wminub (v8qi, v8qi)
6450 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6451 v2si __builtin_arm_wminuw (v2si, v2si)
6452 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6453 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6454 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6455 long long __builtin_arm_wor (long long, long long)
6456 v2si __builtin_arm_wpackdss (long long, long long)
6457 v2si __builtin_arm_wpackdus (long long, long long)
6458 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6459 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6460 v4hi __builtin_arm_wpackwss (v2si, v2si)
6461 v4hi __builtin_arm_wpackwus (v2si, v2si)
6462 long long __builtin_arm_wrord (long long, long long)
6463 long long __builtin_arm_wrordi (long long, int)
6464 v4hi __builtin_arm_wrorh (v4hi, long long)
6465 v4hi __builtin_arm_wrorhi (v4hi, int)
6466 v2si __builtin_arm_wrorw (v2si, long long)
6467 v2si __builtin_arm_wrorwi (v2si, int)
6468 v2si __builtin_arm_wsadb (v8qi, v8qi)
6469 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6470 v2si __builtin_arm_wsadh (v4hi, v4hi)
6471 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6472 v4hi __builtin_arm_wshufh (v4hi, int)
6473 long long __builtin_arm_wslld (long long, long long)
6474 long long __builtin_arm_wslldi (long long, int)
6475 v4hi __builtin_arm_wsllh (v4hi, long long)
6476 v4hi __builtin_arm_wsllhi (v4hi, int)
6477 v2si __builtin_arm_wsllw (v2si, long long)
6478 v2si __builtin_arm_wsllwi (v2si, int)
6479 long long __builtin_arm_wsrad (long long, long long)
6480 long long __builtin_arm_wsradi (long long, int)
6481 v4hi __builtin_arm_wsrah (v4hi, long long)
6482 v4hi __builtin_arm_wsrahi (v4hi, int)
6483 v2si __builtin_arm_wsraw (v2si, long long)
6484 v2si __builtin_arm_wsrawi (v2si, int)
6485 long long __builtin_arm_wsrld (long long, long long)
6486 long long __builtin_arm_wsrldi (long long, int)
6487 v4hi __builtin_arm_wsrlh (v4hi, long long)
6488 v4hi __builtin_arm_wsrlhi (v4hi, int)
6489 v2si __builtin_arm_wsrlw (v2si, long long)
6490 v2si __builtin_arm_wsrlwi (v2si, int)
6491 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6492 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6493 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6494 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6495 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6496 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6497 v2si __builtin_arm_wsubw (v2si, v2si)
6498 v2si __builtin_arm_wsubwss (v2si, v2si)
6499 v2si __builtin_arm_wsubwus (v2si, v2si)
6500 v4hi __builtin_arm_wunpckehsb (v8qi)
6501 v2si __builtin_arm_wunpckehsh (v4hi)
6502 long long __builtin_arm_wunpckehsw (v2si)
6503 v4hi __builtin_arm_wunpckehub (v8qi)
6504 v2si __builtin_arm_wunpckehuh (v4hi)
6505 long long __builtin_arm_wunpckehuw (v2si)
6506 v4hi __builtin_arm_wunpckelsb (v8qi)
6507 v2si __builtin_arm_wunpckelsh (v4hi)
6508 long long __builtin_arm_wunpckelsw (v2si)
6509 v4hi __builtin_arm_wunpckelub (v8qi)
6510 v2si __builtin_arm_wunpckeluh (v4hi)
6511 long long __builtin_arm_wunpckeluw (v2si)
6512 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6513 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6514 v2si __builtin_arm_wunpckihw (v2si, v2si)
6515 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6516 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6517 v2si __builtin_arm_wunpckilw (v2si, v2si)
6518 long long __builtin_arm_wxor (long long, long long)
6519 long long __builtin_arm_wzero ()
6522 @node Blackfin Built-in Functions
6523 @subsection Blackfin Built-in Functions
6525 Currently, there are two Blackfin-specific built-in functions. These are
6526 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6527 using inline assembly; by using these built-in functions the compiler can
6528 automatically add workarounds for hardware errata involving these
6529 instructions. These functions are named as follows:
6532 void __builtin_bfin_csync (void)
6533 void __builtin_bfin_ssync (void)
6536 @node FR-V Built-in Functions
6537 @subsection FR-V Built-in Functions
6539 GCC provides many FR-V-specific built-in functions. In general,
6540 these functions are intended to be compatible with those described
6541 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6542 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6543 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6544 pointer rather than by value.
6546 Most of the functions are named after specific FR-V instructions.
6547 Such functions are said to be ``directly mapped'' and are summarized
6548 here in tabular form.
6552 * Directly-mapped Integer Functions::
6553 * Directly-mapped Media Functions::
6554 * Raw read/write Functions::
6555 * Other Built-in Functions::
6558 @node Argument Types
6559 @subsubsection Argument Types
6561 The arguments to the built-in functions can be divided into three groups:
6562 register numbers, compile-time constants and run-time values. In order
6563 to make this classification clear at a glance, the arguments and return
6564 values are given the following pseudo types:
6566 @multitable @columnfractions .20 .30 .15 .35
6567 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6568 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6569 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6570 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6571 @item @code{uw2} @tab @code{unsigned long long} @tab No
6572 @tab an unsigned doubleword
6573 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6574 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6575 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6576 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6579 These pseudo types are not defined by GCC, they are simply a notational
6580 convenience used in this manual.
6582 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6583 and @code{sw2} are evaluated at run time. They correspond to
6584 register operands in the underlying FR-V instructions.
6586 @code{const} arguments represent immediate operands in the underlying
6587 FR-V instructions. They must be compile-time constants.
6589 @code{acc} arguments are evaluated at compile time and specify the number
6590 of an accumulator register. For example, an @code{acc} argument of 2
6591 will select the ACC2 register.
6593 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6594 number of an IACC register. See @pxref{Other Built-in Functions}
6597 @node Directly-mapped Integer Functions
6598 @subsubsection Directly-mapped Integer Functions
6600 The functions listed below map directly to FR-V I-type instructions.
6602 @multitable @columnfractions .45 .32 .23
6603 @item Function prototype @tab Example usage @tab Assembly output
6604 @item @code{sw1 __ADDSS (sw1, sw1)}
6605 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6606 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6607 @item @code{sw1 __SCAN (sw1, sw1)}
6608 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6609 @tab @code{SCAN @var{a},@var{b},@var{c}}
6610 @item @code{sw1 __SCUTSS (sw1)}
6611 @tab @code{@var{b} = __SCUTSS (@var{a})}
6612 @tab @code{SCUTSS @var{a},@var{b}}
6613 @item @code{sw1 __SLASS (sw1, sw1)}
6614 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6615 @tab @code{SLASS @var{a},@var{b},@var{c}}
6616 @item @code{void __SMASS (sw1, sw1)}
6617 @tab @code{__SMASS (@var{a}, @var{b})}
6618 @tab @code{SMASS @var{a},@var{b}}
6619 @item @code{void __SMSSS (sw1, sw1)}
6620 @tab @code{__SMSSS (@var{a}, @var{b})}
6621 @tab @code{SMSSS @var{a},@var{b}}
6622 @item @code{void __SMU (sw1, sw1)}
6623 @tab @code{__SMU (@var{a}, @var{b})}
6624 @tab @code{SMU @var{a},@var{b}}
6625 @item @code{sw2 __SMUL (sw1, sw1)}
6626 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6627 @tab @code{SMUL @var{a},@var{b},@var{c}}
6628 @item @code{sw1 __SUBSS (sw1, sw1)}
6629 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6630 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6631 @item @code{uw2 __UMUL (uw1, uw1)}
6632 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6633 @tab @code{UMUL @var{a},@var{b},@var{c}}
6636 @node Directly-mapped Media Functions
6637 @subsubsection Directly-mapped Media Functions
6639 The functions listed below map directly to FR-V M-type instructions.
6641 @multitable @columnfractions .45 .32 .23
6642 @item Function prototype @tab Example usage @tab Assembly output
6643 @item @code{uw1 __MABSHS (sw1)}
6644 @tab @code{@var{b} = __MABSHS (@var{a})}
6645 @tab @code{MABSHS @var{a},@var{b}}
6646 @item @code{void __MADDACCS (acc, acc)}
6647 @tab @code{__MADDACCS (@var{b}, @var{a})}
6648 @tab @code{MADDACCS @var{a},@var{b}}
6649 @item @code{sw1 __MADDHSS (sw1, sw1)}
6650 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6651 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6652 @item @code{uw1 __MADDHUS (uw1, uw1)}
6653 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6654 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6655 @item @code{uw1 __MAND (uw1, uw1)}
6656 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6657 @tab @code{MAND @var{a},@var{b},@var{c}}
6658 @item @code{void __MASACCS (acc, acc)}
6659 @tab @code{__MASACCS (@var{b}, @var{a})}
6660 @tab @code{MASACCS @var{a},@var{b}}
6661 @item @code{uw1 __MAVEH (uw1, uw1)}
6662 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6663 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6664 @item @code{uw2 __MBTOH (uw1)}
6665 @tab @code{@var{b} = __MBTOH (@var{a})}
6666 @tab @code{MBTOH @var{a},@var{b}}
6667 @item @code{void __MBTOHE (uw1 *, uw1)}
6668 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6669 @tab @code{MBTOHE @var{a},@var{b}}
6670 @item @code{void __MCLRACC (acc)}
6671 @tab @code{__MCLRACC (@var{a})}
6672 @tab @code{MCLRACC @var{a}}
6673 @item @code{void __MCLRACCA (void)}
6674 @tab @code{__MCLRACCA ()}
6675 @tab @code{MCLRACCA}
6676 @item @code{uw1 __Mcop1 (uw1, uw1)}
6677 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6678 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6679 @item @code{uw1 __Mcop2 (uw1, uw1)}
6680 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6681 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6682 @item @code{uw1 __MCPLHI (uw2, const)}
6683 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6684 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6685 @item @code{uw1 __MCPLI (uw2, const)}
6686 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6687 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6688 @item @code{void __MCPXIS (acc, sw1, sw1)}
6689 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6690 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6691 @item @code{void __MCPXIU (acc, uw1, uw1)}
6692 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6693 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6694 @item @code{void __MCPXRS (acc, sw1, sw1)}
6695 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6696 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6697 @item @code{void __MCPXRU (acc, uw1, uw1)}
6698 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6699 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6700 @item @code{uw1 __MCUT (acc, uw1)}
6701 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6702 @tab @code{MCUT @var{a},@var{b},@var{c}}
6703 @item @code{uw1 __MCUTSS (acc, sw1)}
6704 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6705 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6706 @item @code{void __MDADDACCS (acc, acc)}
6707 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6708 @tab @code{MDADDACCS @var{a},@var{b}}
6709 @item @code{void __MDASACCS (acc, acc)}
6710 @tab @code{__MDASACCS (@var{b}, @var{a})}
6711 @tab @code{MDASACCS @var{a},@var{b}}
6712 @item @code{uw2 __MDCUTSSI (acc, const)}
6713 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6714 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6715 @item @code{uw2 __MDPACKH (uw2, uw2)}
6716 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6717 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6718 @item @code{uw2 __MDROTLI (uw2, const)}
6719 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6720 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6721 @item @code{void __MDSUBACCS (acc, acc)}
6722 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6723 @tab @code{MDSUBACCS @var{a},@var{b}}
6724 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6725 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6726 @tab @code{MDUNPACKH @var{a},@var{b}}
6727 @item @code{uw2 __MEXPDHD (uw1, const)}
6728 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6729 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6730 @item @code{uw1 __MEXPDHW (uw1, const)}
6731 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6732 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6733 @item @code{uw1 __MHDSETH (uw1, const)}
6734 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6735 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6736 @item @code{sw1 __MHDSETS (const)}
6737 @tab @code{@var{b} = __MHDSETS (@var{a})}
6738 @tab @code{MHDSETS #@var{a},@var{b}}
6739 @item @code{uw1 __MHSETHIH (uw1, const)}
6740 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6741 @tab @code{MHSETHIH #@var{a},@var{b}}
6742 @item @code{sw1 __MHSETHIS (sw1, const)}
6743 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6744 @tab @code{MHSETHIS #@var{a},@var{b}}
6745 @item @code{uw1 __MHSETLOH (uw1, const)}
6746 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6747 @tab @code{MHSETLOH #@var{a},@var{b}}
6748 @item @code{sw1 __MHSETLOS (sw1, const)}
6749 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6750 @tab @code{MHSETLOS #@var{a},@var{b}}
6751 @item @code{uw1 __MHTOB (uw2)}
6752 @tab @code{@var{b} = __MHTOB (@var{a})}
6753 @tab @code{MHTOB @var{a},@var{b}}
6754 @item @code{void __MMACHS (acc, sw1, sw1)}
6755 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6756 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6757 @item @code{void __MMACHU (acc, uw1, uw1)}
6758 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6759 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6760 @item @code{void __MMRDHS (acc, sw1, sw1)}
6761 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6762 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6763 @item @code{void __MMRDHU (acc, uw1, uw1)}
6764 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6765 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6766 @item @code{void __MMULHS (acc, sw1, sw1)}
6767 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6768 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6769 @item @code{void __MMULHU (acc, uw1, uw1)}
6770 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6771 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6772 @item @code{void __MMULXHS (acc, sw1, sw1)}
6773 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6774 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6775 @item @code{void __MMULXHU (acc, uw1, uw1)}
6776 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6777 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6778 @item @code{uw1 __MNOT (uw1)}
6779 @tab @code{@var{b} = __MNOT (@var{a})}
6780 @tab @code{MNOT @var{a},@var{b}}
6781 @item @code{uw1 __MOR (uw1, uw1)}
6782 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6783 @tab @code{MOR @var{a},@var{b},@var{c}}
6784 @item @code{uw1 __MPACKH (uh, uh)}
6785 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6786 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6787 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6788 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6789 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6790 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6791 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6792 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6793 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6794 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6795 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6796 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6797 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6798 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6799 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6800 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6801 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6802 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6803 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6804 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6805 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6806 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6807 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6808 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6809 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6810 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6811 @item @code{void __MQMACHS (acc, sw2, sw2)}
6812 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6813 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6814 @item @code{void __MQMACHU (acc, uw2, uw2)}
6815 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6816 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6817 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6818 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6819 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6820 @item @code{void __MQMULHS (acc, sw2, sw2)}
6821 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6822 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6823 @item @code{void __MQMULHU (acc, uw2, uw2)}
6824 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6825 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6826 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6827 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6828 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6829 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6830 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6831 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6832 @item @code{sw2 __MQSATHS (sw2, sw2)}
6833 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6834 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6835 @item @code{uw2 __MQSLLHI (uw2, int)}
6836 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6837 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6838 @item @code{sw2 __MQSRAHI (sw2, int)}
6839 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6840 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6841 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6842 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6843 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6844 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6845 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6846 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6847 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6848 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6849 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6850 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6851 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6852 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6853 @item @code{uw1 __MRDACC (acc)}
6854 @tab @code{@var{b} = __MRDACC (@var{a})}
6855 @tab @code{MRDACC @var{a},@var{b}}
6856 @item @code{uw1 __MRDACCG (acc)}
6857 @tab @code{@var{b} = __MRDACCG (@var{a})}
6858 @tab @code{MRDACCG @var{a},@var{b}}
6859 @item @code{uw1 __MROTLI (uw1, const)}
6860 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6861 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6862 @item @code{uw1 __MROTRI (uw1, const)}
6863 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6864 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6865 @item @code{sw1 __MSATHS (sw1, sw1)}
6866 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6867 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6868 @item @code{uw1 __MSATHU (uw1, uw1)}
6869 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6870 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6871 @item @code{uw1 __MSLLHI (uw1, const)}
6872 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6873 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6874 @item @code{sw1 __MSRAHI (sw1, const)}
6875 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6876 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6877 @item @code{uw1 __MSRLHI (uw1, const)}
6878 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6879 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6880 @item @code{void __MSUBACCS (acc, acc)}
6881 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6882 @tab @code{MSUBACCS @var{a},@var{b}}
6883 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6884 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6885 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6886 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6887 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6888 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6889 @item @code{void __MTRAP (void)}
6890 @tab @code{__MTRAP ()}
6892 @item @code{uw2 __MUNPACKH (uw1)}
6893 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6894 @tab @code{MUNPACKH @var{a},@var{b}}
6895 @item @code{uw1 __MWCUT (uw2, uw1)}
6896 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6897 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6898 @item @code{void __MWTACC (acc, uw1)}
6899 @tab @code{__MWTACC (@var{b}, @var{a})}
6900 @tab @code{MWTACC @var{a},@var{b}}
6901 @item @code{void __MWTACCG (acc, uw1)}
6902 @tab @code{__MWTACCG (@var{b}, @var{a})}
6903 @tab @code{MWTACCG @var{a},@var{b}}
6904 @item @code{uw1 __MXOR (uw1, uw1)}
6905 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6906 @tab @code{MXOR @var{a},@var{b},@var{c}}
6909 @node Raw read/write Functions
6910 @subsubsection Raw read/write Functions
6912 This sections describes built-in functions related to read and write
6913 instructions to access memory. These functions generate
6914 @code{membar} instructions to flush the I/O load and stores where
6915 appropriate, as described in Fujitsu's manual described above.
6919 @item unsigned char __builtin_read8 (void *@var{data})
6920 @item unsigned short __builtin_read16 (void *@var{data})
6921 @item unsigned long __builtin_read32 (void *@var{data})
6922 @item unsigned long long __builtin_read64 (void *@var{data})
6924 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6925 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6926 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6927 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6930 @node Other Built-in Functions
6931 @subsubsection Other Built-in Functions
6933 This section describes built-in functions that are not named after
6934 a specific FR-V instruction.
6937 @item sw2 __IACCreadll (iacc @var{reg})
6938 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6939 for future expansion and must be 0.
6941 @item sw1 __IACCreadl (iacc @var{reg})
6942 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6943 Other values of @var{reg} are rejected as invalid.
6945 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6946 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6947 is reserved for future expansion and must be 0.
6949 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6950 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6951 is 1. Other values of @var{reg} are rejected as invalid.
6953 @item void __data_prefetch0 (const void *@var{x})
6954 Use the @code{dcpl} instruction to load the contents of address @var{x}
6955 into the data cache.
6957 @item void __data_prefetch (const void *@var{x})
6958 Use the @code{nldub} instruction to load the contents of address @var{x}
6959 into the data cache. The instruction will be issued in slot I1@.
6962 @node X86 Built-in Functions
6963 @subsection X86 Built-in Functions
6965 These built-in functions are available for the i386 and x86-64 family
6966 of computers, depending on the command-line switches used.
6968 Note that, if you specify command-line switches such as @option{-msse},
6969 the compiler could use the extended instruction sets even if the built-ins
6970 are not used explicitly in the program. For this reason, applications
6971 which perform runtime CPU detection must compile separate files for each
6972 supported architecture, using the appropriate flags. In particular,
6973 the file containing the CPU detection code should be compiled without
6976 The following machine modes are available for use with MMX built-in functions
6977 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6978 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6979 vector of eight 8-bit integers. Some of the built-in functions operate on
6980 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6982 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6983 of two 32-bit floating point values.
6985 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6986 floating point values. Some instructions use a vector of four 32-bit
6987 integers, these use @code{V4SI}. Finally, some instructions operate on an
6988 entire vector register, interpreting it as a 128-bit integer, these use mode
6991 The following built-in functions are made available by @option{-mmmx}.
6992 All of them generate the machine instruction that is part of the name.
6995 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6996 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6997 v2si __builtin_ia32_paddd (v2si, v2si)
6998 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6999 v4hi __builtin_ia32_psubw (v4hi, v4hi)
7000 v2si __builtin_ia32_psubd (v2si, v2si)
7001 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
7002 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
7003 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
7004 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
7005 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
7006 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
7007 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
7008 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
7009 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
7010 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
7011 di __builtin_ia32_pand (di, di)
7012 di __builtin_ia32_pandn (di,di)
7013 di __builtin_ia32_por (di, di)
7014 di __builtin_ia32_pxor (di, di)
7015 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
7016 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
7017 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
7018 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
7019 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
7020 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
7021 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
7022 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7023 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7024 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7025 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7026 v2si __builtin_ia32_punpckldq (v2si, v2si)
7027 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7028 v4hi __builtin_ia32_packssdw (v2si, v2si)
7029 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7032 The following built-in functions are made available either with
7033 @option{-msse}, or with a combination of @option{-m3dnow} and
7034 @option{-march=athlon}. All of them generate the machine
7035 instruction that is part of the name.
7038 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7039 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7040 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7041 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
7042 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7043 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7044 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7045 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7046 int __builtin_ia32_pextrw (v4hi, int)
7047 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7048 int __builtin_ia32_pmovmskb (v8qi)
7049 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7050 void __builtin_ia32_movntq (di *, di)
7051 void __builtin_ia32_sfence (void)
7054 The following built-in functions are available when @option{-msse} is used.
7055 All of them generate the machine instruction that is part of the name.
7058 int __builtin_ia32_comieq (v4sf, v4sf)
7059 int __builtin_ia32_comineq (v4sf, v4sf)
7060 int __builtin_ia32_comilt (v4sf, v4sf)
7061 int __builtin_ia32_comile (v4sf, v4sf)
7062 int __builtin_ia32_comigt (v4sf, v4sf)
7063 int __builtin_ia32_comige (v4sf, v4sf)
7064 int __builtin_ia32_ucomieq (v4sf, v4sf)
7065 int __builtin_ia32_ucomineq (v4sf, v4sf)
7066 int __builtin_ia32_ucomilt (v4sf, v4sf)
7067 int __builtin_ia32_ucomile (v4sf, v4sf)
7068 int __builtin_ia32_ucomigt (v4sf, v4sf)
7069 int __builtin_ia32_ucomige (v4sf, v4sf)
7070 v4sf __builtin_ia32_addps (v4sf, v4sf)
7071 v4sf __builtin_ia32_subps (v4sf, v4sf)
7072 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7073 v4sf __builtin_ia32_divps (v4sf, v4sf)
7074 v4sf __builtin_ia32_addss (v4sf, v4sf)
7075 v4sf __builtin_ia32_subss (v4sf, v4sf)
7076 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7077 v4sf __builtin_ia32_divss (v4sf, v4sf)
7078 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7079 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7080 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7081 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7082 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7083 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7084 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7085 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7086 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7087 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7088 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7089 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7090 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7091 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7092 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7093 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7094 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7095 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7096 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7097 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7098 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7099 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7100 v4sf __builtin_ia32_minps (v4sf, v4sf)
7101 v4sf __builtin_ia32_minss (v4sf, v4sf)
7102 v4sf __builtin_ia32_andps (v4sf, v4sf)
7103 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7104 v4sf __builtin_ia32_orps (v4sf, v4sf)
7105 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7106 v4sf __builtin_ia32_movss (v4sf, v4sf)
7107 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7108 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7109 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7110 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7111 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7112 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7113 v2si __builtin_ia32_cvtps2pi (v4sf)
7114 int __builtin_ia32_cvtss2si (v4sf)
7115 v2si __builtin_ia32_cvttps2pi (v4sf)
7116 int __builtin_ia32_cvttss2si (v4sf)
7117 v4sf __builtin_ia32_rcpps (v4sf)
7118 v4sf __builtin_ia32_rsqrtps (v4sf)
7119 v4sf __builtin_ia32_sqrtps (v4sf)
7120 v4sf __builtin_ia32_rcpss (v4sf)
7121 v4sf __builtin_ia32_rsqrtss (v4sf)
7122 v4sf __builtin_ia32_sqrtss (v4sf)
7123 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7124 void __builtin_ia32_movntps (float *, v4sf)
7125 int __builtin_ia32_movmskps (v4sf)
7128 The following built-in functions are available when @option{-msse} is used.
7131 @item v4sf __builtin_ia32_loadaps (float *)
7132 Generates the @code{movaps} machine instruction as a load from memory.
7133 @item void __builtin_ia32_storeaps (float *, v4sf)
7134 Generates the @code{movaps} machine instruction as a store to memory.
7135 @item v4sf __builtin_ia32_loadups (float *)
7136 Generates the @code{movups} machine instruction as a load from memory.
7137 @item void __builtin_ia32_storeups (float *, v4sf)
7138 Generates the @code{movups} machine instruction as a store to memory.
7139 @item v4sf __builtin_ia32_loadsss (float *)
7140 Generates the @code{movss} machine instruction as a load from memory.
7141 @item void __builtin_ia32_storess (float *, v4sf)
7142 Generates the @code{movss} machine instruction as a store to memory.
7143 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7144 Generates the @code{movhps} machine instruction as a load from memory.
7145 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7146 Generates the @code{movlps} machine instruction as a load from memory
7147 @item void __builtin_ia32_storehps (v4sf, v2si *)
7148 Generates the @code{movhps} machine instruction as a store to memory.
7149 @item void __builtin_ia32_storelps (v4sf, v2si *)
7150 Generates the @code{movlps} machine instruction as a store to memory.
7153 The following built-in functions are available when @option{-msse2} is used.
7154 All of them generate the machine instruction that is part of the name.
7157 int __builtin_ia32_comisdeq (v2df, v2df)
7158 int __builtin_ia32_comisdlt (v2df, v2df)
7159 int __builtin_ia32_comisdle (v2df, v2df)
7160 int __builtin_ia32_comisdgt (v2df, v2df)
7161 int __builtin_ia32_comisdge (v2df, v2df)
7162 int __builtin_ia32_comisdneq (v2df, v2df)
7163 int __builtin_ia32_ucomisdeq (v2df, v2df)
7164 int __builtin_ia32_ucomisdlt (v2df, v2df)
7165 int __builtin_ia32_ucomisdle (v2df, v2df)
7166 int __builtin_ia32_ucomisdgt (v2df, v2df)
7167 int __builtin_ia32_ucomisdge (v2df, v2df)
7168 int __builtin_ia32_ucomisdneq (v2df, v2df)
7169 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7170 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7171 v2df __builtin_ia32_cmplepd (v2df, v2df)
7172 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7173 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7174 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7175 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7176 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7177 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7178 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7179 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7180 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7181 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7182 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7183 v2df __builtin_ia32_cmplesd (v2df, v2df)
7184 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7185 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7186 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7187 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7188 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7189 v2di __builtin_ia32_paddq (v2di, v2di)
7190 v2di __builtin_ia32_psubq (v2di, v2di)
7191 v2df __builtin_ia32_addpd (v2df, v2df)
7192 v2df __builtin_ia32_subpd (v2df, v2df)
7193 v2df __builtin_ia32_mulpd (v2df, v2df)
7194 v2df __builtin_ia32_divpd (v2df, v2df)
7195 v2df __builtin_ia32_addsd (v2df, v2df)
7196 v2df __builtin_ia32_subsd (v2df, v2df)
7197 v2df __builtin_ia32_mulsd (v2df, v2df)
7198 v2df __builtin_ia32_divsd (v2df, v2df)
7199 v2df __builtin_ia32_minpd (v2df, v2df)
7200 v2df __builtin_ia32_maxpd (v2df, v2df)
7201 v2df __builtin_ia32_minsd (v2df, v2df)
7202 v2df __builtin_ia32_maxsd (v2df, v2df)
7203 v2df __builtin_ia32_andpd (v2df, v2df)
7204 v2df __builtin_ia32_andnpd (v2df, v2df)
7205 v2df __builtin_ia32_orpd (v2df, v2df)
7206 v2df __builtin_ia32_xorpd (v2df, v2df)
7207 v2df __builtin_ia32_movsd (v2df, v2df)
7208 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7209 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7210 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7211 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7212 v4si __builtin_ia32_paddd128 (v4si, v4si)
7213 v2di __builtin_ia32_paddq128 (v2di, v2di)
7214 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7215 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7216 v4si __builtin_ia32_psubd128 (v4si, v4si)
7217 v2di __builtin_ia32_psubq128 (v2di, v2di)
7218 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7219 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7220 v2di __builtin_ia32_pand128 (v2di, v2di)
7221 v2di __builtin_ia32_pandn128 (v2di, v2di)
7222 v2di __builtin_ia32_por128 (v2di, v2di)
7223 v2di __builtin_ia32_pxor128 (v2di, v2di)
7224 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7225 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7226 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7227 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7228 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7229 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7230 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7231 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7232 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7233 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7234 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7235 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7236 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7237 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7238 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7239 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7240 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7241 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7242 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7243 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7244 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7245 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7246 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7247 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7248 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7249 v2df __builtin_ia32_loadupd (double *)
7250 void __builtin_ia32_storeupd (double *, v2df)
7251 v2df __builtin_ia32_loadhpd (v2df, double *)
7252 v2df __builtin_ia32_loadlpd (v2df, double *)
7253 int __builtin_ia32_movmskpd (v2df)
7254 int __builtin_ia32_pmovmskb128 (v16qi)
7255 void __builtin_ia32_movnti (int *, int)
7256 void __builtin_ia32_movntpd (double *, v2df)
7257 void __builtin_ia32_movntdq (v2df *, v2df)
7258 v4si __builtin_ia32_pshufd (v4si, int)
7259 v8hi __builtin_ia32_pshuflw (v8hi, int)
7260 v8hi __builtin_ia32_pshufhw (v8hi, int)
7261 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7262 v2df __builtin_ia32_sqrtpd (v2df)
7263 v2df __builtin_ia32_sqrtsd (v2df)
7264 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7265 v2df __builtin_ia32_cvtdq2pd (v4si)
7266 v4sf __builtin_ia32_cvtdq2ps (v4si)
7267 v4si __builtin_ia32_cvtpd2dq (v2df)
7268 v2si __builtin_ia32_cvtpd2pi (v2df)
7269 v4sf __builtin_ia32_cvtpd2ps (v2df)
7270 v4si __builtin_ia32_cvttpd2dq (v2df)
7271 v2si __builtin_ia32_cvttpd2pi (v2df)
7272 v2df __builtin_ia32_cvtpi2pd (v2si)
7273 int __builtin_ia32_cvtsd2si (v2df)
7274 int __builtin_ia32_cvttsd2si (v2df)
7275 long long __builtin_ia32_cvtsd2si64 (v2df)
7276 long long __builtin_ia32_cvttsd2si64 (v2df)
7277 v4si __builtin_ia32_cvtps2dq (v4sf)
7278 v2df __builtin_ia32_cvtps2pd (v4sf)
7279 v4si __builtin_ia32_cvttps2dq (v4sf)
7280 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7281 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7282 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7283 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7284 void __builtin_ia32_clflush (const void *)
7285 void __builtin_ia32_lfence (void)
7286 void __builtin_ia32_mfence (void)
7287 v16qi __builtin_ia32_loaddqu (const char *)
7288 void __builtin_ia32_storedqu (char *, v16qi)
7289 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7290 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7291 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7292 v4si __builtin_ia32_pslld128 (v4si, v2di)
7293 v2di __builtin_ia32_psllq128 (v4si, v2di)
7294 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7295 v4si __builtin_ia32_psrld128 (v4si, v2di)
7296 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7297 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7298 v4si __builtin_ia32_psrad128 (v4si, v2di)
7299 v2di __builtin_ia32_pslldqi128 (v2di, int)
7300 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7301 v4si __builtin_ia32_pslldi128 (v4si, int)
7302 v2di __builtin_ia32_psllqi128 (v2di, int)
7303 v2di __builtin_ia32_psrldqi128 (v2di, int)
7304 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7305 v4si __builtin_ia32_psrldi128 (v4si, int)
7306 v2di __builtin_ia32_psrlqi128 (v2di, int)
7307 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7308 v4si __builtin_ia32_psradi128 (v4si, int)
7309 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7312 The following built-in functions are available when @option{-msse3} is used.
7313 All of them generate the machine instruction that is part of the name.
7316 v2df __builtin_ia32_addsubpd (v2df, v2df)
7317 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7318 v2df __builtin_ia32_haddpd (v2df, v2df)
7319 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7320 v2df __builtin_ia32_hsubpd (v2df, v2df)
7321 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7322 v16qi __builtin_ia32_lddqu (char const *)
7323 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7324 v2df __builtin_ia32_movddup (v2df)
7325 v4sf __builtin_ia32_movshdup (v4sf)
7326 v4sf __builtin_ia32_movsldup (v4sf)
7327 void __builtin_ia32_mwait (unsigned int, unsigned int)
7330 The following built-in functions are available when @option{-msse3} is used.
7333 @item v2df __builtin_ia32_loadddup (double const *)
7334 Generates the @code{movddup} machine instruction as a load from memory.
7337 The following built-in functions are available when @option{-mssse3} is used.
7338 All of them generate the machine instruction that is part of the name
7342 v2si __builtin_ia32_phaddd (v2si, v2si)
7343 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7344 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7345 v2si __builtin_ia32_phsubd (v2si, v2si)
7346 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7347 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7348 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7349 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7350 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7351 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7352 v2si __builtin_ia32_psignd (v2si, v2si)
7353 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7354 long long __builtin_ia32_palignr (long long, long long, int)
7355 v8qi __builtin_ia32_pabsb (v8qi)
7356 v2si __builtin_ia32_pabsd (v2si)
7357 v4hi __builtin_ia32_pabsw (v4hi)
7360 The following built-in functions are available when @option{-mssse3} is used.
7361 All of them generate the machine instruction that is part of the name
7365 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7366 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7367 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7368 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7369 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7370 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7371 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7372 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7373 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7374 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7375 v4si __builtin_ia32_psignd128 (v4si, v4si)
7376 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7377 v2di __builtin_ia32_palignr (v2di, v2di, int)
7378 v16qi __builtin_ia32_pabsb128 (v16qi)
7379 v4si __builtin_ia32_pabsd128 (v4si)
7380 v8hi __builtin_ia32_pabsw128 (v8hi)
7383 The following built-in functions are available when @option{-msse4a} is used.
7386 void _mm_stream_sd (double*,__m128d);
7387 Generates the @code{movntsd} machine instruction.
7388 void _mm_stream_ss (float*,__m128);
7389 Generates the @code{movntss} machine instruction.
7390 __m128i _mm_extract_si64 (__m128i, __m128i);
7391 Generates the @code{extrq} machine instruction with only SSE register operands.
7392 __m128i _mm_extracti_si64 (__m128i, int, int);
7393 Generates the @code{extrq} machine instruction with SSE register and immediate operands.
7394 __m128i _mm_insert_si64 (__m128i, __m128i);
7395 Generates the @code{insertq} machine instruction with only SSE register operands.
7396 __m128i _mm_inserti_si64 (__m128i, __m128i, int, int);
7397 Generates the @code{insertq} machine instruction with SSE register and immediate operands.
7400 The following built-in functions are available when @option{-m3dnow} is used.
7401 All of them generate the machine instruction that is part of the name.
7404 void __builtin_ia32_femms (void)
7405 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7406 v2si __builtin_ia32_pf2id (v2sf)
7407 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7408 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7409 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7410 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7411 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7412 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7413 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7414 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7415 v2sf __builtin_ia32_pfrcp (v2sf)
7416 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7417 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7418 v2sf __builtin_ia32_pfrsqrt (v2sf)
7419 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7420 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7421 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7422 v2sf __builtin_ia32_pi2fd (v2si)
7423 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7426 The following built-in functions are available when both @option{-m3dnow}
7427 and @option{-march=athlon} are used. All of them generate the machine
7428 instruction that is part of the name.
7431 v2si __builtin_ia32_pf2iw (v2sf)
7432 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7433 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7434 v2sf __builtin_ia32_pi2fw (v2si)
7435 v2sf __builtin_ia32_pswapdsf (v2sf)
7436 v2si __builtin_ia32_pswapdsi (v2si)
7439 @node MIPS DSP Built-in Functions
7440 @subsection MIPS DSP Built-in Functions
7442 The MIPS DSP Application-Specific Extension (ASE) includes new
7443 instructions that are designed to improve the performance of DSP and
7444 media applications. It provides instructions that operate on packed
7445 8-bit/16-bit integer data, Q7, Q15 and Q31 fractional data.
7447 GCC supports MIPS DSP operations using both the generic
7448 vector extensions (@pxref{Vector Extensions}) and a collection of
7449 MIPS-specific built-in functions. Both kinds of support are
7450 enabled by the @option{-mdsp} command-line option.
7452 Revision 2 of the ASE was introduced in the second half of 2006.
7453 This revision adds extra instructions to the original ASE, but is
7454 otherwise backwards-compatible with it. You can select revision 2
7455 using the command-line option @option{-mdspr2}; this option implies
7458 At present, GCC only provides support for operations on 32-bit
7459 vectors. The vector type associated with 8-bit integer data is
7460 usually called @code{v4i8}, the vector type associated with Q7
7461 is usually called @code{v4q7}, the vector type associated with 16-bit
7462 integer data is usually called @code{v2i16}, and the vector type
7463 associated with Q15 is usually called @code{v2q15}. They can be
7464 defined in C as follows:
7467 typedef signed char v4i8 __attribute__ ((vector_size(4)));
7468 typedef signed char v4q7 __attribute__ ((vector_size(4)));
7469 typedef short v2i16 __attribute__ ((vector_size(4)));
7470 typedef short v2q15 __attribute__ ((vector_size(4)));
7473 @code{v4i8}, @code{v4q7}, @code{v2i16} and @code{v2q15} values are
7474 initialized in the same way as aggregates. For example:
7477 v4i8 a = @{1, 2, 3, 4@};
7479 b = (v4i8) @{5, 6, 7, 8@};
7481 v2q15 c = @{0x0fcb, 0x3a75@};
7483 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7486 @emph{Note:} The CPU's endianness determines the order in which values
7487 are packed. On little-endian targets, the first value is the least
7488 significant and the last value is the most significant. The opposite
7489 order applies to big-endian targets. For example, the code above will
7490 set the lowest byte of @code{a} to @code{1} on little-endian targets
7491 and @code{4} on big-endian targets.
7493 @emph{Note:} Q7, Q15 and Q31 values must be initialized with their integer
7494 representation. As shown in this example, the integer representation
7495 of a Q7 value can be obtained by multiplying the fractional value by
7496 @code{0x1.0p7}. The equivalent for Q15 values is to multiply by
7497 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7500 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7501 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7502 and @code{c} and @code{d} are @code{v2q15} values.
7504 @multitable @columnfractions .50 .50
7505 @item C code @tab MIPS instruction
7506 @item @code{a + b} @tab @code{addu.qb}
7507 @item @code{c + d} @tab @code{addq.ph}
7508 @item @code{a - b} @tab @code{subu.qb}
7509 @item @code{c - d} @tab @code{subq.ph}
7512 The table below lists the @code{v2i16} operation for which
7513 hardware support exists for the DSP ASE REV 2. @code{e} and @code{f} are
7514 @code{v2i16} values.
7516 @multitable @columnfractions .50 .50
7517 @item C code @tab MIPS instruction
7518 @item @code{e * f} @tab @code{mul.ph}
7521 It is easier to describe the DSP built-in functions if we first define
7522 the following types:
7527 typedef unsigned int ui32;
7528 typedef long long a64;
7531 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7532 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7533 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7534 @code{long long}, but we use @code{a64} to indicate values that will
7535 be placed in one of the four DSP accumulators (@code{$ac0},
7536 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7538 Also, some built-in functions prefer or require immediate numbers as
7539 parameters, because the corresponding DSP instructions accept both immediate
7540 numbers and register operands, or accept immediate numbers only. The
7541 immediate parameters are listed as follows.
7550 imm_n32_31: -32 to 31.
7551 imm_n512_511: -512 to 511.
7554 The following built-in functions map directly to a particular MIPS DSP
7555 instruction. Please refer to the architecture specification
7556 for details on what each instruction does.
7559 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7560 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7561 q31 __builtin_mips_addq_s_w (q31, q31)
7562 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7563 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7564 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7565 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7566 q31 __builtin_mips_subq_s_w (q31, q31)
7567 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7568 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7569 i32 __builtin_mips_addsc (i32, i32)
7570 i32 __builtin_mips_addwc (i32, i32)
7571 i32 __builtin_mips_modsub (i32, i32)
7572 i32 __builtin_mips_raddu_w_qb (v4i8)
7573 v2q15 __builtin_mips_absq_s_ph (v2q15)
7574 q31 __builtin_mips_absq_s_w (q31)
7575 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7576 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7577 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7578 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7579 q31 __builtin_mips_preceq_w_phl (v2q15)
7580 q31 __builtin_mips_preceq_w_phr (v2q15)
7581 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7582 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7583 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7584 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7585 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7586 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7587 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7588 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7589 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7590 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7591 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7592 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7593 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7594 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7595 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7596 q31 __builtin_mips_shll_s_w (q31, i32)
7597 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7598 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7599 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7600 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7601 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7602 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7603 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7604 q31 __builtin_mips_shra_r_w (q31, i32)
7605 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7606 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7607 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7608 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7609 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7610 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7611 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7612 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7613 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7614 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7615 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7616 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7617 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7618 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7619 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7620 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7621 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7622 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7623 i32 __builtin_mips_bitrev (i32)
7624 i32 __builtin_mips_insv (i32, i32)
7625 v4i8 __builtin_mips_repl_qb (imm0_255)
7626 v4i8 __builtin_mips_repl_qb (i32)
7627 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7628 v2q15 __builtin_mips_repl_ph (i32)
7629 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7630 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7631 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7632 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7633 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7634 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7635 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7636 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7637 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7638 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7639 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7640 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7641 i32 __builtin_mips_extr_w (a64, imm0_31)
7642 i32 __builtin_mips_extr_w (a64, i32)
7643 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7644 i32 __builtin_mips_extr_s_h (a64, i32)
7645 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7646 i32 __builtin_mips_extr_rs_w (a64, i32)
7647 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7648 i32 __builtin_mips_extr_r_w (a64, i32)
7649 i32 __builtin_mips_extp (a64, imm0_31)
7650 i32 __builtin_mips_extp (a64, i32)
7651 i32 __builtin_mips_extpdp (a64, imm0_31)
7652 i32 __builtin_mips_extpdp (a64, i32)
7653 a64 __builtin_mips_shilo (a64, imm_n32_31)
7654 a64 __builtin_mips_shilo (a64, i32)
7655 a64 __builtin_mips_mthlip (a64, i32)
7656 void __builtin_mips_wrdsp (i32, imm0_63)
7657 i32 __builtin_mips_rddsp (imm0_63)
7658 i32 __builtin_mips_lbux (void *, i32)
7659 i32 __builtin_mips_lhx (void *, i32)
7660 i32 __builtin_mips_lwx (void *, i32)
7661 i32 __builtin_mips_bposge32 (void)
7664 The following built-in functions map directly to a particular MIPS DSP REV 2
7665 instruction. Please refer to the architecture specification
7666 for details on what each instruction does.
7669 v4q7 __builtin_mips_absq_s_qb (v4q7);
7670 v2i16 __builtin_mips_addu_ph (v2i16, v2i16);
7671 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16);
7672 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8);
7673 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8);
7674 i32 __builtin_mips_append (i32, i32, imm0_31);
7675 i32 __builtin_mips_balign (i32, i32, imm0_3);
7676 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8);
7677 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8);
7678 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8);
7679 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16);
7680 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16);
7681 a64 __builtin_mips_madd (a64, i32, i32);
7682 a64 __builtin_mips_maddu (a64, ui32, ui32);
7683 a64 __builtin_mips_msub (a64, i32, i32);
7684 a64 __builtin_mips_msubu (a64, ui32, ui32);
7685 v2i16 __builtin_mips_mul_ph (v2i16, v2i16);
7686 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16);
7687 q31 __builtin_mips_mulq_rs_w (q31, q31);
7688 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15);
7689 q31 __builtin_mips_mulq_s_w (q31, q31);
7690 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16);
7691 a64 __builtin_mips_mult (i32, i32);
7692 a64 __builtin_mips_multu (ui32, ui32);
7693 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16);
7694 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31);
7695 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31);
7696 i32 __builtin_mips_prepend (i32, i32, imm0_31);
7697 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7);
7698 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7);
7699 v4i8 __builtin_mips_shra_qb (v4i8, i32);
7700 v4i8 __builtin_mips_shra_r_qb (v4i8, i32);
7701 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15);
7702 v2i16 __builtin_mips_shrl_ph (v2i16, i32);
7703 v2i16 __builtin_mips_subu_ph (v2i16, v2i16);
7704 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16);
7705 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8);
7706 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8);
7707 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15);
7708 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15);
7709 q31 __builtin_mips_addqh_w (q31, q31);
7710 q31 __builtin_mips_addqh_r_w (q31, q31);
7711 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15);
7712 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15);
7713 q31 __builtin_mips_subqh_w (q31, q31);
7714 q31 __builtin_mips_subqh_r_w (q31, q31);
7715 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16);
7716 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16);
7717 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15);
7718 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15);
7719 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15);
7720 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15);
7724 @node MIPS Paired-Single Support
7725 @subsection MIPS Paired-Single Support
7727 The MIPS64 architecture includes a number of instructions that
7728 operate on pairs of single-precision floating-point values.
7729 Each pair is packed into a 64-bit floating-point register,
7730 with one element being designated the ``upper half'' and
7731 the other being designated the ``lower half''.
7733 GCC supports paired-single operations using both the generic
7734 vector extensions (@pxref{Vector Extensions}) and a collection of
7735 MIPS-specific built-in functions. Both kinds of support are
7736 enabled by the @option{-mpaired-single} command-line option.
7738 The vector type associated with paired-single values is usually
7739 called @code{v2sf}. It can be defined in C as follows:
7742 typedef float v2sf __attribute__ ((vector_size (8)));
7745 @code{v2sf} values are initialized in the same way as aggregates.
7749 v2sf a = @{1.5, 9.1@};
7752 b = (v2sf) @{e, f@};
7755 @emph{Note:} The CPU's endianness determines which value is stored in
7756 the upper half of a register and which value is stored in the lower half.
7757 On little-endian targets, the first value is the lower one and the second
7758 value is the upper one. The opposite order applies to big-endian targets.
7759 For example, the code above will set the lower half of @code{a} to
7760 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7763 * Paired-Single Arithmetic::
7764 * Paired-Single Built-in Functions::
7765 * MIPS-3D Built-in Functions::
7768 @node Paired-Single Arithmetic
7769 @subsubsection Paired-Single Arithmetic
7771 The table below lists the @code{v2sf} operations for which hardware
7772 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7773 values and @code{x} is an integral value.
7775 @multitable @columnfractions .50 .50
7776 @item C code @tab MIPS instruction
7777 @item @code{a + b} @tab @code{add.ps}
7778 @item @code{a - b} @tab @code{sub.ps}
7779 @item @code{-a} @tab @code{neg.ps}
7780 @item @code{a * b} @tab @code{mul.ps}
7781 @item @code{a * b + c} @tab @code{madd.ps}
7782 @item @code{a * b - c} @tab @code{msub.ps}
7783 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7784 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7785 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7788 Note that the multiply-accumulate instructions can be disabled
7789 using the command-line option @code{-mno-fused-madd}.
7791 @node Paired-Single Built-in Functions
7792 @subsubsection Paired-Single Built-in Functions
7794 The following paired-single functions map directly to a particular
7795 MIPS instruction. Please refer to the architecture specification
7796 for details on what each instruction does.
7799 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7800 Pair lower lower (@code{pll.ps}).
7802 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7803 Pair upper lower (@code{pul.ps}).
7805 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7806 Pair lower upper (@code{plu.ps}).
7808 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7809 Pair upper upper (@code{puu.ps}).
7811 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7812 Convert pair to paired single (@code{cvt.ps.s}).
7814 @item float __builtin_mips_cvt_s_pl (v2sf)
7815 Convert pair lower to single (@code{cvt.s.pl}).
7817 @item float __builtin_mips_cvt_s_pu (v2sf)
7818 Convert pair upper to single (@code{cvt.s.pu}).
7820 @item v2sf __builtin_mips_abs_ps (v2sf)
7821 Absolute value (@code{abs.ps}).
7823 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7824 Align variable (@code{alnv.ps}).
7826 @emph{Note:} The value of the third parameter must be 0 or 4
7827 modulo 8, otherwise the result will be unpredictable. Please read the
7828 instruction description for details.
7831 The following multi-instruction functions are also available.
7832 In each case, @var{cond} can be any of the 16 floating-point conditions:
7833 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7834 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7835 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7838 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7839 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7840 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7841 @code{movt.ps}/@code{movf.ps}).
7843 The @code{movt} functions return the value @var{x} computed by:
7846 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7847 mov.ps @var{x},@var{c}
7848 movt.ps @var{x},@var{d},@var{cc}
7851 The @code{movf} functions are similar but use @code{movf.ps} instead
7854 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7855 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7856 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7857 @code{bc1t}/@code{bc1f}).
7859 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7860 and return either the upper or lower half of the result. For example:
7864 if (__builtin_mips_upper_c_eq_ps (a, b))
7865 upper_halves_are_equal ();
7867 upper_halves_are_unequal ();
7869 if (__builtin_mips_lower_c_eq_ps (a, b))
7870 lower_halves_are_equal ();
7872 lower_halves_are_unequal ();
7876 @node MIPS-3D Built-in Functions
7877 @subsubsection MIPS-3D Built-in Functions
7879 The MIPS-3D Application-Specific Extension (ASE) includes additional
7880 paired-single instructions that are designed to improve the performance
7881 of 3D graphics operations. Support for these instructions is controlled
7882 by the @option{-mips3d} command-line option.
7884 The functions listed below map directly to a particular MIPS-3D
7885 instruction. Please refer to the architecture specification for
7886 more details on what each instruction does.
7889 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7890 Reduction add (@code{addr.ps}).
7892 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7893 Reduction multiply (@code{mulr.ps}).
7895 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7896 Convert paired single to paired word (@code{cvt.pw.ps}).
7898 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7899 Convert paired word to paired single (@code{cvt.ps.pw}).
7901 @item float __builtin_mips_recip1_s (float)
7902 @itemx double __builtin_mips_recip1_d (double)
7903 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7904 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7906 @item float __builtin_mips_recip2_s (float, float)
7907 @itemx double __builtin_mips_recip2_d (double, double)
7908 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7909 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7911 @item float __builtin_mips_rsqrt1_s (float)
7912 @itemx double __builtin_mips_rsqrt1_d (double)
7913 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7914 Reduced precision reciprocal square root (sequence step 1)
7915 (@code{rsqrt1.@var{fmt}}).
7917 @item float __builtin_mips_rsqrt2_s (float, float)
7918 @itemx double __builtin_mips_rsqrt2_d (double, double)
7919 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7920 Reduced precision reciprocal square root (sequence step 2)
7921 (@code{rsqrt2.@var{fmt}}).
7924 The following multi-instruction functions are also available.
7925 In each case, @var{cond} can be any of the 16 floating-point conditions:
7926 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7927 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7928 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7931 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7932 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7933 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7934 @code{bc1t}/@code{bc1f}).
7936 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7937 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7942 if (__builtin_mips_cabs_eq_s (a, b))
7948 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7949 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7950 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7951 @code{bc1t}/@code{bc1f}).
7953 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7954 and return either the upper or lower half of the result. For example:
7958 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7959 upper_halves_are_equal ();
7961 upper_halves_are_unequal ();
7963 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7964 lower_halves_are_equal ();
7966 lower_halves_are_unequal ();
7969 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7970 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7971 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7972 @code{movt.ps}/@code{movf.ps}).
7974 The @code{movt} functions return the value @var{x} computed by:
7977 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7978 mov.ps @var{x},@var{c}
7979 movt.ps @var{x},@var{d},@var{cc}
7982 The @code{movf} functions are similar but use @code{movf.ps} instead
7985 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7986 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7987 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7988 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7989 Comparison of two paired-single values
7990 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7991 @code{bc1any2t}/@code{bc1any2f}).
7993 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7994 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7995 result is true and the @code{all} forms return true if both results are true.
8000 if (__builtin_mips_any_c_eq_ps (a, b))
8005 if (__builtin_mips_all_c_eq_ps (a, b))
8011 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8012 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8013 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8014 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
8015 Comparison of four paired-single values
8016 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
8017 @code{bc1any4t}/@code{bc1any4f}).
8019 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
8020 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
8021 The @code{any} forms return true if any of the four results are true
8022 and the @code{all} forms return true if all four results are true.
8027 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
8032 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
8039 @node PowerPC AltiVec Built-in Functions
8040 @subsection PowerPC AltiVec Built-in Functions
8042 GCC provides an interface for the PowerPC family of processors to access
8043 the AltiVec operations described in Motorola's AltiVec Programming
8044 Interface Manual. The interface is made available by including
8045 @code{<altivec.h>} and using @option{-maltivec} and
8046 @option{-mabi=altivec}. The interface supports the following vector
8050 vector unsigned char
8054 vector unsigned short
8065 GCC's implementation of the high-level language interface available from
8066 C and C++ code differs from Motorola's documentation in several ways.
8071 A vector constant is a list of constant expressions within curly braces.
8074 A vector initializer requires no cast if the vector constant is of the
8075 same type as the variable it is initializing.
8078 If @code{signed} or @code{unsigned} is omitted, the signedness of the
8079 vector type is the default signedness of the base type. The default
8080 varies depending on the operating system, so a portable program should
8081 always specify the signedness.
8084 Compiling with @option{-maltivec} adds keywords @code{__vector},
8085 @code{__pixel}, and @code{__bool}. Macros @option{vector},
8086 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
8090 GCC allows using a @code{typedef} name as the type specifier for a
8094 For C, overloaded functions are implemented with macros so the following
8098 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
8101 Since @code{vec_add} is a macro, the vector constant in the example
8102 is treated as four separate arguments. Wrap the entire argument in
8103 parentheses for this to work.
8106 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
8107 Internally, GCC uses built-in functions to achieve the functionality in
8108 the aforementioned header file, but they are not supported and are
8109 subject to change without notice.
8111 The following interfaces are supported for the generic and specific
8112 AltiVec operations and the AltiVec predicates. In cases where there
8113 is a direct mapping between generic and specific operations, only the
8114 generic names are shown here, although the specific operations can also
8117 Arguments that are documented as @code{const int} require literal
8118 integral values within the range required for that operation.
8121 vector signed char vec_abs (vector signed char);
8122 vector signed short vec_abs (vector signed short);
8123 vector signed int vec_abs (vector signed int);
8124 vector float vec_abs (vector float);
8126 vector signed char vec_abss (vector signed char);
8127 vector signed short vec_abss (vector signed short);
8128 vector signed int vec_abss (vector signed int);
8130 vector signed char vec_add (vector bool char, vector signed char);
8131 vector signed char vec_add (vector signed char, vector bool char);
8132 vector signed char vec_add (vector signed char, vector signed char);
8133 vector unsigned char vec_add (vector bool char, vector unsigned char);
8134 vector unsigned char vec_add (vector unsigned char, vector bool char);
8135 vector unsigned char vec_add (vector unsigned char,
8136 vector unsigned char);
8137 vector signed short vec_add (vector bool short, vector signed short);
8138 vector signed short vec_add (vector signed short, vector bool short);
8139 vector signed short vec_add (vector signed short, vector signed short);
8140 vector unsigned short vec_add (vector bool short,
8141 vector unsigned short);
8142 vector unsigned short vec_add (vector unsigned short,
8144 vector unsigned short vec_add (vector unsigned short,
8145 vector unsigned short);
8146 vector signed int vec_add (vector bool int, vector signed int);
8147 vector signed int vec_add (vector signed int, vector bool int);
8148 vector signed int vec_add (vector signed int, vector signed int);
8149 vector unsigned int vec_add (vector bool int, vector unsigned int);
8150 vector unsigned int vec_add (vector unsigned int, vector bool int);
8151 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
8152 vector float vec_add (vector float, vector float);
8154 vector float vec_vaddfp (vector float, vector float);
8156 vector signed int vec_vadduwm (vector bool int, vector signed int);
8157 vector signed int vec_vadduwm (vector signed int, vector bool int);
8158 vector signed int vec_vadduwm (vector signed int, vector signed int);
8159 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
8160 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
8161 vector unsigned int vec_vadduwm (vector unsigned int,
8162 vector unsigned int);
8164 vector signed short vec_vadduhm (vector bool short,
8165 vector signed short);
8166 vector signed short vec_vadduhm (vector signed short,
8168 vector signed short vec_vadduhm (vector signed short,
8169 vector signed short);
8170 vector unsigned short vec_vadduhm (vector bool short,
8171 vector unsigned short);
8172 vector unsigned short vec_vadduhm (vector unsigned short,
8174 vector unsigned short vec_vadduhm (vector unsigned short,
8175 vector unsigned short);
8177 vector signed char vec_vaddubm (vector bool char, vector signed char);
8178 vector signed char vec_vaddubm (vector signed char, vector bool char);
8179 vector signed char vec_vaddubm (vector signed char, vector signed char);
8180 vector unsigned char vec_vaddubm (vector bool char,
8181 vector unsigned char);
8182 vector unsigned char vec_vaddubm (vector unsigned char,
8184 vector unsigned char vec_vaddubm (vector unsigned char,
8185 vector unsigned char);
8187 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
8189 vector unsigned char vec_adds (vector bool char, vector unsigned char);
8190 vector unsigned char vec_adds (vector unsigned char, vector bool char);
8191 vector unsigned char vec_adds (vector unsigned char,
8192 vector unsigned char);
8193 vector signed char vec_adds (vector bool char, vector signed char);
8194 vector signed char vec_adds (vector signed char, vector bool char);
8195 vector signed char vec_adds (vector signed char, vector signed char);
8196 vector unsigned short vec_adds (vector bool short,
8197 vector unsigned short);
8198 vector unsigned short vec_adds (vector unsigned short,
8200 vector unsigned short vec_adds (vector unsigned short,
8201 vector unsigned short);
8202 vector signed short vec_adds (vector bool short, vector signed short);
8203 vector signed short vec_adds (vector signed short, vector bool short);
8204 vector signed short vec_adds (vector signed short, vector signed short);
8205 vector unsigned int vec_adds (vector bool int, vector unsigned int);
8206 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8207 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8208 vector signed int vec_adds (vector bool int, vector signed int);
8209 vector signed int vec_adds (vector signed int, vector bool int);
8210 vector signed int vec_adds (vector signed int, vector signed int);
8212 vector signed int vec_vaddsws (vector bool int, vector signed int);
8213 vector signed int vec_vaddsws (vector signed int, vector bool int);
8214 vector signed int vec_vaddsws (vector signed int, vector signed int);
8216 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8217 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8218 vector unsigned int vec_vadduws (vector unsigned int,
8219 vector unsigned int);
8221 vector signed short vec_vaddshs (vector bool short,
8222 vector signed short);
8223 vector signed short vec_vaddshs (vector signed short,
8225 vector signed short vec_vaddshs (vector signed short,
8226 vector signed short);
8228 vector unsigned short vec_vadduhs (vector bool short,
8229 vector unsigned short);
8230 vector unsigned short vec_vadduhs (vector unsigned short,
8232 vector unsigned short vec_vadduhs (vector unsigned short,
8233 vector unsigned short);
8235 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8236 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8237 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8239 vector unsigned char vec_vaddubs (vector bool char,
8240 vector unsigned char);
8241 vector unsigned char vec_vaddubs (vector unsigned char,
8243 vector unsigned char vec_vaddubs (vector unsigned char,
8244 vector unsigned char);
8246 vector float vec_and (vector float, vector float);
8247 vector float vec_and (vector float, vector bool int);
8248 vector float vec_and (vector bool int, vector float);
8249 vector bool int vec_and (vector bool int, vector bool int);
8250 vector signed int vec_and (vector bool int, vector signed int);
8251 vector signed int vec_and (vector signed int, vector bool int);
8252 vector signed int vec_and (vector signed int, vector signed int);
8253 vector unsigned int vec_and (vector bool int, vector unsigned int);
8254 vector unsigned int vec_and (vector unsigned int, vector bool int);
8255 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8256 vector bool short vec_and (vector bool short, vector bool short);
8257 vector signed short vec_and (vector bool short, vector signed short);
8258 vector signed short vec_and (vector signed short, vector bool short);
8259 vector signed short vec_and (vector signed short, vector signed short);
8260 vector unsigned short vec_and (vector bool short,
8261 vector unsigned short);
8262 vector unsigned short vec_and (vector unsigned short,
8264 vector unsigned short vec_and (vector unsigned short,
8265 vector unsigned short);
8266 vector signed char vec_and (vector bool char, vector signed char);
8267 vector bool char vec_and (vector bool char, vector bool char);
8268 vector signed char vec_and (vector signed char, vector bool char);
8269 vector signed char vec_and (vector signed char, vector signed char);
8270 vector unsigned char vec_and (vector bool char, vector unsigned char);
8271 vector unsigned char vec_and (vector unsigned char, vector bool char);
8272 vector unsigned char vec_and (vector unsigned char,
8273 vector unsigned char);
8275 vector float vec_andc (vector float, vector float);
8276 vector float vec_andc (vector float, vector bool int);
8277 vector float vec_andc (vector bool int, vector float);
8278 vector bool int vec_andc (vector bool int, vector bool int);
8279 vector signed int vec_andc (vector bool int, vector signed int);
8280 vector signed int vec_andc (vector signed int, vector bool int);
8281 vector signed int vec_andc (vector signed int, vector signed int);
8282 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8283 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8284 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8285 vector bool short vec_andc (vector bool short, vector bool short);
8286 vector signed short vec_andc (vector bool short, vector signed short);
8287 vector signed short vec_andc (vector signed short, vector bool short);
8288 vector signed short vec_andc (vector signed short, vector signed short);
8289 vector unsigned short vec_andc (vector bool short,
8290 vector unsigned short);
8291 vector unsigned short vec_andc (vector unsigned short,
8293 vector unsigned short vec_andc (vector unsigned short,
8294 vector unsigned short);
8295 vector signed char vec_andc (vector bool char, vector signed char);
8296 vector bool char vec_andc (vector bool char, vector bool char);
8297 vector signed char vec_andc (vector signed char, vector bool char);
8298 vector signed char vec_andc (vector signed char, vector signed char);
8299 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8300 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8301 vector unsigned char vec_andc (vector unsigned char,
8302 vector unsigned char);
8304 vector unsigned char vec_avg (vector unsigned char,
8305 vector unsigned char);
8306 vector signed char vec_avg (vector signed char, vector signed char);
8307 vector unsigned short vec_avg (vector unsigned short,
8308 vector unsigned short);
8309 vector signed short vec_avg (vector signed short, vector signed short);
8310 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8311 vector signed int vec_avg (vector signed int, vector signed int);
8313 vector signed int vec_vavgsw (vector signed int, vector signed int);
8315 vector unsigned int vec_vavguw (vector unsigned int,
8316 vector unsigned int);
8318 vector signed short vec_vavgsh (vector signed short,
8319 vector signed short);
8321 vector unsigned short vec_vavguh (vector unsigned short,
8322 vector unsigned short);
8324 vector signed char vec_vavgsb (vector signed char, vector signed char);
8326 vector unsigned char vec_vavgub (vector unsigned char,
8327 vector unsigned char);
8329 vector float vec_ceil (vector float);
8331 vector signed int vec_cmpb (vector float, vector float);
8333 vector bool char vec_cmpeq (vector signed char, vector signed char);
8334 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8335 vector bool short vec_cmpeq (vector signed short, vector signed short);
8336 vector bool short vec_cmpeq (vector unsigned short,
8337 vector unsigned short);
8338 vector bool int vec_cmpeq (vector signed int, vector signed int);
8339 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8340 vector bool int vec_cmpeq (vector float, vector float);
8342 vector bool int vec_vcmpeqfp (vector float, vector float);
8344 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8345 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8347 vector bool short vec_vcmpequh (vector signed short,
8348 vector signed short);
8349 vector bool short vec_vcmpequh (vector unsigned short,
8350 vector unsigned short);
8352 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8353 vector bool char vec_vcmpequb (vector unsigned char,
8354 vector unsigned char);
8356 vector bool int vec_cmpge (vector float, vector float);
8358 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8359 vector bool char vec_cmpgt (vector signed char, vector signed char);
8360 vector bool short vec_cmpgt (vector unsigned short,
8361 vector unsigned short);
8362 vector bool short vec_cmpgt (vector signed short, vector signed short);
8363 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8364 vector bool int vec_cmpgt (vector signed int, vector signed int);
8365 vector bool int vec_cmpgt (vector float, vector float);
8367 vector bool int vec_vcmpgtfp (vector float, vector float);
8369 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8371 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8373 vector bool short vec_vcmpgtsh (vector signed short,
8374 vector signed short);
8376 vector bool short vec_vcmpgtuh (vector unsigned short,
8377 vector unsigned short);
8379 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8381 vector bool char vec_vcmpgtub (vector unsigned char,
8382 vector unsigned char);
8384 vector bool int vec_cmple (vector float, vector float);
8386 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8387 vector bool char vec_cmplt (vector signed char, vector signed char);
8388 vector bool short vec_cmplt (vector unsigned short,
8389 vector unsigned short);
8390 vector bool short vec_cmplt (vector signed short, vector signed short);
8391 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8392 vector bool int vec_cmplt (vector signed int, vector signed int);
8393 vector bool int vec_cmplt (vector float, vector float);
8395 vector float vec_ctf (vector unsigned int, const int);
8396 vector float vec_ctf (vector signed int, const int);
8398 vector float vec_vcfsx (vector signed int, const int);
8400 vector float vec_vcfux (vector unsigned int, const int);
8402 vector signed int vec_cts (vector float, const int);
8404 vector unsigned int vec_ctu (vector float, const int);
8406 void vec_dss (const int);
8408 void vec_dssall (void);
8410 void vec_dst (const vector unsigned char *, int, const int);
8411 void vec_dst (const vector signed char *, int, const int);
8412 void vec_dst (const vector bool char *, int, const int);
8413 void vec_dst (const vector unsigned short *, int, const int);
8414 void vec_dst (const vector signed short *, int, const int);
8415 void vec_dst (const vector bool short *, int, const int);
8416 void vec_dst (const vector pixel *, int, const int);
8417 void vec_dst (const vector unsigned int *, int, const int);
8418 void vec_dst (const vector signed int *, int, const int);
8419 void vec_dst (const vector bool int *, int, const int);
8420 void vec_dst (const vector float *, int, const int);
8421 void vec_dst (const unsigned char *, int, const int);
8422 void vec_dst (const signed char *, int, const int);
8423 void vec_dst (const unsigned short *, int, const int);
8424 void vec_dst (const short *, int, const int);
8425 void vec_dst (const unsigned int *, int, const int);
8426 void vec_dst (const int *, int, const int);
8427 void vec_dst (const unsigned long *, int, const int);
8428 void vec_dst (const long *, int, const int);
8429 void vec_dst (const float *, int, const int);
8431 void vec_dstst (const vector unsigned char *, int, const int);
8432 void vec_dstst (const vector signed char *, int, const int);
8433 void vec_dstst (const vector bool char *, int, const int);
8434 void vec_dstst (const vector unsigned short *, int, const int);
8435 void vec_dstst (const vector signed short *, int, const int);
8436 void vec_dstst (const vector bool short *, int, const int);
8437 void vec_dstst (const vector pixel *, int, const int);
8438 void vec_dstst (const vector unsigned int *, int, const int);
8439 void vec_dstst (const vector signed int *, int, const int);
8440 void vec_dstst (const vector bool int *, int, const int);
8441 void vec_dstst (const vector float *, int, const int);
8442 void vec_dstst (const unsigned char *, int, const int);
8443 void vec_dstst (const signed char *, int, const int);
8444 void vec_dstst (const unsigned short *, int, const int);
8445 void vec_dstst (const short *, int, const int);
8446 void vec_dstst (const unsigned int *, int, const int);
8447 void vec_dstst (const int *, int, const int);
8448 void vec_dstst (const unsigned long *, int, const int);
8449 void vec_dstst (const long *, int, const int);
8450 void vec_dstst (const float *, int, const int);
8452 void vec_dststt (const vector unsigned char *, int, const int);
8453 void vec_dststt (const vector signed char *, int, const int);
8454 void vec_dststt (const vector bool char *, int, const int);
8455 void vec_dststt (const vector unsigned short *, int, const int);
8456 void vec_dststt (const vector signed short *, int, const int);
8457 void vec_dststt (const vector bool short *, int, const int);
8458 void vec_dststt (const vector pixel *, int, const int);
8459 void vec_dststt (const vector unsigned int *, int, const int);
8460 void vec_dststt (const vector signed int *, int, const int);
8461 void vec_dststt (const vector bool int *, int, const int);
8462 void vec_dststt (const vector float *, int, const int);
8463 void vec_dststt (const unsigned char *, int, const int);
8464 void vec_dststt (const signed char *, int, const int);
8465 void vec_dststt (const unsigned short *, int, const int);
8466 void vec_dststt (const short *, int, const int);
8467 void vec_dststt (const unsigned int *, int, const int);
8468 void vec_dststt (const int *, int, const int);
8469 void vec_dststt (const unsigned long *, int, const int);
8470 void vec_dststt (const long *, int, const int);
8471 void vec_dststt (const float *, int, const int);
8473 void vec_dstt (const vector unsigned char *, int, const int);
8474 void vec_dstt (const vector signed char *, int, const int);
8475 void vec_dstt (const vector bool char *, int, const int);
8476 void vec_dstt (const vector unsigned short *, int, const int);
8477 void vec_dstt (const vector signed short *, int, const int);
8478 void vec_dstt (const vector bool short *, int, const int);
8479 void vec_dstt (const vector pixel *, int, const int);
8480 void vec_dstt (const vector unsigned int *, int, const int);
8481 void vec_dstt (const vector signed int *, int, const int);
8482 void vec_dstt (const vector bool int *, int, const int);
8483 void vec_dstt (const vector float *, int, const int);
8484 void vec_dstt (const unsigned char *, int, const int);
8485 void vec_dstt (const signed char *, int, const int);
8486 void vec_dstt (const unsigned short *, int, const int);
8487 void vec_dstt (const short *, int, const int);
8488 void vec_dstt (const unsigned int *, int, const int);
8489 void vec_dstt (const int *, int, const int);
8490 void vec_dstt (const unsigned long *, int, const int);
8491 void vec_dstt (const long *, int, const int);
8492 void vec_dstt (const float *, int, const int);
8494 vector float vec_expte (vector float);
8496 vector float vec_floor (vector float);
8498 vector float vec_ld (int, const vector float *);
8499 vector float vec_ld (int, const float *);
8500 vector bool int vec_ld (int, const vector bool int *);
8501 vector signed int vec_ld (int, const vector signed int *);
8502 vector signed int vec_ld (int, const int *);
8503 vector signed int vec_ld (int, const long *);
8504 vector unsigned int vec_ld (int, const vector unsigned int *);
8505 vector unsigned int vec_ld (int, const unsigned int *);
8506 vector unsigned int vec_ld (int, const unsigned long *);
8507 vector bool short vec_ld (int, const vector bool short *);
8508 vector pixel vec_ld (int, const vector pixel *);
8509 vector signed short vec_ld (int, const vector signed short *);
8510 vector signed short vec_ld (int, const short *);
8511 vector unsigned short vec_ld (int, const vector unsigned short *);
8512 vector unsigned short vec_ld (int, const unsigned short *);
8513 vector bool char vec_ld (int, const vector bool char *);
8514 vector signed char vec_ld (int, const vector signed char *);
8515 vector signed char vec_ld (int, const signed char *);
8516 vector unsigned char vec_ld (int, const vector unsigned char *);
8517 vector unsigned char vec_ld (int, const unsigned char *);
8519 vector signed char vec_lde (int, const signed char *);
8520 vector unsigned char vec_lde (int, const unsigned char *);
8521 vector signed short vec_lde (int, const short *);
8522 vector unsigned short vec_lde (int, const unsigned short *);
8523 vector float vec_lde (int, const float *);
8524 vector signed int vec_lde (int, const int *);
8525 vector unsigned int vec_lde (int, const unsigned int *);
8526 vector signed int vec_lde (int, const long *);
8527 vector unsigned int vec_lde (int, const unsigned long *);
8529 vector float vec_lvewx (int, float *);
8530 vector signed int vec_lvewx (int, int *);
8531 vector unsigned int vec_lvewx (int, unsigned int *);
8532 vector signed int vec_lvewx (int, long *);
8533 vector unsigned int vec_lvewx (int, unsigned long *);
8535 vector signed short vec_lvehx (int, short *);
8536 vector unsigned short vec_lvehx (int, unsigned short *);
8538 vector signed char vec_lvebx (int, char *);
8539 vector unsigned char vec_lvebx (int, unsigned char *);
8541 vector float vec_ldl (int, const vector float *);
8542 vector float vec_ldl (int, const float *);
8543 vector bool int vec_ldl (int, const vector bool int *);
8544 vector signed int vec_ldl (int, const vector signed int *);
8545 vector signed int vec_ldl (int, const int *);
8546 vector signed int vec_ldl (int, const long *);
8547 vector unsigned int vec_ldl (int, const vector unsigned int *);
8548 vector unsigned int vec_ldl (int, const unsigned int *);
8549 vector unsigned int vec_ldl (int, const unsigned long *);
8550 vector bool short vec_ldl (int, const vector bool short *);
8551 vector pixel vec_ldl (int, const vector pixel *);
8552 vector signed short vec_ldl (int, const vector signed short *);
8553 vector signed short vec_ldl (int, const short *);
8554 vector unsigned short vec_ldl (int, const vector unsigned short *);
8555 vector unsigned short vec_ldl (int, const unsigned short *);
8556 vector bool char vec_ldl (int, const vector bool char *);
8557 vector signed char vec_ldl (int, const vector signed char *);
8558 vector signed char vec_ldl (int, const signed char *);
8559 vector unsigned char vec_ldl (int, const vector unsigned char *);
8560 vector unsigned char vec_ldl (int, const unsigned char *);
8562 vector float vec_loge (vector float);
8564 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8565 vector unsigned char vec_lvsl (int, const volatile signed char *);
8566 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8567 vector unsigned char vec_lvsl (int, const volatile short *);
8568 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8569 vector unsigned char vec_lvsl (int, const volatile int *);
8570 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8571 vector unsigned char vec_lvsl (int, const volatile long *);
8572 vector unsigned char vec_lvsl (int, const volatile float *);
8574 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8575 vector unsigned char vec_lvsr (int, const volatile signed char *);
8576 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8577 vector unsigned char vec_lvsr (int, const volatile short *);
8578 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8579 vector unsigned char vec_lvsr (int, const volatile int *);
8580 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8581 vector unsigned char vec_lvsr (int, const volatile long *);
8582 vector unsigned char vec_lvsr (int, const volatile float *);
8584 vector float vec_madd (vector float, vector float, vector float);
8586 vector signed short vec_madds (vector signed short,
8587 vector signed short,
8588 vector signed short);
8590 vector unsigned char vec_max (vector bool char, vector unsigned char);
8591 vector unsigned char vec_max (vector unsigned char, vector bool char);
8592 vector unsigned char vec_max (vector unsigned char,
8593 vector unsigned char);
8594 vector signed char vec_max (vector bool char, vector signed char);
8595 vector signed char vec_max (vector signed char, vector bool char);
8596 vector signed char vec_max (vector signed char, vector signed char);
8597 vector unsigned short vec_max (vector bool short,
8598 vector unsigned short);
8599 vector unsigned short vec_max (vector unsigned short,
8601 vector unsigned short vec_max (vector unsigned short,
8602 vector unsigned short);
8603 vector signed short vec_max (vector bool short, vector signed short);
8604 vector signed short vec_max (vector signed short, vector bool short);
8605 vector signed short vec_max (vector signed short, vector signed short);
8606 vector unsigned int vec_max (vector bool int, vector unsigned int);
8607 vector unsigned int vec_max (vector unsigned int, vector bool int);
8608 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8609 vector signed int vec_max (vector bool int, vector signed int);
8610 vector signed int vec_max (vector signed int, vector bool int);
8611 vector signed int vec_max (vector signed int, vector signed int);
8612 vector float vec_max (vector float, vector float);
8614 vector float vec_vmaxfp (vector float, vector float);
8616 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8617 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8618 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8620 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8621 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8622 vector unsigned int vec_vmaxuw (vector unsigned int,
8623 vector unsigned int);
8625 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8626 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8627 vector signed short vec_vmaxsh (vector signed short,
8628 vector signed short);
8630 vector unsigned short vec_vmaxuh (vector bool short,
8631 vector unsigned short);
8632 vector unsigned short vec_vmaxuh (vector unsigned short,
8634 vector unsigned short vec_vmaxuh (vector unsigned short,
8635 vector unsigned short);
8637 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8638 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8639 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8641 vector unsigned char vec_vmaxub (vector bool char,
8642 vector unsigned char);
8643 vector unsigned char vec_vmaxub (vector unsigned char,
8645 vector unsigned char vec_vmaxub (vector unsigned char,
8646 vector unsigned char);
8648 vector bool char vec_mergeh (vector bool char, vector bool char);
8649 vector signed char vec_mergeh (vector signed char, vector signed char);
8650 vector unsigned char vec_mergeh (vector unsigned char,
8651 vector unsigned char);
8652 vector bool short vec_mergeh (vector bool short, vector bool short);
8653 vector pixel vec_mergeh (vector pixel, vector pixel);
8654 vector signed short vec_mergeh (vector signed short,
8655 vector signed short);
8656 vector unsigned short vec_mergeh (vector unsigned short,
8657 vector unsigned short);
8658 vector float vec_mergeh (vector float, vector float);
8659 vector bool int vec_mergeh (vector bool int, vector bool int);
8660 vector signed int vec_mergeh (vector signed int, vector signed int);
8661 vector unsigned int vec_mergeh (vector unsigned int,
8662 vector unsigned int);
8664 vector float vec_vmrghw (vector float, vector float);
8665 vector bool int vec_vmrghw (vector bool int, vector bool int);
8666 vector signed int vec_vmrghw (vector signed int, vector signed int);
8667 vector unsigned int vec_vmrghw (vector unsigned int,
8668 vector unsigned int);
8670 vector bool short vec_vmrghh (vector bool short, vector bool short);
8671 vector signed short vec_vmrghh (vector signed short,
8672 vector signed short);
8673 vector unsigned short vec_vmrghh (vector unsigned short,
8674 vector unsigned short);
8675 vector pixel vec_vmrghh (vector pixel, vector pixel);
8677 vector bool char vec_vmrghb (vector bool char, vector bool char);
8678 vector signed char vec_vmrghb (vector signed char, vector signed char);
8679 vector unsigned char vec_vmrghb (vector unsigned char,
8680 vector unsigned char);
8682 vector bool char vec_mergel (vector bool char, vector bool char);
8683 vector signed char vec_mergel (vector signed char, vector signed char);
8684 vector unsigned char vec_mergel (vector unsigned char,
8685 vector unsigned char);
8686 vector bool short vec_mergel (vector bool short, vector bool short);
8687 vector pixel vec_mergel (vector pixel, vector pixel);
8688 vector signed short vec_mergel (vector signed short,
8689 vector signed short);
8690 vector unsigned short vec_mergel (vector unsigned short,
8691 vector unsigned short);
8692 vector float vec_mergel (vector float, vector float);
8693 vector bool int vec_mergel (vector bool int, vector bool int);
8694 vector signed int vec_mergel (vector signed int, vector signed int);
8695 vector unsigned int vec_mergel (vector unsigned int,
8696 vector unsigned int);
8698 vector float vec_vmrglw (vector float, vector float);
8699 vector signed int vec_vmrglw (vector signed int, vector signed int);
8700 vector unsigned int vec_vmrglw (vector unsigned int,
8701 vector unsigned int);
8702 vector bool int vec_vmrglw (vector bool int, vector bool int);
8704 vector bool short vec_vmrglh (vector bool short, vector bool short);
8705 vector signed short vec_vmrglh (vector signed short,
8706 vector signed short);
8707 vector unsigned short vec_vmrglh (vector unsigned short,
8708 vector unsigned short);
8709 vector pixel vec_vmrglh (vector pixel, vector pixel);
8711 vector bool char vec_vmrglb (vector bool char, vector bool char);
8712 vector signed char vec_vmrglb (vector signed char, vector signed char);
8713 vector unsigned char vec_vmrglb (vector unsigned char,
8714 vector unsigned char);
8716 vector unsigned short vec_mfvscr (void);
8718 vector unsigned char vec_min (vector bool char, vector unsigned char);
8719 vector unsigned char vec_min (vector unsigned char, vector bool char);
8720 vector unsigned char vec_min (vector unsigned char,
8721 vector unsigned char);
8722 vector signed char vec_min (vector bool char, vector signed char);
8723 vector signed char vec_min (vector signed char, vector bool char);
8724 vector signed char vec_min (vector signed char, vector signed char);
8725 vector unsigned short vec_min (vector bool short,
8726 vector unsigned short);
8727 vector unsigned short vec_min (vector unsigned short,
8729 vector unsigned short vec_min (vector unsigned short,
8730 vector unsigned short);
8731 vector signed short vec_min (vector bool short, vector signed short);
8732 vector signed short vec_min (vector signed short, vector bool short);
8733 vector signed short vec_min (vector signed short, vector signed short);
8734 vector unsigned int vec_min (vector bool int, vector unsigned int);
8735 vector unsigned int vec_min (vector unsigned int, vector bool int);
8736 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8737 vector signed int vec_min (vector bool int, vector signed int);
8738 vector signed int vec_min (vector signed int, vector bool int);
8739 vector signed int vec_min (vector signed int, vector signed int);
8740 vector float vec_min (vector float, vector float);
8742 vector float vec_vminfp (vector float, vector float);
8744 vector signed int vec_vminsw (vector bool int, vector signed int);
8745 vector signed int vec_vminsw (vector signed int, vector bool int);
8746 vector signed int vec_vminsw (vector signed int, vector signed int);
8748 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8749 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8750 vector unsigned int vec_vminuw (vector unsigned int,
8751 vector unsigned int);
8753 vector signed short vec_vminsh (vector bool short, vector signed short);
8754 vector signed short vec_vminsh (vector signed short, vector bool short);
8755 vector signed short vec_vminsh (vector signed short,
8756 vector signed short);
8758 vector unsigned short vec_vminuh (vector bool short,
8759 vector unsigned short);
8760 vector unsigned short vec_vminuh (vector unsigned short,
8762 vector unsigned short vec_vminuh (vector unsigned short,
8763 vector unsigned short);
8765 vector signed char vec_vminsb (vector bool char, vector signed char);
8766 vector signed char vec_vminsb (vector signed char, vector bool char);
8767 vector signed char vec_vminsb (vector signed char, vector signed char);
8769 vector unsigned char vec_vminub (vector bool char,
8770 vector unsigned char);
8771 vector unsigned char vec_vminub (vector unsigned char,
8773 vector unsigned char vec_vminub (vector unsigned char,
8774 vector unsigned char);
8776 vector signed short vec_mladd (vector signed short,
8777 vector signed short,
8778 vector signed short);
8779 vector signed short vec_mladd (vector signed short,
8780 vector unsigned short,
8781 vector unsigned short);
8782 vector signed short vec_mladd (vector unsigned short,
8783 vector signed short,
8784 vector signed short);
8785 vector unsigned short vec_mladd (vector unsigned short,
8786 vector unsigned short,
8787 vector unsigned short);
8789 vector signed short vec_mradds (vector signed short,
8790 vector signed short,
8791 vector signed short);
8793 vector unsigned int vec_msum (vector unsigned char,
8794 vector unsigned char,
8795 vector unsigned int);
8796 vector signed int vec_msum (vector signed char,
8797 vector unsigned char,
8799 vector unsigned int vec_msum (vector unsigned short,
8800 vector unsigned short,
8801 vector unsigned int);
8802 vector signed int vec_msum (vector signed short,
8803 vector signed short,
8806 vector signed int vec_vmsumshm (vector signed short,
8807 vector signed short,
8810 vector unsigned int vec_vmsumuhm (vector unsigned short,
8811 vector unsigned short,
8812 vector unsigned int);
8814 vector signed int vec_vmsummbm (vector signed char,
8815 vector unsigned char,
8818 vector unsigned int vec_vmsumubm (vector unsigned char,
8819 vector unsigned char,
8820 vector unsigned int);
8822 vector unsigned int vec_msums (vector unsigned short,
8823 vector unsigned short,
8824 vector unsigned int);
8825 vector signed int vec_msums (vector signed short,
8826 vector signed short,
8829 vector signed int vec_vmsumshs (vector signed short,
8830 vector signed short,
8833 vector unsigned int vec_vmsumuhs (vector unsigned short,
8834 vector unsigned short,
8835 vector unsigned int);
8837 void vec_mtvscr (vector signed int);
8838 void vec_mtvscr (vector unsigned int);
8839 void vec_mtvscr (vector bool int);
8840 void vec_mtvscr (vector signed short);
8841 void vec_mtvscr (vector unsigned short);
8842 void vec_mtvscr (vector bool short);
8843 void vec_mtvscr (vector pixel);
8844 void vec_mtvscr (vector signed char);
8845 void vec_mtvscr (vector unsigned char);
8846 void vec_mtvscr (vector bool char);
8848 vector unsigned short vec_mule (vector unsigned char,
8849 vector unsigned char);
8850 vector signed short vec_mule (vector signed char,
8851 vector signed char);
8852 vector unsigned int vec_mule (vector unsigned short,
8853 vector unsigned short);
8854 vector signed int vec_mule (vector signed short, vector signed short);
8856 vector signed int vec_vmulesh (vector signed short,
8857 vector signed short);
8859 vector unsigned int vec_vmuleuh (vector unsigned short,
8860 vector unsigned short);
8862 vector signed short vec_vmulesb (vector signed char,
8863 vector signed char);
8865 vector unsigned short vec_vmuleub (vector unsigned char,
8866 vector unsigned char);
8868 vector unsigned short vec_mulo (vector unsigned char,
8869 vector unsigned char);
8870 vector signed short vec_mulo (vector signed char, vector signed char);
8871 vector unsigned int vec_mulo (vector unsigned short,
8872 vector unsigned short);
8873 vector signed int vec_mulo (vector signed short, vector signed short);
8875 vector signed int vec_vmulosh (vector signed short,
8876 vector signed short);
8878 vector unsigned int vec_vmulouh (vector unsigned short,
8879 vector unsigned short);
8881 vector signed short vec_vmulosb (vector signed char,
8882 vector signed char);
8884 vector unsigned short vec_vmuloub (vector unsigned char,
8885 vector unsigned char);
8887 vector float vec_nmsub (vector float, vector float, vector float);
8889 vector float vec_nor (vector float, vector float);
8890 vector signed int vec_nor (vector signed int, vector signed int);
8891 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8892 vector bool int vec_nor (vector bool int, vector bool int);
8893 vector signed short vec_nor (vector signed short, vector signed short);
8894 vector unsigned short vec_nor (vector unsigned short,
8895 vector unsigned short);
8896 vector bool short vec_nor (vector bool short, vector bool short);
8897 vector signed char vec_nor (vector signed char, vector signed char);
8898 vector unsigned char vec_nor (vector unsigned char,
8899 vector unsigned char);
8900 vector bool char vec_nor (vector bool char, vector bool char);
8902 vector float vec_or (vector float, vector float);
8903 vector float vec_or (vector float, vector bool int);
8904 vector float vec_or (vector bool int, vector float);
8905 vector bool int vec_or (vector bool int, vector bool int);
8906 vector signed int vec_or (vector bool int, vector signed int);
8907 vector signed int vec_or (vector signed int, vector bool int);
8908 vector signed int vec_or (vector signed int, vector signed int);
8909 vector unsigned int vec_or (vector bool int, vector unsigned int);
8910 vector unsigned int vec_or (vector unsigned int, vector bool int);
8911 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8912 vector bool short vec_or (vector bool short, vector bool short);
8913 vector signed short vec_or (vector bool short, vector signed short);
8914 vector signed short vec_or (vector signed short, vector bool short);
8915 vector signed short vec_or (vector signed short, vector signed short);
8916 vector unsigned short vec_or (vector bool short, vector unsigned short);
8917 vector unsigned short vec_or (vector unsigned short, vector bool short);
8918 vector unsigned short vec_or (vector unsigned short,
8919 vector unsigned short);
8920 vector signed char vec_or (vector bool char, vector signed char);
8921 vector bool char vec_or (vector bool char, vector bool char);
8922 vector signed char vec_or (vector signed char, vector bool char);
8923 vector signed char vec_or (vector signed char, vector signed char);
8924 vector unsigned char vec_or (vector bool char, vector unsigned char);
8925 vector unsigned char vec_or (vector unsigned char, vector bool char);
8926 vector unsigned char vec_or (vector unsigned char,
8927 vector unsigned char);
8929 vector signed char vec_pack (vector signed short, vector signed short);
8930 vector unsigned char vec_pack (vector unsigned short,
8931 vector unsigned short);
8932 vector bool char vec_pack (vector bool short, vector bool short);
8933 vector signed short vec_pack (vector signed int, vector signed int);
8934 vector unsigned short vec_pack (vector unsigned int,
8935 vector unsigned int);
8936 vector bool short vec_pack (vector bool int, vector bool int);
8938 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8939 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8940 vector unsigned short vec_vpkuwum (vector unsigned int,
8941 vector unsigned int);
8943 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8944 vector signed char vec_vpkuhum (vector signed short,
8945 vector signed short);
8946 vector unsigned char vec_vpkuhum (vector unsigned short,
8947 vector unsigned short);
8949 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8951 vector unsigned char vec_packs (vector unsigned short,
8952 vector unsigned short);
8953 vector signed char vec_packs (vector signed short, vector signed short);
8954 vector unsigned short vec_packs (vector unsigned int,
8955 vector unsigned int);
8956 vector signed short vec_packs (vector signed int, vector signed int);
8958 vector signed short vec_vpkswss (vector signed int, vector signed int);
8960 vector unsigned short vec_vpkuwus (vector unsigned int,
8961 vector unsigned int);
8963 vector signed char vec_vpkshss (vector signed short,
8964 vector signed short);
8966 vector unsigned char vec_vpkuhus (vector unsigned short,
8967 vector unsigned short);
8969 vector unsigned char vec_packsu (vector unsigned short,
8970 vector unsigned short);
8971 vector unsigned char vec_packsu (vector signed short,
8972 vector signed short);
8973 vector unsigned short vec_packsu (vector unsigned int,
8974 vector unsigned int);
8975 vector unsigned short vec_packsu (vector signed int, vector signed int);
8977 vector unsigned short vec_vpkswus (vector signed int,
8980 vector unsigned char vec_vpkshus (vector signed short,
8981 vector signed short);
8983 vector float vec_perm (vector float,
8985 vector unsigned char);
8986 vector signed int vec_perm (vector signed int,
8988 vector unsigned char);
8989 vector unsigned int vec_perm (vector unsigned int,
8990 vector unsigned int,
8991 vector unsigned char);
8992 vector bool int vec_perm (vector bool int,
8994 vector unsigned char);
8995 vector signed short vec_perm (vector signed short,
8996 vector signed short,
8997 vector unsigned char);
8998 vector unsigned short vec_perm (vector unsigned short,
8999 vector unsigned short,
9000 vector unsigned char);
9001 vector bool short vec_perm (vector bool short,
9003 vector unsigned char);
9004 vector pixel vec_perm (vector pixel,
9006 vector unsigned char);
9007 vector signed char vec_perm (vector signed char,
9009 vector unsigned char);
9010 vector unsigned char vec_perm (vector unsigned char,
9011 vector unsigned char,
9012 vector unsigned char);
9013 vector bool char vec_perm (vector bool char,
9015 vector unsigned char);
9017 vector float vec_re (vector float);
9019 vector signed char vec_rl (vector signed char,
9020 vector unsigned char);
9021 vector unsigned char vec_rl (vector unsigned char,
9022 vector unsigned char);
9023 vector signed short vec_rl (vector signed short, vector unsigned short);
9024 vector unsigned short vec_rl (vector unsigned short,
9025 vector unsigned short);
9026 vector signed int vec_rl (vector signed int, vector unsigned int);
9027 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
9029 vector signed int vec_vrlw (vector signed int, vector unsigned int);
9030 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
9032 vector signed short vec_vrlh (vector signed short,
9033 vector unsigned short);
9034 vector unsigned short vec_vrlh (vector unsigned short,
9035 vector unsigned short);
9037 vector signed char vec_vrlb (vector signed char, vector unsigned char);
9038 vector unsigned char vec_vrlb (vector unsigned char,
9039 vector unsigned char);
9041 vector float vec_round (vector float);
9043 vector float vec_rsqrte (vector float);
9045 vector float vec_sel (vector float, vector float, vector bool int);
9046 vector float vec_sel (vector float, vector float, vector unsigned int);
9047 vector signed int vec_sel (vector signed int,
9050 vector signed int vec_sel (vector signed int,
9052 vector unsigned int);
9053 vector unsigned int vec_sel (vector unsigned int,
9054 vector unsigned int,
9056 vector unsigned int vec_sel (vector unsigned int,
9057 vector unsigned int,
9058 vector unsigned int);
9059 vector bool int vec_sel (vector bool int,
9062 vector bool int vec_sel (vector bool int,
9064 vector unsigned int);
9065 vector signed short vec_sel (vector signed short,
9066 vector signed short,
9068 vector signed short vec_sel (vector signed short,
9069 vector signed short,
9070 vector unsigned short);
9071 vector unsigned short vec_sel (vector unsigned short,
9072 vector unsigned short,
9074 vector unsigned short vec_sel (vector unsigned short,
9075 vector unsigned short,
9076 vector unsigned short);
9077 vector bool short vec_sel (vector bool short,
9080 vector bool short vec_sel (vector bool short,
9082 vector unsigned short);
9083 vector signed char vec_sel (vector signed char,
9086 vector signed char vec_sel (vector signed char,
9088 vector unsigned char);
9089 vector unsigned char vec_sel (vector unsigned char,
9090 vector unsigned char,
9092 vector unsigned char vec_sel (vector unsigned char,
9093 vector unsigned char,
9094 vector unsigned char);
9095 vector bool char vec_sel (vector bool char,
9098 vector bool char vec_sel (vector bool char,
9100 vector unsigned char);
9102 vector signed char vec_sl (vector signed char,
9103 vector unsigned char);
9104 vector unsigned char vec_sl (vector unsigned char,
9105 vector unsigned char);
9106 vector signed short vec_sl (vector signed short, vector unsigned short);
9107 vector unsigned short vec_sl (vector unsigned short,
9108 vector unsigned short);
9109 vector signed int vec_sl (vector signed int, vector unsigned int);
9110 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
9112 vector signed int vec_vslw (vector signed int, vector unsigned int);
9113 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
9115 vector signed short vec_vslh (vector signed short,
9116 vector unsigned short);
9117 vector unsigned short vec_vslh (vector unsigned short,
9118 vector unsigned short);
9120 vector signed char vec_vslb (vector signed char, vector unsigned char);
9121 vector unsigned char vec_vslb (vector unsigned char,
9122 vector unsigned char);
9124 vector float vec_sld (vector float, vector float, const int);
9125 vector signed int vec_sld (vector signed int,
9128 vector unsigned int vec_sld (vector unsigned int,
9129 vector unsigned int,
9131 vector bool int vec_sld (vector bool int,
9134 vector signed short vec_sld (vector signed short,
9135 vector signed short,
9137 vector unsigned short vec_sld (vector unsigned short,
9138 vector unsigned short,
9140 vector bool short vec_sld (vector bool short,
9143 vector pixel vec_sld (vector pixel,
9146 vector signed char vec_sld (vector signed char,
9149 vector unsigned char vec_sld (vector unsigned char,
9150 vector unsigned char,
9152 vector bool char vec_sld (vector bool char,
9156 vector signed int vec_sll (vector signed int,
9157 vector unsigned int);
9158 vector signed int vec_sll (vector signed int,
9159 vector unsigned short);
9160 vector signed int vec_sll (vector signed int,
9161 vector unsigned char);
9162 vector unsigned int vec_sll (vector unsigned int,
9163 vector unsigned int);
9164 vector unsigned int vec_sll (vector unsigned int,
9165 vector unsigned short);
9166 vector unsigned int vec_sll (vector unsigned int,
9167 vector unsigned char);
9168 vector bool int vec_sll (vector bool int,
9169 vector unsigned int);
9170 vector bool int vec_sll (vector bool int,
9171 vector unsigned short);
9172 vector bool int vec_sll (vector bool int,
9173 vector unsigned char);
9174 vector signed short vec_sll (vector signed short,
9175 vector unsigned int);
9176 vector signed short vec_sll (vector signed short,
9177 vector unsigned short);
9178 vector signed short vec_sll (vector signed short,
9179 vector unsigned char);
9180 vector unsigned short vec_sll (vector unsigned short,
9181 vector unsigned int);
9182 vector unsigned short vec_sll (vector unsigned short,
9183 vector unsigned short);
9184 vector unsigned short vec_sll (vector unsigned short,
9185 vector unsigned char);
9186 vector bool short vec_sll (vector bool short, vector unsigned int);
9187 vector bool short vec_sll (vector bool short, vector unsigned short);
9188 vector bool short vec_sll (vector bool short, vector unsigned char);
9189 vector pixel vec_sll (vector pixel, vector unsigned int);
9190 vector pixel vec_sll (vector pixel, vector unsigned short);
9191 vector pixel vec_sll (vector pixel, vector unsigned char);
9192 vector signed char vec_sll (vector signed char, vector unsigned int);
9193 vector signed char vec_sll (vector signed char, vector unsigned short);
9194 vector signed char vec_sll (vector signed char, vector unsigned char);
9195 vector unsigned char vec_sll (vector unsigned char,
9196 vector unsigned int);
9197 vector unsigned char vec_sll (vector unsigned char,
9198 vector unsigned short);
9199 vector unsigned char vec_sll (vector unsigned char,
9200 vector unsigned char);
9201 vector bool char vec_sll (vector bool char, vector unsigned int);
9202 vector bool char vec_sll (vector bool char, vector unsigned short);
9203 vector bool char vec_sll (vector bool char, vector unsigned char);
9205 vector float vec_slo (vector float, vector signed char);
9206 vector float vec_slo (vector float, vector unsigned char);
9207 vector signed int vec_slo (vector signed int, vector signed char);
9208 vector signed int vec_slo (vector signed int, vector unsigned char);
9209 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9210 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9211 vector signed short vec_slo (vector signed short, vector signed char);
9212 vector signed short vec_slo (vector signed short, vector unsigned char);
9213 vector unsigned short vec_slo (vector unsigned short,
9214 vector signed char);
9215 vector unsigned short vec_slo (vector unsigned short,
9216 vector unsigned char);
9217 vector pixel vec_slo (vector pixel, vector signed char);
9218 vector pixel vec_slo (vector pixel, vector unsigned char);
9219 vector signed char vec_slo (vector signed char, vector signed char);
9220 vector signed char vec_slo (vector signed char, vector unsigned char);
9221 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9222 vector unsigned char vec_slo (vector unsigned char,
9223 vector unsigned char);
9225 vector signed char vec_splat (vector signed char, const int);
9226 vector unsigned char vec_splat (vector unsigned char, const int);
9227 vector bool char vec_splat (vector bool char, const int);
9228 vector signed short vec_splat (vector signed short, const int);
9229 vector unsigned short vec_splat (vector unsigned short, const int);
9230 vector bool short vec_splat (vector bool short, const int);
9231 vector pixel vec_splat (vector pixel, const int);
9232 vector float vec_splat (vector float, const int);
9233 vector signed int vec_splat (vector signed int, const int);
9234 vector unsigned int vec_splat (vector unsigned int, const int);
9235 vector bool int vec_splat (vector bool int, const int);
9237 vector float vec_vspltw (vector float, const int);
9238 vector signed int vec_vspltw (vector signed int, const int);
9239 vector unsigned int vec_vspltw (vector unsigned int, const int);
9240 vector bool int vec_vspltw (vector bool int, const int);
9242 vector bool short vec_vsplth (vector bool short, const int);
9243 vector signed short vec_vsplth (vector signed short, const int);
9244 vector unsigned short vec_vsplth (vector unsigned short, const int);
9245 vector pixel vec_vsplth (vector pixel, const int);
9247 vector signed char vec_vspltb (vector signed char, const int);
9248 vector unsigned char vec_vspltb (vector unsigned char, const int);
9249 vector bool char vec_vspltb (vector bool char, const int);
9251 vector signed char vec_splat_s8 (const int);
9253 vector signed short vec_splat_s16 (const int);
9255 vector signed int vec_splat_s32 (const int);
9257 vector unsigned char vec_splat_u8 (const int);
9259 vector unsigned short vec_splat_u16 (const int);
9261 vector unsigned int vec_splat_u32 (const int);
9263 vector signed char vec_sr (vector signed char, vector unsigned char);
9264 vector unsigned char vec_sr (vector unsigned char,
9265 vector unsigned char);
9266 vector signed short vec_sr (vector signed short,
9267 vector unsigned short);
9268 vector unsigned short vec_sr (vector unsigned short,
9269 vector unsigned short);
9270 vector signed int vec_sr (vector signed int, vector unsigned int);
9271 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9273 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9274 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9276 vector signed short vec_vsrh (vector signed short,
9277 vector unsigned short);
9278 vector unsigned short vec_vsrh (vector unsigned short,
9279 vector unsigned short);
9281 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9282 vector unsigned char vec_vsrb (vector unsigned char,
9283 vector unsigned char);
9285 vector signed char vec_sra (vector signed char, vector unsigned char);
9286 vector unsigned char vec_sra (vector unsigned char,
9287 vector unsigned char);
9288 vector signed short vec_sra (vector signed short,
9289 vector unsigned short);
9290 vector unsigned short vec_sra (vector unsigned short,
9291 vector unsigned short);
9292 vector signed int vec_sra (vector signed int, vector unsigned int);
9293 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9295 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9296 vector unsigned int vec_vsraw (vector unsigned int,
9297 vector unsigned int);
9299 vector signed short vec_vsrah (vector signed short,
9300 vector unsigned short);
9301 vector unsigned short vec_vsrah (vector unsigned short,
9302 vector unsigned short);
9304 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9305 vector unsigned char vec_vsrab (vector unsigned char,
9306 vector unsigned char);
9308 vector signed int vec_srl (vector signed int, vector unsigned int);
9309 vector signed int vec_srl (vector signed int, vector unsigned short);
9310 vector signed int vec_srl (vector signed int, vector unsigned char);
9311 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9312 vector unsigned int vec_srl (vector unsigned int,
9313 vector unsigned short);
9314 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9315 vector bool int vec_srl (vector bool int, vector unsigned int);
9316 vector bool int vec_srl (vector bool int, vector unsigned short);
9317 vector bool int vec_srl (vector bool int, vector unsigned char);
9318 vector signed short vec_srl (vector signed short, vector unsigned int);
9319 vector signed short vec_srl (vector signed short,
9320 vector unsigned short);
9321 vector signed short vec_srl (vector signed short, vector unsigned char);
9322 vector unsigned short vec_srl (vector unsigned short,
9323 vector unsigned int);
9324 vector unsigned short vec_srl (vector unsigned short,
9325 vector unsigned short);
9326 vector unsigned short vec_srl (vector unsigned short,
9327 vector unsigned char);
9328 vector bool short vec_srl (vector bool short, vector unsigned int);
9329 vector bool short vec_srl (vector bool short, vector unsigned short);
9330 vector bool short vec_srl (vector bool short, vector unsigned char);
9331 vector pixel vec_srl (vector pixel, vector unsigned int);
9332 vector pixel vec_srl (vector pixel, vector unsigned short);
9333 vector pixel vec_srl (vector pixel, vector unsigned char);
9334 vector signed char vec_srl (vector signed char, vector unsigned int);
9335 vector signed char vec_srl (vector signed char, vector unsigned short);
9336 vector signed char vec_srl (vector signed char, vector unsigned char);
9337 vector unsigned char vec_srl (vector unsigned char,
9338 vector unsigned int);
9339 vector unsigned char vec_srl (vector unsigned char,
9340 vector unsigned short);
9341 vector unsigned char vec_srl (vector unsigned char,
9342 vector unsigned char);
9343 vector bool char vec_srl (vector bool char, vector unsigned int);
9344 vector bool char vec_srl (vector bool char, vector unsigned short);
9345 vector bool char vec_srl (vector bool char, vector unsigned char);
9347 vector float vec_sro (vector float, vector signed char);
9348 vector float vec_sro (vector float, vector unsigned char);
9349 vector signed int vec_sro (vector signed int, vector signed char);
9350 vector signed int vec_sro (vector signed int, vector unsigned char);
9351 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9352 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9353 vector signed short vec_sro (vector signed short, vector signed char);
9354 vector signed short vec_sro (vector signed short, vector unsigned char);
9355 vector unsigned short vec_sro (vector unsigned short,
9356 vector signed char);
9357 vector unsigned short vec_sro (vector unsigned short,
9358 vector unsigned char);
9359 vector pixel vec_sro (vector pixel, vector signed char);
9360 vector pixel vec_sro (vector pixel, vector unsigned char);
9361 vector signed char vec_sro (vector signed char, vector signed char);
9362 vector signed char vec_sro (vector signed char, vector unsigned char);
9363 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9364 vector unsigned char vec_sro (vector unsigned char,
9365 vector unsigned char);
9367 void vec_st (vector float, int, vector float *);
9368 void vec_st (vector float, int, float *);
9369 void vec_st (vector signed int, int, vector signed int *);
9370 void vec_st (vector signed int, int, int *);
9371 void vec_st (vector unsigned int, int, vector unsigned int *);
9372 void vec_st (vector unsigned int, int, unsigned int *);
9373 void vec_st (vector bool int, int, vector bool int *);
9374 void vec_st (vector bool int, int, unsigned int *);
9375 void vec_st (vector bool int, int, int *);
9376 void vec_st (vector signed short, int, vector signed short *);
9377 void vec_st (vector signed short, int, short *);
9378 void vec_st (vector unsigned short, int, vector unsigned short *);
9379 void vec_st (vector unsigned short, int, unsigned short *);
9380 void vec_st (vector bool short, int, vector bool short *);
9381 void vec_st (vector bool short, int, unsigned short *);
9382 void vec_st (vector pixel, int, vector pixel *);
9383 void vec_st (vector pixel, int, unsigned short *);
9384 void vec_st (vector pixel, int, short *);
9385 void vec_st (vector bool short, int, short *);
9386 void vec_st (vector signed char, int, vector signed char *);
9387 void vec_st (vector signed char, int, signed char *);
9388 void vec_st (vector unsigned char, int, vector unsigned char *);
9389 void vec_st (vector unsigned char, int, unsigned char *);
9390 void vec_st (vector bool char, int, vector bool char *);
9391 void vec_st (vector bool char, int, unsigned char *);
9392 void vec_st (vector bool char, int, signed char *);
9394 void vec_ste (vector signed char, int, signed char *);
9395 void vec_ste (vector unsigned char, int, unsigned char *);
9396 void vec_ste (vector bool char, int, signed char *);
9397 void vec_ste (vector bool char, int, unsigned char *);
9398 void vec_ste (vector signed short, int, short *);
9399 void vec_ste (vector unsigned short, int, unsigned short *);
9400 void vec_ste (vector bool short, int, short *);
9401 void vec_ste (vector bool short, int, unsigned short *);
9402 void vec_ste (vector pixel, int, short *);
9403 void vec_ste (vector pixel, int, unsigned short *);
9404 void vec_ste (vector float, int, float *);
9405 void vec_ste (vector signed int, int, int *);
9406 void vec_ste (vector unsigned int, int, unsigned int *);
9407 void vec_ste (vector bool int, int, int *);
9408 void vec_ste (vector bool int, int, unsigned int *);
9410 void vec_stvewx (vector float, int, float *);
9411 void vec_stvewx (vector signed int, int, int *);
9412 void vec_stvewx (vector unsigned int, int, unsigned int *);
9413 void vec_stvewx (vector bool int, int, int *);
9414 void vec_stvewx (vector bool int, int, unsigned int *);
9416 void vec_stvehx (vector signed short, int, short *);
9417 void vec_stvehx (vector unsigned short, int, unsigned short *);
9418 void vec_stvehx (vector bool short, int, short *);
9419 void vec_stvehx (vector bool short, int, unsigned short *);
9420 void vec_stvehx (vector pixel, int, short *);
9421 void vec_stvehx (vector pixel, int, unsigned short *);
9423 void vec_stvebx (vector signed char, int, signed char *);
9424 void vec_stvebx (vector unsigned char, int, unsigned char *);
9425 void vec_stvebx (vector bool char, int, signed char *);
9426 void vec_stvebx (vector bool char, int, unsigned char *);
9428 void vec_stl (vector float, int, vector float *);
9429 void vec_stl (vector float, int, float *);
9430 void vec_stl (vector signed int, int, vector signed int *);
9431 void vec_stl (vector signed int, int, int *);
9432 void vec_stl (vector unsigned int, int, vector unsigned int *);
9433 void vec_stl (vector unsigned int, int, unsigned int *);
9434 void vec_stl (vector bool int, int, vector bool int *);
9435 void vec_stl (vector bool int, int, unsigned int *);
9436 void vec_stl (vector bool int, int, int *);
9437 void vec_stl (vector signed short, int, vector signed short *);
9438 void vec_stl (vector signed short, int, short *);
9439 void vec_stl (vector unsigned short, int, vector unsigned short *);
9440 void vec_stl (vector unsigned short, int, unsigned short *);
9441 void vec_stl (vector bool short, int, vector bool short *);
9442 void vec_stl (vector bool short, int, unsigned short *);
9443 void vec_stl (vector bool short, int, short *);
9444 void vec_stl (vector pixel, int, vector pixel *);
9445 void vec_stl (vector pixel, int, unsigned short *);
9446 void vec_stl (vector pixel, int, short *);
9447 void vec_stl (vector signed char, int, vector signed char *);
9448 void vec_stl (vector signed char, int, signed char *);
9449 void vec_stl (vector unsigned char, int, vector unsigned char *);
9450 void vec_stl (vector unsigned char, int, unsigned char *);
9451 void vec_stl (vector bool char, int, vector bool char *);
9452 void vec_stl (vector bool char, int, unsigned char *);
9453 void vec_stl (vector bool char, int, signed char *);
9455 vector signed char vec_sub (vector bool char, vector signed char);
9456 vector signed char vec_sub (vector signed char, vector bool char);
9457 vector signed char vec_sub (vector signed char, vector signed char);
9458 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9459 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9460 vector unsigned char vec_sub (vector unsigned char,
9461 vector unsigned char);
9462 vector signed short vec_sub (vector bool short, vector signed short);
9463 vector signed short vec_sub (vector signed short, vector bool short);
9464 vector signed short vec_sub (vector signed short, vector signed short);
9465 vector unsigned short vec_sub (vector bool short,
9466 vector unsigned short);
9467 vector unsigned short vec_sub (vector unsigned short,
9469 vector unsigned short vec_sub (vector unsigned short,
9470 vector unsigned short);
9471 vector signed int vec_sub (vector bool int, vector signed int);
9472 vector signed int vec_sub (vector signed int, vector bool int);
9473 vector signed int vec_sub (vector signed int, vector signed int);
9474 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9475 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9476 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9477 vector float vec_sub (vector float, vector float);
9479 vector float vec_vsubfp (vector float, vector float);
9481 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9482 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9483 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9484 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9485 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9486 vector unsigned int vec_vsubuwm (vector unsigned int,
9487 vector unsigned int);
9489 vector signed short vec_vsubuhm (vector bool short,
9490 vector signed short);
9491 vector signed short vec_vsubuhm (vector signed short,
9493 vector signed short vec_vsubuhm (vector signed short,
9494 vector signed short);
9495 vector unsigned short vec_vsubuhm (vector bool short,
9496 vector unsigned short);
9497 vector unsigned short vec_vsubuhm (vector unsigned short,
9499 vector unsigned short vec_vsubuhm (vector unsigned short,
9500 vector unsigned short);
9502 vector signed char vec_vsububm (vector bool char, vector signed char);
9503 vector signed char vec_vsububm (vector signed char, vector bool char);
9504 vector signed char vec_vsububm (vector signed char, vector signed char);
9505 vector unsigned char vec_vsububm (vector bool char,
9506 vector unsigned char);
9507 vector unsigned char vec_vsububm (vector unsigned char,
9509 vector unsigned char vec_vsububm (vector unsigned char,
9510 vector unsigned char);
9512 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9514 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9515 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9516 vector unsigned char vec_subs (vector unsigned char,
9517 vector unsigned char);
9518 vector signed char vec_subs (vector bool char, vector signed char);
9519 vector signed char vec_subs (vector signed char, vector bool char);
9520 vector signed char vec_subs (vector signed char, vector signed char);
9521 vector unsigned short vec_subs (vector bool short,
9522 vector unsigned short);
9523 vector unsigned short vec_subs (vector unsigned short,
9525 vector unsigned short vec_subs (vector unsigned short,
9526 vector unsigned short);
9527 vector signed short vec_subs (vector bool short, vector signed short);
9528 vector signed short vec_subs (vector signed short, vector bool short);
9529 vector signed short vec_subs (vector signed short, vector signed short);
9530 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9531 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9532 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9533 vector signed int vec_subs (vector bool int, vector signed int);
9534 vector signed int vec_subs (vector signed int, vector bool int);
9535 vector signed int vec_subs (vector signed int, vector signed int);
9537 vector signed int vec_vsubsws (vector bool int, vector signed int);
9538 vector signed int vec_vsubsws (vector signed int, vector bool int);
9539 vector signed int vec_vsubsws (vector signed int, vector signed int);
9541 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9542 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9543 vector unsigned int vec_vsubuws (vector unsigned int,
9544 vector unsigned int);
9546 vector signed short vec_vsubshs (vector bool short,
9547 vector signed short);
9548 vector signed short vec_vsubshs (vector signed short,
9550 vector signed short vec_vsubshs (vector signed short,
9551 vector signed short);
9553 vector unsigned short vec_vsubuhs (vector bool short,
9554 vector unsigned short);
9555 vector unsigned short vec_vsubuhs (vector unsigned short,
9557 vector unsigned short vec_vsubuhs (vector unsigned short,
9558 vector unsigned short);
9560 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9561 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9562 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9564 vector unsigned char vec_vsububs (vector bool char,
9565 vector unsigned char);
9566 vector unsigned char vec_vsububs (vector unsigned char,
9568 vector unsigned char vec_vsububs (vector unsigned char,
9569 vector unsigned char);
9571 vector unsigned int vec_sum4s (vector unsigned char,
9572 vector unsigned int);
9573 vector signed int vec_sum4s (vector signed char, vector signed int);
9574 vector signed int vec_sum4s (vector signed short, vector signed int);
9576 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9578 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9580 vector unsigned int vec_vsum4ubs (vector unsigned char,
9581 vector unsigned int);
9583 vector signed int vec_sum2s (vector signed int, vector signed int);
9585 vector signed int vec_sums (vector signed int, vector signed int);
9587 vector float vec_trunc (vector float);
9589 vector signed short vec_unpackh (vector signed char);
9590 vector bool short vec_unpackh (vector bool char);
9591 vector signed int vec_unpackh (vector signed short);
9592 vector bool int vec_unpackh (vector bool short);
9593 vector unsigned int vec_unpackh (vector pixel);
9595 vector bool int vec_vupkhsh (vector bool short);
9596 vector signed int vec_vupkhsh (vector signed short);
9598 vector unsigned int vec_vupkhpx (vector pixel);
9600 vector bool short vec_vupkhsb (vector bool char);
9601 vector signed short vec_vupkhsb (vector signed char);
9603 vector signed short vec_unpackl (vector signed char);
9604 vector bool short vec_unpackl (vector bool char);
9605 vector unsigned int vec_unpackl (vector pixel);
9606 vector signed int vec_unpackl (vector signed short);
9607 vector bool int vec_unpackl (vector bool short);
9609 vector unsigned int vec_vupklpx (vector pixel);
9611 vector bool int vec_vupklsh (vector bool short);
9612 vector signed int vec_vupklsh (vector signed short);
9614 vector bool short vec_vupklsb (vector bool char);
9615 vector signed short vec_vupklsb (vector signed char);
9617 vector float vec_xor (vector float, vector float);
9618 vector float vec_xor (vector float, vector bool int);
9619 vector float vec_xor (vector bool int, vector float);
9620 vector bool int vec_xor (vector bool int, vector bool int);
9621 vector signed int vec_xor (vector bool int, vector signed int);
9622 vector signed int vec_xor (vector signed int, vector bool int);
9623 vector signed int vec_xor (vector signed int, vector signed int);
9624 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9625 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9626 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9627 vector bool short vec_xor (vector bool short, vector bool short);
9628 vector signed short vec_xor (vector bool short, vector signed short);
9629 vector signed short vec_xor (vector signed short, vector bool short);
9630 vector signed short vec_xor (vector signed short, vector signed short);
9631 vector unsigned short vec_xor (vector bool short,
9632 vector unsigned short);
9633 vector unsigned short vec_xor (vector unsigned short,
9635 vector unsigned short vec_xor (vector unsigned short,
9636 vector unsigned short);
9637 vector signed char vec_xor (vector bool char, vector signed char);
9638 vector bool char vec_xor (vector bool char, vector bool char);
9639 vector signed char vec_xor (vector signed char, vector bool char);
9640 vector signed char vec_xor (vector signed char, vector signed char);
9641 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9642 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9643 vector unsigned char vec_xor (vector unsigned char,
9644 vector unsigned char);
9646 int vec_all_eq (vector signed char, vector bool char);
9647 int vec_all_eq (vector signed char, vector signed char);
9648 int vec_all_eq (vector unsigned char, vector bool char);
9649 int vec_all_eq (vector unsigned char, vector unsigned char);
9650 int vec_all_eq (vector bool char, vector bool char);
9651 int vec_all_eq (vector bool char, vector unsigned char);
9652 int vec_all_eq (vector bool char, vector signed char);
9653 int vec_all_eq (vector signed short, vector bool short);
9654 int vec_all_eq (vector signed short, vector signed short);
9655 int vec_all_eq (vector unsigned short, vector bool short);
9656 int vec_all_eq (vector unsigned short, vector unsigned short);
9657 int vec_all_eq (vector bool short, vector bool short);
9658 int vec_all_eq (vector bool short, vector unsigned short);
9659 int vec_all_eq (vector bool short, vector signed short);
9660 int vec_all_eq (vector pixel, vector pixel);
9661 int vec_all_eq (vector signed int, vector bool int);
9662 int vec_all_eq (vector signed int, vector signed int);
9663 int vec_all_eq (vector unsigned int, vector bool int);
9664 int vec_all_eq (vector unsigned int, vector unsigned int);
9665 int vec_all_eq (vector bool int, vector bool int);
9666 int vec_all_eq (vector bool int, vector unsigned int);
9667 int vec_all_eq (vector bool int, vector signed int);
9668 int vec_all_eq (vector float, vector float);
9670 int vec_all_ge (vector bool char, vector unsigned char);
9671 int vec_all_ge (vector unsigned char, vector bool char);
9672 int vec_all_ge (vector unsigned char, vector unsigned char);
9673 int vec_all_ge (vector bool char, vector signed char);
9674 int vec_all_ge (vector signed char, vector bool char);
9675 int vec_all_ge (vector signed char, vector signed char);
9676 int vec_all_ge (vector bool short, vector unsigned short);
9677 int vec_all_ge (vector unsigned short, vector bool short);
9678 int vec_all_ge (vector unsigned short, vector unsigned short);
9679 int vec_all_ge (vector signed short, vector signed short);
9680 int vec_all_ge (vector bool short, vector signed short);
9681 int vec_all_ge (vector signed short, vector bool short);
9682 int vec_all_ge (vector bool int, vector unsigned int);
9683 int vec_all_ge (vector unsigned int, vector bool int);
9684 int vec_all_ge (vector unsigned int, vector unsigned int);
9685 int vec_all_ge (vector bool int, vector signed int);
9686 int vec_all_ge (vector signed int, vector bool int);
9687 int vec_all_ge (vector signed int, vector signed int);
9688 int vec_all_ge (vector float, vector float);
9690 int vec_all_gt (vector bool char, vector unsigned char);
9691 int vec_all_gt (vector unsigned char, vector bool char);
9692 int vec_all_gt (vector unsigned char, vector unsigned char);
9693 int vec_all_gt (vector bool char, vector signed char);
9694 int vec_all_gt (vector signed char, vector bool char);
9695 int vec_all_gt (vector signed char, vector signed char);
9696 int vec_all_gt (vector bool short, vector unsigned short);
9697 int vec_all_gt (vector unsigned short, vector bool short);
9698 int vec_all_gt (vector unsigned short, vector unsigned short);
9699 int vec_all_gt (vector bool short, vector signed short);
9700 int vec_all_gt (vector signed short, vector bool short);
9701 int vec_all_gt (vector signed short, vector signed short);
9702 int vec_all_gt (vector bool int, vector unsigned int);
9703 int vec_all_gt (vector unsigned int, vector bool int);
9704 int vec_all_gt (vector unsigned int, vector unsigned int);
9705 int vec_all_gt (vector bool int, vector signed int);
9706 int vec_all_gt (vector signed int, vector bool int);
9707 int vec_all_gt (vector signed int, vector signed int);
9708 int vec_all_gt (vector float, vector float);
9710 int vec_all_in (vector float, vector float);
9712 int vec_all_le (vector bool char, vector unsigned char);
9713 int vec_all_le (vector unsigned char, vector bool char);
9714 int vec_all_le (vector unsigned char, vector unsigned char);
9715 int vec_all_le (vector bool char, vector signed char);
9716 int vec_all_le (vector signed char, vector bool char);
9717 int vec_all_le (vector signed char, vector signed char);
9718 int vec_all_le (vector bool short, vector unsigned short);
9719 int vec_all_le (vector unsigned short, vector bool short);
9720 int vec_all_le (vector unsigned short, vector unsigned short);
9721 int vec_all_le (vector bool short, vector signed short);
9722 int vec_all_le (vector signed short, vector bool short);
9723 int vec_all_le (vector signed short, vector signed short);
9724 int vec_all_le (vector bool int, vector unsigned int);
9725 int vec_all_le (vector unsigned int, vector bool int);
9726 int vec_all_le (vector unsigned int, vector unsigned int);
9727 int vec_all_le (vector bool int, vector signed int);
9728 int vec_all_le (vector signed int, vector bool int);
9729 int vec_all_le (vector signed int, vector signed int);
9730 int vec_all_le (vector float, vector float);
9732 int vec_all_lt (vector bool char, vector unsigned char);
9733 int vec_all_lt (vector unsigned char, vector bool char);
9734 int vec_all_lt (vector unsigned char, vector unsigned char);
9735 int vec_all_lt (vector bool char, vector signed char);
9736 int vec_all_lt (vector signed char, vector bool char);
9737 int vec_all_lt (vector signed char, vector signed char);
9738 int vec_all_lt (vector bool short, vector unsigned short);
9739 int vec_all_lt (vector unsigned short, vector bool short);
9740 int vec_all_lt (vector unsigned short, vector unsigned short);
9741 int vec_all_lt (vector bool short, vector signed short);
9742 int vec_all_lt (vector signed short, vector bool short);
9743 int vec_all_lt (vector signed short, vector signed short);
9744 int vec_all_lt (vector bool int, vector unsigned int);
9745 int vec_all_lt (vector unsigned int, vector bool int);
9746 int vec_all_lt (vector unsigned int, vector unsigned int);
9747 int vec_all_lt (vector bool int, vector signed int);
9748 int vec_all_lt (vector signed int, vector bool int);
9749 int vec_all_lt (vector signed int, vector signed int);
9750 int vec_all_lt (vector float, vector float);
9752 int vec_all_nan (vector float);
9754 int vec_all_ne (vector signed char, vector bool char);
9755 int vec_all_ne (vector signed char, vector signed char);
9756 int vec_all_ne (vector unsigned char, vector bool char);
9757 int vec_all_ne (vector unsigned char, vector unsigned char);
9758 int vec_all_ne (vector bool char, vector bool char);
9759 int vec_all_ne (vector bool char, vector unsigned char);
9760 int vec_all_ne (vector bool char, vector signed char);
9761 int vec_all_ne (vector signed short, vector bool short);
9762 int vec_all_ne (vector signed short, vector signed short);
9763 int vec_all_ne (vector unsigned short, vector bool short);
9764 int vec_all_ne (vector unsigned short, vector unsigned short);
9765 int vec_all_ne (vector bool short, vector bool short);
9766 int vec_all_ne (vector bool short, vector unsigned short);
9767 int vec_all_ne (vector bool short, vector signed short);
9768 int vec_all_ne (vector pixel, vector pixel);
9769 int vec_all_ne (vector signed int, vector bool int);
9770 int vec_all_ne (vector signed int, vector signed int);
9771 int vec_all_ne (vector unsigned int, vector bool int);
9772 int vec_all_ne (vector unsigned int, vector unsigned int);
9773 int vec_all_ne (vector bool int, vector bool int);
9774 int vec_all_ne (vector bool int, vector unsigned int);
9775 int vec_all_ne (vector bool int, vector signed int);
9776 int vec_all_ne (vector float, vector float);
9778 int vec_all_nge (vector float, vector float);
9780 int vec_all_ngt (vector float, vector float);
9782 int vec_all_nle (vector float, vector float);
9784 int vec_all_nlt (vector float, vector float);
9786 int vec_all_numeric (vector float);
9788 int vec_any_eq (vector signed char, vector bool char);
9789 int vec_any_eq (vector signed char, vector signed char);
9790 int vec_any_eq (vector unsigned char, vector bool char);
9791 int vec_any_eq (vector unsigned char, vector unsigned char);
9792 int vec_any_eq (vector bool char, vector bool char);
9793 int vec_any_eq (vector bool char, vector unsigned char);
9794 int vec_any_eq (vector bool char, vector signed char);
9795 int vec_any_eq (vector signed short, vector bool short);
9796 int vec_any_eq (vector signed short, vector signed short);
9797 int vec_any_eq (vector unsigned short, vector bool short);
9798 int vec_any_eq (vector unsigned short, vector unsigned short);
9799 int vec_any_eq (vector bool short, vector bool short);
9800 int vec_any_eq (vector bool short, vector unsigned short);
9801 int vec_any_eq (vector bool short, vector signed short);
9802 int vec_any_eq (vector pixel, vector pixel);
9803 int vec_any_eq (vector signed int, vector bool int);
9804 int vec_any_eq (vector signed int, vector signed int);
9805 int vec_any_eq (vector unsigned int, vector bool int);
9806 int vec_any_eq (vector unsigned int, vector unsigned int);
9807 int vec_any_eq (vector bool int, vector bool int);
9808 int vec_any_eq (vector bool int, vector unsigned int);
9809 int vec_any_eq (vector bool int, vector signed int);
9810 int vec_any_eq (vector float, vector float);
9812 int vec_any_ge (vector signed char, vector bool char);
9813 int vec_any_ge (vector unsigned char, vector bool char);
9814 int vec_any_ge (vector unsigned char, vector unsigned char);
9815 int vec_any_ge (vector signed char, vector signed char);
9816 int vec_any_ge (vector bool char, vector unsigned char);
9817 int vec_any_ge (vector bool char, vector signed char);
9818 int vec_any_ge (vector unsigned short, vector bool short);
9819 int vec_any_ge (vector unsigned short, vector unsigned short);
9820 int vec_any_ge (vector signed short, vector signed short);
9821 int vec_any_ge (vector signed short, vector bool short);
9822 int vec_any_ge (vector bool short, vector unsigned short);
9823 int vec_any_ge (vector bool short, vector signed short);
9824 int vec_any_ge (vector signed int, vector bool int);
9825 int vec_any_ge (vector unsigned int, vector bool int);
9826 int vec_any_ge (vector unsigned int, vector unsigned int);
9827 int vec_any_ge (vector signed int, vector signed int);
9828 int vec_any_ge (vector bool int, vector unsigned int);
9829 int vec_any_ge (vector bool int, vector signed int);
9830 int vec_any_ge (vector float, vector float);
9832 int vec_any_gt (vector bool char, vector unsigned char);
9833 int vec_any_gt (vector unsigned char, vector bool char);
9834 int vec_any_gt (vector unsigned char, vector unsigned char);
9835 int vec_any_gt (vector bool char, vector signed char);
9836 int vec_any_gt (vector signed char, vector bool char);
9837 int vec_any_gt (vector signed char, vector signed char);
9838 int vec_any_gt (vector bool short, vector unsigned short);
9839 int vec_any_gt (vector unsigned short, vector bool short);
9840 int vec_any_gt (vector unsigned short, vector unsigned short);
9841 int vec_any_gt (vector bool short, vector signed short);
9842 int vec_any_gt (vector signed short, vector bool short);
9843 int vec_any_gt (vector signed short, vector signed short);
9844 int vec_any_gt (vector bool int, vector unsigned int);
9845 int vec_any_gt (vector unsigned int, vector bool int);
9846 int vec_any_gt (vector unsigned int, vector unsigned int);
9847 int vec_any_gt (vector bool int, vector signed int);
9848 int vec_any_gt (vector signed int, vector bool int);
9849 int vec_any_gt (vector signed int, vector signed int);
9850 int vec_any_gt (vector float, vector float);
9852 int vec_any_le (vector bool char, vector unsigned char);
9853 int vec_any_le (vector unsigned char, vector bool char);
9854 int vec_any_le (vector unsigned char, vector unsigned char);
9855 int vec_any_le (vector bool char, vector signed char);
9856 int vec_any_le (vector signed char, vector bool char);
9857 int vec_any_le (vector signed char, vector signed char);
9858 int vec_any_le (vector bool short, vector unsigned short);
9859 int vec_any_le (vector unsigned short, vector bool short);
9860 int vec_any_le (vector unsigned short, vector unsigned short);
9861 int vec_any_le (vector bool short, vector signed short);
9862 int vec_any_le (vector signed short, vector bool short);
9863 int vec_any_le (vector signed short, vector signed short);
9864 int vec_any_le (vector bool int, vector unsigned int);
9865 int vec_any_le (vector unsigned int, vector bool int);
9866 int vec_any_le (vector unsigned int, vector unsigned int);
9867 int vec_any_le (vector bool int, vector signed int);
9868 int vec_any_le (vector signed int, vector bool int);
9869 int vec_any_le (vector signed int, vector signed int);
9870 int vec_any_le (vector float, vector float);
9872 int vec_any_lt (vector bool char, vector unsigned char);
9873 int vec_any_lt (vector unsigned char, vector bool char);
9874 int vec_any_lt (vector unsigned char, vector unsigned char);
9875 int vec_any_lt (vector bool char, vector signed char);
9876 int vec_any_lt (vector signed char, vector bool char);
9877 int vec_any_lt (vector signed char, vector signed char);
9878 int vec_any_lt (vector bool short, vector unsigned short);
9879 int vec_any_lt (vector unsigned short, vector bool short);
9880 int vec_any_lt (vector unsigned short, vector unsigned short);
9881 int vec_any_lt (vector bool short, vector signed short);
9882 int vec_any_lt (vector signed short, vector bool short);
9883 int vec_any_lt (vector signed short, vector signed short);
9884 int vec_any_lt (vector bool int, vector unsigned int);
9885 int vec_any_lt (vector unsigned int, vector bool int);
9886 int vec_any_lt (vector unsigned int, vector unsigned int);
9887 int vec_any_lt (vector bool int, vector signed int);
9888 int vec_any_lt (vector signed int, vector bool int);
9889 int vec_any_lt (vector signed int, vector signed int);
9890 int vec_any_lt (vector float, vector float);
9892 int vec_any_nan (vector float);
9894 int vec_any_ne (vector signed char, vector bool char);
9895 int vec_any_ne (vector signed char, vector signed char);
9896 int vec_any_ne (vector unsigned char, vector bool char);
9897 int vec_any_ne (vector unsigned char, vector unsigned char);
9898 int vec_any_ne (vector bool char, vector bool char);
9899 int vec_any_ne (vector bool char, vector unsigned char);
9900 int vec_any_ne (vector bool char, vector signed char);
9901 int vec_any_ne (vector signed short, vector bool short);
9902 int vec_any_ne (vector signed short, vector signed short);
9903 int vec_any_ne (vector unsigned short, vector bool short);
9904 int vec_any_ne (vector unsigned short, vector unsigned short);
9905 int vec_any_ne (vector bool short, vector bool short);
9906 int vec_any_ne (vector bool short, vector unsigned short);
9907 int vec_any_ne (vector bool short, vector signed short);
9908 int vec_any_ne (vector pixel, vector pixel);
9909 int vec_any_ne (vector signed int, vector bool int);
9910 int vec_any_ne (vector signed int, vector signed int);
9911 int vec_any_ne (vector unsigned int, vector bool int);
9912 int vec_any_ne (vector unsigned int, vector unsigned int);
9913 int vec_any_ne (vector bool int, vector bool int);
9914 int vec_any_ne (vector bool int, vector unsigned int);
9915 int vec_any_ne (vector bool int, vector signed int);
9916 int vec_any_ne (vector float, vector float);
9918 int vec_any_nge (vector float, vector float);
9920 int vec_any_ngt (vector float, vector float);
9922 int vec_any_nle (vector float, vector float);
9924 int vec_any_nlt (vector float, vector float);
9926 int vec_any_numeric (vector float);
9928 int vec_any_out (vector float, vector float);
9931 @node SPARC VIS Built-in Functions
9932 @subsection SPARC VIS Built-in Functions
9934 GCC supports SIMD operations on the SPARC using both the generic vector
9935 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9936 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9937 switch, the VIS extension is exposed as the following built-in functions:
9940 typedef int v2si __attribute__ ((vector_size (8)));
9941 typedef short v4hi __attribute__ ((vector_size (8)));
9942 typedef short v2hi __attribute__ ((vector_size (4)));
9943 typedef char v8qi __attribute__ ((vector_size (8)));
9944 typedef char v4qi __attribute__ ((vector_size (4)));
9946 void * __builtin_vis_alignaddr (void *, long);
9947 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9948 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9949 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9950 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9952 v4hi __builtin_vis_fexpand (v4qi);
9954 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9955 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9956 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9957 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9958 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9959 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9960 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9962 v4qi __builtin_vis_fpack16 (v4hi);
9963 v8qi __builtin_vis_fpack32 (v2si, v2si);
9964 v2hi __builtin_vis_fpackfix (v2si);
9965 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9967 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9970 @node SPU Built-in Functions
9971 @subsection SPU Built-in Functions
9973 GCC provides extensions for the SPU processor as described in the
9974 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
9975 found at @uref{http://cell.scei.co.jp/} or
9976 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
9977 implementation differs in several ways.
9982 The optional extension of specifying vector constants in parentheses is
9986 A vector initializer requires no cast if the vector constant is of the
9987 same type as the variable it is initializing.
9990 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9991 vector type is the default signedness of the base type. The default
9992 varies depending on the operating system, so a portable program should
9993 always specify the signedness.
9996 By default, the keyword @code{__vector} is added. The macro
9997 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
10001 GCC allows using a @code{typedef} name as the type specifier for a
10005 For C, overloaded functions are implemented with macros so the following
10009 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
10012 Since @code{spu_add} is a macro, the vector constant in the example
10013 is treated as four separate arguments. Wrap the entire argument in
10014 parentheses for this to work.
10017 The extended version of @code{__builtin_expect} is not supported.
10021 @emph{Note:} Only the interface described in the aforementioned
10022 specification is supported. Internally, GCC uses built-in functions to
10023 implement the required functionality, but these are not supported and
10024 are subject to change without notice.
10026 @node Target Format Checks
10027 @section Format Checks Specific to Particular Target Machines
10029 For some target machines, GCC supports additional options to the
10031 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
10034 * Solaris Format Checks::
10037 @node Solaris Format Checks
10038 @subsection Solaris Format Checks
10040 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
10041 check. @code{cmn_err} accepts a subset of the standard @code{printf}
10042 conversions, and the two-argument @code{%b} conversion for displaying
10043 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
10046 @section Pragmas Accepted by GCC
10050 GCC supports several types of pragmas, primarily in order to compile
10051 code originally written for other compilers. Note that in general
10052 we do not recommend the use of pragmas; @xref{Function Attributes},
10053 for further explanation.
10058 * RS/6000 and PowerPC Pragmas::
10060 * Solaris Pragmas::
10061 * Symbol-Renaming Pragmas::
10062 * Structure-Packing Pragmas::
10064 * Diagnostic Pragmas::
10065 * Visibility Pragmas::
10069 @subsection ARM Pragmas
10071 The ARM target defines pragmas for controlling the default addition of
10072 @code{long_call} and @code{short_call} attributes to functions.
10073 @xref{Function Attributes}, for information about the effects of these
10078 @cindex pragma, long_calls
10079 Set all subsequent functions to have the @code{long_call} attribute.
10081 @item no_long_calls
10082 @cindex pragma, no_long_calls
10083 Set all subsequent functions to have the @code{short_call} attribute.
10085 @item long_calls_off
10086 @cindex pragma, long_calls_off
10087 Do not affect the @code{long_call} or @code{short_call} attributes of
10088 subsequent functions.
10092 @subsection M32C Pragmas
10095 @item memregs @var{number}
10096 @cindex pragma, memregs
10097 Overrides the command line option @code{-memregs=} for the current
10098 file. Use with care! This pragma must be before any function in the
10099 file, and mixing different memregs values in different objects may
10100 make them incompatible. This pragma is useful when a
10101 performance-critical function uses a memreg for temporary values,
10102 as it may allow you to reduce the number of memregs used.
10106 @node RS/6000 and PowerPC Pragmas
10107 @subsection RS/6000 and PowerPC Pragmas
10109 The RS/6000 and PowerPC targets define one pragma for controlling
10110 whether or not the @code{longcall} attribute is added to function
10111 declarations by default. This pragma overrides the @option{-mlongcall}
10112 option, but not the @code{longcall} and @code{shortcall} attributes.
10113 @xref{RS/6000 and PowerPC Options}, for more information about when long
10114 calls are and are not necessary.
10118 @cindex pragma, longcall
10119 Apply the @code{longcall} attribute to all subsequent function
10123 Do not apply the @code{longcall} attribute to subsequent function
10127 @c Describe c4x pragmas here.
10128 @c Describe h8300 pragmas here.
10129 @c Describe sh pragmas here.
10130 @c Describe v850 pragmas here.
10132 @node Darwin Pragmas
10133 @subsection Darwin Pragmas
10135 The following pragmas are available for all architectures running the
10136 Darwin operating system. These are useful for compatibility with other
10140 @item mark @var{tokens}@dots{}
10141 @cindex pragma, mark
10142 This pragma is accepted, but has no effect.
10144 @item options align=@var{alignment}
10145 @cindex pragma, options align
10146 This pragma sets the alignment of fields in structures. The values of
10147 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
10148 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
10149 properly; to restore the previous setting, use @code{reset} for the
10152 @item segment @var{tokens}@dots{}
10153 @cindex pragma, segment
10154 This pragma is accepted, but has no effect.
10156 @item unused (@var{var} [, @var{var}]@dots{})
10157 @cindex pragma, unused
10158 This pragma declares variables to be possibly unused. GCC will not
10159 produce warnings for the listed variables. The effect is similar to
10160 that of the @code{unused} attribute, except that this pragma may appear
10161 anywhere within the variables' scopes.
10164 @node Solaris Pragmas
10165 @subsection Solaris Pragmas
10167 The Solaris target supports @code{#pragma redefine_extname}
10168 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
10169 @code{#pragma} directives for compatibility with the system compiler.
10172 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
10173 @cindex pragma, align
10175 Increase the minimum alignment of each @var{variable} to @var{alignment}.
10176 This is the same as GCC's @code{aligned} attribute @pxref{Variable
10177 Attributes}). Macro expansion occurs on the arguments to this pragma
10178 when compiling C and Objective-C. It does not currently occur when
10179 compiling C++, but this is a bug which may be fixed in a future
10182 @item fini (@var{function} [, @var{function}]...)
10183 @cindex pragma, fini
10185 This pragma causes each listed @var{function} to be called after
10186 main, or during shared module unloading, by adding a call to the
10187 @code{.fini} section.
10189 @item init (@var{function} [, @var{function}]...)
10190 @cindex pragma, init
10192 This pragma causes each listed @var{function} to be called during
10193 initialization (before @code{main}) or during shared module loading, by
10194 adding a call to the @code{.init} section.
10198 @node Symbol-Renaming Pragmas
10199 @subsection Symbol-Renaming Pragmas
10201 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
10202 supports two @code{#pragma} directives which change the name used in
10203 assembly for a given declaration. These pragmas are only available on
10204 platforms whose system headers need them. To get this effect on all
10205 platforms supported by GCC, use the asm labels extension (@pxref{Asm
10209 @item redefine_extname @var{oldname} @var{newname}
10210 @cindex pragma, redefine_extname
10212 This pragma gives the C function @var{oldname} the assembly symbol
10213 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
10214 will be defined if this pragma is available (currently only on
10217 @item extern_prefix @var{string}
10218 @cindex pragma, extern_prefix
10220 This pragma causes all subsequent external function and variable
10221 declarations to have @var{string} prepended to their assembly symbols.
10222 This effect may be terminated with another @code{extern_prefix} pragma
10223 whose argument is an empty string. The preprocessor macro
10224 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10225 available (currently only on Tru64 UNIX)@.
10228 These pragmas and the asm labels extension interact in a complicated
10229 manner. Here are some corner cases you may want to be aware of.
10232 @item Both pragmas silently apply only to declarations with external
10233 linkage. Asm labels do not have this restriction.
10235 @item In C++, both pragmas silently apply only to declarations with
10236 ``C'' linkage. Again, asm labels do not have this restriction.
10238 @item If any of the three ways of changing the assembly name of a
10239 declaration is applied to a declaration whose assembly name has
10240 already been determined (either by a previous use of one of these
10241 features, or because the compiler needed the assembly name in order to
10242 generate code), and the new name is different, a warning issues and
10243 the name does not change.
10245 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10246 always the C-language name.
10248 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10249 occurs with an asm label attached, the prefix is silently ignored for
10252 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10253 apply to the same declaration, whichever triggered first wins, and a
10254 warning issues if they contradict each other. (We would like to have
10255 @code{#pragma redefine_extname} always win, for consistency with asm
10256 labels, but if @code{#pragma extern_prefix} triggers first we have no
10257 way of knowing that that happened.)
10260 @node Structure-Packing Pragmas
10261 @subsection Structure-Packing Pragmas
10263 For compatibility with Win32, GCC supports a set of @code{#pragma}
10264 directives which change the maximum alignment of members of structures
10265 (other than zero-width bitfields), unions, and classes subsequently
10266 defined. The @var{n} value below always is required to be a small power
10267 of two and specifies the new alignment in bytes.
10270 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10271 @item @code{#pragma pack()} sets the alignment to the one that was in
10272 effect when compilation started (see also command line option
10273 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10274 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10275 setting on an internal stack and then optionally sets the new alignment.
10276 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10277 saved at the top of the internal stack (and removes that stack entry).
10278 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10279 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10280 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10281 @code{#pragma pack(pop)}.
10284 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10285 @code{#pragma} which lays out a structure as the documented
10286 @code{__attribute__ ((ms_struct))}.
10288 @item @code{#pragma ms_struct on} turns on the layout for structures
10290 @item @code{#pragma ms_struct off} turns off the layout for structures
10292 @item @code{#pragma ms_struct reset} goes back to the default layout.
10296 @subsection Weak Pragmas
10298 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10299 directives for declaring symbols to be weak, and defining weak
10303 @item #pragma weak @var{symbol}
10304 @cindex pragma, weak
10305 This pragma declares @var{symbol} to be weak, as if the declaration
10306 had the attribute of the same name. The pragma may appear before
10307 or after the declaration of @var{symbol}, but must appear before
10308 either its first use or its definition. It is not an error for
10309 @var{symbol} to never be defined at all.
10311 @item #pragma weak @var{symbol1} = @var{symbol2}
10312 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10313 It is an error if @var{symbol2} is not defined in the current
10317 @node Diagnostic Pragmas
10318 @subsection Diagnostic Pragmas
10320 GCC allows the user to selectively enable or disable certain types of
10321 diagnostics, and change the kind of the diagnostic. For example, a
10322 project's policy might require that all sources compile with
10323 @option{-Werror} but certain files might have exceptions allowing
10324 specific types of warnings. Or, a project might selectively enable
10325 diagnostics and treat them as errors depending on which preprocessor
10326 macros are defined.
10329 @item #pragma GCC diagnostic @var{kind} @var{option}
10330 @cindex pragma, diagnostic
10332 Modifies the disposition of a diagnostic. Note that not all
10333 diagnostics are modifiable; at the moment only warnings (normally
10334 controlled by @samp{-W...}) can be controlled, and not all of them.
10335 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10336 are controllable and which option controls them.
10338 @var{kind} is @samp{error} to treat this diagnostic as an error,
10339 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10340 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10341 @var{option} is a double quoted string which matches the command line
10345 #pragma GCC diagnostic warning "-Wformat"
10346 #pragma GCC diagnostic error "-Wformat"
10347 #pragma GCC diagnostic ignored "-Wformat"
10350 Note that these pragmas override any command line options. Also,
10351 while it is syntactically valid to put these pragmas anywhere in your
10352 sources, the only supported location for them is before any data or
10353 functions are defined. Doing otherwise may result in unpredictable
10354 results depending on how the optimizer manages your sources. If the
10355 same option is listed multiple times, the last one specified is the
10356 one that is in effect. This pragma is not intended to be a general
10357 purpose replacement for command line options, but for implementing
10358 strict control over project policies.
10362 @node Visibility Pragmas
10363 @subsection Visibility Pragmas
10366 @item #pragma GCC visibility push(@var{visibility})
10367 @itemx #pragma GCC visibility pop
10368 @cindex pragma, visibility
10370 This pragma allows the user to set the visibility for multiple
10371 declarations without having to give each a visibility attribute
10372 @xref{Function Attributes}, for more information about visibility and
10373 the attribute syntax.
10375 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10376 declarations. Class members and template specializations are not
10377 affected; if you want to override the visibility for a particular
10378 member or instantiation, you must use an attribute.
10382 @node Unnamed Fields
10383 @section Unnamed struct/union fields within structs/unions
10387 For compatibility with other compilers, GCC allows you to define
10388 a structure or union that contains, as fields, structures and unions
10389 without names. For example:
10402 In this example, the user would be able to access members of the unnamed
10403 union with code like @samp{foo.b}. Note that only unnamed structs and
10404 unions are allowed, you may not have, for example, an unnamed
10407 You must never create such structures that cause ambiguous field definitions.
10408 For example, this structure:
10419 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10420 Such constructs are not supported and must be avoided. In the future,
10421 such constructs may be detected and treated as compilation errors.
10423 @opindex fms-extensions
10424 Unless @option{-fms-extensions} is used, the unnamed field must be a
10425 structure or union definition without a tag (for example, @samp{struct
10426 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10427 also be a definition with a tag such as @samp{struct foo @{ int a;
10428 @};}, a reference to a previously defined structure or union such as
10429 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10430 previously defined structure or union type.
10433 @section Thread-Local Storage
10434 @cindex Thread-Local Storage
10435 @cindex @acronym{TLS}
10438 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10439 are allocated such that there is one instance of the variable per extant
10440 thread. The run-time model GCC uses to implement this originates
10441 in the IA-64 processor-specific ABI, but has since been migrated
10442 to other processors as well. It requires significant support from
10443 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10444 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10445 is not available everywhere.
10447 At the user level, the extension is visible with a new storage
10448 class keyword: @code{__thread}. For example:
10452 extern __thread struct state s;
10453 static __thread char *p;
10456 The @code{__thread} specifier may be used alone, with the @code{extern}
10457 or @code{static} specifiers, but with no other storage class specifier.
10458 When used with @code{extern} or @code{static}, @code{__thread} must appear
10459 immediately after the other storage class specifier.
10461 The @code{__thread} specifier may be applied to any global, file-scoped
10462 static, function-scoped static, or static data member of a class. It may
10463 not be applied to block-scoped automatic or non-static data member.
10465 When the address-of operator is applied to a thread-local variable, it is
10466 evaluated at run-time and returns the address of the current thread's
10467 instance of that variable. An address so obtained may be used by any
10468 thread. When a thread terminates, any pointers to thread-local variables
10469 in that thread become invalid.
10471 No static initialization may refer to the address of a thread-local variable.
10473 In C++, if an initializer is present for a thread-local variable, it must
10474 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10477 See @uref{http://people.redhat.com/drepper/tls.pdf,
10478 ELF Handling For Thread-Local Storage} for a detailed explanation of
10479 the four thread-local storage addressing models, and how the run-time
10480 is expected to function.
10483 * C99 Thread-Local Edits::
10484 * C++98 Thread-Local Edits::
10487 @node C99 Thread-Local Edits
10488 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10490 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10491 that document the exact semantics of the language extension.
10495 @cite{5.1.2 Execution environments}
10497 Add new text after paragraph 1
10500 Within either execution environment, a @dfn{thread} is a flow of
10501 control within a program. It is implementation defined whether
10502 or not there may be more than one thread associated with a program.
10503 It is implementation defined how threads beyond the first are
10504 created, the name and type of the function called at thread
10505 startup, and how threads may be terminated. However, objects
10506 with thread storage duration shall be initialized before thread
10511 @cite{6.2.4 Storage durations of objects}
10513 Add new text before paragraph 3
10516 An object whose identifier is declared with the storage-class
10517 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10518 Its lifetime is the entire execution of the thread, and its
10519 stored value is initialized only once, prior to thread startup.
10523 @cite{6.4.1 Keywords}
10525 Add @code{__thread}.
10528 @cite{6.7.1 Storage-class specifiers}
10530 Add @code{__thread} to the list of storage class specifiers in
10533 Change paragraph 2 to
10536 With the exception of @code{__thread}, at most one storage-class
10537 specifier may be given [@dots{}]. The @code{__thread} specifier may
10538 be used alone, or immediately following @code{extern} or
10542 Add new text after paragraph 6
10545 The declaration of an identifier for a variable that has
10546 block scope that specifies @code{__thread} shall also
10547 specify either @code{extern} or @code{static}.
10549 The @code{__thread} specifier shall be used only with
10554 @node C++98 Thread-Local Edits
10555 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10557 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10558 that document the exact semantics of the language extension.
10562 @b{[intro.execution]}
10564 New text after paragraph 4
10567 A @dfn{thread} is a flow of control within the abstract machine.
10568 It is implementation defined whether or not there may be more than
10572 New text after paragraph 7
10575 It is unspecified whether additional action must be taken to
10576 ensure when and whether side effects are visible to other threads.
10582 Add @code{__thread}.
10585 @b{[basic.start.main]}
10587 Add after paragraph 5
10590 The thread that begins execution at the @code{main} function is called
10591 the @dfn{main thread}. It is implementation defined how functions
10592 beginning threads other than the main thread are designated or typed.
10593 A function so designated, as well as the @code{main} function, is called
10594 a @dfn{thread startup function}. It is implementation defined what
10595 happens if a thread startup function returns. It is implementation
10596 defined what happens to other threads when any thread calls @code{exit}.
10600 @b{[basic.start.init]}
10602 Add after paragraph 4
10605 The storage for an object of thread storage duration shall be
10606 statically initialized before the first statement of the thread startup
10607 function. An object of thread storage duration shall not require
10608 dynamic initialization.
10612 @b{[basic.start.term]}
10614 Add after paragraph 3
10617 The type of an object with thread storage duration shall not have a
10618 non-trivial destructor, nor shall it be an array type whose elements
10619 (directly or indirectly) have non-trivial destructors.
10625 Add ``thread storage duration'' to the list in paragraph 1.
10630 Thread, static, and automatic storage durations are associated with
10631 objects introduced by declarations [@dots{}].
10634 Add @code{__thread} to the list of specifiers in paragraph 3.
10637 @b{[basic.stc.thread]}
10639 New section before @b{[basic.stc.static]}
10642 The keyword @code{__thread} applied to a non-local object gives the
10643 object thread storage duration.
10645 A local variable or class data member declared both @code{static}
10646 and @code{__thread} gives the variable or member thread storage
10651 @b{[basic.stc.static]}
10656 All objects which have neither thread storage duration, dynamic
10657 storage duration nor are local [@dots{}].
10663 Add @code{__thread} to the list in paragraph 1.
10668 With the exception of @code{__thread}, at most one
10669 @var{storage-class-specifier} shall appear in a given
10670 @var{decl-specifier-seq}. The @code{__thread} specifier may
10671 be used alone, or immediately following the @code{extern} or
10672 @code{static} specifiers. [@dots{}]
10675 Add after paragraph 5
10678 The @code{__thread} specifier can be applied only to the names of objects
10679 and to anonymous unions.
10685 Add after paragraph 6
10688 Non-@code{static} members shall not be @code{__thread}.
10692 @node C++ Extensions
10693 @chapter Extensions to the C++ Language
10694 @cindex extensions, C++ language
10695 @cindex C++ language extensions
10697 The GNU compiler provides these extensions to the C++ language (and you
10698 can also use most of the C language extensions in your C++ programs). If you
10699 want to write code that checks whether these features are available, you can
10700 test for the GNU compiler the same way as for C programs: check for a
10701 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10702 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10703 Predefined Macros,cpp,The GNU C Preprocessor}).
10706 * Volatiles:: What constitutes an access to a volatile object.
10707 * Restricted Pointers:: C99 restricted pointers and references.
10708 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10709 * C++ Interface:: You can use a single C++ header file for both
10710 declarations and definitions.
10711 * Template Instantiation:: Methods for ensuring that exactly one copy of
10712 each needed template instantiation is emitted.
10713 * Bound member functions:: You can extract a function pointer to the
10714 method denoted by a @samp{->*} or @samp{.*} expression.
10715 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10716 * Namespace Association:: Strong using-directives for namespace association.
10717 * Type Traits:: Compiler support for type traits
10718 * Java Exceptions:: Tweaking exception handling to work with Java.
10719 * Deprecated Features:: Things will disappear from g++.
10720 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10724 @section When is a Volatile Object Accessed?
10725 @cindex accessing volatiles
10726 @cindex volatile read
10727 @cindex volatile write
10728 @cindex volatile access
10730 Both the C and C++ standard have the concept of volatile objects. These
10731 are normally accessed by pointers and used for accessing hardware. The
10732 standards encourage compilers to refrain from optimizations concerning
10733 accesses to volatile objects. The C standard leaves it implementation
10734 defined as to what constitutes a volatile access. The C++ standard omits
10735 to specify this, except to say that C++ should behave in a similar manner
10736 to C with respect to volatiles, where possible. The minimum either
10737 standard specifies is that at a sequence point all previous accesses to
10738 volatile objects have stabilized and no subsequent accesses have
10739 occurred. Thus an implementation is free to reorder and combine
10740 volatile accesses which occur between sequence points, but cannot do so
10741 for accesses across a sequence point. The use of volatiles does not
10742 allow you to violate the restriction on updating objects multiple times
10743 within a sequence point.
10745 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10747 The behavior differs slightly between C and C++ in the non-obvious cases:
10750 volatile int *src = @var{somevalue};
10754 With C, such expressions are rvalues, and GCC interprets this either as a
10755 read of the volatile object being pointed to or only as request to evaluate
10756 the side-effects. The C++ standard specifies that such expressions do not
10757 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10758 object may be incomplete. The C++ standard does not specify explicitly
10759 that it is this lvalue to rvalue conversion which may be responsible for
10760 causing an access. However, there is reason to believe that it is,
10761 because otherwise certain simple expressions become undefined. However,
10762 because it would surprise most programmers, G++ treats dereferencing a
10763 pointer to volatile object of complete type when the value is unused as
10764 GCC would do for an equivalent type in C. When the object has incomplete
10765 type, G++ issues a warning; if you wish to force an error, you must
10766 force a conversion to rvalue with, for instance, a static cast.
10768 When using a reference to volatile, G++ does not treat equivalent
10769 expressions as accesses to volatiles, but instead issues a warning that
10770 no volatile is accessed. The rationale for this is that otherwise it
10771 becomes difficult to determine where volatile access occur, and not
10772 possible to ignore the return value from functions returning volatile
10773 references. Again, if you wish to force a read, cast the reference to
10776 @node Restricted Pointers
10777 @section Restricting Pointer Aliasing
10778 @cindex restricted pointers
10779 @cindex restricted references
10780 @cindex restricted this pointer
10782 As with the C front end, G++ understands the C99 feature of restricted pointers,
10783 specified with the @code{__restrict__}, or @code{__restrict} type
10784 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10785 language flag, @code{restrict} is not a keyword in C++.
10787 In addition to allowing restricted pointers, you can specify restricted
10788 references, which indicate that the reference is not aliased in the local
10792 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10799 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10800 @var{rref} refers to a (different) unaliased integer.
10802 You may also specify whether a member function's @var{this} pointer is
10803 unaliased by using @code{__restrict__} as a member function qualifier.
10806 void T::fn () __restrict__
10813 Within the body of @code{T::fn}, @var{this} will have the effective
10814 definition @code{T *__restrict__ const this}. Notice that the
10815 interpretation of a @code{__restrict__} member function qualifier is
10816 different to that of @code{const} or @code{volatile} qualifier, in that it
10817 is applied to the pointer rather than the object. This is consistent with
10818 other compilers which implement restricted pointers.
10820 As with all outermost parameter qualifiers, @code{__restrict__} is
10821 ignored in function definition matching. This means you only need to
10822 specify @code{__restrict__} in a function definition, rather than
10823 in a function prototype as well.
10825 @node Vague Linkage
10826 @section Vague Linkage
10827 @cindex vague linkage
10829 There are several constructs in C++ which require space in the object
10830 file but are not clearly tied to a single translation unit. We say that
10831 these constructs have ``vague linkage''. Typically such constructs are
10832 emitted wherever they are needed, though sometimes we can be more
10836 @item Inline Functions
10837 Inline functions are typically defined in a header file which can be
10838 included in many different compilations. Hopefully they can usually be
10839 inlined, but sometimes an out-of-line copy is necessary, if the address
10840 of the function is taken or if inlining fails. In general, we emit an
10841 out-of-line copy in all translation units where one is needed. As an
10842 exception, we only emit inline virtual functions with the vtable, since
10843 it will always require a copy.
10845 Local static variables and string constants used in an inline function
10846 are also considered to have vague linkage, since they must be shared
10847 between all inlined and out-of-line instances of the function.
10851 C++ virtual functions are implemented in most compilers using a lookup
10852 table, known as a vtable. The vtable contains pointers to the virtual
10853 functions provided by a class, and each object of the class contains a
10854 pointer to its vtable (or vtables, in some multiple-inheritance
10855 situations). If the class declares any non-inline, non-pure virtual
10856 functions, the first one is chosen as the ``key method'' for the class,
10857 and the vtable is only emitted in the translation unit where the key
10860 @emph{Note:} If the chosen key method is later defined as inline, the
10861 vtable will still be emitted in every translation unit which defines it.
10862 Make sure that any inline virtuals are declared inline in the class
10863 body, even if they are not defined there.
10865 @item type_info objects
10868 C++ requires information about types to be written out in order to
10869 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10870 For polymorphic classes (classes with virtual functions), the type_info
10871 object is written out along with the vtable so that @samp{dynamic_cast}
10872 can determine the dynamic type of a class object at runtime. For all
10873 other types, we write out the type_info object when it is used: when
10874 applying @samp{typeid} to an expression, throwing an object, or
10875 referring to a type in a catch clause or exception specification.
10877 @item Template Instantiations
10878 Most everything in this section also applies to template instantiations,
10879 but there are other options as well.
10880 @xref{Template Instantiation,,Where's the Template?}.
10884 When used with GNU ld version 2.8 or later on an ELF system such as
10885 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10886 these constructs will be discarded at link time. This is known as
10889 On targets that don't support COMDAT, but do support weak symbols, GCC
10890 will use them. This way one copy will override all the others, but
10891 the unused copies will still take up space in the executable.
10893 For targets which do not support either COMDAT or weak symbols,
10894 most entities with vague linkage will be emitted as local symbols to
10895 avoid duplicate definition errors from the linker. This will not happen
10896 for local statics in inlines, however, as having multiple copies will
10897 almost certainly break things.
10899 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10900 another way to control placement of these constructs.
10902 @node C++ Interface
10903 @section #pragma interface and implementation
10905 @cindex interface and implementation headers, C++
10906 @cindex C++ interface and implementation headers
10907 @cindex pragmas, interface and implementation
10909 @code{#pragma interface} and @code{#pragma implementation} provide the
10910 user with a way of explicitly directing the compiler to emit entities
10911 with vague linkage (and debugging information) in a particular
10914 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10915 most cases, because of COMDAT support and the ``key method'' heuristic
10916 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10917 program to grow due to unnecessary out-of-line copies of inline
10918 functions. Currently (3.4) the only benefit of these
10919 @code{#pragma}s is reduced duplication of debugging information, and
10920 that should be addressed soon on DWARF 2 targets with the use of
10924 @item #pragma interface
10925 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10926 @kindex #pragma interface
10927 Use this directive in @emph{header files} that define object classes, to save
10928 space in most of the object files that use those classes. Normally,
10929 local copies of certain information (backup copies of inline member
10930 functions, debugging information, and the internal tables that implement
10931 virtual functions) must be kept in each object file that includes class
10932 definitions. You can use this pragma to avoid such duplication. When a
10933 header file containing @samp{#pragma interface} is included in a
10934 compilation, this auxiliary information will not be generated (unless
10935 the main input source file itself uses @samp{#pragma implementation}).
10936 Instead, the object files will contain references to be resolved at link
10939 The second form of this directive is useful for the case where you have
10940 multiple headers with the same name in different directories. If you
10941 use this form, you must specify the same string to @samp{#pragma
10944 @item #pragma implementation
10945 @itemx #pragma implementation "@var{objects}.h"
10946 @kindex #pragma implementation
10947 Use this pragma in a @emph{main input file}, when you want full output from
10948 included header files to be generated (and made globally visible). The
10949 included header file, in turn, should use @samp{#pragma interface}.
10950 Backup copies of inline member functions, debugging information, and the
10951 internal tables used to implement virtual functions are all generated in
10952 implementation files.
10954 @cindex implied @code{#pragma implementation}
10955 @cindex @code{#pragma implementation}, implied
10956 @cindex naming convention, implementation headers
10957 If you use @samp{#pragma implementation} with no argument, it applies to
10958 an include file with the same basename@footnote{A file's @dfn{basename}
10959 was the name stripped of all leading path information and of trailing
10960 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10961 file. For example, in @file{allclass.cc}, giving just
10962 @samp{#pragma implementation}
10963 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10965 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10966 an implementation file whenever you would include it from
10967 @file{allclass.cc} even if you never specified @samp{#pragma
10968 implementation}. This was deemed to be more trouble than it was worth,
10969 however, and disabled.
10971 Use the string argument if you want a single implementation file to
10972 include code from multiple header files. (You must also use
10973 @samp{#include} to include the header file; @samp{#pragma
10974 implementation} only specifies how to use the file---it doesn't actually
10977 There is no way to split up the contents of a single header file into
10978 multiple implementation files.
10981 @cindex inlining and C++ pragmas
10982 @cindex C++ pragmas, effect on inlining
10983 @cindex pragmas in C++, effect on inlining
10984 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10985 effect on function inlining.
10987 If you define a class in a header file marked with @samp{#pragma
10988 interface}, the effect on an inline function defined in that class is
10989 similar to an explicit @code{extern} declaration---the compiler emits
10990 no code at all to define an independent version of the function. Its
10991 definition is used only for inlining with its callers.
10993 @opindex fno-implement-inlines
10994 Conversely, when you include the same header file in a main source file
10995 that declares it as @samp{#pragma implementation}, the compiler emits
10996 code for the function itself; this defines a version of the function
10997 that can be found via pointers (or by callers compiled without
10998 inlining). If all calls to the function can be inlined, you can avoid
10999 emitting the function by compiling with @option{-fno-implement-inlines}.
11000 If any calls were not inlined, you will get linker errors.
11002 @node Template Instantiation
11003 @section Where's the Template?
11004 @cindex template instantiation
11006 C++ templates are the first language feature to require more
11007 intelligence from the environment than one usually finds on a UNIX
11008 system. Somehow the compiler and linker have to make sure that each
11009 template instance occurs exactly once in the executable if it is needed,
11010 and not at all otherwise. There are two basic approaches to this
11011 problem, which are referred to as the Borland model and the Cfront model.
11014 @item Borland model
11015 Borland C++ solved the template instantiation problem by adding the code
11016 equivalent of common blocks to their linker; the compiler emits template
11017 instances in each translation unit that uses them, and the linker
11018 collapses them together. The advantage of this model is that the linker
11019 only has to consider the object files themselves; there is no external
11020 complexity to worry about. This disadvantage is that compilation time
11021 is increased because the template code is being compiled repeatedly.
11022 Code written for this model tends to include definitions of all
11023 templates in the header file, since they must be seen to be
11027 The AT&T C++ translator, Cfront, solved the template instantiation
11028 problem by creating the notion of a template repository, an
11029 automatically maintained place where template instances are stored. A
11030 more modern version of the repository works as follows: As individual
11031 object files are built, the compiler places any template definitions and
11032 instantiations encountered in the repository. At link time, the link
11033 wrapper adds in the objects in the repository and compiles any needed
11034 instances that were not previously emitted. The advantages of this
11035 model are more optimal compilation speed and the ability to use the
11036 system linker; to implement the Borland model a compiler vendor also
11037 needs to replace the linker. The disadvantages are vastly increased
11038 complexity, and thus potential for error; for some code this can be
11039 just as transparent, but in practice it can been very difficult to build
11040 multiple programs in one directory and one program in multiple
11041 directories. Code written for this model tends to separate definitions
11042 of non-inline member templates into a separate file, which should be
11043 compiled separately.
11046 When used with GNU ld version 2.8 or later on an ELF system such as
11047 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
11048 Borland model. On other systems, G++ implements neither automatic
11051 A future version of G++ will support a hybrid model whereby the compiler
11052 will emit any instantiations for which the template definition is
11053 included in the compile, and store template definitions and
11054 instantiation context information into the object file for the rest.
11055 The link wrapper will extract that information as necessary and invoke
11056 the compiler to produce the remaining instantiations. The linker will
11057 then combine duplicate instantiations.
11059 In the mean time, you have the following options for dealing with
11060 template instantiations:
11065 Compile your template-using code with @option{-frepo}. The compiler will
11066 generate files with the extension @samp{.rpo} listing all of the
11067 template instantiations used in the corresponding object files which
11068 could be instantiated there; the link wrapper, @samp{collect2}, will
11069 then update the @samp{.rpo} files to tell the compiler where to place
11070 those instantiations and rebuild any affected object files. The
11071 link-time overhead is negligible after the first pass, as the compiler
11072 will continue to place the instantiations in the same files.
11074 This is your best option for application code written for the Borland
11075 model, as it will just work. Code written for the Cfront model will
11076 need to be modified so that the template definitions are available at
11077 one or more points of instantiation; usually this is as simple as adding
11078 @code{#include <tmethods.cc>} to the end of each template header.
11080 For library code, if you want the library to provide all of the template
11081 instantiations it needs, just try to link all of its object files
11082 together; the link will fail, but cause the instantiations to be
11083 generated as a side effect. Be warned, however, that this may cause
11084 conflicts if multiple libraries try to provide the same instantiations.
11085 For greater control, use explicit instantiation as described in the next
11089 @opindex fno-implicit-templates
11090 Compile your code with @option{-fno-implicit-templates} to disable the
11091 implicit generation of template instances, and explicitly instantiate
11092 all the ones you use. This approach requires more knowledge of exactly
11093 which instances you need than do the others, but it's less
11094 mysterious and allows greater control. You can scatter the explicit
11095 instantiations throughout your program, perhaps putting them in the
11096 translation units where the instances are used or the translation units
11097 that define the templates themselves; you can put all of the explicit
11098 instantiations you need into one big file; or you can create small files
11105 template class Foo<int>;
11106 template ostream& operator <<
11107 (ostream&, const Foo<int>&);
11110 for each of the instances you need, and create a template instantiation
11111 library from those.
11113 If you are using Cfront-model code, you can probably get away with not
11114 using @option{-fno-implicit-templates} when compiling files that don't
11115 @samp{#include} the member template definitions.
11117 If you use one big file to do the instantiations, you may want to
11118 compile it without @option{-fno-implicit-templates} so you get all of the
11119 instances required by your explicit instantiations (but not by any
11120 other files) without having to specify them as well.
11122 G++ has extended the template instantiation syntax given in the ISO
11123 standard to allow forward declaration of explicit instantiations
11124 (with @code{extern}), instantiation of the compiler support data for a
11125 template class (i.e.@: the vtable) without instantiating any of its
11126 members (with @code{inline}), and instantiation of only the static data
11127 members of a template class, without the support data or member
11128 functions (with (@code{static}):
11131 extern template int max (int, int);
11132 inline template class Foo<int>;
11133 static template class Foo<int>;
11137 Do nothing. Pretend G++ does implement automatic instantiation
11138 management. Code written for the Borland model will work fine, but
11139 each translation unit will contain instances of each of the templates it
11140 uses. In a large program, this can lead to an unacceptable amount of code
11144 @node Bound member functions
11145 @section Extracting the function pointer from a bound pointer to member function
11147 @cindex pointer to member function
11148 @cindex bound pointer to member function
11150 In C++, pointer to member functions (PMFs) are implemented using a wide
11151 pointer of sorts to handle all the possible call mechanisms; the PMF
11152 needs to store information about how to adjust the @samp{this} pointer,
11153 and if the function pointed to is virtual, where to find the vtable, and
11154 where in the vtable to look for the member function. If you are using
11155 PMFs in an inner loop, you should really reconsider that decision. If
11156 that is not an option, you can extract the pointer to the function that
11157 would be called for a given object/PMF pair and call it directly inside
11158 the inner loop, to save a bit of time.
11160 Note that you will still be paying the penalty for the call through a
11161 function pointer; on most modern architectures, such a call defeats the
11162 branch prediction features of the CPU@. This is also true of normal
11163 virtual function calls.
11165 The syntax for this extension is
11169 extern int (A::*fp)();
11170 typedef int (*fptr)(A *);
11172 fptr p = (fptr)(a.*fp);
11175 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
11176 no object is needed to obtain the address of the function. They can be
11177 converted to function pointers directly:
11180 fptr p1 = (fptr)(&A::foo);
11183 @opindex Wno-pmf-conversions
11184 You must specify @option{-Wno-pmf-conversions} to use this extension.
11186 @node C++ Attributes
11187 @section C++-Specific Variable, Function, and Type Attributes
11189 Some attributes only make sense for C++ programs.
11192 @item init_priority (@var{priority})
11193 @cindex init_priority attribute
11196 In Standard C++, objects defined at namespace scope are guaranteed to be
11197 initialized in an order in strict accordance with that of their definitions
11198 @emph{in a given translation unit}. No guarantee is made for initializations
11199 across translation units. However, GNU C++ allows users to control the
11200 order of initialization of objects defined at namespace scope with the
11201 @code{init_priority} attribute by specifying a relative @var{priority},
11202 a constant integral expression currently bounded between 101 and 65535
11203 inclusive. Lower numbers indicate a higher priority.
11205 In the following example, @code{A} would normally be created before
11206 @code{B}, but the @code{init_priority} attribute has reversed that order:
11209 Some_Class A __attribute__ ((init_priority (2000)));
11210 Some_Class B __attribute__ ((init_priority (543)));
11214 Note that the particular values of @var{priority} do not matter; only their
11217 @item java_interface
11218 @cindex java_interface attribute
11220 This type attribute informs C++ that the class is a Java interface. It may
11221 only be applied to classes declared within an @code{extern "Java"} block.
11222 Calls to methods declared in this interface will be dispatched using GCJ's
11223 interface table mechanism, instead of regular virtual table dispatch.
11227 See also @xref{Namespace Association}.
11229 @node Namespace Association
11230 @section Namespace Association
11232 @strong{Caution:} The semantics of this extension are not fully
11233 defined. Users should refrain from using this extension as its
11234 semantics may change subtly over time. It is possible that this
11235 extension will be removed in future versions of G++.
11237 A using-directive with @code{__attribute ((strong))} is stronger
11238 than a normal using-directive in two ways:
11242 Templates from the used namespace can be specialized and explicitly
11243 instantiated as though they were members of the using namespace.
11246 The using namespace is considered an associated namespace of all
11247 templates in the used namespace for purposes of argument-dependent
11251 The used namespace must be nested within the using namespace so that
11252 normal unqualified lookup works properly.
11254 This is useful for composing a namespace transparently from
11255 implementation namespaces. For example:
11260 template <class T> struct A @{ @};
11262 using namespace debug __attribute ((__strong__));
11263 template <> struct A<int> @{ @}; // @r{ok to specialize}
11265 template <class T> void f (A<T>);
11270 f (std::A<float>()); // @r{lookup finds} std::f
11276 @section Type Traits
11278 The C++ front-end implements syntactic extensions that allow to
11279 determine at compile time various characteristics of a type (or of a
11283 @item __has_nothrow_assign (type)
11284 If @code{type} is const qualified or is a reference type then the trait is
11285 false. Otherwise if @code{__has_trivial_assign (type)} is true then the trait
11286 is true, else if @code{type} is a cv class or union type with copy assignment
11287 operators that are known not to throw an exception then the trait is true,
11288 else it is false. Requires: @code{type} shall be a complete type, an array
11289 type of unknown bound, or is a @code{void} type.
11291 @item __has_nothrow_copy (type)
11292 If @code{__has_trivial_copy (type)} is true then the trait is true, else if
11293 @code{type} is a cv class or union type with copy constructors that
11294 are known not to throw an exception then the trait is true, else it is false.
11295 Requires: @code{type} shall be a complete type, an array type of
11296 unknown bound, or is a @code{void} type.
11298 @item __has_nothrow_constructor (type)
11299 If @code{__has_trivial_constructor (type)} is true then the trait is
11300 true, else if @code{type} is a cv class or union type (or array
11301 thereof) with a default constructor that is known not to throw an
11302 exception then the trait is true, else it is false. Requires:
11303 @code{type} shall be a complete type, an array type of unknown bound,
11304 or is a @code{void} type.
11306 @item __has_trivial_assign (type)
11307 If @code{type} is const qualified or is a reference type then the trait is
11308 false. Otherwise if @code{__is_pod (type)} is true then the trait is
11309 true, else if @code{type} is a cv class or union type with a trivial
11310 copy assignment ([class.copy]) then the trait is true, else it is
11311 false. Requires: @code{type} shall be a complete type, an array type
11312 of unknown bound, or is a @code{void} type.
11314 @item __has_trivial_copy (type)
11315 If @code{__is_pod (type)} is true or @code{type} is a reference type
11316 then the trait is true, else if @code{type} is a cv class or union type
11317 with a trivial copy constructor ([class.copy]) then the trait
11318 is true, else it is false. Requires: @code{type} shall be a complete
11319 type, an array type of unknown bound, or is a @code{void} type.
11321 @item __has_trivial_constructor (type)
11322 If @code{__is_pod (type)} is true then the trait is true, else if
11323 @code{type} is a cv class or union type (or array thereof) with a
11324 trivial default constructor ([class.ctor]) then the trait is true,
11325 else it is false. Requires: @code{type} shall be a complete type, an
11326 array type of unknown bound, or is a @code{void} type.
11328 @item __has_trivial_destructor (type)
11329 If @code{__is_pod (type)} is true or @code{type} is a reference type then
11330 the trait is true, else if @code{type} is a cv class or union type (or
11331 array thereof) with a trivial destructor ([class.dtor]) then the trait
11332 is true, else it is false. Requires: @code{type} shall be a complete
11333 type, an array type of unknown bound, or is a @code{void} type.
11335 @item __has_virtual_destructor (type)
11336 If @code{type} is a class type with a virtual destructor
11337 ([class.dtor]) then the trait is true, else it is false. Requires:
11338 @code{type} shall be a complete type, an array type of unknown bound,
11339 or is a @code{void} type.
11341 @item __is_abstract (type)
11342 If @code{type} is an abstract class ([class.abstract]) then the trait
11343 is true, else it is false. Requires: @code{type} shall be a complete
11344 type, an array type of unknown bound, or is a @code{void} type.
11346 @item __is_base_of (base_type, derived_type)
11347 If @code{base_type} is a base class of @code{derived_type}
11348 ([class.derived]) then the trait is true, otherwise it is false.
11349 Top-level cv qualifications of @code{base_type} and
11350 @code{derived_type} are ignored. For the purposes of this trait, a
11351 class type is considered is own base. Requires: if @code{__is_class
11352 (base_type)} and @code{__is_class (derived_type)} are true and
11353 @code{base_type} and @code{derived_type} are not the same type
11354 (disregarding cv-qualifiers), @code{derived_type} shall be a complete
11355 type. Diagnostic is produced if this requirement is not met.
11357 @item __is_class (type)
11358 If @code{type} is a cv class type, and not a union type
11359 ([basic.compound]) the the trait is true, else it is false.
11361 @item __is_empty (type)
11362 If @code{__is_class (type)} is false then the trait is false.
11363 Otherwise @code{type} is considered empty if and only if: @code{type}
11364 has no non-static data members, or all non-static data members, if
11365 any, are bit-fields of lenght 0, and @code{type} has no virtual
11366 members, and @code{type} has no virtual base classes, and @code{type}
11367 has no base classes @code{base_type} for which
11368 @code{__is_empty (base_type)} is false. Requires: @code{type} shall
11369 be a complete type, an array type of unknown bound, or is a
11372 @item __is_enum (type)
11373 If @code{type} is a cv enumeration type ([basic.compound]) the the trait is
11374 true, else it is false.
11376 @item __is_pod (type)
11377 If @code{type} is a cv POD type ([basic.types]) then the trait is true,
11378 else it is false. Requires: @code{type} shall be a complete type,
11379 an array type of unknown bound, or is a @code{void} type.
11381 @item __is_polymorphic (type)
11382 If @code{type} is a polymorphic class ([class.virtual]) then the trait
11383 is true, else it is false. Requires: @code{type} shall be a complete
11384 type, an array type of unknown bound, or is a @code{void} type.
11386 @item __is_union (type)
11387 If @code{type} is a cv union type ([basic.compound]) the the trait is
11388 true, else it is false.
11392 @node Java Exceptions
11393 @section Java Exceptions
11395 The Java language uses a slightly different exception handling model
11396 from C++. Normally, GNU C++ will automatically detect when you are
11397 writing C++ code that uses Java exceptions, and handle them
11398 appropriately. However, if C++ code only needs to execute destructors
11399 when Java exceptions are thrown through it, GCC will guess incorrectly.
11400 Sample problematic code is:
11403 struct S @{ ~S(); @};
11404 extern void bar(); // @r{is written in Java, and may throw exceptions}
11413 The usual effect of an incorrect guess is a link failure, complaining of
11414 a missing routine called @samp{__gxx_personality_v0}.
11416 You can inform the compiler that Java exceptions are to be used in a
11417 translation unit, irrespective of what it might think, by writing
11418 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11419 @samp{#pragma} must appear before any functions that throw or catch
11420 exceptions, or run destructors when exceptions are thrown through them.
11422 You cannot mix Java and C++ exceptions in the same translation unit. It
11423 is believed to be safe to throw a C++ exception from one file through
11424 another file compiled for the Java exception model, or vice versa, but
11425 there may be bugs in this area.
11427 @node Deprecated Features
11428 @section Deprecated Features
11430 In the past, the GNU C++ compiler was extended to experiment with new
11431 features, at a time when the C++ language was still evolving. Now that
11432 the C++ standard is complete, some of those features are superseded by
11433 superior alternatives. Using the old features might cause a warning in
11434 some cases that the feature will be dropped in the future. In other
11435 cases, the feature might be gone already.
11437 While the list below is not exhaustive, it documents some of the options
11438 that are now deprecated:
11441 @item -fexternal-templates
11442 @itemx -falt-external-templates
11443 These are two of the many ways for G++ to implement template
11444 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11445 defines how template definitions have to be organized across
11446 implementation units. G++ has an implicit instantiation mechanism that
11447 should work just fine for standard-conforming code.
11449 @item -fstrict-prototype
11450 @itemx -fno-strict-prototype
11451 Previously it was possible to use an empty prototype parameter list to
11452 indicate an unspecified number of parameters (like C), rather than no
11453 parameters, as C++ demands. This feature has been removed, except where
11454 it is required for backwards compatibility @xref{Backwards Compatibility}.
11457 G++ allows a virtual function returning @samp{void *} to be overridden
11458 by one returning a different pointer type. This extension to the
11459 covariant return type rules is now deprecated and will be removed from a
11462 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11463 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11464 and will be removed in a future version. Code using these operators
11465 should be modified to use @code{std::min} and @code{std::max} instead.
11467 The named return value extension has been deprecated, and is now
11470 The use of initializer lists with new expressions has been deprecated,
11471 and is now removed from G++.
11473 Floating and complex non-type template parameters have been deprecated,
11474 and are now removed from G++.
11476 The implicit typename extension has been deprecated and is now
11479 The use of default arguments in function pointers, function typedefs
11480 and other places where they are not permitted by the standard is
11481 deprecated and will be removed from a future version of G++.
11483 G++ allows floating-point literals to appear in integral constant expressions,
11484 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11485 This extension is deprecated and will be removed from a future version.
11487 G++ allows static data members of const floating-point type to be declared
11488 with an initializer in a class definition. The standard only allows
11489 initializers for static members of const integral types and const
11490 enumeration types so this extension has been deprecated and will be removed
11491 from a future version.
11493 @node Backwards Compatibility
11494 @section Backwards Compatibility
11495 @cindex Backwards Compatibility
11496 @cindex ARM [Annotated C++ Reference Manual]
11498 Now that there is a definitive ISO standard C++, G++ has a specification
11499 to adhere to. The C++ language evolved over time, and features that
11500 used to be acceptable in previous drafts of the standard, such as the ARM
11501 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11502 compilation of C++ written to such drafts, G++ contains some backwards
11503 compatibilities. @emph{All such backwards compatibility features are
11504 liable to disappear in future versions of G++.} They should be considered
11505 deprecated @xref{Deprecated Features}.
11509 If a variable is declared at for scope, it used to remain in scope until
11510 the end of the scope which contained the for statement (rather than just
11511 within the for scope). G++ retains this, but issues a warning, if such a
11512 variable is accessed outside the for scope.
11514 @item Implicit C language
11515 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11516 scope to set the language. On such systems, all header files are
11517 implicitly scoped inside a C language scope. Also, an empty prototype
11518 @code{()} will be treated as an unspecified number of arguments, rather
11519 than no arguments, as C++ demands.