1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C and Objective-C@. Most of them are
20 also available in C++. @xref{C++ Extensions,,Extensions to the
21 C++ Language}, for extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Decimal Float:: Decimal Floating Types.
37 * Hex Floats:: Hexadecimal floating-point constants.
38 * Zero Length:: Zero-length arrays.
39 * Variable Length:: Arrays whose length is computed at run time.
40 * Empty Structures:: Structures with no members.
41 * Variadic Macros:: Macros with a variable number of arguments.
42 * Escaped Newlines:: Slightly looser rules for escaped newlines.
43 * Subscripting:: Any array can be subscripted, even if not an lvalue.
44 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
45 * Initializers:: Non-constant initializers.
46 * Compound Literals:: Compound literals give structures, unions
48 * Designated Inits:: Labeling elements of initializers.
49 * Cast to Union:: Casting to union type from any member of the union.
50 * Case Ranges:: `case 1 ... 9' and such.
51 * Mixed Declarations:: Mixing declarations and code.
52 * Function Attributes:: Declaring that functions have no side effects,
53 or that they can never return.
54 * Attribute Syntax:: Formal syntax for attributes.
55 * Function Prototypes:: Prototype declarations and old-style definitions.
56 * C++ Comments:: C++ comments are recognized.
57 * Dollar Signs:: Dollar sign is allowed in identifiers.
58 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Alignment:: Inquiring about the alignment of a type or variable.
62 * Inline:: Defining inline functions (as fast as macros).
63 * Extended Asm:: Assembler instructions with C expressions as operands.
64 (With them you can define ``built-in'' functions.)
65 * Constraints:: Constraints for asm operands
66 * Asm Labels:: Specifying the assembler name to use for a C symbol.
67 * Explicit Reg Vars:: Defining variables residing in specified registers.
68 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
69 * Incomplete Enums:: @code{enum foo;}, with details to follow.
70 * Function Names:: Printable strings which are the name of the current
72 * Return Address:: Getting the return or frame address of a function.
73 * Vector Extensions:: Using vector instructions through built-in functions.
74 * Offsetof:: Special syntax for implementing @code{offsetof}.
75 * Atomic Builtins:: Built-in functions for atomic memory access.
76 * Object Size Checking:: Built-in functions for limited buffer overflow
78 * Other Builtins:: Other built-in functions.
79 * Target Builtins:: Built-in functions specific to particular targets.
80 * Target Format Checks:: Format checks specific to particular targets.
81 * Pragmas:: Pragmas accepted by GCC.
82 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
83 * Thread-Local:: Per-thread variables.
87 @section Statements and Declarations in Expressions
88 @cindex statements inside expressions
89 @cindex declarations inside expressions
90 @cindex expressions containing statements
91 @cindex macros, statements in expressions
93 @c the above section title wrapped and causes an underfull hbox.. i
94 @c changed it from "within" to "in". --mew 4feb93
95 A compound statement enclosed in parentheses may appear as an expression
96 in GNU C@. This allows you to use loops, switches, and local variables
99 Recall that a compound statement is a sequence of statements surrounded
100 by braces; in this construct, parentheses go around the braces. For
104 (@{ int y = foo (); int z;
111 is a valid (though slightly more complex than necessary) expression
112 for the absolute value of @code{foo ()}.
114 The last thing in the compound statement should be an expression
115 followed by a semicolon; the value of this subexpression serves as the
116 value of the entire construct. (If you use some other kind of statement
117 last within the braces, the construct has type @code{void}, and thus
118 effectively no value.)
120 This feature is especially useful in making macro definitions ``safe'' (so
121 that they evaluate each operand exactly once). For example, the
122 ``maximum'' function is commonly defined as a macro in standard C as
126 #define max(a,b) ((a) > (b) ? (a) : (b))
130 @cindex side effects, macro argument
131 But this definition computes either @var{a} or @var{b} twice, with bad
132 results if the operand has side effects. In GNU C, if you know the
133 type of the operands (here taken as @code{int}), you can define
134 the macro safely as follows:
137 #define maxint(a,b) \
138 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
141 Embedded statements are not allowed in constant expressions, such as
142 the value of an enumeration constant, the width of a bit-field, or
143 the initial value of a static variable.
145 If you don't know the type of the operand, you can still do this, but you
146 must use @code{typeof} (@pxref{Typeof}).
148 In G++, the result value of a statement expression undergoes array and
149 function pointer decay, and is returned by value to the enclosing
150 expression. For instance, if @code{A} is a class, then
159 will construct a temporary @code{A} object to hold the result of the
160 statement expression, and that will be used to invoke @code{Foo}.
161 Therefore the @code{this} pointer observed by @code{Foo} will not be the
164 Any temporaries created within a statement within a statement expression
165 will be destroyed at the statement's end. This makes statement
166 expressions inside macros slightly different from function calls. In
167 the latter case temporaries introduced during argument evaluation will
168 be destroyed at the end of the statement that includes the function
169 call. In the statement expression case they will be destroyed during
170 the statement expression. For instance,
173 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
174 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
184 will have different places where temporaries are destroyed. For the
185 @code{macro} case, the temporary @code{X} will be destroyed just after
186 the initialization of @code{b}. In the @code{function} case that
187 temporary will be destroyed when the function returns.
189 These considerations mean that it is probably a bad idea to use
190 statement-expressions of this form in header files that are designed to
191 work with C++. (Note that some versions of the GNU C Library contained
192 header files using statement-expression that lead to precisely this
195 Jumping into a statement expression with @code{goto} or using a
196 @code{switch} statement outside the statement expression with a
197 @code{case} or @code{default} label inside the statement expression is
198 not permitted. Jumping into a statement expression with a computed
199 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
200 Jumping out of a statement expression is permitted, but if the
201 statement expression is part of a larger expression then it is
202 unspecified which other subexpressions of that expression have been
203 evaluated except where the language definition requires certain
204 subexpressions to be evaluated before or after the statement
205 expression. In any case, as with a function call the evaluation of a
206 statement expression is not interleaved with the evaluation of other
207 parts of the containing expression. For example,
210 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
214 will call @code{foo} and @code{bar1} and will not call @code{baz} but
215 may or may not call @code{bar2}. If @code{bar2} is called, it will be
216 called after @code{foo} and before @code{bar1}
219 @section Locally Declared Labels
221 @cindex macros, local labels
223 GCC allows you to declare @dfn{local labels} in any nested block
224 scope. A local label is just like an ordinary label, but you can
225 only reference it (with a @code{goto} statement, or by taking its
226 address) within the block in which it was declared.
228 A local label declaration looks like this:
231 __label__ @var{label};
238 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
241 Local label declarations must come at the beginning of the block,
242 before any ordinary declarations or statements.
244 The label declaration defines the label @emph{name}, but does not define
245 the label itself. You must do this in the usual way, with
246 @code{@var{label}:}, within the statements of the statement expression.
248 The local label feature is useful for complex macros. If a macro
249 contains nested loops, a @code{goto} can be useful for breaking out of
250 them. However, an ordinary label whose scope is the whole function
251 cannot be used: if the macro can be expanded several times in one
252 function, the label will be multiply defined in that function. A
253 local label avoids this problem. For example:
256 #define SEARCH(value, array, target) \
259 typeof (target) _SEARCH_target = (target); \
260 typeof (*(array)) *_SEARCH_array = (array); \
263 for (i = 0; i < max; i++) \
264 for (j = 0; j < max; j++) \
265 if (_SEARCH_array[i][j] == _SEARCH_target) \
266 @{ (value) = i; goto found; @} \
272 This could also be written using a statement-expression:
275 #define SEARCH(array, target) \
278 typeof (target) _SEARCH_target = (target); \
279 typeof (*(array)) *_SEARCH_array = (array); \
282 for (i = 0; i < max; i++) \
283 for (j = 0; j < max; j++) \
284 if (_SEARCH_array[i][j] == _SEARCH_target) \
285 @{ value = i; goto found; @} \
292 Local label declarations also make the labels they declare visible to
293 nested functions, if there are any. @xref{Nested Functions}, for details.
295 @node Labels as Values
296 @section Labels as Values
297 @cindex labels as values
298 @cindex computed gotos
299 @cindex goto with computed label
300 @cindex address of a label
302 You can get the address of a label defined in the current function
303 (or a containing function) with the unary operator @samp{&&}. The
304 value has type @code{void *}. This value is a constant and can be used
305 wherever a constant of that type is valid. For example:
313 To use these values, you need to be able to jump to one. This is done
314 with the computed goto statement@footnote{The analogous feature in
315 Fortran is called an assigned goto, but that name seems inappropriate in
316 C, where one can do more than simply store label addresses in label
317 variables.}, @code{goto *@var{exp};}. For example,
324 Any expression of type @code{void *} is allowed.
326 One way of using these constants is in initializing a static array that
327 will serve as a jump table:
330 static void *array[] = @{ &&foo, &&bar, &&hack @};
333 Then you can select a label with indexing, like this:
340 Note that this does not check whether the subscript is in bounds---array
341 indexing in C never does that.
343 Such an array of label values serves a purpose much like that of the
344 @code{switch} statement. The @code{switch} statement is cleaner, so
345 use that rather than an array unless the problem does not fit a
346 @code{switch} statement very well.
348 Another use of label values is in an interpreter for threaded code.
349 The labels within the interpreter function can be stored in the
350 threaded code for super-fast dispatching.
352 You may not use this mechanism to jump to code in a different function.
353 If you do that, totally unpredictable things will happen. The best way to
354 avoid this is to store the label address only in automatic variables and
355 never pass it as an argument.
357 An alternate way to write the above example is
360 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
362 goto *(&&foo + array[i]);
366 This is more friendly to code living in shared libraries, as it reduces
367 the number of dynamic relocations that are needed, and by consequence,
368 allows the data to be read-only.
370 @node Nested Functions
371 @section Nested Functions
372 @cindex nested functions
373 @cindex downward funargs
376 A @dfn{nested function} is a function defined inside another function.
377 (Nested functions are not supported for GNU C++.) The nested function's
378 name is local to the block where it is defined. For example, here we
379 define a nested function named @code{square}, and call it twice:
383 foo (double a, double b)
385 double square (double z) @{ return z * z; @}
387 return square (a) + square (b);
392 The nested function can access all the variables of the containing
393 function that are visible at the point of its definition. This is
394 called @dfn{lexical scoping}. For example, here we show a nested
395 function which uses an inherited variable named @code{offset}:
399 bar (int *array, int offset, int size)
401 int access (int *array, int index)
402 @{ return array[index + offset]; @}
405 for (i = 0; i < size; i++)
406 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
411 Nested function definitions are permitted within functions in the places
412 where variable definitions are allowed; that is, in any block, mixed
413 with the other declarations and statements in the block.
415 It is possible to call the nested function from outside the scope of its
416 name by storing its address or passing the address to another function:
419 hack (int *array, int size)
421 void store (int index, int value)
422 @{ array[index] = value; @}
424 intermediate (store, size);
428 Here, the function @code{intermediate} receives the address of
429 @code{store} as an argument. If @code{intermediate} calls @code{store},
430 the arguments given to @code{store} are used to store into @code{array}.
431 But this technique works only so long as the containing function
432 (@code{hack}, in this example) does not exit.
434 If you try to call the nested function through its address after the
435 containing function has exited, all hell will break loose. If you try
436 to call it after a containing scope level has exited, and if it refers
437 to some of the variables that are no longer in scope, you may be lucky,
438 but it's not wise to take the risk. If, however, the nested function
439 does not refer to anything that has gone out of scope, you should be
442 GCC implements taking the address of a nested function using a technique
443 called @dfn{trampolines}. A paper describing them is available as
446 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
448 A nested function can jump to a label inherited from a containing
449 function, provided the label was explicitly declared in the containing
450 function (@pxref{Local Labels}). Such a jump returns instantly to the
451 containing function, exiting the nested function which did the
452 @code{goto} and any intermediate functions as well. Here is an example:
456 bar (int *array, int offset, int size)
459 int access (int *array, int index)
463 return array[index + offset];
467 for (i = 0; i < size; i++)
468 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
472 /* @r{Control comes here from @code{access}
473 if it detects an error.} */
480 A nested function always has no linkage. Declaring one with
481 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
482 before its definition, use @code{auto} (which is otherwise meaningless
483 for function declarations).
486 bar (int *array, int offset, int size)
489 auto int access (int *, int);
491 int access (int *array, int index)
495 return array[index + offset];
501 @node Constructing Calls
502 @section Constructing Function Calls
503 @cindex constructing calls
504 @cindex forwarding calls
506 Using the built-in functions described below, you can record
507 the arguments a function received, and call another function
508 with the same arguments, without knowing the number or types
511 You can also record the return value of that function call,
512 and later return that value, without knowing what data type
513 the function tried to return (as long as your caller expects
516 However, these built-in functions may interact badly with some
517 sophisticated features or other extensions of the language. It
518 is, therefore, not recommended to use them outside very simple
519 functions acting as mere forwarders for their arguments.
521 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
522 This built-in function returns a pointer to data
523 describing how to perform a call with the same arguments as were passed
524 to the current function.
526 The function saves the arg pointer register, structure value address,
527 and all registers that might be used to pass arguments to a function
528 into a block of memory allocated on the stack. Then it returns the
529 address of that block.
532 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
533 This built-in function invokes @var{function}
534 with a copy of the parameters described by @var{arguments}
537 The value of @var{arguments} should be the value returned by
538 @code{__builtin_apply_args}. The argument @var{size} specifies the size
539 of the stack argument data, in bytes.
541 This function returns a pointer to data describing
542 how to return whatever value was returned by @var{function}. The data
543 is saved in a block of memory allocated on the stack.
545 It is not always simple to compute the proper value for @var{size}. The
546 value is used by @code{__builtin_apply} to compute the amount of data
547 that should be pushed on the stack and copied from the incoming argument
551 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
552 This built-in function returns the value described by @var{result} from
553 the containing function. You should specify, for @var{result}, a value
554 returned by @code{__builtin_apply}.
558 @section Referring to a Type with @code{typeof}
561 @cindex macros, types of arguments
563 Another way to refer to the type of an expression is with @code{typeof}.
564 The syntax of using of this keyword looks like @code{sizeof}, but the
565 construct acts semantically like a type name defined with @code{typedef}.
567 There are two ways of writing the argument to @code{typeof}: with an
568 expression or with a type. Here is an example with an expression:
575 This assumes that @code{x} is an array of pointers to functions;
576 the type described is that of the values of the functions.
578 Here is an example with a typename as the argument:
585 Here the type described is that of pointers to @code{int}.
587 If you are writing a header file that must work when included in ISO C
588 programs, write @code{__typeof__} instead of @code{typeof}.
589 @xref{Alternate Keywords}.
591 A @code{typeof}-construct can be used anywhere a typedef name could be
592 used. For example, you can use it in a declaration, in a cast, or inside
593 of @code{sizeof} or @code{typeof}.
595 @code{typeof} is often useful in conjunction with the
596 statements-within-expressions feature. Here is how the two together can
597 be used to define a safe ``maximum'' macro that operates on any
598 arithmetic type and evaluates each of its arguments exactly once:
602 (@{ typeof (a) _a = (a); \
603 typeof (b) _b = (b); \
604 _a > _b ? _a : _b; @})
607 @cindex underscores in variables in macros
608 @cindex @samp{_} in variables in macros
609 @cindex local variables in macros
610 @cindex variables, local, in macros
611 @cindex macros, local variables in
613 The reason for using names that start with underscores for the local
614 variables is to avoid conflicts with variable names that occur within the
615 expressions that are substituted for @code{a} and @code{b}. Eventually we
616 hope to design a new form of declaration syntax that allows you to declare
617 variables whose scopes start only after their initializers; this will be a
618 more reliable way to prevent such conflicts.
621 Some more examples of the use of @code{typeof}:
625 This declares @code{y} with the type of what @code{x} points to.
632 This declares @code{y} as an array of such values.
639 This declares @code{y} as an array of pointers to characters:
642 typeof (typeof (char *)[4]) y;
646 It is equivalent to the following traditional C declaration:
652 To see the meaning of the declaration using @code{typeof}, and why it
653 might be a useful way to write, rewrite it with these macros:
656 #define pointer(T) typeof(T *)
657 #define array(T, N) typeof(T [N])
661 Now the declaration can be rewritten this way:
664 array (pointer (char), 4) y;
668 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
669 pointers to @code{char}.
672 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
673 a more limited extension which permitted one to write
676 typedef @var{T} = @var{expr};
680 with the effect of declaring @var{T} to have the type of the expression
681 @var{expr}. This extension does not work with GCC 3 (versions between
682 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
683 relies on it should be rewritten to use @code{typeof}:
686 typedef typeof(@var{expr}) @var{T};
690 This will work with all versions of GCC@.
693 @section Conditionals with Omitted Operands
694 @cindex conditional expressions, extensions
695 @cindex omitted middle-operands
696 @cindex middle-operands, omitted
697 @cindex extensions, @code{?:}
698 @cindex @code{?:} extensions
700 The middle operand in a conditional expression may be omitted. Then
701 if the first operand is nonzero, its value is the value of the conditional
704 Therefore, the expression
711 has the value of @code{x} if that is nonzero; otherwise, the value of
714 This example is perfectly equivalent to
720 @cindex side effect in ?:
721 @cindex ?: side effect
723 In this simple case, the ability to omit the middle operand is not
724 especially useful. When it becomes useful is when the first operand does,
725 or may (if it is a macro argument), contain a side effect. Then repeating
726 the operand in the middle would perform the side effect twice. Omitting
727 the middle operand uses the value already computed without the undesirable
728 effects of recomputing it.
731 @section Double-Word Integers
732 @cindex @code{long long} data types
733 @cindex double-word arithmetic
734 @cindex multiprecision arithmetic
735 @cindex @code{LL} integer suffix
736 @cindex @code{ULL} integer suffix
738 ISO C99 supports data types for integers that are at least 64 bits wide,
739 and as an extension GCC supports them in C89 mode and in C++.
740 Simply write @code{long long int} for a signed integer, or
741 @code{unsigned long long int} for an unsigned integer. To make an
742 integer constant of type @code{long long int}, add the suffix @samp{LL}
743 to the integer. To make an integer constant of type @code{unsigned long
744 long int}, add the suffix @samp{ULL} to the integer.
746 You can use these types in arithmetic like any other integer types.
747 Addition, subtraction, and bitwise boolean operations on these types
748 are open-coded on all types of machines. Multiplication is open-coded
749 if the machine supports fullword-to-doubleword a widening multiply
750 instruction. Division and shifts are open-coded only on machines that
751 provide special support. The operations that are not open-coded use
752 special library routines that come with GCC@.
754 There may be pitfalls when you use @code{long long} types for function
755 arguments, unless you declare function prototypes. If a function
756 expects type @code{int} for its argument, and you pass a value of type
757 @code{long long int}, confusion will result because the caller and the
758 subroutine will disagree about the number of bytes for the argument.
759 Likewise, if the function expects @code{long long int} and you pass
760 @code{int}. The best way to avoid such problems is to use prototypes.
763 @section Complex Numbers
764 @cindex complex numbers
765 @cindex @code{_Complex} keyword
766 @cindex @code{__complex__} keyword
768 ISO C99 supports complex floating data types, and as an extension GCC
769 supports them in C89 mode and in C++, and supports complex integer data
770 types which are not part of ISO C99. You can declare complex types
771 using the keyword @code{_Complex}. As an extension, the older GNU
772 keyword @code{__complex__} is also supported.
774 For example, @samp{_Complex double x;} declares @code{x} as a
775 variable whose real part and imaginary part are both of type
776 @code{double}. @samp{_Complex short int y;} declares @code{y} to
777 have real and imaginary parts of type @code{short int}; this is not
778 likely to be useful, but it shows that the set of complex types is
781 To write a constant with a complex data type, use the suffix @samp{i} or
782 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
783 has type @code{_Complex float} and @code{3i} has type
784 @code{_Complex int}. Such a constant always has a pure imaginary
785 value, but you can form any complex value you like by adding one to a
786 real constant. This is a GNU extension; if you have an ISO C99
787 conforming C library (such as GNU libc), and want to construct complex
788 constants of floating type, you should include @code{<complex.h>} and
789 use the macros @code{I} or @code{_Complex_I} instead.
791 @cindex @code{__real__} keyword
792 @cindex @code{__imag__} keyword
793 To extract the real part of a complex-valued expression @var{exp}, write
794 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
795 extract the imaginary part. This is a GNU extension; for values of
796 floating type, you should use the ISO C99 functions @code{crealf},
797 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
798 @code{cimagl}, declared in @code{<complex.h>} and also provided as
799 built-in functions by GCC@.
801 @cindex complex conjugation
802 The operator @samp{~} performs complex conjugation when used on a value
803 with a complex type. This is a GNU extension; for values of
804 floating type, you should use the ISO C99 functions @code{conjf},
805 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
806 provided as built-in functions by GCC@.
808 GCC can allocate complex automatic variables in a noncontiguous
809 fashion; it's even possible for the real part to be in a register while
810 the imaginary part is on the stack (or vice-versa). Only the DWARF2
811 debug info format can represent this, so use of DWARF2 is recommended.
812 If you are using the stabs debug info format, GCC describes a noncontiguous
813 complex variable as if it were two separate variables of noncomplex type.
814 If the variable's actual name is @code{foo}, the two fictitious
815 variables are named @code{foo$real} and @code{foo$imag}. You can
816 examine and set these two fictitious variables with your debugger.
819 @section Decimal Floating Types
820 @cindex decimal floating types
821 @cindex @code{_Decimal32} data type
822 @cindex @code{_Decimal64} data type
823 @cindex @code{_Decimal128} data type
824 @cindex @code{df} integer suffix
825 @cindex @code{dd} integer suffix
826 @cindex @code{dl} integer suffix
827 @cindex @code{DF} integer suffix
828 @cindex @code{DD} integer suffix
829 @cindex @code{DL} integer suffix
831 As an extension, the GNU C compiler supports decimal floating types as
832 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
833 floating types in GCC will evolve as the draft technical report changes.
834 Calling conventions for any target might also change. Not all targets
835 support decimal floating types.
837 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
838 @code{_Decimal128}. They use a radix of ten, unlike the floating types
839 @code{float}, @code{double}, and @code{long double} whose radix is not
840 specified by the C standard but is usually two.
842 Support for decimal floating types includes the arithmetic operators
843 add, subtract, multiply, divide; unary arithmetic operators;
844 relational operators; equality operators; and conversions to and from
845 integer and other floating types. Use a suffix @samp{df} or
846 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
847 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
850 GCC support of decimal float as specified by the draft technical report
855 Translation time data type (TTDT) is not supported.
858 Characteristics of decimal floating types are defined in header file
859 @file{decfloat.h} rather than @file{float.h}.
862 When the value of a decimal floating type cannot be represented in the
863 integer type to which it is being converted, the result is undefined
864 rather than the result value specified by the draft technical report.
867 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
868 are supported by the DWARF2 debug information format.
874 ISO C99 supports floating-point numbers written not only in the usual
875 decimal notation, such as @code{1.55e1}, but also numbers such as
876 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
877 supports this in C89 mode (except in some cases when strictly
878 conforming) and in C++. In that format the
879 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
880 mandatory. The exponent is a decimal number that indicates the power of
881 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
888 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
889 is the same as @code{1.55e1}.
891 Unlike for floating-point numbers in the decimal notation the exponent
892 is always required in the hexadecimal notation. Otherwise the compiler
893 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
894 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
895 extension for floating-point constants of type @code{float}.
898 @section Arrays of Length Zero
899 @cindex arrays of length zero
900 @cindex zero-length arrays
901 @cindex length-zero arrays
902 @cindex flexible array members
904 Zero-length arrays are allowed in GNU C@. They are very useful as the
905 last element of a structure which is really a header for a variable-length
914 struct line *thisline = (struct line *)
915 malloc (sizeof (struct line) + this_length);
916 thisline->length = this_length;
919 In ISO C90, you would have to give @code{contents} a length of 1, which
920 means either you waste space or complicate the argument to @code{malloc}.
922 In ISO C99, you would use a @dfn{flexible array member}, which is
923 slightly different in syntax and semantics:
927 Flexible array members are written as @code{contents[]} without
931 Flexible array members have incomplete type, and so the @code{sizeof}
932 operator may not be applied. As a quirk of the original implementation
933 of zero-length arrays, @code{sizeof} evaluates to zero.
936 Flexible array members may only appear as the last member of a
937 @code{struct} that is otherwise non-empty.
940 A structure containing a flexible array member, or a union containing
941 such a structure (possibly recursively), may not be a member of a
942 structure or an element of an array. (However, these uses are
943 permitted by GCC as extensions.)
946 GCC versions before 3.0 allowed zero-length arrays to be statically
947 initialized, as if they were flexible arrays. In addition to those
948 cases that were useful, it also allowed initializations in situations
949 that would corrupt later data. Non-empty initialization of zero-length
950 arrays is now treated like any case where there are more initializer
951 elements than the array holds, in that a suitable warning about "excess
952 elements in array" is given, and the excess elements (all of them, in
953 this case) are ignored.
955 Instead GCC allows static initialization of flexible array members.
956 This is equivalent to defining a new structure containing the original
957 structure followed by an array of sufficient size to contain the data.
958 I.e.@: in the following, @code{f1} is constructed as if it were declared
964 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
967 struct f1 f1; int data[3];
968 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
972 The convenience of this extension is that @code{f1} has the desired
973 type, eliminating the need to consistently refer to @code{f2.f1}.
975 This has symmetry with normal static arrays, in that an array of
976 unknown size is also written with @code{[]}.
978 Of course, this extension only makes sense if the extra data comes at
979 the end of a top-level object, as otherwise we would be overwriting
980 data at subsequent offsets. To avoid undue complication and confusion
981 with initialization of deeply nested arrays, we simply disallow any
982 non-empty initialization except when the structure is the top-level
986 struct foo @{ int x; int y[]; @};
987 struct bar @{ struct foo z; @};
989 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
990 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
991 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
992 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
995 @node Empty Structures
996 @section Structures With No Members
997 @cindex empty structures
998 @cindex zero-size structures
1000 GCC permits a C structure to have no members:
1007 The structure will have size zero. In C++, empty structures are part
1008 of the language. G++ treats empty structures as if they had a single
1009 member of type @code{char}.
1011 @node Variable Length
1012 @section Arrays of Variable Length
1013 @cindex variable-length arrays
1014 @cindex arrays of variable length
1017 Variable-length automatic arrays are allowed in ISO C99, and as an
1018 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1019 implementation of variable-length arrays does not yet conform in detail
1020 to the ISO C99 standard.) These arrays are
1021 declared like any other automatic arrays, but with a length that is not
1022 a constant expression. The storage is allocated at the point of
1023 declaration and deallocated when the brace-level is exited. For
1028 concat_fopen (char *s1, char *s2, char *mode)
1030 char str[strlen (s1) + strlen (s2) + 1];
1033 return fopen (str, mode);
1037 @cindex scope of a variable length array
1038 @cindex variable-length array scope
1039 @cindex deallocating variable length arrays
1040 Jumping or breaking out of the scope of the array name deallocates the
1041 storage. Jumping into the scope is not allowed; you get an error
1044 @cindex @code{alloca} vs variable-length arrays
1045 You can use the function @code{alloca} to get an effect much like
1046 variable-length arrays. The function @code{alloca} is available in
1047 many other C implementations (but not in all). On the other hand,
1048 variable-length arrays are more elegant.
1050 There are other differences between these two methods. Space allocated
1051 with @code{alloca} exists until the containing @emph{function} returns.
1052 The space for a variable-length array is deallocated as soon as the array
1053 name's scope ends. (If you use both variable-length arrays and
1054 @code{alloca} in the same function, deallocation of a variable-length array
1055 will also deallocate anything more recently allocated with @code{alloca}.)
1057 You can also use variable-length arrays as arguments to functions:
1061 tester (int len, char data[len][len])
1067 The length of an array is computed once when the storage is allocated
1068 and is remembered for the scope of the array in case you access it with
1071 If you want to pass the array first and the length afterward, you can
1072 use a forward declaration in the parameter list---another GNU extension.
1076 tester (int len; char data[len][len], int len)
1082 @cindex parameter forward declaration
1083 The @samp{int len} before the semicolon is a @dfn{parameter forward
1084 declaration}, and it serves the purpose of making the name @code{len}
1085 known when the declaration of @code{data} is parsed.
1087 You can write any number of such parameter forward declarations in the
1088 parameter list. They can be separated by commas or semicolons, but the
1089 last one must end with a semicolon, which is followed by the ``real''
1090 parameter declarations. Each forward declaration must match a ``real''
1091 declaration in parameter name and data type. ISO C99 does not support
1092 parameter forward declarations.
1094 @node Variadic Macros
1095 @section Macros with a Variable Number of Arguments.
1096 @cindex variable number of arguments
1097 @cindex macro with variable arguments
1098 @cindex rest argument (in macro)
1099 @cindex variadic macros
1101 In the ISO C standard of 1999, a macro can be declared to accept a
1102 variable number of arguments much as a function can. The syntax for
1103 defining the macro is similar to that of a function. Here is an
1107 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1110 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1111 such a macro, it represents the zero or more tokens until the closing
1112 parenthesis that ends the invocation, including any commas. This set of
1113 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1114 wherever it appears. See the CPP manual for more information.
1116 GCC has long supported variadic macros, and used a different syntax that
1117 allowed you to give a name to the variable arguments just like any other
1118 argument. Here is an example:
1121 #define debug(format, args...) fprintf (stderr, format, args)
1124 This is in all ways equivalent to the ISO C example above, but arguably
1125 more readable and descriptive.
1127 GNU CPP has two further variadic macro extensions, and permits them to
1128 be used with either of the above forms of macro definition.
1130 In standard C, you are not allowed to leave the variable argument out
1131 entirely; but you are allowed to pass an empty argument. For example,
1132 this invocation is invalid in ISO C, because there is no comma after
1139 GNU CPP permits you to completely omit the variable arguments in this
1140 way. In the above examples, the compiler would complain, though since
1141 the expansion of the macro still has the extra comma after the format
1144 To help solve this problem, CPP behaves specially for variable arguments
1145 used with the token paste operator, @samp{##}. If instead you write
1148 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1151 and if the variable arguments are omitted or empty, the @samp{##}
1152 operator causes the preprocessor to remove the comma before it. If you
1153 do provide some variable arguments in your macro invocation, GNU CPP
1154 does not complain about the paste operation and instead places the
1155 variable arguments after the comma. Just like any other pasted macro
1156 argument, these arguments are not macro expanded.
1158 @node Escaped Newlines
1159 @section Slightly Looser Rules for Escaped Newlines
1160 @cindex escaped newlines
1161 @cindex newlines (escaped)
1163 Recently, the preprocessor has relaxed its treatment of escaped
1164 newlines. Previously, the newline had to immediately follow a
1165 backslash. The current implementation allows whitespace in the form
1166 of spaces, horizontal and vertical tabs, and form feeds between the
1167 backslash and the subsequent newline. The preprocessor issues a
1168 warning, but treats it as a valid escaped newline and combines the two
1169 lines to form a single logical line. This works within comments and
1170 tokens, as well as between tokens. Comments are @emph{not} treated as
1171 whitespace for the purposes of this relaxation, since they have not
1172 yet been replaced with spaces.
1175 @section Non-Lvalue Arrays May Have Subscripts
1176 @cindex subscripting
1177 @cindex arrays, non-lvalue
1179 @cindex subscripting and function values
1180 In ISO C99, arrays that are not lvalues still decay to pointers, and
1181 may be subscripted, although they may not be modified or used after
1182 the next sequence point and the unary @samp{&} operator may not be
1183 applied to them. As an extension, GCC allows such arrays to be
1184 subscripted in C89 mode, though otherwise they do not decay to
1185 pointers outside C99 mode. For example,
1186 this is valid in GNU C though not valid in C89:
1190 struct foo @{int a[4];@};
1196 return f().a[index];
1202 @section Arithmetic on @code{void}- and Function-Pointers
1203 @cindex void pointers, arithmetic
1204 @cindex void, size of pointer to
1205 @cindex function pointers, arithmetic
1206 @cindex function, size of pointer to
1208 In GNU C, addition and subtraction operations are supported on pointers to
1209 @code{void} and on pointers to functions. This is done by treating the
1210 size of a @code{void} or of a function as 1.
1212 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1213 and on function types, and returns 1.
1215 @opindex Wpointer-arith
1216 The option @option{-Wpointer-arith} requests a warning if these extensions
1220 @section Non-Constant Initializers
1221 @cindex initializers, non-constant
1222 @cindex non-constant initializers
1224 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1225 automatic variable are not required to be constant expressions in GNU C@.
1226 Here is an example of an initializer with run-time varying elements:
1229 foo (float f, float g)
1231 float beat_freqs[2] = @{ f-g, f+g @};
1236 @node Compound Literals
1237 @section Compound Literals
1238 @cindex constructor expressions
1239 @cindex initializations in expressions
1240 @cindex structures, constructor expression
1241 @cindex expressions, constructor
1242 @cindex compound literals
1243 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1245 ISO C99 supports compound literals. A compound literal looks like
1246 a cast containing an initializer. Its value is an object of the
1247 type specified in the cast, containing the elements specified in
1248 the initializer; it is an lvalue. As an extension, GCC supports
1249 compound literals in C89 mode and in C++.
1251 Usually, the specified type is a structure. Assume that
1252 @code{struct foo} and @code{structure} are declared as shown:
1255 struct foo @{int a; char b[2];@} structure;
1259 Here is an example of constructing a @code{struct foo} with a compound literal:
1262 structure = ((struct foo) @{x + y, 'a', 0@});
1266 This is equivalent to writing the following:
1270 struct foo temp = @{x + y, 'a', 0@};
1275 You can also construct an array. If all the elements of the compound literal
1276 are (made up of) simple constant expressions, suitable for use in
1277 initializers of objects of static storage duration, then the compound
1278 literal can be coerced to a pointer to its first element and used in
1279 such an initializer, as shown here:
1282 char **foo = (char *[]) @{ "x", "y", "z" @};
1285 Compound literals for scalar types and union types are is
1286 also allowed, but then the compound literal is equivalent
1289 As a GNU extension, GCC allows initialization of objects with static storage
1290 duration by compound literals (which is not possible in ISO C99, because
1291 the initializer is not a constant).
1292 It is handled as if the object was initialized only with the bracket
1293 enclosed list if compound literal's and object types match.
1294 The initializer list of the compound literal must be constant.
1295 If the object being initialized has array type of unknown size, the size is
1296 determined by compound literal size.
1299 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1300 static int y[] = (int []) @{1, 2, 3@};
1301 static int z[] = (int [3]) @{1@};
1305 The above lines are equivalent to the following:
1307 static struct foo x = @{1, 'a', 'b'@};
1308 static int y[] = @{1, 2, 3@};
1309 static int z[] = @{1, 0, 0@};
1312 @node Designated Inits
1313 @section Designated Initializers
1314 @cindex initializers with labeled elements
1315 @cindex labeled elements in initializers
1316 @cindex case labels in initializers
1317 @cindex designated initializers
1319 Standard C89 requires the elements of an initializer to appear in a fixed
1320 order, the same as the order of the elements in the array or structure
1323 In ISO C99 you can give the elements in any order, specifying the array
1324 indices or structure field names they apply to, and GNU C allows this as
1325 an extension in C89 mode as well. This extension is not
1326 implemented in GNU C++.
1328 To specify an array index, write
1329 @samp{[@var{index}] =} before the element value. For example,
1332 int a[6] = @{ [4] = 29, [2] = 15 @};
1339 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1343 The index values must be constant expressions, even if the array being
1344 initialized is automatic.
1346 An alternative syntax for this which has been obsolete since GCC 2.5 but
1347 GCC still accepts is to write @samp{[@var{index}]} before the element
1348 value, with no @samp{=}.
1350 To initialize a range of elements to the same value, write
1351 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1352 extension. For example,
1355 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1359 If the value in it has side-effects, the side-effects will happen only once,
1360 not for each initialized field by the range initializer.
1363 Note that the length of the array is the highest value specified
1366 In a structure initializer, specify the name of a field to initialize
1367 with @samp{.@var{fieldname} =} before the element value. For example,
1368 given the following structure,
1371 struct point @{ int x, y; @};
1375 the following initialization
1378 struct point p = @{ .y = yvalue, .x = xvalue @};
1385 struct point p = @{ xvalue, yvalue @};
1388 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1389 @samp{@var{fieldname}:}, as shown here:
1392 struct point p = @{ y: yvalue, x: xvalue @};
1396 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1397 @dfn{designator}. You can also use a designator (or the obsolete colon
1398 syntax) when initializing a union, to specify which element of the union
1399 should be used. For example,
1402 union foo @{ int i; double d; @};
1404 union foo f = @{ .d = 4 @};
1408 will convert 4 to a @code{double} to store it in the union using
1409 the second element. By contrast, casting 4 to type @code{union foo}
1410 would store it into the union as the integer @code{i}, since it is
1411 an integer. (@xref{Cast to Union}.)
1413 You can combine this technique of naming elements with ordinary C
1414 initialization of successive elements. Each initializer element that
1415 does not have a designator applies to the next consecutive element of the
1416 array or structure. For example,
1419 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1426 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1429 Labeling the elements of an array initializer is especially useful
1430 when the indices are characters or belong to an @code{enum} type.
1435 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1436 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1439 @cindex designator lists
1440 You can also write a series of @samp{.@var{fieldname}} and
1441 @samp{[@var{index}]} designators before an @samp{=} to specify a
1442 nested subobject to initialize; the list is taken relative to the
1443 subobject corresponding to the closest surrounding brace pair. For
1444 example, with the @samp{struct point} declaration above:
1447 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1451 If the same field is initialized multiple times, it will have value from
1452 the last initialization. If any such overridden initialization has
1453 side-effect, it is unspecified whether the side-effect happens or not.
1454 Currently, GCC will discard them and issue a warning.
1457 @section Case Ranges
1459 @cindex ranges in case statements
1461 You can specify a range of consecutive values in a single @code{case} label,
1465 case @var{low} ... @var{high}:
1469 This has the same effect as the proper number of individual @code{case}
1470 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1472 This feature is especially useful for ranges of ASCII character codes:
1478 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1479 it may be parsed wrong when you use it with integer values. For example,
1494 @section Cast to a Union Type
1495 @cindex cast to a union
1496 @cindex union, casting to a
1498 A cast to union type is similar to other casts, except that the type
1499 specified is a union type. You can specify the type either with
1500 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1501 a constructor though, not a cast, and hence does not yield an lvalue like
1502 normal casts. (@xref{Compound Literals}.)
1504 The types that may be cast to the union type are those of the members
1505 of the union. Thus, given the following union and variables:
1508 union foo @{ int i; double d; @};
1514 both @code{x} and @code{y} can be cast to type @code{union foo}.
1516 Using the cast as the right-hand side of an assignment to a variable of
1517 union type is equivalent to storing in a member of the union:
1522 u = (union foo) x @equiv{} u.i = x
1523 u = (union foo) y @equiv{} u.d = y
1526 You can also use the union cast as a function argument:
1529 void hack (union foo);
1531 hack ((union foo) x);
1534 @node Mixed Declarations
1535 @section Mixed Declarations and Code
1536 @cindex mixed declarations and code
1537 @cindex declarations, mixed with code
1538 @cindex code, mixed with declarations
1540 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1541 within compound statements. As an extension, GCC also allows this in
1542 C89 mode. For example, you could do:
1551 Each identifier is visible from where it is declared until the end of
1552 the enclosing block.
1554 @node Function Attributes
1555 @section Declaring Attributes of Functions
1556 @cindex function attributes
1557 @cindex declaring attributes of functions
1558 @cindex functions that never return
1559 @cindex functions that return more than once
1560 @cindex functions that have no side effects
1561 @cindex functions in arbitrary sections
1562 @cindex functions that behave like malloc
1563 @cindex @code{volatile} applied to function
1564 @cindex @code{const} applied to function
1565 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1566 @cindex functions with non-null pointer arguments
1567 @cindex functions that are passed arguments in registers on the 386
1568 @cindex functions that pop the argument stack on the 386
1569 @cindex functions that do not pop the argument stack on the 386
1571 In GNU C, you declare certain things about functions called in your program
1572 which help the compiler optimize function calls and check your code more
1575 The keyword @code{__attribute__} allows you to specify special
1576 attributes when making a declaration. This keyword is followed by an
1577 attribute specification inside double parentheses. The following
1578 attributes are currently defined for functions on all targets:
1579 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1580 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1581 @code{format}, @code{format_arg}, @code{no_instrument_function},
1582 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1583 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1584 @code{alias}, @code{warn_unused_result}, @code{nonnull}
1585 and @code{externally_visible}. Several other
1586 attributes are defined for functions on particular target systems. Other
1587 attributes, including @code{section} are supported for variables declarations
1588 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1590 You may also specify attributes with @samp{__} preceding and following
1591 each keyword. This allows you to use them in header files without
1592 being concerned about a possible macro of the same name. For example,
1593 you may use @code{__noreturn__} instead of @code{noreturn}.
1595 @xref{Attribute Syntax}, for details of the exact syntax for using
1599 @c Keep this table alphabetized by attribute name. Treat _ as space.
1601 @item alias ("@var{target}")
1602 @cindex @code{alias} attribute
1603 The @code{alias} attribute causes the declaration to be emitted as an
1604 alias for another symbol, which must be specified. For instance,
1607 void __f () @{ /* @r{Do something.} */; @}
1608 void f () __attribute__ ((weak, alias ("__f")));
1611 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1612 mangled name for the target must be used. It is an error if @samp{__f}
1613 is not defined in the same translation unit.
1615 Not all target machines support this attribute.
1618 @cindex @code{always_inline} function attribute
1619 Generally, functions are not inlined unless optimization is specified.
1620 For functions declared inline, this attribute inlines the function even
1621 if no optimization level was specified.
1623 @cindex @code{flatten} function attribute
1625 Generally, inlining into a function is limited. For a function marked with
1626 this attribute, every call inside this function will be inlined, if possible.
1627 Whether the function itself is considered for inlining depends on its size and
1628 the current inlining parameters. The @code{flatten} attribute only works
1629 reliably in unit-at-a-time mode.
1632 @cindex functions that do pop the argument stack on the 386
1634 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1635 assume that the calling function will pop off the stack space used to
1636 pass arguments. This is
1637 useful to override the effects of the @option{-mrtd} switch.
1640 @cindex @code{const} function attribute
1641 Many functions do not examine any values except their arguments, and
1642 have no effects except the return value. Basically this is just slightly
1643 more strict class than the @code{pure} attribute below, since function is not
1644 allowed to read global memory.
1646 @cindex pointer arguments
1647 Note that a function that has pointer arguments and examines the data
1648 pointed to must @emph{not} be declared @code{const}. Likewise, a
1649 function that calls a non-@code{const} function usually must not be
1650 @code{const}. It does not make sense for a @code{const} function to
1653 The attribute @code{const} is not implemented in GCC versions earlier
1654 than 2.5. An alternative way to declare that a function has no side
1655 effects, which works in the current version and in some older versions,
1659 typedef int intfn ();
1661 extern const intfn square;
1664 This approach does not work in GNU C++ from 2.6.0 on, since the language
1665 specifies that the @samp{const} must be attached to the return value.
1669 @cindex @code{constructor} function attribute
1670 @cindex @code{destructor} function attribute
1671 The @code{constructor} attribute causes the function to be called
1672 automatically before execution enters @code{main ()}. Similarly, the
1673 @code{destructor} attribute causes the function to be called
1674 automatically after @code{main ()} has completed or @code{exit ()} has
1675 been called. Functions with these attributes are useful for
1676 initializing data that will be used implicitly during the execution of
1679 These attributes are not currently implemented for Objective-C@.
1682 @cindex @code{deprecated} attribute.
1683 The @code{deprecated} attribute results in a warning if the function
1684 is used anywhere in the source file. This is useful when identifying
1685 functions that are expected to be removed in a future version of a
1686 program. The warning also includes the location of the declaration
1687 of the deprecated function, to enable users to easily find further
1688 information about why the function is deprecated, or what they should
1689 do instead. Note that the warnings only occurs for uses:
1692 int old_fn () __attribute__ ((deprecated));
1694 int (*fn_ptr)() = old_fn;
1697 results in a warning on line 3 but not line 2.
1699 The @code{deprecated} attribute can also be used for variables and
1700 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1703 @cindex @code{__declspec(dllexport)}
1704 On Microsoft Windows targets and Symbian OS targets the
1705 @code{dllexport} attribute causes the compiler to provide a global
1706 pointer to a pointer in a DLL, so that it can be referenced with the
1707 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1708 name is formed by combining @code{_imp__} and the function or variable
1711 You can use @code{__declspec(dllexport)} as a synonym for
1712 @code{__attribute__ ((dllexport))} for compatibility with other
1715 On systems that support the @code{visibility} attribute, this
1716 attribute also implies ``default'' visibility, unless a
1717 @code{visibility} attribute is explicitly specified. You should avoid
1718 the use of @code{dllexport} with ``hidden'' or ``internal''
1719 visibility; in the future GCC may issue an error for those cases.
1721 Currently, the @code{dllexport} attribute is ignored for inlined
1722 functions, unless the @option{-fkeep-inline-functions} flag has been
1723 used. The attribute is also ignored for undefined symbols.
1725 When applied to C++ classes, the attribute marks defined non-inlined
1726 member functions and static data members as exports. Static consts
1727 initialized in-class are not marked unless they are also defined
1730 For Microsoft Windows targets there are alternative methods for
1731 including the symbol in the DLL's export table such as using a
1732 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1733 the @option{--export-all} linker flag.
1736 @cindex @code{__declspec(dllimport)}
1737 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1738 attribute causes the compiler to reference a function or variable via
1739 a global pointer to a pointer that is set up by the DLL exporting the
1740 symbol. The attribute implies @code{extern} storage. On Microsoft
1741 Windows targets, the pointer name is formed by combining @code{_imp__}
1742 and the function or variable name.
1744 You can use @code{__declspec(dllimport)} as a synonym for
1745 @code{__attribute__ ((dllimport))} for compatibility with other
1748 Currently, the attribute is ignored for inlined functions. If the
1749 attribute is applied to a symbol @emph{definition}, an error is reported.
1750 If a symbol previously declared @code{dllimport} is later defined, the
1751 attribute is ignored in subsequent references, and a warning is emitted.
1752 The attribute is also overridden by a subsequent declaration as
1755 When applied to C++ classes, the attribute marks non-inlined
1756 member functions and static data members as imports. However, the
1757 attribute is ignored for virtual methods to allow creation of vtables
1760 On the SH Symbian OS target the @code{dllimport} attribute also has
1761 another affect---it can cause the vtable and run-time type information
1762 for a class to be exported. This happens when the class has a
1763 dllimport'ed constructor or a non-inline, non-pure virtual function
1764 and, for either of those two conditions, the class also has a inline
1765 constructor or destructor and has a key function that is defined in
1766 the current translation unit.
1768 For Microsoft Windows based targets the use of the @code{dllimport}
1769 attribute on functions is not necessary, but provides a small
1770 performance benefit by eliminating a thunk in the DLL@. The use of the
1771 @code{dllimport} attribute on imported variables was required on older
1772 versions of the GNU linker, but can now be avoided by passing the
1773 @option{--enable-auto-import} switch to the GNU linker. As with
1774 functions, using the attribute for a variable eliminates a thunk in
1777 One drawback to using this attribute is that a pointer to a function
1778 or variable marked as @code{dllimport} cannot be used as a constant
1779 address. On Microsoft Windows targets, the attribute can be disabled
1780 for functions by setting the @option{-mnop-fun-dllimport} flag.
1783 @cindex eight bit data on the H8/300, H8/300H, and H8S
1784 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1785 variable should be placed into the eight bit data section.
1786 The compiler will generate more efficient code for certain operations
1787 on data in the eight bit data area. Note the eight bit data area is limited to
1790 You must use GAS and GLD from GNU binutils version 2.7 or later for
1791 this attribute to work correctly.
1793 @item exception_handler
1794 @cindex exception handler functions on the Blackfin processor
1795 Use this attribute on the Blackfin to indicate that the specified function
1796 is an exception handler. The compiler will generate function entry and
1797 exit sequences suitable for use in an exception handler when this
1798 attribute is present.
1801 @cindex functions which handle memory bank switching
1802 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1803 use a calling convention that takes care of switching memory banks when
1804 entering and leaving a function. This calling convention is also the
1805 default when using the @option{-mlong-calls} option.
1807 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1808 to call and return from a function.
1810 On 68HC11 the compiler will generate a sequence of instructions
1811 to invoke a board-specific routine to switch the memory bank and call the
1812 real function. The board-specific routine simulates a @code{call}.
1813 At the end of a function, it will jump to a board-specific routine
1814 instead of using @code{rts}. The board-specific return routine simulates
1818 @cindex functions that pop the argument stack on the 386
1819 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1820 pass the first argument (if of integral type) in the register ECX and
1821 the second argument (if of integral type) in the register EDX@. Subsequent
1822 and other typed arguments are passed on the stack. The called function will
1823 pop the arguments off the stack. If the number of arguments is variable all
1824 arguments are pushed on the stack.
1826 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1827 @cindex @code{format} function attribute
1829 The @code{format} attribute specifies that a function takes @code{printf},
1830 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1831 should be type-checked against a format string. For example, the
1836 my_printf (void *my_object, const char *my_format, ...)
1837 __attribute__ ((format (printf, 2, 3)));
1841 causes the compiler to check the arguments in calls to @code{my_printf}
1842 for consistency with the @code{printf} style format string argument
1845 The parameter @var{archetype} determines how the format string is
1846 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1847 or @code{strfmon}. (You can also use @code{__printf__},
1848 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1849 parameter @var{string-index} specifies which argument is the format
1850 string argument (starting from 1), while @var{first-to-check} is the
1851 number of the first argument to check against the format string. For
1852 functions where the arguments are not available to be checked (such as
1853 @code{vprintf}), specify the third parameter as zero. In this case the
1854 compiler only checks the format string for consistency. For
1855 @code{strftime} formats, the third parameter is required to be zero.
1856 Since non-static C++ methods have an implicit @code{this} argument, the
1857 arguments of such methods should be counted from two, not one, when
1858 giving values for @var{string-index} and @var{first-to-check}.
1860 In the example above, the format string (@code{my_format}) is the second
1861 argument of the function @code{my_print}, and the arguments to check
1862 start with the third argument, so the correct parameters for the format
1863 attribute are 2 and 3.
1865 @opindex ffreestanding
1866 @opindex fno-builtin
1867 The @code{format} attribute allows you to identify your own functions
1868 which take format strings as arguments, so that GCC can check the
1869 calls to these functions for errors. The compiler always (unless
1870 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1871 for the standard library functions @code{printf}, @code{fprintf},
1872 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1873 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1874 warnings are requested (using @option{-Wformat}), so there is no need to
1875 modify the header file @file{stdio.h}. In C99 mode, the functions
1876 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1877 @code{vsscanf} are also checked. Except in strictly conforming C
1878 standard modes, the X/Open function @code{strfmon} is also checked as
1879 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1880 @xref{C Dialect Options,,Options Controlling C Dialect}.
1882 The target may provide additional types of format checks.
1883 @xref{Target Format Checks,,Format Checks Specific to Particular
1886 @item format_arg (@var{string-index})
1887 @cindex @code{format_arg} function attribute
1888 @opindex Wformat-nonliteral
1889 The @code{format_arg} attribute specifies that a function takes a format
1890 string for a @code{printf}, @code{scanf}, @code{strftime} or
1891 @code{strfmon} style function and modifies it (for example, to translate
1892 it into another language), so the result can be passed to a
1893 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1894 function (with the remaining arguments to the format function the same
1895 as they would have been for the unmodified string). For example, the
1900 my_dgettext (char *my_domain, const char *my_format)
1901 __attribute__ ((format_arg (2)));
1905 causes the compiler to check the arguments in calls to a @code{printf},
1906 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1907 format string argument is a call to the @code{my_dgettext} function, for
1908 consistency with the format string argument @code{my_format}. If the
1909 @code{format_arg} attribute had not been specified, all the compiler
1910 could tell in such calls to format functions would be that the format
1911 string argument is not constant; this would generate a warning when
1912 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1913 without the attribute.
1915 The parameter @var{string-index} specifies which argument is the format
1916 string argument (starting from one). Since non-static C++ methods have
1917 an implicit @code{this} argument, the arguments of such methods should
1918 be counted from two.
1920 The @code{format-arg} attribute allows you to identify your own
1921 functions which modify format strings, so that GCC can check the
1922 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1923 type function whose operands are a call to one of your own function.
1924 The compiler always treats @code{gettext}, @code{dgettext}, and
1925 @code{dcgettext} in this manner except when strict ISO C support is
1926 requested by @option{-ansi} or an appropriate @option{-std} option, or
1927 @option{-ffreestanding} or @option{-fno-builtin}
1928 is used. @xref{C Dialect Options,,Options
1929 Controlling C Dialect}.
1931 @item function_vector
1932 @cindex calling functions through the function vector on the H8/300 processors
1933 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1934 function should be called through the function vector. Calling a
1935 function through the function vector will reduce code size, however;
1936 the function vector has a limited size (maximum 128 entries on the H8/300
1937 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1939 You must use GAS and GLD from GNU binutils version 2.7 or later for
1940 this attribute to work correctly.
1943 @cindex interrupt handler functions
1944 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
1945 ports to indicate that the specified function is an interrupt handler.
1946 The compiler will generate function entry and exit sequences suitable
1947 for use in an interrupt handler when this attribute is present.
1949 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1950 SH processors can be specified via the @code{interrupt_handler} attribute.
1952 Note, on the AVR, interrupts will be enabled inside the function.
1954 Note, for the ARM, you can specify the kind of interrupt to be handled by
1955 adding an optional parameter to the interrupt attribute like this:
1958 void f () __attribute__ ((interrupt ("IRQ")));
1961 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
1963 @item interrupt_handler
1964 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
1965 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
1966 indicate that the specified function is an interrupt handler. The compiler
1967 will generate function entry and exit sequences suitable for use in an
1968 interrupt handler when this attribute is present.
1971 @cindex User stack pointer in interrupts on the Blackfin
1972 When used together with @code{interrupt_handler}, @code{exception_handler}
1973 or @code{nmi_handler}, code will be generated to load the stack pointer
1974 from the USP register in the function prologue.
1976 @item long_call/short_call
1977 @cindex indirect calls on ARM
1978 This attribute specifies how a particular function is called on
1979 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
1980 command line switch and @code{#pragma long_calls} settings. The
1981 @code{long_call} attribute indicates that the function might be far
1982 away from the call site and require a different (more expensive)
1983 calling sequence. The @code{short_call} attribute always places
1984 the offset to the function from the call site into the @samp{BL}
1985 instruction directly.
1987 @item longcall/shortcall
1988 @cindex functions called via pointer on the RS/6000 and PowerPC
1989 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
1990 indicates that the function might be far away from the call site and
1991 require a different (more expensive) calling sequence. The
1992 @code{shortcall} attribute indicates that the function is always close
1993 enough for the shorter calling sequence to be used. These attributes
1994 override both the @option{-mlongcall} switch and, on the RS/6000 and
1995 PowerPC, the @code{#pragma longcall} setting.
1997 @xref{RS/6000 and PowerPC Options}, for more information on whether long
1998 calls are necessary.
2001 @cindex indirect calls on MIPS
2002 This attribute specifies how a particular function is called on MIPS@.
2003 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2004 command line switch. This attribute causes the compiler to always call
2005 the function by first loading its address into a register, and then using
2006 the contents of that register.
2009 @cindex @code{malloc} attribute
2010 The @code{malloc} attribute is used to tell the compiler that a function
2011 may be treated as if any non-@code{NULL} pointer it returns cannot
2012 alias any other pointer valid when the function returns.
2013 This will often improve optimization.
2014 Standard functions with this property include @code{malloc} and
2015 @code{calloc}. @code{realloc}-like functions have this property as
2016 long as the old pointer is never referred to (including comparing it
2017 to the new pointer) after the function returns a non-@code{NULL}
2020 @item model (@var{model-name})
2021 @cindex function addressability on the M32R/D
2022 @cindex variable addressability on the IA-64
2024 On the M32R/D, use this attribute to set the addressability of an
2025 object, and of the code generated for a function. The identifier
2026 @var{model-name} is one of @code{small}, @code{medium}, or
2027 @code{large}, representing each of the code models.
2029 Small model objects live in the lower 16MB of memory (so that their
2030 addresses can be loaded with the @code{ld24} instruction), and are
2031 callable with the @code{bl} instruction.
2033 Medium model objects may live anywhere in the 32-bit address space (the
2034 compiler will generate @code{seth/add3} instructions to load their addresses),
2035 and are callable with the @code{bl} instruction.
2037 Large model objects may live anywhere in the 32-bit address space (the
2038 compiler will generate @code{seth/add3} instructions to load their addresses),
2039 and may not be reachable with the @code{bl} instruction (the compiler will
2040 generate the much slower @code{seth/add3/jl} instruction sequence).
2042 On IA-64, use this attribute to set the addressability of an object.
2043 At present, the only supported identifier for @var{model-name} is
2044 @code{small}, indicating addressability via ``small'' (22-bit)
2045 addresses (so that their addresses can be loaded with the @code{addl}
2046 instruction). Caveat: such addressing is by definition not position
2047 independent and hence this attribute must not be used for objects
2048 defined by shared libraries.
2051 @cindex function without a prologue/epilogue code
2052 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2053 specified function does not need prologue/epilogue sequences generated by
2054 the compiler. It is up to the programmer to provide these sequences.
2057 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2058 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2059 use the normal calling convention based on @code{jsr} and @code{rts}.
2060 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2064 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2065 Use this attribute together with @code{interrupt_handler},
2066 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2067 entry code should enable nested interrupts or exceptions.
2070 @cindex NMI handler functions on the Blackfin processor
2071 Use this attribute on the Blackfin to indicate that the specified function
2072 is an NMI handler. The compiler will generate function entry and
2073 exit sequences suitable for use in an NMI handler when this
2074 attribute is present.
2076 @item no_instrument_function
2077 @cindex @code{no_instrument_function} function attribute
2078 @opindex finstrument-functions
2079 If @option{-finstrument-functions} is given, profiling function calls will
2080 be generated at entry and exit of most user-compiled functions.
2081 Functions with this attribute will not be so instrumented.
2084 @cindex @code{noinline} function attribute
2085 This function attribute prevents a function from being considered for
2088 @item nonnull (@var{arg-index}, @dots{})
2089 @cindex @code{nonnull} function attribute
2090 The @code{nonnull} attribute specifies that some function parameters should
2091 be non-null pointers. For instance, the declaration:
2095 my_memcpy (void *dest, const void *src, size_t len)
2096 __attribute__((nonnull (1, 2)));
2100 causes the compiler to check that, in calls to @code{my_memcpy},
2101 arguments @var{dest} and @var{src} are non-null. If the compiler
2102 determines that a null pointer is passed in an argument slot marked
2103 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2104 is issued. The compiler may also choose to make optimizations based
2105 on the knowledge that certain function arguments will not be null.
2107 If no argument index list is given to the @code{nonnull} attribute,
2108 all pointer arguments are marked as non-null. To illustrate, the
2109 following declaration is equivalent to the previous example:
2113 my_memcpy (void *dest, const void *src, size_t len)
2114 __attribute__((nonnull));
2118 @cindex @code{noreturn} function attribute
2119 A few standard library functions, such as @code{abort} and @code{exit},
2120 cannot return. GCC knows this automatically. Some programs define
2121 their own functions that never return. You can declare them
2122 @code{noreturn} to tell the compiler this fact. For example,
2126 void fatal () __attribute__ ((noreturn));
2129 fatal (/* @r{@dots{}} */)
2131 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2137 The @code{noreturn} keyword tells the compiler to assume that
2138 @code{fatal} cannot return. It can then optimize without regard to what
2139 would happen if @code{fatal} ever did return. This makes slightly
2140 better code. More importantly, it helps avoid spurious warnings of
2141 uninitialized variables.
2143 The @code{noreturn} keyword does not affect the exceptional path when that
2144 applies: a @code{noreturn}-marked function may still return to the caller
2145 by throwing an exception or calling @code{longjmp}.
2147 Do not assume that registers saved by the calling function are
2148 restored before calling the @code{noreturn} function.
2150 It does not make sense for a @code{noreturn} function to have a return
2151 type other than @code{void}.
2153 The attribute @code{noreturn} is not implemented in GCC versions
2154 earlier than 2.5. An alternative way to declare that a function does
2155 not return, which works in the current version and in some older
2156 versions, is as follows:
2159 typedef void voidfn ();
2161 volatile voidfn fatal;
2164 This approach does not work in GNU C++.
2167 @cindex @code{nothrow} function attribute
2168 The @code{nothrow} attribute is used to inform the compiler that a
2169 function cannot throw an exception. For example, most functions in
2170 the standard C library can be guaranteed not to throw an exception
2171 with the notable exceptions of @code{qsort} and @code{bsearch} that
2172 take function pointer arguments. The @code{nothrow} attribute is not
2173 implemented in GCC versions earlier than 3.3.
2176 @cindex @code{pure} function attribute
2177 Many functions have no effects except the return value and their
2178 return value depends only on the parameters and/or global variables.
2179 Such a function can be subject
2180 to common subexpression elimination and loop optimization just as an
2181 arithmetic operator would be. These functions should be declared
2182 with the attribute @code{pure}. For example,
2185 int square (int) __attribute__ ((pure));
2189 says that the hypothetical function @code{square} is safe to call
2190 fewer times than the program says.
2192 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2193 Interesting non-pure functions are functions with infinite loops or those
2194 depending on volatile memory or other system resource, that may change between
2195 two consecutive calls (such as @code{feof} in a multithreading environment).
2197 The attribute @code{pure} is not implemented in GCC versions earlier
2200 @item regparm (@var{number})
2201 @cindex @code{regparm} attribute
2202 @cindex functions that are passed arguments in registers on the 386
2203 On the Intel 386, the @code{regparm} attribute causes the compiler to
2204 pass arguments number one to @var{number} if they are of integral type
2205 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2206 take a variable number of arguments will continue to be passed all of their
2207 arguments on the stack.
2209 Beware that on some ELF systems this attribute is unsuitable for
2210 global functions in shared libraries with lazy binding (which is the
2211 default). Lazy binding will send the first call via resolving code in
2212 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2213 per the standard calling conventions. Solaris 8 is affected by this.
2214 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2215 safe since the loaders there save all registers. (Lazy binding can be
2216 disabled with the linker or the loader if desired, to avoid the
2220 @cindex @code{sseregparm} attribute
2221 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2222 causes the compiler to pass up to 8 floating point arguments in
2223 SSE registers instead of on the stack. Functions that take a
2224 variable number of arguments will continue to pass all of their
2225 floating point arguments on the stack.
2227 @item force_align_arg_pointer
2228 @cindex @code{force_align_arg_pointer} attribute
2229 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2230 applied to individual function definitions, generating an alternate
2231 prologue and epilogue that realigns the runtime stack. This supports
2232 mixing legacy codes that run with a 4-byte aligned stack with modern
2233 codes that keep a 16-byte stack for SSE compatibility. The alternate
2234 prologue and epilogue are slower and bigger than the regular ones, and
2235 the alternate prologue requires a scratch register; this lowers the
2236 number of registers available if used in conjunction with the
2237 @code{regparm} attribute. The @code{force_align_arg_pointer}
2238 attribute is incompatible with nested functions; this is considered a
2242 @cindex @code{returns_twice} attribute
2243 The @code{returns_twice} attribute tells the compiler that a function may
2244 return more than one time. The compiler will ensure that all registers
2245 are dead before calling such a function and will emit a warning about
2246 the variables that may be clobbered after the second return from the
2247 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2248 The @code{longjmp}-like counterpart of such function, if any, might need
2249 to be marked with the @code{noreturn} attribute.
2252 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2253 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2254 all registers except the stack pointer should be saved in the prologue
2255 regardless of whether they are used or not.
2257 @item section ("@var{section-name}")
2258 @cindex @code{section} function attribute
2259 Normally, the compiler places the code it generates in the @code{text} section.
2260 Sometimes, however, you need additional sections, or you need certain
2261 particular functions to appear in special sections. The @code{section}
2262 attribute specifies that a function lives in a particular section.
2263 For example, the declaration:
2266 extern void foobar (void) __attribute__ ((section ("bar")));
2270 puts the function @code{foobar} in the @code{bar} section.
2272 Some file formats do not support arbitrary sections so the @code{section}
2273 attribute is not available on all platforms.
2274 If you need to map the entire contents of a module to a particular
2275 section, consider using the facilities of the linker instead.
2278 @cindex @code{sentinel} function attribute
2279 This function attribute ensures that a parameter in a function call is
2280 an explicit @code{NULL}. The attribute is only valid on variadic
2281 functions. By default, the sentinel is located at position zero, the
2282 last parameter of the function call. If an optional integer position
2283 argument P is supplied to the attribute, the sentinel must be located at
2284 position P counting backwards from the end of the argument list.
2287 __attribute__ ((sentinel))
2289 __attribute__ ((sentinel(0)))
2292 The attribute is automatically set with a position of 0 for the built-in
2293 functions @code{execl} and @code{execlp}. The built-in function
2294 @code{execle} has the attribute set with a position of 1.
2296 A valid @code{NULL} in this context is defined as zero with any pointer
2297 type. If your system defines the @code{NULL} macro with an integer type
2298 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2299 with a copy that redefines NULL appropriately.
2301 The warnings for missing or incorrect sentinels are enabled with
2305 See long_call/short_call.
2308 See longcall/shortcall.
2311 @cindex signal handler functions on the AVR processors
2312 Use this attribute on the AVR to indicate that the specified
2313 function is a signal handler. The compiler will generate function
2314 entry and exit sequences suitable for use in a signal handler when this
2315 attribute is present. Interrupts will be disabled inside the function.
2318 Use this attribute on the SH to indicate an @code{interrupt_handler}
2319 function should switch to an alternate stack. It expects a string
2320 argument that names a global variable holding the address of the
2325 void f () __attribute__ ((interrupt_handler,
2326 sp_switch ("alt_stack")));
2330 @cindex functions that pop the argument stack on the 386
2331 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2332 assume that the called function will pop off the stack space used to
2333 pass arguments, unless it takes a variable number of arguments.
2336 @cindex tiny data section on the H8/300H and H8S
2337 Use this attribute on the H8/300H and H8S to indicate that the specified
2338 variable should be placed into the tiny data section.
2339 The compiler will generate more efficient code for loads and stores
2340 on data in the tiny data section. Note the tiny data area is limited to
2341 slightly under 32kbytes of data.
2344 Use this attribute on the SH for an @code{interrupt_handler} to return using
2345 @code{trapa} instead of @code{rte}. This attribute expects an integer
2346 argument specifying the trap number to be used.
2349 @cindex @code{unused} attribute.
2350 This attribute, attached to a function, means that the function is meant
2351 to be possibly unused. GCC will not produce a warning for this
2355 @cindex @code{used} attribute.
2356 This attribute, attached to a function, means that code must be emitted
2357 for the function even if it appears that the function is not referenced.
2358 This is useful, for example, when the function is referenced only in
2361 @item visibility ("@var{visibility_type}")
2362 @cindex @code{visibility} attribute
2363 This attribute affects the linkage of the declaration to which it is attached.
2364 There are four supported @var{visibility_type} values: default,
2365 hidden, protected or internal visibility.
2368 void __attribute__ ((visibility ("protected")))
2369 f () @{ /* @r{Do something.} */; @}
2370 int i __attribute__ ((visibility ("hidden")));
2373 The possible values of @var{visibility_type} correspond to the
2374 visibility settings in the ELF gABI.
2377 @c keep this list of visibilities in alphabetical order.
2380 Default visibility is the normal case for the object file format.
2381 This value is available for the visibility attribute to override other
2382 options that may change the assumed visibility of entities.
2384 On ELF, default visibility means that the declaration is visible to other
2385 modules and, in shared libraries, means that the declared entity may be
2388 On Darwin, default visibility means that the declaration is visible to
2391 Default visibility corresponds to ``external linkage'' in the language.
2394 Hidden visibility indicates that the entity declared will have a new
2395 form of linkage, which we'll call ``hidden linkage''. Two
2396 declarations of an object with hidden linkage refer to the same object
2397 if they are in the same shared object.
2400 Internal visibility is like hidden visibility, but with additional
2401 processor specific semantics. Unless otherwise specified by the
2402 psABI, GCC defines internal visibility to mean that a function is
2403 @emph{never} called from another module. Compare this with hidden
2404 functions which, while they cannot be referenced directly by other
2405 modules, can be referenced indirectly via function pointers. By
2406 indicating that a function cannot be called from outside the module,
2407 GCC may for instance omit the load of a PIC register since it is known
2408 that the calling function loaded the correct value.
2411 Protected visibility is like default visibility except that it
2412 indicates that references within the defining module will bind to the
2413 definition in that module. That is, the declared entity cannot be
2414 overridden by another module.
2418 All visibilities are supported on many, but not all, ELF targets
2419 (supported when the assembler supports the @samp{.visibility}
2420 pseudo-op). Default visibility is supported everywhere. Hidden
2421 visibility is supported on Darwin targets.
2423 The visibility attribute should be applied only to declarations which
2424 would otherwise have external linkage. The attribute should be applied
2425 consistently, so that the same entity should not be declared with
2426 different settings of the attribute.
2428 In C++, the visibility attribute applies to types as well as functions
2429 and objects, because in C++ types have linkage. A class must not have
2430 greater visibility than its non-static data member types and bases,
2431 and class members default to the visibility of their class. Also, a
2432 declaration without explicit visibility is limited to the visibility
2435 In C++, you can mark member functions and static member variables of a
2436 class with the visibility attribute. This is useful if if you know a
2437 particular method or static member variable should only be used from
2438 one shared object; then you can mark it hidden while the rest of the
2439 class has default visibility. Care must be taken to avoid breaking
2440 the One Definition Rule; for example, it is usually not useful to mark
2441 an inline method as hidden without marking the whole class as hidden.
2443 A C++ namespace declaration can also have the visibility attribute.
2444 This attribute applies only to the particular namespace body, not to
2445 other definitions of the same namespace; it is equivalent to using
2446 @samp{#pragma GCC visibility} before and after the namespace
2447 definition (@pxref{Visibility Pragmas}).
2449 In C++, if a template argument has limited visibility, this
2450 restriction is implicitly propagated to the template instantiation.
2451 Otherwise, template instantiations and specializations default to the
2452 visibility of their template.
2454 If both the template and enclosing class have explicit visibility, the
2455 visibility from the template is used.
2457 @item warn_unused_result
2458 @cindex @code{warn_unused_result} attribute
2459 The @code{warn_unused_result} attribute causes a warning to be emitted
2460 if a caller of the function with this attribute does not use its
2461 return value. This is useful for functions where not checking
2462 the result is either a security problem or always a bug, such as
2466 int fn () __attribute__ ((warn_unused_result));
2469 if (fn () < 0) return -1;
2475 results in warning on line 5.
2478 @cindex @code{weak} attribute
2479 The @code{weak} attribute causes the declaration to be emitted as a weak
2480 symbol rather than a global. This is primarily useful in defining
2481 library functions which can be overridden in user code, though it can
2482 also be used with non-function declarations. Weak symbols are supported
2483 for ELF targets, and also for a.out targets when using the GNU assembler
2487 @itemx weakref ("@var{target}")
2488 @cindex @code{weakref} attribute
2489 The @code{weakref} attribute marks a declaration as a weak reference.
2490 Without arguments, it should be accompanied by an @code{alias} attribute
2491 naming the target symbol. Optionally, the @var{target} may be given as
2492 an argument to @code{weakref} itself. In either case, @code{weakref}
2493 implicitly marks the declaration as @code{weak}. Without a
2494 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2495 @code{weakref} is equivalent to @code{weak}.
2498 static int x() __attribute__ ((weakref ("y")));
2499 /* is equivalent to... */
2500 static int x() __attribute__ ((weak, weakref, alias ("y")));
2502 static int x() __attribute__ ((weakref));
2503 static int x() __attribute__ ((alias ("y")));
2506 A weak reference is an alias that does not by itself require a
2507 definition to be given for the target symbol. If the target symbol is
2508 only referenced through weak references, then the becomes a @code{weak}
2509 undefined symbol. If it is directly referenced, however, then such
2510 strong references prevail, and a definition will be required for the
2511 symbol, not necessarily in the same translation unit.
2513 The effect is equivalent to moving all references to the alias to a
2514 separate translation unit, renaming the alias to the aliased symbol,
2515 declaring it as weak, compiling the two separate translation units and
2516 performing a reloadable link on them.
2518 At present, a declaration to which @code{weakref} is attached can
2519 only be @code{static}.
2521 @item externally_visible
2522 @cindex @code{externally_visible} attribute.
2523 This attribute, attached to a global variable or function nullify
2524 effect of @option{-fwhole-program} command line option, so the object
2525 remain visible outside the current compilation unit
2529 You can specify multiple attributes in a declaration by separating them
2530 by commas within the double parentheses or by immediately following an
2531 attribute declaration with another attribute declaration.
2533 @cindex @code{#pragma}, reason for not using
2534 @cindex pragma, reason for not using
2535 Some people object to the @code{__attribute__} feature, suggesting that
2536 ISO C's @code{#pragma} should be used instead. At the time
2537 @code{__attribute__} was designed, there were two reasons for not doing
2542 It is impossible to generate @code{#pragma} commands from a macro.
2545 There is no telling what the same @code{#pragma} might mean in another
2549 These two reasons applied to almost any application that might have been
2550 proposed for @code{#pragma}. It was basically a mistake to use
2551 @code{#pragma} for @emph{anything}.
2553 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2554 to be generated from macros. In addition, a @code{#pragma GCC}
2555 namespace is now in use for GCC-specific pragmas. However, it has been
2556 found convenient to use @code{__attribute__} to achieve a natural
2557 attachment of attributes to their corresponding declarations, whereas
2558 @code{#pragma GCC} is of use for constructs that do not naturally form
2559 part of the grammar. @xref{Other Directives,,Miscellaneous
2560 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2562 @node Attribute Syntax
2563 @section Attribute Syntax
2564 @cindex attribute syntax
2566 This section describes the syntax with which @code{__attribute__} may be
2567 used, and the constructs to which attribute specifiers bind, for the C
2568 language. Some details may vary for C++ and Objective-C@. Because of
2569 infelicities in the grammar for attributes, some forms described here
2570 may not be successfully parsed in all cases.
2572 There are some problems with the semantics of attributes in C++. For
2573 example, there are no manglings for attributes, although they may affect
2574 code generation, so problems may arise when attributed types are used in
2575 conjunction with templates or overloading. Similarly, @code{typeid}
2576 does not distinguish between types with different attributes. Support
2577 for attributes in C++ may be restricted in future to attributes on
2578 declarations only, but not on nested declarators.
2580 @xref{Function Attributes}, for details of the semantics of attributes
2581 applying to functions. @xref{Variable Attributes}, for details of the
2582 semantics of attributes applying to variables. @xref{Type Attributes},
2583 for details of the semantics of attributes applying to structure, union
2584 and enumerated types.
2586 An @dfn{attribute specifier} is of the form
2587 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2588 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2589 each attribute is one of the following:
2593 Empty. Empty attributes are ignored.
2596 A word (which may be an identifier such as @code{unused}, or a reserved
2597 word such as @code{const}).
2600 A word, followed by, in parentheses, parameters for the attribute.
2601 These parameters take one of the following forms:
2605 An identifier. For example, @code{mode} attributes use this form.
2608 An identifier followed by a comma and a non-empty comma-separated list
2609 of expressions. For example, @code{format} attributes use this form.
2612 A possibly empty comma-separated list of expressions. For example,
2613 @code{format_arg} attributes use this form with the list being a single
2614 integer constant expression, and @code{alias} attributes use this form
2615 with the list being a single string constant.
2619 An @dfn{attribute specifier list} is a sequence of one or more attribute
2620 specifiers, not separated by any other tokens.
2622 In GNU C, an attribute specifier list may appear after the colon following a
2623 label, other than a @code{case} or @code{default} label. The only
2624 attribute it makes sense to use after a label is @code{unused}. This
2625 feature is intended for code generated by programs which contains labels
2626 that may be unused but which is compiled with @option{-Wall}. It would
2627 not normally be appropriate to use in it human-written code, though it
2628 could be useful in cases where the code that jumps to the label is
2629 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2630 such placement of attribute lists, as it is permissible for a
2631 declaration, which could begin with an attribute list, to be labelled in
2632 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2633 does not arise there.
2635 An attribute specifier list may appear as part of a @code{struct},
2636 @code{union} or @code{enum} specifier. It may go either immediately
2637 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2638 the closing brace. The former syntax is preferred.
2639 Where attribute specifiers follow the closing brace, they are considered
2640 to relate to the structure, union or enumerated type defined, not to any
2641 enclosing declaration the type specifier appears in, and the type
2642 defined is not complete until after the attribute specifiers.
2643 @c Otherwise, there would be the following problems: a shift/reduce
2644 @c conflict between attributes binding the struct/union/enum and
2645 @c binding to the list of specifiers/qualifiers; and "aligned"
2646 @c attributes could use sizeof for the structure, but the size could be
2647 @c changed later by "packed" attributes.
2649 Otherwise, an attribute specifier appears as part of a declaration,
2650 counting declarations of unnamed parameters and type names, and relates
2651 to that declaration (which may be nested in another declaration, for
2652 example in the case of a parameter declaration), or to a particular declarator
2653 within a declaration. Where an
2654 attribute specifier is applied to a parameter declared as a function or
2655 an array, it should apply to the function or array rather than the
2656 pointer to which the parameter is implicitly converted, but this is not
2657 yet correctly implemented.
2659 Any list of specifiers and qualifiers at the start of a declaration may
2660 contain attribute specifiers, whether or not such a list may in that
2661 context contain storage class specifiers. (Some attributes, however,
2662 are essentially in the nature of storage class specifiers, and only make
2663 sense where storage class specifiers may be used; for example,
2664 @code{section}.) There is one necessary limitation to this syntax: the
2665 first old-style parameter declaration in a function definition cannot
2666 begin with an attribute specifier, because such an attribute applies to
2667 the function instead by syntax described below (which, however, is not
2668 yet implemented in this case). In some other cases, attribute
2669 specifiers are permitted by this grammar but not yet supported by the
2670 compiler. All attribute specifiers in this place relate to the
2671 declaration as a whole. In the obsolescent usage where a type of
2672 @code{int} is implied by the absence of type specifiers, such a list of
2673 specifiers and qualifiers may be an attribute specifier list with no
2674 other specifiers or qualifiers.
2676 At present, the first parameter in a function prototype must have some
2677 type specifier which is not an attribute specifier; this resolves an
2678 ambiguity in the interpretation of @code{void f(int
2679 (__attribute__((foo)) x))}, but is subject to change. At present, if
2680 the parentheses of a function declarator contain only attributes then
2681 those attributes are ignored, rather than yielding an error or warning
2682 or implying a single parameter of type int, but this is subject to
2685 An attribute specifier list may appear immediately before a declarator
2686 (other than the first) in a comma-separated list of declarators in a
2687 declaration of more than one identifier using a single list of
2688 specifiers and qualifiers. Such attribute specifiers apply
2689 only to the identifier before whose declarator they appear. For
2693 __attribute__((noreturn)) void d0 (void),
2694 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2699 the @code{noreturn} attribute applies to all the functions
2700 declared; the @code{format} attribute only applies to @code{d1}.
2702 An attribute specifier list may appear immediately before the comma,
2703 @code{=} or semicolon terminating the declaration of an identifier other
2704 than a function definition. At present, such attribute specifiers apply
2705 to the declared object or function, but in future they may attach to the
2706 outermost adjacent declarator. In simple cases there is no difference,
2707 but, for example, in
2710 void (****f)(void) __attribute__((noreturn));
2714 at present the @code{noreturn} attribute applies to @code{f}, which
2715 causes a warning since @code{f} is not a function, but in future it may
2716 apply to the function @code{****f}. The precise semantics of what
2717 attributes in such cases will apply to are not yet specified. Where an
2718 assembler name for an object or function is specified (@pxref{Asm
2719 Labels}), at present the attribute must follow the @code{asm}
2720 specification; in future, attributes before the @code{asm} specification
2721 may apply to the adjacent declarator, and those after it to the declared
2724 An attribute specifier list may, in future, be permitted to appear after
2725 the declarator in a function definition (before any old-style parameter
2726 declarations or the function body).
2728 Attribute specifiers may be mixed with type qualifiers appearing inside
2729 the @code{[]} of a parameter array declarator, in the C99 construct by
2730 which such qualifiers are applied to the pointer to which the array is
2731 implicitly converted. Such attribute specifiers apply to the pointer,
2732 not to the array, but at present this is not implemented and they are
2735 An attribute specifier list may appear at the start of a nested
2736 declarator. At present, there are some limitations in this usage: the
2737 attributes correctly apply to the declarator, but for most individual
2738 attributes the semantics this implies are not implemented.
2739 When attribute specifiers follow the @code{*} of a pointer
2740 declarator, they may be mixed with any type qualifiers present.
2741 The following describes the formal semantics of this syntax. It will make the
2742 most sense if you are familiar with the formal specification of
2743 declarators in the ISO C standard.
2745 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2746 D1}, where @code{T} contains declaration specifiers that specify a type
2747 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2748 contains an identifier @var{ident}. The type specified for @var{ident}
2749 for derived declarators whose type does not include an attribute
2750 specifier is as in the ISO C standard.
2752 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2753 and the declaration @code{T D} specifies the type
2754 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2755 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2756 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2758 If @code{D1} has the form @code{*
2759 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2760 declaration @code{T D} specifies the type
2761 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2762 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2763 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2769 void (__attribute__((noreturn)) ****f) (void);
2773 specifies the type ``pointer to pointer to pointer to pointer to
2774 non-returning function returning @code{void}''. As another example,
2777 char *__attribute__((aligned(8))) *f;
2781 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2782 Note again that this does not work with most attributes; for example,
2783 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2784 is not yet supported.
2786 For compatibility with existing code written for compiler versions that
2787 did not implement attributes on nested declarators, some laxity is
2788 allowed in the placing of attributes. If an attribute that only applies
2789 to types is applied to a declaration, it will be treated as applying to
2790 the type of that declaration. If an attribute that only applies to
2791 declarations is applied to the type of a declaration, it will be treated
2792 as applying to that declaration; and, for compatibility with code
2793 placing the attributes immediately before the identifier declared, such
2794 an attribute applied to a function return type will be treated as
2795 applying to the function type, and such an attribute applied to an array
2796 element type will be treated as applying to the array type. If an
2797 attribute that only applies to function types is applied to a
2798 pointer-to-function type, it will be treated as applying to the pointer
2799 target type; if such an attribute is applied to a function return type
2800 that is not a pointer-to-function type, it will be treated as applying
2801 to the function type.
2803 @node Function Prototypes
2804 @section Prototypes and Old-Style Function Definitions
2805 @cindex function prototype declarations
2806 @cindex old-style function definitions
2807 @cindex promotion of formal parameters
2809 GNU C extends ISO C to allow a function prototype to override a later
2810 old-style non-prototype definition. Consider the following example:
2813 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2820 /* @r{Prototype function declaration.} */
2821 int isroot P((uid_t));
2823 /* @r{Old-style function definition.} */
2825 isroot (x) /* @r{??? lossage here ???} */
2832 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2833 not allow this example, because subword arguments in old-style
2834 non-prototype definitions are promoted. Therefore in this example the
2835 function definition's argument is really an @code{int}, which does not
2836 match the prototype argument type of @code{short}.
2838 This restriction of ISO C makes it hard to write code that is portable
2839 to traditional C compilers, because the programmer does not know
2840 whether the @code{uid_t} type is @code{short}, @code{int}, or
2841 @code{long}. Therefore, in cases like these GNU C allows a prototype
2842 to override a later old-style definition. More precisely, in GNU C, a
2843 function prototype argument type overrides the argument type specified
2844 by a later old-style definition if the former type is the same as the
2845 latter type before promotion. Thus in GNU C the above example is
2846 equivalent to the following:
2859 GNU C++ does not support old-style function definitions, so this
2860 extension is irrelevant.
2863 @section C++ Style Comments
2865 @cindex C++ comments
2866 @cindex comments, C++ style
2868 In GNU C, you may use C++ style comments, which start with @samp{//} and
2869 continue until the end of the line. Many other C implementations allow
2870 such comments, and they are included in the 1999 C standard. However,
2871 C++ style comments are not recognized if you specify an @option{-std}
2872 option specifying a version of ISO C before C99, or @option{-ansi}
2873 (equivalent to @option{-std=c89}).
2876 @section Dollar Signs in Identifier Names
2878 @cindex dollar signs in identifier names
2879 @cindex identifier names, dollar signs in
2881 In GNU C, you may normally use dollar signs in identifier names.
2882 This is because many traditional C implementations allow such identifiers.
2883 However, dollar signs in identifiers are not supported on a few target
2884 machines, typically because the target assembler does not allow them.
2886 @node Character Escapes
2887 @section The Character @key{ESC} in Constants
2889 You can use the sequence @samp{\e} in a string or character constant to
2890 stand for the ASCII character @key{ESC}.
2893 @section Inquiring on Alignment of Types or Variables
2895 @cindex type alignment
2896 @cindex variable alignment
2898 The keyword @code{__alignof__} allows you to inquire about how an object
2899 is aligned, or the minimum alignment usually required by a type. Its
2900 syntax is just like @code{sizeof}.
2902 For example, if the target machine requires a @code{double} value to be
2903 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2904 This is true on many RISC machines. On more traditional machine
2905 designs, @code{__alignof__ (double)} is 4 or even 2.
2907 Some machines never actually require alignment; they allow reference to any
2908 data type even at an odd address. For these machines, @code{__alignof__}
2909 reports the @emph{recommended} alignment of a type.
2911 If the operand of @code{__alignof__} is an lvalue rather than a type,
2912 its value is the required alignment for its type, taking into account
2913 any minimum alignment specified with GCC's @code{__attribute__}
2914 extension (@pxref{Variable Attributes}). For example, after this
2918 struct foo @{ int x; char y; @} foo1;
2922 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2923 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2925 It is an error to ask for the alignment of an incomplete type.
2927 @node Variable Attributes
2928 @section Specifying Attributes of Variables
2929 @cindex attribute of variables
2930 @cindex variable attributes
2932 The keyword @code{__attribute__} allows you to specify special
2933 attributes of variables or structure fields. This keyword is followed
2934 by an attribute specification inside double parentheses. Some
2935 attributes are currently defined generically for variables.
2936 Other attributes are defined for variables on particular target
2937 systems. Other attributes are available for functions
2938 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2939 Other front ends might define more attributes
2940 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2942 You may also specify attributes with @samp{__} preceding and following
2943 each keyword. This allows you to use them in header files without
2944 being concerned about a possible macro of the same name. For example,
2945 you may use @code{__aligned__} instead of @code{aligned}.
2947 @xref{Attribute Syntax}, for details of the exact syntax for using
2951 @cindex @code{aligned} attribute
2952 @item aligned (@var{alignment})
2953 This attribute specifies a minimum alignment for the variable or
2954 structure field, measured in bytes. For example, the declaration:
2957 int x __attribute__ ((aligned (16))) = 0;
2961 causes the compiler to allocate the global variable @code{x} on a
2962 16-byte boundary. On a 68040, this could be used in conjunction with
2963 an @code{asm} expression to access the @code{move16} instruction which
2964 requires 16-byte aligned operands.
2966 You can also specify the alignment of structure fields. For example, to
2967 create a double-word aligned @code{int} pair, you could write:
2970 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2974 This is an alternative to creating a union with a @code{double} member
2975 that forces the union to be double-word aligned.
2977 As in the preceding examples, you can explicitly specify the alignment
2978 (in bytes) that you wish the compiler to use for a given variable or
2979 structure field. Alternatively, you can leave out the alignment factor
2980 and just ask the compiler to align a variable or field to the maximum
2981 useful alignment for the target machine you are compiling for. For
2982 example, you could write:
2985 short array[3] __attribute__ ((aligned));
2988 Whenever you leave out the alignment factor in an @code{aligned} attribute
2989 specification, the compiler automatically sets the alignment for the declared
2990 variable or field to the largest alignment which is ever used for any data
2991 type on the target machine you are compiling for. Doing this can often make
2992 copy operations more efficient, because the compiler can use whatever
2993 instructions copy the biggest chunks of memory when performing copies to
2994 or from the variables or fields that you have aligned this way.
2996 The @code{aligned} attribute can only increase the alignment; but you
2997 can decrease it by specifying @code{packed} as well. See below.
2999 Note that the effectiveness of @code{aligned} attributes may be limited
3000 by inherent limitations in your linker. On many systems, the linker is
3001 only able to arrange for variables to be aligned up to a certain maximum
3002 alignment. (For some linkers, the maximum supported alignment may
3003 be very very small.) If your linker is only able to align variables
3004 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3005 in an @code{__attribute__} will still only provide you with 8 byte
3006 alignment. See your linker documentation for further information.
3008 @item cleanup (@var{cleanup_function})
3009 @cindex @code{cleanup} attribute
3010 The @code{cleanup} attribute runs a function when the variable goes
3011 out of scope. This attribute can only be applied to auto function
3012 scope variables; it may not be applied to parameters or variables
3013 with static storage duration. The function must take one parameter,
3014 a pointer to a type compatible with the variable. The return value
3015 of the function (if any) is ignored.
3017 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3018 will be run during the stack unwinding that happens during the
3019 processing of the exception. Note that the @code{cleanup} attribute
3020 does not allow the exception to be caught, only to perform an action.
3021 It is undefined what happens if @var{cleanup_function} does not
3026 @cindex @code{common} attribute
3027 @cindex @code{nocommon} attribute
3030 The @code{common} attribute requests GCC to place a variable in
3031 ``common'' storage. The @code{nocommon} attribute requests the
3032 opposite---to allocate space for it directly.
3034 These attributes override the default chosen by the
3035 @option{-fno-common} and @option{-fcommon} flags respectively.
3038 @cindex @code{deprecated} attribute
3039 The @code{deprecated} attribute results in a warning if the variable
3040 is used anywhere in the source file. This is useful when identifying
3041 variables that are expected to be removed in a future version of a
3042 program. The warning also includes the location of the declaration
3043 of the deprecated variable, to enable users to easily find further
3044 information about why the variable is deprecated, or what they should
3045 do instead. Note that the warning only occurs for uses:
3048 extern int old_var __attribute__ ((deprecated));
3050 int new_fn () @{ return old_var; @}
3053 results in a warning on line 3 but not line 2.
3055 The @code{deprecated} attribute can also be used for functions and
3056 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3058 @item mode (@var{mode})
3059 @cindex @code{mode} attribute
3060 This attribute specifies the data type for the declaration---whichever
3061 type corresponds to the mode @var{mode}. This in effect lets you
3062 request an integer or floating point type according to its width.
3064 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3065 indicate the mode corresponding to a one-byte integer, @samp{word} or
3066 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3067 or @samp{__pointer__} for the mode used to represent pointers.
3070 @cindex @code{packed} attribute
3071 The @code{packed} attribute specifies that a variable or structure field
3072 should have the smallest possible alignment---one byte for a variable,
3073 and one bit for a field, unless you specify a larger value with the
3074 @code{aligned} attribute.
3076 Here is a structure in which the field @code{x} is packed, so that it
3077 immediately follows @code{a}:
3083 int x[2] __attribute__ ((packed));
3087 @item section ("@var{section-name}")
3088 @cindex @code{section} variable attribute
3089 Normally, the compiler places the objects it generates in sections like
3090 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3091 or you need certain particular variables to appear in special sections,
3092 for example to map to special hardware. The @code{section}
3093 attribute specifies that a variable (or function) lives in a particular
3094 section. For example, this small program uses several specific section names:
3097 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3098 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3099 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3100 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3104 /* @r{Initialize stack pointer} */
3105 init_sp (stack + sizeof (stack));
3107 /* @r{Initialize initialized data} */
3108 memcpy (&init_data, &data, &edata - &data);
3110 /* @r{Turn on the serial ports} */
3117 Use the @code{section} attribute with an @emph{initialized} definition
3118 of a @emph{global} variable, as shown in the example. GCC issues
3119 a warning and otherwise ignores the @code{section} attribute in
3120 uninitialized variable declarations.
3122 You may only use the @code{section} attribute with a fully initialized
3123 global definition because of the way linkers work. The linker requires
3124 each object be defined once, with the exception that uninitialized
3125 variables tentatively go in the @code{common} (or @code{bss}) section
3126 and can be multiply ``defined''. You can force a variable to be
3127 initialized with the @option{-fno-common} flag or the @code{nocommon}
3130 Some file formats do not support arbitrary sections so the @code{section}
3131 attribute is not available on all platforms.
3132 If you need to map the entire contents of a module to a particular
3133 section, consider using the facilities of the linker instead.
3136 @cindex @code{shared} variable attribute
3137 On Microsoft Windows, in addition to putting variable definitions in a named
3138 section, the section can also be shared among all running copies of an
3139 executable or DLL@. For example, this small program defines shared data
3140 by putting it in a named section @code{shared} and marking the section
3144 int foo __attribute__((section ("shared"), shared)) = 0;
3149 /* @r{Read and write foo. All running
3150 copies see the same value.} */
3156 You may only use the @code{shared} attribute along with @code{section}
3157 attribute with a fully initialized global definition because of the way
3158 linkers work. See @code{section} attribute for more information.
3160 The @code{shared} attribute is only available on Microsoft Windows@.
3162 @item tls_model ("@var{tls_model}")
3163 @cindex @code{tls_model} attribute
3164 The @code{tls_model} attribute sets thread-local storage model
3165 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3166 overriding @option{-ftls-model=} command line switch on a per-variable
3168 The @var{tls_model} argument should be one of @code{global-dynamic},
3169 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3171 Not all targets support this attribute.
3174 This attribute, attached to a variable, means that the variable is meant
3175 to be possibly unused. GCC will not produce a warning for this
3179 This attribute, attached to a variable, means that the variable must be
3180 emitted even if it appears that the variable is not referenced.
3182 @item vector_size (@var{bytes})
3183 This attribute specifies the vector size for the variable, measured in
3184 bytes. For example, the declaration:
3187 int foo __attribute__ ((vector_size (16)));
3191 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3192 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3193 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3195 This attribute is only applicable to integral and float scalars,
3196 although arrays, pointers, and function return values are allowed in
3197 conjunction with this construct.
3199 Aggregates with this attribute are invalid, even if they are of the same
3200 size as a corresponding scalar. For example, the declaration:
3203 struct S @{ int a; @};
3204 struct S __attribute__ ((vector_size (16))) foo;
3208 is invalid even if the size of the structure is the same as the size of
3212 The @code{selectany} attribute causes an initialized global variable to
3213 have link-once semantics. When multiple definitions of the variable are
3214 encountered by the linker, the first is selected and the remainder are
3215 discarded. Following usage by the Microsoft compiler, the linker is told
3216 @emph{not} to warn about size or content differences of the multiple
3219 Although the primary usage of this attribute is for POD types, the
3220 attribute can also be applied to global C++ objects that are initialized
3221 by a constructor. In this case, the static initialization and destruction
3222 code for the object is emitted in each translation defining the object,
3223 but the calls to the constructor and destructor are protected by a
3224 link-once guard variable.
3226 The @code{selectany} attribute is only available on Microsoft Windows
3227 targets. You can use @code{__declspec (selectany)} as a synonym for
3228 @code{__attribute__ ((selectany))} for compatibility with other
3232 The @code{weak} attribute is described in @xref{Function Attributes}.
3235 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3238 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3242 @subsection M32R/D Variable Attributes
3244 One attribute is currently defined for the M32R/D@.
3247 @item model (@var{model-name})
3248 @cindex variable addressability on the M32R/D
3249 Use this attribute on the M32R/D to set the addressability of an object.
3250 The identifier @var{model-name} is one of @code{small}, @code{medium},
3251 or @code{large}, representing each of the code models.
3253 Small model objects live in the lower 16MB of memory (so that their
3254 addresses can be loaded with the @code{ld24} instruction).
3256 Medium and large model objects may live anywhere in the 32-bit address space
3257 (the compiler will generate @code{seth/add3} instructions to load their
3261 @anchor{i386 Variable Attributes}
3262 @subsection i386 Variable Attributes
3264 Two attributes are currently defined for i386 configurations:
3265 @code{ms_struct} and @code{gcc_struct}
3270 @cindex @code{ms_struct} attribute
3271 @cindex @code{gcc_struct} attribute
3273 If @code{packed} is used on a structure, or if bit-fields are used
3274 it may be that the Microsoft ABI packs them differently
3275 than GCC would normally pack them. Particularly when moving packed
3276 data between functions compiled with GCC and the native Microsoft compiler
3277 (either via function call or as data in a file), it may be necessary to access
3280 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3281 compilers to match the native Microsoft compiler.
3283 The Microsoft structure layout algorithm is fairly simple with the exception
3284 of the bitfield packing:
3286 The padding and alignment of members of structures and whether a bit field
3287 can straddle a storage-unit boundary
3290 @item Structure members are stored sequentially in the order in which they are
3291 declared: the first member has the lowest memory address and the last member
3294 @item Every data object has an alignment-requirement. The alignment-requirement
3295 for all data except structures, unions, and arrays is either the size of the
3296 object or the current packing size (specified with either the aligned attribute
3297 or the pack pragma), whichever is less. For structures, unions, and arrays,
3298 the alignment-requirement is the largest alignment-requirement of its members.
3299 Every object is allocated an offset so that:
3301 offset % alignment-requirement == 0
3303 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3304 unit if the integral types are the same size and if the next bit field fits
3305 into the current allocation unit without crossing the boundary imposed by the
3306 common alignment requirements of the bit fields.
3309 Handling of zero-length bitfields:
3311 MSVC interprets zero-length bitfields in the following ways:
3314 @item If a zero-length bitfield is inserted between two bitfields that would
3315 normally be coalesced, the bitfields will not be coalesced.
3322 unsigned long bf_1 : 12;
3324 unsigned long bf_2 : 12;
3328 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3329 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3331 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3332 alignment of the zero-length bitfield is greater than the member that follows it,
3333 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3353 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3354 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3355 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3358 Taking this into account, it is important to note the following:
3361 @item If a zero-length bitfield follows a normal bitfield, the type of the
3362 zero-length bitfield may affect the alignment of the structure as whole. For
3363 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3364 normal bitfield, and is of type short.
3366 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3367 still affect the alignment of the structure:
3377 Here, @code{t4} will take up 4 bytes.
3380 @item Zero-length bitfields following non-bitfield members are ignored:
3391 Here, @code{t5} will take up 2 bytes.
3395 @subsection PowerPC Variable Attributes
3397 Three attributes currently are defined for PowerPC configurations:
3398 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3400 For full documentation of the struct attributes please see the
3401 documentation in the @xref{i386 Variable Attributes}, section.
3403 For documentation of @code{altivec} attribute please see the
3404 documentation in the @xref{PowerPC Type Attributes}, section.
3406 @subsection Xstormy16 Variable Attributes
3408 One attribute is currently defined for xstormy16 configurations:
3413 @cindex @code{below100} attribute
3415 If a variable has the @code{below100} attribute (@code{BELOW100} is
3416 allowed also), GCC will place the variable in the first 0x100 bytes of
3417 memory and use special opcodes to access it. Such variables will be
3418 placed in either the @code{.bss_below100} section or the
3419 @code{.data_below100} section.
3423 @node Type Attributes
3424 @section Specifying Attributes of Types
3425 @cindex attribute of types
3426 @cindex type attributes
3428 The keyword @code{__attribute__} allows you to specify special
3429 attributes of @code{struct} and @code{union} types when you define
3430 such types. This keyword is followed by an attribute specification
3431 inside double parentheses. Seven attributes are currently defined for
3432 types: @code{aligned}, @code{packed}, @code{transparent_union},
3433 @code{unused}, @code{deprecated}, @code{visibility}, and
3434 @code{may_alias}. Other attributes are defined for functions
3435 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3438 You may also specify any one of these attributes with @samp{__}
3439 preceding and following its keyword. This allows you to use these
3440 attributes in header files without being concerned about a possible
3441 macro of the same name. For example, you may use @code{__aligned__}
3442 instead of @code{aligned}.
3444 You may specify type attributes either in a @code{typedef} declaration
3445 or in an enum, struct or union type declaration or definition.
3447 For an enum, struct or union type, you may specify attributes either
3448 between the enum, struct or union tag and the name of the type, or
3449 just past the closing curly brace of the @emph{definition}. The
3450 former syntax is preferred.
3452 @xref{Attribute Syntax}, for details of the exact syntax for using
3456 @cindex @code{aligned} attribute
3457 @item aligned (@var{alignment})
3458 This attribute specifies a minimum alignment (in bytes) for variables
3459 of the specified type. For example, the declarations:
3462 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3463 typedef int more_aligned_int __attribute__ ((aligned (8)));
3467 force the compiler to insure (as far as it can) that each variable whose
3468 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3469 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3470 variables of type @code{struct S} aligned to 8-byte boundaries allows
3471 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3472 store) instructions when copying one variable of type @code{struct S} to
3473 another, thus improving run-time efficiency.
3475 Note that the alignment of any given @code{struct} or @code{union} type
3476 is required by the ISO C standard to be at least a perfect multiple of
3477 the lowest common multiple of the alignments of all of the members of
3478 the @code{struct} or @code{union} in question. This means that you @emph{can}
3479 effectively adjust the alignment of a @code{struct} or @code{union}
3480 type by attaching an @code{aligned} attribute to any one of the members
3481 of such a type, but the notation illustrated in the example above is a
3482 more obvious, intuitive, and readable way to request the compiler to
3483 adjust the alignment of an entire @code{struct} or @code{union} type.
3485 As in the preceding example, you can explicitly specify the alignment
3486 (in bytes) that you wish the compiler to use for a given @code{struct}
3487 or @code{union} type. Alternatively, you can leave out the alignment factor
3488 and just ask the compiler to align a type to the maximum
3489 useful alignment for the target machine you are compiling for. For
3490 example, you could write:
3493 struct S @{ short f[3]; @} __attribute__ ((aligned));
3496 Whenever you leave out the alignment factor in an @code{aligned}
3497 attribute specification, the compiler automatically sets the alignment
3498 for the type to the largest alignment which is ever used for any data
3499 type on the target machine you are compiling for. Doing this can often
3500 make copy operations more efficient, because the compiler can use
3501 whatever instructions copy the biggest chunks of memory when performing
3502 copies to or from the variables which have types that you have aligned
3505 In the example above, if the size of each @code{short} is 2 bytes, then
3506 the size of the entire @code{struct S} type is 6 bytes. The smallest
3507 power of two which is greater than or equal to that is 8, so the
3508 compiler sets the alignment for the entire @code{struct S} type to 8
3511 Note that although you can ask the compiler to select a time-efficient
3512 alignment for a given type and then declare only individual stand-alone
3513 objects of that type, the compiler's ability to select a time-efficient
3514 alignment is primarily useful only when you plan to create arrays of
3515 variables having the relevant (efficiently aligned) type. If you
3516 declare or use arrays of variables of an efficiently-aligned type, then
3517 it is likely that your program will also be doing pointer arithmetic (or
3518 subscripting, which amounts to the same thing) on pointers to the
3519 relevant type, and the code that the compiler generates for these
3520 pointer arithmetic operations will often be more efficient for
3521 efficiently-aligned types than for other types.
3523 The @code{aligned} attribute can only increase the alignment; but you
3524 can decrease it by specifying @code{packed} as well. See below.
3526 Note that the effectiveness of @code{aligned} attributes may be limited
3527 by inherent limitations in your linker. On many systems, the linker is
3528 only able to arrange for variables to be aligned up to a certain maximum
3529 alignment. (For some linkers, the maximum supported alignment may
3530 be very very small.) If your linker is only able to align variables
3531 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3532 in an @code{__attribute__} will still only provide you with 8 byte
3533 alignment. See your linker documentation for further information.
3536 This attribute, attached to @code{struct} or @code{union} type
3537 definition, specifies that each member (other than zero-width bitfields)
3538 of the structure or union is placed to minimize the memory required. When
3539 attached to an @code{enum} definition, it indicates that the smallest
3540 integral type should be used.
3542 @opindex fshort-enums
3543 Specifying this attribute for @code{struct} and @code{union} types is
3544 equivalent to specifying the @code{packed} attribute on each of the
3545 structure or union members. Specifying the @option{-fshort-enums}
3546 flag on the line is equivalent to specifying the @code{packed}
3547 attribute on all @code{enum} definitions.
3549 In the following example @code{struct my_packed_struct}'s members are
3550 packed closely together, but the internal layout of its @code{s} member
3551 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3555 struct my_unpacked_struct
3561 struct __attribute__ ((__packed__)) my_packed_struct
3565 struct my_unpacked_struct s;
3569 You may only specify this attribute on the definition of a @code{enum},
3570 @code{struct} or @code{union}, not on a @code{typedef} which does not
3571 also define the enumerated type, structure or union.
3573 @item transparent_union
3574 This attribute, attached to a @code{union} type definition, indicates
3575 that any function parameter having that union type causes calls to that
3576 function to be treated in a special way.
3578 First, the argument corresponding to a transparent union type can be of
3579 any type in the union; no cast is required. Also, if the union contains
3580 a pointer type, the corresponding argument can be a null pointer
3581 constant or a void pointer expression; and if the union contains a void
3582 pointer type, the corresponding argument can be any pointer expression.
3583 If the union member type is a pointer, qualifiers like @code{const} on
3584 the referenced type must be respected, just as with normal pointer
3587 Second, the argument is passed to the function using the calling
3588 conventions of the first member of the transparent union, not the calling
3589 conventions of the union itself. All members of the union must have the
3590 same machine representation; this is necessary for this argument passing
3593 Transparent unions are designed for library functions that have multiple
3594 interfaces for compatibility reasons. For example, suppose the
3595 @code{wait} function must accept either a value of type @code{int *} to
3596 comply with Posix, or a value of type @code{union wait *} to comply with
3597 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3598 @code{wait} would accept both kinds of arguments, but it would also
3599 accept any other pointer type and this would make argument type checking
3600 less useful. Instead, @code{<sys/wait.h>} might define the interface
3608 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3610 pid_t wait (wait_status_ptr_t);
3613 This interface allows either @code{int *} or @code{union wait *}
3614 arguments to be passed, using the @code{int *} calling convention.
3615 The program can call @code{wait} with arguments of either type:
3618 int w1 () @{ int w; return wait (&w); @}
3619 int w2 () @{ union wait w; return wait (&w); @}
3622 With this interface, @code{wait}'s implementation might look like this:
3625 pid_t wait (wait_status_ptr_t p)
3627 return waitpid (-1, p.__ip, 0);
3632 When attached to a type (including a @code{union} or a @code{struct}),
3633 this attribute means that variables of that type are meant to appear
3634 possibly unused. GCC will not produce a warning for any variables of
3635 that type, even if the variable appears to do nothing. This is often
3636 the case with lock or thread classes, which are usually defined and then
3637 not referenced, but contain constructors and destructors that have
3638 nontrivial bookkeeping functions.
3641 The @code{deprecated} attribute results in a warning if the type
3642 is used anywhere in the source file. This is useful when identifying
3643 types that are expected to be removed in a future version of a program.
3644 If possible, the warning also includes the location of the declaration
3645 of the deprecated type, to enable users to easily find further
3646 information about why the type is deprecated, or what they should do
3647 instead. Note that the warnings only occur for uses and then only
3648 if the type is being applied to an identifier that itself is not being
3649 declared as deprecated.
3652 typedef int T1 __attribute__ ((deprecated));
3656 typedef T1 T3 __attribute__ ((deprecated));
3657 T3 z __attribute__ ((deprecated));
3660 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3661 warning is issued for line 4 because T2 is not explicitly
3662 deprecated. Line 5 has no warning because T3 is explicitly
3663 deprecated. Similarly for line 6.
3665 The @code{deprecated} attribute can also be used for functions and
3666 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3669 Accesses to objects with types with this attribute are not subjected to
3670 type-based alias analysis, but are instead assumed to be able to alias
3671 any other type of objects, just like the @code{char} type. See
3672 @option{-fstrict-aliasing} for more information on aliasing issues.
3677 typedef short __attribute__((__may_alias__)) short_a;
3683 short_a *b = (short_a *) &a;
3687 if (a == 0x12345678)
3694 If you replaced @code{short_a} with @code{short} in the variable
3695 declaration, the above program would abort when compiled with
3696 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3697 above in recent GCC versions.
3700 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3701 applied to class, struct, union and enum types. Unlike other type
3702 attributes, the attribute must appear between the initial keyword and
3703 the name of the type; it cannot appear after the body of the type.
3705 Note that the type visibility is applied to vague linkage entities
3706 associated with the class (vtable, typeinfo node, etc.). In
3707 particular, if a class is thrown as an exception in one shared object
3708 and caught in another, the class must have default visibility.
3709 Otherwise the two shared objects will be unable to use the same
3710 typeinfo node and exception handling will break.
3712 @subsection ARM Type Attributes
3714 On those ARM targets that support @code{dllimport} (such as Symbian
3715 OS), you can use the @code{notshared} attribute to indicate that the
3716 virtual table and other similar data for a class should not be
3717 exported from a DLL@. For example:
3720 class __declspec(notshared) C @{
3722 __declspec(dllimport) C();
3726 __declspec(dllexport)
3730 In this code, @code{C::C} is exported from the current DLL, but the
3731 virtual table for @code{C} is not exported. (You can use
3732 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3733 most Symbian OS code uses @code{__declspec}.)
3735 @anchor{i386 Type Attributes}
3736 @subsection i386 Type Attributes
3738 Two attributes are currently defined for i386 configurations:
3739 @code{ms_struct} and @code{gcc_struct}
3743 @cindex @code{ms_struct}
3744 @cindex @code{gcc_struct}
3746 If @code{packed} is used on a structure, or if bit-fields are used
3747 it may be that the Microsoft ABI packs them differently
3748 than GCC would normally pack them. Particularly when moving packed
3749 data between functions compiled with GCC and the native Microsoft compiler
3750 (either via function call or as data in a file), it may be necessary to access
3753 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3754 compilers to match the native Microsoft compiler.
3757 To specify multiple attributes, separate them by commas within the
3758 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3761 @anchor{PowerPC Type Attributes}
3762 @subsection PowerPC Type Attributes
3764 Three attributes currently are defined for PowerPC configurations:
3765 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3767 For full documentation of the struct attributes please see the
3768 documentation in the @xref{i386 Type Attributes}, section.
3770 The @code{altivec} attribute allows one to declare AltiVec vector data
3771 types supported by the AltiVec Programming Interface Manual. The
3772 attribute requires an argument to specify one of three vector types:
3773 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3774 and @code{bool__} (always followed by unsigned).
3777 __attribute__((altivec(vector__)))
3778 __attribute__((altivec(pixel__))) unsigned short
3779 __attribute__((altivec(bool__))) unsigned
3782 These attributes mainly are intended to support the @code{__vector},
3783 @code{__pixel}, and @code{__bool} AltiVec keywords.
3786 @section An Inline Function is As Fast As a Macro
3787 @cindex inline functions
3788 @cindex integrating function code
3790 @cindex macros, inline alternative
3792 By declaring a function @code{inline}, you can direct GCC to
3793 integrate that function's code into the code for its callers. This
3794 makes execution faster by eliminating the function-call overhead; in
3795 addition, if any of the actual argument values are constant, their known
3796 values may permit simplifications at compile time so that not all of the
3797 inline function's code needs to be included. The effect on code size is
3798 less predictable; object code may be larger or smaller with function
3799 inlining, depending on the particular case. Inlining of functions is an
3800 optimization and it really ``works'' only in optimizing compilation. If
3801 you don't use @option{-O}, no function is really inline.
3803 Inline functions are included in the ISO C99 standard, but there are
3804 currently substantial differences between what GCC implements and what
3805 the ISO C99 standard requires.
3807 To declare a function inline, use the @code{inline} keyword in its
3808 declaration, like this:
3818 (If you are writing a header file to be included in ISO C programs, write
3819 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3820 You can also make all ``simple enough'' functions inline with the option
3821 @option{-finline-functions}.
3824 Note that certain usages in a function definition can make it unsuitable
3825 for inline substitution. Among these usages are: use of varargs, use of
3826 alloca, use of variable sized data types (@pxref{Variable Length}),
3827 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3828 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3829 will warn when a function marked @code{inline} could not be substituted,
3830 and will give the reason for the failure.
3832 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3833 does not affect the linkage of the function.
3835 @cindex automatic @code{inline} for C++ member fns
3836 @cindex @code{inline} automatic for C++ member fns
3837 @cindex member fns, automatically @code{inline}
3838 @cindex C++ member fns, automatically @code{inline}
3839 @opindex fno-default-inline
3840 GCC automatically inlines member functions defined within the class
3841 body of C++ programs even if they are not explicitly declared
3842 @code{inline}. (You can override this with @option{-fno-default-inline};
3843 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3845 @cindex inline functions, omission of
3846 @opindex fkeep-inline-functions
3847 When a function is both inline and @code{static}, if all calls to the
3848 function are integrated into the caller, and the function's address is
3849 never used, then the function's own assembler code is never referenced.
3850 In this case, GCC does not actually output assembler code for the
3851 function, unless you specify the option @option{-fkeep-inline-functions}.
3852 Some calls cannot be integrated for various reasons (in particular,
3853 calls that precede the function's definition cannot be integrated, and
3854 neither can recursive calls within the definition). If there is a
3855 nonintegrated call, then the function is compiled to assembler code as
3856 usual. The function must also be compiled as usual if the program
3857 refers to its address, because that can't be inlined.
3859 @cindex non-static inline function
3860 When an inline function is not @code{static}, then the compiler must assume
3861 that there may be calls from other source files; since a global symbol can
3862 be defined only once in any program, the function must not be defined in
3863 the other source files, so the calls therein cannot be integrated.
3864 Therefore, a non-@code{static} inline function is always compiled on its
3865 own in the usual fashion.
3867 If you specify both @code{inline} and @code{extern} in the function
3868 definition, then the definition is used only for inlining. In no case
3869 is the function compiled on its own, not even if you refer to its
3870 address explicitly. Such an address becomes an external reference, as
3871 if you had only declared the function, and had not defined it.
3873 This combination of @code{inline} and @code{extern} has almost the
3874 effect of a macro. The way to use it is to put a function definition in
3875 a header file with these keywords, and put another copy of the
3876 definition (lacking @code{inline} and @code{extern}) in a library file.
3877 The definition in the header file will cause most calls to the function
3878 to be inlined. If any uses of the function remain, they will refer to
3879 the single copy in the library.
3881 Since GCC eventually will implement ISO C99 semantics for
3882 inline functions, it is best to use @code{static inline} only
3883 to guarantee compatibility. (The
3884 existing semantics will remain available when @option{-std=gnu89} is
3885 specified, but eventually the default will be @option{-std=gnu99} and
3886 that will implement the C99 semantics, though it does not do so yet.)
3888 GCC does not inline any functions when not optimizing unless you specify
3889 the @samp{always_inline} attribute for the function, like this:
3892 /* @r{Prototype.} */
3893 inline void foo (const char) __attribute__((always_inline));
3897 @section Assembler Instructions with C Expression Operands
3898 @cindex extended @code{asm}
3899 @cindex @code{asm} expressions
3900 @cindex assembler instructions
3903 In an assembler instruction using @code{asm}, you can specify the
3904 operands of the instruction using C expressions. This means you need not
3905 guess which registers or memory locations will contain the data you want
3908 You must specify an assembler instruction template much like what
3909 appears in a machine description, plus an operand constraint string for
3912 For example, here is how to use the 68881's @code{fsinx} instruction:
3915 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3919 Here @code{angle} is the C expression for the input operand while
3920 @code{result} is that of the output operand. Each has @samp{"f"} as its
3921 operand constraint, saying that a floating point register is required.
3922 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3923 output operands' constraints must use @samp{=}. The constraints use the
3924 same language used in the machine description (@pxref{Constraints}).
3926 Each operand is described by an operand-constraint string followed by
3927 the C expression in parentheses. A colon separates the assembler
3928 template from the first output operand and another separates the last
3929 output operand from the first input, if any. Commas separate the
3930 operands within each group. The total number of operands is currently
3931 limited to 30; this limitation may be lifted in some future version of
3934 If there are no output operands but there are input operands, you must
3935 place two consecutive colons surrounding the place where the output
3938 As of GCC version 3.1, it is also possible to specify input and output
3939 operands using symbolic names which can be referenced within the
3940 assembler code. These names are specified inside square brackets
3941 preceding the constraint string, and can be referenced inside the
3942 assembler code using @code{%[@var{name}]} instead of a percentage sign
3943 followed by the operand number. Using named operands the above example
3947 asm ("fsinx %[angle],%[output]"
3948 : [output] "=f" (result)
3949 : [angle] "f" (angle));
3953 Note that the symbolic operand names have no relation whatsoever to
3954 other C identifiers. You may use any name you like, even those of
3955 existing C symbols, but you must ensure that no two operands within the same
3956 assembler construct use the same symbolic name.
3958 Output operand expressions must be lvalues; the compiler can check this.
3959 The input operands need not be lvalues. The compiler cannot check
3960 whether the operands have data types that are reasonable for the
3961 instruction being executed. It does not parse the assembler instruction
3962 template and does not know what it means or even whether it is valid
3963 assembler input. The extended @code{asm} feature is most often used for
3964 machine instructions the compiler itself does not know exist. If
3965 the output expression cannot be directly addressed (for example, it is a
3966 bit-field), your constraint must allow a register. In that case, GCC
3967 will use the register as the output of the @code{asm}, and then store
3968 that register into the output.
3970 The ordinary output operands must be write-only; GCC will assume that
3971 the values in these operands before the instruction are dead and need
3972 not be generated. Extended asm supports input-output or read-write
3973 operands. Use the constraint character @samp{+} to indicate such an
3974 operand and list it with the output operands. You should only use
3975 read-write operands when the constraints for the operand (or the
3976 operand in which only some of the bits are to be changed) allow a
3979 You may, as an alternative, logically split its function into two
3980 separate operands, one input operand and one write-only output
3981 operand. The connection between them is expressed by constraints
3982 which say they need to be in the same location when the instruction
3983 executes. You can use the same C expression for both operands, or
3984 different expressions. For example, here we write the (fictitious)
3985 @samp{combine} instruction with @code{bar} as its read-only source
3986 operand and @code{foo} as its read-write destination:
3989 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3993 The constraint @samp{"0"} for operand 1 says that it must occupy the
3994 same location as operand 0. A number in constraint is allowed only in
3995 an input operand and it must refer to an output operand.
3997 Only a number in the constraint can guarantee that one operand will be in
3998 the same place as another. The mere fact that @code{foo} is the value
3999 of both operands is not enough to guarantee that they will be in the
4000 same place in the generated assembler code. The following would not
4004 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4007 Various optimizations or reloading could cause operands 0 and 1 to be in
4008 different registers; GCC knows no reason not to do so. For example, the
4009 compiler might find a copy of the value of @code{foo} in one register and
4010 use it for operand 1, but generate the output operand 0 in a different
4011 register (copying it afterward to @code{foo}'s own address). Of course,
4012 since the register for operand 1 is not even mentioned in the assembler
4013 code, the result will not work, but GCC can't tell that.
4015 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4016 the operand number for a matching constraint. For example:
4019 asm ("cmoveq %1,%2,%[result]"
4020 : [result] "=r"(result)
4021 : "r" (test), "r"(new), "[result]"(old));
4024 Sometimes you need to make an @code{asm} operand be a specific register,
4025 but there's no matching constraint letter for that register @emph{by
4026 itself}. To force the operand into that register, use a local variable
4027 for the operand and specify the register in the variable declaration.
4028 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4029 register constraint letter that matches the register:
4032 register int *p1 asm ("r0") = @dots{};
4033 register int *p2 asm ("r1") = @dots{};
4034 register int *result asm ("r0");
4035 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4038 @anchor{Example of asm with clobbered asm reg}
4039 In the above example, beware that a register that is call-clobbered by
4040 the target ABI will be overwritten by any function call in the
4041 assignment, including library calls for arithmetic operators.
4042 Assuming it is a call-clobbered register, this may happen to @code{r0}
4043 above by the assignment to @code{p2}. If you have to use such a
4044 register, use temporary variables for expressions between the register
4049 register int *p1 asm ("r0") = @dots{};
4050 register int *p2 asm ("r1") = t1;
4051 register int *result asm ("r0");
4052 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4055 Some instructions clobber specific hard registers. To describe this,
4056 write a third colon after the input operands, followed by the names of
4057 the clobbered hard registers (given as strings). Here is a realistic
4058 example for the VAX:
4061 asm volatile ("movc3 %0,%1,%2"
4062 : /* @r{no outputs} */
4063 : "g" (from), "g" (to), "g" (count)
4064 : "r0", "r1", "r2", "r3", "r4", "r5");
4067 You may not write a clobber description in a way that overlaps with an
4068 input or output operand. For example, you may not have an operand
4069 describing a register class with one member if you mention that register
4070 in the clobber list. Variables declared to live in specific registers
4071 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4072 have no part mentioned in the clobber description.
4073 There is no way for you to specify that an input
4074 operand is modified without also specifying it as an output
4075 operand. Note that if all the output operands you specify are for this
4076 purpose (and hence unused), you will then also need to specify
4077 @code{volatile} for the @code{asm} construct, as described below, to
4078 prevent GCC from deleting the @code{asm} statement as unused.
4080 If you refer to a particular hardware register from the assembler code,
4081 you will probably have to list the register after the third colon to
4082 tell the compiler the register's value is modified. In some assemblers,
4083 the register names begin with @samp{%}; to produce one @samp{%} in the
4084 assembler code, you must write @samp{%%} in the input.
4086 If your assembler instruction can alter the condition code register, add
4087 @samp{cc} to the list of clobbered registers. GCC on some machines
4088 represents the condition codes as a specific hardware register;
4089 @samp{cc} serves to name this register. On other machines, the
4090 condition code is handled differently, and specifying @samp{cc} has no
4091 effect. But it is valid no matter what the machine.
4093 If your assembler instructions access memory in an unpredictable
4094 fashion, add @samp{memory} to the list of clobbered registers. This
4095 will cause GCC to not keep memory values cached in registers across the
4096 assembler instruction and not optimize stores or loads to that memory.
4097 You will also want to add the @code{volatile} keyword if the memory
4098 affected is not listed in the inputs or outputs of the @code{asm}, as
4099 the @samp{memory} clobber does not count as a side-effect of the
4100 @code{asm}. If you know how large the accessed memory is, you can add
4101 it as input or output but if this is not known, you should add
4102 @samp{memory}. As an example, if you access ten bytes of a string, you
4103 can use a memory input like:
4106 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4109 Note that in the following example the memory input is necessary,
4110 otherwise GCC might optimize the store to @code{x} away:
4117 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4118 "=&d" (r) : "a" (y), "m" (*y));
4123 You can put multiple assembler instructions together in a single
4124 @code{asm} template, separated by the characters normally used in assembly
4125 code for the system. A combination that works in most places is a newline
4126 to break the line, plus a tab character to move to the instruction field
4127 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4128 assembler allows semicolons as a line-breaking character. Note that some
4129 assembler dialects use semicolons to start a comment.
4130 The input operands are guaranteed not to use any of the clobbered
4131 registers, and neither will the output operands' addresses, so you can
4132 read and write the clobbered registers as many times as you like. Here
4133 is an example of multiple instructions in a template; it assumes the
4134 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4137 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4139 : "g" (from), "g" (to)
4143 Unless an output operand has the @samp{&} constraint modifier, GCC
4144 may allocate it in the same register as an unrelated input operand, on
4145 the assumption the inputs are consumed before the outputs are produced.
4146 This assumption may be false if the assembler code actually consists of
4147 more than one instruction. In such a case, use @samp{&} for each output
4148 operand that may not overlap an input. @xref{Modifiers}.
4150 If you want to test the condition code produced by an assembler
4151 instruction, you must include a branch and a label in the @code{asm}
4152 construct, as follows:
4155 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4161 This assumes your assembler supports local labels, as the GNU assembler
4162 and most Unix assemblers do.
4164 Speaking of labels, jumps from one @code{asm} to another are not
4165 supported. The compiler's optimizers do not know about these jumps, and
4166 therefore they cannot take account of them when deciding how to
4169 @cindex macros containing @code{asm}
4170 Usually the most convenient way to use these @code{asm} instructions is to
4171 encapsulate them in macros that look like functions. For example,
4175 (@{ double __value, __arg = (x); \
4176 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4181 Here the variable @code{__arg} is used to make sure that the instruction
4182 operates on a proper @code{double} value, and to accept only those
4183 arguments @code{x} which can convert automatically to a @code{double}.
4185 Another way to make sure the instruction operates on the correct data
4186 type is to use a cast in the @code{asm}. This is different from using a
4187 variable @code{__arg} in that it converts more different types. For
4188 example, if the desired type were @code{int}, casting the argument to
4189 @code{int} would accept a pointer with no complaint, while assigning the
4190 argument to an @code{int} variable named @code{__arg} would warn about
4191 using a pointer unless the caller explicitly casts it.
4193 If an @code{asm} has output operands, GCC assumes for optimization
4194 purposes the instruction has no side effects except to change the output
4195 operands. This does not mean instructions with a side effect cannot be
4196 used, but you must be careful, because the compiler may eliminate them
4197 if the output operands aren't used, or move them out of loops, or
4198 replace two with one if they constitute a common subexpression. Also,
4199 if your instruction does have a side effect on a variable that otherwise
4200 appears not to change, the old value of the variable may be reused later
4201 if it happens to be found in a register.
4203 You can prevent an @code{asm} instruction from being deleted
4204 by writing the keyword @code{volatile} after
4205 the @code{asm}. For example:
4208 #define get_and_set_priority(new) \
4210 asm volatile ("get_and_set_priority %0, %1" \
4211 : "=g" (__old) : "g" (new)); \
4216 The @code{volatile} keyword indicates that the instruction has
4217 important side-effects. GCC will not delete a volatile @code{asm} if
4218 it is reachable. (The instruction can still be deleted if GCC can
4219 prove that control-flow will never reach the location of the
4220 instruction.) Note that even a volatile @code{asm} instruction
4221 can be moved relative to other code, including across jump
4222 instructions. For example, on many targets there is a system
4223 register which can be set to control the rounding mode of
4224 floating point operations. You might try
4225 setting it with a volatile @code{asm}, like this PowerPC example:
4228 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4233 This will not work reliably, as the compiler may move the addition back
4234 before the volatile @code{asm}. To make it work you need to add an
4235 artificial dependency to the @code{asm} referencing a variable in the code
4236 you don't want moved, for example:
4239 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4243 Similarly, you can't expect a
4244 sequence of volatile @code{asm} instructions to remain perfectly
4245 consecutive. If you want consecutive output, use a single @code{asm}.
4246 Also, GCC will perform some optimizations across a volatile @code{asm}
4247 instruction; GCC does not ``forget everything'' when it encounters
4248 a volatile @code{asm} instruction the way some other compilers do.
4250 An @code{asm} instruction without any output operands will be treated
4251 identically to a volatile @code{asm} instruction.
4253 It is a natural idea to look for a way to give access to the condition
4254 code left by the assembler instruction. However, when we attempted to
4255 implement this, we found no way to make it work reliably. The problem
4256 is that output operands might need reloading, which would result in
4257 additional following ``store'' instructions. On most machines, these
4258 instructions would alter the condition code before there was time to
4259 test it. This problem doesn't arise for ordinary ``test'' and
4260 ``compare'' instructions because they don't have any output operands.
4262 For reasons similar to those described above, it is not possible to give
4263 an assembler instruction access to the condition code left by previous
4266 If you are writing a header file that should be includable in ISO C
4267 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4270 @subsection Size of an @code{asm}
4272 Some targets require that GCC track the size of each instruction used in
4273 order to generate correct code. Because the final length of an
4274 @code{asm} is only known by the assembler, GCC must make an estimate as
4275 to how big it will be. The estimate is formed by counting the number of
4276 statements in the pattern of the @code{asm} and multiplying that by the
4277 length of the longest instruction on that processor. Statements in the
4278 @code{asm} are identified by newline characters and whatever statement
4279 separator characters are supported by the assembler; on most processors
4280 this is the `@code{;}' character.
4282 Normally, GCC's estimate is perfectly adequate to ensure that correct
4283 code is generated, but it is possible to confuse the compiler if you use
4284 pseudo instructions or assembler macros that expand into multiple real
4285 instructions or if you use assembler directives that expand to more
4286 space in the object file than would be needed for a single instruction.
4287 If this happens then the assembler will produce a diagnostic saying that
4288 a label is unreachable.
4290 @subsection i386 floating point asm operands
4292 There are several rules on the usage of stack-like regs in
4293 asm_operands insns. These rules apply only to the operands that are
4298 Given a set of input regs that die in an asm_operands, it is
4299 necessary to know which are implicitly popped by the asm, and
4300 which must be explicitly popped by gcc.
4302 An input reg that is implicitly popped by the asm must be
4303 explicitly clobbered, unless it is constrained to match an
4307 For any input reg that is implicitly popped by an asm, it is
4308 necessary to know how to adjust the stack to compensate for the pop.
4309 If any non-popped input is closer to the top of the reg-stack than
4310 the implicitly popped reg, it would not be possible to know what the
4311 stack looked like---it's not clear how the rest of the stack ``slides
4314 All implicitly popped input regs must be closer to the top of
4315 the reg-stack than any input that is not implicitly popped.
4317 It is possible that if an input dies in an insn, reload might
4318 use the input reg for an output reload. Consider this example:
4321 asm ("foo" : "=t" (a) : "f" (b));
4324 This asm says that input B is not popped by the asm, and that
4325 the asm pushes a result onto the reg-stack, i.e., the stack is one
4326 deeper after the asm than it was before. But, it is possible that
4327 reload will think that it can use the same reg for both the input and
4328 the output, if input B dies in this insn.
4330 If any input operand uses the @code{f} constraint, all output reg
4331 constraints must use the @code{&} earlyclobber.
4333 The asm above would be written as
4336 asm ("foo" : "=&t" (a) : "f" (b));
4340 Some operands need to be in particular places on the stack. All
4341 output operands fall in this category---there is no other way to
4342 know which regs the outputs appear in unless the user indicates
4343 this in the constraints.
4345 Output operands must specifically indicate which reg an output
4346 appears in after an asm. @code{=f} is not allowed: the operand
4347 constraints must select a class with a single reg.
4350 Output operands may not be ``inserted'' between existing stack regs.
4351 Since no 387 opcode uses a read/write operand, all output operands
4352 are dead before the asm_operands, and are pushed by the asm_operands.
4353 It makes no sense to push anywhere but the top of the reg-stack.
4355 Output operands must start at the top of the reg-stack: output
4356 operands may not ``skip'' a reg.
4359 Some asm statements may need extra stack space for internal
4360 calculations. This can be guaranteed by clobbering stack registers
4361 unrelated to the inputs and outputs.
4365 Here are a couple of reasonable asms to want to write. This asm
4366 takes one input, which is internally popped, and produces two outputs.
4369 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4372 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4373 and replaces them with one output. The user must code the @code{st(1)}
4374 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4377 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4383 @section Controlling Names Used in Assembler Code
4384 @cindex assembler names for identifiers
4385 @cindex names used in assembler code
4386 @cindex identifiers, names in assembler code
4388 You can specify the name to be used in the assembler code for a C
4389 function or variable by writing the @code{asm} (or @code{__asm__})
4390 keyword after the declarator as follows:
4393 int foo asm ("myfoo") = 2;
4397 This specifies that the name to be used for the variable @code{foo} in
4398 the assembler code should be @samp{myfoo} rather than the usual
4401 On systems where an underscore is normally prepended to the name of a C
4402 function or variable, this feature allows you to define names for the
4403 linker that do not start with an underscore.
4405 It does not make sense to use this feature with a non-static local
4406 variable since such variables do not have assembler names. If you are
4407 trying to put the variable in a particular register, see @ref{Explicit
4408 Reg Vars}. GCC presently accepts such code with a warning, but will
4409 probably be changed to issue an error, rather than a warning, in the
4412 You cannot use @code{asm} in this way in a function @emph{definition}; but
4413 you can get the same effect by writing a declaration for the function
4414 before its definition and putting @code{asm} there, like this:
4417 extern func () asm ("FUNC");
4424 It is up to you to make sure that the assembler names you choose do not
4425 conflict with any other assembler symbols. Also, you must not use a
4426 register name; that would produce completely invalid assembler code. GCC
4427 does not as yet have the ability to store static variables in registers.
4428 Perhaps that will be added.
4430 @node Explicit Reg Vars
4431 @section Variables in Specified Registers
4432 @cindex explicit register variables
4433 @cindex variables in specified registers
4434 @cindex specified registers
4435 @cindex registers, global allocation
4437 GNU C allows you to put a few global variables into specified hardware
4438 registers. You can also specify the register in which an ordinary
4439 register variable should be allocated.
4443 Global register variables reserve registers throughout the program.
4444 This may be useful in programs such as programming language
4445 interpreters which have a couple of global variables that are accessed
4449 Local register variables in specific registers do not reserve the
4450 registers, except at the point where they are used as input or output
4451 operands in an @code{asm} statement and the @code{asm} statement itself is
4452 not deleted. The compiler's data flow analysis is capable of determining
4453 where the specified registers contain live values, and where they are
4454 available for other uses. Stores into local register variables may be deleted
4455 when they appear to be dead according to dataflow analysis. References
4456 to local register variables may be deleted or moved or simplified.
4458 These local variables are sometimes convenient for use with the extended
4459 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4460 output of the assembler instruction directly into a particular register.
4461 (This will work provided the register you specify fits the constraints
4462 specified for that operand in the @code{asm}.)
4470 @node Global Reg Vars
4471 @subsection Defining Global Register Variables
4472 @cindex global register variables
4473 @cindex registers, global variables in
4475 You can define a global register variable in GNU C like this:
4478 register int *foo asm ("a5");
4482 Here @code{a5} is the name of the register which should be used. Choose a
4483 register which is normally saved and restored by function calls on your
4484 machine, so that library routines will not clobber it.
4486 Naturally the register name is cpu-dependent, so you would need to
4487 conditionalize your program according to cpu type. The register
4488 @code{a5} would be a good choice on a 68000 for a variable of pointer
4489 type. On machines with register windows, be sure to choose a ``global''
4490 register that is not affected magically by the function call mechanism.
4492 In addition, operating systems on one type of cpu may differ in how they
4493 name the registers; then you would need additional conditionals. For
4494 example, some 68000 operating systems call this register @code{%a5}.
4496 Eventually there may be a way of asking the compiler to choose a register
4497 automatically, but first we need to figure out how it should choose and
4498 how to enable you to guide the choice. No solution is evident.
4500 Defining a global register variable in a certain register reserves that
4501 register entirely for this use, at least within the current compilation.
4502 The register will not be allocated for any other purpose in the functions
4503 in the current compilation. The register will not be saved and restored by
4504 these functions. Stores into this register are never deleted even if they
4505 would appear to be dead, but references may be deleted or moved or
4508 It is not safe to access the global register variables from signal
4509 handlers, or from more than one thread of control, because the system
4510 library routines may temporarily use the register for other things (unless
4511 you recompile them specially for the task at hand).
4513 @cindex @code{qsort}, and global register variables
4514 It is not safe for one function that uses a global register variable to
4515 call another such function @code{foo} by way of a third function
4516 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4517 different source file in which the variable wasn't declared). This is
4518 because @code{lose} might save the register and put some other value there.
4519 For example, you can't expect a global register variable to be available in
4520 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4521 might have put something else in that register. (If you are prepared to
4522 recompile @code{qsort} with the same global register variable, you can
4523 solve this problem.)
4525 If you want to recompile @code{qsort} or other source files which do not
4526 actually use your global register variable, so that they will not use that
4527 register for any other purpose, then it suffices to specify the compiler
4528 option @option{-ffixed-@var{reg}}. You need not actually add a global
4529 register declaration to their source code.
4531 A function which can alter the value of a global register variable cannot
4532 safely be called from a function compiled without this variable, because it
4533 could clobber the value the caller expects to find there on return.
4534 Therefore, the function which is the entry point into the part of the
4535 program that uses the global register variable must explicitly save and
4536 restore the value which belongs to its caller.
4538 @cindex register variable after @code{longjmp}
4539 @cindex global register after @code{longjmp}
4540 @cindex value after @code{longjmp}
4543 On most machines, @code{longjmp} will restore to each global register
4544 variable the value it had at the time of the @code{setjmp}. On some
4545 machines, however, @code{longjmp} will not change the value of global
4546 register variables. To be portable, the function that called @code{setjmp}
4547 should make other arrangements to save the values of the global register
4548 variables, and to restore them in a @code{longjmp}. This way, the same
4549 thing will happen regardless of what @code{longjmp} does.
4551 All global register variable declarations must precede all function
4552 definitions. If such a declaration could appear after function
4553 definitions, the declaration would be too late to prevent the register from
4554 being used for other purposes in the preceding functions.
4556 Global register variables may not have initial values, because an
4557 executable file has no means to supply initial contents for a register.
4559 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4560 registers, but certain library functions, such as @code{getwd}, as well
4561 as the subroutines for division and remainder, modify g3 and g4. g1 and
4562 g2 are local temporaries.
4564 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4565 Of course, it will not do to use more than a few of those.
4567 @node Local Reg Vars
4568 @subsection Specifying Registers for Local Variables
4569 @cindex local variables, specifying registers
4570 @cindex specifying registers for local variables
4571 @cindex registers for local variables
4573 You can define a local register variable with a specified register
4577 register int *foo asm ("a5");
4581 Here @code{a5} is the name of the register which should be used. Note
4582 that this is the same syntax used for defining global register
4583 variables, but for a local variable it would appear within a function.
4585 Naturally the register name is cpu-dependent, but this is not a
4586 problem, since specific registers are most often useful with explicit
4587 assembler instructions (@pxref{Extended Asm}). Both of these things
4588 generally require that you conditionalize your program according to
4591 In addition, operating systems on one type of cpu may differ in how they
4592 name the registers; then you would need additional conditionals. For
4593 example, some 68000 operating systems call this register @code{%a5}.
4595 Defining such a register variable does not reserve the register; it
4596 remains available for other uses in places where flow control determines
4597 the variable's value is not live.
4599 This option does not guarantee that GCC will generate code that has
4600 this variable in the register you specify at all times. You may not
4601 code an explicit reference to this register in the @emph{assembler
4602 instruction template} part of an @code{asm} statement and assume it will
4603 always refer to this variable. However, using the variable as an
4604 @code{asm} @emph{operand} guarantees that the specified register is used
4607 Stores into local register variables may be deleted when they appear to be dead
4608 according to dataflow analysis. References to local register variables may
4609 be deleted or moved or simplified.
4611 As for global register variables, it's recommended that you choose a
4612 register which is normally saved and restored by function calls on
4613 your machine, so that library routines will not clobber it. A common
4614 pitfall is to initialize multiple call-clobbered registers with
4615 arbitrary expressions, where a function call or library call for an
4616 arithmetic operator will overwrite a register value from a previous
4617 assignment, for example @code{r0} below:
4619 register int *p1 asm ("r0") = @dots{};
4620 register int *p2 asm ("r1") = @dots{};
4622 In those cases, a solution is to use a temporary variable for
4623 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4625 @node Alternate Keywords
4626 @section Alternate Keywords
4627 @cindex alternate keywords
4628 @cindex keywords, alternate
4630 @option{-ansi} and the various @option{-std} options disable certain
4631 keywords. This causes trouble when you want to use GNU C extensions, or
4632 a general-purpose header file that should be usable by all programs,
4633 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4634 @code{inline} are not available in programs compiled with
4635 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4636 program compiled with @option{-std=c99}). The ISO C99 keyword
4637 @code{restrict} is only available when @option{-std=gnu99} (which will
4638 eventually be the default) or @option{-std=c99} (or the equivalent
4639 @option{-std=iso9899:1999}) is used.
4641 The way to solve these problems is to put @samp{__} at the beginning and
4642 end of each problematical keyword. For example, use @code{__asm__}
4643 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4645 Other C compilers won't accept these alternative keywords; if you want to
4646 compile with another compiler, you can define the alternate keywords as
4647 macros to replace them with the customary keywords. It looks like this:
4655 @findex __extension__
4657 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4659 prevent such warnings within one expression by writing
4660 @code{__extension__} before the expression. @code{__extension__} has no
4661 effect aside from this.
4663 @node Incomplete Enums
4664 @section Incomplete @code{enum} Types
4666 You can define an @code{enum} tag without specifying its possible values.
4667 This results in an incomplete type, much like what you get if you write
4668 @code{struct foo} without describing the elements. A later declaration
4669 which does specify the possible values completes the type.
4671 You can't allocate variables or storage using the type while it is
4672 incomplete. However, you can work with pointers to that type.
4674 This extension may not be very useful, but it makes the handling of
4675 @code{enum} more consistent with the way @code{struct} and @code{union}
4678 This extension is not supported by GNU C++.
4680 @node Function Names
4681 @section Function Names as Strings
4682 @cindex @code{__func__} identifier
4683 @cindex @code{__FUNCTION__} identifier
4684 @cindex @code{__PRETTY_FUNCTION__} identifier
4686 GCC provides three magic variables which hold the name of the current
4687 function, as a string. The first of these is @code{__func__}, which
4688 is part of the C99 standard:
4691 The identifier @code{__func__} is implicitly declared by the translator
4692 as if, immediately following the opening brace of each function
4693 definition, the declaration
4696 static const char __func__[] = "function-name";
4699 appeared, where function-name is the name of the lexically-enclosing
4700 function. This name is the unadorned name of the function.
4703 @code{__FUNCTION__} is another name for @code{__func__}. Older
4704 versions of GCC recognize only this name. However, it is not
4705 standardized. For maximum portability, we recommend you use
4706 @code{__func__}, but provide a fallback definition with the
4710 #if __STDC_VERSION__ < 199901L
4712 # define __func__ __FUNCTION__
4714 # define __func__ "<unknown>"
4719 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4720 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4721 the type signature of the function as well as its bare name. For
4722 example, this program:
4726 extern int printf (char *, ...);
4733 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4734 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4752 __PRETTY_FUNCTION__ = void a::sub(int)
4755 These identifiers are not preprocessor macros. In GCC 3.3 and
4756 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4757 were treated as string literals; they could be used to initialize
4758 @code{char} arrays, and they could be concatenated with other string
4759 literals. GCC 3.4 and later treat them as variables, like
4760 @code{__func__}. In C++, @code{__FUNCTION__} and
4761 @code{__PRETTY_FUNCTION__} have always been variables.
4763 @node Return Address
4764 @section Getting the Return or Frame Address of a Function
4766 These functions may be used to get information about the callers of a
4769 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4770 This function returns the return address of the current function, or of
4771 one of its callers. The @var{level} argument is number of frames to
4772 scan up the call stack. A value of @code{0} yields the return address
4773 of the current function, a value of @code{1} yields the return address
4774 of the caller of the current function, and so forth. When inlining
4775 the expected behavior is that the function will return the address of
4776 the function that will be returned to. To work around this behavior use
4777 the @code{noinline} function attribute.
4779 The @var{level} argument must be a constant integer.
4781 On some machines it may be impossible to determine the return address of
4782 any function other than the current one; in such cases, or when the top
4783 of the stack has been reached, this function will return @code{0} or a
4784 random value. In addition, @code{__builtin_frame_address} may be used
4785 to determine if the top of the stack has been reached.
4787 This function should only be used with a nonzero argument for debugging
4791 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4792 This function is similar to @code{__builtin_return_address}, but it
4793 returns the address of the function frame rather than the return address
4794 of the function. Calling @code{__builtin_frame_address} with a value of
4795 @code{0} yields the frame address of the current function, a value of
4796 @code{1} yields the frame address of the caller of the current function,
4799 The frame is the area on the stack which holds local variables and saved
4800 registers. The frame address is normally the address of the first word
4801 pushed on to the stack by the function. However, the exact definition
4802 depends upon the processor and the calling convention. If the processor
4803 has a dedicated frame pointer register, and the function has a frame,
4804 then @code{__builtin_frame_address} will return the value of the frame
4807 On some machines it may be impossible to determine the frame address of
4808 any function other than the current one; in such cases, or when the top
4809 of the stack has been reached, this function will return @code{0} if
4810 the first frame pointer is properly initialized by the startup code.
4812 This function should only be used with a nonzero argument for debugging
4816 @node Vector Extensions
4817 @section Using vector instructions through built-in functions
4819 On some targets, the instruction set contains SIMD vector instructions that
4820 operate on multiple values contained in one large register at the same time.
4821 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4824 The first step in using these extensions is to provide the necessary data
4825 types. This should be done using an appropriate @code{typedef}:
4828 typedef int v4si __attribute__ ((vector_size (16)));
4831 The @code{int} type specifies the base type, while the attribute specifies
4832 the vector size for the variable, measured in bytes. For example, the
4833 declaration above causes the compiler to set the mode for the @code{v4si}
4834 type to be 16 bytes wide and divided into @code{int} sized units. For
4835 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4836 corresponding mode of @code{foo} will be @acronym{V4SI}.
4838 The @code{vector_size} attribute is only applicable to integral and
4839 float scalars, although arrays, pointers, and function return values
4840 are allowed in conjunction with this construct.
4842 All the basic integer types can be used as base types, both as signed
4843 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4844 @code{long long}. In addition, @code{float} and @code{double} can be
4845 used to build floating-point vector types.
4847 Specifying a combination that is not valid for the current architecture
4848 will cause GCC to synthesize the instructions using a narrower mode.
4849 For example, if you specify a variable of type @code{V4SI} and your
4850 architecture does not allow for this specific SIMD type, GCC will
4851 produce code that uses 4 @code{SIs}.
4853 The types defined in this manner can be used with a subset of normal C
4854 operations. Currently, GCC will allow using the following operators
4855 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4857 The operations behave like C++ @code{valarrays}. Addition is defined as
4858 the addition of the corresponding elements of the operands. For
4859 example, in the code below, each of the 4 elements in @var{a} will be
4860 added to the corresponding 4 elements in @var{b} and the resulting
4861 vector will be stored in @var{c}.
4864 typedef int v4si __attribute__ ((vector_size (16)));
4871 Subtraction, multiplication, division, and the logical operations
4872 operate in a similar manner. Likewise, the result of using the unary
4873 minus or complement operators on a vector type is a vector whose
4874 elements are the negative or complemented values of the corresponding
4875 elements in the operand.
4877 You can declare variables and use them in function calls and returns, as
4878 well as in assignments and some casts. You can specify a vector type as
4879 a return type for a function. Vector types can also be used as function
4880 arguments. It is possible to cast from one vector type to another,
4881 provided they are of the same size (in fact, you can also cast vectors
4882 to and from other datatypes of the same size).
4884 You cannot operate between vectors of different lengths or different
4885 signedness without a cast.
4887 A port that supports hardware vector operations, usually provides a set
4888 of built-in functions that can be used to operate on vectors. For
4889 example, a function to add two vectors and multiply the result by a
4890 third could look like this:
4893 v4si f (v4si a, v4si b, v4si c)
4895 v4si tmp = __builtin_addv4si (a, b);
4896 return __builtin_mulv4si (tmp, c);
4903 @findex __builtin_offsetof
4905 GCC implements for both C and C++ a syntactic extension to implement
4906 the @code{offsetof} macro.
4910 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4912 offsetof_member_designator:
4914 | offsetof_member_designator "." @code{identifier}
4915 | offsetof_member_designator "[" @code{expr} "]"
4918 This extension is sufficient such that
4921 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4924 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4925 may be dependent. In either case, @var{member} may consist of a single
4926 identifier, or a sequence of member accesses and array references.
4928 @node Atomic Builtins
4929 @section Built-in functions for atomic memory access
4931 The following builtins are intended to be compatible with those described
4932 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
4933 section 7.4. As such, they depart from the normal GCC practice of using
4934 the ``__builtin_'' prefix, and further that they are overloaded such that
4935 they work on multiple types.
4937 The definition given in the Intel documentation allows only for the use of
4938 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
4939 counterparts. GCC will allow any integral scalar or pointer type that is
4940 1, 2, 4 or 8 bytes in length.
4942 Not all operations are supported by all target processors. If a particular
4943 operation cannot be implemented on the target processor, a warning will be
4944 generated and a call an external function will be generated. The external
4945 function will carry the same name as the builtin, with an additional suffix
4946 @samp{_@var{n}} where @var{n} is the size of the data type.
4948 @c ??? Should we have a mechanism to suppress this warning? This is almost
4949 @c useful for implementing the operation under the control of an external
4952 In most cases, these builtins are considered a @dfn{full barrier}. That is,
4953 no memory operand will be moved across the operation, either forward or
4954 backward. Further, instructions will be issued as necessary to prevent the
4955 processor from speculating loads across the operation and from queuing stores
4956 after the operation.
4958 All of the routines are are described in the Intel documentation to take
4959 ``an optional list of variables protected by the memory barrier''. It's
4960 not clear what is meant by that; it could mean that @emph{only} the
4961 following variables are protected, or it could mean that these variables
4962 should in addition be protected. At present GCC ignores this list and
4963 protects all variables which are globally accessible. If in the future
4964 we make some use of this list, an empty list will continue to mean all
4965 globally accessible variables.
4968 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
4969 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
4970 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
4971 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
4972 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
4973 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
4974 @findex __sync_fetch_and_add
4975 @findex __sync_fetch_and_sub
4976 @findex __sync_fetch_and_or
4977 @findex __sync_fetch_and_and
4978 @findex __sync_fetch_and_xor
4979 @findex __sync_fetch_and_nand
4980 These builtins perform the operation suggested by the name, and
4981 returns the value that had previously been in memory. That is,
4984 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
4985 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
4988 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
4989 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
4990 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
4991 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
4992 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
4993 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
4994 @findex __sync_add_and_fetch
4995 @findex __sync_sub_and_fetch
4996 @findex __sync_or_and_fetch
4997 @findex __sync_and_and_fetch
4998 @findex __sync_xor_and_fetch
4999 @findex __sync_nand_and_fetch
5000 These builtins perform the operation suggested by the name, and
5001 return the new value. That is,
5004 @{ *ptr @var{op}= value; return *ptr; @}
5005 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5008 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5009 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5010 @findex __sync_bool_compare_and_swap
5011 @findex __sync_val_compare_and_swap
5012 These builtins perform an atomic compare and swap. That is, if the current
5013 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5016 The ``bool'' version returns true if the comparison is successful and
5017 @var{newval} was written. The ``val'' version returns the contents
5018 of @code{*@var{ptr}} before the operation.
5020 @item __sync_synchronize (...)
5021 @findex __sync_synchronize
5022 This builtin issues a full memory barrier.
5024 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5025 @findex __sync_lock_test_and_set
5026 This builtin, as described by Intel, is not a traditional test-and-set
5027 operation, but rather an atomic exchange operation. It writes @var{value}
5028 into @code{*@var{ptr}}, and returns the previous contents of
5031 Many targets have only minimal support for such locks, and do not support
5032 a full exchange operation. In this case, a target may support reduced
5033 functionality here by which the @emph{only} valid value to store is the
5034 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5035 is implementation defined.
5037 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5038 This means that references after the builtin cannot move to (or be
5039 speculated to) before the builtin, but previous memory stores may not
5040 be globally visible yet, and previous memory loads may not yet be
5043 @item void __sync_lock_release (@var{type} *ptr, ...)
5044 @findex __sync_lock_release
5045 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5046 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5048 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5049 This means that all previous memory stores are globally visible, and all
5050 previous memory loads have been satisfied, but following memory reads
5051 are not prevented from being speculated to before the barrier.
5054 @node Object Size Checking
5055 @section Object Size Checking Builtins
5056 @findex __builtin_object_size
5057 @findex __builtin___memcpy_chk
5058 @findex __builtin___mempcpy_chk
5059 @findex __builtin___memmove_chk
5060 @findex __builtin___memset_chk
5061 @findex __builtin___strcpy_chk
5062 @findex __builtin___stpcpy_chk
5063 @findex __builtin___strncpy_chk
5064 @findex __builtin___strcat_chk
5065 @findex __builtin___strncat_chk
5066 @findex __builtin___sprintf_chk
5067 @findex __builtin___snprintf_chk
5068 @findex __builtin___vsprintf_chk
5069 @findex __builtin___vsnprintf_chk
5070 @findex __builtin___printf_chk
5071 @findex __builtin___vprintf_chk
5072 @findex __builtin___fprintf_chk
5073 @findex __builtin___vfprintf_chk
5075 GCC implements a limited buffer overflow protection mechanism
5076 that can prevent some buffer overflow attacks.
5078 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5079 is a built-in construct that returns a constant number of bytes from
5080 @var{ptr} to the end of the object @var{ptr} pointer points to
5081 (if known at compile time). @code{__builtin_object_size} never evaluates
5082 its arguments for side-effects. If there are any side-effects in them, it
5083 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5084 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5085 point to and all of them are known at compile time, the returned number
5086 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5087 0 and minimum if nonzero. If it is not possible to determine which objects
5088 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5089 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5090 for @var{type} 2 or 3.
5092 @var{type} is an integer constant from 0 to 3. If the least significant
5093 bit is clear, objects are whole variables, if it is set, a closest
5094 surrounding subobject is considered the object a pointer points to.
5095 The second bit determines if maximum or minimum of remaining bytes
5099 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5100 char *p = &var.buf1[1], *q = &var.b;
5102 /* Here the object p points to is var. */
5103 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5104 /* The subobject p points to is var.buf1. */
5105 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5106 /* The object q points to is var. */
5107 assert (__builtin_object_size (q, 0)
5108 == (char *) (&var + 1) - (char *) &var.b);
5109 /* The subobject q points to is var.b. */
5110 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5114 There are built-in functions added for many common string operation
5115 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5116 built-in is provided. This built-in has an additional last argument,
5117 which is the number of bytes remaining in object the @var{dest}
5118 argument points to or @code{(size_t) -1} if the size is not known.
5120 The built-in functions are optimized into the normal string functions
5121 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5122 it is known at compile time that the destination object will not
5123 be overflown. If the compiler can determine at compile time the
5124 object will be always overflown, it issues a warning.
5126 The intended use can be e.g.
5130 #define bos0(dest) __builtin_object_size (dest, 0)
5131 #define memcpy(dest, src, n) \
5132 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5136 /* It is unknown what object p points to, so this is optimized
5137 into plain memcpy - no checking is possible. */
5138 memcpy (p, "abcde", n);
5139 /* Destination is known and length too. It is known at compile
5140 time there will be no overflow. */
5141 memcpy (&buf[5], "abcde", 5);
5142 /* Destination is known, but the length is not known at compile time.
5143 This will result in __memcpy_chk call that can check for overflow
5145 memcpy (&buf[5], "abcde", n);
5146 /* Destination is known and it is known at compile time there will
5147 be overflow. There will be a warning and __memcpy_chk call that
5148 will abort the program at runtime. */
5149 memcpy (&buf[6], "abcde", 5);
5152 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5153 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5154 @code{strcat} and @code{strncat}.
5156 There are also checking built-in functions for formatted output functions.
5158 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5159 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5160 const char *fmt, ...);
5161 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5163 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5164 const char *fmt, va_list ap);
5167 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5168 etc. functions and can contain implementation specific flags on what
5169 additional security measures the checking function might take, such as
5170 handling @code{%n} differently.
5172 The @var{os} argument is the object size @var{s} points to, like in the
5173 other built-in functions. There is a small difference in the behavior
5174 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5175 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5176 the checking function is called with @var{os} argument set to
5179 In addition to this, there are checking built-in functions
5180 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5181 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5182 These have just one additional argument, @var{flag}, right before
5183 format string @var{fmt}. If the compiler is able to optimize them to
5184 @code{fputc} etc. functions, it will, otherwise the checking function
5185 should be called and the @var{flag} argument passed to it.
5187 @node Other Builtins
5188 @section Other built-in functions provided by GCC
5189 @cindex built-in functions
5190 @findex __builtin_isgreater
5191 @findex __builtin_isgreaterequal
5192 @findex __builtin_isless
5193 @findex __builtin_islessequal
5194 @findex __builtin_islessgreater
5195 @findex __builtin_isunordered
5196 @findex __builtin_powi
5197 @findex __builtin_powif
5198 @findex __builtin_powil
5356 @findex fprintf_unlocked
5358 @findex fputs_unlocked
5468 @findex printf_unlocked
5497 @findex significandf
5498 @findex significandl
5569 GCC provides a large number of built-in functions other than the ones
5570 mentioned above. Some of these are for internal use in the processing
5571 of exceptions or variable-length argument lists and will not be
5572 documented here because they may change from time to time; we do not
5573 recommend general use of these functions.
5575 The remaining functions are provided for optimization purposes.
5577 @opindex fno-builtin
5578 GCC includes built-in versions of many of the functions in the standard
5579 C library. The versions prefixed with @code{__builtin_} will always be
5580 treated as having the same meaning as the C library function even if you
5581 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5582 Many of these functions are only optimized in certain cases; if they are
5583 not optimized in a particular case, a call to the library function will
5588 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5589 @option{-std=c99}), the functions
5590 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5591 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5592 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5593 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5594 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5595 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5596 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5597 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5598 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5599 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5600 @code{significandf}, @code{significandl}, @code{significand},
5601 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5602 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5603 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5604 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5605 @code{ynl} and @code{yn}
5606 may be handled as built-in functions.
5607 All these functions have corresponding versions
5608 prefixed with @code{__builtin_}, which may be used even in strict C89
5611 The ISO C99 functions
5612 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5613 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5614 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5615 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5616 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5617 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5618 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5619 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5620 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5621 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5622 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5623 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5624 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5625 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5626 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5627 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5628 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5629 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5630 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5631 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5632 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5633 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5634 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5635 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5636 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5637 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5638 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5639 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5640 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5641 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5642 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5643 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5644 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5645 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5646 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5647 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5648 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5649 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5650 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5651 are handled as built-in functions
5652 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5654 There are also built-in versions of the ISO C99 functions
5655 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5656 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5657 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5658 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5659 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5660 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5661 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5662 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5663 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5664 that are recognized in any mode since ISO C90 reserves these names for
5665 the purpose to which ISO C99 puts them. All these functions have
5666 corresponding versions prefixed with @code{__builtin_}.
5668 The ISO C94 functions
5669 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5670 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5671 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5673 are handled as built-in functions
5674 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5676 The ISO C90 functions
5677 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5678 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5679 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5680 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5681 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5682 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5683 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5684 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5685 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5686 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5687 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5688 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5689 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5690 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5691 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5692 @code{vprintf} and @code{vsprintf}
5693 are all recognized as built-in functions unless
5694 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5695 is specified for an individual function). All of these functions have
5696 corresponding versions prefixed with @code{__builtin_}.
5698 GCC provides built-in versions of the ISO C99 floating point comparison
5699 macros that avoid raising exceptions for unordered operands. They have
5700 the same names as the standard macros ( @code{isgreater},
5701 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5702 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5703 prefixed. We intend for a library implementor to be able to simply
5704 @code{#define} each standard macro to its built-in equivalent.
5706 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5708 You can use the built-in function @code{__builtin_types_compatible_p} to
5709 determine whether two types are the same.
5711 This built-in function returns 1 if the unqualified versions of the
5712 types @var{type1} and @var{type2} (which are types, not expressions) are
5713 compatible, 0 otherwise. The result of this built-in function can be
5714 used in integer constant expressions.
5716 This built-in function ignores top level qualifiers (e.g., @code{const},
5717 @code{volatile}). For example, @code{int} is equivalent to @code{const
5720 The type @code{int[]} and @code{int[5]} are compatible. On the other
5721 hand, @code{int} and @code{char *} are not compatible, even if the size
5722 of their types, on the particular architecture are the same. Also, the
5723 amount of pointer indirection is taken into account when determining
5724 similarity. Consequently, @code{short *} is not similar to
5725 @code{short **}. Furthermore, two types that are typedefed are
5726 considered compatible if their underlying types are compatible.
5728 An @code{enum} type is not considered to be compatible with another
5729 @code{enum} type even if both are compatible with the same integer
5730 type; this is what the C standard specifies.
5731 For example, @code{enum @{foo, bar@}} is not similar to
5732 @code{enum @{hot, dog@}}.
5734 You would typically use this function in code whose execution varies
5735 depending on the arguments' types. For example:
5740 typeof (x) tmp = (x); \
5741 if (__builtin_types_compatible_p (typeof (x), long double)) \
5742 tmp = foo_long_double (tmp); \
5743 else if (__builtin_types_compatible_p (typeof (x), double)) \
5744 tmp = foo_double (tmp); \
5745 else if (__builtin_types_compatible_p (typeof (x), float)) \
5746 tmp = foo_float (tmp); \
5753 @emph{Note:} This construct is only available for C@.
5757 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5759 You can use the built-in function @code{__builtin_choose_expr} to
5760 evaluate code depending on the value of a constant expression. This
5761 built-in function returns @var{exp1} if @var{const_exp}, which is a
5762 constant expression that must be able to be determined at compile time,
5763 is nonzero. Otherwise it returns 0.
5765 This built-in function is analogous to the @samp{? :} operator in C,
5766 except that the expression returned has its type unaltered by promotion
5767 rules. Also, the built-in function does not evaluate the expression
5768 that was not chosen. For example, if @var{const_exp} evaluates to true,
5769 @var{exp2} is not evaluated even if it has side-effects.
5771 This built-in function can return an lvalue if the chosen argument is an
5774 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5775 type. Similarly, if @var{exp2} is returned, its return type is the same
5782 __builtin_choose_expr ( \
5783 __builtin_types_compatible_p (typeof (x), double), \
5785 __builtin_choose_expr ( \
5786 __builtin_types_compatible_p (typeof (x), float), \
5788 /* @r{The void expression results in a compile-time error} \
5789 @r{when assigning the result to something.} */ \
5793 @emph{Note:} This construct is only available for C@. Furthermore, the
5794 unused expression (@var{exp1} or @var{exp2} depending on the value of
5795 @var{const_exp}) may still generate syntax errors. This may change in
5800 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5801 You can use the built-in function @code{__builtin_constant_p} to
5802 determine if a value is known to be constant at compile-time and hence
5803 that GCC can perform constant-folding on expressions involving that
5804 value. The argument of the function is the value to test. The function
5805 returns the integer 1 if the argument is known to be a compile-time
5806 constant and 0 if it is not known to be a compile-time constant. A
5807 return of 0 does not indicate that the value is @emph{not} a constant,
5808 but merely that GCC cannot prove it is a constant with the specified
5809 value of the @option{-O} option.
5811 You would typically use this function in an embedded application where
5812 memory was a critical resource. If you have some complex calculation,
5813 you may want it to be folded if it involves constants, but need to call
5814 a function if it does not. For example:
5817 #define Scale_Value(X) \
5818 (__builtin_constant_p (X) \
5819 ? ((X) * SCALE + OFFSET) : Scale (X))
5822 You may use this built-in function in either a macro or an inline
5823 function. However, if you use it in an inlined function and pass an
5824 argument of the function as the argument to the built-in, GCC will
5825 never return 1 when you call the inline function with a string constant
5826 or compound literal (@pxref{Compound Literals}) and will not return 1
5827 when you pass a constant numeric value to the inline function unless you
5828 specify the @option{-O} option.
5830 You may also use @code{__builtin_constant_p} in initializers for static
5831 data. For instance, you can write
5834 static const int table[] = @{
5835 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5841 This is an acceptable initializer even if @var{EXPRESSION} is not a
5842 constant expression. GCC must be more conservative about evaluating the
5843 built-in in this case, because it has no opportunity to perform
5846 Previous versions of GCC did not accept this built-in in data
5847 initializers. The earliest version where it is completely safe is
5851 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5852 @opindex fprofile-arcs
5853 You may use @code{__builtin_expect} to provide the compiler with
5854 branch prediction information. In general, you should prefer to
5855 use actual profile feedback for this (@option{-fprofile-arcs}), as
5856 programmers are notoriously bad at predicting how their programs
5857 actually perform. However, there are applications in which this
5858 data is hard to collect.
5860 The return value is the value of @var{exp}, which should be an
5861 integral expression. The value of @var{c} must be a compile-time
5862 constant. The semantics of the built-in are that it is expected
5863 that @var{exp} == @var{c}. For example:
5866 if (__builtin_expect (x, 0))
5871 would indicate that we do not expect to call @code{foo}, since
5872 we expect @code{x} to be zero. Since you are limited to integral
5873 expressions for @var{exp}, you should use constructions such as
5876 if (__builtin_expect (ptr != NULL, 1))
5881 when testing pointer or floating-point values.
5884 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5885 This function is used to minimize cache-miss latency by moving data into
5886 a cache before it is accessed.
5887 You can insert calls to @code{__builtin_prefetch} into code for which
5888 you know addresses of data in memory that is likely to be accessed soon.
5889 If the target supports them, data prefetch instructions will be generated.
5890 If the prefetch is done early enough before the access then the data will
5891 be in the cache by the time it is accessed.
5893 The value of @var{addr} is the address of the memory to prefetch.
5894 There are two optional arguments, @var{rw} and @var{locality}.
5895 The value of @var{rw} is a compile-time constant one or zero; one
5896 means that the prefetch is preparing for a write to the memory address
5897 and zero, the default, means that the prefetch is preparing for a read.
5898 The value @var{locality} must be a compile-time constant integer between
5899 zero and three. A value of zero means that the data has no temporal
5900 locality, so it need not be left in the cache after the access. A value
5901 of three means that the data has a high degree of temporal locality and
5902 should be left in all levels of cache possible. Values of one and two
5903 mean, respectively, a low or moderate degree of temporal locality. The
5907 for (i = 0; i < n; i++)
5910 __builtin_prefetch (&a[i+j], 1, 1);
5911 __builtin_prefetch (&b[i+j], 0, 1);
5916 Data prefetch does not generate faults if @var{addr} is invalid, but
5917 the address expression itself must be valid. For example, a prefetch
5918 of @code{p->next} will not fault if @code{p->next} is not a valid
5919 address, but evaluation will fault if @code{p} is not a valid address.
5921 If the target does not support data prefetch, the address expression
5922 is evaluated if it includes side effects but no other code is generated
5923 and GCC does not issue a warning.
5926 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5927 Returns a positive infinity, if supported by the floating-point format,
5928 else @code{DBL_MAX}. This function is suitable for implementing the
5929 ISO C macro @code{HUGE_VAL}.
5932 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5933 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5936 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5937 Similar to @code{__builtin_huge_val}, except the return
5938 type is @code{long double}.
5941 @deftypefn {Built-in Function} double __builtin_inf (void)
5942 Similar to @code{__builtin_huge_val}, except a warning is generated
5943 if the target floating-point format does not support infinities.
5946 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
5947 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
5950 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
5951 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
5954 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
5955 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
5958 @deftypefn {Built-in Function} float __builtin_inff (void)
5959 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5960 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5963 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5964 Similar to @code{__builtin_inf}, except the return
5965 type is @code{long double}.
5968 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5969 This is an implementation of the ISO C99 function @code{nan}.
5971 Since ISO C99 defines this function in terms of @code{strtod}, which we
5972 do not implement, a description of the parsing is in order. The string
5973 is parsed as by @code{strtol}; that is, the base is recognized by
5974 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5975 in the significand such that the least significant bit of the number
5976 is at the least significant bit of the significand. The number is
5977 truncated to fit the significand field provided. The significand is
5978 forced to be a quiet NaN@.
5980 This function, if given a string literal all of which would have been
5981 consumed by strtol, is evaluated early enough that it is considered a
5982 compile-time constant.
5985 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
5986 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
5989 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
5990 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
5993 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
5994 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
5997 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5998 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6001 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6002 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6005 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6006 Similar to @code{__builtin_nan}, except the significand is forced
6007 to be a signaling NaN@. The @code{nans} function is proposed by
6008 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6011 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6012 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6015 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6016 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6019 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6020 Returns one plus the index of the least significant 1-bit of @var{x}, or
6021 if @var{x} is zero, returns zero.
6024 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6025 Returns the number of leading 0-bits in @var{x}, starting at the most
6026 significant bit position. If @var{x} is 0, the result is undefined.
6029 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6030 Returns the number of trailing 0-bits in @var{x}, starting at the least
6031 significant bit position. If @var{x} is 0, the result is undefined.
6034 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6035 Returns the number of 1-bits in @var{x}.
6038 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6039 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6043 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6044 Similar to @code{__builtin_ffs}, except the argument type is
6045 @code{unsigned long}.
6048 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6049 Similar to @code{__builtin_clz}, except the argument type is
6050 @code{unsigned long}.
6053 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6054 Similar to @code{__builtin_ctz}, except the argument type is
6055 @code{unsigned long}.
6058 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6059 Similar to @code{__builtin_popcount}, except the argument type is
6060 @code{unsigned long}.
6063 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6064 Similar to @code{__builtin_parity}, except the argument type is
6065 @code{unsigned long}.
6068 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6069 Similar to @code{__builtin_ffs}, except the argument type is
6070 @code{unsigned long long}.
6073 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6074 Similar to @code{__builtin_clz}, except the argument type is
6075 @code{unsigned long long}.
6078 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6079 Similar to @code{__builtin_ctz}, except the argument type is
6080 @code{unsigned long long}.
6083 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6084 Similar to @code{__builtin_popcount}, except the argument type is
6085 @code{unsigned long long}.
6088 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6089 Similar to @code{__builtin_parity}, except the argument type is
6090 @code{unsigned long long}.
6093 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6094 Returns the first argument raised to the power of the second. Unlike the
6095 @code{pow} function no guarantees about precision and rounding are made.
6098 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6099 Similar to @code{__builtin_powi}, except the argument and return types
6103 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6104 Similar to @code{__builtin_powi}, except the argument and return types
6105 are @code{long double}.
6109 @node Target Builtins
6110 @section Built-in Functions Specific to Particular Target Machines
6112 On some target machines, GCC supports many built-in functions specific
6113 to those machines. Generally these generate calls to specific machine
6114 instructions, but allow the compiler to schedule those calls.
6117 * Alpha Built-in Functions::
6118 * ARM Built-in Functions::
6119 * Blackfin Built-in Functions::
6120 * FR-V Built-in Functions::
6121 * X86 Built-in Functions::
6122 * MIPS DSP Built-in Functions::
6123 * MIPS Paired-Single Support::
6124 * PowerPC AltiVec Built-in Functions::
6125 * SPARC VIS Built-in Functions::
6128 @node Alpha Built-in Functions
6129 @subsection Alpha Built-in Functions
6131 These built-in functions are available for the Alpha family of
6132 processors, depending on the command-line switches used.
6134 The following built-in functions are always available. They
6135 all generate the machine instruction that is part of the name.
6138 long __builtin_alpha_implver (void)
6139 long __builtin_alpha_rpcc (void)
6140 long __builtin_alpha_amask (long)
6141 long __builtin_alpha_cmpbge (long, long)
6142 long __builtin_alpha_extbl (long, long)
6143 long __builtin_alpha_extwl (long, long)
6144 long __builtin_alpha_extll (long, long)
6145 long __builtin_alpha_extql (long, long)
6146 long __builtin_alpha_extwh (long, long)
6147 long __builtin_alpha_extlh (long, long)
6148 long __builtin_alpha_extqh (long, long)
6149 long __builtin_alpha_insbl (long, long)
6150 long __builtin_alpha_inswl (long, long)
6151 long __builtin_alpha_insll (long, long)
6152 long __builtin_alpha_insql (long, long)
6153 long __builtin_alpha_inswh (long, long)
6154 long __builtin_alpha_inslh (long, long)
6155 long __builtin_alpha_insqh (long, long)
6156 long __builtin_alpha_mskbl (long, long)
6157 long __builtin_alpha_mskwl (long, long)
6158 long __builtin_alpha_mskll (long, long)
6159 long __builtin_alpha_mskql (long, long)
6160 long __builtin_alpha_mskwh (long, long)
6161 long __builtin_alpha_msklh (long, long)
6162 long __builtin_alpha_mskqh (long, long)
6163 long __builtin_alpha_umulh (long, long)
6164 long __builtin_alpha_zap (long, long)
6165 long __builtin_alpha_zapnot (long, long)
6168 The following built-in functions are always with @option{-mmax}
6169 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6170 later. They all generate the machine instruction that is part
6174 long __builtin_alpha_pklb (long)
6175 long __builtin_alpha_pkwb (long)
6176 long __builtin_alpha_unpkbl (long)
6177 long __builtin_alpha_unpkbw (long)
6178 long __builtin_alpha_minub8 (long, long)
6179 long __builtin_alpha_minsb8 (long, long)
6180 long __builtin_alpha_minuw4 (long, long)
6181 long __builtin_alpha_minsw4 (long, long)
6182 long __builtin_alpha_maxub8 (long, long)
6183 long __builtin_alpha_maxsb8 (long, long)
6184 long __builtin_alpha_maxuw4 (long, long)
6185 long __builtin_alpha_maxsw4 (long, long)
6186 long __builtin_alpha_perr (long, long)
6189 The following built-in functions are always with @option{-mcix}
6190 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6191 later. They all generate the machine instruction that is part
6195 long __builtin_alpha_cttz (long)
6196 long __builtin_alpha_ctlz (long)
6197 long __builtin_alpha_ctpop (long)
6200 The following builtins are available on systems that use the OSF/1
6201 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6202 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6203 @code{rdval} and @code{wrval}.
6206 void *__builtin_thread_pointer (void)
6207 void __builtin_set_thread_pointer (void *)
6210 @node ARM Built-in Functions
6211 @subsection ARM Built-in Functions
6213 These built-in functions are available for the ARM family of
6214 processors, when the @option{-mcpu=iwmmxt} switch is used:
6217 typedef int v2si __attribute__ ((vector_size (8)));
6218 typedef short v4hi __attribute__ ((vector_size (8)));
6219 typedef char v8qi __attribute__ ((vector_size (8)));
6221 int __builtin_arm_getwcx (int)
6222 void __builtin_arm_setwcx (int, int)
6223 int __builtin_arm_textrmsb (v8qi, int)
6224 int __builtin_arm_textrmsh (v4hi, int)
6225 int __builtin_arm_textrmsw (v2si, int)
6226 int __builtin_arm_textrmub (v8qi, int)
6227 int __builtin_arm_textrmuh (v4hi, int)
6228 int __builtin_arm_textrmuw (v2si, int)
6229 v8qi __builtin_arm_tinsrb (v8qi, int)
6230 v4hi __builtin_arm_tinsrh (v4hi, int)
6231 v2si __builtin_arm_tinsrw (v2si, int)
6232 long long __builtin_arm_tmia (long long, int, int)
6233 long long __builtin_arm_tmiabb (long long, int, int)
6234 long long __builtin_arm_tmiabt (long long, int, int)
6235 long long __builtin_arm_tmiaph (long long, int, int)
6236 long long __builtin_arm_tmiatb (long long, int, int)
6237 long long __builtin_arm_tmiatt (long long, int, int)
6238 int __builtin_arm_tmovmskb (v8qi)
6239 int __builtin_arm_tmovmskh (v4hi)
6240 int __builtin_arm_tmovmskw (v2si)
6241 long long __builtin_arm_waccb (v8qi)
6242 long long __builtin_arm_wacch (v4hi)
6243 long long __builtin_arm_waccw (v2si)
6244 v8qi __builtin_arm_waddb (v8qi, v8qi)
6245 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6246 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6247 v4hi __builtin_arm_waddh (v4hi, v4hi)
6248 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6249 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6250 v2si __builtin_arm_waddw (v2si, v2si)
6251 v2si __builtin_arm_waddwss (v2si, v2si)
6252 v2si __builtin_arm_waddwus (v2si, v2si)
6253 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6254 long long __builtin_arm_wand(long long, long long)
6255 long long __builtin_arm_wandn (long long, long long)
6256 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6257 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6258 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6259 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6260 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6261 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6262 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6263 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6264 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6265 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6266 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6267 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6268 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6269 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6270 long long __builtin_arm_wmacsz (v4hi, v4hi)
6271 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6272 long long __builtin_arm_wmacuz (v4hi, v4hi)
6273 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6274 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6275 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6276 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6277 v2si __builtin_arm_wmaxsw (v2si, v2si)
6278 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6279 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6280 v2si __builtin_arm_wmaxuw (v2si, v2si)
6281 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6282 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6283 v2si __builtin_arm_wminsw (v2si, v2si)
6284 v8qi __builtin_arm_wminub (v8qi, v8qi)
6285 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6286 v2si __builtin_arm_wminuw (v2si, v2si)
6287 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6288 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6289 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6290 long long __builtin_arm_wor (long long, long long)
6291 v2si __builtin_arm_wpackdss (long long, long long)
6292 v2si __builtin_arm_wpackdus (long long, long long)
6293 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6294 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6295 v4hi __builtin_arm_wpackwss (v2si, v2si)
6296 v4hi __builtin_arm_wpackwus (v2si, v2si)
6297 long long __builtin_arm_wrord (long long, long long)
6298 long long __builtin_arm_wrordi (long long, int)
6299 v4hi __builtin_arm_wrorh (v4hi, long long)
6300 v4hi __builtin_arm_wrorhi (v4hi, int)
6301 v2si __builtin_arm_wrorw (v2si, long long)
6302 v2si __builtin_arm_wrorwi (v2si, int)
6303 v2si __builtin_arm_wsadb (v8qi, v8qi)
6304 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6305 v2si __builtin_arm_wsadh (v4hi, v4hi)
6306 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6307 v4hi __builtin_arm_wshufh (v4hi, int)
6308 long long __builtin_arm_wslld (long long, long long)
6309 long long __builtin_arm_wslldi (long long, int)
6310 v4hi __builtin_arm_wsllh (v4hi, long long)
6311 v4hi __builtin_arm_wsllhi (v4hi, int)
6312 v2si __builtin_arm_wsllw (v2si, long long)
6313 v2si __builtin_arm_wsllwi (v2si, int)
6314 long long __builtin_arm_wsrad (long long, long long)
6315 long long __builtin_arm_wsradi (long long, int)
6316 v4hi __builtin_arm_wsrah (v4hi, long long)
6317 v4hi __builtin_arm_wsrahi (v4hi, int)
6318 v2si __builtin_arm_wsraw (v2si, long long)
6319 v2si __builtin_arm_wsrawi (v2si, int)
6320 long long __builtin_arm_wsrld (long long, long long)
6321 long long __builtin_arm_wsrldi (long long, int)
6322 v4hi __builtin_arm_wsrlh (v4hi, long long)
6323 v4hi __builtin_arm_wsrlhi (v4hi, int)
6324 v2si __builtin_arm_wsrlw (v2si, long long)
6325 v2si __builtin_arm_wsrlwi (v2si, int)
6326 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6327 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6328 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6329 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6330 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6331 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6332 v2si __builtin_arm_wsubw (v2si, v2si)
6333 v2si __builtin_arm_wsubwss (v2si, v2si)
6334 v2si __builtin_arm_wsubwus (v2si, v2si)
6335 v4hi __builtin_arm_wunpckehsb (v8qi)
6336 v2si __builtin_arm_wunpckehsh (v4hi)
6337 long long __builtin_arm_wunpckehsw (v2si)
6338 v4hi __builtin_arm_wunpckehub (v8qi)
6339 v2si __builtin_arm_wunpckehuh (v4hi)
6340 long long __builtin_arm_wunpckehuw (v2si)
6341 v4hi __builtin_arm_wunpckelsb (v8qi)
6342 v2si __builtin_arm_wunpckelsh (v4hi)
6343 long long __builtin_arm_wunpckelsw (v2si)
6344 v4hi __builtin_arm_wunpckelub (v8qi)
6345 v2si __builtin_arm_wunpckeluh (v4hi)
6346 long long __builtin_arm_wunpckeluw (v2si)
6347 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6348 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6349 v2si __builtin_arm_wunpckihw (v2si, v2si)
6350 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6351 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6352 v2si __builtin_arm_wunpckilw (v2si, v2si)
6353 long long __builtin_arm_wxor (long long, long long)
6354 long long __builtin_arm_wzero ()
6357 @node Blackfin Built-in Functions
6358 @subsection Blackfin Built-in Functions
6360 Currently, there are two Blackfin-specific built-in functions. These are
6361 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6362 using inline assembly; by using these built-in functions the compiler can
6363 automatically add workarounds for hardware errata involving these
6364 instructions. These functions are named as follows:
6367 void __builtin_bfin_csync (void)
6368 void __builtin_bfin_ssync (void)
6371 @node FR-V Built-in Functions
6372 @subsection FR-V Built-in Functions
6374 GCC provides many FR-V-specific built-in functions. In general,
6375 these functions are intended to be compatible with those described
6376 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6377 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6378 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6379 pointer rather than by value.
6381 Most of the functions are named after specific FR-V instructions.
6382 Such functions are said to be ``directly mapped'' and are summarized
6383 here in tabular form.
6387 * Directly-mapped Integer Functions::
6388 * Directly-mapped Media Functions::
6389 * Raw read/write Functions::
6390 * Other Built-in Functions::
6393 @node Argument Types
6394 @subsubsection Argument Types
6396 The arguments to the built-in functions can be divided into three groups:
6397 register numbers, compile-time constants and run-time values. In order
6398 to make this classification clear at a glance, the arguments and return
6399 values are given the following pseudo types:
6401 @multitable @columnfractions .20 .30 .15 .35
6402 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6403 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6404 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6405 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6406 @item @code{uw2} @tab @code{unsigned long long} @tab No
6407 @tab an unsigned doubleword
6408 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6409 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6410 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6411 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6414 These pseudo types are not defined by GCC, they are simply a notational
6415 convenience used in this manual.
6417 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6418 and @code{sw2} are evaluated at run time. They correspond to
6419 register operands in the underlying FR-V instructions.
6421 @code{const} arguments represent immediate operands in the underlying
6422 FR-V instructions. They must be compile-time constants.
6424 @code{acc} arguments are evaluated at compile time and specify the number
6425 of an accumulator register. For example, an @code{acc} argument of 2
6426 will select the ACC2 register.
6428 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6429 number of an IACC register. See @pxref{Other Built-in Functions}
6432 @node Directly-mapped Integer Functions
6433 @subsubsection Directly-mapped Integer Functions
6435 The functions listed below map directly to FR-V I-type instructions.
6437 @multitable @columnfractions .45 .32 .23
6438 @item Function prototype @tab Example usage @tab Assembly output
6439 @item @code{sw1 __ADDSS (sw1, sw1)}
6440 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6441 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6442 @item @code{sw1 __SCAN (sw1, sw1)}
6443 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6444 @tab @code{SCAN @var{a},@var{b},@var{c}}
6445 @item @code{sw1 __SCUTSS (sw1)}
6446 @tab @code{@var{b} = __SCUTSS (@var{a})}
6447 @tab @code{SCUTSS @var{a},@var{b}}
6448 @item @code{sw1 __SLASS (sw1, sw1)}
6449 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6450 @tab @code{SLASS @var{a},@var{b},@var{c}}
6451 @item @code{void __SMASS (sw1, sw1)}
6452 @tab @code{__SMASS (@var{a}, @var{b})}
6453 @tab @code{SMASS @var{a},@var{b}}
6454 @item @code{void __SMSSS (sw1, sw1)}
6455 @tab @code{__SMSSS (@var{a}, @var{b})}
6456 @tab @code{SMSSS @var{a},@var{b}}
6457 @item @code{void __SMU (sw1, sw1)}
6458 @tab @code{__SMU (@var{a}, @var{b})}
6459 @tab @code{SMU @var{a},@var{b}}
6460 @item @code{sw2 __SMUL (sw1, sw1)}
6461 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6462 @tab @code{SMUL @var{a},@var{b},@var{c}}
6463 @item @code{sw1 __SUBSS (sw1, sw1)}
6464 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6465 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6466 @item @code{uw2 __UMUL (uw1, uw1)}
6467 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6468 @tab @code{UMUL @var{a},@var{b},@var{c}}
6471 @node Directly-mapped Media Functions
6472 @subsubsection Directly-mapped Media Functions
6474 The functions listed below map directly to FR-V M-type instructions.
6476 @multitable @columnfractions .45 .32 .23
6477 @item Function prototype @tab Example usage @tab Assembly output
6478 @item @code{uw1 __MABSHS (sw1)}
6479 @tab @code{@var{b} = __MABSHS (@var{a})}
6480 @tab @code{MABSHS @var{a},@var{b}}
6481 @item @code{void __MADDACCS (acc, acc)}
6482 @tab @code{__MADDACCS (@var{b}, @var{a})}
6483 @tab @code{MADDACCS @var{a},@var{b}}
6484 @item @code{sw1 __MADDHSS (sw1, sw1)}
6485 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6486 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6487 @item @code{uw1 __MADDHUS (uw1, uw1)}
6488 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6489 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6490 @item @code{uw1 __MAND (uw1, uw1)}
6491 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6492 @tab @code{MAND @var{a},@var{b},@var{c}}
6493 @item @code{void __MASACCS (acc, acc)}
6494 @tab @code{__MASACCS (@var{b}, @var{a})}
6495 @tab @code{MASACCS @var{a},@var{b}}
6496 @item @code{uw1 __MAVEH (uw1, uw1)}
6497 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6498 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6499 @item @code{uw2 __MBTOH (uw1)}
6500 @tab @code{@var{b} = __MBTOH (@var{a})}
6501 @tab @code{MBTOH @var{a},@var{b}}
6502 @item @code{void __MBTOHE (uw1 *, uw1)}
6503 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6504 @tab @code{MBTOHE @var{a},@var{b}}
6505 @item @code{void __MCLRACC (acc)}
6506 @tab @code{__MCLRACC (@var{a})}
6507 @tab @code{MCLRACC @var{a}}
6508 @item @code{void __MCLRACCA (void)}
6509 @tab @code{__MCLRACCA ()}
6510 @tab @code{MCLRACCA}
6511 @item @code{uw1 __Mcop1 (uw1, uw1)}
6512 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6513 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6514 @item @code{uw1 __Mcop2 (uw1, uw1)}
6515 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6516 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6517 @item @code{uw1 __MCPLHI (uw2, const)}
6518 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6519 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6520 @item @code{uw1 __MCPLI (uw2, const)}
6521 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6522 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6523 @item @code{void __MCPXIS (acc, sw1, sw1)}
6524 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6525 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6526 @item @code{void __MCPXIU (acc, uw1, uw1)}
6527 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6528 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6529 @item @code{void __MCPXRS (acc, sw1, sw1)}
6530 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6531 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6532 @item @code{void __MCPXRU (acc, uw1, uw1)}
6533 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6534 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6535 @item @code{uw1 __MCUT (acc, uw1)}
6536 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6537 @tab @code{MCUT @var{a},@var{b},@var{c}}
6538 @item @code{uw1 __MCUTSS (acc, sw1)}
6539 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6540 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6541 @item @code{void __MDADDACCS (acc, acc)}
6542 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6543 @tab @code{MDADDACCS @var{a},@var{b}}
6544 @item @code{void __MDASACCS (acc, acc)}
6545 @tab @code{__MDASACCS (@var{b}, @var{a})}
6546 @tab @code{MDASACCS @var{a},@var{b}}
6547 @item @code{uw2 __MDCUTSSI (acc, const)}
6548 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6549 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6550 @item @code{uw2 __MDPACKH (uw2, uw2)}
6551 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6552 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6553 @item @code{uw2 __MDROTLI (uw2, const)}
6554 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6555 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6556 @item @code{void __MDSUBACCS (acc, acc)}
6557 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6558 @tab @code{MDSUBACCS @var{a},@var{b}}
6559 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6560 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6561 @tab @code{MDUNPACKH @var{a},@var{b}}
6562 @item @code{uw2 __MEXPDHD (uw1, const)}
6563 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6564 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6565 @item @code{uw1 __MEXPDHW (uw1, const)}
6566 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6567 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6568 @item @code{uw1 __MHDSETH (uw1, const)}
6569 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6570 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6571 @item @code{sw1 __MHDSETS (const)}
6572 @tab @code{@var{b} = __MHDSETS (@var{a})}
6573 @tab @code{MHDSETS #@var{a},@var{b}}
6574 @item @code{uw1 __MHSETHIH (uw1, const)}
6575 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6576 @tab @code{MHSETHIH #@var{a},@var{b}}
6577 @item @code{sw1 __MHSETHIS (sw1, const)}
6578 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6579 @tab @code{MHSETHIS #@var{a},@var{b}}
6580 @item @code{uw1 __MHSETLOH (uw1, const)}
6581 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6582 @tab @code{MHSETLOH #@var{a},@var{b}}
6583 @item @code{sw1 __MHSETLOS (sw1, const)}
6584 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6585 @tab @code{MHSETLOS #@var{a},@var{b}}
6586 @item @code{uw1 __MHTOB (uw2)}
6587 @tab @code{@var{b} = __MHTOB (@var{a})}
6588 @tab @code{MHTOB @var{a},@var{b}}
6589 @item @code{void __MMACHS (acc, sw1, sw1)}
6590 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6591 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6592 @item @code{void __MMACHU (acc, uw1, uw1)}
6593 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6594 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6595 @item @code{void __MMRDHS (acc, sw1, sw1)}
6596 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6597 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6598 @item @code{void __MMRDHU (acc, uw1, uw1)}
6599 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6600 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6601 @item @code{void __MMULHS (acc, sw1, sw1)}
6602 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6603 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6604 @item @code{void __MMULHU (acc, uw1, uw1)}
6605 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6606 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6607 @item @code{void __MMULXHS (acc, sw1, sw1)}
6608 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6609 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6610 @item @code{void __MMULXHU (acc, uw1, uw1)}
6611 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6612 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6613 @item @code{uw1 __MNOT (uw1)}
6614 @tab @code{@var{b} = __MNOT (@var{a})}
6615 @tab @code{MNOT @var{a},@var{b}}
6616 @item @code{uw1 __MOR (uw1, uw1)}
6617 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6618 @tab @code{MOR @var{a},@var{b},@var{c}}
6619 @item @code{uw1 __MPACKH (uh, uh)}
6620 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6621 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6622 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6623 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6624 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6625 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6626 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6627 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6628 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6629 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6630 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6631 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6632 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6633 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6634 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6635 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6636 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6637 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6638 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6639 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6640 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6641 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6642 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6643 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6644 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6645 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6646 @item @code{void __MQMACHS (acc, sw2, sw2)}
6647 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6648 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6649 @item @code{void __MQMACHU (acc, uw2, uw2)}
6650 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6651 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6652 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6653 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6654 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6655 @item @code{void __MQMULHS (acc, sw2, sw2)}
6656 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6657 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6658 @item @code{void __MQMULHU (acc, uw2, uw2)}
6659 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6660 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6661 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6662 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6663 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6664 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6665 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6666 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6667 @item @code{sw2 __MQSATHS (sw2, sw2)}
6668 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6669 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6670 @item @code{uw2 __MQSLLHI (uw2, int)}
6671 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6672 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6673 @item @code{sw2 __MQSRAHI (sw2, int)}
6674 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6675 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6676 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6677 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6678 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6679 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6680 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6681 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6682 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6683 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6684 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6685 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6686 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6687 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6688 @item @code{uw1 __MRDACC (acc)}
6689 @tab @code{@var{b} = __MRDACC (@var{a})}
6690 @tab @code{MRDACC @var{a},@var{b}}
6691 @item @code{uw1 __MRDACCG (acc)}
6692 @tab @code{@var{b} = __MRDACCG (@var{a})}
6693 @tab @code{MRDACCG @var{a},@var{b}}
6694 @item @code{uw1 __MROTLI (uw1, const)}
6695 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6696 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6697 @item @code{uw1 __MROTRI (uw1, const)}
6698 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6699 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6700 @item @code{sw1 __MSATHS (sw1, sw1)}
6701 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6702 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6703 @item @code{uw1 __MSATHU (uw1, uw1)}
6704 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6705 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6706 @item @code{uw1 __MSLLHI (uw1, const)}
6707 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6708 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6709 @item @code{sw1 __MSRAHI (sw1, const)}
6710 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6711 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6712 @item @code{uw1 __MSRLHI (uw1, const)}
6713 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6714 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6715 @item @code{void __MSUBACCS (acc, acc)}
6716 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6717 @tab @code{MSUBACCS @var{a},@var{b}}
6718 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6719 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6720 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6721 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6722 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6723 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6724 @item @code{void __MTRAP (void)}
6725 @tab @code{__MTRAP ()}
6727 @item @code{uw2 __MUNPACKH (uw1)}
6728 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6729 @tab @code{MUNPACKH @var{a},@var{b}}
6730 @item @code{uw1 __MWCUT (uw2, uw1)}
6731 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6732 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6733 @item @code{void __MWTACC (acc, uw1)}
6734 @tab @code{__MWTACC (@var{b}, @var{a})}
6735 @tab @code{MWTACC @var{a},@var{b}}
6736 @item @code{void __MWTACCG (acc, uw1)}
6737 @tab @code{__MWTACCG (@var{b}, @var{a})}
6738 @tab @code{MWTACCG @var{a},@var{b}}
6739 @item @code{uw1 __MXOR (uw1, uw1)}
6740 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6741 @tab @code{MXOR @var{a},@var{b},@var{c}}
6744 @node Raw read/write Functions
6745 @subsubsection Raw read/write Functions
6747 This sections describes built-in functions related to read and write
6748 instructions to access memory. These functions generate
6749 @code{membar} instructions to flush the I/O load and stores where
6750 appropriate, as described in Fujitsu's manual described above.
6754 @item unsigned char __builtin_read8 (void *@var{data})
6755 @item unsigned short __builtin_read16 (void *@var{data})
6756 @item unsigned long __builtin_read32 (void *@var{data})
6757 @item unsigned long long __builtin_read64 (void *@var{data})
6759 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6760 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6761 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6762 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6765 @node Other Built-in Functions
6766 @subsubsection Other Built-in Functions
6768 This section describes built-in functions that are not named after
6769 a specific FR-V instruction.
6772 @item sw2 __IACCreadll (iacc @var{reg})
6773 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6774 for future expansion and must be 0.
6776 @item sw1 __IACCreadl (iacc @var{reg})
6777 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6778 Other values of @var{reg} are rejected as invalid.
6780 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6781 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6782 is reserved for future expansion and must be 0.
6784 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6785 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6786 is 1. Other values of @var{reg} are rejected as invalid.
6788 @item void __data_prefetch0 (const void *@var{x})
6789 Use the @code{dcpl} instruction to load the contents of address @var{x}
6790 into the data cache.
6792 @item void __data_prefetch (const void *@var{x})
6793 Use the @code{nldub} instruction to load the contents of address @var{x}
6794 into the data cache. The instruction will be issued in slot I1@.
6797 @node X86 Built-in Functions
6798 @subsection X86 Built-in Functions
6800 These built-in functions are available for the i386 and x86-64 family
6801 of computers, depending on the command-line switches used.
6803 Note that, if you specify command-line switches such as @option{-msse},
6804 the compiler could use the extended instruction sets even if the built-ins
6805 are not used explicitly in the program. For this reason, applications
6806 which perform runtime CPU detection must compile separate files for each
6807 supported architecture, using the appropriate flags. In particular,
6808 the file containing the CPU detection code should be compiled without
6811 The following machine modes are available for use with MMX built-in functions
6812 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6813 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6814 vector of eight 8-bit integers. Some of the built-in functions operate on
6815 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6817 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6818 of two 32-bit floating point values.
6820 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6821 floating point values. Some instructions use a vector of four 32-bit
6822 integers, these use @code{V4SI}. Finally, some instructions operate on an
6823 entire vector register, interpreting it as a 128-bit integer, these use mode
6826 The following built-in functions are made available by @option{-mmmx}.
6827 All of them generate the machine instruction that is part of the name.
6830 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6831 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6832 v2si __builtin_ia32_paddd (v2si, v2si)
6833 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6834 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6835 v2si __builtin_ia32_psubd (v2si, v2si)
6836 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6837 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6838 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6839 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6840 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6841 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6842 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6843 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6844 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6845 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6846 di __builtin_ia32_pand (di, di)
6847 di __builtin_ia32_pandn (di,di)
6848 di __builtin_ia32_por (di, di)
6849 di __builtin_ia32_pxor (di, di)
6850 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6851 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6852 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6853 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6854 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6855 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6856 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6857 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6858 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6859 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6860 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6861 v2si __builtin_ia32_punpckldq (v2si, v2si)
6862 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6863 v4hi __builtin_ia32_packssdw (v2si, v2si)
6864 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6867 The following built-in functions are made available either with
6868 @option{-msse}, or with a combination of @option{-m3dnow} and
6869 @option{-march=athlon}. All of them generate the machine
6870 instruction that is part of the name.
6873 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6874 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6875 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6876 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6877 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6878 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6879 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6880 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6881 int __builtin_ia32_pextrw (v4hi, int)
6882 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6883 int __builtin_ia32_pmovmskb (v8qi)
6884 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6885 void __builtin_ia32_movntq (di *, di)
6886 void __builtin_ia32_sfence (void)
6889 The following built-in functions are available when @option{-msse} is used.
6890 All of them generate the machine instruction that is part of the name.
6893 int __builtin_ia32_comieq (v4sf, v4sf)
6894 int __builtin_ia32_comineq (v4sf, v4sf)
6895 int __builtin_ia32_comilt (v4sf, v4sf)
6896 int __builtin_ia32_comile (v4sf, v4sf)
6897 int __builtin_ia32_comigt (v4sf, v4sf)
6898 int __builtin_ia32_comige (v4sf, v4sf)
6899 int __builtin_ia32_ucomieq (v4sf, v4sf)
6900 int __builtin_ia32_ucomineq (v4sf, v4sf)
6901 int __builtin_ia32_ucomilt (v4sf, v4sf)
6902 int __builtin_ia32_ucomile (v4sf, v4sf)
6903 int __builtin_ia32_ucomigt (v4sf, v4sf)
6904 int __builtin_ia32_ucomige (v4sf, v4sf)
6905 v4sf __builtin_ia32_addps (v4sf, v4sf)
6906 v4sf __builtin_ia32_subps (v4sf, v4sf)
6907 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6908 v4sf __builtin_ia32_divps (v4sf, v4sf)
6909 v4sf __builtin_ia32_addss (v4sf, v4sf)
6910 v4sf __builtin_ia32_subss (v4sf, v4sf)
6911 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6912 v4sf __builtin_ia32_divss (v4sf, v4sf)
6913 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6914 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6915 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6916 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6917 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6918 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6919 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6920 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6921 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6922 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6923 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6924 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6925 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6926 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6927 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6928 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6929 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6930 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6931 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6932 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6933 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6934 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6935 v4sf __builtin_ia32_minps (v4sf, v4sf)
6936 v4sf __builtin_ia32_minss (v4sf, v4sf)
6937 v4sf __builtin_ia32_andps (v4sf, v4sf)
6938 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6939 v4sf __builtin_ia32_orps (v4sf, v4sf)
6940 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6941 v4sf __builtin_ia32_movss (v4sf, v4sf)
6942 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6943 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6944 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6945 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6946 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6947 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6948 v2si __builtin_ia32_cvtps2pi (v4sf)
6949 int __builtin_ia32_cvtss2si (v4sf)
6950 v2si __builtin_ia32_cvttps2pi (v4sf)
6951 int __builtin_ia32_cvttss2si (v4sf)
6952 v4sf __builtin_ia32_rcpps (v4sf)
6953 v4sf __builtin_ia32_rsqrtps (v4sf)
6954 v4sf __builtin_ia32_sqrtps (v4sf)
6955 v4sf __builtin_ia32_rcpss (v4sf)
6956 v4sf __builtin_ia32_rsqrtss (v4sf)
6957 v4sf __builtin_ia32_sqrtss (v4sf)
6958 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6959 void __builtin_ia32_movntps (float *, v4sf)
6960 int __builtin_ia32_movmskps (v4sf)
6963 The following built-in functions are available when @option{-msse} is used.
6966 @item v4sf __builtin_ia32_loadaps (float *)
6967 Generates the @code{movaps} machine instruction as a load from memory.
6968 @item void __builtin_ia32_storeaps (float *, v4sf)
6969 Generates the @code{movaps} machine instruction as a store to memory.
6970 @item v4sf __builtin_ia32_loadups (float *)
6971 Generates the @code{movups} machine instruction as a load from memory.
6972 @item void __builtin_ia32_storeups (float *, v4sf)
6973 Generates the @code{movups} machine instruction as a store to memory.
6974 @item v4sf __builtin_ia32_loadsss (float *)
6975 Generates the @code{movss} machine instruction as a load from memory.
6976 @item void __builtin_ia32_storess (float *, v4sf)
6977 Generates the @code{movss} machine instruction as a store to memory.
6978 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6979 Generates the @code{movhps} machine instruction as a load from memory.
6980 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6981 Generates the @code{movlps} machine instruction as a load from memory
6982 @item void __builtin_ia32_storehps (v4sf, v2si *)
6983 Generates the @code{movhps} machine instruction as a store to memory.
6984 @item void __builtin_ia32_storelps (v4sf, v2si *)
6985 Generates the @code{movlps} machine instruction as a store to memory.
6988 The following built-in functions are available when @option{-msse3} is used.
6989 All of them generate the machine instruction that is part of the name.
6992 v2df __builtin_ia32_addsubpd (v2df, v2df)
6993 v2df __builtin_ia32_addsubps (v2df, v2df)
6994 v2df __builtin_ia32_haddpd (v2df, v2df)
6995 v2df __builtin_ia32_haddps (v2df, v2df)
6996 v2df __builtin_ia32_hsubpd (v2df, v2df)
6997 v2df __builtin_ia32_hsubps (v2df, v2df)
6998 v16qi __builtin_ia32_lddqu (char const *)
6999 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7000 v2df __builtin_ia32_movddup (v2df)
7001 v4sf __builtin_ia32_movshdup (v4sf)
7002 v4sf __builtin_ia32_movsldup (v4sf)
7003 void __builtin_ia32_mwait (unsigned int, unsigned int)
7006 The following built-in functions are available when @option{-msse3} is used.
7009 @item v2df __builtin_ia32_loadddup (double const *)
7010 Generates the @code{movddup} machine instruction as a load from memory.
7013 The following built-in functions are available when @option{-m3dnow} is used.
7014 All of them generate the machine instruction that is part of the name.
7017 void __builtin_ia32_femms (void)
7018 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7019 v2si __builtin_ia32_pf2id (v2sf)
7020 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7021 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7022 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7023 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7024 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7025 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7026 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7027 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7028 v2sf __builtin_ia32_pfrcp (v2sf)
7029 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7030 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7031 v2sf __builtin_ia32_pfrsqrt (v2sf)
7032 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7033 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7034 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7035 v2sf __builtin_ia32_pi2fd (v2si)
7036 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7039 The following built-in functions are available when both @option{-m3dnow}
7040 and @option{-march=athlon} are used. All of them generate the machine
7041 instruction that is part of the name.
7044 v2si __builtin_ia32_pf2iw (v2sf)
7045 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7046 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7047 v2sf __builtin_ia32_pi2fw (v2si)
7048 v2sf __builtin_ia32_pswapdsf (v2sf)
7049 v2si __builtin_ia32_pswapdsi (v2si)
7052 @node MIPS DSP Built-in Functions
7053 @subsection MIPS DSP Built-in Functions
7055 The MIPS DSP Application-Specific Extension (ASE) includes new
7056 instructions that are designed to improve the performance of DSP and
7057 media applications. It provides instructions that operate on packed
7058 8-bit integer data, Q15 fractional data and Q31 fractional data.
7060 GCC supports MIPS DSP operations using both the generic
7061 vector extensions (@pxref{Vector Extensions}) and a collection of
7062 MIPS-specific built-in functions. Both kinds of support are
7063 enabled by the @option{-mdsp} command-line option.
7065 At present, GCC only provides support for operations on 32-bit
7066 vectors. The vector type associated with 8-bit integer data is
7067 usually called @code{v4i8} and the vector type associated with Q15 is
7068 usually called @code{v2q15}. They can be defined in C as follows:
7071 typedef char v4i8 __attribute__ ((vector_size(4)));
7072 typedef short v2q15 __attribute__ ((vector_size(4)));
7075 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7076 aggregates. For example:
7079 v4i8 a = @{1, 2, 3, 4@};
7081 b = (v4i8) @{5, 6, 7, 8@};
7083 v2q15 c = @{0x0fcb, 0x3a75@};
7085 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7088 @emph{Note:} The CPU's endianness determines the order in which values
7089 are packed. On little-endian targets, the first value is the least
7090 significant and the last value is the most significant. The opposite
7091 order applies to big-endian targets. For example, the code above will
7092 set the lowest byte of @code{a} to @code{1} on little-endian targets
7093 and @code{4} on big-endian targets.
7095 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7096 representation. As shown in this example, the integer representation
7097 of a Q15 value can be obtained by multiplying the fractional value by
7098 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7101 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7102 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7103 and @code{c} and @code{d} are @code{v2q15} values.
7105 @multitable @columnfractions .50 .50
7106 @item C code @tab MIPS instruction
7107 @item @code{a + b} @tab @code{addu.qb}
7108 @item @code{c + d} @tab @code{addq.ph}
7109 @item @code{a - b} @tab @code{subu.qb}
7110 @item @code{c - d} @tab @code{subq.ph}
7113 It is easier to describe the DSP built-in functions if we first define
7114 the following types:
7119 typedef long long a64;
7122 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7123 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7124 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7125 @code{long long}, but we use @code{a64} to indicate values that will
7126 be placed in one of the four DSP accumulators (@code{$ac0},
7127 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7129 Also, some built-in functions prefer or require immediate numbers as
7130 parameters, because the corresponding DSP instructions accept both immediate
7131 numbers and register operands, or accept immediate numbers only. The
7132 immediate parameters are listed as follows.
7140 imm_n32_31: -32 to 31.
7141 imm_n512_511: -512 to 511.
7144 The following built-in functions map directly to a particular MIPS DSP
7145 instruction. Please refer to the architecture specification
7146 for details on what each instruction does.
7149 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7150 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7151 q31 __builtin_mips_addq_s_w (q31, q31)
7152 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7153 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7154 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7155 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7156 q31 __builtin_mips_subq_s_w (q31, q31)
7157 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7158 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7159 i32 __builtin_mips_addsc (i32, i32)
7160 i32 __builtin_mips_addwc (i32, i32)
7161 i32 __builtin_mips_modsub (i32, i32)
7162 i32 __builtin_mips_raddu_w_qb (v4i8)
7163 v2q15 __builtin_mips_absq_s_ph (v2q15)
7164 q31 __builtin_mips_absq_s_w (q31)
7165 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7166 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7167 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7168 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7169 q31 __builtin_mips_preceq_w_phl (v2q15)
7170 q31 __builtin_mips_preceq_w_phr (v2q15)
7171 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7172 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7173 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7174 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7175 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7176 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7177 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7178 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7179 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7180 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7181 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7182 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7183 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7184 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7185 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7186 q31 __builtin_mips_shll_s_w (q31, i32)
7187 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7188 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7189 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7190 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7191 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7192 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7193 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7194 q31 __builtin_mips_shra_r_w (q31, i32)
7195 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7196 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7197 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7198 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7199 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7200 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7201 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7202 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7203 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7204 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7205 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7206 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7207 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7208 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7209 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7210 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7211 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7212 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7213 i32 __builtin_mips_bitrev (i32)
7214 i32 __builtin_mips_insv (i32, i32)
7215 v4i8 __builtin_mips_repl_qb (imm0_255)
7216 v4i8 __builtin_mips_repl_qb (i32)
7217 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7218 v2q15 __builtin_mips_repl_ph (i32)
7219 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7220 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7221 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7222 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7223 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7224 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7225 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7226 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7227 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7228 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7229 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7230 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7231 i32 __builtin_mips_extr_w (a64, imm0_31)
7232 i32 __builtin_mips_extr_w (a64, i32)
7233 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7234 i32 __builtin_mips_extr_s_h (a64, i32)
7235 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7236 i32 __builtin_mips_extr_rs_w (a64, i32)
7237 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7238 i32 __builtin_mips_extr_r_w (a64, i32)
7239 i32 __builtin_mips_extp (a64, imm0_31)
7240 i32 __builtin_mips_extp (a64, i32)
7241 i32 __builtin_mips_extpdp (a64, imm0_31)
7242 i32 __builtin_mips_extpdp (a64, i32)
7243 a64 __builtin_mips_shilo (a64, imm_n32_31)
7244 a64 __builtin_mips_shilo (a64, i32)
7245 a64 __builtin_mips_mthlip (a64, i32)
7246 void __builtin_mips_wrdsp (i32, imm0_63)
7247 i32 __builtin_mips_rddsp (imm0_63)
7248 i32 __builtin_mips_lbux (void *, i32)
7249 i32 __builtin_mips_lhx (void *, i32)
7250 i32 __builtin_mips_lwx (void *, i32)
7251 i32 __builtin_mips_bposge32 (void)
7254 @node MIPS Paired-Single Support
7255 @subsection MIPS Paired-Single Support
7257 The MIPS64 architecture includes a number of instructions that
7258 operate on pairs of single-precision floating-point values.
7259 Each pair is packed into a 64-bit floating-point register,
7260 with one element being designated the ``upper half'' and
7261 the other being designated the ``lower half''.
7263 GCC supports paired-single operations using both the generic
7264 vector extensions (@pxref{Vector Extensions}) and a collection of
7265 MIPS-specific built-in functions. Both kinds of support are
7266 enabled by the @option{-mpaired-single} command-line option.
7268 The vector type associated with paired-single values is usually
7269 called @code{v2sf}. It can be defined in C as follows:
7272 typedef float v2sf __attribute__ ((vector_size (8)));
7275 @code{v2sf} values are initialized in the same way as aggregates.
7279 v2sf a = @{1.5, 9.1@};
7282 b = (v2sf) @{e, f@};
7285 @emph{Note:} The CPU's endianness determines which value is stored in
7286 the upper half of a register and which value is stored in the lower half.
7287 On little-endian targets, the first value is the lower one and the second
7288 value is the upper one. The opposite order applies to big-endian targets.
7289 For example, the code above will set the lower half of @code{a} to
7290 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7293 * Paired-Single Arithmetic::
7294 * Paired-Single Built-in Functions::
7295 * MIPS-3D Built-in Functions::
7298 @node Paired-Single Arithmetic
7299 @subsubsection Paired-Single Arithmetic
7301 The table below lists the @code{v2sf} operations for which hardware
7302 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7303 values and @code{x} is an integral value.
7305 @multitable @columnfractions .50 .50
7306 @item C code @tab MIPS instruction
7307 @item @code{a + b} @tab @code{add.ps}
7308 @item @code{a - b} @tab @code{sub.ps}
7309 @item @code{-a} @tab @code{neg.ps}
7310 @item @code{a * b} @tab @code{mul.ps}
7311 @item @code{a * b + c} @tab @code{madd.ps}
7312 @item @code{a * b - c} @tab @code{msub.ps}
7313 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7314 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7315 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7318 Note that the multiply-accumulate instructions can be disabled
7319 using the command-line option @code{-mno-fused-madd}.
7321 @node Paired-Single Built-in Functions
7322 @subsubsection Paired-Single Built-in Functions
7324 The following paired-single functions map directly to a particular
7325 MIPS instruction. Please refer to the architecture specification
7326 for details on what each instruction does.
7329 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7330 Pair lower lower (@code{pll.ps}).
7332 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7333 Pair upper lower (@code{pul.ps}).
7335 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7336 Pair lower upper (@code{plu.ps}).
7338 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7339 Pair upper upper (@code{puu.ps}).
7341 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7342 Convert pair to paired single (@code{cvt.ps.s}).
7344 @item float __builtin_mips_cvt_s_pl (v2sf)
7345 Convert pair lower to single (@code{cvt.s.pl}).
7347 @item float __builtin_mips_cvt_s_pu (v2sf)
7348 Convert pair upper to single (@code{cvt.s.pu}).
7350 @item v2sf __builtin_mips_abs_ps (v2sf)
7351 Absolute value (@code{abs.ps}).
7353 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7354 Align variable (@code{alnv.ps}).
7356 @emph{Note:} The value of the third parameter must be 0 or 4
7357 modulo 8, otherwise the result will be unpredictable. Please read the
7358 instruction description for details.
7361 The following multi-instruction functions are also available.
7362 In each case, @var{cond} can be any of the 16 floating-point conditions:
7363 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7364 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7365 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7368 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7369 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7370 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7371 @code{movt.ps}/@code{movf.ps}).
7373 The @code{movt} functions return the value @var{x} computed by:
7376 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7377 mov.ps @var{x},@var{c}
7378 movt.ps @var{x},@var{d},@var{cc}
7381 The @code{movf} functions are similar but use @code{movf.ps} instead
7384 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7385 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7386 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7387 @code{bc1t}/@code{bc1f}).
7389 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7390 and return either the upper or lower half of the result. For example:
7394 if (__builtin_mips_upper_c_eq_ps (a, b))
7395 upper_halves_are_equal ();
7397 upper_halves_are_unequal ();
7399 if (__builtin_mips_lower_c_eq_ps (a, b))
7400 lower_halves_are_equal ();
7402 lower_halves_are_unequal ();
7406 @node MIPS-3D Built-in Functions
7407 @subsubsection MIPS-3D Built-in Functions
7409 The MIPS-3D Application-Specific Extension (ASE) includes additional
7410 paired-single instructions that are designed to improve the performance
7411 of 3D graphics operations. Support for these instructions is controlled
7412 by the @option{-mips3d} command-line option.
7414 The functions listed below map directly to a particular MIPS-3D
7415 instruction. Please refer to the architecture specification for
7416 more details on what each instruction does.
7419 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7420 Reduction add (@code{addr.ps}).
7422 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7423 Reduction multiply (@code{mulr.ps}).
7425 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7426 Convert paired single to paired word (@code{cvt.pw.ps}).
7428 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7429 Convert paired word to paired single (@code{cvt.ps.pw}).
7431 @item float __builtin_mips_recip1_s (float)
7432 @itemx double __builtin_mips_recip1_d (double)
7433 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7434 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7436 @item float __builtin_mips_recip2_s (float, float)
7437 @itemx double __builtin_mips_recip2_d (double, double)
7438 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7439 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7441 @item float __builtin_mips_rsqrt1_s (float)
7442 @itemx double __builtin_mips_rsqrt1_d (double)
7443 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7444 Reduced precision reciprocal square root (sequence step 1)
7445 (@code{rsqrt1.@var{fmt}}).
7447 @item float __builtin_mips_rsqrt2_s (float, float)
7448 @itemx double __builtin_mips_rsqrt2_d (double, double)
7449 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7450 Reduced precision reciprocal square root (sequence step 2)
7451 (@code{rsqrt2.@var{fmt}}).
7454 The following multi-instruction functions are also available.
7455 In each case, @var{cond} can be any of the 16 floating-point conditions:
7456 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7457 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7458 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7461 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7462 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7463 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7464 @code{bc1t}/@code{bc1f}).
7466 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7467 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7472 if (__builtin_mips_cabs_eq_s (a, b))
7478 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7479 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7480 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7481 @code{bc1t}/@code{bc1f}).
7483 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7484 and return either the upper or lower half of the result. For example:
7488 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7489 upper_halves_are_equal ();
7491 upper_halves_are_unequal ();
7493 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7494 lower_halves_are_equal ();
7496 lower_halves_are_unequal ();
7499 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7500 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7501 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7502 @code{movt.ps}/@code{movf.ps}).
7504 The @code{movt} functions return the value @var{x} computed by:
7507 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7508 mov.ps @var{x},@var{c}
7509 movt.ps @var{x},@var{d},@var{cc}
7512 The @code{movf} functions are similar but use @code{movf.ps} instead
7515 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7516 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7517 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7518 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7519 Comparison of two paired-single values
7520 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7521 @code{bc1any2t}/@code{bc1any2f}).
7523 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7524 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7525 result is true and the @code{all} forms return true if both results are true.
7530 if (__builtin_mips_any_c_eq_ps (a, b))
7535 if (__builtin_mips_all_c_eq_ps (a, b))
7541 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7542 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7543 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7544 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7545 Comparison of four paired-single values
7546 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7547 @code{bc1any4t}/@code{bc1any4f}).
7549 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7550 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7551 The @code{any} forms return true if any of the four results are true
7552 and the @code{all} forms return true if all four results are true.
7557 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7562 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7569 @node PowerPC AltiVec Built-in Functions
7570 @subsection PowerPC AltiVec Built-in Functions
7572 GCC provides an interface for the PowerPC family of processors to access
7573 the AltiVec operations described in Motorola's AltiVec Programming
7574 Interface Manual. The interface is made available by including
7575 @code{<altivec.h>} and using @option{-maltivec} and
7576 @option{-mabi=altivec}. The interface supports the following vector
7580 vector unsigned char
7584 vector unsigned short
7595 GCC's implementation of the high-level language interface available from
7596 C and C++ code differs from Motorola's documentation in several ways.
7601 A vector constant is a list of constant expressions within curly braces.
7604 A vector initializer requires no cast if the vector constant is of the
7605 same type as the variable it is initializing.
7608 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7609 vector type is the default signedness of the base type. The default
7610 varies depending on the operating system, so a portable program should
7611 always specify the signedness.
7614 Compiling with @option{-maltivec} adds keywords @code{__vector},
7615 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7616 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7620 GCC allows using a @code{typedef} name as the type specifier for a
7624 For C, overloaded functions are implemented with macros so the following
7628 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7631 Since @code{vec_add} is a macro, the vector constant in the example
7632 is treated as four separate arguments. Wrap the entire argument in
7633 parentheses for this to work.
7636 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
7637 Internally, GCC uses built-in functions to achieve the functionality in
7638 the aforementioned header file, but they are not supported and are
7639 subject to change without notice.
7641 The following interfaces are supported for the generic and specific
7642 AltiVec operations and the AltiVec predicates. In cases where there
7643 is a direct mapping between generic and specific operations, only the
7644 generic names are shown here, although the specific operations can also
7647 Arguments that are documented as @code{const int} require literal
7648 integral values within the range required for that operation.
7651 vector signed char vec_abs (vector signed char);
7652 vector signed short vec_abs (vector signed short);
7653 vector signed int vec_abs (vector signed int);
7654 vector float vec_abs (vector float);
7656 vector signed char vec_abss (vector signed char);
7657 vector signed short vec_abss (vector signed short);
7658 vector signed int vec_abss (vector signed int);
7660 vector signed char vec_add (vector bool char, vector signed char);
7661 vector signed char vec_add (vector signed char, vector bool char);
7662 vector signed char vec_add (vector signed char, vector signed char);
7663 vector unsigned char vec_add (vector bool char, vector unsigned char);
7664 vector unsigned char vec_add (vector unsigned char, vector bool char);
7665 vector unsigned char vec_add (vector unsigned char,
7666 vector unsigned char);
7667 vector signed short vec_add (vector bool short, vector signed short);
7668 vector signed short vec_add (vector signed short, vector bool short);
7669 vector signed short vec_add (vector signed short, vector signed short);
7670 vector unsigned short vec_add (vector bool short,
7671 vector unsigned short);
7672 vector unsigned short vec_add (vector unsigned short,
7674 vector unsigned short vec_add (vector unsigned short,
7675 vector unsigned short);
7676 vector signed int vec_add (vector bool int, vector signed int);
7677 vector signed int vec_add (vector signed int, vector bool int);
7678 vector signed int vec_add (vector signed int, vector signed int);
7679 vector unsigned int vec_add (vector bool int, vector unsigned int);
7680 vector unsigned int vec_add (vector unsigned int, vector bool int);
7681 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
7682 vector float vec_add (vector float, vector float);
7684 vector float vec_vaddfp (vector float, vector float);
7686 vector signed int vec_vadduwm (vector bool int, vector signed int);
7687 vector signed int vec_vadduwm (vector signed int, vector bool int);
7688 vector signed int vec_vadduwm (vector signed int, vector signed int);
7689 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
7690 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
7691 vector unsigned int vec_vadduwm (vector unsigned int,
7692 vector unsigned int);
7694 vector signed short vec_vadduhm (vector bool short,
7695 vector signed short);
7696 vector signed short vec_vadduhm (vector signed short,
7698 vector signed short vec_vadduhm (vector signed short,
7699 vector signed short);
7700 vector unsigned short vec_vadduhm (vector bool short,
7701 vector unsigned short);
7702 vector unsigned short vec_vadduhm (vector unsigned short,
7704 vector unsigned short vec_vadduhm (vector unsigned short,
7705 vector unsigned short);
7707 vector signed char vec_vaddubm (vector bool char, vector signed char);
7708 vector signed char vec_vaddubm (vector signed char, vector bool char);
7709 vector signed char vec_vaddubm (vector signed char, vector signed char);
7710 vector unsigned char vec_vaddubm (vector bool char,
7711 vector unsigned char);
7712 vector unsigned char vec_vaddubm (vector unsigned char,
7714 vector unsigned char vec_vaddubm (vector unsigned char,
7715 vector unsigned char);
7717 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
7719 vector unsigned char vec_adds (vector bool char, vector unsigned char);
7720 vector unsigned char vec_adds (vector unsigned char, vector bool char);
7721 vector unsigned char vec_adds (vector unsigned char,
7722 vector unsigned char);
7723 vector signed char vec_adds (vector bool char, vector signed char);
7724 vector signed char vec_adds (vector signed char, vector bool char);
7725 vector signed char vec_adds (vector signed char, vector signed char);
7726 vector unsigned short vec_adds (vector bool short,
7727 vector unsigned short);
7728 vector unsigned short vec_adds (vector unsigned short,
7730 vector unsigned short vec_adds (vector unsigned short,
7731 vector unsigned short);
7732 vector signed short vec_adds (vector bool short, vector signed short);
7733 vector signed short vec_adds (vector signed short, vector bool short);
7734 vector signed short vec_adds (vector signed short, vector signed short);
7735 vector unsigned int vec_adds (vector bool int, vector unsigned int);
7736 vector unsigned int vec_adds (vector unsigned int, vector bool int);
7737 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
7738 vector signed int vec_adds (vector bool int, vector signed int);
7739 vector signed int vec_adds (vector signed int, vector bool int);
7740 vector signed int vec_adds (vector signed int, vector signed int);
7742 vector signed int vec_vaddsws (vector bool int, vector signed int);
7743 vector signed int vec_vaddsws (vector signed int, vector bool int);
7744 vector signed int vec_vaddsws (vector signed int, vector signed int);
7746 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
7747 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
7748 vector unsigned int vec_vadduws (vector unsigned int,
7749 vector unsigned int);
7751 vector signed short vec_vaddshs (vector bool short,
7752 vector signed short);
7753 vector signed short vec_vaddshs (vector signed short,
7755 vector signed short vec_vaddshs (vector signed short,
7756 vector signed short);
7758 vector unsigned short vec_vadduhs (vector bool short,
7759 vector unsigned short);
7760 vector unsigned short vec_vadduhs (vector unsigned short,
7762 vector unsigned short vec_vadduhs (vector unsigned short,
7763 vector unsigned short);
7765 vector signed char vec_vaddsbs (vector bool char, vector signed char);
7766 vector signed char vec_vaddsbs (vector signed char, vector bool char);
7767 vector signed char vec_vaddsbs (vector signed char, vector signed char);
7769 vector unsigned char vec_vaddubs (vector bool char,
7770 vector unsigned char);
7771 vector unsigned char vec_vaddubs (vector unsigned char,
7773 vector unsigned char vec_vaddubs (vector unsigned char,
7774 vector unsigned char);
7776 vector float vec_and (vector float, vector float);
7777 vector float vec_and (vector float, vector bool int);
7778 vector float vec_and (vector bool int, vector float);
7779 vector bool int vec_and (vector bool int, vector bool int);
7780 vector signed int vec_and (vector bool int, vector signed int);
7781 vector signed int vec_and (vector signed int, vector bool int);
7782 vector signed int vec_and (vector signed int, vector signed int);
7783 vector unsigned int vec_and (vector bool int, vector unsigned int);
7784 vector unsigned int vec_and (vector unsigned int, vector bool int);
7785 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
7786 vector bool short vec_and (vector bool short, vector bool short);
7787 vector signed short vec_and (vector bool short, vector signed short);
7788 vector signed short vec_and (vector signed short, vector bool short);
7789 vector signed short vec_and (vector signed short, vector signed short);
7790 vector unsigned short vec_and (vector bool short,
7791 vector unsigned short);
7792 vector unsigned short vec_and (vector unsigned short,
7794 vector unsigned short vec_and (vector unsigned short,
7795 vector unsigned short);
7796 vector signed char vec_and (vector bool char, vector signed char);
7797 vector bool char vec_and (vector bool char, vector bool char);
7798 vector signed char vec_and (vector signed char, vector bool char);
7799 vector signed char vec_and (vector signed char, vector signed char);
7800 vector unsigned char vec_and (vector bool char, vector unsigned char);
7801 vector unsigned char vec_and (vector unsigned char, vector bool char);
7802 vector unsigned char vec_and (vector unsigned char,
7803 vector unsigned char);
7805 vector float vec_andc (vector float, vector float);
7806 vector float vec_andc (vector float, vector bool int);
7807 vector float vec_andc (vector bool int, vector float);
7808 vector bool int vec_andc (vector bool int, vector bool int);
7809 vector signed int vec_andc (vector bool int, vector signed int);
7810 vector signed int vec_andc (vector signed int, vector bool int);
7811 vector signed int vec_andc (vector signed int, vector signed int);
7812 vector unsigned int vec_andc (vector bool int, vector unsigned int);
7813 vector unsigned int vec_andc (vector unsigned int, vector bool int);
7814 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
7815 vector bool short vec_andc (vector bool short, vector bool short);
7816 vector signed short vec_andc (vector bool short, vector signed short);
7817 vector signed short vec_andc (vector signed short, vector bool short);
7818 vector signed short vec_andc (vector signed short, vector signed short);
7819 vector unsigned short vec_andc (vector bool short,
7820 vector unsigned short);
7821 vector unsigned short vec_andc (vector unsigned short,
7823 vector unsigned short vec_andc (vector unsigned short,
7824 vector unsigned short);
7825 vector signed char vec_andc (vector bool char, vector signed char);
7826 vector bool char vec_andc (vector bool char, vector bool char);
7827 vector signed char vec_andc (vector signed char, vector bool char);
7828 vector signed char vec_andc (vector signed char, vector signed char);
7829 vector unsigned char vec_andc (vector bool char, vector unsigned char);
7830 vector unsigned char vec_andc (vector unsigned char, vector bool char);
7831 vector unsigned char vec_andc (vector unsigned char,
7832 vector unsigned char);
7834 vector unsigned char vec_avg (vector unsigned char,
7835 vector unsigned char);
7836 vector signed char vec_avg (vector signed char, vector signed char);
7837 vector unsigned short vec_avg (vector unsigned short,
7838 vector unsigned short);
7839 vector signed short vec_avg (vector signed short, vector signed short);
7840 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
7841 vector signed int vec_avg (vector signed int, vector signed int);
7843 vector signed int vec_vavgsw (vector signed int, vector signed int);
7845 vector unsigned int vec_vavguw (vector unsigned int,
7846 vector unsigned int);
7848 vector signed short vec_vavgsh (vector signed short,
7849 vector signed short);
7851 vector unsigned short vec_vavguh (vector unsigned short,
7852 vector unsigned short);
7854 vector signed char vec_vavgsb (vector signed char, vector signed char);
7856 vector unsigned char vec_vavgub (vector unsigned char,
7857 vector unsigned char);
7859 vector float vec_ceil (vector float);
7861 vector signed int vec_cmpb (vector float, vector float);
7863 vector bool char vec_cmpeq (vector signed char, vector signed char);
7864 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
7865 vector bool short vec_cmpeq (vector signed short, vector signed short);
7866 vector bool short vec_cmpeq (vector unsigned short,
7867 vector unsigned short);
7868 vector bool int vec_cmpeq (vector signed int, vector signed int);
7869 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
7870 vector bool int vec_cmpeq (vector float, vector float);
7872 vector bool int vec_vcmpeqfp (vector float, vector float);
7874 vector bool int vec_vcmpequw (vector signed int, vector signed int);
7875 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
7877 vector bool short vec_vcmpequh (vector signed short,
7878 vector signed short);
7879 vector bool short vec_vcmpequh (vector unsigned short,
7880 vector unsigned short);
7882 vector bool char vec_vcmpequb (vector signed char, vector signed char);
7883 vector bool char vec_vcmpequb (vector unsigned char,
7884 vector unsigned char);
7886 vector bool int vec_cmpge (vector float, vector float);
7888 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
7889 vector bool char vec_cmpgt (vector signed char, vector signed char);
7890 vector bool short vec_cmpgt (vector unsigned short,
7891 vector unsigned short);
7892 vector bool short vec_cmpgt (vector signed short, vector signed short);
7893 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
7894 vector bool int vec_cmpgt (vector signed int, vector signed int);
7895 vector bool int vec_cmpgt (vector float, vector float);
7897 vector bool int vec_vcmpgtfp (vector float, vector float);
7899 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
7901 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
7903 vector bool short vec_vcmpgtsh (vector signed short,
7904 vector signed short);
7906 vector bool short vec_vcmpgtuh (vector unsigned short,
7907 vector unsigned short);
7909 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
7911 vector bool char vec_vcmpgtub (vector unsigned char,
7912 vector unsigned char);
7914 vector bool int vec_cmple (vector float, vector float);
7916 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
7917 vector bool char vec_cmplt (vector signed char, vector signed char);
7918 vector bool short vec_cmplt (vector unsigned short,
7919 vector unsigned short);
7920 vector bool short vec_cmplt (vector signed short, vector signed short);
7921 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
7922 vector bool int vec_cmplt (vector signed int, vector signed int);
7923 vector bool int vec_cmplt (vector float, vector float);
7925 vector float vec_ctf (vector unsigned int, const int);
7926 vector float vec_ctf (vector signed int, const int);
7928 vector float vec_vcfsx (vector signed int, const int);
7930 vector float vec_vcfux (vector unsigned int, const int);
7932 vector signed int vec_cts (vector float, const int);
7934 vector unsigned int vec_ctu (vector float, const int);
7936 void vec_dss (const int);
7938 void vec_dssall (void);
7940 void vec_dst (const vector unsigned char *, int, const int);
7941 void vec_dst (const vector signed char *, int, const int);
7942 void vec_dst (const vector bool char *, int, const int);
7943 void vec_dst (const vector unsigned short *, int, const int);
7944 void vec_dst (const vector signed short *, int, const int);
7945 void vec_dst (const vector bool short *, int, const int);
7946 void vec_dst (const vector pixel *, int, const int);
7947 void vec_dst (const vector unsigned int *, int, const int);
7948 void vec_dst (const vector signed int *, int, const int);
7949 void vec_dst (const vector bool int *, int, const int);
7950 void vec_dst (const vector float *, int, const int);
7951 void vec_dst (const unsigned char *, int, const int);
7952 void vec_dst (const signed char *, int, const int);
7953 void vec_dst (const unsigned short *, int, const int);
7954 void vec_dst (const short *, int, const int);
7955 void vec_dst (const unsigned int *, int, const int);
7956 void vec_dst (const int *, int, const int);
7957 void vec_dst (const unsigned long *, int, const int);
7958 void vec_dst (const long *, int, const int);
7959 void vec_dst (const float *, int, const int);
7961 void vec_dstst (const vector unsigned char *, int, const int);
7962 void vec_dstst (const vector signed char *, int, const int);
7963 void vec_dstst (const vector bool char *, int, const int);
7964 void vec_dstst (const vector unsigned short *, int, const int);
7965 void vec_dstst (const vector signed short *, int, const int);
7966 void vec_dstst (const vector bool short *, int, const int);
7967 void vec_dstst (const vector pixel *, int, const int);
7968 void vec_dstst (const vector unsigned int *, int, const int);
7969 void vec_dstst (const vector signed int *, int, const int);
7970 void vec_dstst (const vector bool int *, int, const int);
7971 void vec_dstst (const vector float *, int, const int);
7972 void vec_dstst (const unsigned char *, int, const int);
7973 void vec_dstst (const signed char *, int, const int);
7974 void vec_dstst (const unsigned short *, int, const int);
7975 void vec_dstst (const short *, int, const int);
7976 void vec_dstst (const unsigned int *, int, const int);
7977 void vec_dstst (const int *, int, const int);
7978 void vec_dstst (const unsigned long *, int, const int);
7979 void vec_dstst (const long *, int, const int);
7980 void vec_dstst (const float *, int, const int);
7982 void vec_dststt (const vector unsigned char *, int, const int);
7983 void vec_dststt (const vector signed char *, int, const int);
7984 void vec_dststt (const vector bool char *, int, const int);
7985 void vec_dststt (const vector unsigned short *, int, const int);
7986 void vec_dststt (const vector signed short *, int, const int);
7987 void vec_dststt (const vector bool short *, int, const int);
7988 void vec_dststt (const vector pixel *, int, const int);
7989 void vec_dststt (const vector unsigned int *, int, const int);
7990 void vec_dststt (const vector signed int *, int, const int);
7991 void vec_dststt (const vector bool int *, int, const int);
7992 void vec_dststt (const vector float *, int, const int);
7993 void vec_dststt (const unsigned char *, int, const int);
7994 void vec_dststt (const signed char *, int, const int);
7995 void vec_dststt (const unsigned short *, int, const int);
7996 void vec_dststt (const short *, int, const int);
7997 void vec_dststt (const unsigned int *, int, const int);
7998 void vec_dststt (const int *, int, const int);
7999 void vec_dststt (const unsigned long *, int, const int);
8000 void vec_dststt (const long *, int, const int);
8001 void vec_dststt (const float *, int, const int);
8003 void vec_dstt (const vector unsigned char *, int, const int);
8004 void vec_dstt (const vector signed char *, int, const int);
8005 void vec_dstt (const vector bool char *, int, const int);
8006 void vec_dstt (const vector unsigned short *, int, const int);
8007 void vec_dstt (const vector signed short *, int, const int);
8008 void vec_dstt (const vector bool short *, int, const int);
8009 void vec_dstt (const vector pixel *, int, const int);
8010 void vec_dstt (const vector unsigned int *, int, const int);
8011 void vec_dstt (const vector signed int *, int, const int);
8012 void vec_dstt (const vector bool int *, int, const int);
8013 void vec_dstt (const vector float *, int, const int);
8014 void vec_dstt (const unsigned char *, int, const int);
8015 void vec_dstt (const signed char *, int, const int);
8016 void vec_dstt (const unsigned short *, int, const int);
8017 void vec_dstt (const short *, int, const int);
8018 void vec_dstt (const unsigned int *, int, const int);
8019 void vec_dstt (const int *, int, const int);
8020 void vec_dstt (const unsigned long *, int, const int);
8021 void vec_dstt (const long *, int, const int);
8022 void vec_dstt (const float *, int, const int);
8024 vector float vec_expte (vector float);
8026 vector float vec_floor (vector float);
8028 vector float vec_ld (int, const vector float *);
8029 vector float vec_ld (int, const float *);
8030 vector bool int vec_ld (int, const vector bool int *);
8031 vector signed int vec_ld (int, const vector signed int *);
8032 vector signed int vec_ld (int, const int *);
8033 vector signed int vec_ld (int, const long *);
8034 vector unsigned int vec_ld (int, const vector unsigned int *);
8035 vector unsigned int vec_ld (int, const unsigned int *);
8036 vector unsigned int vec_ld (int, const unsigned long *);
8037 vector bool short vec_ld (int, const vector bool short *);
8038 vector pixel vec_ld (int, const vector pixel *);
8039 vector signed short vec_ld (int, const vector signed short *);
8040 vector signed short vec_ld (int, const short *);
8041 vector unsigned short vec_ld (int, const vector unsigned short *);
8042 vector unsigned short vec_ld (int, const unsigned short *);
8043 vector bool char vec_ld (int, const vector bool char *);
8044 vector signed char vec_ld (int, const vector signed char *);
8045 vector signed char vec_ld (int, const signed char *);
8046 vector unsigned char vec_ld (int, const vector unsigned char *);
8047 vector unsigned char vec_ld (int, const unsigned char *);
8049 vector signed char vec_lde (int, const signed char *);
8050 vector unsigned char vec_lde (int, const unsigned char *);
8051 vector signed short vec_lde (int, const short *);
8052 vector unsigned short vec_lde (int, const unsigned short *);
8053 vector float vec_lde (int, const float *);
8054 vector signed int vec_lde (int, const int *);
8055 vector unsigned int vec_lde (int, const unsigned int *);
8056 vector signed int vec_lde (int, const long *);
8057 vector unsigned int vec_lde (int, const unsigned long *);
8059 vector float vec_lvewx (int, float *);
8060 vector signed int vec_lvewx (int, int *);
8061 vector unsigned int vec_lvewx (int, unsigned int *);
8062 vector signed int vec_lvewx (int, long *);
8063 vector unsigned int vec_lvewx (int, unsigned long *);
8065 vector signed short vec_lvehx (int, short *);
8066 vector unsigned short vec_lvehx (int, unsigned short *);
8068 vector signed char vec_lvebx (int, char *);
8069 vector unsigned char vec_lvebx (int, unsigned char *);
8071 vector float vec_ldl (int, const vector float *);
8072 vector float vec_ldl (int, const float *);
8073 vector bool int vec_ldl (int, const vector bool int *);
8074 vector signed int vec_ldl (int, const vector signed int *);
8075 vector signed int vec_ldl (int, const int *);
8076 vector signed int vec_ldl (int, const long *);
8077 vector unsigned int vec_ldl (int, const vector unsigned int *);
8078 vector unsigned int vec_ldl (int, const unsigned int *);
8079 vector unsigned int vec_ldl (int, const unsigned long *);
8080 vector bool short vec_ldl (int, const vector bool short *);
8081 vector pixel vec_ldl (int, const vector pixel *);
8082 vector signed short vec_ldl (int, const vector signed short *);
8083 vector signed short vec_ldl (int, const short *);
8084 vector unsigned short vec_ldl (int, const vector unsigned short *);
8085 vector unsigned short vec_ldl (int, const unsigned short *);
8086 vector bool char vec_ldl (int, const vector bool char *);
8087 vector signed char vec_ldl (int, const vector signed char *);
8088 vector signed char vec_ldl (int, const signed char *);
8089 vector unsigned char vec_ldl (int, const vector unsigned char *);
8090 vector unsigned char vec_ldl (int, const unsigned char *);
8092 vector float vec_loge (vector float);
8094 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8095 vector unsigned char vec_lvsl (int, const volatile signed char *);
8096 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8097 vector unsigned char vec_lvsl (int, const volatile short *);
8098 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8099 vector unsigned char vec_lvsl (int, const volatile int *);
8100 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8101 vector unsigned char vec_lvsl (int, const volatile long *);
8102 vector unsigned char vec_lvsl (int, const volatile float *);
8104 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8105 vector unsigned char vec_lvsr (int, const volatile signed char *);
8106 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8107 vector unsigned char vec_lvsr (int, const volatile short *);
8108 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8109 vector unsigned char vec_lvsr (int, const volatile int *);
8110 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8111 vector unsigned char vec_lvsr (int, const volatile long *);
8112 vector unsigned char vec_lvsr (int, const volatile float *);
8114 vector float vec_madd (vector float, vector float, vector float);
8116 vector signed short vec_madds (vector signed short,
8117 vector signed short,
8118 vector signed short);
8120 vector unsigned char vec_max (vector bool char, vector unsigned char);
8121 vector unsigned char vec_max (vector unsigned char, vector bool char);
8122 vector unsigned char vec_max (vector unsigned char,
8123 vector unsigned char);
8124 vector signed char vec_max (vector bool char, vector signed char);
8125 vector signed char vec_max (vector signed char, vector bool char);
8126 vector signed char vec_max (vector signed char, vector signed char);
8127 vector unsigned short vec_max (vector bool short,
8128 vector unsigned short);
8129 vector unsigned short vec_max (vector unsigned short,
8131 vector unsigned short vec_max (vector unsigned short,
8132 vector unsigned short);
8133 vector signed short vec_max (vector bool short, vector signed short);
8134 vector signed short vec_max (vector signed short, vector bool short);
8135 vector signed short vec_max (vector signed short, vector signed short);
8136 vector unsigned int vec_max (vector bool int, vector unsigned int);
8137 vector unsigned int vec_max (vector unsigned int, vector bool int);
8138 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8139 vector signed int vec_max (vector bool int, vector signed int);
8140 vector signed int vec_max (vector signed int, vector bool int);
8141 vector signed int vec_max (vector signed int, vector signed int);
8142 vector float vec_max (vector float, vector float);
8144 vector float vec_vmaxfp (vector float, vector float);
8146 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8147 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8148 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8150 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8151 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8152 vector unsigned int vec_vmaxuw (vector unsigned int,
8153 vector unsigned int);
8155 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8156 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8157 vector signed short vec_vmaxsh (vector signed short,
8158 vector signed short);
8160 vector unsigned short vec_vmaxuh (vector bool short,
8161 vector unsigned short);
8162 vector unsigned short vec_vmaxuh (vector unsigned short,
8164 vector unsigned short vec_vmaxuh (vector unsigned short,
8165 vector unsigned short);
8167 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8168 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8169 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8171 vector unsigned char vec_vmaxub (vector bool char,
8172 vector unsigned char);
8173 vector unsigned char vec_vmaxub (vector unsigned char,
8175 vector unsigned char vec_vmaxub (vector unsigned char,
8176 vector unsigned char);
8178 vector bool char vec_mergeh (vector bool char, vector bool char);
8179 vector signed char vec_mergeh (vector signed char, vector signed char);
8180 vector unsigned char vec_mergeh (vector unsigned char,
8181 vector unsigned char);
8182 vector bool short vec_mergeh (vector bool short, vector bool short);
8183 vector pixel vec_mergeh (vector pixel, vector pixel);
8184 vector signed short vec_mergeh (vector signed short,
8185 vector signed short);
8186 vector unsigned short vec_mergeh (vector unsigned short,
8187 vector unsigned short);
8188 vector float vec_mergeh (vector float, vector float);
8189 vector bool int vec_mergeh (vector bool int, vector bool int);
8190 vector signed int vec_mergeh (vector signed int, vector signed int);
8191 vector unsigned int vec_mergeh (vector unsigned int,
8192 vector unsigned int);
8194 vector float vec_vmrghw (vector float, vector float);
8195 vector bool int vec_vmrghw (vector bool int, vector bool int);
8196 vector signed int vec_vmrghw (vector signed int, vector signed int);
8197 vector unsigned int vec_vmrghw (vector unsigned int,
8198 vector unsigned int);
8200 vector bool short vec_vmrghh (vector bool short, vector bool short);
8201 vector signed short vec_vmrghh (vector signed short,
8202 vector signed short);
8203 vector unsigned short vec_vmrghh (vector unsigned short,
8204 vector unsigned short);
8205 vector pixel vec_vmrghh (vector pixel, vector pixel);
8207 vector bool char vec_vmrghb (vector bool char, vector bool char);
8208 vector signed char vec_vmrghb (vector signed char, vector signed char);
8209 vector unsigned char vec_vmrghb (vector unsigned char,
8210 vector unsigned char);
8212 vector bool char vec_mergel (vector bool char, vector bool char);
8213 vector signed char vec_mergel (vector signed char, vector signed char);
8214 vector unsigned char vec_mergel (vector unsigned char,
8215 vector unsigned char);
8216 vector bool short vec_mergel (vector bool short, vector bool short);
8217 vector pixel vec_mergel (vector pixel, vector pixel);
8218 vector signed short vec_mergel (vector signed short,
8219 vector signed short);
8220 vector unsigned short vec_mergel (vector unsigned short,
8221 vector unsigned short);
8222 vector float vec_mergel (vector float, vector float);
8223 vector bool int vec_mergel (vector bool int, vector bool int);
8224 vector signed int vec_mergel (vector signed int, vector signed int);
8225 vector unsigned int vec_mergel (vector unsigned int,
8226 vector unsigned int);
8228 vector float vec_vmrglw (vector float, vector float);
8229 vector signed int vec_vmrglw (vector signed int, vector signed int);
8230 vector unsigned int vec_vmrglw (vector unsigned int,
8231 vector unsigned int);
8232 vector bool int vec_vmrglw (vector bool int, vector bool int);
8234 vector bool short vec_vmrglh (vector bool short, vector bool short);
8235 vector signed short vec_vmrglh (vector signed short,
8236 vector signed short);
8237 vector unsigned short vec_vmrglh (vector unsigned short,
8238 vector unsigned short);
8239 vector pixel vec_vmrglh (vector pixel, vector pixel);
8241 vector bool char vec_vmrglb (vector bool char, vector bool char);
8242 vector signed char vec_vmrglb (vector signed char, vector signed char);
8243 vector unsigned char vec_vmrglb (vector unsigned char,
8244 vector unsigned char);
8246 vector unsigned short vec_mfvscr (void);
8248 vector unsigned char vec_min (vector bool char, vector unsigned char);
8249 vector unsigned char vec_min (vector unsigned char, vector bool char);
8250 vector unsigned char vec_min (vector unsigned char,
8251 vector unsigned char);
8252 vector signed char vec_min (vector bool char, vector signed char);
8253 vector signed char vec_min (vector signed char, vector bool char);
8254 vector signed char vec_min (vector signed char, vector signed char);
8255 vector unsigned short vec_min (vector bool short,
8256 vector unsigned short);
8257 vector unsigned short vec_min (vector unsigned short,
8259 vector unsigned short vec_min (vector unsigned short,
8260 vector unsigned short);
8261 vector signed short vec_min (vector bool short, vector signed short);
8262 vector signed short vec_min (vector signed short, vector bool short);
8263 vector signed short vec_min (vector signed short, vector signed short);
8264 vector unsigned int vec_min (vector bool int, vector unsigned int);
8265 vector unsigned int vec_min (vector unsigned int, vector bool int);
8266 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8267 vector signed int vec_min (vector bool int, vector signed int);
8268 vector signed int vec_min (vector signed int, vector bool int);
8269 vector signed int vec_min (vector signed int, vector signed int);
8270 vector float vec_min (vector float, vector float);
8272 vector float vec_vminfp (vector float, vector float);
8274 vector signed int vec_vminsw (vector bool int, vector signed int);
8275 vector signed int vec_vminsw (vector signed int, vector bool int);
8276 vector signed int vec_vminsw (vector signed int, vector signed int);
8278 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8279 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8280 vector unsigned int vec_vminuw (vector unsigned int,
8281 vector unsigned int);
8283 vector signed short vec_vminsh (vector bool short, vector signed short);
8284 vector signed short vec_vminsh (vector signed short, vector bool short);
8285 vector signed short vec_vminsh (vector signed short,
8286 vector signed short);
8288 vector unsigned short vec_vminuh (vector bool short,
8289 vector unsigned short);
8290 vector unsigned short vec_vminuh (vector unsigned short,
8292 vector unsigned short vec_vminuh (vector unsigned short,
8293 vector unsigned short);
8295 vector signed char vec_vminsb (vector bool char, vector signed char);
8296 vector signed char vec_vminsb (vector signed char, vector bool char);
8297 vector signed char vec_vminsb (vector signed char, vector signed char);
8299 vector unsigned char vec_vminub (vector bool char,
8300 vector unsigned char);
8301 vector unsigned char vec_vminub (vector unsigned char,
8303 vector unsigned char vec_vminub (vector unsigned char,
8304 vector unsigned char);
8306 vector signed short vec_mladd (vector signed short,
8307 vector signed short,
8308 vector signed short);
8309 vector signed short vec_mladd (vector signed short,
8310 vector unsigned short,
8311 vector unsigned short);
8312 vector signed short vec_mladd (vector unsigned short,
8313 vector signed short,
8314 vector signed short);
8315 vector unsigned short vec_mladd (vector unsigned short,
8316 vector unsigned short,
8317 vector unsigned short);
8319 vector signed short vec_mradds (vector signed short,
8320 vector signed short,
8321 vector signed short);
8323 vector unsigned int vec_msum (vector unsigned char,
8324 vector unsigned char,
8325 vector unsigned int);
8326 vector signed int vec_msum (vector signed char,
8327 vector unsigned char,
8329 vector unsigned int vec_msum (vector unsigned short,
8330 vector unsigned short,
8331 vector unsigned int);
8332 vector signed int vec_msum (vector signed short,
8333 vector signed short,
8336 vector signed int vec_vmsumshm (vector signed short,
8337 vector signed short,
8340 vector unsigned int vec_vmsumuhm (vector unsigned short,
8341 vector unsigned short,
8342 vector unsigned int);
8344 vector signed int vec_vmsummbm (vector signed char,
8345 vector unsigned char,
8348 vector unsigned int vec_vmsumubm (vector unsigned char,
8349 vector unsigned char,
8350 vector unsigned int);
8352 vector unsigned int vec_msums (vector unsigned short,
8353 vector unsigned short,
8354 vector unsigned int);
8355 vector signed int vec_msums (vector signed short,
8356 vector signed short,
8359 vector signed int vec_vmsumshs (vector signed short,
8360 vector signed short,
8363 vector unsigned int vec_vmsumuhs (vector unsigned short,
8364 vector unsigned short,
8365 vector unsigned int);
8367 void vec_mtvscr (vector signed int);
8368 void vec_mtvscr (vector unsigned int);
8369 void vec_mtvscr (vector bool int);
8370 void vec_mtvscr (vector signed short);
8371 void vec_mtvscr (vector unsigned short);
8372 void vec_mtvscr (vector bool short);
8373 void vec_mtvscr (vector pixel);
8374 void vec_mtvscr (vector signed char);
8375 void vec_mtvscr (vector unsigned char);
8376 void vec_mtvscr (vector bool char);
8378 vector unsigned short vec_mule (vector unsigned char,
8379 vector unsigned char);
8380 vector signed short vec_mule (vector signed char,
8381 vector signed char);
8382 vector unsigned int vec_mule (vector unsigned short,
8383 vector unsigned short);
8384 vector signed int vec_mule (vector signed short, vector signed short);
8386 vector signed int vec_vmulesh (vector signed short,
8387 vector signed short);
8389 vector unsigned int vec_vmuleuh (vector unsigned short,
8390 vector unsigned short);
8392 vector signed short vec_vmulesb (vector signed char,
8393 vector signed char);
8395 vector unsigned short vec_vmuleub (vector unsigned char,
8396 vector unsigned char);
8398 vector unsigned short vec_mulo (vector unsigned char,
8399 vector unsigned char);
8400 vector signed short vec_mulo (vector signed char, vector signed char);
8401 vector unsigned int vec_mulo (vector unsigned short,
8402 vector unsigned short);
8403 vector signed int vec_mulo (vector signed short, vector signed short);
8405 vector signed int vec_vmulosh (vector signed short,
8406 vector signed short);
8408 vector unsigned int vec_vmulouh (vector unsigned short,
8409 vector unsigned short);
8411 vector signed short vec_vmulosb (vector signed char,
8412 vector signed char);
8414 vector unsigned short vec_vmuloub (vector unsigned char,
8415 vector unsigned char);
8417 vector float vec_nmsub (vector float, vector float, vector float);
8419 vector float vec_nor (vector float, vector float);
8420 vector signed int vec_nor (vector signed int, vector signed int);
8421 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8422 vector bool int vec_nor (vector bool int, vector bool int);
8423 vector signed short vec_nor (vector signed short, vector signed short);
8424 vector unsigned short vec_nor (vector unsigned short,
8425 vector unsigned short);
8426 vector bool short vec_nor (vector bool short, vector bool short);
8427 vector signed char vec_nor (vector signed char, vector signed char);
8428 vector unsigned char vec_nor (vector unsigned char,
8429 vector unsigned char);
8430 vector bool char vec_nor (vector bool char, vector bool char);
8432 vector float vec_or (vector float, vector float);
8433 vector float vec_or (vector float, vector bool int);
8434 vector float vec_or (vector bool int, vector float);
8435 vector bool int vec_or (vector bool int, vector bool int);
8436 vector signed int vec_or (vector bool int, vector signed int);
8437 vector signed int vec_or (vector signed int, vector bool int);
8438 vector signed int vec_or (vector signed int, vector signed int);
8439 vector unsigned int vec_or (vector bool int, vector unsigned int);
8440 vector unsigned int vec_or (vector unsigned int, vector bool int);
8441 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8442 vector bool short vec_or (vector bool short, vector bool short);
8443 vector signed short vec_or (vector bool short, vector signed short);
8444 vector signed short vec_or (vector signed short, vector bool short);
8445 vector signed short vec_or (vector signed short, vector signed short);
8446 vector unsigned short vec_or (vector bool short, vector unsigned short);
8447 vector unsigned short vec_or (vector unsigned short, vector bool short);
8448 vector unsigned short vec_or (vector unsigned short,
8449 vector unsigned short);
8450 vector signed char vec_or (vector bool char, vector signed char);
8451 vector bool char vec_or (vector bool char, vector bool char);
8452 vector signed char vec_or (vector signed char, vector bool char);
8453 vector signed char vec_or (vector signed char, vector signed char);
8454 vector unsigned char vec_or (vector bool char, vector unsigned char);
8455 vector unsigned char vec_or (vector unsigned char, vector bool char);
8456 vector unsigned char vec_or (vector unsigned char,
8457 vector unsigned char);
8459 vector signed char vec_pack (vector signed short, vector signed short);
8460 vector unsigned char vec_pack (vector unsigned short,
8461 vector unsigned short);
8462 vector bool char vec_pack (vector bool short, vector bool short);
8463 vector signed short vec_pack (vector signed int, vector signed int);
8464 vector unsigned short vec_pack (vector unsigned int,
8465 vector unsigned int);
8466 vector bool short vec_pack (vector bool int, vector bool int);
8468 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8469 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8470 vector unsigned short vec_vpkuwum (vector unsigned int,
8471 vector unsigned int);
8473 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8474 vector signed char vec_vpkuhum (vector signed short,
8475 vector signed short);
8476 vector unsigned char vec_vpkuhum (vector unsigned short,
8477 vector unsigned short);
8479 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8481 vector unsigned char vec_packs (vector unsigned short,
8482 vector unsigned short);
8483 vector signed char vec_packs (vector signed short, vector signed short);
8484 vector unsigned short vec_packs (vector unsigned int,
8485 vector unsigned int);
8486 vector signed short vec_packs (vector signed int, vector signed int);
8488 vector signed short vec_vpkswss (vector signed int, vector signed int);
8490 vector unsigned short vec_vpkuwus (vector unsigned int,
8491 vector unsigned int);
8493 vector signed char vec_vpkshss (vector signed short,
8494 vector signed short);
8496 vector unsigned char vec_vpkuhus (vector unsigned short,
8497 vector unsigned short);
8499 vector unsigned char vec_packsu (vector unsigned short,
8500 vector unsigned short);
8501 vector unsigned char vec_packsu (vector signed short,
8502 vector signed short);
8503 vector unsigned short vec_packsu (vector unsigned int,
8504 vector unsigned int);
8505 vector unsigned short vec_packsu (vector signed int, vector signed int);
8507 vector unsigned short vec_vpkswus (vector signed int,
8510 vector unsigned char vec_vpkshus (vector signed short,
8511 vector signed short);
8513 vector float vec_perm (vector float,
8515 vector unsigned char);
8516 vector signed int vec_perm (vector signed int,
8518 vector unsigned char);
8519 vector unsigned int vec_perm (vector unsigned int,
8520 vector unsigned int,
8521 vector unsigned char);
8522 vector bool int vec_perm (vector bool int,
8524 vector unsigned char);
8525 vector signed short vec_perm (vector signed short,
8526 vector signed short,
8527 vector unsigned char);
8528 vector unsigned short vec_perm (vector unsigned short,
8529 vector unsigned short,
8530 vector unsigned char);
8531 vector bool short vec_perm (vector bool short,
8533 vector unsigned char);
8534 vector pixel vec_perm (vector pixel,
8536 vector unsigned char);
8537 vector signed char vec_perm (vector signed char,
8539 vector unsigned char);
8540 vector unsigned char vec_perm (vector unsigned char,
8541 vector unsigned char,
8542 vector unsigned char);
8543 vector bool char vec_perm (vector bool char,
8545 vector unsigned char);
8547 vector float vec_re (vector float);
8549 vector signed char vec_rl (vector signed char,
8550 vector unsigned char);
8551 vector unsigned char vec_rl (vector unsigned char,
8552 vector unsigned char);
8553 vector signed short vec_rl (vector signed short, vector unsigned short);
8554 vector unsigned short vec_rl (vector unsigned short,
8555 vector unsigned short);
8556 vector signed int vec_rl (vector signed int, vector unsigned int);
8557 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8559 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8560 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8562 vector signed short vec_vrlh (vector signed short,
8563 vector unsigned short);
8564 vector unsigned short vec_vrlh (vector unsigned short,
8565 vector unsigned short);
8567 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8568 vector unsigned char vec_vrlb (vector unsigned char,
8569 vector unsigned char);
8571 vector float vec_round (vector float);
8573 vector float vec_rsqrte (vector float);
8575 vector float vec_sel (vector float, vector float, vector bool int);
8576 vector float vec_sel (vector float, vector float, vector unsigned int);
8577 vector signed int vec_sel (vector signed int,
8580 vector signed int vec_sel (vector signed int,
8582 vector unsigned int);
8583 vector unsigned int vec_sel (vector unsigned int,
8584 vector unsigned int,
8586 vector unsigned int vec_sel (vector unsigned int,
8587 vector unsigned int,
8588 vector unsigned int);
8589 vector bool int vec_sel (vector bool int,
8592 vector bool int vec_sel (vector bool int,
8594 vector unsigned int);
8595 vector signed short vec_sel (vector signed short,
8596 vector signed short,
8598 vector signed short vec_sel (vector signed short,
8599 vector signed short,
8600 vector unsigned short);
8601 vector unsigned short vec_sel (vector unsigned short,
8602 vector unsigned short,
8604 vector unsigned short vec_sel (vector unsigned short,
8605 vector unsigned short,
8606 vector unsigned short);
8607 vector bool short vec_sel (vector bool short,
8610 vector bool short vec_sel (vector bool short,
8612 vector unsigned short);
8613 vector signed char vec_sel (vector signed char,
8616 vector signed char vec_sel (vector signed char,
8618 vector unsigned char);
8619 vector unsigned char vec_sel (vector unsigned char,
8620 vector unsigned char,
8622 vector unsigned char vec_sel (vector unsigned char,
8623 vector unsigned char,
8624 vector unsigned char);
8625 vector bool char vec_sel (vector bool char,
8628 vector bool char vec_sel (vector bool char,
8630 vector unsigned char);
8632 vector signed char vec_sl (vector signed char,
8633 vector unsigned char);
8634 vector unsigned char vec_sl (vector unsigned char,
8635 vector unsigned char);
8636 vector signed short vec_sl (vector signed short, vector unsigned short);
8637 vector unsigned short vec_sl (vector unsigned short,
8638 vector unsigned short);
8639 vector signed int vec_sl (vector signed int, vector unsigned int);
8640 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
8642 vector signed int vec_vslw (vector signed int, vector unsigned int);
8643 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
8645 vector signed short vec_vslh (vector signed short,
8646 vector unsigned short);
8647 vector unsigned short vec_vslh (vector unsigned short,
8648 vector unsigned short);
8650 vector signed char vec_vslb (vector signed char, vector unsigned char);
8651 vector unsigned char vec_vslb (vector unsigned char,
8652 vector unsigned char);
8654 vector float vec_sld (vector float, vector float, const int);
8655 vector signed int vec_sld (vector signed int,
8658 vector unsigned int vec_sld (vector unsigned int,
8659 vector unsigned int,
8661 vector bool int vec_sld (vector bool int,
8664 vector signed short vec_sld (vector signed short,
8665 vector signed short,
8667 vector unsigned short vec_sld (vector unsigned short,
8668 vector unsigned short,
8670 vector bool short vec_sld (vector bool short,
8673 vector pixel vec_sld (vector pixel,
8676 vector signed char vec_sld (vector signed char,
8679 vector unsigned char vec_sld (vector unsigned char,
8680 vector unsigned char,
8682 vector bool char vec_sld (vector bool char,
8686 vector signed int vec_sll (vector signed int,
8687 vector unsigned int);
8688 vector signed int vec_sll (vector signed int,
8689 vector unsigned short);
8690 vector signed int vec_sll (vector signed int,
8691 vector unsigned char);
8692 vector unsigned int vec_sll (vector unsigned int,
8693 vector unsigned int);
8694 vector unsigned int vec_sll (vector unsigned int,
8695 vector unsigned short);
8696 vector unsigned int vec_sll (vector unsigned int,
8697 vector unsigned char);
8698 vector bool int vec_sll (vector bool int,
8699 vector unsigned int);
8700 vector bool int vec_sll (vector bool int,
8701 vector unsigned short);
8702 vector bool int vec_sll (vector bool int,
8703 vector unsigned char);
8704 vector signed short vec_sll (vector signed short,
8705 vector unsigned int);
8706 vector signed short vec_sll (vector signed short,
8707 vector unsigned short);
8708 vector signed short vec_sll (vector signed short,
8709 vector unsigned char);
8710 vector unsigned short vec_sll (vector unsigned short,
8711 vector unsigned int);
8712 vector unsigned short vec_sll (vector unsigned short,
8713 vector unsigned short);
8714 vector unsigned short vec_sll (vector unsigned short,
8715 vector unsigned char);
8716 vector bool short vec_sll (vector bool short, vector unsigned int);
8717 vector bool short vec_sll (vector bool short, vector unsigned short);
8718 vector bool short vec_sll (vector bool short, vector unsigned char);
8719 vector pixel vec_sll (vector pixel, vector unsigned int);
8720 vector pixel vec_sll (vector pixel, vector unsigned short);
8721 vector pixel vec_sll (vector pixel, vector unsigned char);
8722 vector signed char vec_sll (vector signed char, vector unsigned int);
8723 vector signed char vec_sll (vector signed char, vector unsigned short);
8724 vector signed char vec_sll (vector signed char, vector unsigned char);
8725 vector unsigned char vec_sll (vector unsigned char,
8726 vector unsigned int);
8727 vector unsigned char vec_sll (vector unsigned char,
8728 vector unsigned short);
8729 vector unsigned char vec_sll (vector unsigned char,
8730 vector unsigned char);
8731 vector bool char vec_sll (vector bool char, vector unsigned int);
8732 vector bool char vec_sll (vector bool char, vector unsigned short);
8733 vector bool char vec_sll (vector bool char, vector unsigned char);
8735 vector float vec_slo (vector float, vector signed char);
8736 vector float vec_slo (vector float, vector unsigned char);
8737 vector signed int vec_slo (vector signed int, vector signed char);
8738 vector signed int vec_slo (vector signed int, vector unsigned char);
8739 vector unsigned int vec_slo (vector unsigned int, vector signed char);
8740 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
8741 vector signed short vec_slo (vector signed short, vector signed char);
8742 vector signed short vec_slo (vector signed short, vector unsigned char);
8743 vector unsigned short vec_slo (vector unsigned short,
8744 vector signed char);
8745 vector unsigned short vec_slo (vector unsigned short,
8746 vector unsigned char);
8747 vector pixel vec_slo (vector pixel, vector signed char);
8748 vector pixel vec_slo (vector pixel, vector unsigned char);
8749 vector signed char vec_slo (vector signed char, vector signed char);
8750 vector signed char vec_slo (vector signed char, vector unsigned char);
8751 vector unsigned char vec_slo (vector unsigned char, vector signed char);
8752 vector unsigned char vec_slo (vector unsigned char,
8753 vector unsigned char);
8755 vector signed char vec_splat (vector signed char, const int);
8756 vector unsigned char vec_splat (vector unsigned char, const int);
8757 vector bool char vec_splat (vector bool char, const int);
8758 vector signed short vec_splat (vector signed short, const int);
8759 vector unsigned short vec_splat (vector unsigned short, const int);
8760 vector bool short vec_splat (vector bool short, const int);
8761 vector pixel vec_splat (vector pixel, const int);
8762 vector float vec_splat (vector float, const int);
8763 vector signed int vec_splat (vector signed int, const int);
8764 vector unsigned int vec_splat (vector unsigned int, const int);
8765 vector bool int vec_splat (vector bool int, const int);
8767 vector float vec_vspltw (vector float, const int);
8768 vector signed int vec_vspltw (vector signed int, const int);
8769 vector unsigned int vec_vspltw (vector unsigned int, const int);
8770 vector bool int vec_vspltw (vector bool int, const int);
8772 vector bool short vec_vsplth (vector bool short, const int);
8773 vector signed short vec_vsplth (vector signed short, const int);
8774 vector unsigned short vec_vsplth (vector unsigned short, const int);
8775 vector pixel vec_vsplth (vector pixel, const int);
8777 vector signed char vec_vspltb (vector signed char, const int);
8778 vector unsigned char vec_vspltb (vector unsigned char, const int);
8779 vector bool char vec_vspltb (vector bool char, const int);
8781 vector signed char vec_splat_s8 (const int);
8783 vector signed short vec_splat_s16 (const int);
8785 vector signed int vec_splat_s32 (const int);
8787 vector unsigned char vec_splat_u8 (const int);
8789 vector unsigned short vec_splat_u16 (const int);
8791 vector unsigned int vec_splat_u32 (const int);
8793 vector signed char vec_sr (vector signed char, vector unsigned char);
8794 vector unsigned char vec_sr (vector unsigned char,
8795 vector unsigned char);
8796 vector signed short vec_sr (vector signed short,
8797 vector unsigned short);
8798 vector unsigned short vec_sr (vector unsigned short,
8799 vector unsigned short);
8800 vector signed int vec_sr (vector signed int, vector unsigned int);
8801 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
8803 vector signed int vec_vsrw (vector signed int, vector unsigned int);
8804 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
8806 vector signed short vec_vsrh (vector signed short,
8807 vector unsigned short);
8808 vector unsigned short vec_vsrh (vector unsigned short,
8809 vector unsigned short);
8811 vector signed char vec_vsrb (vector signed char, vector unsigned char);
8812 vector unsigned char vec_vsrb (vector unsigned char,
8813 vector unsigned char);
8815 vector signed char vec_sra (vector signed char, vector unsigned char);
8816 vector unsigned char vec_sra (vector unsigned char,
8817 vector unsigned char);
8818 vector signed short vec_sra (vector signed short,
8819 vector unsigned short);
8820 vector unsigned short vec_sra (vector unsigned short,
8821 vector unsigned short);
8822 vector signed int vec_sra (vector signed int, vector unsigned int);
8823 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
8825 vector signed int vec_vsraw (vector signed int, vector unsigned int);
8826 vector unsigned int vec_vsraw (vector unsigned int,
8827 vector unsigned int);
8829 vector signed short vec_vsrah (vector signed short,
8830 vector unsigned short);
8831 vector unsigned short vec_vsrah (vector unsigned short,
8832 vector unsigned short);
8834 vector signed char vec_vsrab (vector signed char, vector unsigned char);
8835 vector unsigned char vec_vsrab (vector unsigned char,
8836 vector unsigned char);
8838 vector signed int vec_srl (vector signed int, vector unsigned int);
8839 vector signed int vec_srl (vector signed int, vector unsigned short);
8840 vector signed int vec_srl (vector signed int, vector unsigned char);
8841 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
8842 vector unsigned int vec_srl (vector unsigned int,
8843 vector unsigned short);
8844 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
8845 vector bool int vec_srl (vector bool int, vector unsigned int);
8846 vector bool int vec_srl (vector bool int, vector unsigned short);
8847 vector bool int vec_srl (vector bool int, vector unsigned char);
8848 vector signed short vec_srl (vector signed short, vector unsigned int);
8849 vector signed short vec_srl (vector signed short,
8850 vector unsigned short);
8851 vector signed short vec_srl (vector signed short, vector unsigned char);
8852 vector unsigned short vec_srl (vector unsigned short,
8853 vector unsigned int);
8854 vector unsigned short vec_srl (vector unsigned short,
8855 vector unsigned short);
8856 vector unsigned short vec_srl (vector unsigned short,
8857 vector unsigned char);
8858 vector bool short vec_srl (vector bool short, vector unsigned int);
8859 vector bool short vec_srl (vector bool short, vector unsigned short);
8860 vector bool short vec_srl (vector bool short, vector unsigned char);
8861 vector pixel vec_srl (vector pixel, vector unsigned int);
8862 vector pixel vec_srl (vector pixel, vector unsigned short);
8863 vector pixel vec_srl (vector pixel, vector unsigned char);
8864 vector signed char vec_srl (vector signed char, vector unsigned int);
8865 vector signed char vec_srl (vector signed char, vector unsigned short);
8866 vector signed char vec_srl (vector signed char, vector unsigned char);
8867 vector unsigned char vec_srl (vector unsigned char,
8868 vector unsigned int);
8869 vector unsigned char vec_srl (vector unsigned char,
8870 vector unsigned short);
8871 vector unsigned char vec_srl (vector unsigned char,
8872 vector unsigned char);
8873 vector bool char vec_srl (vector bool char, vector unsigned int);
8874 vector bool char vec_srl (vector bool char, vector unsigned short);
8875 vector bool char vec_srl (vector bool char, vector unsigned char);
8877 vector float vec_sro (vector float, vector signed char);
8878 vector float vec_sro (vector float, vector unsigned char);
8879 vector signed int vec_sro (vector signed int, vector signed char);
8880 vector signed int vec_sro (vector signed int, vector unsigned char);
8881 vector unsigned int vec_sro (vector unsigned int, vector signed char);
8882 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
8883 vector signed short vec_sro (vector signed short, vector signed char);
8884 vector signed short vec_sro (vector signed short, vector unsigned char);
8885 vector unsigned short vec_sro (vector unsigned short,
8886 vector signed char);
8887 vector unsigned short vec_sro (vector unsigned short,
8888 vector unsigned char);
8889 vector pixel vec_sro (vector pixel, vector signed char);
8890 vector pixel vec_sro (vector pixel, vector unsigned char);
8891 vector signed char vec_sro (vector signed char, vector signed char);
8892 vector signed char vec_sro (vector signed char, vector unsigned char);
8893 vector unsigned char vec_sro (vector unsigned char, vector signed char);
8894 vector unsigned char vec_sro (vector unsigned char,
8895 vector unsigned char);
8897 void vec_st (vector float, int, vector float *);
8898 void vec_st (vector float, int, float *);
8899 void vec_st (vector signed int, int, vector signed int *);
8900 void vec_st (vector signed int, int, int *);
8901 void vec_st (vector unsigned int, int, vector unsigned int *);
8902 void vec_st (vector unsigned int, int, unsigned int *);
8903 void vec_st (vector bool int, int, vector bool int *);
8904 void vec_st (vector bool int, int, unsigned int *);
8905 void vec_st (vector bool int, int, int *);
8906 void vec_st (vector signed short, int, vector signed short *);
8907 void vec_st (vector signed short, int, short *);
8908 void vec_st (vector unsigned short, int, vector unsigned short *);
8909 void vec_st (vector unsigned short, int, unsigned short *);
8910 void vec_st (vector bool short, int, vector bool short *);
8911 void vec_st (vector bool short, int, unsigned short *);
8912 void vec_st (vector pixel, int, vector pixel *);
8913 void vec_st (vector pixel, int, unsigned short *);
8914 void vec_st (vector pixel, int, short *);
8915 void vec_st (vector bool short, int, short *);
8916 void vec_st (vector signed char, int, vector signed char *);
8917 void vec_st (vector signed char, int, signed char *);
8918 void vec_st (vector unsigned char, int, vector unsigned char *);
8919 void vec_st (vector unsigned char, int, unsigned char *);
8920 void vec_st (vector bool char, int, vector bool char *);
8921 void vec_st (vector bool char, int, unsigned char *);
8922 void vec_st (vector bool char, int, signed char *);
8924 void vec_ste (vector signed char, int, signed char *);
8925 void vec_ste (vector unsigned char, int, unsigned char *);
8926 void vec_ste (vector bool char, int, signed char *);
8927 void vec_ste (vector bool char, int, unsigned char *);
8928 void vec_ste (vector signed short, int, short *);
8929 void vec_ste (vector unsigned short, int, unsigned short *);
8930 void vec_ste (vector bool short, int, short *);
8931 void vec_ste (vector bool short, int, unsigned short *);
8932 void vec_ste (vector pixel, int, short *);
8933 void vec_ste (vector pixel, int, unsigned short *);
8934 void vec_ste (vector float, int, float *);
8935 void vec_ste (vector signed int, int, int *);
8936 void vec_ste (vector unsigned int, int, unsigned int *);
8937 void vec_ste (vector bool int, int, int *);
8938 void vec_ste (vector bool int, int, unsigned int *);
8940 void vec_stvewx (vector float, int, float *);
8941 void vec_stvewx (vector signed int, int, int *);
8942 void vec_stvewx (vector unsigned int, int, unsigned int *);
8943 void vec_stvewx (vector bool int, int, int *);
8944 void vec_stvewx (vector bool int, int, unsigned int *);
8946 void vec_stvehx (vector signed short, int, short *);
8947 void vec_stvehx (vector unsigned short, int, unsigned short *);
8948 void vec_stvehx (vector bool short, int, short *);
8949 void vec_stvehx (vector bool short, int, unsigned short *);
8950 void vec_stvehx (vector pixel, int, short *);
8951 void vec_stvehx (vector pixel, int, unsigned short *);
8953 void vec_stvebx (vector signed char, int, signed char *);
8954 void vec_stvebx (vector unsigned char, int, unsigned char *);
8955 void vec_stvebx (vector bool char, int, signed char *);
8956 void vec_stvebx (vector bool char, int, unsigned char *);
8958 void vec_stl (vector float, int, vector float *);
8959 void vec_stl (vector float, int, float *);
8960 void vec_stl (vector signed int, int, vector signed int *);
8961 void vec_stl (vector signed int, int, int *);
8962 void vec_stl (vector unsigned int, int, vector unsigned int *);
8963 void vec_stl (vector unsigned int, int, unsigned int *);
8964 void vec_stl (vector bool int, int, vector bool int *);
8965 void vec_stl (vector bool int, int, unsigned int *);
8966 void vec_stl (vector bool int, int, int *);
8967 void vec_stl (vector signed short, int, vector signed short *);
8968 void vec_stl (vector signed short, int, short *);
8969 void vec_stl (vector unsigned short, int, vector unsigned short *);
8970 void vec_stl (vector unsigned short, int, unsigned short *);
8971 void vec_stl (vector bool short, int, vector bool short *);
8972 void vec_stl (vector bool short, int, unsigned short *);
8973 void vec_stl (vector bool short, int, short *);
8974 void vec_stl (vector pixel, int, vector pixel *);
8975 void vec_stl (vector pixel, int, unsigned short *);
8976 void vec_stl (vector pixel, int, short *);
8977 void vec_stl (vector signed char, int, vector signed char *);
8978 void vec_stl (vector signed char, int, signed char *);
8979 void vec_stl (vector unsigned char, int, vector unsigned char *);
8980 void vec_stl (vector unsigned char, int, unsigned char *);
8981 void vec_stl (vector bool char, int, vector bool char *);
8982 void vec_stl (vector bool char, int, unsigned char *);
8983 void vec_stl (vector bool char, int, signed char *);
8985 vector signed char vec_sub (vector bool char, vector signed char);
8986 vector signed char vec_sub (vector signed char, vector bool char);
8987 vector signed char vec_sub (vector signed char, vector signed char);
8988 vector unsigned char vec_sub (vector bool char, vector unsigned char);
8989 vector unsigned char vec_sub (vector unsigned char, vector bool char);
8990 vector unsigned char vec_sub (vector unsigned char,
8991 vector unsigned char);
8992 vector signed short vec_sub (vector bool short, vector signed short);
8993 vector signed short vec_sub (vector signed short, vector bool short);
8994 vector signed short vec_sub (vector signed short, vector signed short);
8995 vector unsigned short vec_sub (vector bool short,
8996 vector unsigned short);
8997 vector unsigned short vec_sub (vector unsigned short,
8999 vector unsigned short vec_sub (vector unsigned short,
9000 vector unsigned short);
9001 vector signed int vec_sub (vector bool int, vector signed int);
9002 vector signed int vec_sub (vector signed int, vector bool int);
9003 vector signed int vec_sub (vector signed int, vector signed int);
9004 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9005 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9006 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9007 vector float vec_sub (vector float, vector float);
9009 vector float vec_vsubfp (vector float, vector float);
9011 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9012 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9013 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9014 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9015 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9016 vector unsigned int vec_vsubuwm (vector unsigned int,
9017 vector unsigned int);
9019 vector signed short vec_vsubuhm (vector bool short,
9020 vector signed short);
9021 vector signed short vec_vsubuhm (vector signed short,
9023 vector signed short vec_vsubuhm (vector signed short,
9024 vector signed short);
9025 vector unsigned short vec_vsubuhm (vector bool short,
9026 vector unsigned short);
9027 vector unsigned short vec_vsubuhm (vector unsigned short,
9029 vector unsigned short vec_vsubuhm (vector unsigned short,
9030 vector unsigned short);
9032 vector signed char vec_vsububm (vector bool char, vector signed char);
9033 vector signed char vec_vsububm (vector signed char, vector bool char);
9034 vector signed char vec_vsububm (vector signed char, vector signed char);
9035 vector unsigned char vec_vsububm (vector bool char,
9036 vector unsigned char);
9037 vector unsigned char vec_vsububm (vector unsigned char,
9039 vector unsigned char vec_vsububm (vector unsigned char,
9040 vector unsigned char);
9042 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9044 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9045 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9046 vector unsigned char vec_subs (vector unsigned char,
9047 vector unsigned char);
9048 vector signed char vec_subs (vector bool char, vector signed char);
9049 vector signed char vec_subs (vector signed char, vector bool char);
9050 vector signed char vec_subs (vector signed char, vector signed char);
9051 vector unsigned short vec_subs (vector bool short,
9052 vector unsigned short);
9053 vector unsigned short vec_subs (vector unsigned short,
9055 vector unsigned short vec_subs (vector unsigned short,
9056 vector unsigned short);
9057 vector signed short vec_subs (vector bool short, vector signed short);
9058 vector signed short vec_subs (vector signed short, vector bool short);
9059 vector signed short vec_subs (vector signed short, vector signed short);
9060 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9061 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9062 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9063 vector signed int vec_subs (vector bool int, vector signed int);
9064 vector signed int vec_subs (vector signed int, vector bool int);
9065 vector signed int vec_subs (vector signed int, vector signed int);
9067 vector signed int vec_vsubsws (vector bool int, vector signed int);
9068 vector signed int vec_vsubsws (vector signed int, vector bool int);
9069 vector signed int vec_vsubsws (vector signed int, vector signed int);
9071 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9072 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9073 vector unsigned int vec_vsubuws (vector unsigned int,
9074 vector unsigned int);
9076 vector signed short vec_vsubshs (vector bool short,
9077 vector signed short);
9078 vector signed short vec_vsubshs (vector signed short,
9080 vector signed short vec_vsubshs (vector signed short,
9081 vector signed short);
9083 vector unsigned short vec_vsubuhs (vector bool short,
9084 vector unsigned short);
9085 vector unsigned short vec_vsubuhs (vector unsigned short,
9087 vector unsigned short vec_vsubuhs (vector unsigned short,
9088 vector unsigned short);
9090 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9091 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9092 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9094 vector unsigned char vec_vsububs (vector bool char,
9095 vector unsigned char);
9096 vector unsigned char vec_vsububs (vector unsigned char,
9098 vector unsigned char vec_vsububs (vector unsigned char,
9099 vector unsigned char);
9101 vector unsigned int vec_sum4s (vector unsigned char,
9102 vector unsigned int);
9103 vector signed int vec_sum4s (vector signed char, vector signed int);
9104 vector signed int vec_sum4s (vector signed short, vector signed int);
9106 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9108 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9110 vector unsigned int vec_vsum4ubs (vector unsigned char,
9111 vector unsigned int);
9113 vector signed int vec_sum2s (vector signed int, vector signed int);
9115 vector signed int vec_sums (vector signed int, vector signed int);
9117 vector float vec_trunc (vector float);
9119 vector signed short vec_unpackh (vector signed char);
9120 vector bool short vec_unpackh (vector bool char);
9121 vector signed int vec_unpackh (vector signed short);
9122 vector bool int vec_unpackh (vector bool short);
9123 vector unsigned int vec_unpackh (vector pixel);
9125 vector bool int vec_vupkhsh (vector bool short);
9126 vector signed int vec_vupkhsh (vector signed short);
9128 vector unsigned int vec_vupkhpx (vector pixel);
9130 vector bool short vec_vupkhsb (vector bool char);
9131 vector signed short vec_vupkhsb (vector signed char);
9133 vector signed short vec_unpackl (vector signed char);
9134 vector bool short vec_unpackl (vector bool char);
9135 vector unsigned int vec_unpackl (vector pixel);
9136 vector signed int vec_unpackl (vector signed short);
9137 vector bool int vec_unpackl (vector bool short);
9139 vector unsigned int vec_vupklpx (vector pixel);
9141 vector bool int vec_vupklsh (vector bool short);
9142 vector signed int vec_vupklsh (vector signed short);
9144 vector bool short vec_vupklsb (vector bool char);
9145 vector signed short vec_vupklsb (vector signed char);
9147 vector float vec_xor (vector float, vector float);
9148 vector float vec_xor (vector float, vector bool int);
9149 vector float vec_xor (vector bool int, vector float);
9150 vector bool int vec_xor (vector bool int, vector bool int);
9151 vector signed int vec_xor (vector bool int, vector signed int);
9152 vector signed int vec_xor (vector signed int, vector bool int);
9153 vector signed int vec_xor (vector signed int, vector signed int);
9154 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9155 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9156 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9157 vector bool short vec_xor (vector bool short, vector bool short);
9158 vector signed short vec_xor (vector bool short, vector signed short);
9159 vector signed short vec_xor (vector signed short, vector bool short);
9160 vector signed short vec_xor (vector signed short, vector signed short);
9161 vector unsigned short vec_xor (vector bool short,
9162 vector unsigned short);
9163 vector unsigned short vec_xor (vector unsigned short,
9165 vector unsigned short vec_xor (vector unsigned short,
9166 vector unsigned short);
9167 vector signed char vec_xor (vector bool char, vector signed char);
9168 vector bool char vec_xor (vector bool char, vector bool char);
9169 vector signed char vec_xor (vector signed char, vector bool char);
9170 vector signed char vec_xor (vector signed char, vector signed char);
9171 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9172 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9173 vector unsigned char vec_xor (vector unsigned char,
9174 vector unsigned char);
9176 int vec_all_eq (vector signed char, vector bool char);
9177 int vec_all_eq (vector signed char, vector signed char);
9178 int vec_all_eq (vector unsigned char, vector bool char);
9179 int vec_all_eq (vector unsigned char, vector unsigned char);
9180 int vec_all_eq (vector bool char, vector bool char);
9181 int vec_all_eq (vector bool char, vector unsigned char);
9182 int vec_all_eq (vector bool char, vector signed char);
9183 int vec_all_eq (vector signed short, vector bool short);
9184 int vec_all_eq (vector signed short, vector signed short);
9185 int vec_all_eq (vector unsigned short, vector bool short);
9186 int vec_all_eq (vector unsigned short, vector unsigned short);
9187 int vec_all_eq (vector bool short, vector bool short);
9188 int vec_all_eq (vector bool short, vector unsigned short);
9189 int vec_all_eq (vector bool short, vector signed short);
9190 int vec_all_eq (vector pixel, vector pixel);
9191 int vec_all_eq (vector signed int, vector bool int);
9192 int vec_all_eq (vector signed int, vector signed int);
9193 int vec_all_eq (vector unsigned int, vector bool int);
9194 int vec_all_eq (vector unsigned int, vector unsigned int);
9195 int vec_all_eq (vector bool int, vector bool int);
9196 int vec_all_eq (vector bool int, vector unsigned int);
9197 int vec_all_eq (vector bool int, vector signed int);
9198 int vec_all_eq (vector float, vector float);
9200 int vec_all_ge (vector bool char, vector unsigned char);
9201 int vec_all_ge (vector unsigned char, vector bool char);
9202 int vec_all_ge (vector unsigned char, vector unsigned char);
9203 int vec_all_ge (vector bool char, vector signed char);
9204 int vec_all_ge (vector signed char, vector bool char);
9205 int vec_all_ge (vector signed char, vector signed char);
9206 int vec_all_ge (vector bool short, vector unsigned short);
9207 int vec_all_ge (vector unsigned short, vector bool short);
9208 int vec_all_ge (vector unsigned short, vector unsigned short);
9209 int vec_all_ge (vector signed short, vector signed short);
9210 int vec_all_ge (vector bool short, vector signed short);
9211 int vec_all_ge (vector signed short, vector bool short);
9212 int vec_all_ge (vector bool int, vector unsigned int);
9213 int vec_all_ge (vector unsigned int, vector bool int);
9214 int vec_all_ge (vector unsigned int, vector unsigned int);
9215 int vec_all_ge (vector bool int, vector signed int);
9216 int vec_all_ge (vector signed int, vector bool int);
9217 int vec_all_ge (vector signed int, vector signed int);
9218 int vec_all_ge (vector float, vector float);
9220 int vec_all_gt (vector bool char, vector unsigned char);
9221 int vec_all_gt (vector unsigned char, vector bool char);
9222 int vec_all_gt (vector unsigned char, vector unsigned char);
9223 int vec_all_gt (vector bool char, vector signed char);
9224 int vec_all_gt (vector signed char, vector bool char);
9225 int vec_all_gt (vector signed char, vector signed char);
9226 int vec_all_gt (vector bool short, vector unsigned short);
9227 int vec_all_gt (vector unsigned short, vector bool short);
9228 int vec_all_gt (vector unsigned short, vector unsigned short);
9229 int vec_all_gt (vector bool short, vector signed short);
9230 int vec_all_gt (vector signed short, vector bool short);
9231 int vec_all_gt (vector signed short, vector signed short);
9232 int vec_all_gt (vector bool int, vector unsigned int);
9233 int vec_all_gt (vector unsigned int, vector bool int);
9234 int vec_all_gt (vector unsigned int, vector unsigned int);
9235 int vec_all_gt (vector bool int, vector signed int);
9236 int vec_all_gt (vector signed int, vector bool int);
9237 int vec_all_gt (vector signed int, vector signed int);
9238 int vec_all_gt (vector float, vector float);
9240 int vec_all_in (vector float, vector float);
9242 int vec_all_le (vector bool char, vector unsigned char);
9243 int vec_all_le (vector unsigned char, vector bool char);
9244 int vec_all_le (vector unsigned char, vector unsigned char);
9245 int vec_all_le (vector bool char, vector signed char);
9246 int vec_all_le (vector signed char, vector bool char);
9247 int vec_all_le (vector signed char, vector signed char);
9248 int vec_all_le (vector bool short, vector unsigned short);
9249 int vec_all_le (vector unsigned short, vector bool short);
9250 int vec_all_le (vector unsigned short, vector unsigned short);
9251 int vec_all_le (vector bool short, vector signed short);
9252 int vec_all_le (vector signed short, vector bool short);
9253 int vec_all_le (vector signed short, vector signed short);
9254 int vec_all_le (vector bool int, vector unsigned int);
9255 int vec_all_le (vector unsigned int, vector bool int);
9256 int vec_all_le (vector unsigned int, vector unsigned int);
9257 int vec_all_le (vector bool int, vector signed int);
9258 int vec_all_le (vector signed int, vector bool int);
9259 int vec_all_le (vector signed int, vector signed int);
9260 int vec_all_le (vector float, vector float);
9262 int vec_all_lt (vector bool char, vector unsigned char);
9263 int vec_all_lt (vector unsigned char, vector bool char);
9264 int vec_all_lt (vector unsigned char, vector unsigned char);
9265 int vec_all_lt (vector bool char, vector signed char);
9266 int vec_all_lt (vector signed char, vector bool char);
9267 int vec_all_lt (vector signed char, vector signed char);
9268 int vec_all_lt (vector bool short, vector unsigned short);
9269 int vec_all_lt (vector unsigned short, vector bool short);
9270 int vec_all_lt (vector unsigned short, vector unsigned short);
9271 int vec_all_lt (vector bool short, vector signed short);
9272 int vec_all_lt (vector signed short, vector bool short);
9273 int vec_all_lt (vector signed short, vector signed short);
9274 int vec_all_lt (vector bool int, vector unsigned int);
9275 int vec_all_lt (vector unsigned int, vector bool int);
9276 int vec_all_lt (vector unsigned int, vector unsigned int);
9277 int vec_all_lt (vector bool int, vector signed int);
9278 int vec_all_lt (vector signed int, vector bool int);
9279 int vec_all_lt (vector signed int, vector signed int);
9280 int vec_all_lt (vector float, vector float);
9282 int vec_all_nan (vector float);
9284 int vec_all_ne (vector signed char, vector bool char);
9285 int vec_all_ne (vector signed char, vector signed char);
9286 int vec_all_ne (vector unsigned char, vector bool char);
9287 int vec_all_ne (vector unsigned char, vector unsigned char);
9288 int vec_all_ne (vector bool char, vector bool char);
9289 int vec_all_ne (vector bool char, vector unsigned char);
9290 int vec_all_ne (vector bool char, vector signed char);
9291 int vec_all_ne (vector signed short, vector bool short);
9292 int vec_all_ne (vector signed short, vector signed short);
9293 int vec_all_ne (vector unsigned short, vector bool short);
9294 int vec_all_ne (vector unsigned short, vector unsigned short);
9295 int vec_all_ne (vector bool short, vector bool short);
9296 int vec_all_ne (vector bool short, vector unsigned short);
9297 int vec_all_ne (vector bool short, vector signed short);
9298 int vec_all_ne (vector pixel, vector pixel);
9299 int vec_all_ne (vector signed int, vector bool int);
9300 int vec_all_ne (vector signed int, vector signed int);
9301 int vec_all_ne (vector unsigned int, vector bool int);
9302 int vec_all_ne (vector unsigned int, vector unsigned int);
9303 int vec_all_ne (vector bool int, vector bool int);
9304 int vec_all_ne (vector bool int, vector unsigned int);
9305 int vec_all_ne (vector bool int, vector signed int);
9306 int vec_all_ne (vector float, vector float);
9308 int vec_all_nge (vector float, vector float);
9310 int vec_all_ngt (vector float, vector float);
9312 int vec_all_nle (vector float, vector float);
9314 int vec_all_nlt (vector float, vector float);
9316 int vec_all_numeric (vector float);
9318 int vec_any_eq (vector signed char, vector bool char);
9319 int vec_any_eq (vector signed char, vector signed char);
9320 int vec_any_eq (vector unsigned char, vector bool char);
9321 int vec_any_eq (vector unsigned char, vector unsigned char);
9322 int vec_any_eq (vector bool char, vector bool char);
9323 int vec_any_eq (vector bool char, vector unsigned char);
9324 int vec_any_eq (vector bool char, vector signed char);
9325 int vec_any_eq (vector signed short, vector bool short);
9326 int vec_any_eq (vector signed short, vector signed short);
9327 int vec_any_eq (vector unsigned short, vector bool short);
9328 int vec_any_eq (vector unsigned short, vector unsigned short);
9329 int vec_any_eq (vector bool short, vector bool short);
9330 int vec_any_eq (vector bool short, vector unsigned short);
9331 int vec_any_eq (vector bool short, vector signed short);
9332 int vec_any_eq (vector pixel, vector pixel);
9333 int vec_any_eq (vector signed int, vector bool int);
9334 int vec_any_eq (vector signed int, vector signed int);
9335 int vec_any_eq (vector unsigned int, vector bool int);
9336 int vec_any_eq (vector unsigned int, vector unsigned int);
9337 int vec_any_eq (vector bool int, vector bool int);
9338 int vec_any_eq (vector bool int, vector unsigned int);
9339 int vec_any_eq (vector bool int, vector signed int);
9340 int vec_any_eq (vector float, vector float);
9342 int vec_any_ge (vector signed char, vector bool char);
9343 int vec_any_ge (vector unsigned char, vector bool char);
9344 int vec_any_ge (vector unsigned char, vector unsigned char);
9345 int vec_any_ge (vector signed char, vector signed char);
9346 int vec_any_ge (vector bool char, vector unsigned char);
9347 int vec_any_ge (vector bool char, vector signed char);
9348 int vec_any_ge (vector unsigned short, vector bool short);
9349 int vec_any_ge (vector unsigned short, vector unsigned short);
9350 int vec_any_ge (vector signed short, vector signed short);
9351 int vec_any_ge (vector signed short, vector bool short);
9352 int vec_any_ge (vector bool short, vector unsigned short);
9353 int vec_any_ge (vector bool short, vector signed short);
9354 int vec_any_ge (vector signed int, vector bool int);
9355 int vec_any_ge (vector unsigned int, vector bool int);
9356 int vec_any_ge (vector unsigned int, vector unsigned int);
9357 int vec_any_ge (vector signed int, vector signed int);
9358 int vec_any_ge (vector bool int, vector unsigned int);
9359 int vec_any_ge (vector bool int, vector signed int);
9360 int vec_any_ge (vector float, vector float);
9362 int vec_any_gt (vector bool char, vector unsigned char);
9363 int vec_any_gt (vector unsigned char, vector bool char);
9364 int vec_any_gt (vector unsigned char, vector unsigned char);
9365 int vec_any_gt (vector bool char, vector signed char);
9366 int vec_any_gt (vector signed char, vector bool char);
9367 int vec_any_gt (vector signed char, vector signed char);
9368 int vec_any_gt (vector bool short, vector unsigned short);
9369 int vec_any_gt (vector unsigned short, vector bool short);
9370 int vec_any_gt (vector unsigned short, vector unsigned short);
9371 int vec_any_gt (vector bool short, vector signed short);
9372 int vec_any_gt (vector signed short, vector bool short);
9373 int vec_any_gt (vector signed short, vector signed short);
9374 int vec_any_gt (vector bool int, vector unsigned int);
9375 int vec_any_gt (vector unsigned int, vector bool int);
9376 int vec_any_gt (vector unsigned int, vector unsigned int);
9377 int vec_any_gt (vector bool int, vector signed int);
9378 int vec_any_gt (vector signed int, vector bool int);
9379 int vec_any_gt (vector signed int, vector signed int);
9380 int vec_any_gt (vector float, vector float);
9382 int vec_any_le (vector bool char, vector unsigned char);
9383 int vec_any_le (vector unsigned char, vector bool char);
9384 int vec_any_le (vector unsigned char, vector unsigned char);
9385 int vec_any_le (vector bool char, vector signed char);
9386 int vec_any_le (vector signed char, vector bool char);
9387 int vec_any_le (vector signed char, vector signed char);
9388 int vec_any_le (vector bool short, vector unsigned short);
9389 int vec_any_le (vector unsigned short, vector bool short);
9390 int vec_any_le (vector unsigned short, vector unsigned short);
9391 int vec_any_le (vector bool short, vector signed short);
9392 int vec_any_le (vector signed short, vector bool short);
9393 int vec_any_le (vector signed short, vector signed short);
9394 int vec_any_le (vector bool int, vector unsigned int);
9395 int vec_any_le (vector unsigned int, vector bool int);
9396 int vec_any_le (vector unsigned int, vector unsigned int);
9397 int vec_any_le (vector bool int, vector signed int);
9398 int vec_any_le (vector signed int, vector bool int);
9399 int vec_any_le (vector signed int, vector signed int);
9400 int vec_any_le (vector float, vector float);
9402 int vec_any_lt (vector bool char, vector unsigned char);
9403 int vec_any_lt (vector unsigned char, vector bool char);
9404 int vec_any_lt (vector unsigned char, vector unsigned char);
9405 int vec_any_lt (vector bool char, vector signed char);
9406 int vec_any_lt (vector signed char, vector bool char);
9407 int vec_any_lt (vector signed char, vector signed char);
9408 int vec_any_lt (vector bool short, vector unsigned short);
9409 int vec_any_lt (vector unsigned short, vector bool short);
9410 int vec_any_lt (vector unsigned short, vector unsigned short);
9411 int vec_any_lt (vector bool short, vector signed short);
9412 int vec_any_lt (vector signed short, vector bool short);
9413 int vec_any_lt (vector signed short, vector signed short);
9414 int vec_any_lt (vector bool int, vector unsigned int);
9415 int vec_any_lt (vector unsigned int, vector bool int);
9416 int vec_any_lt (vector unsigned int, vector unsigned int);
9417 int vec_any_lt (vector bool int, vector signed int);
9418 int vec_any_lt (vector signed int, vector bool int);
9419 int vec_any_lt (vector signed int, vector signed int);
9420 int vec_any_lt (vector float, vector float);
9422 int vec_any_nan (vector float);
9424 int vec_any_ne (vector signed char, vector bool char);
9425 int vec_any_ne (vector signed char, vector signed char);
9426 int vec_any_ne (vector unsigned char, vector bool char);
9427 int vec_any_ne (vector unsigned char, vector unsigned char);
9428 int vec_any_ne (vector bool char, vector bool char);
9429 int vec_any_ne (vector bool char, vector unsigned char);
9430 int vec_any_ne (vector bool char, vector signed char);
9431 int vec_any_ne (vector signed short, vector bool short);
9432 int vec_any_ne (vector signed short, vector signed short);
9433 int vec_any_ne (vector unsigned short, vector bool short);
9434 int vec_any_ne (vector unsigned short, vector unsigned short);
9435 int vec_any_ne (vector bool short, vector bool short);
9436 int vec_any_ne (vector bool short, vector unsigned short);
9437 int vec_any_ne (vector bool short, vector signed short);
9438 int vec_any_ne (vector pixel, vector pixel);
9439 int vec_any_ne (vector signed int, vector bool int);
9440 int vec_any_ne (vector signed int, vector signed int);
9441 int vec_any_ne (vector unsigned int, vector bool int);
9442 int vec_any_ne (vector unsigned int, vector unsigned int);
9443 int vec_any_ne (vector bool int, vector bool int);
9444 int vec_any_ne (vector bool int, vector unsigned int);
9445 int vec_any_ne (vector bool int, vector signed int);
9446 int vec_any_ne (vector float, vector float);
9448 int vec_any_nge (vector float, vector float);
9450 int vec_any_ngt (vector float, vector float);
9452 int vec_any_nle (vector float, vector float);
9454 int vec_any_nlt (vector float, vector float);
9456 int vec_any_numeric (vector float);
9458 int vec_any_out (vector float, vector float);
9461 @node SPARC VIS Built-in Functions
9462 @subsection SPARC VIS Built-in Functions
9464 GCC supports SIMD operations on the SPARC using both the generic vector
9465 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9466 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9467 switch, the VIS extension is exposed as the following built-in functions:
9470 typedef int v2si __attribute__ ((vector_size (8)));
9471 typedef short v4hi __attribute__ ((vector_size (8)));
9472 typedef short v2hi __attribute__ ((vector_size (4)));
9473 typedef char v8qi __attribute__ ((vector_size (8)));
9474 typedef char v4qi __attribute__ ((vector_size (4)));
9476 void * __builtin_vis_alignaddr (void *, long);
9477 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9478 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9479 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9480 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9482 v4hi __builtin_vis_fexpand (v4qi);
9484 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9485 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9486 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9487 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9488 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9489 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9490 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9492 v4qi __builtin_vis_fpack16 (v4hi);
9493 v8qi __builtin_vis_fpack32 (v2si, v2si);
9494 v2hi __builtin_vis_fpackfix (v2si);
9495 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9497 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9500 @node Target Format Checks
9501 @section Format Checks Specific to Particular Target Machines
9503 For some target machines, GCC supports additional options to the
9505 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9508 * Solaris Format Checks::
9511 @node Solaris Format Checks
9512 @subsection Solaris Format Checks
9514 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9515 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9516 conversions, and the two-argument @code{%b} conversion for displaying
9517 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9520 @section Pragmas Accepted by GCC
9524 GCC supports several types of pragmas, primarily in order to compile
9525 code originally written for other compilers. Note that in general
9526 we do not recommend the use of pragmas; @xref{Function Attributes},
9527 for further explanation.
9532 * RS/6000 and PowerPC Pragmas::
9535 * Symbol-Renaming Pragmas::
9536 * Structure-Packing Pragmas::
9538 * Diagnostic Pragmas::
9539 * Visibility Pragmas::
9543 @subsection ARM Pragmas
9545 The ARM target defines pragmas for controlling the default addition of
9546 @code{long_call} and @code{short_call} attributes to functions.
9547 @xref{Function Attributes}, for information about the effects of these
9552 @cindex pragma, long_calls
9553 Set all subsequent functions to have the @code{long_call} attribute.
9556 @cindex pragma, no_long_calls
9557 Set all subsequent functions to have the @code{short_call} attribute.
9559 @item long_calls_off
9560 @cindex pragma, long_calls_off
9561 Do not affect the @code{long_call} or @code{short_call} attributes of
9562 subsequent functions.
9566 @subsection M32C Pragmas
9569 @item memregs @var{number}
9570 @cindex pragma, memregs
9571 Overrides the command line option @code{-memregs=} for the current
9572 file. Use with care! This pragma must be before any function in the
9573 file, and mixing different memregs values in different objects may
9574 make them incompatible. This pragma is useful when a
9575 performance-critical function uses a memreg for temporary values,
9576 as it may allow you to reduce the number of memregs used.
9580 @node RS/6000 and PowerPC Pragmas
9581 @subsection RS/6000 and PowerPC Pragmas
9583 The RS/6000 and PowerPC targets define one pragma for controlling
9584 whether or not the @code{longcall} attribute is added to function
9585 declarations by default. This pragma overrides the @option{-mlongcall}
9586 option, but not the @code{longcall} and @code{shortcall} attributes.
9587 @xref{RS/6000 and PowerPC Options}, for more information about when long
9588 calls are and are not necessary.
9592 @cindex pragma, longcall
9593 Apply the @code{longcall} attribute to all subsequent function
9597 Do not apply the @code{longcall} attribute to subsequent function
9601 @c Describe c4x pragmas here.
9602 @c Describe h8300 pragmas here.
9603 @c Describe sh pragmas here.
9604 @c Describe v850 pragmas here.
9606 @node Darwin Pragmas
9607 @subsection Darwin Pragmas
9609 The following pragmas are available for all architectures running the
9610 Darwin operating system. These are useful for compatibility with other
9614 @item mark @var{tokens}@dots{}
9615 @cindex pragma, mark
9616 This pragma is accepted, but has no effect.
9618 @item options align=@var{alignment}
9619 @cindex pragma, options align
9620 This pragma sets the alignment of fields in structures. The values of
9621 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9622 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9623 properly; to restore the previous setting, use @code{reset} for the
9626 @item segment @var{tokens}@dots{}
9627 @cindex pragma, segment
9628 This pragma is accepted, but has no effect.
9630 @item unused (@var{var} [, @var{var}]@dots{})
9631 @cindex pragma, unused
9632 This pragma declares variables to be possibly unused. GCC will not
9633 produce warnings for the listed variables. The effect is similar to
9634 that of the @code{unused} attribute, except that this pragma may appear
9635 anywhere within the variables' scopes.
9638 @node Solaris Pragmas
9639 @subsection Solaris Pragmas
9641 The Solaris target supports @code{#pragma redefine_extname}
9642 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
9643 @code{#pragma} directives for compatibility with the system compiler.
9646 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
9647 @cindex pragma, align
9649 Increase the minimum alignment of each @var{variable} to @var{alignment}.
9650 This is the same as GCC's @code{aligned} attribute @pxref{Variable
9651 Attributes}). Macro expansion occurs on the arguments to this pragma
9652 when compiling C and Objective-C. It does not currently occur when
9653 compiling C++, but this is a bug which may be fixed in a future
9656 @item fini (@var{function} [, @var{function}]...)
9657 @cindex pragma, fini
9659 This pragma causes each listed @var{function} to be called after
9660 main, or during shared module unloading, by adding a call to the
9661 @code{.fini} section.
9663 @item init (@var{function} [, @var{function}]...)
9664 @cindex pragma, init
9666 This pragma causes each listed @var{function} to be called during
9667 initialization (before @code{main}) or during shared module loading, by
9668 adding a call to the @code{.init} section.
9672 @node Symbol-Renaming Pragmas
9673 @subsection Symbol-Renaming Pragmas
9675 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
9676 supports two @code{#pragma} directives which change the name used in
9677 assembly for a given declaration. These pragmas are only available on
9678 platforms whose system headers need them. To get this effect on all
9679 platforms supported by GCC, use the asm labels extension (@pxref{Asm
9683 @item redefine_extname @var{oldname} @var{newname}
9684 @cindex pragma, redefine_extname
9686 This pragma gives the C function @var{oldname} the assembly symbol
9687 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
9688 will be defined if this pragma is available (currently only on
9691 @item extern_prefix @var{string}
9692 @cindex pragma, extern_prefix
9694 This pragma causes all subsequent external function and variable
9695 declarations to have @var{string} prepended to their assembly symbols.
9696 This effect may be terminated with another @code{extern_prefix} pragma
9697 whose argument is an empty string. The preprocessor macro
9698 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
9699 available (currently only on Tru64 UNIX)@.
9702 These pragmas and the asm labels extension interact in a complicated
9703 manner. Here are some corner cases you may want to be aware of.
9706 @item Both pragmas silently apply only to declarations with external
9707 linkage. Asm labels do not have this restriction.
9709 @item In C++, both pragmas silently apply only to declarations with
9710 ``C'' linkage. Again, asm labels do not have this restriction.
9712 @item If any of the three ways of changing the assembly name of a
9713 declaration is applied to a declaration whose assembly name has
9714 already been determined (either by a previous use of one of these
9715 features, or because the compiler needed the assembly name in order to
9716 generate code), and the new name is different, a warning issues and
9717 the name does not change.
9719 @item The @var{oldname} used by @code{#pragma redefine_extname} is
9720 always the C-language name.
9722 @item If @code{#pragma extern_prefix} is in effect, and a declaration
9723 occurs with an asm label attached, the prefix is silently ignored for
9726 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
9727 apply to the same declaration, whichever triggered first wins, and a
9728 warning issues if they contradict each other. (We would like to have
9729 @code{#pragma redefine_extname} always win, for consistency with asm
9730 labels, but if @code{#pragma extern_prefix} triggers first we have no
9731 way of knowing that that happened.)
9734 @node Structure-Packing Pragmas
9735 @subsection Structure-Packing Pragmas
9737 For compatibility with Win32, GCC supports a set of @code{#pragma}
9738 directives which change the maximum alignment of members of structures
9739 (other than zero-width bitfields), unions, and classes subsequently
9740 defined. The @var{n} value below always is required to be a small power
9741 of two and specifies the new alignment in bytes.
9744 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
9745 @item @code{#pragma pack()} sets the alignment to the one that was in
9746 effect when compilation started (see also command line option
9747 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
9748 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
9749 setting on an internal stack and then optionally sets the new alignment.
9750 @item @code{#pragma pack(pop)} restores the alignment setting to the one
9751 saved at the top of the internal stack (and removes that stack entry).
9752 Note that @code{#pragma pack([@var{n}])} does not influence this internal
9753 stack; thus it is possible to have @code{#pragma pack(push)} followed by
9754 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
9755 @code{#pragma pack(pop)}.
9758 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
9759 @code{#pragma} which lays out a structure as the documented
9760 @code{__attribute__ ((ms_struct))}.
9762 @item @code{#pragma ms_struct on} turns on the layout for structures
9764 @item @code{#pragma ms_struct off} turns off the layout for structures
9766 @item @code{#pragma ms_struct reset} goes back to the default layout.
9770 @subsection Weak Pragmas
9772 For compatibility with SVR4, GCC supports a set of @code{#pragma}
9773 directives for declaring symbols to be weak, and defining weak
9777 @item #pragma weak @var{symbol}
9778 @cindex pragma, weak
9779 This pragma declares @var{symbol} to be weak, as if the declaration
9780 had the attribute of the same name. The pragma may appear before
9781 or after the declaration of @var{symbol}, but must appear before
9782 either its first use or its definition. It is not an error for
9783 @var{symbol} to never be defined at all.
9785 @item #pragma weak @var{symbol1} = @var{symbol2}
9786 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
9787 It is an error if @var{symbol2} is not defined in the current
9791 @node Diagnostic Pragmas
9792 @subsection Diagnostic Pragmas
9794 GCC allows the user to selectively enable or disable certain types of
9795 diagnostics, and change the kind of the diagnostic. For example, a
9796 project's policy might require that all sources compile with
9797 @option{-Werror} but certain files might have exceptions allowing
9798 specific types of warnings. Or, a project might selectively enable
9799 diagnostics and treat them as errors depending on which preprocessor
9803 @item #pragma GCC diagnostic @var{kind} @var{option}
9804 @cindex pragma, diagnostic
9806 Modifies the disposition of a diagnostic. Note that not all
9807 diagnostics are modifyiable; at the moment only warnings (normally
9808 controlled by @samp{-W...}) can be controlled, and not all of them.
9809 Use @option{-fdiagnostics-show-option} to determine which diagnostics
9810 are controllable and which option controls them.
9812 @var{kind} is @samp{error} to treat this diagnostic as an error,
9813 @samp{warning} to treat it like a warning (even if @option{-Werror} is
9814 in effect), or @samp{ignored} if the diagnostic is to be ignored.
9815 @var{option} is a double quoted string which matches the command line
9819 #pragma GCC diagnostic warning "-Wformat"
9820 #pragma GCC diagnostic error "-Walways-true"
9821 #pragma GCC diagnostic ignored "-Walways-true"
9824 Note that these pragmas override any command line options. Also,
9825 while it is syntactically valid to put these pragmas anywhere in your
9826 sources, the only supported location for them is before any data or
9827 functions are defined. Doing otherwise may result in unpredictable
9828 results depending on how the optimizer manages your sources. If the
9829 same option is listed multiple times, the last one specified is the
9830 one that is in effect. This pragma is not intended to be a general
9831 purpose replacement for command line options, but for implementing
9832 strict control over project policies.
9836 @node Visibility Pragmas
9837 @subsection Visibility Pragmas
9840 @item #pragma GCC visibility push(@var{visibility})
9841 @itemx #pragma GCC visibility pop
9842 @cindex pragma, visibility
9844 This pragma allows the user to set the visibility for multiple
9845 declarations without having to give each a visibility attribute
9846 @xref{Function Attributes}, for more information about visibility and
9847 the attribute syntax.
9849 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
9850 declarations. Class members and template specializations are not
9851 affected; if you want to override the visibility for a particular
9852 member or instantiation, you must use an attribute.
9856 @node Unnamed Fields
9857 @section Unnamed struct/union fields within structs/unions
9861 For compatibility with other compilers, GCC allows you to define
9862 a structure or union that contains, as fields, structures and unions
9863 without names. For example:
9876 In this example, the user would be able to access members of the unnamed
9877 union with code like @samp{foo.b}. Note that only unnamed structs and
9878 unions are allowed, you may not have, for example, an unnamed
9881 You must never create such structures that cause ambiguous field definitions.
9882 For example, this structure:
9893 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
9894 Such constructs are not supported and must be avoided. In the future,
9895 such constructs may be detected and treated as compilation errors.
9897 @opindex fms-extensions
9898 Unless @option{-fms-extensions} is used, the unnamed field must be a
9899 structure or union definition without a tag (for example, @samp{struct
9900 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
9901 also be a definition with a tag such as @samp{struct foo @{ int a;
9902 @};}, a reference to a previously defined structure or union such as
9903 @samp{struct foo;}, or a reference to a @code{typedef} name for a
9904 previously defined structure or union type.
9907 @section Thread-Local Storage
9908 @cindex Thread-Local Storage
9909 @cindex @acronym{TLS}
9912 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
9913 are allocated such that there is one instance of the variable per extant
9914 thread. The run-time model GCC uses to implement this originates
9915 in the IA-64 processor-specific ABI, but has since been migrated
9916 to other processors as well. It requires significant support from
9917 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
9918 system libraries (@file{libc.so} and @file{libpthread.so}), so it
9919 is not available everywhere.
9921 At the user level, the extension is visible with a new storage
9922 class keyword: @code{__thread}. For example:
9926 extern __thread struct state s;
9927 static __thread char *p;
9930 The @code{__thread} specifier may be used alone, with the @code{extern}
9931 or @code{static} specifiers, but with no other storage class specifier.
9932 When used with @code{extern} or @code{static}, @code{__thread} must appear
9933 immediately after the other storage class specifier.
9935 The @code{__thread} specifier may be applied to any global, file-scoped
9936 static, function-scoped static, or static data member of a class. It may
9937 not be applied to block-scoped automatic or non-static data member.
9939 When the address-of operator is applied to a thread-local variable, it is
9940 evaluated at run-time and returns the address of the current thread's
9941 instance of that variable. An address so obtained may be used by any
9942 thread. When a thread terminates, any pointers to thread-local variables
9943 in that thread become invalid.
9945 No static initialization may refer to the address of a thread-local variable.
9947 In C++, if an initializer is present for a thread-local variable, it must
9948 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
9951 See @uref{http://people.redhat.com/drepper/tls.pdf,
9952 ELF Handling For Thread-Local Storage} for a detailed explanation of
9953 the four thread-local storage addressing models, and how the run-time
9954 is expected to function.
9957 * C99 Thread-Local Edits::
9958 * C++98 Thread-Local Edits::
9961 @node C99 Thread-Local Edits
9962 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
9964 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
9965 that document the exact semantics of the language extension.
9969 @cite{5.1.2 Execution environments}
9971 Add new text after paragraph 1
9974 Within either execution environment, a @dfn{thread} is a flow of
9975 control within a program. It is implementation defined whether
9976 or not there may be more than one thread associated with a program.
9977 It is implementation defined how threads beyond the first are
9978 created, the name and type of the function called at thread
9979 startup, and how threads may be terminated. However, objects
9980 with thread storage duration shall be initialized before thread
9985 @cite{6.2.4 Storage durations of objects}
9987 Add new text before paragraph 3
9990 An object whose identifier is declared with the storage-class
9991 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
9992 Its lifetime is the entire execution of the thread, and its
9993 stored value is initialized only once, prior to thread startup.
9997 @cite{6.4.1 Keywords}
9999 Add @code{__thread}.
10002 @cite{6.7.1 Storage-class specifiers}
10004 Add @code{__thread} to the list of storage class specifiers in
10007 Change paragraph 2 to
10010 With the exception of @code{__thread}, at most one storage-class
10011 specifier may be given [@dots{}]. The @code{__thread} specifier may
10012 be used alone, or immediately following @code{extern} or
10016 Add new text after paragraph 6
10019 The declaration of an identifier for a variable that has
10020 block scope that specifies @code{__thread} shall also
10021 specify either @code{extern} or @code{static}.
10023 The @code{__thread} specifier shall be used only with
10028 @node C++98 Thread-Local Edits
10029 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10031 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10032 that document the exact semantics of the language extension.
10036 @b{[intro.execution]}
10038 New text after paragraph 4
10041 A @dfn{thread} is a flow of control within the abstract machine.
10042 It is implementation defined whether or not there may be more than
10046 New text after paragraph 7
10049 It is unspecified whether additional action must be taken to
10050 ensure when and whether side effects are visible to other threads.
10056 Add @code{__thread}.
10059 @b{[basic.start.main]}
10061 Add after paragraph 5
10064 The thread that begins execution at the @code{main} function is called
10065 the @dfn{main thread}. It is implementation defined how functions
10066 beginning threads other than the main thread are designated or typed.
10067 A function so designated, as well as the @code{main} function, is called
10068 a @dfn{thread startup function}. It is implementation defined what
10069 happens if a thread startup function returns. It is implementation
10070 defined what happens to other threads when any thread calls @code{exit}.
10074 @b{[basic.start.init]}
10076 Add after paragraph 4
10079 The storage for an object of thread storage duration shall be
10080 statically initialized before the first statement of the thread startup
10081 function. An object of thread storage duration shall not require
10082 dynamic initialization.
10086 @b{[basic.start.term]}
10088 Add after paragraph 3
10091 The type of an object with thread storage duration shall not have a
10092 non-trivial destructor, nor shall it be an array type whose elements
10093 (directly or indirectly) have non-trivial destructors.
10099 Add ``thread storage duration'' to the list in paragraph 1.
10104 Thread, static, and automatic storage durations are associated with
10105 objects introduced by declarations [@dots{}].
10108 Add @code{__thread} to the list of specifiers in paragraph 3.
10111 @b{[basic.stc.thread]}
10113 New section before @b{[basic.stc.static]}
10116 The keyword @code{__thread} applied to a non-local object gives the
10117 object thread storage duration.
10119 A local variable or class data member declared both @code{static}
10120 and @code{__thread} gives the variable or member thread storage
10125 @b{[basic.stc.static]}
10130 All objects which have neither thread storage duration, dynamic
10131 storage duration nor are local [@dots{}].
10137 Add @code{__thread} to the list in paragraph 1.
10142 With the exception of @code{__thread}, at most one
10143 @var{storage-class-specifier} shall appear in a given
10144 @var{decl-specifier-seq}. The @code{__thread} specifier may
10145 be used alone, or immediately following the @code{extern} or
10146 @code{static} specifiers. [@dots{}]
10149 Add after paragraph 5
10152 The @code{__thread} specifier can be applied only to the names of objects
10153 and to anonymous unions.
10159 Add after paragraph 6
10162 Non-@code{static} members shall not be @code{__thread}.
10166 @node C++ Extensions
10167 @chapter Extensions to the C++ Language
10168 @cindex extensions, C++ language
10169 @cindex C++ language extensions
10171 The GNU compiler provides these extensions to the C++ language (and you
10172 can also use most of the C language extensions in your C++ programs). If you
10173 want to write code that checks whether these features are available, you can
10174 test for the GNU compiler the same way as for C programs: check for a
10175 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10176 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10177 Predefined Macros,cpp,The GNU C Preprocessor}).
10180 * Volatiles:: What constitutes an access to a volatile object.
10181 * Restricted Pointers:: C99 restricted pointers and references.
10182 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10183 * C++ Interface:: You can use a single C++ header file for both
10184 declarations and definitions.
10185 * Template Instantiation:: Methods for ensuring that exactly one copy of
10186 each needed template instantiation is emitted.
10187 * Bound member functions:: You can extract a function pointer to the
10188 method denoted by a @samp{->*} or @samp{.*} expression.
10189 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10190 * Namespace Association:: Strong using-directives for namespace association.
10191 * Java Exceptions:: Tweaking exception handling to work with Java.
10192 * Deprecated Features:: Things will disappear from g++.
10193 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10197 @section When is a Volatile Object Accessed?
10198 @cindex accessing volatiles
10199 @cindex volatile read
10200 @cindex volatile write
10201 @cindex volatile access
10203 Both the C and C++ standard have the concept of volatile objects. These
10204 are normally accessed by pointers and used for accessing hardware. The
10205 standards encourage compilers to refrain from optimizations
10206 concerning accesses to volatile objects that it might perform on
10207 non-volatile objects. The C standard leaves it implementation defined
10208 as to what constitutes a volatile access. The C++ standard omits to
10209 specify this, except to say that C++ should behave in a similar manner
10210 to C with respect to volatiles, where possible. The minimum either
10211 standard specifies is that at a sequence point all previous accesses to
10212 volatile objects have stabilized and no subsequent accesses have
10213 occurred. Thus an implementation is free to reorder and combine
10214 volatile accesses which occur between sequence points, but cannot do so
10215 for accesses across a sequence point. The use of volatiles does not
10216 allow you to violate the restriction on updating objects multiple times
10217 within a sequence point.
10219 In most expressions, it is intuitively obvious what is a read and what is
10220 a write. For instance
10223 volatile int *dst = @var{somevalue};
10224 volatile int *src = @var{someothervalue};
10229 will cause a read of the volatile object pointed to by @var{src} and stores the
10230 value into the volatile object pointed to by @var{dst}. There is no
10231 guarantee that these reads and writes are atomic, especially for objects
10232 larger than @code{int}.
10234 Less obvious expressions are where something which looks like an access
10235 is used in a void context. An example would be,
10238 volatile int *src = @var{somevalue};
10242 With C, such expressions are rvalues, and as rvalues cause a read of
10243 the object, GCC interprets this as a read of the volatile being pointed
10244 to. The C++ standard specifies that such expressions do not undergo
10245 lvalue to rvalue conversion, and that the type of the dereferenced
10246 object may be incomplete. The C++ standard does not specify explicitly
10247 that it is this lvalue to rvalue conversion which is responsible for
10248 causing an access. However, there is reason to believe that it is,
10249 because otherwise certain simple expressions become undefined. However,
10250 because it would surprise most programmers, G++ treats dereferencing a
10251 pointer to volatile object of complete type in a void context as a read
10252 of the object. When the object has incomplete type, G++ issues a
10257 struct T @{int m;@};
10258 volatile S *ptr1 = @var{somevalue};
10259 volatile T *ptr2 = @var{somevalue};
10264 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
10265 causes a read of the object pointed to. If you wish to force an error on
10266 the first case, you must force a conversion to rvalue with, for instance
10267 a static cast, @code{static_cast<S>(*ptr1)}.
10269 When using a reference to volatile, G++ does not treat equivalent
10270 expressions as accesses to volatiles, but instead issues a warning that
10271 no volatile is accessed. The rationale for this is that otherwise it
10272 becomes difficult to determine where volatile access occur, and not
10273 possible to ignore the return value from functions returning volatile
10274 references. Again, if you wish to force a read, cast the reference to
10277 @node Restricted Pointers
10278 @section Restricting Pointer Aliasing
10279 @cindex restricted pointers
10280 @cindex restricted references
10281 @cindex restricted this pointer
10283 As with the C front end, G++ understands the C99 feature of restricted pointers,
10284 specified with the @code{__restrict__}, or @code{__restrict} type
10285 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10286 language flag, @code{restrict} is not a keyword in C++.
10288 In addition to allowing restricted pointers, you can specify restricted
10289 references, which indicate that the reference is not aliased in the local
10293 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10300 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10301 @var{rref} refers to a (different) unaliased integer.
10303 You may also specify whether a member function's @var{this} pointer is
10304 unaliased by using @code{__restrict__} as a member function qualifier.
10307 void T::fn () __restrict__
10314 Within the body of @code{T::fn}, @var{this} will have the effective
10315 definition @code{T *__restrict__ const this}. Notice that the
10316 interpretation of a @code{__restrict__} member function qualifier is
10317 different to that of @code{const} or @code{volatile} qualifier, in that it
10318 is applied to the pointer rather than the object. This is consistent with
10319 other compilers which implement restricted pointers.
10321 As with all outermost parameter qualifiers, @code{__restrict__} is
10322 ignored in function definition matching. This means you only need to
10323 specify @code{__restrict__} in a function definition, rather than
10324 in a function prototype as well.
10326 @node Vague Linkage
10327 @section Vague Linkage
10328 @cindex vague linkage
10330 There are several constructs in C++ which require space in the object
10331 file but are not clearly tied to a single translation unit. We say that
10332 these constructs have ``vague linkage''. Typically such constructs are
10333 emitted wherever they are needed, though sometimes we can be more
10337 @item Inline Functions
10338 Inline functions are typically defined in a header file which can be
10339 included in many different compilations. Hopefully they can usually be
10340 inlined, but sometimes an out-of-line copy is necessary, if the address
10341 of the function is taken or if inlining fails. In general, we emit an
10342 out-of-line copy in all translation units where one is needed. As an
10343 exception, we only emit inline virtual functions with the vtable, since
10344 it will always require a copy.
10346 Local static variables and string constants used in an inline function
10347 are also considered to have vague linkage, since they must be shared
10348 between all inlined and out-of-line instances of the function.
10352 C++ virtual functions are implemented in most compilers using a lookup
10353 table, known as a vtable. The vtable contains pointers to the virtual
10354 functions provided by a class, and each object of the class contains a
10355 pointer to its vtable (or vtables, in some multiple-inheritance
10356 situations). If the class declares any non-inline, non-pure virtual
10357 functions, the first one is chosen as the ``key method'' for the class,
10358 and the vtable is only emitted in the translation unit where the key
10361 @emph{Note:} If the chosen key method is later defined as inline, the
10362 vtable will still be emitted in every translation unit which defines it.
10363 Make sure that any inline virtuals are declared inline in the class
10364 body, even if they are not defined there.
10366 @item type_info objects
10369 C++ requires information about types to be written out in order to
10370 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10371 For polymorphic classes (classes with virtual functions), the type_info
10372 object is written out along with the vtable so that @samp{dynamic_cast}
10373 can determine the dynamic type of a class object at runtime. For all
10374 other types, we write out the type_info object when it is used: when
10375 applying @samp{typeid} to an expression, throwing an object, or
10376 referring to a type in a catch clause or exception specification.
10378 @item Template Instantiations
10379 Most everything in this section also applies to template instantiations,
10380 but there are other options as well.
10381 @xref{Template Instantiation,,Where's the Template?}.
10385 When used with GNU ld version 2.8 or later on an ELF system such as
10386 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10387 these constructs will be discarded at link time. This is known as
10390 On targets that don't support COMDAT, but do support weak symbols, GCC
10391 will use them. This way one copy will override all the others, but
10392 the unused copies will still take up space in the executable.
10394 For targets which do not support either COMDAT or weak symbols,
10395 most entities with vague linkage will be emitted as local symbols to
10396 avoid duplicate definition errors from the linker. This will not happen
10397 for local statics in inlines, however, as having multiple copies will
10398 almost certainly break things.
10400 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10401 another way to control placement of these constructs.
10403 @node C++ Interface
10404 @section #pragma interface and implementation
10406 @cindex interface and implementation headers, C++
10407 @cindex C++ interface and implementation headers
10408 @cindex pragmas, interface and implementation
10410 @code{#pragma interface} and @code{#pragma implementation} provide the
10411 user with a way of explicitly directing the compiler to emit entities
10412 with vague linkage (and debugging information) in a particular
10415 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10416 most cases, because of COMDAT support and the ``key method'' heuristic
10417 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10418 program to grow due to unnecessary out-of-line copies of inline
10419 functions. Currently (3.4) the only benefit of these
10420 @code{#pragma}s is reduced duplication of debugging information, and
10421 that should be addressed soon on DWARF 2 targets with the use of
10425 @item #pragma interface
10426 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10427 @kindex #pragma interface
10428 Use this directive in @emph{header files} that define object classes, to save
10429 space in most of the object files that use those classes. Normally,
10430 local copies of certain information (backup copies of inline member
10431 functions, debugging information, and the internal tables that implement
10432 virtual functions) must be kept in each object file that includes class
10433 definitions. You can use this pragma to avoid such duplication. When a
10434 header file containing @samp{#pragma interface} is included in a
10435 compilation, this auxiliary information will not be generated (unless
10436 the main input source file itself uses @samp{#pragma implementation}).
10437 Instead, the object files will contain references to be resolved at link
10440 The second form of this directive is useful for the case where you have
10441 multiple headers with the same name in different directories. If you
10442 use this form, you must specify the same string to @samp{#pragma
10445 @item #pragma implementation
10446 @itemx #pragma implementation "@var{objects}.h"
10447 @kindex #pragma implementation
10448 Use this pragma in a @emph{main input file}, when you want full output from
10449 included header files to be generated (and made globally visible). The
10450 included header file, in turn, should use @samp{#pragma interface}.
10451 Backup copies of inline member functions, debugging information, and the
10452 internal tables used to implement virtual functions are all generated in
10453 implementation files.
10455 @cindex implied @code{#pragma implementation}
10456 @cindex @code{#pragma implementation}, implied
10457 @cindex naming convention, implementation headers
10458 If you use @samp{#pragma implementation} with no argument, it applies to
10459 an include file with the same basename@footnote{A file's @dfn{basename}
10460 was the name stripped of all leading path information and of trailing
10461 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10462 file. For example, in @file{allclass.cc}, giving just
10463 @samp{#pragma implementation}
10464 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10466 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10467 an implementation file whenever you would include it from
10468 @file{allclass.cc} even if you never specified @samp{#pragma
10469 implementation}. This was deemed to be more trouble than it was worth,
10470 however, and disabled.
10472 Use the string argument if you want a single implementation file to
10473 include code from multiple header files. (You must also use
10474 @samp{#include} to include the header file; @samp{#pragma
10475 implementation} only specifies how to use the file---it doesn't actually
10478 There is no way to split up the contents of a single header file into
10479 multiple implementation files.
10482 @cindex inlining and C++ pragmas
10483 @cindex C++ pragmas, effect on inlining
10484 @cindex pragmas in C++, effect on inlining
10485 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10486 effect on function inlining.
10488 If you define a class in a header file marked with @samp{#pragma
10489 interface}, the effect on an inline function defined in that class is
10490 similar to an explicit @code{extern} declaration---the compiler emits
10491 no code at all to define an independent version of the function. Its
10492 definition is used only for inlining with its callers.
10494 @opindex fno-implement-inlines
10495 Conversely, when you include the same header file in a main source file
10496 that declares it as @samp{#pragma implementation}, the compiler emits
10497 code for the function itself; this defines a version of the function
10498 that can be found via pointers (or by callers compiled without
10499 inlining). If all calls to the function can be inlined, you can avoid
10500 emitting the function by compiling with @option{-fno-implement-inlines}.
10501 If any calls were not inlined, you will get linker errors.
10503 @node Template Instantiation
10504 @section Where's the Template?
10505 @cindex template instantiation
10507 C++ templates are the first language feature to require more
10508 intelligence from the environment than one usually finds on a UNIX
10509 system. Somehow the compiler and linker have to make sure that each
10510 template instance occurs exactly once in the executable if it is needed,
10511 and not at all otherwise. There are two basic approaches to this
10512 problem, which are referred to as the Borland model and the Cfront model.
10515 @item Borland model
10516 Borland C++ solved the template instantiation problem by adding the code
10517 equivalent of common blocks to their linker; the compiler emits template
10518 instances in each translation unit that uses them, and the linker
10519 collapses them together. The advantage of this model is that the linker
10520 only has to consider the object files themselves; there is no external
10521 complexity to worry about. This disadvantage is that compilation time
10522 is increased because the template code is being compiled repeatedly.
10523 Code written for this model tends to include definitions of all
10524 templates in the header file, since they must be seen to be
10528 The AT&T C++ translator, Cfront, solved the template instantiation
10529 problem by creating the notion of a template repository, an
10530 automatically maintained place where template instances are stored. A
10531 more modern version of the repository works as follows: As individual
10532 object files are built, the compiler places any template definitions and
10533 instantiations encountered in the repository. At link time, the link
10534 wrapper adds in the objects in the repository and compiles any needed
10535 instances that were not previously emitted. The advantages of this
10536 model are more optimal compilation speed and the ability to use the
10537 system linker; to implement the Borland model a compiler vendor also
10538 needs to replace the linker. The disadvantages are vastly increased
10539 complexity, and thus potential for error; for some code this can be
10540 just as transparent, but in practice it can been very difficult to build
10541 multiple programs in one directory and one program in multiple
10542 directories. Code written for this model tends to separate definitions
10543 of non-inline member templates into a separate file, which should be
10544 compiled separately.
10547 When used with GNU ld version 2.8 or later on an ELF system such as
10548 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10549 Borland model. On other systems, G++ implements neither automatic
10552 A future version of G++ will support a hybrid model whereby the compiler
10553 will emit any instantiations for which the template definition is
10554 included in the compile, and store template definitions and
10555 instantiation context information into the object file for the rest.
10556 The link wrapper will extract that information as necessary and invoke
10557 the compiler to produce the remaining instantiations. The linker will
10558 then combine duplicate instantiations.
10560 In the mean time, you have the following options for dealing with
10561 template instantiations:
10566 Compile your template-using code with @option{-frepo}. The compiler will
10567 generate files with the extension @samp{.rpo} listing all of the
10568 template instantiations used in the corresponding object files which
10569 could be instantiated there; the link wrapper, @samp{collect2}, will
10570 then update the @samp{.rpo} files to tell the compiler where to place
10571 those instantiations and rebuild any affected object files. The
10572 link-time overhead is negligible after the first pass, as the compiler
10573 will continue to place the instantiations in the same files.
10575 This is your best option for application code written for the Borland
10576 model, as it will just work. Code written for the Cfront model will
10577 need to be modified so that the template definitions are available at
10578 one or more points of instantiation; usually this is as simple as adding
10579 @code{#include <tmethods.cc>} to the end of each template header.
10581 For library code, if you want the library to provide all of the template
10582 instantiations it needs, just try to link all of its object files
10583 together; the link will fail, but cause the instantiations to be
10584 generated as a side effect. Be warned, however, that this may cause
10585 conflicts if multiple libraries try to provide the same instantiations.
10586 For greater control, use explicit instantiation as described in the next
10590 @opindex fno-implicit-templates
10591 Compile your code with @option{-fno-implicit-templates} to disable the
10592 implicit generation of template instances, and explicitly instantiate
10593 all the ones you use. This approach requires more knowledge of exactly
10594 which instances you need than do the others, but it's less
10595 mysterious and allows greater control. You can scatter the explicit
10596 instantiations throughout your program, perhaps putting them in the
10597 translation units where the instances are used or the translation units
10598 that define the templates themselves; you can put all of the explicit
10599 instantiations you need into one big file; or you can create small files
10606 template class Foo<int>;
10607 template ostream& operator <<
10608 (ostream&, const Foo<int>&);
10611 for each of the instances you need, and create a template instantiation
10612 library from those.
10614 If you are using Cfront-model code, you can probably get away with not
10615 using @option{-fno-implicit-templates} when compiling files that don't
10616 @samp{#include} the member template definitions.
10618 If you use one big file to do the instantiations, you may want to
10619 compile it without @option{-fno-implicit-templates} so you get all of the
10620 instances required by your explicit instantiations (but not by any
10621 other files) without having to specify them as well.
10623 G++ has extended the template instantiation syntax given in the ISO
10624 standard to allow forward declaration of explicit instantiations
10625 (with @code{extern}), instantiation of the compiler support data for a
10626 template class (i.e.@: the vtable) without instantiating any of its
10627 members (with @code{inline}), and instantiation of only the static data
10628 members of a template class, without the support data or member
10629 functions (with (@code{static}):
10632 extern template int max (int, int);
10633 inline template class Foo<int>;
10634 static template class Foo<int>;
10638 Do nothing. Pretend G++ does implement automatic instantiation
10639 management. Code written for the Borland model will work fine, but
10640 each translation unit will contain instances of each of the templates it
10641 uses. In a large program, this can lead to an unacceptable amount of code
10645 @node Bound member functions
10646 @section Extracting the function pointer from a bound pointer to member function
10648 @cindex pointer to member function
10649 @cindex bound pointer to member function
10651 In C++, pointer to member functions (PMFs) are implemented using a wide
10652 pointer of sorts to handle all the possible call mechanisms; the PMF
10653 needs to store information about how to adjust the @samp{this} pointer,
10654 and if the function pointed to is virtual, where to find the vtable, and
10655 where in the vtable to look for the member function. If you are using
10656 PMFs in an inner loop, you should really reconsider that decision. If
10657 that is not an option, you can extract the pointer to the function that
10658 would be called for a given object/PMF pair and call it directly inside
10659 the inner loop, to save a bit of time.
10661 Note that you will still be paying the penalty for the call through a
10662 function pointer; on most modern architectures, such a call defeats the
10663 branch prediction features of the CPU@. This is also true of normal
10664 virtual function calls.
10666 The syntax for this extension is
10670 extern int (A::*fp)();
10671 typedef int (*fptr)(A *);
10673 fptr p = (fptr)(a.*fp);
10676 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
10677 no object is needed to obtain the address of the function. They can be
10678 converted to function pointers directly:
10681 fptr p1 = (fptr)(&A::foo);
10684 @opindex Wno-pmf-conversions
10685 You must specify @option{-Wno-pmf-conversions} to use this extension.
10687 @node C++ Attributes
10688 @section C++-Specific Variable, Function, and Type Attributes
10690 Some attributes only make sense for C++ programs.
10693 @item init_priority (@var{priority})
10694 @cindex init_priority attribute
10697 In Standard C++, objects defined at namespace scope are guaranteed to be
10698 initialized in an order in strict accordance with that of their definitions
10699 @emph{in a given translation unit}. No guarantee is made for initializations
10700 across translation units. However, GNU C++ allows users to control the
10701 order of initialization of objects defined at namespace scope with the
10702 @code{init_priority} attribute by specifying a relative @var{priority},
10703 a constant integral expression currently bounded between 101 and 65535
10704 inclusive. Lower numbers indicate a higher priority.
10706 In the following example, @code{A} would normally be created before
10707 @code{B}, but the @code{init_priority} attribute has reversed that order:
10710 Some_Class A __attribute__ ((init_priority (2000)));
10711 Some_Class B __attribute__ ((init_priority (543)));
10715 Note that the particular values of @var{priority} do not matter; only their
10718 @item java_interface
10719 @cindex java_interface attribute
10721 This type attribute informs C++ that the class is a Java interface. It may
10722 only be applied to classes declared within an @code{extern "Java"} block.
10723 Calls to methods declared in this interface will be dispatched using GCJ's
10724 interface table mechanism, instead of regular virtual table dispatch.
10728 See also @xref{Namespace Association}.
10730 @node Namespace Association
10731 @section Namespace Association
10733 @strong{Caution:} The semantics of this extension are not fully
10734 defined. Users should refrain from using this extension as its
10735 semantics may change subtly over time. It is possible that this
10736 extension will be removed in future versions of G++.
10738 A using-directive with @code{__attribute ((strong))} is stronger
10739 than a normal using-directive in two ways:
10743 Templates from the used namespace can be specialized and explicitly
10744 instantiated as though they were members of the using namespace.
10747 The using namespace is considered an associated namespace of all
10748 templates in the used namespace for purposes of argument-dependent
10752 The used namespace must be nested within the using namespace so that
10753 normal unqualified lookup works properly.
10755 This is useful for composing a namespace transparently from
10756 implementation namespaces. For example:
10761 template <class T> struct A @{ @};
10763 using namespace debug __attribute ((__strong__));
10764 template <> struct A<int> @{ @}; // @r{ok to specialize}
10766 template <class T> void f (A<T>);
10771 f (std::A<float>()); // @r{lookup finds} std::f
10776 @node Java Exceptions
10777 @section Java Exceptions
10779 The Java language uses a slightly different exception handling model
10780 from C++. Normally, GNU C++ will automatically detect when you are
10781 writing C++ code that uses Java exceptions, and handle them
10782 appropriately. However, if C++ code only needs to execute destructors
10783 when Java exceptions are thrown through it, GCC will guess incorrectly.
10784 Sample problematic code is:
10787 struct S @{ ~S(); @};
10788 extern void bar(); // @r{is written in Java, and may throw exceptions}
10797 The usual effect of an incorrect guess is a link failure, complaining of
10798 a missing routine called @samp{__gxx_personality_v0}.
10800 You can inform the compiler that Java exceptions are to be used in a
10801 translation unit, irrespective of what it might think, by writing
10802 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
10803 @samp{#pragma} must appear before any functions that throw or catch
10804 exceptions, or run destructors when exceptions are thrown through them.
10806 You cannot mix Java and C++ exceptions in the same translation unit. It
10807 is believed to be safe to throw a C++ exception from one file through
10808 another file compiled for the Java exception model, or vice versa, but
10809 there may be bugs in this area.
10811 @node Deprecated Features
10812 @section Deprecated Features
10814 In the past, the GNU C++ compiler was extended to experiment with new
10815 features, at a time when the C++ language was still evolving. Now that
10816 the C++ standard is complete, some of those features are superseded by
10817 superior alternatives. Using the old features might cause a warning in
10818 some cases that the feature will be dropped in the future. In other
10819 cases, the feature might be gone already.
10821 While the list below is not exhaustive, it documents some of the options
10822 that are now deprecated:
10825 @item -fexternal-templates
10826 @itemx -falt-external-templates
10827 These are two of the many ways for G++ to implement template
10828 instantiation. @xref{Template Instantiation}. The C++ standard clearly
10829 defines how template definitions have to be organized across
10830 implementation units. G++ has an implicit instantiation mechanism that
10831 should work just fine for standard-conforming code.
10833 @item -fstrict-prototype
10834 @itemx -fno-strict-prototype
10835 Previously it was possible to use an empty prototype parameter list to
10836 indicate an unspecified number of parameters (like C), rather than no
10837 parameters, as C++ demands. This feature has been removed, except where
10838 it is required for backwards compatibility @xref{Backwards Compatibility}.
10841 G++ allows a virtual function returning @samp{void *} to be overridden
10842 by one returning a different pointer type. This extension to the
10843 covariant return type rules is now deprecated and will be removed from a
10846 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
10847 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
10848 and will be removed in a future version. Code using these operators
10849 should be modified to use @code{std::min} and @code{std::max} instead.
10851 The named return value extension has been deprecated, and is now
10854 The use of initializer lists with new expressions has been deprecated,
10855 and is now removed from G++.
10857 Floating and complex non-type template parameters have been deprecated,
10858 and are now removed from G++.
10860 The implicit typename extension has been deprecated and is now
10863 The use of default arguments in function pointers, function typedefs and
10864 and other places where they are not permitted by the standard is
10865 deprecated and will be removed from a future version of G++.
10867 G++ allows floating-point literals to appear in integral constant expressions,
10868 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
10869 This extension is deprecated and will be removed from a future version.
10871 G++ allows static data members of const floating-point type to be declared
10872 with an initializer in a class definition. The standard only allows
10873 initializers for static members of const integral types and const
10874 enumeration types so this extension has been deprecated and will be removed
10875 from a future version.
10877 @node Backwards Compatibility
10878 @section Backwards Compatibility
10879 @cindex Backwards Compatibility
10880 @cindex ARM [Annotated C++ Reference Manual]
10882 Now that there is a definitive ISO standard C++, G++ has a specification
10883 to adhere to. The C++ language evolved over time, and features that
10884 used to be acceptable in previous drafts of the standard, such as the ARM
10885 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
10886 compilation of C++ written to such drafts, G++ contains some backwards
10887 compatibilities. @emph{All such backwards compatibility features are
10888 liable to disappear in future versions of G++.} They should be considered
10889 deprecated @xref{Deprecated Features}.
10893 If a variable is declared at for scope, it used to remain in scope until
10894 the end of the scope which contained the for statement (rather than just
10895 within the for scope). G++ retains this, but issues a warning, if such a
10896 variable is accessed outside the for scope.
10898 @item Implicit C language
10899 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
10900 scope to set the language. On such systems, all header files are
10901 implicitly scoped inside a C language scope. Also, an empty prototype
10902 @code{()} will be treated as an unspecified number of arguments, rather
10903 than no arguments, as C++ demands.