1 @c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002,2003,2004,2005
2 @c Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Extensions to the C Language Family
8 @cindex extensions, C language
9 @cindex C language extensions
12 GNU C provides several language features not found in ISO standard C@.
13 (The @option{-pedantic} option directs GCC to print a warning message if
14 any of these features is used.) To test for the availability of these
15 features in conditional compilation, check for a predefined macro
16 @code{__GNUC__}, which is always defined under GCC@.
18 These extensions are available in C and Objective-C@. Most of them are
19 also available in C++. @xref{C++ Extensions,,Extensions to the
20 C++ Language}, for extensions that apply @emph{only} to C++.
22 Some features that are in ISO C99 but not C89 or C++ are also, as
23 extensions, accepted by GCC in C89 mode and in C++.
26 * Statement Exprs:: Putting statements and declarations inside expressions.
27 * Local Labels:: Labels local to a block.
28 * Labels as Values:: Getting pointers to labels, and computed gotos.
29 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
30 * Constructing Calls:: Dispatching a call to another function.
31 * Typeof:: @code{typeof}: referring to the type of an expression.
32 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
33 * Long Long:: Double-word integers---@code{long long int}.
34 * Complex:: Data types for complex numbers.
35 * Hex Floats:: Hexadecimal floating-point constants.
36 * Zero Length:: Zero-length arrays.
37 * Variable Length:: Arrays whose length is computed at run time.
38 * Empty Structures:: Structures with no members.
39 * Variadic Macros:: Macros with a variable number of arguments.
40 * Escaped Newlines:: Slightly looser rules for escaped newlines.
41 * Subscripting:: Any array can be subscripted, even if not an lvalue.
42 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
43 * Initializers:: Non-constant initializers.
44 * Compound Literals:: Compound literals give structures, unions
46 * Designated Inits:: Labeling elements of initializers.
47 * Cast to Union:: Casting to union type from any member of the union.
48 * Case Ranges:: `case 1 ... 9' and such.
49 * Mixed Declarations:: Mixing declarations and code.
50 * Function Attributes:: Declaring that functions have no side effects,
51 or that they can never return.
52 * Attribute Syntax:: Formal syntax for attributes.
53 * Function Prototypes:: Prototype declarations and old-style definitions.
54 * C++ Comments:: C++ comments are recognized.
55 * Dollar Signs:: Dollar sign is allowed in identifiers.
56 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
57 * Variable Attributes:: Specifying attributes of variables.
58 * Type Attributes:: Specifying attributes of types.
59 * Alignment:: Inquiring about the alignment of a type or variable.
60 * Inline:: Defining inline functions (as fast as macros).
61 * Extended Asm:: Assembler instructions with C expressions as operands.
62 (With them you can define ``built-in'' functions.)
63 * Constraints:: Constraints for asm operands
64 * Asm Labels:: Specifying the assembler name to use for a C symbol.
65 * Explicit Reg Vars:: Defining variables residing in specified registers.
66 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
67 * Incomplete Enums:: @code{enum foo;}, with details to follow.
68 * Function Names:: Printable strings which are the name of the current
70 * Return Address:: Getting the return or frame address of a function.
71 * Vector Extensions:: Using vector instructions through built-in functions.
72 * Offsetof:: Special syntax for implementing @code{offsetof}.
73 * Atomic Builtins:: Built-in functions for atomic memory access.
74 * Other Builtins:: Other built-in functions.
75 * Target Builtins:: Built-in functions specific to particular targets.
76 * Target Format Checks:: Format checks specific to particular targets.
77 * Pragmas:: Pragmas accepted by GCC.
78 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
79 * Thread-Local:: Per-thread variables.
83 @section Statements and Declarations in Expressions
84 @cindex statements inside expressions
85 @cindex declarations inside expressions
86 @cindex expressions containing statements
87 @cindex macros, statements in expressions
89 @c the above section title wrapped and causes an underfull hbox.. i
90 @c changed it from "within" to "in". --mew 4feb93
91 A compound statement enclosed in parentheses may appear as an expression
92 in GNU C@. This allows you to use loops, switches, and local variables
95 Recall that a compound statement is a sequence of statements surrounded
96 by braces; in this construct, parentheses go around the braces. For
100 (@{ int y = foo (); int z;
107 is a valid (though slightly more complex than necessary) expression
108 for the absolute value of @code{foo ()}.
110 The last thing in the compound statement should be an expression
111 followed by a semicolon; the value of this subexpression serves as the
112 value of the entire construct. (If you use some other kind of statement
113 last within the braces, the construct has type @code{void}, and thus
114 effectively no value.)
116 This feature is especially useful in making macro definitions ``safe'' (so
117 that they evaluate each operand exactly once). For example, the
118 ``maximum'' function is commonly defined as a macro in standard C as
122 #define max(a,b) ((a) > (b) ? (a) : (b))
126 @cindex side effects, macro argument
127 But this definition computes either @var{a} or @var{b} twice, with bad
128 results if the operand has side effects. In GNU C, if you know the
129 type of the operands (here taken as @code{int}), you can define
130 the macro safely as follows:
133 #define maxint(a,b) \
134 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
137 Embedded statements are not allowed in constant expressions, such as
138 the value of an enumeration constant, the width of a bit-field, or
139 the initial value of a static variable.
141 If you don't know the type of the operand, you can still do this, but you
142 must use @code{typeof} (@pxref{Typeof}).
144 In G++, the result value of a statement expression undergoes array and
145 function pointer decay, and is returned by value to the enclosing
146 expression. For instance, if @code{A} is a class, then
155 will construct a temporary @code{A} object to hold the result of the
156 statement expression, and that will be used to invoke @code{Foo}.
157 Therefore the @code{this} pointer observed by @code{Foo} will not be the
160 Any temporaries created within a statement within a statement expression
161 will be destroyed at the statement's end. This makes statement
162 expressions inside macros slightly different from function calls. In
163 the latter case temporaries introduced during argument evaluation will
164 be destroyed at the end of the statement that includes the function
165 call. In the statement expression case they will be destroyed during
166 the statement expression. For instance,
169 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
170 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
180 will have different places where temporaries are destroyed. For the
181 @code{macro} case, the temporary @code{X} will be destroyed just after
182 the initialization of @code{b}. In the @code{function} case that
183 temporary will be destroyed when the function returns.
185 These considerations mean that it is probably a bad idea to use
186 statement-expressions of this form in header files that are designed to
187 work with C++. (Note that some versions of the GNU C Library contained
188 header files using statement-expression that lead to precisely this
191 Jumping into a statement expression with @code{goto} or using a
192 @code{switch} statement outside the statement expression with a
193 @code{case} or @code{default} label inside the statement expression is
194 not permitted. Jumping into a statement expression with a computed
195 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
196 Jumping out of a statement expression is permitted, but if the
197 statement expression is part of a larger expression then it is
198 unspecified which other subexpressions of that expression have been
199 evaluated except where the language definition requires certain
200 subexpressions to be evaluated before or after the statement
201 expression. In any case, as with a function call the evaluation of a
202 statement expression is not interleaved with the evaluation of other
203 parts of the containing expression. For example,
206 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
210 will call @code{foo} and @code{bar1} and will not call @code{baz} but
211 may or may not call @code{bar2}. If @code{bar2} is called, it will be
212 called after @code{foo} and before @code{bar1}
215 @section Locally Declared Labels
217 @cindex macros, local labels
219 GCC allows you to declare @dfn{local labels} in any nested block
220 scope. A local label is just like an ordinary label, but you can
221 only reference it (with a @code{goto} statement, or by taking its
222 address) within the block in which it was declared.
224 A local label declaration looks like this:
227 __label__ @var{label};
234 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
237 Local label declarations must come at the beginning of the block,
238 before any ordinary declarations or statements.
240 The label declaration defines the label @emph{name}, but does not define
241 the label itself. You must do this in the usual way, with
242 @code{@var{label}:}, within the statements of the statement expression.
244 The local label feature is useful for complex macros. If a macro
245 contains nested loops, a @code{goto} can be useful for breaking out of
246 them. However, an ordinary label whose scope is the whole function
247 cannot be used: if the macro can be expanded several times in one
248 function, the label will be multiply defined in that function. A
249 local label avoids this problem. For example:
252 #define SEARCH(value, array, target) \
255 typeof (target) _SEARCH_target = (target); \
256 typeof (*(array)) *_SEARCH_array = (array); \
259 for (i = 0; i < max; i++) \
260 for (j = 0; j < max; j++) \
261 if (_SEARCH_array[i][j] == _SEARCH_target) \
262 @{ (value) = i; goto found; @} \
268 This could also be written using a statement-expression:
271 #define SEARCH(array, target) \
274 typeof (target) _SEARCH_target = (target); \
275 typeof (*(array)) *_SEARCH_array = (array); \
278 for (i = 0; i < max; i++) \
279 for (j = 0; j < max; j++) \
280 if (_SEARCH_array[i][j] == _SEARCH_target) \
281 @{ value = i; goto found; @} \
288 Local label declarations also make the labels they declare visible to
289 nested functions, if there are any. @xref{Nested Functions}, for details.
291 @node Labels as Values
292 @section Labels as Values
293 @cindex labels as values
294 @cindex computed gotos
295 @cindex goto with computed label
296 @cindex address of a label
298 You can get the address of a label defined in the current function
299 (or a containing function) with the unary operator @samp{&&}. The
300 value has type @code{void *}. This value is a constant and can be used
301 wherever a constant of that type is valid. For example:
309 To use these values, you need to be able to jump to one. This is done
310 with the computed goto statement@footnote{The analogous feature in
311 Fortran is called an assigned goto, but that name seems inappropriate in
312 C, where one can do more than simply store label addresses in label
313 variables.}, @code{goto *@var{exp};}. For example,
320 Any expression of type @code{void *} is allowed.
322 One way of using these constants is in initializing a static array that
323 will serve as a jump table:
326 static void *array[] = @{ &&foo, &&bar, &&hack @};
329 Then you can select a label with indexing, like this:
336 Note that this does not check whether the subscript is in bounds---array
337 indexing in C never does that.
339 Such an array of label values serves a purpose much like that of the
340 @code{switch} statement. The @code{switch} statement is cleaner, so
341 use that rather than an array unless the problem does not fit a
342 @code{switch} statement very well.
344 Another use of label values is in an interpreter for threaded code.
345 The labels within the interpreter function can be stored in the
346 threaded code for super-fast dispatching.
348 You may not use this mechanism to jump to code in a different function.
349 If you do that, totally unpredictable things will happen. The best way to
350 avoid this is to store the label address only in automatic variables and
351 never pass it as an argument.
353 An alternate way to write the above example is
356 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
358 goto *(&&foo + array[i]);
362 This is more friendly to code living in shared libraries, as it reduces
363 the number of dynamic relocations that are needed, and by consequence,
364 allows the data to be read-only.
366 @node Nested Functions
367 @section Nested Functions
368 @cindex nested functions
369 @cindex downward funargs
372 A @dfn{nested function} is a function defined inside another function.
373 (Nested functions are not supported for GNU C++.) The nested function's
374 name is local to the block where it is defined. For example, here we
375 define a nested function named @code{square}, and call it twice:
379 foo (double a, double b)
381 double square (double z) @{ return z * z; @}
383 return square (a) + square (b);
388 The nested function can access all the variables of the containing
389 function that are visible at the point of its definition. This is
390 called @dfn{lexical scoping}. For example, here we show a nested
391 function which uses an inherited variable named @code{offset}:
395 bar (int *array, int offset, int size)
397 int access (int *array, int index)
398 @{ return array[index + offset]; @}
401 for (i = 0; i < size; i++)
402 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
407 Nested function definitions are permitted within functions in the places
408 where variable definitions are allowed; that is, in any block, mixed
409 with the other declarations and statements in the block.
411 It is possible to call the nested function from outside the scope of its
412 name by storing its address or passing the address to another function:
415 hack (int *array, int size)
417 void store (int index, int value)
418 @{ array[index] = value; @}
420 intermediate (store, size);
424 Here, the function @code{intermediate} receives the address of
425 @code{store} as an argument. If @code{intermediate} calls @code{store},
426 the arguments given to @code{store} are used to store into @code{array}.
427 But this technique works only so long as the containing function
428 (@code{hack}, in this example) does not exit.
430 If you try to call the nested function through its address after the
431 containing function has exited, all hell will break loose. If you try
432 to call it after a containing scope level has exited, and if it refers
433 to some of the variables that are no longer in scope, you may be lucky,
434 but it's not wise to take the risk. If, however, the nested function
435 does not refer to anything that has gone out of scope, you should be
438 GCC implements taking the address of a nested function using a technique
439 called @dfn{trampolines}. A paper describing them is available as
442 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
444 A nested function can jump to a label inherited from a containing
445 function, provided the label was explicitly declared in the containing
446 function (@pxref{Local Labels}). Such a jump returns instantly to the
447 containing function, exiting the nested function which did the
448 @code{goto} and any intermediate functions as well. Here is an example:
452 bar (int *array, int offset, int size)
455 int access (int *array, int index)
459 return array[index + offset];
463 for (i = 0; i < size; i++)
464 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
468 /* @r{Control comes here from @code{access}
469 if it detects an error.} */
476 A nested function always has no linkage. Declaring one with
477 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
478 before its definition, use @code{auto} (which is otherwise meaningless
479 for function declarations).
482 bar (int *array, int offset, int size)
485 auto int access (int *, int);
487 int access (int *array, int index)
491 return array[index + offset];
497 @node Constructing Calls
498 @section Constructing Function Calls
499 @cindex constructing calls
500 @cindex forwarding calls
502 Using the built-in functions described below, you can record
503 the arguments a function received, and call another function
504 with the same arguments, without knowing the number or types
507 You can also record the return value of that function call,
508 and later return that value, without knowing what data type
509 the function tried to return (as long as your caller expects
512 However, these built-in functions may interact badly with some
513 sophisticated features or other extensions of the language. It
514 is, therefore, not recommended to use them outside very simple
515 functions acting as mere forwarders for their arguments.
517 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
518 This built-in function returns a pointer to data
519 describing how to perform a call with the same arguments as were passed
520 to the current function.
522 The function saves the arg pointer register, structure value address,
523 and all registers that might be used to pass arguments to a function
524 into a block of memory allocated on the stack. Then it returns the
525 address of that block.
528 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
529 This built-in function invokes @var{function}
530 with a copy of the parameters described by @var{arguments}
533 The value of @var{arguments} should be the value returned by
534 @code{__builtin_apply_args}. The argument @var{size} specifies the size
535 of the stack argument data, in bytes.
537 This function returns a pointer to data describing
538 how to return whatever value was returned by @var{function}. The data
539 is saved in a block of memory allocated on the stack.
541 It is not always simple to compute the proper value for @var{size}. The
542 value is used by @code{__builtin_apply} to compute the amount of data
543 that should be pushed on the stack and copied from the incoming argument
547 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
548 This built-in function returns the value described by @var{result} from
549 the containing function. You should specify, for @var{result}, a value
550 returned by @code{__builtin_apply}.
554 @section Referring to a Type with @code{typeof}
557 @cindex macros, types of arguments
559 Another way to refer to the type of an expression is with @code{typeof}.
560 The syntax of using of this keyword looks like @code{sizeof}, but the
561 construct acts semantically like a type name defined with @code{typedef}.
563 There are two ways of writing the argument to @code{typeof}: with an
564 expression or with a type. Here is an example with an expression:
571 This assumes that @code{x} is an array of pointers to functions;
572 the type described is that of the values of the functions.
574 Here is an example with a typename as the argument:
581 Here the type described is that of pointers to @code{int}.
583 If you are writing a header file that must work when included in ISO C
584 programs, write @code{__typeof__} instead of @code{typeof}.
585 @xref{Alternate Keywords}.
587 A @code{typeof}-construct can be used anywhere a typedef name could be
588 used. For example, you can use it in a declaration, in a cast, or inside
589 of @code{sizeof} or @code{typeof}.
591 @code{typeof} is often useful in conjunction with the
592 statements-within-expressions feature. Here is how the two together can
593 be used to define a safe ``maximum'' macro that operates on any
594 arithmetic type and evaluates each of its arguments exactly once:
598 (@{ typeof (a) _a = (a); \
599 typeof (b) _b = (b); \
600 _a > _b ? _a : _b; @})
603 @cindex underscores in variables in macros
604 @cindex @samp{_} in variables in macros
605 @cindex local variables in macros
606 @cindex variables, local, in macros
607 @cindex macros, local variables in
609 The reason for using names that start with underscores for the local
610 variables is to avoid conflicts with variable names that occur within the
611 expressions that are substituted for @code{a} and @code{b}. Eventually we
612 hope to design a new form of declaration syntax that allows you to declare
613 variables whose scopes start only after their initializers; this will be a
614 more reliable way to prevent such conflicts.
617 Some more examples of the use of @code{typeof}:
621 This declares @code{y} with the type of what @code{x} points to.
628 This declares @code{y} as an array of such values.
635 This declares @code{y} as an array of pointers to characters:
638 typeof (typeof (char *)[4]) y;
642 It is equivalent to the following traditional C declaration:
648 To see the meaning of the declaration using @code{typeof}, and why it
649 might be a useful way to write, rewrite it with these macros:
652 #define pointer(T) typeof(T *)
653 #define array(T, N) typeof(T [N])
657 Now the declaration can be rewritten this way:
660 array (pointer (char), 4) y;
664 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
665 pointers to @code{char}.
668 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
669 a more limited extension which permitted one to write
672 typedef @var{T} = @var{expr};
676 with the effect of declaring @var{T} to have the type of the expression
677 @var{expr}. This extension does not work with GCC 3 (versions between
678 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
679 relies on it should be rewritten to use @code{typeof}:
682 typedef typeof(@var{expr}) @var{T};
686 This will work with all versions of GCC@.
689 @section Conditionals with Omitted Operands
690 @cindex conditional expressions, extensions
691 @cindex omitted middle-operands
692 @cindex middle-operands, omitted
693 @cindex extensions, @code{?:}
694 @cindex @code{?:} extensions
696 The middle operand in a conditional expression may be omitted. Then
697 if the first operand is nonzero, its value is the value of the conditional
700 Therefore, the expression
707 has the value of @code{x} if that is nonzero; otherwise, the value of
710 This example is perfectly equivalent to
716 @cindex side effect in ?:
717 @cindex ?: side effect
719 In this simple case, the ability to omit the middle operand is not
720 especially useful. When it becomes useful is when the first operand does,
721 or may (if it is a macro argument), contain a side effect. Then repeating
722 the operand in the middle would perform the side effect twice. Omitting
723 the middle operand uses the value already computed without the undesirable
724 effects of recomputing it.
727 @section Double-Word Integers
728 @cindex @code{long long} data types
729 @cindex double-word arithmetic
730 @cindex multiprecision arithmetic
731 @cindex @code{LL} integer suffix
732 @cindex @code{ULL} integer suffix
734 ISO C99 supports data types for integers that are at least 64 bits wide,
735 and as an extension GCC supports them in C89 mode and in C++.
736 Simply write @code{long long int} for a signed integer, or
737 @code{unsigned long long int} for an unsigned integer. To make an
738 integer constant of type @code{long long int}, add the suffix @samp{LL}
739 to the integer. To make an integer constant of type @code{unsigned long
740 long int}, add the suffix @samp{ULL} to the integer.
742 You can use these types in arithmetic like any other integer types.
743 Addition, subtraction, and bitwise boolean operations on these types
744 are open-coded on all types of machines. Multiplication is open-coded
745 if the machine supports fullword-to-doubleword a widening multiply
746 instruction. Division and shifts are open-coded only on machines that
747 provide special support. The operations that are not open-coded use
748 special library routines that come with GCC@.
750 There may be pitfalls when you use @code{long long} types for function
751 arguments, unless you declare function prototypes. If a function
752 expects type @code{int} for its argument, and you pass a value of type
753 @code{long long int}, confusion will result because the caller and the
754 subroutine will disagree about the number of bytes for the argument.
755 Likewise, if the function expects @code{long long int} and you pass
756 @code{int}. The best way to avoid such problems is to use prototypes.
759 @section Complex Numbers
760 @cindex complex numbers
761 @cindex @code{_Complex} keyword
762 @cindex @code{__complex__} keyword
764 ISO C99 supports complex floating data types, and as an extension GCC
765 supports them in C89 mode and in C++, and supports complex integer data
766 types which are not part of ISO C99. You can declare complex types
767 using the keyword @code{_Complex}. As an extension, the older GNU
768 keyword @code{__complex__} is also supported.
770 For example, @samp{_Complex double x;} declares @code{x} as a
771 variable whose real part and imaginary part are both of type
772 @code{double}. @samp{_Complex short int y;} declares @code{y} to
773 have real and imaginary parts of type @code{short int}; this is not
774 likely to be useful, but it shows that the set of complex types is
777 To write a constant with a complex data type, use the suffix @samp{i} or
778 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
779 has type @code{_Complex float} and @code{3i} has type
780 @code{_Complex int}. Such a constant always has a pure imaginary
781 value, but you can form any complex value you like by adding one to a
782 real constant. This is a GNU extension; if you have an ISO C99
783 conforming C library (such as GNU libc), and want to construct complex
784 constants of floating type, you should include @code{<complex.h>} and
785 use the macros @code{I} or @code{_Complex_I} instead.
787 @cindex @code{__real__} keyword
788 @cindex @code{__imag__} keyword
789 To extract the real part of a complex-valued expression @var{exp}, write
790 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
791 extract the imaginary part. This is a GNU extension; for values of
792 floating type, you should use the ISO C99 functions @code{crealf},
793 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
794 @code{cimagl}, declared in @code{<complex.h>} and also provided as
795 built-in functions by GCC@.
797 @cindex complex conjugation
798 The operator @samp{~} performs complex conjugation when used on a value
799 with a complex type. This is a GNU extension; for values of
800 floating type, you should use the ISO C99 functions @code{conjf},
801 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
802 provided as built-in functions by GCC@.
804 GCC can allocate complex automatic variables in a noncontiguous
805 fashion; it's even possible for the real part to be in a register while
806 the imaginary part is on the stack (or vice-versa). Only the DWARF2
807 debug info format can represent this, so use of DWARF2 is recommended.
808 If you are using the stabs debug info format, GCC describes a noncontiguous
809 complex variable as if it were two separate variables of noncomplex type.
810 If the variable's actual name is @code{foo}, the two fictitious
811 variables are named @code{foo$real} and @code{foo$imag}. You can
812 examine and set these two fictitious variables with your debugger.
818 ISO C99 supports floating-point numbers written not only in the usual
819 decimal notation, such as @code{1.55e1}, but also numbers such as
820 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
821 supports this in C89 mode (except in some cases when strictly
822 conforming) and in C++. In that format the
823 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
824 mandatory. The exponent is a decimal number that indicates the power of
825 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
832 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
833 is the same as @code{1.55e1}.
835 Unlike for floating-point numbers in the decimal notation the exponent
836 is always required in the hexadecimal notation. Otherwise the compiler
837 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
838 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
839 extension for floating-point constants of type @code{float}.
842 @section Arrays of Length Zero
843 @cindex arrays of length zero
844 @cindex zero-length arrays
845 @cindex length-zero arrays
846 @cindex flexible array members
848 Zero-length arrays are allowed in GNU C@. They are very useful as the
849 last element of a structure which is really a header for a variable-length
858 struct line *thisline = (struct line *)
859 malloc (sizeof (struct line) + this_length);
860 thisline->length = this_length;
863 In ISO C90, you would have to give @code{contents} a length of 1, which
864 means either you waste space or complicate the argument to @code{malloc}.
866 In ISO C99, you would use a @dfn{flexible array member}, which is
867 slightly different in syntax and semantics:
871 Flexible array members are written as @code{contents[]} without
875 Flexible array members have incomplete type, and so the @code{sizeof}
876 operator may not be applied. As a quirk of the original implementation
877 of zero-length arrays, @code{sizeof} evaluates to zero.
880 Flexible array members may only appear as the last member of a
881 @code{struct} that is otherwise non-empty.
884 A structure containing a flexible array member, or a union containing
885 such a structure (possibly recursively), may not be a member of a
886 structure or an element of an array. (However, these uses are
887 permitted by GCC as extensions.)
890 GCC versions before 3.0 allowed zero-length arrays to be statically
891 initialized, as if they were flexible arrays. In addition to those
892 cases that were useful, it also allowed initializations in situations
893 that would corrupt later data. Non-empty initialization of zero-length
894 arrays is now treated like any case where there are more initializer
895 elements than the array holds, in that a suitable warning about "excess
896 elements in array" is given, and the excess elements (all of them, in
897 this case) are ignored.
899 Instead GCC allows static initialization of flexible array members.
900 This is equivalent to defining a new structure containing the original
901 structure followed by an array of sufficient size to contain the data.
902 I.e.@: in the following, @code{f1} is constructed as if it were declared
908 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
911 struct f1 f1; int data[3];
912 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
916 The convenience of this extension is that @code{f1} has the desired
917 type, eliminating the need to consistently refer to @code{f2.f1}.
919 This has symmetry with normal static arrays, in that an array of
920 unknown size is also written with @code{[]}.
922 Of course, this extension only makes sense if the extra data comes at
923 the end of a top-level object, as otherwise we would be overwriting
924 data at subsequent offsets. To avoid undue complication and confusion
925 with initialization of deeply nested arrays, we simply disallow any
926 non-empty initialization except when the structure is the top-level
930 struct foo @{ int x; int y[]; @};
931 struct bar @{ struct foo z; @};
933 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
934 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
935 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
936 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
939 @node Empty Structures
940 @section Structures With No Members
941 @cindex empty structures
942 @cindex zero-size structures
944 GCC permits a C structure to have no members:
951 The structure will have size zero. In C++, empty structures are part
952 of the language. G++ treats empty structures as if they had a single
953 member of type @code{char}.
955 @node Variable Length
956 @section Arrays of Variable Length
957 @cindex variable-length arrays
958 @cindex arrays of variable length
961 Variable-length automatic arrays are allowed in ISO C99, and as an
962 extension GCC accepts them in C89 mode and in C++. (However, GCC's
963 implementation of variable-length arrays does not yet conform in detail
964 to the ISO C99 standard.) These arrays are
965 declared like any other automatic arrays, but with a length that is not
966 a constant expression. The storage is allocated at the point of
967 declaration and deallocated when the brace-level is exited. For
972 concat_fopen (char *s1, char *s2, char *mode)
974 char str[strlen (s1) + strlen (s2) + 1];
977 return fopen (str, mode);
981 @cindex scope of a variable length array
982 @cindex variable-length array scope
983 @cindex deallocating variable length arrays
984 Jumping or breaking out of the scope of the array name deallocates the
985 storage. Jumping into the scope is not allowed; you get an error
988 @cindex @code{alloca} vs variable-length arrays
989 You can use the function @code{alloca} to get an effect much like
990 variable-length arrays. The function @code{alloca} is available in
991 many other C implementations (but not in all). On the other hand,
992 variable-length arrays are more elegant.
994 There are other differences between these two methods. Space allocated
995 with @code{alloca} exists until the containing @emph{function} returns.
996 The space for a variable-length array is deallocated as soon as the array
997 name's scope ends. (If you use both variable-length arrays and
998 @code{alloca} in the same function, deallocation of a variable-length array
999 will also deallocate anything more recently allocated with @code{alloca}.)
1001 You can also use variable-length arrays as arguments to functions:
1005 tester (int len, char data[len][len])
1011 The length of an array is computed once when the storage is allocated
1012 and is remembered for the scope of the array in case you access it with
1015 If you want to pass the array first and the length afterward, you can
1016 use a forward declaration in the parameter list---another GNU extension.
1020 tester (int len; char data[len][len], int len)
1026 @cindex parameter forward declaration
1027 The @samp{int len} before the semicolon is a @dfn{parameter forward
1028 declaration}, and it serves the purpose of making the name @code{len}
1029 known when the declaration of @code{data} is parsed.
1031 You can write any number of such parameter forward declarations in the
1032 parameter list. They can be separated by commas or semicolons, but the
1033 last one must end with a semicolon, which is followed by the ``real''
1034 parameter declarations. Each forward declaration must match a ``real''
1035 declaration in parameter name and data type. ISO C99 does not support
1036 parameter forward declarations.
1038 @node Variadic Macros
1039 @section Macros with a Variable Number of Arguments.
1040 @cindex variable number of arguments
1041 @cindex macro with variable arguments
1042 @cindex rest argument (in macro)
1043 @cindex variadic macros
1045 In the ISO C standard of 1999, a macro can be declared to accept a
1046 variable number of arguments much as a function can. The syntax for
1047 defining the macro is similar to that of a function. Here is an
1051 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1054 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1055 such a macro, it represents the zero or more tokens until the closing
1056 parenthesis that ends the invocation, including any commas. This set of
1057 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1058 wherever it appears. See the CPP manual for more information.
1060 GCC has long supported variadic macros, and used a different syntax that
1061 allowed you to give a name to the variable arguments just like any other
1062 argument. Here is an example:
1065 #define debug(format, args...) fprintf (stderr, format, args)
1068 This is in all ways equivalent to the ISO C example above, but arguably
1069 more readable and descriptive.
1071 GNU CPP has two further variadic macro extensions, and permits them to
1072 be used with either of the above forms of macro definition.
1074 In standard C, you are not allowed to leave the variable argument out
1075 entirely; but you are allowed to pass an empty argument. For example,
1076 this invocation is invalid in ISO C, because there is no comma after
1083 GNU CPP permits you to completely omit the variable arguments in this
1084 way. In the above examples, the compiler would complain, though since
1085 the expansion of the macro still has the extra comma after the format
1088 To help solve this problem, CPP behaves specially for variable arguments
1089 used with the token paste operator, @samp{##}. If instead you write
1092 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1095 and if the variable arguments are omitted or empty, the @samp{##}
1096 operator causes the preprocessor to remove the comma before it. If you
1097 do provide some variable arguments in your macro invocation, GNU CPP
1098 does not complain about the paste operation and instead places the
1099 variable arguments after the comma. Just like any other pasted macro
1100 argument, these arguments are not macro expanded.
1102 @node Escaped Newlines
1103 @section Slightly Looser Rules for Escaped Newlines
1104 @cindex escaped newlines
1105 @cindex newlines (escaped)
1107 Recently, the preprocessor has relaxed its treatment of escaped
1108 newlines. Previously, the newline had to immediately follow a
1109 backslash. The current implementation allows whitespace in the form
1110 of spaces, horizontal and vertical tabs, and form feeds between the
1111 backslash and the subsequent newline. The preprocessor issues a
1112 warning, but treats it as a valid escaped newline and combines the two
1113 lines to form a single logical line. This works within comments and
1114 tokens, as well as between tokens. Comments are @emph{not} treated as
1115 whitespace for the purposes of this relaxation, since they have not
1116 yet been replaced with spaces.
1119 @section Non-Lvalue Arrays May Have Subscripts
1120 @cindex subscripting
1121 @cindex arrays, non-lvalue
1123 @cindex subscripting and function values
1124 In ISO C99, arrays that are not lvalues still decay to pointers, and
1125 may be subscripted, although they may not be modified or used after
1126 the next sequence point and the unary @samp{&} operator may not be
1127 applied to them. As an extension, GCC allows such arrays to be
1128 subscripted in C89 mode, though otherwise they do not decay to
1129 pointers outside C99 mode. For example,
1130 this is valid in GNU C though not valid in C89:
1134 struct foo @{int a[4];@};
1140 return f().a[index];
1146 @section Arithmetic on @code{void}- and Function-Pointers
1147 @cindex void pointers, arithmetic
1148 @cindex void, size of pointer to
1149 @cindex function pointers, arithmetic
1150 @cindex function, size of pointer to
1152 In GNU C, addition and subtraction operations are supported on pointers to
1153 @code{void} and on pointers to functions. This is done by treating the
1154 size of a @code{void} or of a function as 1.
1156 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1157 and on function types, and returns 1.
1159 @opindex Wpointer-arith
1160 The option @option{-Wpointer-arith} requests a warning if these extensions
1164 @section Non-Constant Initializers
1165 @cindex initializers, non-constant
1166 @cindex non-constant initializers
1168 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1169 automatic variable are not required to be constant expressions in GNU C@.
1170 Here is an example of an initializer with run-time varying elements:
1173 foo (float f, float g)
1175 float beat_freqs[2] = @{ f-g, f+g @};
1180 @node Compound Literals
1181 @section Compound Literals
1182 @cindex constructor expressions
1183 @cindex initializations in expressions
1184 @cindex structures, constructor expression
1185 @cindex expressions, constructor
1186 @cindex compound literals
1187 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1189 ISO C99 supports compound literals. A compound literal looks like
1190 a cast containing an initializer. Its value is an object of the
1191 type specified in the cast, containing the elements specified in
1192 the initializer; it is an lvalue. As an extension, GCC supports
1193 compound literals in C89 mode and in C++.
1195 Usually, the specified type is a structure. Assume that
1196 @code{struct foo} and @code{structure} are declared as shown:
1199 struct foo @{int a; char b[2];@} structure;
1203 Here is an example of constructing a @code{struct foo} with a compound literal:
1206 structure = ((struct foo) @{x + y, 'a', 0@});
1210 This is equivalent to writing the following:
1214 struct foo temp = @{x + y, 'a', 0@};
1219 You can also construct an array. If all the elements of the compound literal
1220 are (made up of) simple constant expressions, suitable for use in
1221 initializers of objects of static storage duration, then the compound
1222 literal can be coerced to a pointer to its first element and used in
1223 such an initializer, as shown here:
1226 char **foo = (char *[]) @{ "x", "y", "z" @};
1229 Compound literals for scalar types and union types are is
1230 also allowed, but then the compound literal is equivalent
1233 As a GNU extension, GCC allows initialization of objects with static storage
1234 duration by compound literals (which is not possible in ISO C99, because
1235 the initializer is not a constant).
1236 It is handled as if the object was initialized only with the bracket
1237 enclosed list if compound literal's and object types match.
1238 The initializer list of the compound literal must be constant.
1239 If the object being initialized has array type of unknown size, the size is
1240 determined by compound literal size.
1243 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1244 static int y[] = (int []) @{1, 2, 3@};
1245 static int z[] = (int [3]) @{1@};
1249 The above lines are equivalent to the following:
1251 static struct foo x = @{1, 'a', 'b'@};
1252 static int y[] = @{1, 2, 3@};
1253 static int z[] = @{1, 0, 0@};
1256 @node Designated Inits
1257 @section Designated Initializers
1258 @cindex initializers with labeled elements
1259 @cindex labeled elements in initializers
1260 @cindex case labels in initializers
1261 @cindex designated initializers
1263 Standard C89 requires the elements of an initializer to appear in a fixed
1264 order, the same as the order of the elements in the array or structure
1267 In ISO C99 you can give the elements in any order, specifying the array
1268 indices or structure field names they apply to, and GNU C allows this as
1269 an extension in C89 mode as well. This extension is not
1270 implemented in GNU C++.
1272 To specify an array index, write
1273 @samp{[@var{index}] =} before the element value. For example,
1276 int a[6] = @{ [4] = 29, [2] = 15 @};
1283 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1287 The index values must be constant expressions, even if the array being
1288 initialized is automatic.
1290 An alternative syntax for this which has been obsolete since GCC 2.5 but
1291 GCC still accepts is to write @samp{[@var{index}]} before the element
1292 value, with no @samp{=}.
1294 To initialize a range of elements to the same value, write
1295 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1296 extension. For example,
1299 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1303 If the value in it has side-effects, the side-effects will happen only once,
1304 not for each initialized field by the range initializer.
1307 Note that the length of the array is the highest value specified
1310 In a structure initializer, specify the name of a field to initialize
1311 with @samp{.@var{fieldname} =} before the element value. For example,
1312 given the following structure,
1315 struct point @{ int x, y; @};
1319 the following initialization
1322 struct point p = @{ .y = yvalue, .x = xvalue @};
1329 struct point p = @{ xvalue, yvalue @};
1332 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1333 @samp{@var{fieldname}:}, as shown here:
1336 struct point p = @{ y: yvalue, x: xvalue @};
1340 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1341 @dfn{designator}. You can also use a designator (or the obsolete colon
1342 syntax) when initializing a union, to specify which element of the union
1343 should be used. For example,
1346 union foo @{ int i; double d; @};
1348 union foo f = @{ .d = 4 @};
1352 will convert 4 to a @code{double} to store it in the union using
1353 the second element. By contrast, casting 4 to type @code{union foo}
1354 would store it into the union as the integer @code{i}, since it is
1355 an integer. (@xref{Cast to Union}.)
1357 You can combine this technique of naming elements with ordinary C
1358 initialization of successive elements. Each initializer element that
1359 does not have a designator applies to the next consecutive element of the
1360 array or structure. For example,
1363 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1370 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1373 Labeling the elements of an array initializer is especially useful
1374 when the indices are characters or belong to an @code{enum} type.
1379 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1380 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1383 @cindex designator lists
1384 You can also write a series of @samp{.@var{fieldname}} and
1385 @samp{[@var{index}]} designators before an @samp{=} to specify a
1386 nested subobject to initialize; the list is taken relative to the
1387 subobject corresponding to the closest surrounding brace pair. For
1388 example, with the @samp{struct point} declaration above:
1391 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1395 If the same field is initialized multiple times, it will have value from
1396 the last initialization. If any such overridden initialization has
1397 side-effect, it is unspecified whether the side-effect happens or not.
1398 Currently, GCC will discard them and issue a warning.
1401 @section Case Ranges
1403 @cindex ranges in case statements
1405 You can specify a range of consecutive values in a single @code{case} label,
1409 case @var{low} ... @var{high}:
1413 This has the same effect as the proper number of individual @code{case}
1414 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1416 This feature is especially useful for ranges of ASCII character codes:
1422 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1423 it may be parsed wrong when you use it with integer values. For example,
1438 @section Cast to a Union Type
1439 @cindex cast to a union
1440 @cindex union, casting to a
1442 A cast to union type is similar to other casts, except that the type
1443 specified is a union type. You can specify the type either with
1444 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1445 a constructor though, not a cast, and hence does not yield an lvalue like
1446 normal casts. (@xref{Compound Literals}.)
1448 The types that may be cast to the union type are those of the members
1449 of the union. Thus, given the following union and variables:
1452 union foo @{ int i; double d; @};
1458 both @code{x} and @code{y} can be cast to type @code{union foo}.
1460 Using the cast as the right-hand side of an assignment to a variable of
1461 union type is equivalent to storing in a member of the union:
1466 u = (union foo) x @equiv{} u.i = x
1467 u = (union foo) y @equiv{} u.d = y
1470 You can also use the union cast as a function argument:
1473 void hack (union foo);
1475 hack ((union foo) x);
1478 @node Mixed Declarations
1479 @section Mixed Declarations and Code
1480 @cindex mixed declarations and code
1481 @cindex declarations, mixed with code
1482 @cindex code, mixed with declarations
1484 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1485 within compound statements. As an extension, GCC also allows this in
1486 C89 mode. For example, you could do:
1495 Each identifier is visible from where it is declared until the end of
1496 the enclosing block.
1498 @node Function Attributes
1499 @section Declaring Attributes of Functions
1500 @cindex function attributes
1501 @cindex declaring attributes of functions
1502 @cindex functions that never return
1503 @cindex functions that return more than once
1504 @cindex functions that have no side effects
1505 @cindex functions in arbitrary sections
1506 @cindex functions that behave like malloc
1507 @cindex @code{volatile} applied to function
1508 @cindex @code{const} applied to function
1509 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1510 @cindex functions with non-null pointer arguments
1511 @cindex functions that are passed arguments in registers on the 386
1512 @cindex functions that pop the argument stack on the 386
1513 @cindex functions that do not pop the argument stack on the 386
1515 In GNU C, you declare certain things about functions called in your program
1516 which help the compiler optimize function calls and check your code more
1519 The keyword @code{__attribute__} allows you to specify special
1520 attributes when making a declaration. This keyword is followed by an
1521 attribute specification inside double parentheses. The following
1522 attributes are currently defined for functions on all targets:
1523 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1524 @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1525 @code{format}, @code{format_arg}, @code{no_instrument_function},
1526 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1527 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1528 @code{alias}, @code{warn_unused_result} and @code{nonnull}. Several other
1529 attributes are defined for functions on particular target systems. Other
1530 attributes, including @code{section} are supported for variables declarations
1531 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1533 You may also specify attributes with @samp{__} preceding and following
1534 each keyword. This allows you to use them in header files without
1535 being concerned about a possible macro of the same name. For example,
1536 you may use @code{__noreturn__} instead of @code{noreturn}.
1538 @xref{Attribute Syntax}, for details of the exact syntax for using
1542 @c Keep this table alphabetized by attribute name. Treat _ as space.
1544 @item alias ("@var{target}")
1545 @cindex @code{alias} attribute
1546 The @code{alias} attribute causes the declaration to be emitted as an
1547 alias for another symbol, which must be specified. For instance,
1550 void __f () @{ /* @r{Do something.} */; @}
1551 void f () __attribute__ ((weak, alias ("__f")));
1554 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
1555 mangled name for the target must be used. It is an error if @samp{__f}
1556 is not defined in the same translation unit.
1558 Not all target machines support this attribute.
1561 @cindex @code{always_inline} function attribute
1562 Generally, functions are not inlined unless optimization is specified.
1563 For functions declared inline, this attribute inlines the function even
1564 if no optimization level was specified.
1567 @cindex functions that do pop the argument stack on the 386
1569 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1570 assume that the calling function will pop off the stack space used to
1571 pass arguments. This is
1572 useful to override the effects of the @option{-mrtd} switch.
1575 @cindex @code{const} function attribute
1576 Many functions do not examine any values except their arguments, and
1577 have no effects except the return value. Basically this is just slightly
1578 more strict class than the @code{pure} attribute below, since function is not
1579 allowed to read global memory.
1581 @cindex pointer arguments
1582 Note that a function that has pointer arguments and examines the data
1583 pointed to must @emph{not} be declared @code{const}. Likewise, a
1584 function that calls a non-@code{const} function usually must not be
1585 @code{const}. It does not make sense for a @code{const} function to
1588 The attribute @code{const} is not implemented in GCC versions earlier
1589 than 2.5. An alternative way to declare that a function has no side
1590 effects, which works in the current version and in some older versions,
1594 typedef int intfn ();
1596 extern const intfn square;
1599 This approach does not work in GNU C++ from 2.6.0 on, since the language
1600 specifies that the @samp{const} must be attached to the return value.
1604 @cindex @code{constructor} function attribute
1605 @cindex @code{destructor} function attribute
1606 The @code{constructor} attribute causes the function to be called
1607 automatically before execution enters @code{main ()}. Similarly, the
1608 @code{destructor} attribute causes the function to be called
1609 automatically after @code{main ()} has completed or @code{exit ()} has
1610 been called. Functions with these attributes are useful for
1611 initializing data that will be used implicitly during the execution of
1614 These attributes are not currently implemented for Objective-C@.
1617 @cindex @code{deprecated} attribute.
1618 The @code{deprecated} attribute results in a warning if the function
1619 is used anywhere in the source file. This is useful when identifying
1620 functions that are expected to be removed in a future version of a
1621 program. The warning also includes the location of the declaration
1622 of the deprecated function, to enable users to easily find further
1623 information about why the function is deprecated, or what they should
1624 do instead. Note that the warnings only occurs for uses:
1627 int old_fn () __attribute__ ((deprecated));
1629 int (*fn_ptr)() = old_fn;
1632 results in a warning on line 3 but not line 2.
1634 The @code{deprecated} attribute can also be used for variables and
1635 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1638 @cindex @code{__declspec(dllexport)}
1639 On Microsoft Windows targets and Symbian OS targets the
1640 @code{dllexport} attribute causes the compiler to provide a global
1641 pointer to a pointer in a DLL, so that it can be referenced with the
1642 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1643 name is formed by combining @code{_imp__} and the function or variable
1646 You can use @code{__declspec(dllexport)} as a synonym for
1647 @code{__attribute__ ((dllexport))} for compatibility with other
1650 On systems that support the @code{visibility} attribute, this
1651 attribute also implies ``default'' visibility, unless a
1652 @code{visibility} attribute is explicitly specified. You should avoid
1653 the use of @code{dllexport} with ``hidden'' or ``internal''
1654 visibility; in the future GCC may issue an error for those cases.
1656 Currently, the @code{dllexport} attribute is ignored for inlined
1657 functions, unless the @option{-fkeep-inline-functions} flag has been
1658 used. The attribute is also ignored for undefined symbols.
1660 When applied to C++ classes, the attribute marks defined non-inlined
1661 member functions and static data members as exports. Static consts
1662 initialized in-class are not marked unless they are also defined
1665 For Microsoft Windows targets there are alternative methods for
1666 including the symbol in the DLL's export table such as using a
1667 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1668 the @option{--export-all} linker flag.
1671 @cindex @code{__declspec(dllimport)}
1672 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1673 attribute causes the compiler to reference a function or variable via
1674 a global pointer to a pointer that is set up by the DLL exporting the
1675 symbol. The attribute implies @code{extern} storage. On Microsoft
1676 Windows targets, the pointer name is formed by combining @code{_imp__}
1677 and the function or variable name.
1679 You can use @code{__declspec(dllimport)} as a synonym for
1680 @code{__attribute__ ((dllimport))} for compatibility with other
1683 Currently, the attribute is ignored for inlined functions. If the
1684 attribute is applied to a symbol @emph{definition}, an error is reported.
1685 If a symbol previously declared @code{dllimport} is later defined, the
1686 attribute is ignored in subsequent references, and a warning is emitted.
1687 The attribute is also overridden by a subsequent declaration as
1690 When applied to C++ classes, the attribute marks non-inlined
1691 member functions and static data members as imports. However, the
1692 attribute is ignored for virtual methods to allow creation of vtables
1695 On the SH Symbian OS target the @code{dllimport} attribute also has
1696 another affect---it can cause the vtable and run-time type information
1697 for a class to be exported. This happens when the class has a
1698 dllimport'ed constructor or a non-inline, non-pure virtual function
1699 and, for either of those two conditions, the class also has a inline
1700 constructor or destructor and has a key function that is defined in
1701 the current translation unit.
1703 For Microsoft Windows based targets the use of the @code{dllimport}
1704 attribute on functions is not necessary, but provides a small
1705 performance benefit by eliminating a thunk in the DLL@. The use of the
1706 @code{dllimport} attribute on imported variables was required on older
1707 versions of the GNU linker, but can now be avoided by passing the
1708 @option{--enable-auto-import} switch to the GNU linker. As with
1709 functions, using the attribute for a variable eliminates a thunk in
1712 One drawback to using this attribute is that a pointer to a function
1713 or variable marked as @code{dllimport} cannot be used as a constant
1714 address. On Microsoft Windows targets, the attribute can be disabled
1715 for functions by setting the @option{-mnop-fun-dllimport} flag.
1718 @cindex eight bit data on the H8/300, H8/300H, and H8S
1719 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1720 variable should be placed into the eight bit data section.
1721 The compiler will generate more efficient code for certain operations
1722 on data in the eight bit data area. Note the eight bit data area is limited to
1725 You must use GAS and GLD from GNU binutils version 2.7 or later for
1726 this attribute to work correctly.
1728 @item exception_handler
1729 @cindex exception handler functions on the Blackfin processor
1730 Use this attribute on the Blackfin to indicate that the specified function
1731 is an exception handler. The compiler will generate function entry and
1732 exit sequences suitable for use in an exception handler when this
1733 attribute is present.
1736 @cindex functions which handle memory bank switching
1737 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1738 use a calling convention that takes care of switching memory banks when
1739 entering and leaving a function. This calling convention is also the
1740 default when using the @option{-mlong-calls} option.
1742 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1743 to call and return from a function.
1745 On 68HC11 the compiler will generate a sequence of instructions
1746 to invoke a board-specific routine to switch the memory bank and call the
1747 real function. The board-specific routine simulates a @code{call}.
1748 At the end of a function, it will jump to a board-specific routine
1749 instead of using @code{rts}. The board-specific return routine simulates
1753 @cindex functions that pop the argument stack on the 386
1754 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1755 pass the first argument (if of integral type) in the register ECX and
1756 the second argument (if of integral type) in the register EDX@. Subsequent
1757 and other typed arguments are passed on the stack. The called function will
1758 pop the arguments off the stack. If the number of arguments is variable all
1759 arguments are pushed on the stack.
1761 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1762 @cindex @code{format} function attribute
1764 The @code{format} attribute specifies that a function takes @code{printf},
1765 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1766 should be type-checked against a format string. For example, the
1771 my_printf (void *my_object, const char *my_format, ...)
1772 __attribute__ ((format (printf, 2, 3)));
1776 causes the compiler to check the arguments in calls to @code{my_printf}
1777 for consistency with the @code{printf} style format string argument
1780 The parameter @var{archetype} determines how the format string is
1781 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1782 or @code{strfmon}. (You can also use @code{__printf__},
1783 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1784 parameter @var{string-index} specifies which argument is the format
1785 string argument (starting from 1), while @var{first-to-check} is the
1786 number of the first argument to check against the format string. For
1787 functions where the arguments are not available to be checked (such as
1788 @code{vprintf}), specify the third parameter as zero. In this case the
1789 compiler only checks the format string for consistency. For
1790 @code{strftime} formats, the third parameter is required to be zero.
1791 Since non-static C++ methods have an implicit @code{this} argument, the
1792 arguments of such methods should be counted from two, not one, when
1793 giving values for @var{string-index} and @var{first-to-check}.
1795 In the example above, the format string (@code{my_format}) is the second
1796 argument of the function @code{my_print}, and the arguments to check
1797 start with the third argument, so the correct parameters for the format
1798 attribute are 2 and 3.
1800 @opindex ffreestanding
1801 @opindex fno-builtin
1802 The @code{format} attribute allows you to identify your own functions
1803 which take format strings as arguments, so that GCC can check the
1804 calls to these functions for errors. The compiler always (unless
1805 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1806 for the standard library functions @code{printf}, @code{fprintf},
1807 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1808 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1809 warnings are requested (using @option{-Wformat}), so there is no need to
1810 modify the header file @file{stdio.h}. In C99 mode, the functions
1811 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1812 @code{vsscanf} are also checked. Except in strictly conforming C
1813 standard modes, the X/Open function @code{strfmon} is also checked as
1814 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1815 @xref{C Dialect Options,,Options Controlling C Dialect}.
1817 The target may provide additional types of format checks.
1818 @xref{Target Format Checks,,Format Checks Specific to Particular
1821 @item format_arg (@var{string-index})
1822 @cindex @code{format_arg} function attribute
1823 @opindex Wformat-nonliteral
1824 The @code{format_arg} attribute specifies that a function takes a format
1825 string for a @code{printf}, @code{scanf}, @code{strftime} or
1826 @code{strfmon} style function and modifies it (for example, to translate
1827 it into another language), so the result can be passed to a
1828 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1829 function (with the remaining arguments to the format function the same
1830 as they would have been for the unmodified string). For example, the
1835 my_dgettext (char *my_domain, const char *my_format)
1836 __attribute__ ((format_arg (2)));
1840 causes the compiler to check the arguments in calls to a @code{printf},
1841 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1842 format string argument is a call to the @code{my_dgettext} function, for
1843 consistency with the format string argument @code{my_format}. If the
1844 @code{format_arg} attribute had not been specified, all the compiler
1845 could tell in such calls to format functions would be that the format
1846 string argument is not constant; this would generate a warning when
1847 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1848 without the attribute.
1850 The parameter @var{string-index} specifies which argument is the format
1851 string argument (starting from one). Since non-static C++ methods have
1852 an implicit @code{this} argument, the arguments of such methods should
1853 be counted from two.
1855 The @code{format-arg} attribute allows you to identify your own
1856 functions which modify format strings, so that GCC can check the
1857 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1858 type function whose operands are a call to one of your own function.
1859 The compiler always treats @code{gettext}, @code{dgettext}, and
1860 @code{dcgettext} in this manner except when strict ISO C support is
1861 requested by @option{-ansi} or an appropriate @option{-std} option, or
1862 @option{-ffreestanding} or @option{-fno-builtin}
1863 is used. @xref{C Dialect Options,,Options
1864 Controlling C Dialect}.
1866 @item function_vector
1867 @cindex calling functions through the function vector on the H8/300 processors
1868 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1869 function should be called through the function vector. Calling a
1870 function through the function vector will reduce code size, however;
1871 the function vector has a limited size (maximum 128 entries on the H8/300
1872 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1874 You must use GAS and GLD from GNU binutils version 2.7 or later for
1875 this attribute to work correctly.
1878 @cindex interrupt handler functions
1879 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
1880 that the specified function is an interrupt handler. The compiler will
1881 generate function entry and exit sequences suitable for use in an
1882 interrupt handler when this attribute is present.
1884 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1885 SH processors can be specified via the @code{interrupt_handler} attribute.
1887 Note, on the AVR, interrupts will be enabled inside the function.
1889 Note, for the ARM, you can specify the kind of interrupt to be handled by
1890 adding an optional parameter to the interrupt attribute like this:
1893 void f () __attribute__ ((interrupt ("IRQ")));
1896 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
1898 @item interrupt_handler
1899 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
1900 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
1901 indicate that the specified function is an interrupt handler. The compiler
1902 will generate function entry and exit sequences suitable for use in an
1903 interrupt handler when this attribute is present.
1906 @cindex User stack pointer in interrupts on the Blackfin
1907 When used together with @code{interrupt_handler}, @code{exception_handler}
1908 or @code{nmi_handler}, code will be generated to load the stack pointer
1909 from the USP register in the function prologue.
1911 @item long_call/short_call
1912 @cindex indirect calls on ARM
1913 This attribute specifies how a particular function is called on
1914 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
1915 command line switch and @code{#pragma long_calls} settings. The
1916 @code{long_call} attribute causes the compiler to always call the
1917 function by first loading its address into a register and then using the
1918 contents of that register. The @code{short_call} attribute always places
1919 the offset to the function from the call site into the @samp{BL}
1920 instruction directly.
1922 @item longcall/shortcall
1923 @cindex functions called via pointer on the RS/6000 and PowerPC
1924 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
1925 compiler to always call this function via a pointer, just as it would if
1926 the @option{-mlongcall} option had been specified. The @code{shortcall}
1927 attribute causes the compiler not to do this. These attributes override
1928 both the @option{-mlongcall} switch and the @code{#pragma longcall}
1931 @xref{RS/6000 and PowerPC Options}, for more information on whether long
1932 calls are necessary.
1935 @cindex @code{malloc} attribute
1936 The @code{malloc} attribute is used to tell the compiler that a function
1937 may be treated as if any non-@code{NULL} pointer it returns cannot
1938 alias any other pointer valid when the function returns.
1939 This will often improve optimization.
1940 Standard functions with this property include @code{malloc} and
1941 @code{calloc}. @code{realloc}-like functions have this property as
1942 long as the old pointer is never referred to (including comparing it
1943 to the new pointer) after the function returns a non-@code{NULL}
1946 @item model (@var{model-name})
1947 @cindex function addressability on the M32R/D
1948 @cindex variable addressability on the IA-64
1950 On the M32R/D, use this attribute to set the addressability of an
1951 object, and of the code generated for a function. The identifier
1952 @var{model-name} is one of @code{small}, @code{medium}, or
1953 @code{large}, representing each of the code models.
1955 Small model objects live in the lower 16MB of memory (so that their
1956 addresses can be loaded with the @code{ld24} instruction), and are
1957 callable with the @code{bl} instruction.
1959 Medium model objects may live anywhere in the 32-bit address space (the
1960 compiler will generate @code{seth/add3} instructions to load their addresses),
1961 and are callable with the @code{bl} instruction.
1963 Large model objects may live anywhere in the 32-bit address space (the
1964 compiler will generate @code{seth/add3} instructions to load their addresses),
1965 and may not be reachable with the @code{bl} instruction (the compiler will
1966 generate the much slower @code{seth/add3/jl} instruction sequence).
1968 On IA-64, use this attribute to set the addressability of an object.
1969 At present, the only supported identifier for @var{model-name} is
1970 @code{small}, indicating addressability via ``small'' (22-bit)
1971 addresses (so that their addresses can be loaded with the @code{addl}
1972 instruction). Caveat: such addressing is by definition not position
1973 independent and hence this attribute must not be used for objects
1974 defined by shared libraries.
1977 @cindex function without a prologue/epilogue code
1978 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
1979 specified function does not need prologue/epilogue sequences generated by
1980 the compiler. It is up to the programmer to provide these sequences.
1983 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
1984 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
1985 use the normal calling convention based on @code{jsr} and @code{rts}.
1986 This attribute can be used to cancel the effect of the @option{-mlong-calls}
1990 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
1991 Use this attribute together with @code{interrupt_handler},
1992 @code{exception_handler} or @code{nmi_handler} to indicate that the function
1993 entry code should enable nested interrupts or exceptions.
1996 @cindex NMI handler functions on the Blackfin processor
1997 Use this attribute on the Blackfin to indicate that the specified function
1998 is an NMI handler. The compiler will generate function entry and
1999 exit sequences suitable for use in an NMI handler when this
2000 attribute is present.
2002 @item no_instrument_function
2003 @cindex @code{no_instrument_function} function attribute
2004 @opindex finstrument-functions
2005 If @option{-finstrument-functions} is given, profiling function calls will
2006 be generated at entry and exit of most user-compiled functions.
2007 Functions with this attribute will not be so instrumented.
2010 @cindex @code{noinline} function attribute
2011 This function attribute prevents a function from being considered for
2014 @item nonnull (@var{arg-index}, @dots{})
2015 @cindex @code{nonnull} function attribute
2016 The @code{nonnull} attribute specifies that some function parameters should
2017 be non-null pointers. For instance, the declaration:
2021 my_memcpy (void *dest, const void *src, size_t len)
2022 __attribute__((nonnull (1, 2)));
2026 causes the compiler to check that, in calls to @code{my_memcpy},
2027 arguments @var{dest} and @var{src} are non-null. If the compiler
2028 determines that a null pointer is passed in an argument slot marked
2029 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2030 is issued. The compiler may also choose to make optimizations based
2031 on the knowledge that certain function arguments will not be null.
2033 If no argument index list is given to the @code{nonnull} attribute,
2034 all pointer arguments are marked as non-null. To illustrate, the
2035 following declaration is equivalent to the previous example:
2039 my_memcpy (void *dest, const void *src, size_t len)
2040 __attribute__((nonnull));
2044 @cindex @code{noreturn} function attribute
2045 A few standard library functions, such as @code{abort} and @code{exit},
2046 cannot return. GCC knows this automatically. Some programs define
2047 their own functions that never return. You can declare them
2048 @code{noreturn} to tell the compiler this fact. For example,
2052 void fatal () __attribute__ ((noreturn));
2055 fatal (/* @r{@dots{}} */)
2057 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2063 The @code{noreturn} keyword tells the compiler to assume that
2064 @code{fatal} cannot return. It can then optimize without regard to what
2065 would happen if @code{fatal} ever did return. This makes slightly
2066 better code. More importantly, it helps avoid spurious warnings of
2067 uninitialized variables.
2069 The @code{noreturn} keyword does not affect the exceptional path when that
2070 applies: a @code{noreturn}-marked function may still return to the caller
2071 by throwing an exception or calling @code{longjmp}.
2073 Do not assume that registers saved by the calling function are
2074 restored before calling the @code{noreturn} function.
2076 It does not make sense for a @code{noreturn} function to have a return
2077 type other than @code{void}.
2079 The attribute @code{noreturn} is not implemented in GCC versions
2080 earlier than 2.5. An alternative way to declare that a function does
2081 not return, which works in the current version and in some older
2082 versions, is as follows:
2085 typedef void voidfn ();
2087 volatile voidfn fatal;
2090 This approach does not work in GNU C++.
2093 @cindex @code{nothrow} function attribute
2094 The @code{nothrow} attribute is used to inform the compiler that a
2095 function cannot throw an exception. For example, most functions in
2096 the standard C library can be guaranteed not to throw an exception
2097 with the notable exceptions of @code{qsort} and @code{bsearch} that
2098 take function pointer arguments. The @code{nothrow} attribute is not
2099 implemented in GCC versions earlier than 3.3.
2102 @cindex @code{pure} function attribute
2103 Many functions have no effects except the return value and their
2104 return value depends only on the parameters and/or global variables.
2105 Such a function can be subject
2106 to common subexpression elimination and loop optimization just as an
2107 arithmetic operator would be. These functions should be declared
2108 with the attribute @code{pure}. For example,
2111 int square (int) __attribute__ ((pure));
2115 says that the hypothetical function @code{square} is safe to call
2116 fewer times than the program says.
2118 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2119 Interesting non-pure functions are functions with infinite loops or those
2120 depending on volatile memory or other system resource, that may change between
2121 two consecutive calls (such as @code{feof} in a multithreading environment).
2123 The attribute @code{pure} is not implemented in GCC versions earlier
2126 @item regparm (@var{number})
2127 @cindex @code{regparm} attribute
2128 @cindex functions that are passed arguments in registers on the 386
2129 On the Intel 386, the @code{regparm} attribute causes the compiler to
2130 pass arguments number one to @var{number} if they are of integral type
2131 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2132 take a variable number of arguments will continue to be passed all of their
2133 arguments on the stack.
2135 Beware that on some ELF systems this attribute is unsuitable for
2136 global functions in shared libraries with lazy binding (which is the
2137 default). Lazy binding will send the first call via resolving code in
2138 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2139 per the standard calling conventions. Solaris 8 is affected by this.
2140 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2141 safe since the loaders there save all registers. (Lazy binding can be
2142 disabled with the linker or the loader if desired, to avoid the
2146 @cindex @code{sseregparm} attribute
2147 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2148 causes the compiler to pass up to 8 floating point arguments in
2149 SSE registers instead of on the stack. Functions that take a
2150 variable number of arguments will continue to pass all of their
2151 floating point arguments on the stack.
2154 @cindex @code{returns_twice} attribute
2155 The @code{returns_twice} attribute tells the compiler that a function may
2156 return more than one time. The compiler will ensure that all registers
2157 are dead before calling such a function and will emit a warning about
2158 the variables that may be clobbered after the second return from the
2159 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2160 The @code{longjmp}-like counterpart of such function, if any, might need
2161 to be marked with the @code{noreturn} attribute.
2164 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2165 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2166 all registers except the stack pointer should be saved in the prologue
2167 regardless of whether they are used or not.
2169 @item section ("@var{section-name}")
2170 @cindex @code{section} function attribute
2171 Normally, the compiler places the code it generates in the @code{text} section.
2172 Sometimes, however, you need additional sections, or you need certain
2173 particular functions to appear in special sections. The @code{section}
2174 attribute specifies that a function lives in a particular section.
2175 For example, the declaration:
2178 extern void foobar (void) __attribute__ ((section ("bar")));
2182 puts the function @code{foobar} in the @code{bar} section.
2184 Some file formats do not support arbitrary sections so the @code{section}
2185 attribute is not available on all platforms.
2186 If you need to map the entire contents of a module to a particular
2187 section, consider using the facilities of the linker instead.
2190 @cindex @code{sentinel} function attribute
2191 This function attribute ensures that a parameter in a function call is
2192 an explicit @code{NULL}. The attribute is only valid on variadic
2193 functions. By default, the sentinel is located at position zero, the
2194 last parameter of the function call. If an optional integer position
2195 argument P is supplied to the attribute, the sentinel must be located at
2196 position P counting backwards from the end of the argument list.
2199 __attribute__ ((sentinel))
2201 __attribute__ ((sentinel(0)))
2204 The attribute is automatically set with a position of 0 for the built-in
2205 functions @code{execl} and @code{execlp}. The built-in function
2206 @code{execle} has the attribute set with a position of 1.
2208 A valid @code{NULL} in this context is defined as zero with any pointer
2209 type. If your system defines the @code{NULL} macro with an integer type
2210 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2211 with a copy that redefines NULL appropriately.
2213 The warnings for missing or incorrect sentinels are enabled with
2217 See long_call/short_call.
2220 See longcall/shortcall.
2223 @cindex signal handler functions on the AVR processors
2224 Use this attribute on the AVR to indicate that the specified
2225 function is a signal handler. The compiler will generate function
2226 entry and exit sequences suitable for use in a signal handler when this
2227 attribute is present. Interrupts will be disabled inside the function.
2230 Use this attribute on the SH to indicate an @code{interrupt_handler}
2231 function should switch to an alternate stack. It expects a string
2232 argument that names a global variable holding the address of the
2237 void f () __attribute__ ((interrupt_handler,
2238 sp_switch ("alt_stack")));
2242 @cindex functions that pop the argument stack on the 386
2243 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2244 assume that the called function will pop off the stack space used to
2245 pass arguments, unless it takes a variable number of arguments.
2248 @cindex tiny data section on the H8/300H and H8S
2249 Use this attribute on the H8/300H and H8S to indicate that the specified
2250 variable should be placed into the tiny data section.
2251 The compiler will generate more efficient code for loads and stores
2252 on data in the tiny data section. Note the tiny data area is limited to
2253 slightly under 32kbytes of data.
2256 Use this attribute on the SH for an @code{interrupt_handler} to return using
2257 @code{trapa} instead of @code{rte}. This attribute expects an integer
2258 argument specifying the trap number to be used.
2261 @cindex @code{unused} attribute.
2262 This attribute, attached to a function, means that the function is meant
2263 to be possibly unused. GCC will not produce a warning for this
2267 @cindex @code{used} attribute.
2268 This attribute, attached to a function, means that code must be emitted
2269 for the function even if it appears that the function is not referenced.
2270 This is useful, for example, when the function is referenced only in
2273 @item visibility ("@var{visibility_type}")
2274 @cindex @code{visibility} attribute
2275 The @code{visibility} attribute on ELF targets causes the declaration
2276 to be emitted with default, hidden, protected or internal visibility.
2279 void __attribute__ ((visibility ("protected")))
2280 f () @{ /* @r{Do something.} */; @}
2281 int i __attribute__ ((visibility ("hidden")));
2284 See the ELF gABI for complete details, but the short story is:
2287 @c keep this list of visibilities in alphabetical order.
2290 Default visibility is the normal case for ELF@. This value is
2291 available for the visibility attribute to override other options
2292 that may change the assumed visibility of symbols.
2295 Hidden visibility indicates that the symbol will not be placed into
2296 the dynamic symbol table, so no other @dfn{module} (executable or
2297 shared library) can reference it directly.
2300 Internal visibility is like hidden visibility, but with additional
2301 processor specific semantics. Unless otherwise specified by the psABI,
2302 GCC defines internal visibility to mean that the function is @emph{never}
2303 called from another module. Note that hidden symbols, while they cannot
2304 be referenced directly by other modules, can be referenced indirectly via
2305 function pointers. By indicating that a symbol cannot be called from
2306 outside the module, GCC may for instance omit the load of a PIC register
2307 since it is known that the calling function loaded the correct value.
2310 Protected visibility indicates that the symbol will be placed in the
2311 dynamic symbol table, but that references within the defining module
2312 will bind to the local symbol. That is, the symbol cannot be overridden
2317 Not all ELF targets support this attribute.
2319 @item warn_unused_result
2320 @cindex @code{warn_unused_result} attribute
2321 The @code{warn_unused_result} attribute causes a warning to be emitted
2322 if a caller of the function with this attribute does not use its
2323 return value. This is useful for functions where not checking
2324 the result is either a security problem or always a bug, such as
2328 int fn () __attribute__ ((warn_unused_result));
2331 if (fn () < 0) return -1;
2337 results in warning on line 5.
2340 @cindex @code{weak} attribute
2341 The @code{weak} attribute causes the declaration to be emitted as a weak
2342 symbol rather than a global. This is primarily useful in defining
2343 library functions which can be overridden in user code, though it can
2344 also be used with non-function declarations. Weak symbols are supported
2345 for ELF targets, and also for a.out targets when using the GNU assembler
2350 You can specify multiple attributes in a declaration by separating them
2351 by commas within the double parentheses or by immediately following an
2352 attribute declaration with another attribute declaration.
2354 @cindex @code{#pragma}, reason for not using
2355 @cindex pragma, reason for not using
2356 Some people object to the @code{__attribute__} feature, suggesting that
2357 ISO C's @code{#pragma} should be used instead. At the time
2358 @code{__attribute__} was designed, there were two reasons for not doing
2363 It is impossible to generate @code{#pragma} commands from a macro.
2366 There is no telling what the same @code{#pragma} might mean in another
2370 These two reasons applied to almost any application that might have been
2371 proposed for @code{#pragma}. It was basically a mistake to use
2372 @code{#pragma} for @emph{anything}.
2374 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2375 to be generated from macros. In addition, a @code{#pragma GCC}
2376 namespace is now in use for GCC-specific pragmas. However, it has been
2377 found convenient to use @code{__attribute__} to achieve a natural
2378 attachment of attributes to their corresponding declarations, whereas
2379 @code{#pragma GCC} is of use for constructs that do not naturally form
2380 part of the grammar. @xref{Other Directives,,Miscellaneous
2381 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2383 @node Attribute Syntax
2384 @section Attribute Syntax
2385 @cindex attribute syntax
2387 This section describes the syntax with which @code{__attribute__} may be
2388 used, and the constructs to which attribute specifiers bind, for the C
2389 language. Some details may vary for C++ and Objective-C@. Because of
2390 infelicities in the grammar for attributes, some forms described here
2391 may not be successfully parsed in all cases.
2393 There are some problems with the semantics of attributes in C++. For
2394 example, there are no manglings for attributes, although they may affect
2395 code generation, so problems may arise when attributed types are used in
2396 conjunction with templates or overloading. Similarly, @code{typeid}
2397 does not distinguish between types with different attributes. Support
2398 for attributes in C++ may be restricted in future to attributes on
2399 declarations only, but not on nested declarators.
2401 @xref{Function Attributes}, for details of the semantics of attributes
2402 applying to functions. @xref{Variable Attributes}, for details of the
2403 semantics of attributes applying to variables. @xref{Type Attributes},
2404 for details of the semantics of attributes applying to structure, union
2405 and enumerated types.
2407 An @dfn{attribute specifier} is of the form
2408 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2409 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2410 each attribute is one of the following:
2414 Empty. Empty attributes are ignored.
2417 A word (which may be an identifier such as @code{unused}, or a reserved
2418 word such as @code{const}).
2421 A word, followed by, in parentheses, parameters for the attribute.
2422 These parameters take one of the following forms:
2426 An identifier. For example, @code{mode} attributes use this form.
2429 An identifier followed by a comma and a non-empty comma-separated list
2430 of expressions. For example, @code{format} attributes use this form.
2433 A possibly empty comma-separated list of expressions. For example,
2434 @code{format_arg} attributes use this form with the list being a single
2435 integer constant expression, and @code{alias} attributes use this form
2436 with the list being a single string constant.
2440 An @dfn{attribute specifier list} is a sequence of one or more attribute
2441 specifiers, not separated by any other tokens.
2443 In GNU C, an attribute specifier list may appear after the colon following a
2444 label, other than a @code{case} or @code{default} label. The only
2445 attribute it makes sense to use after a label is @code{unused}. This
2446 feature is intended for code generated by programs which contains labels
2447 that may be unused but which is compiled with @option{-Wall}. It would
2448 not normally be appropriate to use in it human-written code, though it
2449 could be useful in cases where the code that jumps to the label is
2450 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2451 such placement of attribute lists, as it is permissible for a
2452 declaration, which could begin with an attribute list, to be labelled in
2453 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2454 does not arise there.
2456 An attribute specifier list may appear as part of a @code{struct},
2457 @code{union} or @code{enum} specifier. It may go either immediately
2458 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2459 the closing brace. It is ignored if the content of the structure, union
2460 or enumerated type is not defined in the specifier in which the
2461 attribute specifier list is used---that is, in usages such as
2462 @code{struct __attribute__((foo)) bar} with no following opening brace.
2463 Where attribute specifiers follow the closing brace, they are considered
2464 to relate to the structure, union or enumerated type defined, not to any
2465 enclosing declaration the type specifier appears in, and the type
2466 defined is not complete until after the attribute specifiers.
2467 @c Otherwise, there would be the following problems: a shift/reduce
2468 @c conflict between attributes binding the struct/union/enum and
2469 @c binding to the list of specifiers/qualifiers; and "aligned"
2470 @c attributes could use sizeof for the structure, but the size could be
2471 @c changed later by "packed" attributes.
2473 Otherwise, an attribute specifier appears as part of a declaration,
2474 counting declarations of unnamed parameters and type names, and relates
2475 to that declaration (which may be nested in another declaration, for
2476 example in the case of a parameter declaration), or to a particular declarator
2477 within a declaration. Where an
2478 attribute specifier is applied to a parameter declared as a function or
2479 an array, it should apply to the function or array rather than the
2480 pointer to which the parameter is implicitly converted, but this is not
2481 yet correctly implemented.
2483 Any list of specifiers and qualifiers at the start of a declaration may
2484 contain attribute specifiers, whether or not such a list may in that
2485 context contain storage class specifiers. (Some attributes, however,
2486 are essentially in the nature of storage class specifiers, and only make
2487 sense where storage class specifiers may be used; for example,
2488 @code{section}.) There is one necessary limitation to this syntax: the
2489 first old-style parameter declaration in a function definition cannot
2490 begin with an attribute specifier, because such an attribute applies to
2491 the function instead by syntax described below (which, however, is not
2492 yet implemented in this case). In some other cases, attribute
2493 specifiers are permitted by this grammar but not yet supported by the
2494 compiler. All attribute specifiers in this place relate to the
2495 declaration as a whole. In the obsolescent usage where a type of
2496 @code{int} is implied by the absence of type specifiers, such a list of
2497 specifiers and qualifiers may be an attribute specifier list with no
2498 other specifiers or qualifiers.
2500 At present, the first parameter in a function prototype must have some
2501 type specifier which is not an attribute specifier; this resolves an
2502 ambiguity in the interpretation of @code{void f(int
2503 (__attribute__((foo)) x))}, but is subject to change. At present, if
2504 the parentheses of a function declarator contain only attributes then
2505 those attributes are ignored, rather than yielding an error or warning
2506 or implying a single parameter of type int, but this is subject to
2509 An attribute specifier list may appear immediately before a declarator
2510 (other than the first) in a comma-separated list of declarators in a
2511 declaration of more than one identifier using a single list of
2512 specifiers and qualifiers. Such attribute specifiers apply
2513 only to the identifier before whose declarator they appear. For
2517 __attribute__((noreturn)) void d0 (void),
2518 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2523 the @code{noreturn} attribute applies to all the functions
2524 declared; the @code{format} attribute only applies to @code{d1}.
2526 An attribute specifier list may appear immediately before the comma,
2527 @code{=} or semicolon terminating the declaration of an identifier other
2528 than a function definition. At present, such attribute specifiers apply
2529 to the declared object or function, but in future they may attach to the
2530 outermost adjacent declarator. In simple cases there is no difference,
2531 but, for example, in
2534 void (****f)(void) __attribute__((noreturn));
2538 at present the @code{noreturn} attribute applies to @code{f}, which
2539 causes a warning since @code{f} is not a function, but in future it may
2540 apply to the function @code{****f}. The precise semantics of what
2541 attributes in such cases will apply to are not yet specified. Where an
2542 assembler name for an object or function is specified (@pxref{Asm
2543 Labels}), at present the attribute must follow the @code{asm}
2544 specification; in future, attributes before the @code{asm} specification
2545 may apply to the adjacent declarator, and those after it to the declared
2548 An attribute specifier list may, in future, be permitted to appear after
2549 the declarator in a function definition (before any old-style parameter
2550 declarations or the function body).
2552 Attribute specifiers may be mixed with type qualifiers appearing inside
2553 the @code{[]} of a parameter array declarator, in the C99 construct by
2554 which such qualifiers are applied to the pointer to which the array is
2555 implicitly converted. Such attribute specifiers apply to the pointer,
2556 not to the array, but at present this is not implemented and they are
2559 An attribute specifier list may appear at the start of a nested
2560 declarator. At present, there are some limitations in this usage: the
2561 attributes correctly apply to the declarator, but for most individual
2562 attributes the semantics this implies are not implemented.
2563 When attribute specifiers follow the @code{*} of a pointer
2564 declarator, they may be mixed with any type qualifiers present.
2565 The following describes the formal semantics of this syntax. It will make the
2566 most sense if you are familiar with the formal specification of
2567 declarators in the ISO C standard.
2569 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2570 D1}, where @code{T} contains declaration specifiers that specify a type
2571 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2572 contains an identifier @var{ident}. The type specified for @var{ident}
2573 for derived declarators whose type does not include an attribute
2574 specifier is as in the ISO C standard.
2576 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2577 and the declaration @code{T D} specifies the type
2578 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2579 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2580 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2582 If @code{D1} has the form @code{*
2583 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2584 declaration @code{T D} specifies the type
2585 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2586 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2587 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2593 void (__attribute__((noreturn)) ****f) (void);
2597 specifies the type ``pointer to pointer to pointer to pointer to
2598 non-returning function returning @code{void}''. As another example,
2601 char *__attribute__((aligned(8))) *f;
2605 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2606 Note again that this does not work with most attributes; for example,
2607 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2608 is not yet supported.
2610 For compatibility with existing code written for compiler versions that
2611 did not implement attributes on nested declarators, some laxity is
2612 allowed in the placing of attributes. If an attribute that only applies
2613 to types is applied to a declaration, it will be treated as applying to
2614 the type of that declaration. If an attribute that only applies to
2615 declarations is applied to the type of a declaration, it will be treated
2616 as applying to that declaration; and, for compatibility with code
2617 placing the attributes immediately before the identifier declared, such
2618 an attribute applied to a function return type will be treated as
2619 applying to the function type, and such an attribute applied to an array
2620 element type will be treated as applying to the array type. If an
2621 attribute that only applies to function types is applied to a
2622 pointer-to-function type, it will be treated as applying to the pointer
2623 target type; if such an attribute is applied to a function return type
2624 that is not a pointer-to-function type, it will be treated as applying
2625 to the function type.
2627 @node Function Prototypes
2628 @section Prototypes and Old-Style Function Definitions
2629 @cindex function prototype declarations
2630 @cindex old-style function definitions
2631 @cindex promotion of formal parameters
2633 GNU C extends ISO C to allow a function prototype to override a later
2634 old-style non-prototype definition. Consider the following example:
2637 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2644 /* @r{Prototype function declaration.} */
2645 int isroot P((uid_t));
2647 /* @r{Old-style function definition.} */
2649 isroot (x) /* @r{??? lossage here ???} */
2656 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2657 not allow this example, because subword arguments in old-style
2658 non-prototype definitions are promoted. Therefore in this example the
2659 function definition's argument is really an @code{int}, which does not
2660 match the prototype argument type of @code{short}.
2662 This restriction of ISO C makes it hard to write code that is portable
2663 to traditional C compilers, because the programmer does not know
2664 whether the @code{uid_t} type is @code{short}, @code{int}, or
2665 @code{long}. Therefore, in cases like these GNU C allows a prototype
2666 to override a later old-style definition. More precisely, in GNU C, a
2667 function prototype argument type overrides the argument type specified
2668 by a later old-style definition if the former type is the same as the
2669 latter type before promotion. Thus in GNU C the above example is
2670 equivalent to the following:
2683 GNU C++ does not support old-style function definitions, so this
2684 extension is irrelevant.
2687 @section C++ Style Comments
2689 @cindex C++ comments
2690 @cindex comments, C++ style
2692 In GNU C, you may use C++ style comments, which start with @samp{//} and
2693 continue until the end of the line. Many other C implementations allow
2694 such comments, and they are included in the 1999 C standard. However,
2695 C++ style comments are not recognized if you specify an @option{-std}
2696 option specifying a version of ISO C before C99, or @option{-ansi}
2697 (equivalent to @option{-std=c89}).
2700 @section Dollar Signs in Identifier Names
2702 @cindex dollar signs in identifier names
2703 @cindex identifier names, dollar signs in
2705 In GNU C, you may normally use dollar signs in identifier names.
2706 This is because many traditional C implementations allow such identifiers.
2707 However, dollar signs in identifiers are not supported on a few target
2708 machines, typically because the target assembler does not allow them.
2710 @node Character Escapes
2711 @section The Character @key{ESC} in Constants
2713 You can use the sequence @samp{\e} in a string or character constant to
2714 stand for the ASCII character @key{ESC}.
2717 @section Inquiring on Alignment of Types or Variables
2719 @cindex type alignment
2720 @cindex variable alignment
2722 The keyword @code{__alignof__} allows you to inquire about how an object
2723 is aligned, or the minimum alignment usually required by a type. Its
2724 syntax is just like @code{sizeof}.
2726 For example, if the target machine requires a @code{double} value to be
2727 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2728 This is true on many RISC machines. On more traditional machine
2729 designs, @code{__alignof__ (double)} is 4 or even 2.
2731 Some machines never actually require alignment; they allow reference to any
2732 data type even at an odd address. For these machines, @code{__alignof__}
2733 reports the @emph{recommended} alignment of a type.
2735 If the operand of @code{__alignof__} is an lvalue rather than a type,
2736 its value is the required alignment for its type, taking into account
2737 any minimum alignment specified with GCC's @code{__attribute__}
2738 extension (@pxref{Variable Attributes}). For example, after this
2742 struct foo @{ int x; char y; @} foo1;
2746 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2747 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2749 It is an error to ask for the alignment of an incomplete type.
2751 @node Variable Attributes
2752 @section Specifying Attributes of Variables
2753 @cindex attribute of variables
2754 @cindex variable attributes
2756 The keyword @code{__attribute__} allows you to specify special
2757 attributes of variables or structure fields. This keyword is followed
2758 by an attribute specification inside double parentheses. Some
2759 attributes are currently defined generically for variables.
2760 Other attributes are defined for variables on particular target
2761 systems. Other attributes are available for functions
2762 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2763 Other front ends might define more attributes
2764 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2766 You may also specify attributes with @samp{__} preceding and following
2767 each keyword. This allows you to use them in header files without
2768 being concerned about a possible macro of the same name. For example,
2769 you may use @code{__aligned__} instead of @code{aligned}.
2771 @xref{Attribute Syntax}, for details of the exact syntax for using
2775 @cindex @code{aligned} attribute
2776 @item aligned (@var{alignment})
2777 This attribute specifies a minimum alignment for the variable or
2778 structure field, measured in bytes. For example, the declaration:
2781 int x __attribute__ ((aligned (16))) = 0;
2785 causes the compiler to allocate the global variable @code{x} on a
2786 16-byte boundary. On a 68040, this could be used in conjunction with
2787 an @code{asm} expression to access the @code{move16} instruction which
2788 requires 16-byte aligned operands.
2790 You can also specify the alignment of structure fields. For example, to
2791 create a double-word aligned @code{int} pair, you could write:
2794 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2798 This is an alternative to creating a union with a @code{double} member
2799 that forces the union to be double-word aligned.
2801 As in the preceding examples, you can explicitly specify the alignment
2802 (in bytes) that you wish the compiler to use for a given variable or
2803 structure field. Alternatively, you can leave out the alignment factor
2804 and just ask the compiler to align a variable or field to the maximum
2805 useful alignment for the target machine you are compiling for. For
2806 example, you could write:
2809 short array[3] __attribute__ ((aligned));
2812 Whenever you leave out the alignment factor in an @code{aligned} attribute
2813 specification, the compiler automatically sets the alignment for the declared
2814 variable or field to the largest alignment which is ever used for any data
2815 type on the target machine you are compiling for. Doing this can often make
2816 copy operations more efficient, because the compiler can use whatever
2817 instructions copy the biggest chunks of memory when performing copies to
2818 or from the variables or fields that you have aligned this way.
2820 The @code{aligned} attribute can only increase the alignment; but you
2821 can decrease it by specifying @code{packed} as well. See below.
2823 Note that the effectiveness of @code{aligned} attributes may be limited
2824 by inherent limitations in your linker. On many systems, the linker is
2825 only able to arrange for variables to be aligned up to a certain maximum
2826 alignment. (For some linkers, the maximum supported alignment may
2827 be very very small.) If your linker is only able to align variables
2828 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
2829 in an @code{__attribute__} will still only provide you with 8 byte
2830 alignment. See your linker documentation for further information.
2832 @item cleanup (@var{cleanup_function})
2833 @cindex @code{cleanup} attribute
2834 The @code{cleanup} attribute runs a function when the variable goes
2835 out of scope. This attribute can only be applied to auto function
2836 scope variables; it may not be applied to parameters or variables
2837 with static storage duration. The function must take one parameter,
2838 a pointer to a type compatible with the variable. The return value
2839 of the function (if any) is ignored.
2841 If @option{-fexceptions} is enabled, then @var{cleanup_function}
2842 will be run during the stack unwinding that happens during the
2843 processing of the exception. Note that the @code{cleanup} attribute
2844 does not allow the exception to be caught, only to perform an action.
2845 It is undefined what happens if @var{cleanup_function} does not
2850 @cindex @code{common} attribute
2851 @cindex @code{nocommon} attribute
2854 The @code{common} attribute requests GCC to place a variable in
2855 ``common'' storage. The @code{nocommon} attribute requests the
2856 opposite---to allocate space for it directly.
2858 These attributes override the default chosen by the
2859 @option{-fno-common} and @option{-fcommon} flags respectively.
2862 @cindex @code{deprecated} attribute
2863 The @code{deprecated} attribute results in a warning if the variable
2864 is used anywhere in the source file. This is useful when identifying
2865 variables that are expected to be removed in a future version of a
2866 program. The warning also includes the location of the declaration
2867 of the deprecated variable, to enable users to easily find further
2868 information about why the variable is deprecated, or what they should
2869 do instead. Note that the warning only occurs for uses:
2872 extern int old_var __attribute__ ((deprecated));
2874 int new_fn () @{ return old_var; @}
2877 results in a warning on line 3 but not line 2.
2879 The @code{deprecated} attribute can also be used for functions and
2880 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
2882 @item mode (@var{mode})
2883 @cindex @code{mode} attribute
2884 This attribute specifies the data type for the declaration---whichever
2885 type corresponds to the mode @var{mode}. This in effect lets you
2886 request an integer or floating point type according to its width.
2888 You may also specify a mode of @samp{byte} or @samp{__byte__} to
2889 indicate the mode corresponding to a one-byte integer, @samp{word} or
2890 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
2891 or @samp{__pointer__} for the mode used to represent pointers.
2894 @cindex @code{packed} attribute
2895 The @code{packed} attribute specifies that a variable or structure field
2896 should have the smallest possible alignment---one byte for a variable,
2897 and one bit for a field, unless you specify a larger value with the
2898 @code{aligned} attribute.
2900 Here is a structure in which the field @code{x} is packed, so that it
2901 immediately follows @code{a}:
2907 int x[2] __attribute__ ((packed));
2911 @item section ("@var{section-name}")
2912 @cindex @code{section} variable attribute
2913 Normally, the compiler places the objects it generates in sections like
2914 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
2915 or you need certain particular variables to appear in special sections,
2916 for example to map to special hardware. The @code{section}
2917 attribute specifies that a variable (or function) lives in a particular
2918 section. For example, this small program uses several specific section names:
2921 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
2922 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
2923 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
2924 int init_data __attribute__ ((section ("INITDATA"))) = 0;
2928 /* @r{Initialize stack pointer} */
2929 init_sp (stack + sizeof (stack));
2931 /* @r{Initialize initialized data} */
2932 memcpy (&init_data, &data, &edata - &data);
2934 /* @r{Turn on the serial ports} */
2941 Use the @code{section} attribute with an @emph{initialized} definition
2942 of a @emph{global} variable, as shown in the example. GCC issues
2943 a warning and otherwise ignores the @code{section} attribute in
2944 uninitialized variable declarations.
2946 You may only use the @code{section} attribute with a fully initialized
2947 global definition because of the way linkers work. The linker requires
2948 each object be defined once, with the exception that uninitialized
2949 variables tentatively go in the @code{common} (or @code{bss}) section
2950 and can be multiply ``defined''. You can force a variable to be
2951 initialized with the @option{-fno-common} flag or the @code{nocommon}
2954 Some file formats do not support arbitrary sections so the @code{section}
2955 attribute is not available on all platforms.
2956 If you need to map the entire contents of a module to a particular
2957 section, consider using the facilities of the linker instead.
2960 @cindex @code{shared} variable attribute
2961 On Microsoft Windows, in addition to putting variable definitions in a named
2962 section, the section can also be shared among all running copies of an
2963 executable or DLL@. For example, this small program defines shared data
2964 by putting it in a named section @code{shared} and marking the section
2968 int foo __attribute__((section ("shared"), shared)) = 0;
2973 /* @r{Read and write foo. All running
2974 copies see the same value.} */
2980 You may only use the @code{shared} attribute along with @code{section}
2981 attribute with a fully initialized global definition because of the way
2982 linkers work. See @code{section} attribute for more information.
2984 The @code{shared} attribute is only available on Microsoft Windows@.
2986 @item tls_model ("@var{tls_model}")
2987 @cindex @code{tls_model} attribute
2988 The @code{tls_model} attribute sets thread-local storage model
2989 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
2990 overriding @option{-ftls-model=} command line switch on a per-variable
2992 The @var{tls_model} argument should be one of @code{global-dynamic},
2993 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
2995 Not all targets support this attribute.
2997 @item transparent_union
2998 This attribute, attached to a function parameter which is a union, means
2999 that the corresponding argument may have the type of any union member,
3000 but the argument is passed as if its type were that of the first union
3001 member. For more details see @xref{Type Attributes}. You can also use
3002 this attribute on a @code{typedef} for a union data type; then it
3003 applies to all function parameters with that type.
3006 This attribute, attached to a variable, means that the variable is meant
3007 to be possibly unused. GCC will not produce a warning for this
3010 @item vector_size (@var{bytes})
3011 This attribute specifies the vector size for the variable, measured in
3012 bytes. For example, the declaration:
3015 int foo __attribute__ ((vector_size (16)));
3019 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3020 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3021 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3023 This attribute is only applicable to integral and float scalars,
3024 although arrays, pointers, and function return values are allowed in
3025 conjunction with this construct.
3027 Aggregates with this attribute are invalid, even if they are of the same
3028 size as a corresponding scalar. For example, the declaration:
3031 struct S @{ int a; @};
3032 struct S __attribute__ ((vector_size (16))) foo;
3036 is invalid even if the size of the structure is the same as the size of
3040 The @code{selectany} attribute causes an initialized global variable to
3041 have link-once semantics. When multiple definitions of the variable are
3042 encountered by the linker, the first is selected and the remainder are
3043 discarded. Following usage by the Microsoft compiler, the linker is told
3044 @emph{not} to warn about size or content differences of the multiple
3047 Although the primary usage of this attribute is for POD types, the
3048 attribute can also be applied to global C++ objects that are initialized
3049 by a constructor. In this case, the static initialization and destruction
3050 code for the object is emitted in each translation defining the object,
3051 but the calls to the constructor and destructor are protected by a
3052 link-once guard variable.
3054 The @code{selectany} attribute is only available on Microsoft Windows
3055 targets. You can use @code{__declspec (selectany)} as a synonym for
3056 @code{__attribute__ ((selectany))} for compatibility with other
3060 The @code{weak} attribute is described in @xref{Function Attributes}.
3063 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3066 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3070 @subsection M32R/D Variable Attributes
3072 One attribute is currently defined for the M32R/D@.
3075 @item model (@var{model-name})
3076 @cindex variable addressability on the M32R/D
3077 Use this attribute on the M32R/D to set the addressability of an object.
3078 The identifier @var{model-name} is one of @code{small}, @code{medium},
3079 or @code{large}, representing each of the code models.
3081 Small model objects live in the lower 16MB of memory (so that their
3082 addresses can be loaded with the @code{ld24} instruction).
3084 Medium and large model objects may live anywhere in the 32-bit address space
3085 (the compiler will generate @code{seth/add3} instructions to load their
3089 @subsection i386 Variable Attributes
3091 Two attributes are currently defined for i386 configurations:
3092 @code{ms_struct} and @code{gcc_struct}
3097 @cindex @code{ms_struct} attribute
3098 @cindex @code{gcc_struct} attribute
3100 If @code{packed} is used on a structure, or if bit-fields are used
3101 it may be that the Microsoft ABI packs them differently
3102 than GCC would normally pack them. Particularly when moving packed
3103 data between functions compiled with GCC and the native Microsoft compiler
3104 (either via function call or as data in a file), it may be necessary to access
3107 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3108 compilers to match the native Microsoft compiler.
3111 @subsection Xstormy16 Variable Attributes
3113 One attribute is currently defined for xstormy16 configurations:
3118 @cindex @code{below100} attribute
3120 If a variable has the @code{below100} attribute (@code{BELOW100} is
3121 allowed also), GCC will place the variable in the first 0x100 bytes of
3122 memory and use special opcodes to access it. Such variables will be
3123 placed in either the @code{.bss_below100} section or the
3124 @code{.data_below100} section.
3128 @node Type Attributes
3129 @section Specifying Attributes of Types
3130 @cindex attribute of types
3131 @cindex type attributes
3133 The keyword @code{__attribute__} allows you to specify special
3134 attributes of @code{struct} and @code{union} types when you define such
3135 types. This keyword is followed by an attribute specification inside
3136 double parentheses. Six attributes are currently defined for types:
3137 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3138 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3139 functions (@pxref{Function Attributes}) and for variables
3140 (@pxref{Variable Attributes}).
3142 You may also specify any one of these attributes with @samp{__}
3143 preceding and following its keyword. This allows you to use these
3144 attributes in header files without being concerned about a possible
3145 macro of the same name. For example, you may use @code{__aligned__}
3146 instead of @code{aligned}.
3148 You may specify the @code{aligned} and @code{transparent_union}
3149 attributes either in a @code{typedef} declaration or just past the
3150 closing curly brace of a complete enum, struct or union type
3151 @emph{definition} and the @code{packed} attribute only past the closing
3152 brace of a definition.
3154 You may also specify attributes between the enum, struct or union
3155 tag and the name of the type rather than after the closing brace.
3157 @xref{Attribute Syntax}, for details of the exact syntax for using
3161 @cindex @code{aligned} attribute
3162 @item aligned (@var{alignment})
3163 This attribute specifies a minimum alignment (in bytes) for variables
3164 of the specified type. For example, the declarations:
3167 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3168 typedef int more_aligned_int __attribute__ ((aligned (8)));
3172 force the compiler to insure (as far as it can) that each variable whose
3173 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3174 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3175 variables of type @code{struct S} aligned to 8-byte boundaries allows
3176 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3177 store) instructions when copying one variable of type @code{struct S} to
3178 another, thus improving run-time efficiency.
3180 Note that the alignment of any given @code{struct} or @code{union} type
3181 is required by the ISO C standard to be at least a perfect multiple of
3182 the lowest common multiple of the alignments of all of the members of
3183 the @code{struct} or @code{union} in question. This means that you @emph{can}
3184 effectively adjust the alignment of a @code{struct} or @code{union}
3185 type by attaching an @code{aligned} attribute to any one of the members
3186 of such a type, but the notation illustrated in the example above is a
3187 more obvious, intuitive, and readable way to request the compiler to
3188 adjust the alignment of an entire @code{struct} or @code{union} type.
3190 As in the preceding example, you can explicitly specify the alignment
3191 (in bytes) that you wish the compiler to use for a given @code{struct}
3192 or @code{union} type. Alternatively, you can leave out the alignment factor
3193 and just ask the compiler to align a type to the maximum
3194 useful alignment for the target machine you are compiling for. For
3195 example, you could write:
3198 struct S @{ short f[3]; @} __attribute__ ((aligned));
3201 Whenever you leave out the alignment factor in an @code{aligned}
3202 attribute specification, the compiler automatically sets the alignment
3203 for the type to the largest alignment which is ever used for any data
3204 type on the target machine you are compiling for. Doing this can often
3205 make copy operations more efficient, because the compiler can use
3206 whatever instructions copy the biggest chunks of memory when performing
3207 copies to or from the variables which have types that you have aligned
3210 In the example above, if the size of each @code{short} is 2 bytes, then
3211 the size of the entire @code{struct S} type is 6 bytes. The smallest
3212 power of two which is greater than or equal to that is 8, so the
3213 compiler sets the alignment for the entire @code{struct S} type to 8
3216 Note that although you can ask the compiler to select a time-efficient
3217 alignment for a given type and then declare only individual stand-alone
3218 objects of that type, the compiler's ability to select a time-efficient
3219 alignment is primarily useful only when you plan to create arrays of
3220 variables having the relevant (efficiently aligned) type. If you
3221 declare or use arrays of variables of an efficiently-aligned type, then
3222 it is likely that your program will also be doing pointer arithmetic (or
3223 subscripting, which amounts to the same thing) on pointers to the
3224 relevant type, and the code that the compiler generates for these
3225 pointer arithmetic operations will often be more efficient for
3226 efficiently-aligned types than for other types.
3228 The @code{aligned} attribute can only increase the alignment; but you
3229 can decrease it by specifying @code{packed} as well. See below.
3231 Note that the effectiveness of @code{aligned} attributes may be limited
3232 by inherent limitations in your linker. On many systems, the linker is
3233 only able to arrange for variables to be aligned up to a certain maximum
3234 alignment. (For some linkers, the maximum supported alignment may
3235 be very very small.) If your linker is only able to align variables
3236 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3237 in an @code{__attribute__} will still only provide you with 8 byte
3238 alignment. See your linker documentation for further information.
3241 This attribute, attached to @code{struct} or @code{union} type
3242 definition, specifies that each member of the structure or union is
3243 placed to minimize the memory required. When attached to an @code{enum}
3244 definition, it indicates that the smallest integral type should be used.
3246 @opindex fshort-enums
3247 Specifying this attribute for @code{struct} and @code{union} types is
3248 equivalent to specifying the @code{packed} attribute on each of the
3249 structure or union members. Specifying the @option{-fshort-enums}
3250 flag on the line is equivalent to specifying the @code{packed}
3251 attribute on all @code{enum} definitions.
3253 In the following example @code{struct my_packed_struct}'s members are
3254 packed closely together, but the internal layout of its @code{s} member
3255 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3259 struct my_unpacked_struct
3265 struct __attribute__ ((__packed__)) my_packed_struct
3269 struct my_unpacked_struct s;
3273 You may only specify this attribute on the definition of a @code{enum},
3274 @code{struct} or @code{union}, not on a @code{typedef} which does not
3275 also define the enumerated type, structure or union.
3277 @item transparent_union
3278 This attribute, attached to a @code{union} type definition, indicates
3279 that any function parameter having that union type causes calls to that
3280 function to be treated in a special way.
3282 First, the argument corresponding to a transparent union type can be of
3283 any type in the union; no cast is required. Also, if the union contains
3284 a pointer type, the corresponding argument can be a null pointer
3285 constant or a void pointer expression; and if the union contains a void
3286 pointer type, the corresponding argument can be any pointer expression.
3287 If the union member type is a pointer, qualifiers like @code{const} on
3288 the referenced type must be respected, just as with normal pointer
3291 Second, the argument is passed to the function using the calling
3292 conventions of the first member of the transparent union, not the calling
3293 conventions of the union itself. All members of the union must have the
3294 same machine representation; this is necessary for this argument passing
3297 Transparent unions are designed for library functions that have multiple
3298 interfaces for compatibility reasons. For example, suppose the
3299 @code{wait} function must accept either a value of type @code{int *} to
3300 comply with Posix, or a value of type @code{union wait *} to comply with
3301 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3302 @code{wait} would accept both kinds of arguments, but it would also
3303 accept any other pointer type and this would make argument type checking
3304 less useful. Instead, @code{<sys/wait.h>} might define the interface
3312 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3314 pid_t wait (wait_status_ptr_t);
3317 This interface allows either @code{int *} or @code{union wait *}
3318 arguments to be passed, using the @code{int *} calling convention.
3319 The program can call @code{wait} with arguments of either type:
3322 int w1 () @{ int w; return wait (&w); @}
3323 int w2 () @{ union wait w; return wait (&w); @}
3326 With this interface, @code{wait}'s implementation might look like this:
3329 pid_t wait (wait_status_ptr_t p)
3331 return waitpid (-1, p.__ip, 0);
3336 When attached to a type (including a @code{union} or a @code{struct}),
3337 this attribute means that variables of that type are meant to appear
3338 possibly unused. GCC will not produce a warning for any variables of
3339 that type, even if the variable appears to do nothing. This is often
3340 the case with lock or thread classes, which are usually defined and then
3341 not referenced, but contain constructors and destructors that have
3342 nontrivial bookkeeping functions.
3345 The @code{deprecated} attribute results in a warning if the type
3346 is used anywhere in the source file. This is useful when identifying
3347 types that are expected to be removed in a future version of a program.
3348 If possible, the warning also includes the location of the declaration
3349 of the deprecated type, to enable users to easily find further
3350 information about why the type is deprecated, or what they should do
3351 instead. Note that the warnings only occur for uses and then only
3352 if the type is being applied to an identifier that itself is not being
3353 declared as deprecated.
3356 typedef int T1 __attribute__ ((deprecated));
3360 typedef T1 T3 __attribute__ ((deprecated));
3361 T3 z __attribute__ ((deprecated));
3364 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3365 warning is issued for line 4 because T2 is not explicitly
3366 deprecated. Line 5 has no warning because T3 is explicitly
3367 deprecated. Similarly for line 6.
3369 The @code{deprecated} attribute can also be used for functions and
3370 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3373 Accesses to objects with types with this attribute are not subjected to
3374 type-based alias analysis, but are instead assumed to be able to alias
3375 any other type of objects, just like the @code{char} type. See
3376 @option{-fstrict-aliasing} for more information on aliasing issues.
3381 typedef short __attribute__((__may_alias__)) short_a;
3387 short_a *b = (short_a *) &a;
3391 if (a == 0x12345678)
3398 If you replaced @code{short_a} with @code{short} in the variable
3399 declaration, the above program would abort when compiled with
3400 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3401 above in recent GCC versions.
3403 @subsection ARM Type Attributes
3405 On those ARM targets that support @code{dllimport} (such as Symbian
3406 OS), you can use the @code{notshared} attribute to indicate that the
3407 virtual table and other similar data for a class should not be
3408 exported from a DLL@. For example:
3411 class __declspec(notshared) C @{
3413 __declspec(dllimport) C();
3417 __declspec(dllexport)
3421 In this code, @code{C::C} is exported from the current DLL, but the
3422 virtual table for @code{C} is not exported. (You can use
3423 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3424 most Symbian OS code uses @code{__declspec}.)
3426 @subsection i386 Type Attributes
3428 Two attributes are currently defined for i386 configurations:
3429 @code{ms_struct} and @code{gcc_struct}
3433 @cindex @code{ms_struct}
3434 @cindex @code{gcc_struct}
3436 If @code{packed} is used on a structure, or if bit-fields are used
3437 it may be that the Microsoft ABI packs them differently
3438 than GCC would normally pack them. Particularly when moving packed
3439 data between functions compiled with GCC and the native Microsoft compiler
3440 (either via function call or as data in a file), it may be necessary to access
3443 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3444 compilers to match the native Microsoft compiler.
3447 To specify multiple attributes, separate them by commas within the
3448 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3452 @section An Inline Function is As Fast As a Macro
3453 @cindex inline functions
3454 @cindex integrating function code
3456 @cindex macros, inline alternative
3458 By declaring a function @code{inline}, you can direct GCC to
3459 integrate that function's code into the code for its callers. This
3460 makes execution faster by eliminating the function-call overhead; in
3461 addition, if any of the actual argument values are constant, their known
3462 values may permit simplifications at compile time so that not all of the
3463 inline function's code needs to be included. The effect on code size is
3464 less predictable; object code may be larger or smaller with function
3465 inlining, depending on the particular case. Inlining of functions is an
3466 optimization and it really ``works'' only in optimizing compilation. If
3467 you don't use @option{-O}, no function is really inline.
3469 Inline functions are included in the ISO C99 standard, but there are
3470 currently substantial differences between what GCC implements and what
3471 the ISO C99 standard requires.
3473 To declare a function inline, use the @code{inline} keyword in its
3474 declaration, like this:
3484 (If you are writing a header file to be included in ISO C programs, write
3485 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3486 You can also make all ``simple enough'' functions inline with the option
3487 @option{-finline-functions}.
3490 Note that certain usages in a function definition can make it unsuitable
3491 for inline substitution. Among these usages are: use of varargs, use of
3492 alloca, use of variable sized data types (@pxref{Variable Length}),
3493 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3494 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3495 will warn when a function marked @code{inline} could not be substituted,
3496 and will give the reason for the failure.
3498 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3499 does not affect the linkage of the function.
3501 @cindex automatic @code{inline} for C++ member fns
3502 @cindex @code{inline} automatic for C++ member fns
3503 @cindex member fns, automatically @code{inline}
3504 @cindex C++ member fns, automatically @code{inline}
3505 @opindex fno-default-inline
3506 GCC automatically inlines member functions defined within the class
3507 body of C++ programs even if they are not explicitly declared
3508 @code{inline}. (You can override this with @option{-fno-default-inline};
3509 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3511 @cindex inline functions, omission of
3512 @opindex fkeep-inline-functions
3513 When a function is both inline and @code{static}, if all calls to the
3514 function are integrated into the caller, and the function's address is
3515 never used, then the function's own assembler code is never referenced.
3516 In this case, GCC does not actually output assembler code for the
3517 function, unless you specify the option @option{-fkeep-inline-functions}.
3518 Some calls cannot be integrated for various reasons (in particular,
3519 calls that precede the function's definition cannot be integrated, and
3520 neither can recursive calls within the definition). If there is a
3521 nonintegrated call, then the function is compiled to assembler code as
3522 usual. The function must also be compiled as usual if the program
3523 refers to its address, because that can't be inlined.
3525 @cindex non-static inline function
3526 When an inline function is not @code{static}, then the compiler must assume
3527 that there may be calls from other source files; since a global symbol can
3528 be defined only once in any program, the function must not be defined in
3529 the other source files, so the calls therein cannot be integrated.
3530 Therefore, a non-@code{static} inline function is always compiled on its
3531 own in the usual fashion.
3533 If you specify both @code{inline} and @code{extern} in the function
3534 definition, then the definition is used only for inlining. In no case
3535 is the function compiled on its own, not even if you refer to its
3536 address explicitly. Such an address becomes an external reference, as
3537 if you had only declared the function, and had not defined it.
3539 This combination of @code{inline} and @code{extern} has almost the
3540 effect of a macro. The way to use it is to put a function definition in
3541 a header file with these keywords, and put another copy of the
3542 definition (lacking @code{inline} and @code{extern}) in a library file.
3543 The definition in the header file will cause most calls to the function
3544 to be inlined. If any uses of the function remain, they will refer to
3545 the single copy in the library.
3547 Since GCC eventually will implement ISO C99 semantics for
3548 inline functions, it is best to use @code{static inline} only
3549 to guarantee compatibility. (The
3550 existing semantics will remain available when @option{-std=gnu89} is
3551 specified, but eventually the default will be @option{-std=gnu99} and
3552 that will implement the C99 semantics, though it does not do so yet.)
3554 GCC does not inline any functions when not optimizing unless you specify
3555 the @samp{always_inline} attribute for the function, like this:
3558 /* @r{Prototype.} */
3559 inline void foo (const char) __attribute__((always_inline));
3563 @section Assembler Instructions with C Expression Operands
3564 @cindex extended @code{asm}
3565 @cindex @code{asm} expressions
3566 @cindex assembler instructions
3569 In an assembler instruction using @code{asm}, you can specify the
3570 operands of the instruction using C expressions. This means you need not
3571 guess which registers or memory locations will contain the data you want
3574 You must specify an assembler instruction template much like what
3575 appears in a machine description, plus an operand constraint string for
3578 For example, here is how to use the 68881's @code{fsinx} instruction:
3581 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3585 Here @code{angle} is the C expression for the input operand while
3586 @code{result} is that of the output operand. Each has @samp{"f"} as its
3587 operand constraint, saying that a floating point register is required.
3588 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3589 output operands' constraints must use @samp{=}. The constraints use the
3590 same language used in the machine description (@pxref{Constraints}).
3592 Each operand is described by an operand-constraint string followed by
3593 the C expression in parentheses. A colon separates the assembler
3594 template from the first output operand and another separates the last
3595 output operand from the first input, if any. Commas separate the
3596 operands within each group. The total number of operands is currently
3597 limited to 30; this limitation may be lifted in some future version of
3600 If there are no output operands but there are input operands, you must
3601 place two consecutive colons surrounding the place where the output
3604 As of GCC version 3.1, it is also possible to specify input and output
3605 operands using symbolic names which can be referenced within the
3606 assembler code. These names are specified inside square brackets
3607 preceding the constraint string, and can be referenced inside the
3608 assembler code using @code{%[@var{name}]} instead of a percentage sign
3609 followed by the operand number. Using named operands the above example
3613 asm ("fsinx %[angle],%[output]"
3614 : [output] "=f" (result)
3615 : [angle] "f" (angle));
3619 Note that the symbolic operand names have no relation whatsoever to
3620 other C identifiers. You may use any name you like, even those of
3621 existing C symbols, but you must ensure that no two operands within the same
3622 assembler construct use the same symbolic name.
3624 Output operand expressions must be lvalues; the compiler can check this.
3625 The input operands need not be lvalues. The compiler cannot check
3626 whether the operands have data types that are reasonable for the
3627 instruction being executed. It does not parse the assembler instruction
3628 template and does not know what it means or even whether it is valid
3629 assembler input. The extended @code{asm} feature is most often used for
3630 machine instructions the compiler itself does not know exist. If
3631 the output expression cannot be directly addressed (for example, it is a
3632 bit-field), your constraint must allow a register. In that case, GCC
3633 will use the register as the output of the @code{asm}, and then store
3634 that register into the output.
3636 The ordinary output operands must be write-only; GCC will assume that
3637 the values in these operands before the instruction are dead and need
3638 not be generated. Extended asm supports input-output or read-write
3639 operands. Use the constraint character @samp{+} to indicate such an
3640 operand and list it with the output operands. You should only use
3641 read-write operands when the constraints for the operand (or the
3642 operand in which only some of the bits are to be changed) allow a
3645 You may, as an alternative, logically split its function into two
3646 separate operands, one input operand and one write-only output
3647 operand. The connection between them is expressed by constraints
3648 which say they need to be in the same location when the instruction
3649 executes. You can use the same C expression for both operands, or
3650 different expressions. For example, here we write the (fictitious)
3651 @samp{combine} instruction with @code{bar} as its read-only source
3652 operand and @code{foo} as its read-write destination:
3655 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3659 The constraint @samp{"0"} for operand 1 says that it must occupy the
3660 same location as operand 0. A number in constraint is allowed only in
3661 an input operand and it must refer to an output operand.
3663 Only a number in the constraint can guarantee that one operand will be in
3664 the same place as another. The mere fact that @code{foo} is the value
3665 of both operands is not enough to guarantee that they will be in the
3666 same place in the generated assembler code. The following would not
3670 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3673 Various optimizations or reloading could cause operands 0 and 1 to be in
3674 different registers; GCC knows no reason not to do so. For example, the
3675 compiler might find a copy of the value of @code{foo} in one register and
3676 use it for operand 1, but generate the output operand 0 in a different
3677 register (copying it afterward to @code{foo}'s own address). Of course,
3678 since the register for operand 1 is not even mentioned in the assembler
3679 code, the result will not work, but GCC can't tell that.
3681 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3682 the operand number for a matching constraint. For example:
3685 asm ("cmoveq %1,%2,%[result]"
3686 : [result] "=r"(result)
3687 : "r" (test), "r"(new), "[result]"(old));
3690 Sometimes you need to make an @code{asm} operand be a specific register,
3691 but there's no matching constraint letter for that register @emph{by
3692 itself}. To force the operand into that register, use a local variable
3693 for the operand and specify the register in the variable declaration.
3694 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
3695 register constraint letter that matches the register:
3698 register int *p1 asm ("r0") = @dots{};
3699 register int *p2 asm ("r1") = @dots{};
3700 register int *result asm ("r0");
3701 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
3704 @anchor{Example of asm with clobbered asm reg}
3705 In the above example, beware that a register that is call-clobbered by
3706 the target ABI will be overwritten by any function call in the
3707 assignment, including library calls for arithmetic operators.
3708 Assuming it is a call-clobbered register, this may happen to @code{r0}
3709 above by the assignment to @code{p2}. If you have to use such a
3710 register, use temporary variables for expressions between the register
3715 register int *p1 asm ("r0") = @dots{};
3716 register int *p2 asm ("r1") = t1;
3717 register int *result asm ("r0");
3718 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
3721 Some instructions clobber specific hard registers. To describe this,
3722 write a third colon after the input operands, followed by the names of
3723 the clobbered hard registers (given as strings). Here is a realistic
3724 example for the VAX:
3727 asm volatile ("movc3 %0,%1,%2"
3728 : /* @r{no outputs} */
3729 : "g" (from), "g" (to), "g" (count)
3730 : "r0", "r1", "r2", "r3", "r4", "r5");
3733 You may not write a clobber description in a way that overlaps with an
3734 input or output operand. For example, you may not have an operand
3735 describing a register class with one member if you mention that register
3736 in the clobber list. Variables declared to live in specific registers
3737 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3738 have no part mentioned in the clobber description.
3739 There is no way for you to specify that an input
3740 operand is modified without also specifying it as an output
3741 operand. Note that if all the output operands you specify are for this
3742 purpose (and hence unused), you will then also need to specify
3743 @code{volatile} for the @code{asm} construct, as described below, to
3744 prevent GCC from deleting the @code{asm} statement as unused.
3746 If you refer to a particular hardware register from the assembler code,
3747 you will probably have to list the register after the third colon to
3748 tell the compiler the register's value is modified. In some assemblers,
3749 the register names begin with @samp{%}; to produce one @samp{%} in the
3750 assembler code, you must write @samp{%%} in the input.
3752 If your assembler instruction can alter the condition code register, add
3753 @samp{cc} to the list of clobbered registers. GCC on some machines
3754 represents the condition codes as a specific hardware register;
3755 @samp{cc} serves to name this register. On other machines, the
3756 condition code is handled differently, and specifying @samp{cc} has no
3757 effect. But it is valid no matter what the machine.
3759 If your assembler instructions access memory in an unpredictable
3760 fashion, add @samp{memory} to the list of clobbered registers. This
3761 will cause GCC to not keep memory values cached in registers across the
3762 assembler instruction and not optimize stores or loads to that memory.
3763 You will also want to add the @code{volatile} keyword if the memory
3764 affected is not listed in the inputs or outputs of the @code{asm}, as
3765 the @samp{memory} clobber does not count as a side-effect of the
3766 @code{asm}. If you know how large the accessed memory is, you can add
3767 it as input or output but if this is not known, you should add
3768 @samp{memory}. As an example, if you access ten bytes of a string, you
3769 can use a memory input like:
3772 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
3775 Note that in the following example the memory input is necessary,
3776 otherwise GCC might optimize the store to @code{x} away:
3783 asm ("magic stuff accessing an 'int' pointed to by '%1'"
3784 "=&d" (r) : "a" (y), "m" (*y));
3789 You can put multiple assembler instructions together in a single
3790 @code{asm} template, separated by the characters normally used in assembly
3791 code for the system. A combination that works in most places is a newline
3792 to break the line, plus a tab character to move to the instruction field
3793 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3794 assembler allows semicolons as a line-breaking character. Note that some
3795 assembler dialects use semicolons to start a comment.
3796 The input operands are guaranteed not to use any of the clobbered
3797 registers, and neither will the output operands' addresses, so you can
3798 read and write the clobbered registers as many times as you like. Here
3799 is an example of multiple instructions in a template; it assumes the
3800 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3803 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3805 : "g" (from), "g" (to)
3809 Unless an output operand has the @samp{&} constraint modifier, GCC
3810 may allocate it in the same register as an unrelated input operand, on
3811 the assumption the inputs are consumed before the outputs are produced.
3812 This assumption may be false if the assembler code actually consists of
3813 more than one instruction. In such a case, use @samp{&} for each output
3814 operand that may not overlap an input. @xref{Modifiers}.
3816 If you want to test the condition code produced by an assembler
3817 instruction, you must include a branch and a label in the @code{asm}
3818 construct, as follows:
3821 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3827 This assumes your assembler supports local labels, as the GNU assembler
3828 and most Unix assemblers do.
3830 Speaking of labels, jumps from one @code{asm} to another are not
3831 supported. The compiler's optimizers do not know about these jumps, and
3832 therefore they cannot take account of them when deciding how to
3835 @cindex macros containing @code{asm}
3836 Usually the most convenient way to use these @code{asm} instructions is to
3837 encapsulate them in macros that look like functions. For example,
3841 (@{ double __value, __arg = (x); \
3842 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
3847 Here the variable @code{__arg} is used to make sure that the instruction
3848 operates on a proper @code{double} value, and to accept only those
3849 arguments @code{x} which can convert automatically to a @code{double}.
3851 Another way to make sure the instruction operates on the correct data
3852 type is to use a cast in the @code{asm}. This is different from using a
3853 variable @code{__arg} in that it converts more different types. For
3854 example, if the desired type were @code{int}, casting the argument to
3855 @code{int} would accept a pointer with no complaint, while assigning the
3856 argument to an @code{int} variable named @code{__arg} would warn about
3857 using a pointer unless the caller explicitly casts it.
3859 If an @code{asm} has output operands, GCC assumes for optimization
3860 purposes the instruction has no side effects except to change the output
3861 operands. This does not mean instructions with a side effect cannot be
3862 used, but you must be careful, because the compiler may eliminate them
3863 if the output operands aren't used, or move them out of loops, or
3864 replace two with one if they constitute a common subexpression. Also,
3865 if your instruction does have a side effect on a variable that otherwise
3866 appears not to change, the old value of the variable may be reused later
3867 if it happens to be found in a register.
3869 You can prevent an @code{asm} instruction from being deleted
3870 by writing the keyword @code{volatile} after
3871 the @code{asm}. For example:
3874 #define get_and_set_priority(new) \
3876 asm volatile ("get_and_set_priority %0, %1" \
3877 : "=g" (__old) : "g" (new)); \
3882 The @code{volatile} keyword indicates that the instruction has
3883 important side-effects. GCC will not delete a volatile @code{asm} if
3884 it is reachable. (The instruction can still be deleted if GCC can
3885 prove that control-flow will never reach the location of the
3886 instruction.) Note that even a volatile @code{asm} instruction
3887 can be moved relative to other code, including across jump
3888 instructions. For example, on many targets there is a system
3889 register which can be set to control the rounding mode of
3890 floating point operations. You might try
3891 setting it with a volatile @code{asm}, like this PowerPC example:
3894 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
3899 This will not work reliably, as the compiler may move the addition back
3900 before the volatile @code{asm}. To make it work you need to add an
3901 artificial dependency to the @code{asm} referencing a variable in the code
3902 you don't want moved, for example:
3905 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
3909 Similarly, you can't expect a
3910 sequence of volatile @code{asm} instructions to remain perfectly
3911 consecutive. If you want consecutive output, use a single @code{asm}.
3912 Also, GCC will perform some optimizations across a volatile @code{asm}
3913 instruction; GCC does not ``forget everything'' when it encounters
3914 a volatile @code{asm} instruction the way some other compilers do.
3916 An @code{asm} instruction without any output operands will be treated
3917 identically to a volatile @code{asm} instruction.
3919 It is a natural idea to look for a way to give access to the condition
3920 code left by the assembler instruction. However, when we attempted to
3921 implement this, we found no way to make it work reliably. The problem
3922 is that output operands might need reloading, which would result in
3923 additional following ``store'' instructions. On most machines, these
3924 instructions would alter the condition code before there was time to
3925 test it. This problem doesn't arise for ordinary ``test'' and
3926 ``compare'' instructions because they don't have any output operands.
3928 For reasons similar to those described above, it is not possible to give
3929 an assembler instruction access to the condition code left by previous
3932 If you are writing a header file that should be includable in ISO C
3933 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
3936 @subsection Size of an @code{asm}
3938 Some targets require that GCC track the size of each instruction used in
3939 order to generate correct code. Because the final length of an
3940 @code{asm} is only known by the assembler, GCC must make an estimate as
3941 to how big it will be. The estimate is formed by counting the number of
3942 statements in the pattern of the @code{asm} and multiplying that by the
3943 length of the longest instruction on that processor. Statements in the
3944 @code{asm} are identified by newline characters and whatever statement
3945 separator characters are supported by the assembler; on most processors
3946 this is the `@code{;}' character.
3948 Normally, GCC's estimate is perfectly adequate to ensure that correct
3949 code is generated, but it is possible to confuse the compiler if you use
3950 pseudo instructions or assembler macros that expand into multiple real
3951 instructions or if you use assembler directives that expand to more
3952 space in the object file than would be needed for a single instruction.
3953 If this happens then the assembler will produce a diagnostic saying that
3954 a label is unreachable.
3956 @subsection i386 floating point asm operands
3958 There are several rules on the usage of stack-like regs in
3959 asm_operands insns. These rules apply only to the operands that are
3964 Given a set of input regs that die in an asm_operands, it is
3965 necessary to know which are implicitly popped by the asm, and
3966 which must be explicitly popped by gcc.
3968 An input reg that is implicitly popped by the asm must be
3969 explicitly clobbered, unless it is constrained to match an
3973 For any input reg that is implicitly popped by an asm, it is
3974 necessary to know how to adjust the stack to compensate for the pop.
3975 If any non-popped input is closer to the top of the reg-stack than
3976 the implicitly popped reg, it would not be possible to know what the
3977 stack looked like---it's not clear how the rest of the stack ``slides
3980 All implicitly popped input regs must be closer to the top of
3981 the reg-stack than any input that is not implicitly popped.
3983 It is possible that if an input dies in an insn, reload might
3984 use the input reg for an output reload. Consider this example:
3987 asm ("foo" : "=t" (a) : "f" (b));
3990 This asm says that input B is not popped by the asm, and that
3991 the asm pushes a result onto the reg-stack, i.e., the stack is one
3992 deeper after the asm than it was before. But, it is possible that
3993 reload will think that it can use the same reg for both the input and
3994 the output, if input B dies in this insn.
3996 If any input operand uses the @code{f} constraint, all output reg
3997 constraints must use the @code{&} earlyclobber.
3999 The asm above would be written as
4002 asm ("foo" : "=&t" (a) : "f" (b));
4006 Some operands need to be in particular places on the stack. All
4007 output operands fall in this category---there is no other way to
4008 know which regs the outputs appear in unless the user indicates
4009 this in the constraints.
4011 Output operands must specifically indicate which reg an output
4012 appears in after an asm. @code{=f} is not allowed: the operand
4013 constraints must select a class with a single reg.
4016 Output operands may not be ``inserted'' between existing stack regs.
4017 Since no 387 opcode uses a read/write operand, all output operands
4018 are dead before the asm_operands, and are pushed by the asm_operands.
4019 It makes no sense to push anywhere but the top of the reg-stack.
4021 Output operands must start at the top of the reg-stack: output
4022 operands may not ``skip'' a reg.
4025 Some asm statements may need extra stack space for internal
4026 calculations. This can be guaranteed by clobbering stack registers
4027 unrelated to the inputs and outputs.
4031 Here are a couple of reasonable asms to want to write. This asm
4032 takes one input, which is internally popped, and produces two outputs.
4035 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4038 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4039 and replaces them with one output. The user must code the @code{st(1)}
4040 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4043 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4049 @section Controlling Names Used in Assembler Code
4050 @cindex assembler names for identifiers
4051 @cindex names used in assembler code
4052 @cindex identifiers, names in assembler code
4054 You can specify the name to be used in the assembler code for a C
4055 function or variable by writing the @code{asm} (or @code{__asm__})
4056 keyword after the declarator as follows:
4059 int foo asm ("myfoo") = 2;
4063 This specifies that the name to be used for the variable @code{foo} in
4064 the assembler code should be @samp{myfoo} rather than the usual
4067 On systems where an underscore is normally prepended to the name of a C
4068 function or variable, this feature allows you to define names for the
4069 linker that do not start with an underscore.
4071 It does not make sense to use this feature with a non-static local
4072 variable since such variables do not have assembler names. If you are
4073 trying to put the variable in a particular register, see @ref{Explicit
4074 Reg Vars}. GCC presently accepts such code with a warning, but will
4075 probably be changed to issue an error, rather than a warning, in the
4078 You cannot use @code{asm} in this way in a function @emph{definition}; but
4079 you can get the same effect by writing a declaration for the function
4080 before its definition and putting @code{asm} there, like this:
4083 extern func () asm ("FUNC");
4090 It is up to you to make sure that the assembler names you choose do not
4091 conflict with any other assembler symbols. Also, you must not use a
4092 register name; that would produce completely invalid assembler code. GCC
4093 does not as yet have the ability to store static variables in registers.
4094 Perhaps that will be added.
4096 @node Explicit Reg Vars
4097 @section Variables in Specified Registers
4098 @cindex explicit register variables
4099 @cindex variables in specified registers
4100 @cindex specified registers
4101 @cindex registers, global allocation
4103 GNU C allows you to put a few global variables into specified hardware
4104 registers. You can also specify the register in which an ordinary
4105 register variable should be allocated.
4109 Global register variables reserve registers throughout the program.
4110 This may be useful in programs such as programming language
4111 interpreters which have a couple of global variables that are accessed
4115 Local register variables in specific registers do not reserve the
4116 registers, except at the point where they are used as input or output
4117 operands in an @code{asm} statement and the @code{asm} statement itself is
4118 not deleted. The compiler's data flow analysis is capable of determining
4119 where the specified registers contain live values, and where they are
4120 available for other uses. Stores into local register variables may be deleted
4121 when they appear to be dead according to dataflow analysis. References
4122 to local register variables may be deleted or moved or simplified.
4124 These local variables are sometimes convenient for use with the extended
4125 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4126 output of the assembler instruction directly into a particular register.
4127 (This will work provided the register you specify fits the constraints
4128 specified for that operand in the @code{asm}.)
4136 @node Global Reg Vars
4137 @subsection Defining Global Register Variables
4138 @cindex global register variables
4139 @cindex registers, global variables in
4141 You can define a global register variable in GNU C like this:
4144 register int *foo asm ("a5");
4148 Here @code{a5} is the name of the register which should be used. Choose a
4149 register which is normally saved and restored by function calls on your
4150 machine, so that library routines will not clobber it.
4152 Naturally the register name is cpu-dependent, so you would need to
4153 conditionalize your program according to cpu type. The register
4154 @code{a5} would be a good choice on a 68000 for a variable of pointer
4155 type. On machines with register windows, be sure to choose a ``global''
4156 register that is not affected magically by the function call mechanism.
4158 In addition, operating systems on one type of cpu may differ in how they
4159 name the registers; then you would need additional conditionals. For
4160 example, some 68000 operating systems call this register @code{%a5}.
4162 Eventually there may be a way of asking the compiler to choose a register
4163 automatically, but first we need to figure out how it should choose and
4164 how to enable you to guide the choice. No solution is evident.
4166 Defining a global register variable in a certain register reserves that
4167 register entirely for this use, at least within the current compilation.
4168 The register will not be allocated for any other purpose in the functions
4169 in the current compilation. The register will not be saved and restored by
4170 these functions. Stores into this register are never deleted even if they
4171 would appear to be dead, but references may be deleted or moved or
4174 It is not safe to access the global register variables from signal
4175 handlers, or from more than one thread of control, because the system
4176 library routines may temporarily use the register for other things (unless
4177 you recompile them specially for the task at hand).
4179 @cindex @code{qsort}, and global register variables
4180 It is not safe for one function that uses a global register variable to
4181 call another such function @code{foo} by way of a third function
4182 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4183 different source file in which the variable wasn't declared). This is
4184 because @code{lose} might save the register and put some other value there.
4185 For example, you can't expect a global register variable to be available in
4186 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4187 might have put something else in that register. (If you are prepared to
4188 recompile @code{qsort} with the same global register variable, you can
4189 solve this problem.)
4191 If you want to recompile @code{qsort} or other source files which do not
4192 actually use your global register variable, so that they will not use that
4193 register for any other purpose, then it suffices to specify the compiler
4194 option @option{-ffixed-@var{reg}}. You need not actually add a global
4195 register declaration to their source code.
4197 A function which can alter the value of a global register variable cannot
4198 safely be called from a function compiled without this variable, because it
4199 could clobber the value the caller expects to find there on return.
4200 Therefore, the function which is the entry point into the part of the
4201 program that uses the global register variable must explicitly save and
4202 restore the value which belongs to its caller.
4204 @cindex register variable after @code{longjmp}
4205 @cindex global register after @code{longjmp}
4206 @cindex value after @code{longjmp}
4209 On most machines, @code{longjmp} will restore to each global register
4210 variable the value it had at the time of the @code{setjmp}. On some
4211 machines, however, @code{longjmp} will not change the value of global
4212 register variables. To be portable, the function that called @code{setjmp}
4213 should make other arrangements to save the values of the global register
4214 variables, and to restore them in a @code{longjmp}. This way, the same
4215 thing will happen regardless of what @code{longjmp} does.
4217 All global register variable declarations must precede all function
4218 definitions. If such a declaration could appear after function
4219 definitions, the declaration would be too late to prevent the register from
4220 being used for other purposes in the preceding functions.
4222 Global register variables may not have initial values, because an
4223 executable file has no means to supply initial contents for a register.
4225 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4226 registers, but certain library functions, such as @code{getwd}, as well
4227 as the subroutines for division and remainder, modify g3 and g4. g1 and
4228 g2 are local temporaries.
4230 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4231 Of course, it will not do to use more than a few of those.
4233 @node Local Reg Vars
4234 @subsection Specifying Registers for Local Variables
4235 @cindex local variables, specifying registers
4236 @cindex specifying registers for local variables
4237 @cindex registers for local variables
4239 You can define a local register variable with a specified register
4243 register int *foo asm ("a5");
4247 Here @code{a5} is the name of the register which should be used. Note
4248 that this is the same syntax used for defining global register
4249 variables, but for a local variable it would appear within a function.
4251 Naturally the register name is cpu-dependent, but this is not a
4252 problem, since specific registers are most often useful with explicit
4253 assembler instructions (@pxref{Extended Asm}). Both of these things
4254 generally require that you conditionalize your program according to
4257 In addition, operating systems on one type of cpu may differ in how they
4258 name the registers; then you would need additional conditionals. For
4259 example, some 68000 operating systems call this register @code{%a5}.
4261 Defining such a register variable does not reserve the register; it
4262 remains available for other uses in places where flow control determines
4263 the variable's value is not live.
4265 This option does not guarantee that GCC will generate code that has
4266 this variable in the register you specify at all times. You may not
4267 code an explicit reference to this register in the @emph{assembler
4268 instruction template} part of an @code{asm} statement and assume it will
4269 always refer to this variable. However, using the variable as an
4270 @code{asm} @emph{operand} guarantees that the specified register is used
4273 Stores into local register variables may be deleted when they appear to be dead
4274 according to dataflow analysis. References to local register variables may
4275 be deleted or moved or simplified.
4277 As for global register variables, it's recommended that you choose a
4278 register which is normally saved and restored by function calls on
4279 your machine, so that library routines will not clobber it. A common
4280 pitfall is to initialize multiple call-clobbered registers with
4281 arbitrary expressions, where a function call or library call for an
4282 arithmetic operator will overwrite a register value from a previous
4283 assignment, for example @code{r0} below:
4285 register int *p1 asm ("r0") = @dots{};
4286 register int *p2 asm ("r1") = @dots{};
4288 In those cases, a solution is to use a temporary variable for
4289 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4291 @node Alternate Keywords
4292 @section Alternate Keywords
4293 @cindex alternate keywords
4294 @cindex keywords, alternate
4296 @option{-ansi} and the various @option{-std} options disable certain
4297 keywords. This causes trouble when you want to use GNU C extensions, or
4298 a general-purpose header file that should be usable by all programs,
4299 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4300 @code{inline} are not available in programs compiled with
4301 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4302 program compiled with @option{-std=c99}). The ISO C99 keyword
4303 @code{restrict} is only available when @option{-std=gnu99} (which will
4304 eventually be the default) or @option{-std=c99} (or the equivalent
4305 @option{-std=iso9899:1999}) is used.
4307 The way to solve these problems is to put @samp{__} at the beginning and
4308 end of each problematical keyword. For example, use @code{__asm__}
4309 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4311 Other C compilers won't accept these alternative keywords; if you want to
4312 compile with another compiler, you can define the alternate keywords as
4313 macros to replace them with the customary keywords. It looks like this:
4321 @findex __extension__
4323 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4325 prevent such warnings within one expression by writing
4326 @code{__extension__} before the expression. @code{__extension__} has no
4327 effect aside from this.
4329 @node Incomplete Enums
4330 @section Incomplete @code{enum} Types
4332 You can define an @code{enum} tag without specifying its possible values.
4333 This results in an incomplete type, much like what you get if you write
4334 @code{struct foo} without describing the elements. A later declaration
4335 which does specify the possible values completes the type.
4337 You can't allocate variables or storage using the type while it is
4338 incomplete. However, you can work with pointers to that type.
4340 This extension may not be very useful, but it makes the handling of
4341 @code{enum} more consistent with the way @code{struct} and @code{union}
4344 This extension is not supported by GNU C++.
4346 @node Function Names
4347 @section Function Names as Strings
4348 @cindex @code{__func__} identifier
4349 @cindex @code{__FUNCTION__} identifier
4350 @cindex @code{__PRETTY_FUNCTION__} identifier
4352 GCC provides three magic variables which hold the name of the current
4353 function, as a string. The first of these is @code{__func__}, which
4354 is part of the C99 standard:
4357 The identifier @code{__func__} is implicitly declared by the translator
4358 as if, immediately following the opening brace of each function
4359 definition, the declaration
4362 static const char __func__[] = "function-name";
4365 appeared, where function-name is the name of the lexically-enclosing
4366 function. This name is the unadorned name of the function.
4369 @code{__FUNCTION__} is another name for @code{__func__}. Older
4370 versions of GCC recognize only this name. However, it is not
4371 standardized. For maximum portability, we recommend you use
4372 @code{__func__}, but provide a fallback definition with the
4376 #if __STDC_VERSION__ < 199901L
4378 # define __func__ __FUNCTION__
4380 # define __func__ "<unknown>"
4385 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4386 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4387 the type signature of the function as well as its bare name. For
4388 example, this program:
4392 extern int printf (char *, ...);
4399 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4400 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4418 __PRETTY_FUNCTION__ = void a::sub(int)
4421 These identifiers are not preprocessor macros. In GCC 3.3 and
4422 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4423 were treated as string literals; they could be used to initialize
4424 @code{char} arrays, and they could be concatenated with other string
4425 literals. GCC 3.4 and later treat them as variables, like
4426 @code{__func__}. In C++, @code{__FUNCTION__} and
4427 @code{__PRETTY_FUNCTION__} have always been variables.
4429 @node Return Address
4430 @section Getting the Return or Frame Address of a Function
4432 These functions may be used to get information about the callers of a
4435 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4436 This function returns the return address of the current function, or of
4437 one of its callers. The @var{level} argument is number of frames to
4438 scan up the call stack. A value of @code{0} yields the return address
4439 of the current function, a value of @code{1} yields the return address
4440 of the caller of the current function, and so forth. When inlining
4441 the expected behavior is that the function will return the address of
4442 the function that will be returned to. To work around this behavior use
4443 the @code{noinline} function attribute.
4445 The @var{level} argument must be a constant integer.
4447 On some machines it may be impossible to determine the return address of
4448 any function other than the current one; in such cases, or when the top
4449 of the stack has been reached, this function will return @code{0} or a
4450 random value. In addition, @code{__builtin_frame_address} may be used
4451 to determine if the top of the stack has been reached.
4453 This function should only be used with a nonzero argument for debugging
4457 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4458 This function is similar to @code{__builtin_return_address}, but it
4459 returns the address of the function frame rather than the return address
4460 of the function. Calling @code{__builtin_frame_address} with a value of
4461 @code{0} yields the frame address of the current function, a value of
4462 @code{1} yields the frame address of the caller of the current function,
4465 The frame is the area on the stack which holds local variables and saved
4466 registers. The frame address is normally the address of the first word
4467 pushed on to the stack by the function. However, the exact definition
4468 depends upon the processor and the calling convention. If the processor
4469 has a dedicated frame pointer register, and the function has a frame,
4470 then @code{__builtin_frame_address} will return the value of the frame
4473 On some machines it may be impossible to determine the frame address of
4474 any function other than the current one; in such cases, or when the top
4475 of the stack has been reached, this function will return @code{0} if
4476 the first frame pointer is properly initialized by the startup code.
4478 This function should only be used with a nonzero argument for debugging
4482 @node Vector Extensions
4483 @section Using vector instructions through built-in functions
4485 On some targets, the instruction set contains SIMD vector instructions that
4486 operate on multiple values contained in one large register at the same time.
4487 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4490 The first step in using these extensions is to provide the necessary data
4491 types. This should be done using an appropriate @code{typedef}:
4494 typedef int v4si __attribute__ ((vector_size (16)));
4497 The @code{int} type specifies the base type, while the attribute specifies
4498 the vector size for the variable, measured in bytes. For example, the
4499 declaration above causes the compiler to set the mode for the @code{v4si}
4500 type to be 16 bytes wide and divided into @code{int} sized units. For
4501 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4502 corresponding mode of @code{foo} will be @acronym{V4SI}.
4504 The @code{vector_size} attribute is only applicable to integral and
4505 float scalars, although arrays, pointers, and function return values
4506 are allowed in conjunction with this construct.
4508 All the basic integer types can be used as base types, both as signed
4509 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4510 @code{long long}. In addition, @code{float} and @code{double} can be
4511 used to build floating-point vector types.
4513 Specifying a combination that is not valid for the current architecture
4514 will cause GCC to synthesize the instructions using a narrower mode.
4515 For example, if you specify a variable of type @code{V4SI} and your
4516 architecture does not allow for this specific SIMD type, GCC will
4517 produce code that uses 4 @code{SIs}.
4519 The types defined in this manner can be used with a subset of normal C
4520 operations. Currently, GCC will allow using the following operators
4521 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4523 The operations behave like C++ @code{valarrays}. Addition is defined as
4524 the addition of the corresponding elements of the operands. For
4525 example, in the code below, each of the 4 elements in @var{a} will be
4526 added to the corresponding 4 elements in @var{b} and the resulting
4527 vector will be stored in @var{c}.
4530 typedef int v4si __attribute__ ((vector_size (16)));
4537 Subtraction, multiplication, division, and the logical operations
4538 operate in a similar manner. Likewise, the result of using the unary
4539 minus or complement operators on a vector type is a vector whose
4540 elements are the negative or complemented values of the corresponding
4541 elements in the operand.
4543 You can declare variables and use them in function calls and returns, as
4544 well as in assignments and some casts. You can specify a vector type as
4545 a return type for a function. Vector types can also be used as function
4546 arguments. It is possible to cast from one vector type to another,
4547 provided they are of the same size (in fact, you can also cast vectors
4548 to and from other datatypes of the same size).
4550 You cannot operate between vectors of different lengths or different
4551 signedness without a cast.
4553 A port that supports hardware vector operations, usually provides a set
4554 of built-in functions that can be used to operate on vectors. For
4555 example, a function to add two vectors and multiply the result by a
4556 third could look like this:
4559 v4si f (v4si a, v4si b, v4si c)
4561 v4si tmp = __builtin_addv4si (a, b);
4562 return __builtin_mulv4si (tmp, c);
4569 @findex __builtin_offsetof
4571 GCC implements for both C and C++ a syntactic extension to implement
4572 the @code{offsetof} macro.
4576 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4578 offsetof_member_designator:
4580 | offsetof_member_designator "." @code{identifier}
4581 | offsetof_member_designator "[" @code{expr} "]"
4584 This extension is sufficient such that
4587 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4590 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4591 may be dependent. In either case, @var{member} may consist of a single
4592 identifier, or a sequence of member accesses and array references.
4594 @node Atomic Builtins
4595 @section Built-in functions for atomic memory access
4597 The following builtins are intended to be compatible with those described
4598 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
4599 section 7.4. As such, they depart from the normal GCC practice of using
4600 the ``__builtin_'' prefix, and further that they are overloaded such that
4601 they work on multiple types.
4603 The definition given in the Intel documentation allows only for the use of
4604 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
4605 counterparts. GCC will allow any integral scalar or pointer type that is
4606 1, 2, 4 or 8 bytes in length.
4608 Not all operations are supported by all target processors. If a particular
4609 operation cannot be implemented on the target processor, a warning will be
4610 generated and a call an external function will be generated. The external
4611 function will carry the same name as the builtin, with an additional suffix
4612 @samp{_@var{n}} where @var{n} is the size of the data type.
4614 @c ??? Should we have a mechanism to suppress this warning? This is almost
4615 @c useful for implementing the operation under the control of an external
4618 In most cases, these builtins are considered a @dfn{full barrier}. That is,
4619 no memory operand will be moved across the operation, either forward or
4620 backward. Further, instructions will be issued as necessary to prevent the
4621 processor from speculating loads across the operation and from queuing stores
4622 after the operation.
4624 All of the routines are are described in the Intel documentation to take
4625 ``an optional list of variables protected by the memory barrier''. It's
4626 not clear what is meant by that; it could mean that @emph{only} the
4627 following variables are protected, or it could mean that these variables
4628 should in addition be protected. At present GCC ignores this list and
4629 protects all variables which are globally accessible. If in the future
4630 we make some use of this list, an empty list will continue to mean all
4631 globally accessible variables.
4634 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
4635 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
4636 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
4637 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
4638 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
4639 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
4640 @findex __sync_fetch_and_add
4641 @findex __sync_fetch_and_sub
4642 @findex __sync_fetch_and_or
4643 @findex __sync_fetch_and_and
4644 @findex __sync_fetch_and_xor
4645 @findex __sync_fetch_and_nand
4646 These builtins perform the operation suggested by the name, and
4647 returns the value that had previously been in memory. That is,
4650 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
4651 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
4654 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
4655 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
4656 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
4657 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
4658 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
4659 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
4660 @findex __sync_add_and_fetch
4661 @findex __sync_sub_and_fetch
4662 @findex __sync_or_and_fetch
4663 @findex __sync_and_and_fetch
4664 @findex __sync_xor_and_fetch
4665 @findex __sync_nand_and_fetch
4666 These builtins perform the operation suggested by the name, and
4667 return the new value. That is,
4670 @{ *ptr @var{op}= value; return *ptr; @}
4671 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
4674 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
4675 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
4676 @findex __sync_bool_compare_and_swap
4677 @findex __sync_val_compare_and_swap
4678 These builtins perform an atomic compare and swap. That is, if the current
4679 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
4682 The ``bool'' version returns true if the comparison is successful and
4683 @var{newval} was written. The ``val'' version returns the contents
4684 of @code{*@var{ptr}} before the operation.
4686 @item __sync_synchronize (...)
4687 @findex __sync_synchronize
4688 This builtin issues a full memory barrier.
4690 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
4691 @findex __sync_lock_test_and_set
4692 This builtin, as described by Intel, is not a traditional test-and-set
4693 operation, but rather an atomic exchange operation. It writes @var{value}
4694 into @code{*@var{ptr}}, and returns the previous contents of
4697 Many targets have only minimal support for such locks, and do not support
4698 a full exchange operation. In this case, a target may support reduced
4699 functionality here by which the @emph{only} valid value to store is the
4700 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
4701 is implementation defined.
4703 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
4704 This means that references after the builtin cannot move to (or be
4705 speculated to) before the builtin, but previous memory stores may not
4706 be globally visible yet, and previous memory loads may not yet be
4709 @item void __sync_lock_release (@var{type} *ptr, ...)
4710 @findex __sync_lock_release
4711 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
4712 Normally this means writing the constant 0 to @code{*@var{ptr}}.
4714 This builtin is not a full barrier, but rather a @dfn{release barrier}.
4715 This means that all previous memory stores are globally visible, and all
4716 previous memory loads have been satisfied, but following memory reads
4717 are not prevented from being speculated to before the barrier.
4720 @node Other Builtins
4721 @section Other built-in functions provided by GCC
4722 @cindex built-in functions
4723 @findex __builtin_isgreater
4724 @findex __builtin_isgreaterequal
4725 @findex __builtin_isless
4726 @findex __builtin_islessequal
4727 @findex __builtin_islessgreater
4728 @findex __builtin_isunordered
4729 @findex __builtin_powi
4730 @findex __builtin_powif
4731 @findex __builtin_powil
4889 @findex fprintf_unlocked
4891 @findex fputs_unlocked
5001 @findex printf_unlocked
5030 @findex significandf
5031 @findex significandl
5102 GCC provides a large number of built-in functions other than the ones
5103 mentioned above. Some of these are for internal use in the processing
5104 of exceptions or variable-length argument lists and will not be
5105 documented here because they may change from time to time; we do not
5106 recommend general use of these functions.
5108 The remaining functions are provided for optimization purposes.
5110 @opindex fno-builtin
5111 GCC includes built-in versions of many of the functions in the standard
5112 C library. The versions prefixed with @code{__builtin_} will always be
5113 treated as having the same meaning as the C library function even if you
5114 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5115 Many of these functions are only optimized in certain cases; if they are
5116 not optimized in a particular case, a call to the library function will
5121 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5122 @option{-std=c99}), the functions
5123 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5124 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5125 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5126 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5127 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5128 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5129 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5130 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5131 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5132 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5133 @code{significandf}, @code{significandl}, @code{significand},
5134 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5135 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5136 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5137 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5138 @code{ynl} and @code{yn}
5139 may be handled as built-in functions.
5140 All these functions have corresponding versions
5141 prefixed with @code{__builtin_}, which may be used even in strict C89
5144 The ISO C99 functions
5145 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5146 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5147 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5148 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5149 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5150 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5151 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5152 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5153 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5154 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5155 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5156 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5157 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5158 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5159 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5160 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5161 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5162 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5163 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5164 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5165 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5166 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5167 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5168 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5169 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5170 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5171 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5172 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5173 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5174 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5175 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5176 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5177 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5178 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5179 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5180 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5181 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5182 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5183 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5184 are handled as built-in functions
5185 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5187 There are also built-in versions of the ISO C99 functions
5188 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5189 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5190 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5191 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5192 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5193 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5194 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5195 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5196 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5197 that are recognized in any mode since ISO C90 reserves these names for
5198 the purpose to which ISO C99 puts them. All these functions have
5199 corresponding versions prefixed with @code{__builtin_}.
5201 The ISO C94 functions
5202 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5203 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5204 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5206 are handled as built-in functions
5207 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5209 The ISO C90 functions
5210 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5211 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5212 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5213 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5214 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5215 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5216 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5217 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5218 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5219 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5220 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5221 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5222 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5223 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5224 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5225 @code{vprintf} and @code{vsprintf}
5226 are all recognized as built-in functions unless
5227 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5228 is specified for an individual function). All of these functions have
5229 corresponding versions prefixed with @code{__builtin_}.
5231 GCC provides built-in versions of the ISO C99 floating point comparison
5232 macros that avoid raising exceptions for unordered operands. They have
5233 the same names as the standard macros ( @code{isgreater},
5234 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5235 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5236 prefixed. We intend for a library implementor to be able to simply
5237 @code{#define} each standard macro to its built-in equivalent.
5239 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5241 You can use the built-in function @code{__builtin_types_compatible_p} to
5242 determine whether two types are the same.
5244 This built-in function returns 1 if the unqualified versions of the
5245 types @var{type1} and @var{type2} (which are types, not expressions) are
5246 compatible, 0 otherwise. The result of this built-in function can be
5247 used in integer constant expressions.
5249 This built-in function ignores top level qualifiers (e.g., @code{const},
5250 @code{volatile}). For example, @code{int} is equivalent to @code{const
5253 The type @code{int[]} and @code{int[5]} are compatible. On the other
5254 hand, @code{int} and @code{char *} are not compatible, even if the size
5255 of their types, on the particular architecture are the same. Also, the
5256 amount of pointer indirection is taken into account when determining
5257 similarity. Consequently, @code{short *} is not similar to
5258 @code{short **}. Furthermore, two types that are typedefed are
5259 considered compatible if their underlying types are compatible.
5261 An @code{enum} type is not considered to be compatible with another
5262 @code{enum} type even if both are compatible with the same integer
5263 type; this is what the C standard specifies.
5264 For example, @code{enum @{foo, bar@}} is not similar to
5265 @code{enum @{hot, dog@}}.
5267 You would typically use this function in code whose execution varies
5268 depending on the arguments' types. For example:
5274 if (__builtin_types_compatible_p (typeof (x), long double)) \
5275 tmp = foo_long_double (tmp); \
5276 else if (__builtin_types_compatible_p (typeof (x), double)) \
5277 tmp = foo_double (tmp); \
5278 else if (__builtin_types_compatible_p (typeof (x), float)) \
5279 tmp = foo_float (tmp); \
5286 @emph{Note:} This construct is only available for C@.
5290 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5292 You can use the built-in function @code{__builtin_choose_expr} to
5293 evaluate code depending on the value of a constant expression. This
5294 built-in function returns @var{exp1} if @var{const_exp}, which is a
5295 constant expression that must be able to be determined at compile time,
5296 is nonzero. Otherwise it returns 0.
5298 This built-in function is analogous to the @samp{? :} operator in C,
5299 except that the expression returned has its type unaltered by promotion
5300 rules. Also, the built-in function does not evaluate the expression
5301 that was not chosen. For example, if @var{const_exp} evaluates to true,
5302 @var{exp2} is not evaluated even if it has side-effects.
5304 This built-in function can return an lvalue if the chosen argument is an
5307 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5308 type. Similarly, if @var{exp2} is returned, its return type is the same
5315 __builtin_choose_expr ( \
5316 __builtin_types_compatible_p (typeof (x), double), \
5318 __builtin_choose_expr ( \
5319 __builtin_types_compatible_p (typeof (x), float), \
5321 /* @r{The void expression results in a compile-time error} \
5322 @r{when assigning the result to something.} */ \
5326 @emph{Note:} This construct is only available for C@. Furthermore, the
5327 unused expression (@var{exp1} or @var{exp2} depending on the value of
5328 @var{const_exp}) may still generate syntax errors. This may change in
5333 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5334 You can use the built-in function @code{__builtin_constant_p} to
5335 determine if a value is known to be constant at compile-time and hence
5336 that GCC can perform constant-folding on expressions involving that
5337 value. The argument of the function is the value to test. The function
5338 returns the integer 1 if the argument is known to be a compile-time
5339 constant and 0 if it is not known to be a compile-time constant. A
5340 return of 0 does not indicate that the value is @emph{not} a constant,
5341 but merely that GCC cannot prove it is a constant with the specified
5342 value of the @option{-O} option.
5344 You would typically use this function in an embedded application where
5345 memory was a critical resource. If you have some complex calculation,
5346 you may want it to be folded if it involves constants, but need to call
5347 a function if it does not. For example:
5350 #define Scale_Value(X) \
5351 (__builtin_constant_p (X) \
5352 ? ((X) * SCALE + OFFSET) : Scale (X))
5355 You may use this built-in function in either a macro or an inline
5356 function. However, if you use it in an inlined function and pass an
5357 argument of the function as the argument to the built-in, GCC will
5358 never return 1 when you call the inline function with a string constant
5359 or compound literal (@pxref{Compound Literals}) and will not return 1
5360 when you pass a constant numeric value to the inline function unless you
5361 specify the @option{-O} option.
5363 You may also use @code{__builtin_constant_p} in initializers for static
5364 data. For instance, you can write
5367 static const int table[] = @{
5368 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5374 This is an acceptable initializer even if @var{EXPRESSION} is not a
5375 constant expression. GCC must be more conservative about evaluating the
5376 built-in in this case, because it has no opportunity to perform
5379 Previous versions of GCC did not accept this built-in in data
5380 initializers. The earliest version where it is completely safe is
5384 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5385 @opindex fprofile-arcs
5386 You may use @code{__builtin_expect} to provide the compiler with
5387 branch prediction information. In general, you should prefer to
5388 use actual profile feedback for this (@option{-fprofile-arcs}), as
5389 programmers are notoriously bad at predicting how their programs
5390 actually perform. However, there are applications in which this
5391 data is hard to collect.
5393 The return value is the value of @var{exp}, which should be an
5394 integral expression. The value of @var{c} must be a compile-time
5395 constant. The semantics of the built-in are that it is expected
5396 that @var{exp} == @var{c}. For example:
5399 if (__builtin_expect (x, 0))
5404 would indicate that we do not expect to call @code{foo}, since
5405 we expect @code{x} to be zero. Since you are limited to integral
5406 expressions for @var{exp}, you should use constructions such as
5409 if (__builtin_expect (ptr != NULL, 1))
5414 when testing pointer or floating-point values.
5417 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5418 This function is used to minimize cache-miss latency by moving data into
5419 a cache before it is accessed.
5420 You can insert calls to @code{__builtin_prefetch} into code for which
5421 you know addresses of data in memory that is likely to be accessed soon.
5422 If the target supports them, data prefetch instructions will be generated.
5423 If the prefetch is done early enough before the access then the data will
5424 be in the cache by the time it is accessed.
5426 The value of @var{addr} is the address of the memory to prefetch.
5427 There are two optional arguments, @var{rw} and @var{locality}.
5428 The value of @var{rw} is a compile-time constant one or zero; one
5429 means that the prefetch is preparing for a write to the memory address
5430 and zero, the default, means that the prefetch is preparing for a read.
5431 The value @var{locality} must be a compile-time constant integer between
5432 zero and three. A value of zero means that the data has no temporal
5433 locality, so it need not be left in the cache after the access. A value
5434 of three means that the data has a high degree of temporal locality and
5435 should be left in all levels of cache possible. Values of one and two
5436 mean, respectively, a low or moderate degree of temporal locality. The
5440 for (i = 0; i < n; i++)
5443 __builtin_prefetch (&a[i+j], 1, 1);
5444 __builtin_prefetch (&b[i+j], 0, 1);
5449 Data prefetch does not generate faults if @var{addr} is invalid, but
5450 the address expression itself must be valid. For example, a prefetch
5451 of @code{p->next} will not fault if @code{p->next} is not a valid
5452 address, but evaluation will fault if @code{p} is not a valid address.
5454 If the target does not support data prefetch, the address expression
5455 is evaluated if it includes side effects but no other code is generated
5456 and GCC does not issue a warning.
5459 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5460 Returns a positive infinity, if supported by the floating-point format,
5461 else @code{DBL_MAX}. This function is suitable for implementing the
5462 ISO C macro @code{HUGE_VAL}.
5465 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5466 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5469 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5470 Similar to @code{__builtin_huge_val}, except the return
5471 type is @code{long double}.
5474 @deftypefn {Built-in Function} double __builtin_inf (void)
5475 Similar to @code{__builtin_huge_val}, except a warning is generated
5476 if the target floating-point format does not support infinities.
5479 @deftypefn {Built-in Function} float __builtin_inff (void)
5480 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5481 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5484 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5485 Similar to @code{__builtin_inf}, except the return
5486 type is @code{long double}.
5489 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5490 This is an implementation of the ISO C99 function @code{nan}.
5492 Since ISO C99 defines this function in terms of @code{strtod}, which we
5493 do not implement, a description of the parsing is in order. The string
5494 is parsed as by @code{strtol}; that is, the base is recognized by
5495 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5496 in the significand such that the least significant bit of the number
5497 is at the least significant bit of the significand. The number is
5498 truncated to fit the significand field provided. The significand is
5499 forced to be a quiet NaN@.
5501 This function, if given a string literal, is evaluated early enough
5502 that it is considered a compile-time constant.
5505 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5506 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5509 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5510 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5513 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5514 Similar to @code{__builtin_nan}, except the significand is forced
5515 to be a signaling NaN@. The @code{nans} function is proposed by
5516 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
5519 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5520 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5523 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5524 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5527 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5528 Returns one plus the index of the least significant 1-bit of @var{x}, or
5529 if @var{x} is zero, returns zero.
5532 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5533 Returns the number of leading 0-bits in @var{x}, starting at the most
5534 significant bit position. If @var{x} is 0, the result is undefined.
5537 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5538 Returns the number of trailing 0-bits in @var{x}, starting at the least
5539 significant bit position. If @var{x} is 0, the result is undefined.
5542 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5543 Returns the number of 1-bits in @var{x}.
5546 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5547 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
5551 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5552 Similar to @code{__builtin_ffs}, except the argument type is
5553 @code{unsigned long}.
5556 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5557 Similar to @code{__builtin_clz}, except the argument type is
5558 @code{unsigned long}.
5561 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5562 Similar to @code{__builtin_ctz}, except the argument type is
5563 @code{unsigned long}.
5566 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5567 Similar to @code{__builtin_popcount}, except the argument type is
5568 @code{unsigned long}.
5571 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5572 Similar to @code{__builtin_parity}, except the argument type is
5573 @code{unsigned long}.
5576 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5577 Similar to @code{__builtin_ffs}, except the argument type is
5578 @code{unsigned long long}.
5581 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5582 Similar to @code{__builtin_clz}, except the argument type is
5583 @code{unsigned long long}.
5586 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5587 Similar to @code{__builtin_ctz}, except the argument type is
5588 @code{unsigned long long}.
5591 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5592 Similar to @code{__builtin_popcount}, except the argument type is
5593 @code{unsigned long long}.
5596 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5597 Similar to @code{__builtin_parity}, except the argument type is
5598 @code{unsigned long long}.
5601 @deftypefn {Built-in Function} double __builtin_powi (double, int)
5602 Returns the first argument raised to the power of the second. Unlike the
5603 @code{pow} function no guarantees about precision and rounding are made.
5606 @deftypefn {Built-in Function} float __builtin_powif (float, int)
5607 Similar to @code{__builtin_powi}, except the argument and return types
5611 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
5612 Similar to @code{__builtin_powi}, except the argument and return types
5613 are @code{long double}.
5617 @node Target Builtins
5618 @section Built-in Functions Specific to Particular Target Machines
5620 On some target machines, GCC supports many built-in functions specific
5621 to those machines. Generally these generate calls to specific machine
5622 instructions, but allow the compiler to schedule those calls.
5625 * Alpha Built-in Functions::
5626 * ARM Built-in Functions::
5627 * FR-V Built-in Functions::
5628 * X86 Built-in Functions::
5629 * MIPS Paired-Single Support::
5630 * PowerPC AltiVec Built-in Functions::
5631 * SPARC VIS Built-in Functions::
5634 @node Alpha Built-in Functions
5635 @subsection Alpha Built-in Functions
5637 These built-in functions are available for the Alpha family of
5638 processors, depending on the command-line switches used.
5640 The following built-in functions are always available. They
5641 all generate the machine instruction that is part of the name.
5644 long __builtin_alpha_implver (void)
5645 long __builtin_alpha_rpcc (void)
5646 long __builtin_alpha_amask (long)
5647 long __builtin_alpha_cmpbge (long, long)
5648 long __builtin_alpha_extbl (long, long)
5649 long __builtin_alpha_extwl (long, long)
5650 long __builtin_alpha_extll (long, long)
5651 long __builtin_alpha_extql (long, long)
5652 long __builtin_alpha_extwh (long, long)
5653 long __builtin_alpha_extlh (long, long)
5654 long __builtin_alpha_extqh (long, long)
5655 long __builtin_alpha_insbl (long, long)
5656 long __builtin_alpha_inswl (long, long)
5657 long __builtin_alpha_insll (long, long)
5658 long __builtin_alpha_insql (long, long)
5659 long __builtin_alpha_inswh (long, long)
5660 long __builtin_alpha_inslh (long, long)
5661 long __builtin_alpha_insqh (long, long)
5662 long __builtin_alpha_mskbl (long, long)
5663 long __builtin_alpha_mskwl (long, long)
5664 long __builtin_alpha_mskll (long, long)
5665 long __builtin_alpha_mskql (long, long)
5666 long __builtin_alpha_mskwh (long, long)
5667 long __builtin_alpha_msklh (long, long)
5668 long __builtin_alpha_mskqh (long, long)
5669 long __builtin_alpha_umulh (long, long)
5670 long __builtin_alpha_zap (long, long)
5671 long __builtin_alpha_zapnot (long, long)
5674 The following built-in functions are always with @option{-mmax}
5675 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5676 later. They all generate the machine instruction that is part
5680 long __builtin_alpha_pklb (long)
5681 long __builtin_alpha_pkwb (long)
5682 long __builtin_alpha_unpkbl (long)
5683 long __builtin_alpha_unpkbw (long)
5684 long __builtin_alpha_minub8 (long, long)
5685 long __builtin_alpha_minsb8 (long, long)
5686 long __builtin_alpha_minuw4 (long, long)
5687 long __builtin_alpha_minsw4 (long, long)
5688 long __builtin_alpha_maxub8 (long, long)
5689 long __builtin_alpha_maxsb8 (long, long)
5690 long __builtin_alpha_maxuw4 (long, long)
5691 long __builtin_alpha_maxsw4 (long, long)
5692 long __builtin_alpha_perr (long, long)
5695 The following built-in functions are always with @option{-mcix}
5696 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5697 later. They all generate the machine instruction that is part
5701 long __builtin_alpha_cttz (long)
5702 long __builtin_alpha_ctlz (long)
5703 long __builtin_alpha_ctpop (long)
5706 The following builtins are available on systems that use the OSF/1
5707 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5708 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5709 @code{rdval} and @code{wrval}.
5712 void *__builtin_thread_pointer (void)
5713 void __builtin_set_thread_pointer (void *)
5716 @node ARM Built-in Functions
5717 @subsection ARM Built-in Functions
5719 These built-in functions are available for the ARM family of
5720 processors, when the @option{-mcpu=iwmmxt} switch is used:
5723 typedef int v2si __attribute__ ((vector_size (8)));
5724 typedef short v4hi __attribute__ ((vector_size (8)));
5725 typedef char v8qi __attribute__ ((vector_size (8)));
5727 int __builtin_arm_getwcx (int)
5728 void __builtin_arm_setwcx (int, int)
5729 int __builtin_arm_textrmsb (v8qi, int)
5730 int __builtin_arm_textrmsh (v4hi, int)
5731 int __builtin_arm_textrmsw (v2si, int)
5732 int __builtin_arm_textrmub (v8qi, int)
5733 int __builtin_arm_textrmuh (v4hi, int)
5734 int __builtin_arm_textrmuw (v2si, int)
5735 v8qi __builtin_arm_tinsrb (v8qi, int)
5736 v4hi __builtin_arm_tinsrh (v4hi, int)
5737 v2si __builtin_arm_tinsrw (v2si, int)
5738 long long __builtin_arm_tmia (long long, int, int)
5739 long long __builtin_arm_tmiabb (long long, int, int)
5740 long long __builtin_arm_tmiabt (long long, int, int)
5741 long long __builtin_arm_tmiaph (long long, int, int)
5742 long long __builtin_arm_tmiatb (long long, int, int)
5743 long long __builtin_arm_tmiatt (long long, int, int)
5744 int __builtin_arm_tmovmskb (v8qi)
5745 int __builtin_arm_tmovmskh (v4hi)
5746 int __builtin_arm_tmovmskw (v2si)
5747 long long __builtin_arm_waccb (v8qi)
5748 long long __builtin_arm_wacch (v4hi)
5749 long long __builtin_arm_waccw (v2si)
5750 v8qi __builtin_arm_waddb (v8qi, v8qi)
5751 v8qi __builtin_arm_waddbss (v8qi, v8qi)
5752 v8qi __builtin_arm_waddbus (v8qi, v8qi)
5753 v4hi __builtin_arm_waddh (v4hi, v4hi)
5754 v4hi __builtin_arm_waddhss (v4hi, v4hi)
5755 v4hi __builtin_arm_waddhus (v4hi, v4hi)
5756 v2si __builtin_arm_waddw (v2si, v2si)
5757 v2si __builtin_arm_waddwss (v2si, v2si)
5758 v2si __builtin_arm_waddwus (v2si, v2si)
5759 v8qi __builtin_arm_walign (v8qi, v8qi, int)
5760 long long __builtin_arm_wand(long long, long long)
5761 long long __builtin_arm_wandn (long long, long long)
5762 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5763 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5764 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5765 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5766 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5767 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5768 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5769 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5770 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5771 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5772 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5773 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5774 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5775 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5776 long long __builtin_arm_wmacsz (v4hi, v4hi)
5777 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5778 long long __builtin_arm_wmacuz (v4hi, v4hi)
5779 v4hi __builtin_arm_wmadds (v4hi, v4hi)
5780 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5781 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5782 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5783 v2si __builtin_arm_wmaxsw (v2si, v2si)
5784 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5785 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5786 v2si __builtin_arm_wmaxuw (v2si, v2si)
5787 v8qi __builtin_arm_wminsb (v8qi, v8qi)
5788 v4hi __builtin_arm_wminsh (v4hi, v4hi)
5789 v2si __builtin_arm_wminsw (v2si, v2si)
5790 v8qi __builtin_arm_wminub (v8qi, v8qi)
5791 v4hi __builtin_arm_wminuh (v4hi, v4hi)
5792 v2si __builtin_arm_wminuw (v2si, v2si)
5793 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5794 v4hi __builtin_arm_wmulul (v4hi, v4hi)
5795 v4hi __builtin_arm_wmulum (v4hi, v4hi)
5796 long long __builtin_arm_wor (long long, long long)
5797 v2si __builtin_arm_wpackdss (long long, long long)
5798 v2si __builtin_arm_wpackdus (long long, long long)
5799 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5800 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5801 v4hi __builtin_arm_wpackwss (v2si, v2si)
5802 v4hi __builtin_arm_wpackwus (v2si, v2si)
5803 long long __builtin_arm_wrord (long long, long long)
5804 long long __builtin_arm_wrordi (long long, int)
5805 v4hi __builtin_arm_wrorh (v4hi, long long)
5806 v4hi __builtin_arm_wrorhi (v4hi, int)
5807 v2si __builtin_arm_wrorw (v2si, long long)
5808 v2si __builtin_arm_wrorwi (v2si, int)
5809 v2si __builtin_arm_wsadb (v8qi, v8qi)
5810 v2si __builtin_arm_wsadbz (v8qi, v8qi)
5811 v2si __builtin_arm_wsadh (v4hi, v4hi)
5812 v2si __builtin_arm_wsadhz (v4hi, v4hi)
5813 v4hi __builtin_arm_wshufh (v4hi, int)
5814 long long __builtin_arm_wslld (long long, long long)
5815 long long __builtin_arm_wslldi (long long, int)
5816 v4hi __builtin_arm_wsllh (v4hi, long long)
5817 v4hi __builtin_arm_wsllhi (v4hi, int)
5818 v2si __builtin_arm_wsllw (v2si, long long)
5819 v2si __builtin_arm_wsllwi (v2si, int)
5820 long long __builtin_arm_wsrad (long long, long long)
5821 long long __builtin_arm_wsradi (long long, int)
5822 v4hi __builtin_arm_wsrah (v4hi, long long)
5823 v4hi __builtin_arm_wsrahi (v4hi, int)
5824 v2si __builtin_arm_wsraw (v2si, long long)
5825 v2si __builtin_arm_wsrawi (v2si, int)
5826 long long __builtin_arm_wsrld (long long, long long)
5827 long long __builtin_arm_wsrldi (long long, int)
5828 v4hi __builtin_arm_wsrlh (v4hi, long long)
5829 v4hi __builtin_arm_wsrlhi (v4hi, int)
5830 v2si __builtin_arm_wsrlw (v2si, long long)
5831 v2si __builtin_arm_wsrlwi (v2si, int)
5832 v8qi __builtin_arm_wsubb (v8qi, v8qi)
5833 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5834 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5835 v4hi __builtin_arm_wsubh (v4hi, v4hi)
5836 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5837 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5838 v2si __builtin_arm_wsubw (v2si, v2si)
5839 v2si __builtin_arm_wsubwss (v2si, v2si)
5840 v2si __builtin_arm_wsubwus (v2si, v2si)
5841 v4hi __builtin_arm_wunpckehsb (v8qi)
5842 v2si __builtin_arm_wunpckehsh (v4hi)
5843 long long __builtin_arm_wunpckehsw (v2si)
5844 v4hi __builtin_arm_wunpckehub (v8qi)
5845 v2si __builtin_arm_wunpckehuh (v4hi)
5846 long long __builtin_arm_wunpckehuw (v2si)
5847 v4hi __builtin_arm_wunpckelsb (v8qi)
5848 v2si __builtin_arm_wunpckelsh (v4hi)
5849 long long __builtin_arm_wunpckelsw (v2si)
5850 v4hi __builtin_arm_wunpckelub (v8qi)
5851 v2si __builtin_arm_wunpckeluh (v4hi)
5852 long long __builtin_arm_wunpckeluw (v2si)
5853 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5854 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5855 v2si __builtin_arm_wunpckihw (v2si, v2si)
5856 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5857 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5858 v2si __builtin_arm_wunpckilw (v2si, v2si)
5859 long long __builtin_arm_wxor (long long, long long)
5860 long long __builtin_arm_wzero ()
5863 @node FR-V Built-in Functions
5864 @subsection FR-V Built-in Functions
5866 GCC provides many FR-V-specific built-in functions. In general,
5867 these functions are intended to be compatible with those described
5868 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
5869 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
5870 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
5871 pointer rather than by value.
5873 Most of the functions are named after specific FR-V instructions.
5874 Such functions are said to be ``directly mapped'' and are summarized
5875 here in tabular form.
5879 * Directly-mapped Integer Functions::
5880 * Directly-mapped Media Functions::
5881 * Other Built-in Functions::
5884 @node Argument Types
5885 @subsubsection Argument Types
5887 The arguments to the built-in functions can be divided into three groups:
5888 register numbers, compile-time constants and run-time values. In order
5889 to make this classification clear at a glance, the arguments and return
5890 values are given the following pseudo types:
5892 @multitable @columnfractions .20 .30 .15 .35
5893 @item Pseudo type @tab Real C type @tab Constant? @tab Description
5894 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
5895 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
5896 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
5897 @item @code{uw2} @tab @code{unsigned long long} @tab No
5898 @tab an unsigned doubleword
5899 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
5900 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
5901 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
5902 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
5905 These pseudo types are not defined by GCC, they are simply a notational
5906 convenience used in this manual.
5908 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
5909 and @code{sw2} are evaluated at run time. They correspond to
5910 register operands in the underlying FR-V instructions.
5912 @code{const} arguments represent immediate operands in the underlying
5913 FR-V instructions. They must be compile-time constants.
5915 @code{acc} arguments are evaluated at compile time and specify the number
5916 of an accumulator register. For example, an @code{acc} argument of 2
5917 will select the ACC2 register.
5919 @code{iacc} arguments are similar to @code{acc} arguments but specify the
5920 number of an IACC register. See @pxref{Other Built-in Functions}
5923 @node Directly-mapped Integer Functions
5924 @subsubsection Directly-mapped Integer Functions
5926 The functions listed below map directly to FR-V I-type instructions.
5928 @multitable @columnfractions .45 .32 .23
5929 @item Function prototype @tab Example usage @tab Assembly output
5930 @item @code{sw1 __ADDSS (sw1, sw1)}
5931 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
5932 @tab @code{ADDSS @var{a},@var{b},@var{c}}
5933 @item @code{sw1 __SCAN (sw1, sw1)}
5934 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
5935 @tab @code{SCAN @var{a},@var{b},@var{c}}
5936 @item @code{sw1 __SCUTSS (sw1)}
5937 @tab @code{@var{b} = __SCUTSS (@var{a})}
5938 @tab @code{SCUTSS @var{a},@var{b}}
5939 @item @code{sw1 __SLASS (sw1, sw1)}
5940 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
5941 @tab @code{SLASS @var{a},@var{b},@var{c}}
5942 @item @code{void __SMASS (sw1, sw1)}
5943 @tab @code{__SMASS (@var{a}, @var{b})}
5944 @tab @code{SMASS @var{a},@var{b}}
5945 @item @code{void __SMSSS (sw1, sw1)}
5946 @tab @code{__SMSSS (@var{a}, @var{b})}
5947 @tab @code{SMSSS @var{a},@var{b}}
5948 @item @code{void __SMU (sw1, sw1)}
5949 @tab @code{__SMU (@var{a}, @var{b})}
5950 @tab @code{SMU @var{a},@var{b}}
5951 @item @code{sw2 __SMUL (sw1, sw1)}
5952 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
5953 @tab @code{SMUL @var{a},@var{b},@var{c}}
5954 @item @code{sw1 __SUBSS (sw1, sw1)}
5955 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
5956 @tab @code{SUBSS @var{a},@var{b},@var{c}}
5957 @item @code{uw2 __UMUL (uw1, uw1)}
5958 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
5959 @tab @code{UMUL @var{a},@var{b},@var{c}}
5962 @node Directly-mapped Media Functions
5963 @subsubsection Directly-mapped Media Functions
5965 The functions listed below map directly to FR-V M-type instructions.
5967 @multitable @columnfractions .45 .32 .23
5968 @item Function prototype @tab Example usage @tab Assembly output
5969 @item @code{uw1 __MABSHS (sw1)}
5970 @tab @code{@var{b} = __MABSHS (@var{a})}
5971 @tab @code{MABSHS @var{a},@var{b}}
5972 @item @code{void __MADDACCS (acc, acc)}
5973 @tab @code{__MADDACCS (@var{b}, @var{a})}
5974 @tab @code{MADDACCS @var{a},@var{b}}
5975 @item @code{sw1 __MADDHSS (sw1, sw1)}
5976 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
5977 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
5978 @item @code{uw1 __MADDHUS (uw1, uw1)}
5979 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
5980 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
5981 @item @code{uw1 __MAND (uw1, uw1)}
5982 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
5983 @tab @code{MAND @var{a},@var{b},@var{c}}
5984 @item @code{void __MASACCS (acc, acc)}
5985 @tab @code{__MASACCS (@var{b}, @var{a})}
5986 @tab @code{MASACCS @var{a},@var{b}}
5987 @item @code{uw1 __MAVEH (uw1, uw1)}
5988 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
5989 @tab @code{MAVEH @var{a},@var{b},@var{c}}
5990 @item @code{uw2 __MBTOH (uw1)}
5991 @tab @code{@var{b} = __MBTOH (@var{a})}
5992 @tab @code{MBTOH @var{a},@var{b}}
5993 @item @code{void __MBTOHE (uw1 *, uw1)}
5994 @tab @code{__MBTOHE (&@var{b}, @var{a})}
5995 @tab @code{MBTOHE @var{a},@var{b}}
5996 @item @code{void __MCLRACC (acc)}
5997 @tab @code{__MCLRACC (@var{a})}
5998 @tab @code{MCLRACC @var{a}}
5999 @item @code{void __MCLRACCA (void)}
6000 @tab @code{__MCLRACCA ()}
6001 @tab @code{MCLRACCA}
6002 @item @code{uw1 __Mcop1 (uw1, uw1)}
6003 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6004 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6005 @item @code{uw1 __Mcop2 (uw1, uw1)}
6006 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6007 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6008 @item @code{uw1 __MCPLHI (uw2, const)}
6009 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6010 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6011 @item @code{uw1 __MCPLI (uw2, const)}
6012 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6013 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6014 @item @code{void __MCPXIS (acc, sw1, sw1)}
6015 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6016 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6017 @item @code{void __MCPXIU (acc, uw1, uw1)}
6018 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6019 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6020 @item @code{void __MCPXRS (acc, sw1, sw1)}
6021 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6022 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6023 @item @code{void __MCPXRU (acc, uw1, uw1)}
6024 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6025 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6026 @item @code{uw1 __MCUT (acc, uw1)}
6027 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6028 @tab @code{MCUT @var{a},@var{b},@var{c}}
6029 @item @code{uw1 __MCUTSS (acc, sw1)}
6030 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6031 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6032 @item @code{void __MDADDACCS (acc, acc)}
6033 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6034 @tab @code{MDADDACCS @var{a},@var{b}}
6035 @item @code{void __MDASACCS (acc, acc)}
6036 @tab @code{__MDASACCS (@var{b}, @var{a})}
6037 @tab @code{MDASACCS @var{a},@var{b}}
6038 @item @code{uw2 __MDCUTSSI (acc, const)}
6039 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6040 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6041 @item @code{uw2 __MDPACKH (uw2, uw2)}
6042 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6043 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6044 @item @code{uw2 __MDROTLI (uw2, const)}
6045 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6046 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6047 @item @code{void __MDSUBACCS (acc, acc)}
6048 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6049 @tab @code{MDSUBACCS @var{a},@var{b}}
6050 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6051 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6052 @tab @code{MDUNPACKH @var{a},@var{b}}
6053 @item @code{uw2 __MEXPDHD (uw1, const)}
6054 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6055 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6056 @item @code{uw1 __MEXPDHW (uw1, const)}
6057 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6058 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6059 @item @code{uw1 __MHDSETH (uw1, const)}
6060 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6061 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6062 @item @code{sw1 __MHDSETS (const)}
6063 @tab @code{@var{b} = __MHDSETS (@var{a})}
6064 @tab @code{MHDSETS #@var{a},@var{b}}
6065 @item @code{uw1 __MHSETHIH (uw1, const)}
6066 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6067 @tab @code{MHSETHIH #@var{a},@var{b}}
6068 @item @code{sw1 __MHSETHIS (sw1, const)}
6069 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6070 @tab @code{MHSETHIS #@var{a},@var{b}}
6071 @item @code{uw1 __MHSETLOH (uw1, const)}
6072 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6073 @tab @code{MHSETLOH #@var{a},@var{b}}
6074 @item @code{sw1 __MHSETLOS (sw1, const)}
6075 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6076 @tab @code{MHSETLOS #@var{a},@var{b}}
6077 @item @code{uw1 __MHTOB (uw2)}
6078 @tab @code{@var{b} = __MHTOB (@var{a})}
6079 @tab @code{MHTOB @var{a},@var{b}}
6080 @item @code{void __MMACHS (acc, sw1, sw1)}
6081 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6082 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6083 @item @code{void __MMACHU (acc, uw1, uw1)}
6084 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6085 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6086 @item @code{void __MMRDHS (acc, sw1, sw1)}
6087 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6088 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6089 @item @code{void __MMRDHU (acc, uw1, uw1)}
6090 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6091 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6092 @item @code{void __MMULHS (acc, sw1, sw1)}
6093 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6094 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6095 @item @code{void __MMULHU (acc, uw1, uw1)}
6096 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6097 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6098 @item @code{void __MMULXHS (acc, sw1, sw1)}
6099 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6100 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6101 @item @code{void __MMULXHU (acc, uw1, uw1)}
6102 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6103 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6104 @item @code{uw1 __MNOT (uw1)}
6105 @tab @code{@var{b} = __MNOT (@var{a})}
6106 @tab @code{MNOT @var{a},@var{b}}
6107 @item @code{uw1 __MOR (uw1, uw1)}
6108 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6109 @tab @code{MOR @var{a},@var{b},@var{c}}
6110 @item @code{uw1 __MPACKH (uh, uh)}
6111 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6112 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6113 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6114 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6115 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6116 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6117 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6118 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6119 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6120 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6121 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6122 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6123 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6124 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6125 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6126 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6127 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6128 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6129 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6130 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6131 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6132 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6133 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6134 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6135 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6136 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6137 @item @code{void __MQMACHS (acc, sw2, sw2)}
6138 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6139 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6140 @item @code{void __MQMACHU (acc, uw2, uw2)}
6141 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6142 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6143 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6144 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6145 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6146 @item @code{void __MQMULHS (acc, sw2, sw2)}
6147 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6148 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6149 @item @code{void __MQMULHU (acc, uw2, uw2)}
6150 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6151 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6152 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6153 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6154 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6155 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6156 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6157 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6158 @item @code{sw2 __MQSATHS (sw2, sw2)}
6159 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6160 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6161 @item @code{uw2 __MQSLLHI (uw2, int)}
6162 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6163 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6164 @item @code{sw2 __MQSRAHI (sw2, int)}
6165 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6166 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6167 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6168 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6169 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6170 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6171 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6172 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6173 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6174 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6175 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6176 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6177 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6178 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6179 @item @code{uw1 __MRDACC (acc)}
6180 @tab @code{@var{b} = __MRDACC (@var{a})}
6181 @tab @code{MRDACC @var{a},@var{b}}
6182 @item @code{uw1 __MRDACCG (acc)}
6183 @tab @code{@var{b} = __MRDACCG (@var{a})}
6184 @tab @code{MRDACCG @var{a},@var{b}}
6185 @item @code{uw1 __MROTLI (uw1, const)}
6186 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6187 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6188 @item @code{uw1 __MROTRI (uw1, const)}
6189 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6190 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6191 @item @code{sw1 __MSATHS (sw1, sw1)}
6192 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6193 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6194 @item @code{uw1 __MSATHU (uw1, uw1)}
6195 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6196 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6197 @item @code{uw1 __MSLLHI (uw1, const)}
6198 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6199 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6200 @item @code{sw1 __MSRAHI (sw1, const)}
6201 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6202 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6203 @item @code{uw1 __MSRLHI (uw1, const)}
6204 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6205 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6206 @item @code{void __MSUBACCS (acc, acc)}
6207 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6208 @tab @code{MSUBACCS @var{a},@var{b}}
6209 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6210 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6211 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6212 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6213 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6214 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6215 @item @code{void __MTRAP (void)}
6216 @tab @code{__MTRAP ()}
6218 @item @code{uw2 __MUNPACKH (uw1)}
6219 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6220 @tab @code{MUNPACKH @var{a},@var{b}}
6221 @item @code{uw1 __MWCUT (uw2, uw1)}
6222 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6223 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6224 @item @code{void __MWTACC (acc, uw1)}
6225 @tab @code{__MWTACC (@var{b}, @var{a})}
6226 @tab @code{MWTACC @var{a},@var{b}}
6227 @item @code{void __MWTACCG (acc, uw1)}
6228 @tab @code{__MWTACCG (@var{b}, @var{a})}
6229 @tab @code{MWTACCG @var{a},@var{b}}
6230 @item @code{uw1 __MXOR (uw1, uw1)}
6231 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6232 @tab @code{MXOR @var{a},@var{b},@var{c}}
6235 @node Other Built-in Functions
6236 @subsubsection Other Built-in Functions
6238 This section describes built-in functions that are not named after
6239 a specific FR-V instruction.
6242 @item sw2 __IACCreadll (iacc @var{reg})
6243 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6244 for future expansion and must be 0.
6246 @item sw1 __IACCreadl (iacc @var{reg})
6247 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6248 Other values of @var{reg} are rejected as invalid.
6250 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6251 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6252 is reserved for future expansion and must be 0.
6254 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6255 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6256 is 1. Other values of @var{reg} are rejected as invalid.
6258 @item void __data_prefetch0 (const void *@var{x})
6259 Use the @code{dcpl} instruction to load the contents of address @var{x}
6260 into the data cache.
6262 @item void __data_prefetch (const void *@var{x})
6263 Use the @code{nldub} instruction to load the contents of address @var{x}
6264 into the data cache. The instruction will be issued in slot I1@.
6267 @node X86 Built-in Functions
6268 @subsection X86 Built-in Functions
6270 These built-in functions are available for the i386 and x86-64 family
6271 of computers, depending on the command-line switches used.
6273 The following machine modes are available for use with MMX built-in functions
6274 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6275 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6276 vector of eight 8-bit integers. Some of the built-in functions operate on
6277 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6279 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6280 of two 32-bit floating point values.
6282 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6283 floating point values. Some instructions use a vector of four 32-bit
6284 integers, these use @code{V4SI}. Finally, some instructions operate on an
6285 entire vector register, interpreting it as a 128-bit integer, these use mode
6288 The following built-in functions are made available by @option{-mmmx}.
6289 All of them generate the machine instruction that is part of the name.
6292 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6293 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6294 v2si __builtin_ia32_paddd (v2si, v2si)
6295 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6296 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6297 v2si __builtin_ia32_psubd (v2si, v2si)
6298 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6299 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6300 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6301 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6302 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6303 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6304 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6305 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6306 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6307 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6308 di __builtin_ia32_pand (di, di)
6309 di __builtin_ia32_pandn (di,di)
6310 di __builtin_ia32_por (di, di)
6311 di __builtin_ia32_pxor (di, di)
6312 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6313 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6314 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6315 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6316 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6317 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6318 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6319 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6320 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6321 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6322 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6323 v2si __builtin_ia32_punpckldq (v2si, v2si)
6324 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6325 v4hi __builtin_ia32_packssdw (v2si, v2si)
6326 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6329 The following built-in functions are made available either with
6330 @option{-msse}, or with a combination of @option{-m3dnow} and
6331 @option{-march=athlon}. All of them generate the machine
6332 instruction that is part of the name.
6335 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6336 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6337 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6338 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6339 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6340 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6341 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6342 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6343 int __builtin_ia32_pextrw (v4hi, int)
6344 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6345 int __builtin_ia32_pmovmskb (v8qi)
6346 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6347 void __builtin_ia32_movntq (di *, di)
6348 void __builtin_ia32_sfence (void)
6351 The following built-in functions are available when @option{-msse} is used.
6352 All of them generate the machine instruction that is part of the name.
6355 int __builtin_ia32_comieq (v4sf, v4sf)
6356 int __builtin_ia32_comineq (v4sf, v4sf)
6357 int __builtin_ia32_comilt (v4sf, v4sf)
6358 int __builtin_ia32_comile (v4sf, v4sf)
6359 int __builtin_ia32_comigt (v4sf, v4sf)
6360 int __builtin_ia32_comige (v4sf, v4sf)
6361 int __builtin_ia32_ucomieq (v4sf, v4sf)
6362 int __builtin_ia32_ucomineq (v4sf, v4sf)
6363 int __builtin_ia32_ucomilt (v4sf, v4sf)
6364 int __builtin_ia32_ucomile (v4sf, v4sf)
6365 int __builtin_ia32_ucomigt (v4sf, v4sf)
6366 int __builtin_ia32_ucomige (v4sf, v4sf)
6367 v4sf __builtin_ia32_addps (v4sf, v4sf)
6368 v4sf __builtin_ia32_subps (v4sf, v4sf)
6369 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6370 v4sf __builtin_ia32_divps (v4sf, v4sf)
6371 v4sf __builtin_ia32_addss (v4sf, v4sf)
6372 v4sf __builtin_ia32_subss (v4sf, v4sf)
6373 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6374 v4sf __builtin_ia32_divss (v4sf, v4sf)
6375 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6376 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6377 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6378 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6379 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6380 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6381 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6382 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6383 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6384 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6385 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6386 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6387 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6388 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6389 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6390 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6391 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6392 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6393 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6394 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6395 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6396 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6397 v4sf __builtin_ia32_minps (v4sf, v4sf)
6398 v4sf __builtin_ia32_minss (v4sf, v4sf)
6399 v4sf __builtin_ia32_andps (v4sf, v4sf)
6400 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6401 v4sf __builtin_ia32_orps (v4sf, v4sf)
6402 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6403 v4sf __builtin_ia32_movss (v4sf, v4sf)
6404 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6405 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6406 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6407 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6408 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6409 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6410 v2si __builtin_ia32_cvtps2pi (v4sf)
6411 int __builtin_ia32_cvtss2si (v4sf)
6412 v2si __builtin_ia32_cvttps2pi (v4sf)
6413 int __builtin_ia32_cvttss2si (v4sf)
6414 v4sf __builtin_ia32_rcpps (v4sf)
6415 v4sf __builtin_ia32_rsqrtps (v4sf)
6416 v4sf __builtin_ia32_sqrtps (v4sf)
6417 v4sf __builtin_ia32_rcpss (v4sf)
6418 v4sf __builtin_ia32_rsqrtss (v4sf)
6419 v4sf __builtin_ia32_sqrtss (v4sf)
6420 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6421 void __builtin_ia32_movntps (float *, v4sf)
6422 int __builtin_ia32_movmskps (v4sf)
6425 The following built-in functions are available when @option{-msse} is used.
6428 @item v4sf __builtin_ia32_loadaps (float *)
6429 Generates the @code{movaps} machine instruction as a load from memory.
6430 @item void __builtin_ia32_storeaps (float *, v4sf)
6431 Generates the @code{movaps} machine instruction as a store to memory.
6432 @item v4sf __builtin_ia32_loadups (float *)
6433 Generates the @code{movups} machine instruction as a load from memory.
6434 @item void __builtin_ia32_storeups (float *, v4sf)
6435 Generates the @code{movups} machine instruction as a store to memory.
6436 @item v4sf __builtin_ia32_loadsss (float *)
6437 Generates the @code{movss} machine instruction as a load from memory.
6438 @item void __builtin_ia32_storess (float *, v4sf)
6439 Generates the @code{movss} machine instruction as a store to memory.
6440 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6441 Generates the @code{movhps} machine instruction as a load from memory.
6442 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6443 Generates the @code{movlps} machine instruction as a load from memory
6444 @item void __builtin_ia32_storehps (v4sf, v2si *)
6445 Generates the @code{movhps} machine instruction as a store to memory.
6446 @item void __builtin_ia32_storelps (v4sf, v2si *)
6447 Generates the @code{movlps} machine instruction as a store to memory.
6450 The following built-in functions are available when @option{-msse3} is used.
6451 All of them generate the machine instruction that is part of the name.
6454 v2df __builtin_ia32_addsubpd (v2df, v2df)
6455 v2df __builtin_ia32_addsubps (v2df, v2df)
6456 v2df __builtin_ia32_haddpd (v2df, v2df)
6457 v2df __builtin_ia32_haddps (v2df, v2df)
6458 v2df __builtin_ia32_hsubpd (v2df, v2df)
6459 v2df __builtin_ia32_hsubps (v2df, v2df)
6460 v16qi __builtin_ia32_lddqu (char const *)
6461 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6462 v2df __builtin_ia32_movddup (v2df)
6463 v4sf __builtin_ia32_movshdup (v4sf)
6464 v4sf __builtin_ia32_movsldup (v4sf)
6465 void __builtin_ia32_mwait (unsigned int, unsigned int)
6468 The following built-in functions are available when @option{-msse3} is used.
6471 @item v2df __builtin_ia32_loadddup (double const *)
6472 Generates the @code{movddup} machine instruction as a load from memory.
6475 The following built-in functions are available when @option{-m3dnow} is used.
6476 All of them generate the machine instruction that is part of the name.
6479 void __builtin_ia32_femms (void)
6480 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6481 v2si __builtin_ia32_pf2id (v2sf)
6482 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6483 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6484 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6485 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6486 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6487 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6488 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6489 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6490 v2sf __builtin_ia32_pfrcp (v2sf)
6491 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6492 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6493 v2sf __builtin_ia32_pfrsqrt (v2sf)
6494 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6495 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6496 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6497 v2sf __builtin_ia32_pi2fd (v2si)
6498 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6501 The following built-in functions are available when both @option{-m3dnow}
6502 and @option{-march=athlon} are used. All of them generate the machine
6503 instruction that is part of the name.
6506 v2si __builtin_ia32_pf2iw (v2sf)
6507 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6508 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6509 v2sf __builtin_ia32_pi2fw (v2si)
6510 v2sf __builtin_ia32_pswapdsf (v2sf)
6511 v2si __builtin_ia32_pswapdsi (v2si)
6514 @node MIPS Paired-Single Support
6515 @subsection MIPS Paired-Single Support
6517 The MIPS64 architecture includes a number of instructions that
6518 operate on pairs of single-precision floating-point values.
6519 Each pair is packed into a 64-bit floating-point register,
6520 with one element being designated the ``upper half'' and
6521 the other being designated the ``lower half''.
6523 GCC supports paired-single operations using both the generic
6524 vector extensions (@pxref{Vector Extensions}) and a collection of
6525 MIPS-specific built-in functions. Both kinds of support are
6526 enabled by the @option{-mpaired-single} command-line option.
6528 The vector type associated with paired-single values is usually
6529 called @code{v2sf}. It can be defined in C as follows:
6532 typedef float v2sf __attribute__ ((vector_size (8)));
6535 @code{v2sf} values are initialized in the same way as aggregates.
6539 v2sf a = @{1.5, 9.1@};
6542 b = (v2sf) @{e, f@};
6545 @emph{Note:} The CPU's endianness determines which value is stored in
6546 the upper half of a register and which value is stored in the lower half.
6547 On little-endian targets, the first value is the lower one and the second
6548 value is the upper one. The opposite order applies to big-endian targets.
6549 For example, the code above will set the lower half of @code{a} to
6550 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
6553 * Paired-Single Arithmetic::
6554 * Paired-Single Built-in Functions::
6555 * MIPS-3D Built-in Functions::
6558 @node Paired-Single Arithmetic
6559 @subsubsection Paired-Single Arithmetic
6561 The table below lists the @code{v2sf} operations for which hardware
6562 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
6563 values and @code{x} is an integral value.
6565 @multitable @columnfractions .50 .50
6566 @item C code @tab MIPS instruction
6567 @item @code{a + b} @tab @code{add.ps}
6568 @item @code{a - b} @tab @code{sub.ps}
6569 @item @code{-a} @tab @code{neg.ps}
6570 @item @code{a * b} @tab @code{mul.ps}
6571 @item @code{a * b + c} @tab @code{madd.ps}
6572 @item @code{a * b - c} @tab @code{msub.ps}
6573 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
6574 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
6575 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
6578 Note that the multiply-accumulate instructions can be disabled
6579 using the command-line option @code{-mno-fused-madd}.
6581 @node Paired-Single Built-in Functions
6582 @subsubsection Paired-Single Built-in Functions
6584 The following paired-single functions map directly to a particular
6585 MIPS instruction. Please refer to the architecture specification
6586 for details on what each instruction does.
6589 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
6590 Pair lower lower (@code{pll.ps}).
6592 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
6593 Pair upper lower (@code{pul.ps}).
6595 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
6596 Pair lower upper (@code{plu.ps}).
6598 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
6599 Pair upper upper (@code{puu.ps}).
6601 @item v2sf __builtin_mips_cvt_ps_s (float, float)
6602 Convert pair to paired single (@code{cvt.ps.s}).
6604 @item float __builtin_mips_cvt_s_pl (v2sf)
6605 Convert pair lower to single (@code{cvt.s.pl}).
6607 @item float __builtin_mips_cvt_s_pu (v2sf)
6608 Convert pair upper to single (@code{cvt.s.pu}).
6610 @item v2sf __builtin_mips_abs_ps (v2sf)
6611 Absolute value (@code{abs.ps}).
6613 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
6614 Align variable (@code{alnv.ps}).
6616 @emph{Note:} The value of the third parameter must be 0 or 4
6617 modulo 8, otherwise the result will be unpredictable. Please read the
6618 instruction description for details.
6621 The following multi-instruction functions are also available.
6622 In each case, @var{cond} can be any of the 16 floating-point conditions:
6623 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
6624 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
6625 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
6628 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6629 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6630 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
6631 @code{movt.ps}/@code{movf.ps}).
6633 The @code{movt} functions return the value @var{x} computed by:
6636 c.@var{cond}.ps @var{cc},@var{a},@var{b}
6637 mov.ps @var{x},@var{c}
6638 movt.ps @var{x},@var{d},@var{cc}
6641 The @code{movf} functions are similar but use @code{movf.ps} instead
6644 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6645 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6646 Comparison of two paired-single values (@code{c.@var{cond}.ps},
6647 @code{bc1t}/@code{bc1f}).
6649 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
6650 and return either the upper or lower half of the result. For example:
6654 if (__builtin_mips_upper_c_eq_ps (a, b))
6655 upper_halves_are_equal ();
6657 upper_halves_are_unequal ();
6659 if (__builtin_mips_lower_c_eq_ps (a, b))
6660 lower_halves_are_equal ();
6662 lower_halves_are_unequal ();
6666 @node MIPS-3D Built-in Functions
6667 @subsubsection MIPS-3D Built-in Functions
6669 The MIPS-3D Application-Specific Extension (ASE) includes additional
6670 paired-single instructions that are designed to improve the performance
6671 of 3D graphics operations. Support for these instructions is controlled
6672 by the @option{-mips3d} command-line option.
6674 The functions listed below map directly to a particular MIPS-3D
6675 instruction. Please refer to the architecture specification for
6676 more details on what each instruction does.
6679 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
6680 Reduction add (@code{addr.ps}).
6682 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
6683 Reduction multiply (@code{mulr.ps}).
6685 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
6686 Convert paired single to paired word (@code{cvt.pw.ps}).
6688 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
6689 Convert paired word to paired single (@code{cvt.ps.pw}).
6691 @item float __builtin_mips_recip1_s (float)
6692 @itemx double __builtin_mips_recip1_d (double)
6693 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
6694 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
6696 @item float __builtin_mips_recip2_s (float, float)
6697 @itemx double __builtin_mips_recip2_d (double, double)
6698 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
6699 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
6701 @item float __builtin_mips_rsqrt1_s (float)
6702 @itemx double __builtin_mips_rsqrt1_d (double)
6703 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
6704 Reduced precision reciprocal square root (sequence step 1)
6705 (@code{rsqrt1.@var{fmt}}).
6707 @item float __builtin_mips_rsqrt2_s (float, float)
6708 @itemx double __builtin_mips_rsqrt2_d (double, double)
6709 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
6710 Reduced precision reciprocal square root (sequence step 2)
6711 (@code{rsqrt2.@var{fmt}}).
6714 The following multi-instruction functions are also available.
6715 In each case, @var{cond} can be any of the 16 floating-point conditions:
6716 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
6717 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
6718 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
6721 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
6722 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
6723 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
6724 @code{bc1t}/@code{bc1f}).
6726 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
6727 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
6732 if (__builtin_mips_cabs_eq_s (a, b))
6738 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6739 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6740 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
6741 @code{bc1t}/@code{bc1f}).
6743 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
6744 and return either the upper or lower half of the result. For example:
6748 if (__builtin_mips_upper_cabs_eq_ps (a, b))
6749 upper_halves_are_equal ();
6751 upper_halves_are_unequal ();
6753 if (__builtin_mips_lower_cabs_eq_ps (a, b))
6754 lower_halves_are_equal ();
6756 lower_halves_are_unequal ();
6759 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6760 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6761 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
6762 @code{movt.ps}/@code{movf.ps}).
6764 The @code{movt} functions return the value @var{x} computed by:
6767 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
6768 mov.ps @var{x},@var{c}
6769 movt.ps @var{x},@var{d},@var{cc}
6772 The @code{movf} functions are similar but use @code{movf.ps} instead
6775 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6776 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6777 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6778 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6779 Comparison of two paired-single values
6780 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
6781 @code{bc1any2t}/@code{bc1any2f}).
6783 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
6784 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
6785 result is true and the @code{all} forms return true if both results are true.
6790 if (__builtin_mips_any_c_eq_ps (a, b))
6795 if (__builtin_mips_all_c_eq_ps (a, b))
6801 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6802 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6803 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6804 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6805 Comparison of four paired-single values
6806 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
6807 @code{bc1any4t}/@code{bc1any4f}).
6809 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
6810 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
6811 The @code{any} forms return true if any of the four results are true
6812 and the @code{all} forms return true if all four results are true.
6817 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
6822 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
6829 @node PowerPC AltiVec Built-in Functions
6830 @subsection PowerPC AltiVec Built-in Functions
6832 GCC provides an interface for the PowerPC family of processors to access
6833 the AltiVec operations described in Motorola's AltiVec Programming
6834 Interface Manual. The interface is made available by including
6835 @code{<altivec.h>} and using @option{-maltivec} and
6836 @option{-mabi=altivec}. The interface supports the following vector
6840 vector unsigned char
6844 vector unsigned short
6855 GCC's implementation of the high-level language interface available from
6856 C and C++ code differs from Motorola's documentation in several ways.
6861 A vector constant is a list of constant expressions within curly braces.
6864 A vector initializer requires no cast if the vector constant is of the
6865 same type as the variable it is initializing.
6868 If @code{signed} or @code{unsigned} is omitted, the signedness of the
6869 vector type is the default signedness of the base type. The default
6870 varies depending on the operating system, so a portable program should
6871 always specify the signedness.
6874 Compiling with @option{-maltivec} adds keywords @code{__vector},
6875 @code{__pixel}, and @code{__bool}. Macros @option{vector},
6876 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
6880 GCC allows using a @code{typedef} name as the type specifier for a
6884 For C, overloaded functions are implemented with macros so the following
6888 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6891 Since @code{vec_add} is a macro, the vector constant in the example
6892 is treated as four separate arguments. Wrap the entire argument in
6893 parentheses for this to work.
6896 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
6897 Internally, GCC uses built-in functions to achieve the functionality in
6898 the aforementioned header file, but they are not supported and are
6899 subject to change without notice.
6901 The following interfaces are supported for the generic and specific
6902 AltiVec operations and the AltiVec predicates. In cases where there
6903 is a direct mapping between generic and specific operations, only the
6904 generic names are shown here, although the specific operations can also
6907 Arguments that are documented as @code{const int} require literal
6908 integral values within the range required for that operation.
6911 vector signed char vec_abs (vector signed char);
6912 vector signed short vec_abs (vector signed short);
6913 vector signed int vec_abs (vector signed int);
6914 vector float vec_abs (vector float);
6916 vector signed char vec_abss (vector signed char);
6917 vector signed short vec_abss (vector signed short);
6918 vector signed int vec_abss (vector signed int);
6920 vector signed char vec_add (vector bool char, vector signed char);
6921 vector signed char vec_add (vector signed char, vector bool char);
6922 vector signed char vec_add (vector signed char, vector signed char);
6923 vector unsigned char vec_add (vector bool char, vector unsigned char);
6924 vector unsigned char vec_add (vector unsigned char, vector bool char);
6925 vector unsigned char vec_add (vector unsigned char,
6926 vector unsigned char);
6927 vector signed short vec_add (vector bool short, vector signed short);
6928 vector signed short vec_add (vector signed short, vector bool short);
6929 vector signed short vec_add (vector signed short, vector signed short);
6930 vector unsigned short vec_add (vector bool short,
6931 vector unsigned short);
6932 vector unsigned short vec_add (vector unsigned short,
6934 vector unsigned short vec_add (vector unsigned short,
6935 vector unsigned short);
6936 vector signed int vec_add (vector bool int, vector signed int);
6937 vector signed int vec_add (vector signed int, vector bool int);
6938 vector signed int vec_add (vector signed int, vector signed int);
6939 vector unsigned int vec_add (vector bool int, vector unsigned int);
6940 vector unsigned int vec_add (vector unsigned int, vector bool int);
6941 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6942 vector float vec_add (vector float, vector float);
6944 vector float vec_vaddfp (vector float, vector float);
6946 vector signed int vec_vadduwm (vector bool int, vector signed int);
6947 vector signed int vec_vadduwm (vector signed int, vector bool int);
6948 vector signed int vec_vadduwm (vector signed int, vector signed int);
6949 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
6950 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
6951 vector unsigned int vec_vadduwm (vector unsigned int,
6952 vector unsigned int);
6954 vector signed short vec_vadduhm (vector bool short,
6955 vector signed short);
6956 vector signed short vec_vadduhm (vector signed short,
6958 vector signed short vec_vadduhm (vector signed short,
6959 vector signed short);
6960 vector unsigned short vec_vadduhm (vector bool short,
6961 vector unsigned short);
6962 vector unsigned short vec_vadduhm (vector unsigned short,
6964 vector unsigned short vec_vadduhm (vector unsigned short,
6965 vector unsigned short);
6967 vector signed char vec_vaddubm (vector bool char, vector signed char);
6968 vector signed char vec_vaddubm (vector signed char, vector bool char);
6969 vector signed char vec_vaddubm (vector signed char, vector signed char);
6970 vector unsigned char vec_vaddubm (vector bool char,
6971 vector unsigned char);
6972 vector unsigned char vec_vaddubm (vector unsigned char,
6974 vector unsigned char vec_vaddubm (vector unsigned char,
6975 vector unsigned char);
6977 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6979 vector unsigned char vec_adds (vector bool char, vector unsigned char);
6980 vector unsigned char vec_adds (vector unsigned char, vector bool char);
6981 vector unsigned char vec_adds (vector unsigned char,
6982 vector unsigned char);
6983 vector signed char vec_adds (vector bool char, vector signed char);
6984 vector signed char vec_adds (vector signed char, vector bool char);
6985 vector signed char vec_adds (vector signed char, vector signed char);
6986 vector unsigned short vec_adds (vector bool short,
6987 vector unsigned short);
6988 vector unsigned short vec_adds (vector unsigned short,
6990 vector unsigned short vec_adds (vector unsigned short,
6991 vector unsigned short);
6992 vector signed short vec_adds (vector bool short, vector signed short);
6993 vector signed short vec_adds (vector signed short, vector bool short);
6994 vector signed short vec_adds (vector signed short, vector signed short);
6995 vector unsigned int vec_adds (vector bool int, vector unsigned int);
6996 vector unsigned int vec_adds (vector unsigned int, vector bool int);
6997 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6998 vector signed int vec_adds (vector bool int, vector signed int);
6999 vector signed int vec_adds (vector signed int, vector bool int);
7000 vector signed int vec_adds (vector signed int, vector signed int);
7002 vector signed int vec_vaddsws (vector bool int, vector signed int);
7003 vector signed int vec_vaddsws (vector signed int, vector bool int);
7004 vector signed int vec_vaddsws (vector signed int, vector signed int);
7006 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
7007 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
7008 vector unsigned int vec_vadduws (vector unsigned int,
7009 vector unsigned int);
7011 vector signed short vec_vaddshs (vector bool short,
7012 vector signed short);
7013 vector signed short vec_vaddshs (vector signed short,
7015 vector signed short vec_vaddshs (vector signed short,
7016 vector signed short);
7018 vector unsigned short vec_vadduhs (vector bool short,
7019 vector unsigned short);
7020 vector unsigned short vec_vadduhs (vector unsigned short,
7022 vector unsigned short vec_vadduhs (vector unsigned short,
7023 vector unsigned short);
7025 vector signed char vec_vaddsbs (vector bool char, vector signed char);
7026 vector signed char vec_vaddsbs (vector signed char, vector bool char);
7027 vector signed char vec_vaddsbs (vector signed char, vector signed char);
7029 vector unsigned char vec_vaddubs (vector bool char,
7030 vector unsigned char);
7031 vector unsigned char vec_vaddubs (vector unsigned char,
7033 vector unsigned char vec_vaddubs (vector unsigned char,
7034 vector unsigned char);
7036 vector float vec_and (vector float, vector float);
7037 vector float vec_and (vector float, vector bool int);
7038 vector float vec_and (vector bool int, vector float);
7039 vector bool int vec_and (vector bool int, vector bool int);
7040 vector signed int vec_and (vector bool int, vector signed int);
7041 vector signed int vec_and (vector signed int, vector bool int);
7042 vector signed int vec_and (vector signed int, vector signed int);
7043 vector unsigned int vec_and (vector bool int, vector unsigned int);
7044 vector unsigned int vec_and (vector unsigned int, vector bool int);
7045 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
7046 vector bool short vec_and (vector bool short, vector bool short);
7047 vector signed short vec_and (vector bool short, vector signed short);
7048 vector signed short vec_and (vector signed short, vector bool short);
7049 vector signed short vec_and (vector signed short, vector signed short);
7050 vector unsigned short vec_and (vector bool short,
7051 vector unsigned short);
7052 vector unsigned short vec_and (vector unsigned short,
7054 vector unsigned short vec_and (vector unsigned short,
7055 vector unsigned short);
7056 vector signed char vec_and (vector bool char, vector signed char);
7057 vector bool char vec_and (vector bool char, vector bool char);
7058 vector signed char vec_and (vector signed char, vector bool char);
7059 vector signed char vec_and (vector signed char, vector signed char);
7060 vector unsigned char vec_and (vector bool char, vector unsigned char);
7061 vector unsigned char vec_and (vector unsigned char, vector bool char);
7062 vector unsigned char vec_and (vector unsigned char,
7063 vector unsigned char);
7065 vector float vec_andc (vector float, vector float);
7066 vector float vec_andc (vector float, vector bool int);
7067 vector float vec_andc (vector bool int, vector float);
7068 vector bool int vec_andc (vector bool int, vector bool int);
7069 vector signed int vec_andc (vector bool int, vector signed int);
7070 vector signed int vec_andc (vector signed int, vector bool int);
7071 vector signed int vec_andc (vector signed int, vector signed int);
7072 vector unsigned int vec_andc (vector bool int, vector unsigned int);
7073 vector unsigned int vec_andc (vector unsigned int, vector bool int);
7074 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
7075 vector bool short vec_andc (vector bool short, vector bool short);
7076 vector signed short vec_andc (vector bool short, vector signed short);
7077 vector signed short vec_andc (vector signed short, vector bool short);
7078 vector signed short vec_andc (vector signed short, vector signed short);
7079 vector unsigned short vec_andc (vector bool short,
7080 vector unsigned short);
7081 vector unsigned short vec_andc (vector unsigned short,
7083 vector unsigned short vec_andc (vector unsigned short,
7084 vector unsigned short);
7085 vector signed char vec_andc (vector bool char, vector signed char);
7086 vector bool char vec_andc (vector bool char, vector bool char);
7087 vector signed char vec_andc (vector signed char, vector bool char);
7088 vector signed char vec_andc (vector signed char, vector signed char);
7089 vector unsigned char vec_andc (vector bool char, vector unsigned char);
7090 vector unsigned char vec_andc (vector unsigned char, vector bool char);
7091 vector unsigned char vec_andc (vector unsigned char,
7092 vector unsigned char);
7094 vector unsigned char vec_avg (vector unsigned char,
7095 vector unsigned char);
7096 vector signed char vec_avg (vector signed char, vector signed char);
7097 vector unsigned short vec_avg (vector unsigned short,
7098 vector unsigned short);
7099 vector signed short vec_avg (vector signed short, vector signed short);
7100 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
7101 vector signed int vec_avg (vector signed int, vector signed int);
7103 vector signed int vec_vavgsw (vector signed int, vector signed int);
7105 vector unsigned int vec_vavguw (vector unsigned int,
7106 vector unsigned int);
7108 vector signed short vec_vavgsh (vector signed short,
7109 vector signed short);
7111 vector unsigned short vec_vavguh (vector unsigned short,
7112 vector unsigned short);
7114 vector signed char vec_vavgsb (vector signed char, vector signed char);
7116 vector unsigned char vec_vavgub (vector unsigned char,
7117 vector unsigned char);
7119 vector float vec_ceil (vector float);
7121 vector signed int vec_cmpb (vector float, vector float);
7123 vector bool char vec_cmpeq (vector signed char, vector signed char);
7124 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
7125 vector bool short vec_cmpeq (vector signed short, vector signed short);
7126 vector bool short vec_cmpeq (vector unsigned short,
7127 vector unsigned short);
7128 vector bool int vec_cmpeq (vector signed int, vector signed int);
7129 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
7130 vector bool int vec_cmpeq (vector float, vector float);
7132 vector bool int vec_vcmpeqfp (vector float, vector float);
7134 vector bool int vec_vcmpequw (vector signed int, vector signed int);
7135 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
7137 vector bool short vec_vcmpequh (vector signed short,
7138 vector signed short);
7139 vector bool short vec_vcmpequh (vector unsigned short,
7140 vector unsigned short);
7142 vector bool char vec_vcmpequb (vector signed char, vector signed char);
7143 vector bool char vec_vcmpequb (vector unsigned char,
7144 vector unsigned char);
7146 vector bool int vec_cmpge (vector float, vector float);
7148 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
7149 vector bool char vec_cmpgt (vector signed char, vector signed char);
7150 vector bool short vec_cmpgt (vector unsigned short,
7151 vector unsigned short);
7152 vector bool short vec_cmpgt (vector signed short, vector signed short);
7153 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
7154 vector bool int vec_cmpgt (vector signed int, vector signed int);
7155 vector bool int vec_cmpgt (vector float, vector float);
7157 vector bool int vec_vcmpgtfp (vector float, vector float);
7159 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
7161 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
7163 vector bool short vec_vcmpgtsh (vector signed short,
7164 vector signed short);
7166 vector bool short vec_vcmpgtuh (vector unsigned short,
7167 vector unsigned short);
7169 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
7171 vector bool char vec_vcmpgtub (vector unsigned char,
7172 vector unsigned char);
7174 vector bool int vec_cmple (vector float, vector float);
7176 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
7177 vector bool char vec_cmplt (vector signed char, vector signed char);
7178 vector bool short vec_cmplt (vector unsigned short,
7179 vector unsigned short);
7180 vector bool short vec_cmplt (vector signed short, vector signed short);
7181 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
7182 vector bool int vec_cmplt (vector signed int, vector signed int);
7183 vector bool int vec_cmplt (vector float, vector float);
7185 vector float vec_ctf (vector unsigned int, const int);
7186 vector float vec_ctf (vector signed int, const int);
7188 vector float vec_vcfsx (vector signed int, const int);
7190 vector float vec_vcfux (vector unsigned int, const int);
7192 vector signed int vec_cts (vector float, const int);
7194 vector unsigned int vec_ctu (vector float, const int);
7196 void vec_dss (const int);
7198 void vec_dssall (void);
7200 void vec_dst (const vector unsigned char *, int, const int);
7201 void vec_dst (const vector signed char *, int, const int);
7202 void vec_dst (const vector bool char *, int, const int);
7203 void vec_dst (const vector unsigned short *, int, const int);
7204 void vec_dst (const vector signed short *, int, const int);
7205 void vec_dst (const vector bool short *, int, const int);
7206 void vec_dst (const vector pixel *, int, const int);
7207 void vec_dst (const vector unsigned int *, int, const int);
7208 void vec_dst (const vector signed int *, int, const int);
7209 void vec_dst (const vector bool int *, int, const int);
7210 void vec_dst (const vector float *, int, const int);
7211 void vec_dst (const unsigned char *, int, const int);
7212 void vec_dst (const signed char *, int, const int);
7213 void vec_dst (const unsigned short *, int, const int);
7214 void vec_dst (const short *, int, const int);
7215 void vec_dst (const unsigned int *, int, const int);
7216 void vec_dst (const int *, int, const int);
7217 void vec_dst (const unsigned long *, int, const int);
7218 void vec_dst (const long *, int, const int);
7219 void vec_dst (const float *, int, const int);
7221 void vec_dstst (const vector unsigned char *, int, const int);
7222 void vec_dstst (const vector signed char *, int, const int);
7223 void vec_dstst (const vector bool char *, int, const int);
7224 void vec_dstst (const vector unsigned short *, int, const int);
7225 void vec_dstst (const vector signed short *, int, const int);
7226 void vec_dstst (const vector bool short *, int, const int);
7227 void vec_dstst (const vector pixel *, int, const int);
7228 void vec_dstst (const vector unsigned int *, int, const int);
7229 void vec_dstst (const vector signed int *, int, const int);
7230 void vec_dstst (const vector bool int *, int, const int);
7231 void vec_dstst (const vector float *, int, const int);
7232 void vec_dstst (const unsigned char *, int, const int);
7233 void vec_dstst (const signed char *, int, const int);
7234 void vec_dstst (const unsigned short *, int, const int);
7235 void vec_dstst (const short *, int, const int);
7236 void vec_dstst (const unsigned int *, int, const int);
7237 void vec_dstst (const int *, int, const int);
7238 void vec_dstst (const unsigned long *, int, const int);
7239 void vec_dstst (const long *, int, const int);
7240 void vec_dstst (const float *, int, const int);
7242 void vec_dststt (const vector unsigned char *, int, const int);
7243 void vec_dststt (const vector signed char *, int, const int);
7244 void vec_dststt (const vector bool char *, int, const int);
7245 void vec_dststt (const vector unsigned short *, int, const int);
7246 void vec_dststt (const vector signed short *, int, const int);
7247 void vec_dststt (const vector bool short *, int, const int);
7248 void vec_dststt (const vector pixel *, int, const int);
7249 void vec_dststt (const vector unsigned int *, int, const int);
7250 void vec_dststt (const vector signed int *, int, const int);
7251 void vec_dststt (const vector bool int *, int, const int);
7252 void vec_dststt (const vector float *, int, const int);
7253 void vec_dststt (const unsigned char *, int, const int);
7254 void vec_dststt (const signed char *, int, const int);
7255 void vec_dststt (const unsigned short *, int, const int);
7256 void vec_dststt (const short *, int, const int);
7257 void vec_dststt (const unsigned int *, int, const int);
7258 void vec_dststt (const int *, int, const int);
7259 void vec_dststt (const unsigned long *, int, const int);
7260 void vec_dststt (const long *, int, const int);
7261 void vec_dststt (const float *, int, const int);
7263 void vec_dstt (const vector unsigned char *, int, const int);
7264 void vec_dstt (const vector signed char *, int, const int);
7265 void vec_dstt (const vector bool char *, int, const int);
7266 void vec_dstt (const vector unsigned short *, int, const int);
7267 void vec_dstt (const vector signed short *, int, const int);
7268 void vec_dstt (const vector bool short *, int, const int);
7269 void vec_dstt (const vector pixel *, int, const int);
7270 void vec_dstt (const vector unsigned int *, int, const int);
7271 void vec_dstt (const vector signed int *, int, const int);
7272 void vec_dstt (const vector bool int *, int, const int);
7273 void vec_dstt (const vector float *, int, const int);
7274 void vec_dstt (const unsigned char *, int, const int);
7275 void vec_dstt (const signed char *, int, const int);
7276 void vec_dstt (const unsigned short *, int, const int);
7277 void vec_dstt (const short *, int, const int);
7278 void vec_dstt (const unsigned int *, int, const int);
7279 void vec_dstt (const int *, int, const int);
7280 void vec_dstt (const unsigned long *, int, const int);
7281 void vec_dstt (const long *, int, const int);
7282 void vec_dstt (const float *, int, const int);
7284 vector float vec_expte (vector float);
7286 vector float vec_floor (vector float);
7288 vector float vec_ld (int, const vector float *);
7289 vector float vec_ld (int, const float *);
7290 vector bool int vec_ld (int, const vector bool int *);
7291 vector signed int vec_ld (int, const vector signed int *);
7292 vector signed int vec_ld (int, const int *);
7293 vector signed int vec_ld (int, const long *);
7294 vector unsigned int vec_ld (int, const vector unsigned int *);
7295 vector unsigned int vec_ld (int, const unsigned int *);
7296 vector unsigned int vec_ld (int, const unsigned long *);
7297 vector bool short vec_ld (int, const vector bool short *);
7298 vector pixel vec_ld (int, const vector pixel *);
7299 vector signed short vec_ld (int, const vector signed short *);
7300 vector signed short vec_ld (int, const short *);
7301 vector unsigned short vec_ld (int, const vector unsigned short *);
7302 vector unsigned short vec_ld (int, const unsigned short *);
7303 vector bool char vec_ld (int, const vector bool char *);
7304 vector signed char vec_ld (int, const vector signed char *);
7305 vector signed char vec_ld (int, const signed char *);
7306 vector unsigned char vec_ld (int, const vector unsigned char *);
7307 vector unsigned char vec_ld (int, const unsigned char *);
7309 vector signed char vec_lde (int, const signed char *);
7310 vector unsigned char vec_lde (int, const unsigned char *);
7311 vector signed short vec_lde (int, const short *);
7312 vector unsigned short vec_lde (int, const unsigned short *);
7313 vector float vec_lde (int, const float *);
7314 vector signed int vec_lde (int, const int *);
7315 vector unsigned int vec_lde (int, const unsigned int *);
7316 vector signed int vec_lde (int, const long *);
7317 vector unsigned int vec_lde (int, const unsigned long *);
7319 vector float vec_lvewx (int, float *);
7320 vector signed int vec_lvewx (int, int *);
7321 vector unsigned int vec_lvewx (int, unsigned int *);
7322 vector signed int vec_lvewx (int, long *);
7323 vector unsigned int vec_lvewx (int, unsigned long *);
7325 vector signed short vec_lvehx (int, short *);
7326 vector unsigned short vec_lvehx (int, unsigned short *);
7328 vector signed char vec_lvebx (int, char *);
7329 vector unsigned char vec_lvebx (int, unsigned char *);
7331 vector float vec_ldl (int, const vector float *);
7332 vector float vec_ldl (int, const float *);
7333 vector bool int vec_ldl (int, const vector bool int *);
7334 vector signed int vec_ldl (int, const vector signed int *);
7335 vector signed int vec_ldl (int, const int *);
7336 vector signed int vec_ldl (int, const long *);
7337 vector unsigned int vec_ldl (int, const vector unsigned int *);
7338 vector unsigned int vec_ldl (int, const unsigned int *);
7339 vector unsigned int vec_ldl (int, const unsigned long *);
7340 vector bool short vec_ldl (int, const vector bool short *);
7341 vector pixel vec_ldl (int, const vector pixel *);
7342 vector signed short vec_ldl (int, const vector signed short *);
7343 vector signed short vec_ldl (int, const short *);
7344 vector unsigned short vec_ldl (int, const vector unsigned short *);
7345 vector unsigned short vec_ldl (int, const unsigned short *);
7346 vector bool char vec_ldl (int, const vector bool char *);
7347 vector signed char vec_ldl (int, const vector signed char *);
7348 vector signed char vec_ldl (int, const signed char *);
7349 vector unsigned char vec_ldl (int, const vector unsigned char *);
7350 vector unsigned char vec_ldl (int, const unsigned char *);
7352 vector float vec_loge (vector float);
7354 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
7355 vector unsigned char vec_lvsl (int, const volatile signed char *);
7356 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
7357 vector unsigned char vec_lvsl (int, const volatile short *);
7358 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
7359 vector unsigned char vec_lvsl (int, const volatile int *);
7360 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
7361 vector unsigned char vec_lvsl (int, const volatile long *);
7362 vector unsigned char vec_lvsl (int, const volatile float *);
7364 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
7365 vector unsigned char vec_lvsr (int, const volatile signed char *);
7366 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
7367 vector unsigned char vec_lvsr (int, const volatile short *);
7368 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
7369 vector unsigned char vec_lvsr (int, const volatile int *);
7370 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
7371 vector unsigned char vec_lvsr (int, const volatile long *);
7372 vector unsigned char vec_lvsr (int, const volatile float *);
7374 vector float vec_madd (vector float, vector float, vector float);
7376 vector signed short vec_madds (vector signed short,
7377 vector signed short,
7378 vector signed short);
7380 vector unsigned char vec_max (vector bool char, vector unsigned char);
7381 vector unsigned char vec_max (vector unsigned char, vector bool char);
7382 vector unsigned char vec_max (vector unsigned char,
7383 vector unsigned char);
7384 vector signed char vec_max (vector bool char, vector signed char);
7385 vector signed char vec_max (vector signed char, vector bool char);
7386 vector signed char vec_max (vector signed char, vector signed char);
7387 vector unsigned short vec_max (vector bool short,
7388 vector unsigned short);
7389 vector unsigned short vec_max (vector unsigned short,
7391 vector unsigned short vec_max (vector unsigned short,
7392 vector unsigned short);
7393 vector signed short vec_max (vector bool short, vector signed short);
7394 vector signed short vec_max (vector signed short, vector bool short);
7395 vector signed short vec_max (vector signed short, vector signed short);
7396 vector unsigned int vec_max (vector bool int, vector unsigned int);
7397 vector unsigned int vec_max (vector unsigned int, vector bool int);
7398 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
7399 vector signed int vec_max (vector bool int, vector signed int);
7400 vector signed int vec_max (vector signed int, vector bool int);
7401 vector signed int vec_max (vector signed int, vector signed int);
7402 vector float vec_max (vector float, vector float);
7404 vector float vec_vmaxfp (vector float, vector float);
7406 vector signed int vec_vmaxsw (vector bool int, vector signed int);
7407 vector signed int vec_vmaxsw (vector signed int, vector bool int);
7408 vector signed int vec_vmaxsw (vector signed int, vector signed int);
7410 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
7411 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
7412 vector unsigned int vec_vmaxuw (vector unsigned int,
7413 vector unsigned int);
7415 vector signed short vec_vmaxsh (vector bool short, vector signed short);
7416 vector signed short vec_vmaxsh (vector signed short, vector bool short);
7417 vector signed short vec_vmaxsh (vector signed short,
7418 vector signed short);
7420 vector unsigned short vec_vmaxuh (vector bool short,
7421 vector unsigned short);
7422 vector unsigned short vec_vmaxuh (vector unsigned short,
7424 vector unsigned short vec_vmaxuh (vector unsigned short,
7425 vector unsigned short);
7427 vector signed char vec_vmaxsb (vector bool char, vector signed char);
7428 vector signed char vec_vmaxsb (vector signed char, vector bool char);
7429 vector signed char vec_vmaxsb (vector signed char, vector signed char);
7431 vector unsigned char vec_vmaxub (vector bool char,
7432 vector unsigned char);
7433 vector unsigned char vec_vmaxub (vector unsigned char,
7435 vector unsigned char vec_vmaxub (vector unsigned char,
7436 vector unsigned char);
7438 vector bool char vec_mergeh (vector bool char, vector bool char);
7439 vector signed char vec_mergeh (vector signed char, vector signed char);
7440 vector unsigned char vec_mergeh (vector unsigned char,
7441 vector unsigned char);
7442 vector bool short vec_mergeh (vector bool short, vector bool short);
7443 vector pixel vec_mergeh (vector pixel, vector pixel);
7444 vector signed short vec_mergeh (vector signed short,
7445 vector signed short);
7446 vector unsigned short vec_mergeh (vector unsigned short,
7447 vector unsigned short);
7448 vector float vec_mergeh (vector float, vector float);
7449 vector bool int vec_mergeh (vector bool int, vector bool int);
7450 vector signed int vec_mergeh (vector signed int, vector signed int);
7451 vector unsigned int vec_mergeh (vector unsigned int,
7452 vector unsigned int);
7454 vector float vec_vmrghw (vector float, vector float);
7455 vector bool int vec_vmrghw (vector bool int, vector bool int);
7456 vector signed int vec_vmrghw (vector signed int, vector signed int);
7457 vector unsigned int vec_vmrghw (vector unsigned int,
7458 vector unsigned int);
7460 vector bool short vec_vmrghh (vector bool short, vector bool short);
7461 vector signed short vec_vmrghh (vector signed short,
7462 vector signed short);
7463 vector unsigned short vec_vmrghh (vector unsigned short,
7464 vector unsigned short);
7465 vector pixel vec_vmrghh (vector pixel, vector pixel);
7467 vector bool char vec_vmrghb (vector bool char, vector bool char);
7468 vector signed char vec_vmrghb (vector signed char, vector signed char);
7469 vector unsigned char vec_vmrghb (vector unsigned char,
7470 vector unsigned char);
7472 vector bool char vec_mergel (vector bool char, vector bool char);
7473 vector signed char vec_mergel (vector signed char, vector signed char);
7474 vector unsigned char vec_mergel (vector unsigned char,
7475 vector unsigned char);
7476 vector bool short vec_mergel (vector bool short, vector bool short);
7477 vector pixel vec_mergel (vector pixel, vector pixel);
7478 vector signed short vec_mergel (vector signed short,
7479 vector signed short);
7480 vector unsigned short vec_mergel (vector unsigned short,
7481 vector unsigned short);
7482 vector float vec_mergel (vector float, vector float);
7483 vector bool int vec_mergel (vector bool int, vector bool int);
7484 vector signed int vec_mergel (vector signed int, vector signed int);
7485 vector unsigned int vec_mergel (vector unsigned int,
7486 vector unsigned int);
7488 vector float vec_vmrglw (vector float, vector float);
7489 vector signed int vec_vmrglw (vector signed int, vector signed int);
7490 vector unsigned int vec_vmrglw (vector unsigned int,
7491 vector unsigned int);
7492 vector bool int vec_vmrglw (vector bool int, vector bool int);
7494 vector bool short vec_vmrglh (vector bool short, vector bool short);
7495 vector signed short vec_vmrglh (vector signed short,
7496 vector signed short);
7497 vector unsigned short vec_vmrglh (vector unsigned short,
7498 vector unsigned short);
7499 vector pixel vec_vmrglh (vector pixel, vector pixel);
7501 vector bool char vec_vmrglb (vector bool char, vector bool char);
7502 vector signed char vec_vmrglb (vector signed char, vector signed char);
7503 vector unsigned char vec_vmrglb (vector unsigned char,
7504 vector unsigned char);
7506 vector unsigned short vec_mfvscr (void);
7508 vector unsigned char vec_min (vector bool char, vector unsigned char);
7509 vector unsigned char vec_min (vector unsigned char, vector bool char);
7510 vector unsigned char vec_min (vector unsigned char,
7511 vector unsigned char);
7512 vector signed char vec_min (vector bool char, vector signed char);
7513 vector signed char vec_min (vector signed char, vector bool char);
7514 vector signed char vec_min (vector signed char, vector signed char);
7515 vector unsigned short vec_min (vector bool short,
7516 vector unsigned short);
7517 vector unsigned short vec_min (vector unsigned short,
7519 vector unsigned short vec_min (vector unsigned short,
7520 vector unsigned short);
7521 vector signed short vec_min (vector bool short, vector signed short);
7522 vector signed short vec_min (vector signed short, vector bool short);
7523 vector signed short vec_min (vector signed short, vector signed short);
7524 vector unsigned int vec_min (vector bool int, vector unsigned int);
7525 vector unsigned int vec_min (vector unsigned int, vector bool int);
7526 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
7527 vector signed int vec_min (vector bool int, vector signed int);
7528 vector signed int vec_min (vector signed int, vector bool int);
7529 vector signed int vec_min (vector signed int, vector signed int);
7530 vector float vec_min (vector float, vector float);
7532 vector float vec_vminfp (vector float, vector float);
7534 vector signed int vec_vminsw (vector bool int, vector signed int);
7535 vector signed int vec_vminsw (vector signed int, vector bool int);
7536 vector signed int vec_vminsw (vector signed int, vector signed int);
7538 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
7539 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
7540 vector unsigned int vec_vminuw (vector unsigned int,
7541 vector unsigned int);
7543 vector signed short vec_vminsh (vector bool short, vector signed short);
7544 vector signed short vec_vminsh (vector signed short, vector bool short);
7545 vector signed short vec_vminsh (vector signed short,
7546 vector signed short);
7548 vector unsigned short vec_vminuh (vector bool short,
7549 vector unsigned short);
7550 vector unsigned short vec_vminuh (vector unsigned short,
7552 vector unsigned short vec_vminuh (vector unsigned short,
7553 vector unsigned short);
7555 vector signed char vec_vminsb (vector bool char, vector signed char);
7556 vector signed char vec_vminsb (vector signed char, vector bool char);
7557 vector signed char vec_vminsb (vector signed char, vector signed char);
7559 vector unsigned char vec_vminub (vector bool char,
7560 vector unsigned char);
7561 vector unsigned char vec_vminub (vector unsigned char,
7563 vector unsigned char vec_vminub (vector unsigned char,
7564 vector unsigned char);
7566 vector signed short vec_mladd (vector signed short,
7567 vector signed short,
7568 vector signed short);
7569 vector signed short vec_mladd (vector signed short,
7570 vector unsigned short,
7571 vector unsigned short);
7572 vector signed short vec_mladd (vector unsigned short,
7573 vector signed short,
7574 vector signed short);
7575 vector unsigned short vec_mladd (vector unsigned short,
7576 vector unsigned short,
7577 vector unsigned short);
7579 vector signed short vec_mradds (vector signed short,
7580 vector signed short,
7581 vector signed short);
7583 vector unsigned int vec_msum (vector unsigned char,
7584 vector unsigned char,
7585 vector unsigned int);
7586 vector signed int vec_msum (vector signed char,
7587 vector unsigned char,
7589 vector unsigned int vec_msum (vector unsigned short,
7590 vector unsigned short,
7591 vector unsigned int);
7592 vector signed int vec_msum (vector signed short,
7593 vector signed short,
7596 vector signed int vec_vmsumshm (vector signed short,
7597 vector signed short,
7600 vector unsigned int vec_vmsumuhm (vector unsigned short,
7601 vector unsigned short,
7602 vector unsigned int);
7604 vector signed int vec_vmsummbm (vector signed char,
7605 vector unsigned char,
7608 vector unsigned int vec_vmsumubm (vector unsigned char,
7609 vector unsigned char,
7610 vector unsigned int);
7612 vector unsigned int vec_msums (vector unsigned short,
7613 vector unsigned short,
7614 vector unsigned int);
7615 vector signed int vec_msums (vector signed short,
7616 vector signed short,
7619 vector signed int vec_vmsumshs (vector signed short,
7620 vector signed short,
7623 vector unsigned int vec_vmsumuhs (vector unsigned short,
7624 vector unsigned short,
7625 vector unsigned int);
7627 void vec_mtvscr (vector signed int);
7628 void vec_mtvscr (vector unsigned int);
7629 void vec_mtvscr (vector bool int);
7630 void vec_mtvscr (vector signed short);
7631 void vec_mtvscr (vector unsigned short);
7632 void vec_mtvscr (vector bool short);
7633 void vec_mtvscr (vector pixel);
7634 void vec_mtvscr (vector signed char);
7635 void vec_mtvscr (vector unsigned char);
7636 void vec_mtvscr (vector bool char);
7638 vector unsigned short vec_mule (vector unsigned char,
7639 vector unsigned char);
7640 vector signed short vec_mule (vector signed char,
7641 vector signed char);
7642 vector unsigned int vec_mule (vector unsigned short,
7643 vector unsigned short);
7644 vector signed int vec_mule (vector signed short, vector signed short);
7646 vector signed int vec_vmulesh (vector signed short,
7647 vector signed short);
7649 vector unsigned int vec_vmuleuh (vector unsigned short,
7650 vector unsigned short);
7652 vector signed short vec_vmulesb (vector signed char,
7653 vector signed char);
7655 vector unsigned short vec_vmuleub (vector unsigned char,
7656 vector unsigned char);
7658 vector unsigned short vec_mulo (vector unsigned char,
7659 vector unsigned char);
7660 vector signed short vec_mulo (vector signed char, vector signed char);
7661 vector unsigned int vec_mulo (vector unsigned short,
7662 vector unsigned short);
7663 vector signed int vec_mulo (vector signed short, vector signed short);
7665 vector signed int vec_vmulosh (vector signed short,
7666 vector signed short);
7668 vector unsigned int vec_vmulouh (vector unsigned short,
7669 vector unsigned short);
7671 vector signed short vec_vmulosb (vector signed char,
7672 vector signed char);
7674 vector unsigned short vec_vmuloub (vector unsigned char,
7675 vector unsigned char);
7677 vector float vec_nmsub (vector float, vector float, vector float);
7679 vector float vec_nor (vector float, vector float);
7680 vector signed int vec_nor (vector signed int, vector signed int);
7681 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
7682 vector bool int vec_nor (vector bool int, vector bool int);
7683 vector signed short vec_nor (vector signed short, vector signed short);
7684 vector unsigned short vec_nor (vector unsigned short,
7685 vector unsigned short);
7686 vector bool short vec_nor (vector bool short, vector bool short);
7687 vector signed char vec_nor (vector signed char, vector signed char);
7688 vector unsigned char vec_nor (vector unsigned char,
7689 vector unsigned char);
7690 vector bool char vec_nor (vector bool char, vector bool char);
7692 vector float vec_or (vector float, vector float);
7693 vector float vec_or (vector float, vector bool int);
7694 vector float vec_or (vector bool int, vector float);
7695 vector bool int vec_or (vector bool int, vector bool int);
7696 vector signed int vec_or (vector bool int, vector signed int);
7697 vector signed int vec_or (vector signed int, vector bool int);
7698 vector signed int vec_or (vector signed int, vector signed int);
7699 vector unsigned int vec_or (vector bool int, vector unsigned int);
7700 vector unsigned int vec_or (vector unsigned int, vector bool int);
7701 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
7702 vector bool short vec_or (vector bool short, vector bool short);
7703 vector signed short vec_or (vector bool short, vector signed short);
7704 vector signed short vec_or (vector signed short, vector bool short);
7705 vector signed short vec_or (vector signed short, vector signed short);
7706 vector unsigned short vec_or (vector bool short, vector unsigned short);
7707 vector unsigned short vec_or (vector unsigned short, vector bool short);
7708 vector unsigned short vec_or (vector unsigned short,
7709 vector unsigned short);
7710 vector signed char vec_or (vector bool char, vector signed char);
7711 vector bool char vec_or (vector bool char, vector bool char);
7712 vector signed char vec_or (vector signed char, vector bool char);
7713 vector signed char vec_or (vector signed char, vector signed char);
7714 vector unsigned char vec_or (vector bool char, vector unsigned char);
7715 vector unsigned char vec_or (vector unsigned char, vector bool char);
7716 vector unsigned char vec_or (vector unsigned char,
7717 vector unsigned char);
7719 vector signed char vec_pack (vector signed short, vector signed short);
7720 vector unsigned char vec_pack (vector unsigned short,
7721 vector unsigned short);
7722 vector bool char vec_pack (vector bool short, vector bool short);
7723 vector signed short vec_pack (vector signed int, vector signed int);
7724 vector unsigned short vec_pack (vector unsigned int,
7725 vector unsigned int);
7726 vector bool short vec_pack (vector bool int, vector bool int);
7728 vector bool short vec_vpkuwum (vector bool int, vector bool int);
7729 vector signed short vec_vpkuwum (vector signed int, vector signed int);
7730 vector unsigned short vec_vpkuwum (vector unsigned int,
7731 vector unsigned int);
7733 vector bool char vec_vpkuhum (vector bool short, vector bool short);
7734 vector signed char vec_vpkuhum (vector signed short,
7735 vector signed short);
7736 vector unsigned char vec_vpkuhum (vector unsigned short,
7737 vector unsigned short);
7739 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
7741 vector unsigned char vec_packs (vector unsigned short,
7742 vector unsigned short);
7743 vector signed char vec_packs (vector signed short, vector signed short);
7744 vector unsigned short vec_packs (vector unsigned int,
7745 vector unsigned int);
7746 vector signed short vec_packs (vector signed int, vector signed int);
7748 vector signed short vec_vpkswss (vector signed int, vector signed int);
7750 vector unsigned short vec_vpkuwus (vector unsigned int,
7751 vector unsigned int);
7753 vector signed char vec_vpkshss (vector signed short,
7754 vector signed short);
7756 vector unsigned char vec_vpkuhus (vector unsigned short,
7757 vector unsigned short);
7759 vector unsigned char vec_packsu (vector unsigned short,
7760 vector unsigned short);
7761 vector unsigned char vec_packsu (vector signed short,
7762 vector signed short);
7763 vector unsigned short vec_packsu (vector unsigned int,
7764 vector unsigned int);
7765 vector unsigned short vec_packsu (vector signed int, vector signed int);
7767 vector unsigned short vec_vpkswus (vector signed int,
7770 vector unsigned char vec_vpkshus (vector signed short,
7771 vector signed short);
7773 vector float vec_perm (vector float,
7775 vector unsigned char);
7776 vector signed int vec_perm (vector signed int,
7778 vector unsigned char);
7779 vector unsigned int vec_perm (vector unsigned int,
7780 vector unsigned int,
7781 vector unsigned char);
7782 vector bool int vec_perm (vector bool int,
7784 vector unsigned char);
7785 vector signed short vec_perm (vector signed short,
7786 vector signed short,
7787 vector unsigned char);
7788 vector unsigned short vec_perm (vector unsigned short,
7789 vector unsigned short,
7790 vector unsigned char);
7791 vector bool short vec_perm (vector bool short,
7793 vector unsigned char);
7794 vector pixel vec_perm (vector pixel,
7796 vector unsigned char);
7797 vector signed char vec_perm (vector signed char,
7799 vector unsigned char);
7800 vector unsigned char vec_perm (vector unsigned char,
7801 vector unsigned char,
7802 vector unsigned char);
7803 vector bool char vec_perm (vector bool char,
7805 vector unsigned char);
7807 vector float vec_re (vector float);
7809 vector signed char vec_rl (vector signed char,
7810 vector unsigned char);
7811 vector unsigned char vec_rl (vector unsigned char,
7812 vector unsigned char);
7813 vector signed short vec_rl (vector signed short, vector unsigned short);
7814 vector unsigned short vec_rl (vector unsigned short,
7815 vector unsigned short);
7816 vector signed int vec_rl (vector signed int, vector unsigned int);
7817 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
7819 vector signed int vec_vrlw (vector signed int, vector unsigned int);
7820 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
7822 vector signed short vec_vrlh (vector signed short,
7823 vector unsigned short);
7824 vector unsigned short vec_vrlh (vector unsigned short,
7825 vector unsigned short);
7827 vector signed char vec_vrlb (vector signed char, vector unsigned char);
7828 vector unsigned char vec_vrlb (vector unsigned char,
7829 vector unsigned char);
7831 vector float vec_round (vector float);
7833 vector float vec_rsqrte (vector float);
7835 vector float vec_sel (vector float, vector float, vector bool int);
7836 vector float vec_sel (vector float, vector float, vector unsigned int);
7837 vector signed int vec_sel (vector signed int,
7840 vector signed int vec_sel (vector signed int,
7842 vector unsigned int);
7843 vector unsigned int vec_sel (vector unsigned int,
7844 vector unsigned int,
7846 vector unsigned int vec_sel (vector unsigned int,
7847 vector unsigned int,
7848 vector unsigned int);
7849 vector bool int vec_sel (vector bool int,
7852 vector bool int vec_sel (vector bool int,
7854 vector unsigned int);
7855 vector signed short vec_sel (vector signed short,
7856 vector signed short,
7858 vector signed short vec_sel (vector signed short,
7859 vector signed short,
7860 vector unsigned short);
7861 vector unsigned short vec_sel (vector unsigned short,
7862 vector unsigned short,
7864 vector unsigned short vec_sel (vector unsigned short,
7865 vector unsigned short,
7866 vector unsigned short);
7867 vector bool short vec_sel (vector bool short,
7870 vector bool short vec_sel (vector bool short,
7872 vector unsigned short);
7873 vector signed char vec_sel (vector signed char,
7876 vector signed char vec_sel (vector signed char,
7878 vector unsigned char);
7879 vector unsigned char vec_sel (vector unsigned char,
7880 vector unsigned char,
7882 vector unsigned char vec_sel (vector unsigned char,
7883 vector unsigned char,
7884 vector unsigned char);
7885 vector bool char vec_sel (vector bool char,
7888 vector bool char vec_sel (vector bool char,
7890 vector unsigned char);
7892 vector signed char vec_sl (vector signed char,
7893 vector unsigned char);
7894 vector unsigned char vec_sl (vector unsigned char,
7895 vector unsigned char);
7896 vector signed short vec_sl (vector signed short, vector unsigned short);
7897 vector unsigned short vec_sl (vector unsigned short,
7898 vector unsigned short);
7899 vector signed int vec_sl (vector signed int, vector unsigned int);
7900 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
7902 vector signed int vec_vslw (vector signed int, vector unsigned int);
7903 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
7905 vector signed short vec_vslh (vector signed short,
7906 vector unsigned short);
7907 vector unsigned short vec_vslh (vector unsigned short,
7908 vector unsigned short);
7910 vector signed char vec_vslb (vector signed char, vector unsigned char);
7911 vector unsigned char vec_vslb (vector unsigned char,
7912 vector unsigned char);
7914 vector float vec_sld (vector float, vector float, const int);
7915 vector signed int vec_sld (vector signed int,
7918 vector unsigned int vec_sld (vector unsigned int,
7919 vector unsigned int,
7921 vector bool int vec_sld (vector bool int,
7924 vector signed short vec_sld (vector signed short,
7925 vector signed short,
7927 vector unsigned short vec_sld (vector unsigned short,
7928 vector unsigned short,
7930 vector bool short vec_sld (vector bool short,
7933 vector pixel vec_sld (vector pixel,
7936 vector signed char vec_sld (vector signed char,
7939 vector unsigned char vec_sld (vector unsigned char,
7940 vector unsigned char,
7942 vector bool char vec_sld (vector bool char,
7946 vector signed int vec_sll (vector signed int,
7947 vector unsigned int);
7948 vector signed int vec_sll (vector signed int,
7949 vector unsigned short);
7950 vector signed int vec_sll (vector signed int,
7951 vector unsigned char);
7952 vector unsigned int vec_sll (vector unsigned int,
7953 vector unsigned int);
7954 vector unsigned int vec_sll (vector unsigned int,
7955 vector unsigned short);
7956 vector unsigned int vec_sll (vector unsigned int,
7957 vector unsigned char);
7958 vector bool int vec_sll (vector bool int,
7959 vector unsigned int);
7960 vector bool int vec_sll (vector bool int,
7961 vector unsigned short);
7962 vector bool int vec_sll (vector bool int,
7963 vector unsigned char);
7964 vector signed short vec_sll (vector signed short,
7965 vector unsigned int);
7966 vector signed short vec_sll (vector signed short,
7967 vector unsigned short);
7968 vector signed short vec_sll (vector signed short,
7969 vector unsigned char);
7970 vector unsigned short vec_sll (vector unsigned short,
7971 vector unsigned int);
7972 vector unsigned short vec_sll (vector unsigned short,
7973 vector unsigned short);
7974 vector unsigned short vec_sll (vector unsigned short,
7975 vector unsigned char);
7976 vector bool short vec_sll (vector bool short, vector unsigned int);
7977 vector bool short vec_sll (vector bool short, vector unsigned short);
7978 vector bool short vec_sll (vector bool short, vector unsigned char);
7979 vector pixel vec_sll (vector pixel, vector unsigned int);
7980 vector pixel vec_sll (vector pixel, vector unsigned short);
7981 vector pixel vec_sll (vector pixel, vector unsigned char);
7982 vector signed char vec_sll (vector signed char, vector unsigned int);
7983 vector signed char vec_sll (vector signed char, vector unsigned short);
7984 vector signed char vec_sll (vector signed char, vector unsigned char);
7985 vector unsigned char vec_sll (vector unsigned char,
7986 vector unsigned int);
7987 vector unsigned char vec_sll (vector unsigned char,
7988 vector unsigned short);
7989 vector unsigned char vec_sll (vector unsigned char,
7990 vector unsigned char);
7991 vector bool char vec_sll (vector bool char, vector unsigned int);
7992 vector bool char vec_sll (vector bool char, vector unsigned short);
7993 vector bool char vec_sll (vector bool char, vector unsigned char);
7995 vector float vec_slo (vector float, vector signed char);
7996 vector float vec_slo (vector float, vector unsigned char);
7997 vector signed int vec_slo (vector signed int, vector signed char);
7998 vector signed int vec_slo (vector signed int, vector unsigned char);
7999 vector unsigned int vec_slo (vector unsigned int, vector signed char);
8000 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
8001 vector signed short vec_slo (vector signed short, vector signed char);
8002 vector signed short vec_slo (vector signed short, vector unsigned char);
8003 vector unsigned short vec_slo (vector unsigned short,
8004 vector signed char);
8005 vector unsigned short vec_slo (vector unsigned short,
8006 vector unsigned char);
8007 vector pixel vec_slo (vector pixel, vector signed char);
8008 vector pixel vec_slo (vector pixel, vector unsigned char);
8009 vector signed char vec_slo (vector signed char, vector signed char);
8010 vector signed char vec_slo (vector signed char, vector unsigned char);
8011 vector unsigned char vec_slo (vector unsigned char, vector signed char);
8012 vector unsigned char vec_slo (vector unsigned char,
8013 vector unsigned char);
8015 vector signed char vec_splat (vector signed char, const int);
8016 vector unsigned char vec_splat (vector unsigned char, const int);
8017 vector bool char vec_splat (vector bool char, const int);
8018 vector signed short vec_splat (vector signed short, const int);
8019 vector unsigned short vec_splat (vector unsigned short, const int);
8020 vector bool short vec_splat (vector bool short, const int);
8021 vector pixel vec_splat (vector pixel, const int);
8022 vector float vec_splat (vector float, const int);
8023 vector signed int vec_splat (vector signed int, const int);
8024 vector unsigned int vec_splat (vector unsigned int, const int);
8025 vector bool int vec_splat (vector bool int, const int);
8027 vector float vec_vspltw (vector float, const int);
8028 vector signed int vec_vspltw (vector signed int, const int);
8029 vector unsigned int vec_vspltw (vector unsigned int, const int);
8030 vector bool int vec_vspltw (vector bool int, const int);
8032 vector bool short vec_vsplth (vector bool short, const int);
8033 vector signed short vec_vsplth (vector signed short, const int);
8034 vector unsigned short vec_vsplth (vector unsigned short, const int);
8035 vector pixel vec_vsplth (vector pixel, const int);
8037 vector signed char vec_vspltb (vector signed char, const int);
8038 vector unsigned char vec_vspltb (vector unsigned char, const int);
8039 vector bool char vec_vspltb (vector bool char, const int);
8041 vector signed char vec_splat_s8 (const int);
8043 vector signed short vec_splat_s16 (const int);
8045 vector signed int vec_splat_s32 (const int);
8047 vector unsigned char vec_splat_u8 (const int);
8049 vector unsigned short vec_splat_u16 (const int);
8051 vector unsigned int vec_splat_u32 (const int);
8053 vector signed char vec_sr (vector signed char, vector unsigned char);
8054 vector unsigned char vec_sr (vector unsigned char,
8055 vector unsigned char);
8056 vector signed short vec_sr (vector signed short,
8057 vector unsigned short);
8058 vector unsigned short vec_sr (vector unsigned short,
8059 vector unsigned short);
8060 vector signed int vec_sr (vector signed int, vector unsigned int);
8061 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
8063 vector signed int vec_vsrw (vector signed int, vector unsigned int);
8064 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
8066 vector signed short vec_vsrh (vector signed short,
8067 vector unsigned short);
8068 vector unsigned short vec_vsrh (vector unsigned short,
8069 vector unsigned short);
8071 vector signed char vec_vsrb (vector signed char, vector unsigned char);
8072 vector unsigned char vec_vsrb (vector unsigned char,
8073 vector unsigned char);
8075 vector signed char vec_sra (vector signed char, vector unsigned char);
8076 vector unsigned char vec_sra (vector unsigned char,
8077 vector unsigned char);
8078 vector signed short vec_sra (vector signed short,
8079 vector unsigned short);
8080 vector unsigned short vec_sra (vector unsigned short,
8081 vector unsigned short);
8082 vector signed int vec_sra (vector signed int, vector unsigned int);
8083 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
8085 vector signed int vec_vsraw (vector signed int, vector unsigned int);
8086 vector unsigned int vec_vsraw (vector unsigned int,
8087 vector unsigned int);
8089 vector signed short vec_vsrah (vector signed short,
8090 vector unsigned short);
8091 vector unsigned short vec_vsrah (vector unsigned short,
8092 vector unsigned short);
8094 vector signed char vec_vsrab (vector signed char, vector unsigned char);
8095 vector unsigned char vec_vsrab (vector unsigned char,
8096 vector unsigned char);
8098 vector signed int vec_srl (vector signed int, vector unsigned int);
8099 vector signed int vec_srl (vector signed int, vector unsigned short);
8100 vector signed int vec_srl (vector signed int, vector unsigned char);
8101 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
8102 vector unsigned int vec_srl (vector unsigned int,
8103 vector unsigned short);
8104 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
8105 vector bool int vec_srl (vector bool int, vector unsigned int);
8106 vector bool int vec_srl (vector bool int, vector unsigned short);
8107 vector bool int vec_srl (vector bool int, vector unsigned char);
8108 vector signed short vec_srl (vector signed short, vector unsigned int);
8109 vector signed short vec_srl (vector signed short,
8110 vector unsigned short);
8111 vector signed short vec_srl (vector signed short, vector unsigned char);
8112 vector unsigned short vec_srl (vector unsigned short,
8113 vector unsigned int);
8114 vector unsigned short vec_srl (vector unsigned short,
8115 vector unsigned short);
8116 vector unsigned short vec_srl (vector unsigned short,
8117 vector unsigned char);
8118 vector bool short vec_srl (vector bool short, vector unsigned int);
8119 vector bool short vec_srl (vector bool short, vector unsigned short);
8120 vector bool short vec_srl (vector bool short, vector unsigned char);
8121 vector pixel vec_srl (vector pixel, vector unsigned int);
8122 vector pixel vec_srl (vector pixel, vector unsigned short);
8123 vector pixel vec_srl (vector pixel, vector unsigned char);
8124 vector signed char vec_srl (vector signed char, vector unsigned int);
8125 vector signed char vec_srl (vector signed char, vector unsigned short);
8126 vector signed char vec_srl (vector signed char, vector unsigned char);
8127 vector unsigned char vec_srl (vector unsigned char,
8128 vector unsigned int);
8129 vector unsigned char vec_srl (vector unsigned char,
8130 vector unsigned short);
8131 vector unsigned char vec_srl (vector unsigned char,
8132 vector unsigned char);
8133 vector bool char vec_srl (vector bool char, vector unsigned int);
8134 vector bool char vec_srl (vector bool char, vector unsigned short);
8135 vector bool char vec_srl (vector bool char, vector unsigned char);
8137 vector float vec_sro (vector float, vector signed char);
8138 vector float vec_sro (vector float, vector unsigned char);
8139 vector signed int vec_sro (vector signed int, vector signed char);
8140 vector signed int vec_sro (vector signed int, vector unsigned char);
8141 vector unsigned int vec_sro (vector unsigned int, vector signed char);
8142 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
8143 vector signed short vec_sro (vector signed short, vector signed char);
8144 vector signed short vec_sro (vector signed short, vector unsigned char);
8145 vector unsigned short vec_sro (vector unsigned short,
8146 vector signed char);
8147 vector unsigned short vec_sro (vector unsigned short,
8148 vector unsigned char);
8149 vector pixel vec_sro (vector pixel, vector signed char);
8150 vector pixel vec_sro (vector pixel, vector unsigned char);
8151 vector signed char vec_sro (vector signed char, vector signed char);
8152 vector signed char vec_sro (vector signed char, vector unsigned char);
8153 vector unsigned char vec_sro (vector unsigned char, vector signed char);
8154 vector unsigned char vec_sro (vector unsigned char,
8155 vector unsigned char);
8157 void vec_st (vector float, int, vector float *);
8158 void vec_st (vector float, int, float *);
8159 void vec_st (vector signed int, int, vector signed int *);
8160 void vec_st (vector signed int, int, int *);
8161 void vec_st (vector unsigned int, int, vector unsigned int *);
8162 void vec_st (vector unsigned int, int, unsigned int *);
8163 void vec_st (vector bool int, int, vector bool int *);
8164 void vec_st (vector bool int, int, unsigned int *);
8165 void vec_st (vector bool int, int, int *);
8166 void vec_st (vector signed short, int, vector signed short *);
8167 void vec_st (vector signed short, int, short *);
8168 void vec_st (vector unsigned short, int, vector unsigned short *);
8169 void vec_st (vector unsigned short, int, unsigned short *);
8170 void vec_st (vector bool short, int, vector bool short *);
8171 void vec_st (vector bool short, int, unsigned short *);
8172 void vec_st (vector pixel, int, vector pixel *);
8173 void vec_st (vector pixel, int, unsigned short *);
8174 void vec_st (vector pixel, int, short *);
8175 void vec_st (vector bool short, int, short *);
8176 void vec_st (vector signed char, int, vector signed char *);
8177 void vec_st (vector signed char, int, signed char *);
8178 void vec_st (vector unsigned char, int, vector unsigned char *);
8179 void vec_st (vector unsigned char, int, unsigned char *);
8180 void vec_st (vector bool char, int, vector bool char *);
8181 void vec_st (vector bool char, int, unsigned char *);
8182 void vec_st (vector bool char, int, signed char *);
8184 void vec_ste (vector signed char, int, signed char *);
8185 void vec_ste (vector unsigned char, int, unsigned char *);
8186 void vec_ste (vector bool char, int, signed char *);
8187 void vec_ste (vector bool char, int, unsigned char *);
8188 void vec_ste (vector signed short, int, short *);
8189 void vec_ste (vector unsigned short, int, unsigned short *);
8190 void vec_ste (vector bool short, int, short *);
8191 void vec_ste (vector bool short, int, unsigned short *);
8192 void vec_ste (vector pixel, int, short *);
8193 void vec_ste (vector pixel, int, unsigned short *);
8194 void vec_ste (vector float, int, float *);
8195 void vec_ste (vector signed int, int, int *);
8196 void vec_ste (vector unsigned int, int, unsigned int *);
8197 void vec_ste (vector bool int, int, int *);
8198 void vec_ste (vector bool int, int, unsigned int *);
8200 void vec_stvewx (vector float, int, float *);
8201 void vec_stvewx (vector signed int, int, int *);
8202 void vec_stvewx (vector unsigned int, int, unsigned int *);
8203 void vec_stvewx (vector bool int, int, int *);
8204 void vec_stvewx (vector bool int, int, unsigned int *);
8206 void vec_stvehx (vector signed short, int, short *);
8207 void vec_stvehx (vector unsigned short, int, unsigned short *);
8208 void vec_stvehx (vector bool short, int, short *);
8209 void vec_stvehx (vector bool short, int, unsigned short *);
8210 void vec_stvehx (vector pixel, int, short *);
8211 void vec_stvehx (vector pixel, int, unsigned short *);
8213 void vec_stvebx (vector signed char, int, signed char *);
8214 void vec_stvebx (vector unsigned char, int, unsigned char *);
8215 void vec_stvebx (vector bool char, int, signed char *);
8216 void vec_stvebx (vector bool char, int, unsigned char *);
8218 void vec_stl (vector float, int, vector float *);
8219 void vec_stl (vector float, int, float *);
8220 void vec_stl (vector signed int, int, vector signed int *);
8221 void vec_stl (vector signed int, int, int *);
8222 void vec_stl (vector unsigned int, int, vector unsigned int *);
8223 void vec_stl (vector unsigned int, int, unsigned int *);
8224 void vec_stl (vector bool int, int, vector bool int *);
8225 void vec_stl (vector bool int, int, unsigned int *);
8226 void vec_stl (vector bool int, int, int *);
8227 void vec_stl (vector signed short, int, vector signed short *);
8228 void vec_stl (vector signed short, int, short *);
8229 void vec_stl (vector unsigned short, int, vector unsigned short *);
8230 void vec_stl (vector unsigned short, int, unsigned short *);
8231 void vec_stl (vector bool short, int, vector bool short *);
8232 void vec_stl (vector bool short, int, unsigned short *);
8233 void vec_stl (vector bool short, int, short *);
8234 void vec_stl (vector pixel, int, vector pixel *);
8235 void vec_stl (vector pixel, int, unsigned short *);
8236 void vec_stl (vector pixel, int, short *);
8237 void vec_stl (vector signed char, int, vector signed char *);
8238 void vec_stl (vector signed char, int, signed char *);
8239 void vec_stl (vector unsigned char, int, vector unsigned char *);
8240 void vec_stl (vector unsigned char, int, unsigned char *);
8241 void vec_stl (vector bool char, int, vector bool char *);
8242 void vec_stl (vector bool char, int, unsigned char *);
8243 void vec_stl (vector bool char, int, signed char *);
8245 vector signed char vec_sub (vector bool char, vector signed char);
8246 vector signed char vec_sub (vector signed char, vector bool char);
8247 vector signed char vec_sub (vector signed char, vector signed char);
8248 vector unsigned char vec_sub (vector bool char, vector unsigned char);
8249 vector unsigned char vec_sub (vector unsigned char, vector bool char);
8250 vector unsigned char vec_sub (vector unsigned char,
8251 vector unsigned char);
8252 vector signed short vec_sub (vector bool short, vector signed short);
8253 vector signed short vec_sub (vector signed short, vector bool short);
8254 vector signed short vec_sub (vector signed short, vector signed short);
8255 vector unsigned short vec_sub (vector bool short,
8256 vector unsigned short);
8257 vector unsigned short vec_sub (vector unsigned short,
8259 vector unsigned short vec_sub (vector unsigned short,
8260 vector unsigned short);
8261 vector signed int vec_sub (vector bool int, vector signed int);
8262 vector signed int vec_sub (vector signed int, vector bool int);
8263 vector signed int vec_sub (vector signed int, vector signed int);
8264 vector unsigned int vec_sub (vector bool int, vector unsigned int);
8265 vector unsigned int vec_sub (vector unsigned int, vector bool int);
8266 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
8267 vector float vec_sub (vector float, vector float);
8269 vector float vec_vsubfp (vector float, vector float);
8271 vector signed int vec_vsubuwm (vector bool int, vector signed int);
8272 vector signed int vec_vsubuwm (vector signed int, vector bool int);
8273 vector signed int vec_vsubuwm (vector signed int, vector signed int);
8274 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
8275 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
8276 vector unsigned int vec_vsubuwm (vector unsigned int,
8277 vector unsigned int);
8279 vector signed short vec_vsubuhm (vector bool short,
8280 vector signed short);
8281 vector signed short vec_vsubuhm (vector signed short,
8283 vector signed short vec_vsubuhm (vector signed short,
8284 vector signed short);
8285 vector unsigned short vec_vsubuhm (vector bool short,
8286 vector unsigned short);
8287 vector unsigned short vec_vsubuhm (vector unsigned short,
8289 vector unsigned short vec_vsubuhm (vector unsigned short,
8290 vector unsigned short);
8292 vector signed char vec_vsububm (vector bool char, vector signed char);
8293 vector signed char vec_vsububm (vector signed char, vector bool char);
8294 vector signed char vec_vsububm (vector signed char, vector signed char);
8295 vector unsigned char vec_vsububm (vector bool char,
8296 vector unsigned char);
8297 vector unsigned char vec_vsububm (vector unsigned char,
8299 vector unsigned char vec_vsububm (vector unsigned char,
8300 vector unsigned char);
8302 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
8304 vector unsigned char vec_subs (vector bool char, vector unsigned char);
8305 vector unsigned char vec_subs (vector unsigned char, vector bool char);
8306 vector unsigned char vec_subs (vector unsigned char,
8307 vector unsigned char);
8308 vector signed char vec_subs (vector bool char, vector signed char);
8309 vector signed char vec_subs (vector signed char, vector bool char);
8310 vector signed char vec_subs (vector signed char, vector signed char);
8311 vector unsigned short vec_subs (vector bool short,
8312 vector unsigned short);
8313 vector unsigned short vec_subs (vector unsigned short,
8315 vector unsigned short vec_subs (vector unsigned short,
8316 vector unsigned short);
8317 vector signed short vec_subs (vector bool short, vector signed short);
8318 vector signed short vec_subs (vector signed short, vector bool short);
8319 vector signed short vec_subs (vector signed short, vector signed short);
8320 vector unsigned int vec_subs (vector bool int, vector unsigned int);
8321 vector unsigned int vec_subs (vector unsigned int, vector bool int);
8322 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
8323 vector signed int vec_subs (vector bool int, vector signed int);
8324 vector signed int vec_subs (vector signed int, vector bool int);
8325 vector signed int vec_subs (vector signed int, vector signed int);
8327 vector signed int vec_vsubsws (vector bool int, vector signed int);
8328 vector signed int vec_vsubsws (vector signed int, vector bool int);
8329 vector signed int vec_vsubsws (vector signed int, vector signed int);
8331 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
8332 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
8333 vector unsigned int vec_vsubuws (vector unsigned int,
8334 vector unsigned int);
8336 vector signed short vec_vsubshs (vector bool short,
8337 vector signed short);
8338 vector signed short vec_vsubshs (vector signed short,
8340 vector signed short vec_vsubshs (vector signed short,
8341 vector signed short);
8343 vector unsigned short vec_vsubuhs (vector bool short,
8344 vector unsigned short);
8345 vector unsigned short vec_vsubuhs (vector unsigned short,
8347 vector unsigned short vec_vsubuhs (vector unsigned short,
8348 vector unsigned short);
8350 vector signed char vec_vsubsbs (vector bool char, vector signed char);
8351 vector signed char vec_vsubsbs (vector signed char, vector bool char);
8352 vector signed char vec_vsubsbs (vector signed char, vector signed char);
8354 vector unsigned char vec_vsububs (vector bool char,
8355 vector unsigned char);
8356 vector unsigned char vec_vsububs (vector unsigned char,
8358 vector unsigned char vec_vsububs (vector unsigned char,
8359 vector unsigned char);
8361 vector unsigned int vec_sum4s (vector unsigned char,
8362 vector unsigned int);
8363 vector signed int vec_sum4s (vector signed char, vector signed int);
8364 vector signed int vec_sum4s (vector signed short, vector signed int);
8366 vector signed int vec_vsum4shs (vector signed short, vector signed int);
8368 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
8370 vector unsigned int vec_vsum4ubs (vector unsigned char,
8371 vector unsigned int);
8373 vector signed int vec_sum2s (vector signed int, vector signed int);
8375 vector signed int vec_sums (vector signed int, vector signed int);
8377 vector float vec_trunc (vector float);
8379 vector signed short vec_unpackh (vector signed char);
8380 vector bool short vec_unpackh (vector bool char);
8381 vector signed int vec_unpackh (vector signed short);
8382 vector bool int vec_unpackh (vector bool short);
8383 vector unsigned int vec_unpackh (vector pixel);
8385 vector bool int vec_vupkhsh (vector bool short);
8386 vector signed int vec_vupkhsh (vector signed short);
8388 vector unsigned int vec_vupkhpx (vector pixel);
8390 vector bool short vec_vupkhsb (vector bool char);
8391 vector signed short vec_vupkhsb (vector signed char);
8393 vector signed short vec_unpackl (vector signed char);
8394 vector bool short vec_unpackl (vector bool char);
8395 vector unsigned int vec_unpackl (vector pixel);
8396 vector signed int vec_unpackl (vector signed short);
8397 vector bool int vec_unpackl (vector bool short);
8399 vector unsigned int vec_vupklpx (vector pixel);
8401 vector bool int vec_vupklsh (vector bool short);
8402 vector signed int vec_vupklsh (vector signed short);
8404 vector bool short vec_vupklsb (vector bool char);
8405 vector signed short vec_vupklsb (vector signed char);
8407 vector float vec_xor (vector float, vector float);
8408 vector float vec_xor (vector float, vector bool int);
8409 vector float vec_xor (vector bool int, vector float);
8410 vector bool int vec_xor (vector bool int, vector bool int);
8411 vector signed int vec_xor (vector bool int, vector signed int);
8412 vector signed int vec_xor (vector signed int, vector bool int);
8413 vector signed int vec_xor (vector signed int, vector signed int);
8414 vector unsigned int vec_xor (vector bool int, vector unsigned int);
8415 vector unsigned int vec_xor (vector unsigned int, vector bool int);
8416 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
8417 vector bool short vec_xor (vector bool short, vector bool short);
8418 vector signed short vec_xor (vector bool short, vector signed short);
8419 vector signed short vec_xor (vector signed short, vector bool short);
8420 vector signed short vec_xor (vector signed short, vector signed short);
8421 vector unsigned short vec_xor (vector bool short,
8422 vector unsigned short);
8423 vector unsigned short vec_xor (vector unsigned short,
8425 vector unsigned short vec_xor (vector unsigned short,
8426 vector unsigned short);
8427 vector signed char vec_xor (vector bool char, vector signed char);
8428 vector bool char vec_xor (vector bool char, vector bool char);
8429 vector signed char vec_xor (vector signed char, vector bool char);
8430 vector signed char vec_xor (vector signed char, vector signed char);
8431 vector unsigned char vec_xor (vector bool char, vector unsigned char);
8432 vector unsigned char vec_xor (vector unsigned char, vector bool char);
8433 vector unsigned char vec_xor (vector unsigned char,
8434 vector unsigned char);
8436 int vec_all_eq (vector signed char, vector bool char);
8437 int vec_all_eq (vector signed char, vector signed char);
8438 int vec_all_eq (vector unsigned char, vector bool char);
8439 int vec_all_eq (vector unsigned char, vector unsigned char);
8440 int vec_all_eq (vector bool char, vector bool char);
8441 int vec_all_eq (vector bool char, vector unsigned char);
8442 int vec_all_eq (vector bool char, vector signed char);
8443 int vec_all_eq (vector signed short, vector bool short);
8444 int vec_all_eq (vector signed short, vector signed short);
8445 int vec_all_eq (vector unsigned short, vector bool short);
8446 int vec_all_eq (vector unsigned short, vector unsigned short);
8447 int vec_all_eq (vector bool short, vector bool short);
8448 int vec_all_eq (vector bool short, vector unsigned short);
8449 int vec_all_eq (vector bool short, vector signed short);
8450 int vec_all_eq (vector pixel, vector pixel);
8451 int vec_all_eq (vector signed int, vector bool int);
8452 int vec_all_eq (vector signed int, vector signed int);
8453 int vec_all_eq (vector unsigned int, vector bool int);
8454 int vec_all_eq (vector unsigned int, vector unsigned int);
8455 int vec_all_eq (vector bool int, vector bool int);
8456 int vec_all_eq (vector bool int, vector unsigned int);
8457 int vec_all_eq (vector bool int, vector signed int);
8458 int vec_all_eq (vector float, vector float);
8460 int vec_all_ge (vector bool char, vector unsigned char);
8461 int vec_all_ge (vector unsigned char, vector bool char);
8462 int vec_all_ge (vector unsigned char, vector unsigned char);
8463 int vec_all_ge (vector bool char, vector signed char);
8464 int vec_all_ge (vector signed char, vector bool char);
8465 int vec_all_ge (vector signed char, vector signed char);
8466 int vec_all_ge (vector bool short, vector unsigned short);
8467 int vec_all_ge (vector unsigned short, vector bool short);
8468 int vec_all_ge (vector unsigned short, vector unsigned short);
8469 int vec_all_ge (vector signed short, vector signed short);
8470 int vec_all_ge (vector bool short, vector signed short);
8471 int vec_all_ge (vector signed short, vector bool short);
8472 int vec_all_ge (vector bool int, vector unsigned int);
8473 int vec_all_ge (vector unsigned int, vector bool int);
8474 int vec_all_ge (vector unsigned int, vector unsigned int);
8475 int vec_all_ge (vector bool int, vector signed int);
8476 int vec_all_ge (vector signed int, vector bool int);
8477 int vec_all_ge (vector signed int, vector signed int);
8478 int vec_all_ge (vector float, vector float);
8480 int vec_all_gt (vector bool char, vector unsigned char);
8481 int vec_all_gt (vector unsigned char, vector bool char);
8482 int vec_all_gt (vector unsigned char, vector unsigned char);
8483 int vec_all_gt (vector bool char, vector signed char);
8484 int vec_all_gt (vector signed char, vector bool char);
8485 int vec_all_gt (vector signed char, vector signed char);
8486 int vec_all_gt (vector bool short, vector unsigned short);
8487 int vec_all_gt (vector unsigned short, vector bool short);
8488 int vec_all_gt (vector unsigned short, vector unsigned short);
8489 int vec_all_gt (vector bool short, vector signed short);
8490 int vec_all_gt (vector signed short, vector bool short);
8491 int vec_all_gt (vector signed short, vector signed short);
8492 int vec_all_gt (vector bool int, vector unsigned int);
8493 int vec_all_gt (vector unsigned int, vector bool int);
8494 int vec_all_gt (vector unsigned int, vector unsigned int);
8495 int vec_all_gt (vector bool int, vector signed int);
8496 int vec_all_gt (vector signed int, vector bool int);
8497 int vec_all_gt (vector signed int, vector signed int);
8498 int vec_all_gt (vector float, vector float);
8500 int vec_all_in (vector float, vector float);
8502 int vec_all_le (vector bool char, vector unsigned char);
8503 int vec_all_le (vector unsigned char, vector bool char);
8504 int vec_all_le (vector unsigned char, vector unsigned char);
8505 int vec_all_le (vector bool char, vector signed char);
8506 int vec_all_le (vector signed char, vector bool char);
8507 int vec_all_le (vector signed char, vector signed char);
8508 int vec_all_le (vector bool short, vector unsigned short);
8509 int vec_all_le (vector unsigned short, vector bool short);
8510 int vec_all_le (vector unsigned short, vector unsigned short);
8511 int vec_all_le (vector bool short, vector signed short);
8512 int vec_all_le (vector signed short, vector bool short);
8513 int vec_all_le (vector signed short, vector signed short);
8514 int vec_all_le (vector bool int, vector unsigned int);
8515 int vec_all_le (vector unsigned int, vector bool int);
8516 int vec_all_le (vector unsigned int, vector unsigned int);
8517 int vec_all_le (vector bool int, vector signed int);
8518 int vec_all_le (vector signed int, vector bool int);
8519 int vec_all_le (vector signed int, vector signed int);
8520 int vec_all_le (vector float, vector float);
8522 int vec_all_lt (vector bool char, vector unsigned char);
8523 int vec_all_lt (vector unsigned char, vector bool char);
8524 int vec_all_lt (vector unsigned char, vector unsigned char);
8525 int vec_all_lt (vector bool char, vector signed char);
8526 int vec_all_lt (vector signed char, vector bool char);
8527 int vec_all_lt (vector signed char, vector signed char);
8528 int vec_all_lt (vector bool short, vector unsigned short);
8529 int vec_all_lt (vector unsigned short, vector bool short);
8530 int vec_all_lt (vector unsigned short, vector unsigned short);
8531 int vec_all_lt (vector bool short, vector signed short);
8532 int vec_all_lt (vector signed short, vector bool short);
8533 int vec_all_lt (vector signed short, vector signed short);
8534 int vec_all_lt (vector bool int, vector unsigned int);
8535 int vec_all_lt (vector unsigned int, vector bool int);
8536 int vec_all_lt (vector unsigned int, vector unsigned int);
8537 int vec_all_lt (vector bool int, vector signed int);
8538 int vec_all_lt (vector signed int, vector bool int);
8539 int vec_all_lt (vector signed int, vector signed int);
8540 int vec_all_lt (vector float, vector float);
8542 int vec_all_nan (vector float);
8544 int vec_all_ne (vector signed char, vector bool char);
8545 int vec_all_ne (vector signed char, vector signed char);
8546 int vec_all_ne (vector unsigned char, vector bool char);
8547 int vec_all_ne (vector unsigned char, vector unsigned char);
8548 int vec_all_ne (vector bool char, vector bool char);
8549 int vec_all_ne (vector bool char, vector unsigned char);
8550 int vec_all_ne (vector bool char, vector signed char);
8551 int vec_all_ne (vector signed short, vector bool short);
8552 int vec_all_ne (vector signed short, vector signed short);
8553 int vec_all_ne (vector unsigned short, vector bool short);
8554 int vec_all_ne (vector unsigned short, vector unsigned short);
8555 int vec_all_ne (vector bool short, vector bool short);
8556 int vec_all_ne (vector bool short, vector unsigned short);
8557 int vec_all_ne (vector bool short, vector signed short);
8558 int vec_all_ne (vector pixel, vector pixel);
8559 int vec_all_ne (vector signed int, vector bool int);
8560 int vec_all_ne (vector signed int, vector signed int);
8561 int vec_all_ne (vector unsigned int, vector bool int);
8562 int vec_all_ne (vector unsigned int, vector unsigned int);
8563 int vec_all_ne (vector bool int, vector bool int);
8564 int vec_all_ne (vector bool int, vector unsigned int);
8565 int vec_all_ne (vector bool int, vector signed int);
8566 int vec_all_ne (vector float, vector float);
8568 int vec_all_nge (vector float, vector float);
8570 int vec_all_ngt (vector float, vector float);
8572 int vec_all_nle (vector float, vector float);
8574 int vec_all_nlt (vector float, vector float);
8576 int vec_all_numeric (vector float);
8578 int vec_any_eq (vector signed char, vector bool char);
8579 int vec_any_eq (vector signed char, vector signed char);
8580 int vec_any_eq (vector unsigned char, vector bool char);
8581 int vec_any_eq (vector unsigned char, vector unsigned char);
8582 int vec_any_eq (vector bool char, vector bool char);
8583 int vec_any_eq (vector bool char, vector unsigned char);
8584 int vec_any_eq (vector bool char, vector signed char);
8585 int vec_any_eq (vector signed short, vector bool short);
8586 int vec_any_eq (vector signed short, vector signed short);
8587 int vec_any_eq (vector unsigned short, vector bool short);
8588 int vec_any_eq (vector unsigned short, vector unsigned short);
8589 int vec_any_eq (vector bool short, vector bool short);
8590 int vec_any_eq (vector bool short, vector unsigned short);
8591 int vec_any_eq (vector bool short, vector signed short);
8592 int vec_any_eq (vector pixel, vector pixel);
8593 int vec_any_eq (vector signed int, vector bool int);
8594 int vec_any_eq (vector signed int, vector signed int);
8595 int vec_any_eq (vector unsigned int, vector bool int);
8596 int vec_any_eq (vector unsigned int, vector unsigned int);
8597 int vec_any_eq (vector bool int, vector bool int);
8598 int vec_any_eq (vector bool int, vector unsigned int);
8599 int vec_any_eq (vector bool int, vector signed int);
8600 int vec_any_eq (vector float, vector float);
8602 int vec_any_ge (vector signed char, vector bool char);
8603 int vec_any_ge (vector unsigned char, vector bool char);
8604 int vec_any_ge (vector unsigned char, vector unsigned char);
8605 int vec_any_ge (vector signed char, vector signed char);
8606 int vec_any_ge (vector bool char, vector unsigned char);
8607 int vec_any_ge (vector bool char, vector signed char);
8608 int vec_any_ge (vector unsigned short, vector bool short);
8609 int vec_any_ge (vector unsigned short, vector unsigned short);
8610 int vec_any_ge (vector signed short, vector signed short);
8611 int vec_any_ge (vector signed short, vector bool short);
8612 int vec_any_ge (vector bool short, vector unsigned short);
8613 int vec_any_ge (vector bool short, vector signed short);
8614 int vec_any_ge (vector signed int, vector bool int);
8615 int vec_any_ge (vector unsigned int, vector bool int);
8616 int vec_any_ge (vector unsigned int, vector unsigned int);
8617 int vec_any_ge (vector signed int, vector signed int);
8618 int vec_any_ge (vector bool int, vector unsigned int);
8619 int vec_any_ge (vector bool int, vector signed int);
8620 int vec_any_ge (vector float, vector float);
8622 int vec_any_gt (vector bool char, vector unsigned char);
8623 int vec_any_gt (vector unsigned char, vector bool char);
8624 int vec_any_gt (vector unsigned char, vector unsigned char);
8625 int vec_any_gt (vector bool char, vector signed char);
8626 int vec_any_gt (vector signed char, vector bool char);
8627 int vec_any_gt (vector signed char, vector signed char);
8628 int vec_any_gt (vector bool short, vector unsigned short);
8629 int vec_any_gt (vector unsigned short, vector bool short);
8630 int vec_any_gt (vector unsigned short, vector unsigned short);
8631 int vec_any_gt (vector bool short, vector signed short);
8632 int vec_any_gt (vector signed short, vector bool short);
8633 int vec_any_gt (vector signed short, vector signed short);
8634 int vec_any_gt (vector bool int, vector unsigned int);
8635 int vec_any_gt (vector unsigned int, vector bool int);
8636 int vec_any_gt (vector unsigned int, vector unsigned int);
8637 int vec_any_gt (vector bool int, vector signed int);
8638 int vec_any_gt (vector signed int, vector bool int);
8639 int vec_any_gt (vector signed int, vector signed int);
8640 int vec_any_gt (vector float, vector float);
8642 int vec_any_le (vector bool char, vector unsigned char);
8643 int vec_any_le (vector unsigned char, vector bool char);
8644 int vec_any_le (vector unsigned char, vector unsigned char);
8645 int vec_any_le (vector bool char, vector signed char);
8646 int vec_any_le (vector signed char, vector bool char);
8647 int vec_any_le (vector signed char, vector signed char);
8648 int vec_any_le (vector bool short, vector unsigned short);
8649 int vec_any_le (vector unsigned short, vector bool short);
8650 int vec_any_le (vector unsigned short, vector unsigned short);
8651 int vec_any_le (vector bool short, vector signed short);
8652 int vec_any_le (vector signed short, vector bool short);
8653 int vec_any_le (vector signed short, vector signed short);
8654 int vec_any_le (vector bool int, vector unsigned int);
8655 int vec_any_le (vector unsigned int, vector bool int);
8656 int vec_any_le (vector unsigned int, vector unsigned int);
8657 int vec_any_le (vector bool int, vector signed int);
8658 int vec_any_le (vector signed int, vector bool int);
8659 int vec_any_le (vector signed int, vector signed int);
8660 int vec_any_le (vector float, vector float);
8662 int vec_any_lt (vector bool char, vector unsigned char);
8663 int vec_any_lt (vector unsigned char, vector bool char);
8664 int vec_any_lt (vector unsigned char, vector unsigned char);
8665 int vec_any_lt (vector bool char, vector signed char);
8666 int vec_any_lt (vector signed char, vector bool char);
8667 int vec_any_lt (vector signed char, vector signed char);
8668 int vec_any_lt (vector bool short, vector unsigned short);
8669 int vec_any_lt (vector unsigned short, vector bool short);
8670 int vec_any_lt (vector unsigned short, vector unsigned short);
8671 int vec_any_lt (vector bool short, vector signed short);
8672 int vec_any_lt (vector signed short, vector bool short);
8673 int vec_any_lt (vector signed short, vector signed short);
8674 int vec_any_lt (vector bool int, vector unsigned int);
8675 int vec_any_lt (vector unsigned int, vector bool int);
8676 int vec_any_lt (vector unsigned int, vector unsigned int);
8677 int vec_any_lt (vector bool int, vector signed int);
8678 int vec_any_lt (vector signed int, vector bool int);
8679 int vec_any_lt (vector signed int, vector signed int);
8680 int vec_any_lt (vector float, vector float);
8682 int vec_any_nan (vector float);
8684 int vec_any_ne (vector signed char, vector bool char);
8685 int vec_any_ne (vector signed char, vector signed char);
8686 int vec_any_ne (vector unsigned char, vector bool char);
8687 int vec_any_ne (vector unsigned char, vector unsigned char);
8688 int vec_any_ne (vector bool char, vector bool char);
8689 int vec_any_ne (vector bool char, vector unsigned char);
8690 int vec_any_ne (vector bool char, vector signed char);
8691 int vec_any_ne (vector signed short, vector bool short);
8692 int vec_any_ne (vector signed short, vector signed short);
8693 int vec_any_ne (vector unsigned short, vector bool short);
8694 int vec_any_ne (vector unsigned short, vector unsigned short);
8695 int vec_any_ne (vector bool short, vector bool short);
8696 int vec_any_ne (vector bool short, vector unsigned short);
8697 int vec_any_ne (vector bool short, vector signed short);
8698 int vec_any_ne (vector pixel, vector pixel);
8699 int vec_any_ne (vector signed int, vector bool int);
8700 int vec_any_ne (vector signed int, vector signed int);
8701 int vec_any_ne (vector unsigned int, vector bool int);
8702 int vec_any_ne (vector unsigned int, vector unsigned int);
8703 int vec_any_ne (vector bool int, vector bool int);
8704 int vec_any_ne (vector bool int, vector unsigned int);
8705 int vec_any_ne (vector bool int, vector signed int);
8706 int vec_any_ne (vector float, vector float);
8708 int vec_any_nge (vector float, vector float);
8710 int vec_any_ngt (vector float, vector float);
8712 int vec_any_nle (vector float, vector float);
8714 int vec_any_nlt (vector float, vector float);
8716 int vec_any_numeric (vector float);
8718 int vec_any_out (vector float, vector float);
8721 @node SPARC VIS Built-in Functions
8722 @subsection SPARC VIS Built-in Functions
8724 GCC supports SIMD operations on the SPARC using both the generic vector
8725 extensions (@pxref{Vector Extensions}) as well as built-in functions for
8726 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
8727 switch, the VIS extension is exposed as the following built-in functions:
8730 typedef int v2si __attribute__ ((vector_size (8)));
8731 typedef short v4hi __attribute__ ((vector_size (8)));
8732 typedef short v2hi __attribute__ ((vector_size (4)));
8733 typedef char v8qi __attribute__ ((vector_size (8)));
8734 typedef char v4qi __attribute__ ((vector_size (4)));
8736 void * __builtin_vis_alignaddr (void *, long);
8737 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
8738 v2si __builtin_vis_faligndatav2si (v2si, v2si);
8739 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
8740 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
8742 v4hi __builtin_vis_fexpand (v4qi);
8744 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
8745 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
8746 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
8747 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
8748 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
8749 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
8750 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
8752 v4qi __builtin_vis_fpack16 (v4hi);
8753 v8qi __builtin_vis_fpack32 (v2si, v2si);
8754 v2hi __builtin_vis_fpackfix (v2si);
8755 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
8757 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
8760 @node Target Format Checks
8761 @section Format Checks Specific to Particular Target Machines
8763 For some target machines, GCC supports additional options to the
8765 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
8768 * Solaris Format Checks::
8771 @node Solaris Format Checks
8772 @subsection Solaris Format Checks
8774 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
8775 check. @code{cmn_err} accepts a subset of the standard @code{printf}
8776 conversions, and the two-argument @code{%b} conversion for displaying
8777 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
8780 @section Pragmas Accepted by GCC
8784 GCC supports several types of pragmas, primarily in order to compile
8785 code originally written for other compilers. Note that in general
8786 we do not recommend the use of pragmas; @xref{Function Attributes},
8787 for further explanation.
8791 * RS/6000 and PowerPC Pragmas::
8794 * Symbol-Renaming Pragmas::
8795 * Structure-Packing Pragmas::
8800 @subsection ARM Pragmas
8802 The ARM target defines pragmas for controlling the default addition of
8803 @code{long_call} and @code{short_call} attributes to functions.
8804 @xref{Function Attributes}, for information about the effects of these
8809 @cindex pragma, long_calls
8810 Set all subsequent functions to have the @code{long_call} attribute.
8813 @cindex pragma, no_long_calls
8814 Set all subsequent functions to have the @code{short_call} attribute.
8816 @item long_calls_off
8817 @cindex pragma, long_calls_off
8818 Do not affect the @code{long_call} or @code{short_call} attributes of
8819 subsequent functions.
8822 @node RS/6000 and PowerPC Pragmas
8823 @subsection RS/6000 and PowerPC Pragmas
8825 The RS/6000 and PowerPC targets define one pragma for controlling
8826 whether or not the @code{longcall} attribute is added to function
8827 declarations by default. This pragma overrides the @option{-mlongcall}
8828 option, but not the @code{longcall} and @code{shortcall} attributes.
8829 @xref{RS/6000 and PowerPC Options}, for more information about when long
8830 calls are and are not necessary.
8834 @cindex pragma, longcall
8835 Apply the @code{longcall} attribute to all subsequent function
8839 Do not apply the @code{longcall} attribute to subsequent function
8843 @c Describe c4x pragmas here.
8844 @c Describe h8300 pragmas here.
8845 @c Describe sh pragmas here.
8846 @c Describe v850 pragmas here.
8848 @node Darwin Pragmas
8849 @subsection Darwin Pragmas
8851 The following pragmas are available for all architectures running the
8852 Darwin operating system. These are useful for compatibility with other
8856 @item mark @var{tokens}@dots{}
8857 @cindex pragma, mark
8858 This pragma is accepted, but has no effect.
8860 @item options align=@var{alignment}
8861 @cindex pragma, options align
8862 This pragma sets the alignment of fields in structures. The values of
8863 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
8864 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
8865 properly; to restore the previous setting, use @code{reset} for the
8868 @item segment @var{tokens}@dots{}
8869 @cindex pragma, segment
8870 This pragma is accepted, but has no effect.
8872 @item unused (@var{var} [, @var{var}]@dots{})
8873 @cindex pragma, unused
8874 This pragma declares variables to be possibly unused. GCC will not
8875 produce warnings for the listed variables. The effect is similar to
8876 that of the @code{unused} attribute, except that this pragma may appear
8877 anywhere within the variables' scopes.
8880 @node Solaris Pragmas
8881 @subsection Solaris Pragmas
8883 The Solaris target supports @code{#pragma redefine_extname}
8884 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
8885 @code{#pragma} directives for compatibility with the system compiler.
8888 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
8889 @cindex pragma, align
8891 Increase the minimum alignment of each @var{variable} to @var{alignment}.
8892 This is the same as GCC's @code{aligned} attribute @pxref{Variable
8893 Attributes}). Macro expansion occurs on the arguments to this pragma
8894 when compiling C and Objective-C. It does not currently occur when
8895 compiling C++, but this is a bug which may be fixed in a future
8898 @item fini (@var{function} [, @var{function}]...)
8899 @cindex pragma, fini
8901 This pragma causes each listed @var{function} to be called after
8902 main, or during shared module unloading, by adding a call to the
8903 @code{.fini} section.
8905 @item init (@var{function} [, @var{function}]...)
8906 @cindex pragma, init
8908 This pragma causes each listed @var{function} to be called during
8909 initialization (before @code{main}) or during shared module loading, by
8910 adding a call to the @code{.init} section.
8914 @node Symbol-Renaming Pragmas
8915 @subsection Symbol-Renaming Pragmas
8917 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
8918 supports two @code{#pragma} directives which change the name used in
8919 assembly for a given declaration. These pragmas are only available on
8920 platforms whose system headers need them. To get this effect on all
8921 platforms supported by GCC, use the asm labels extension (@pxref{Asm
8925 @item redefine_extname @var{oldname} @var{newname}
8926 @cindex pragma, redefine_extname
8928 This pragma gives the C function @var{oldname} the assembly symbol
8929 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
8930 will be defined if this pragma is available (currently only on
8933 @item extern_prefix @var{string}
8934 @cindex pragma, extern_prefix
8936 This pragma causes all subsequent external function and variable
8937 declarations to have @var{string} prepended to their assembly symbols.
8938 This effect may be terminated with another @code{extern_prefix} pragma
8939 whose argument is an empty string. The preprocessor macro
8940 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
8941 available (currently only on Tru64 UNIX)@.
8944 These pragmas and the asm labels extension interact in a complicated
8945 manner. Here are some corner cases you may want to be aware of.
8948 @item Both pragmas silently apply only to declarations with external
8949 linkage. Asm labels do not have this restriction.
8951 @item In C++, both pragmas silently apply only to declarations with
8952 ``C'' linkage. Again, asm labels do not have this restriction.
8954 @item If any of the three ways of changing the assembly name of a
8955 declaration is applied to a declaration whose assembly name has
8956 already been determined (either by a previous use of one of these
8957 features, or because the compiler needed the assembly name in order to
8958 generate code), and the new name is different, a warning issues and
8959 the name does not change.
8961 @item The @var{oldname} used by @code{#pragma redefine_extname} is
8962 always the C-language name.
8964 @item If @code{#pragma extern_prefix} is in effect, and a declaration
8965 occurs with an asm label attached, the prefix is silently ignored for
8968 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
8969 apply to the same declaration, whichever triggered first wins, and a
8970 warning issues if they contradict each other. (We would like to have
8971 @code{#pragma redefine_extname} always win, for consistency with asm
8972 labels, but if @code{#pragma extern_prefix} triggers first we have no
8973 way of knowing that that happened.)
8976 @node Structure-Packing Pragmas
8977 @subsection Structure-Packing Pragmas
8979 For compatibility with Win32, GCC supports a set of @code{#pragma}
8980 directives which change the maximum alignment of members of structures,
8981 unions, and classes subsequently defined. The @var{n} value below always
8982 is required to be a small power of two and specifies the new alignment
8986 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
8987 @item @code{#pragma pack()} sets the alignment to the one that was in
8988 effect when compilation started (see also command line option
8989 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
8990 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
8991 setting on an internal stack and then optionally sets the new alignment.
8992 @item @code{#pragma pack(pop)} restores the alignment setting to the one
8993 saved at the top of the internal stack (and removes that stack entry).
8994 Note that @code{#pragma pack([@var{n}])} does not influence this internal
8995 stack; thus it is possible to have @code{#pragma pack(push)} followed by
8996 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
8997 @code{#pragma pack(pop)}.
9001 @subsection Weak Pragmas
9003 For compatibility with SVR4, GCC supports a set of @code{#pragma}
9004 directives for declaring symbols to be weak, and defining weak
9008 @item #pragma weak @var{symbol}
9009 @cindex pragma, weak
9010 This pragma declares @var{symbol} to be weak, as if the declaration
9011 had the attribute of the same name. The pragma may appear before
9012 or after the declaration of @var{symbol}, but must appear before
9013 either its first use or its definition. It is not an error for
9014 @var{symbol} to never be defined at all.
9016 @item #pragma weak @var{symbol1} = @var{symbol2}
9017 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
9018 It is an error if @var{symbol2} is not defined in the current
9022 @node Unnamed Fields
9023 @section Unnamed struct/union fields within structs/unions
9027 For compatibility with other compilers, GCC allows you to define
9028 a structure or union that contains, as fields, structures and unions
9029 without names. For example:
9042 In this example, the user would be able to access members of the unnamed
9043 union with code like @samp{foo.b}. Note that only unnamed structs and
9044 unions are allowed, you may not have, for example, an unnamed
9047 You must never create such structures that cause ambiguous field definitions.
9048 For example, this structure:
9059 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
9060 Such constructs are not supported and must be avoided. In the future,
9061 such constructs may be detected and treated as compilation errors.
9063 @opindex fms-extensions
9064 Unless @option{-fms-extensions} is used, the unnamed field must be a
9065 structure or union definition without a tag (for example, @samp{struct
9066 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
9067 also be a definition with a tag such as @samp{struct foo @{ int a;
9068 @};}, a reference to a previously defined structure or union such as
9069 @samp{struct foo;}, or a reference to a @code{typedef} name for a
9070 previously defined structure or union type.
9073 @section Thread-Local Storage
9074 @cindex Thread-Local Storage
9075 @cindex @acronym{TLS}
9078 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
9079 are allocated such that there is one instance of the variable per extant
9080 thread. The run-time model GCC uses to implement this originates
9081 in the IA-64 processor-specific ABI, but has since been migrated
9082 to other processors as well. It requires significant support from
9083 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
9084 system libraries (@file{libc.so} and @file{libpthread.so}), so it
9085 is not available everywhere.
9087 At the user level, the extension is visible with a new storage
9088 class keyword: @code{__thread}. For example:
9092 extern __thread struct state s;
9093 static __thread char *p;
9096 The @code{__thread} specifier may be used alone, with the @code{extern}
9097 or @code{static} specifiers, but with no other storage class specifier.
9098 When used with @code{extern} or @code{static}, @code{__thread} must appear
9099 immediately after the other storage class specifier.
9101 The @code{__thread} specifier may be applied to any global, file-scoped
9102 static, function-scoped static, or static data member of a class. It may
9103 not be applied to block-scoped automatic or non-static data member.
9105 When the address-of operator is applied to a thread-local variable, it is
9106 evaluated at run-time and returns the address of the current thread's
9107 instance of that variable. An address so obtained may be used by any
9108 thread. When a thread terminates, any pointers to thread-local variables
9109 in that thread become invalid.
9111 No static initialization may refer to the address of a thread-local variable.
9113 In C++, if an initializer is present for a thread-local variable, it must
9114 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
9117 See @uref{http://people.redhat.com/drepper/tls.pdf,
9118 ELF Handling For Thread-Local Storage} for a detailed explanation of
9119 the four thread-local storage addressing models, and how the run-time
9120 is expected to function.
9123 * C99 Thread-Local Edits::
9124 * C++98 Thread-Local Edits::
9127 @node C99 Thread-Local Edits
9128 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
9130 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
9131 that document the exact semantics of the language extension.
9135 @cite{5.1.2 Execution environments}
9137 Add new text after paragraph 1
9140 Within either execution environment, a @dfn{thread} is a flow of
9141 control within a program. It is implementation defined whether
9142 or not there may be more than one thread associated with a program.
9143 It is implementation defined how threads beyond the first are
9144 created, the name and type of the function called at thread
9145 startup, and how threads may be terminated. However, objects
9146 with thread storage duration shall be initialized before thread
9151 @cite{6.2.4 Storage durations of objects}
9153 Add new text before paragraph 3
9156 An object whose identifier is declared with the storage-class
9157 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
9158 Its lifetime is the entire execution of the thread, and its
9159 stored value is initialized only once, prior to thread startup.
9163 @cite{6.4.1 Keywords}
9165 Add @code{__thread}.
9168 @cite{6.7.1 Storage-class specifiers}
9170 Add @code{__thread} to the list of storage class specifiers in
9173 Change paragraph 2 to
9176 With the exception of @code{__thread}, at most one storage-class
9177 specifier may be given [@dots{}]. The @code{__thread} specifier may
9178 be used alone, or immediately following @code{extern} or
9182 Add new text after paragraph 6
9185 The declaration of an identifier for a variable that has
9186 block scope that specifies @code{__thread} shall also
9187 specify either @code{extern} or @code{static}.
9189 The @code{__thread} specifier shall be used only with
9194 @node C++98 Thread-Local Edits
9195 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
9197 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
9198 that document the exact semantics of the language extension.
9202 @b{[intro.execution]}
9204 New text after paragraph 4
9207 A @dfn{thread} is a flow of control within the abstract machine.
9208 It is implementation defined whether or not there may be more than
9212 New text after paragraph 7
9215 It is unspecified whether additional action must be taken to
9216 ensure when and whether side effects are visible to other threads.
9222 Add @code{__thread}.
9225 @b{[basic.start.main]}
9227 Add after paragraph 5
9230 The thread that begins execution at the @code{main} function is called
9231 the @dfn{main thread}. It is implementation defined how functions
9232 beginning threads other than the main thread are designated or typed.
9233 A function so designated, as well as the @code{main} function, is called
9234 a @dfn{thread startup function}. It is implementation defined what
9235 happens if a thread startup function returns. It is implementation
9236 defined what happens to other threads when any thread calls @code{exit}.
9240 @b{[basic.start.init]}
9242 Add after paragraph 4
9245 The storage for an object of thread storage duration shall be
9246 statically initialized before the first statement of the thread startup
9247 function. An object of thread storage duration shall not require
9248 dynamic initialization.
9252 @b{[basic.start.term]}
9254 Add after paragraph 3
9257 The type of an object with thread storage duration shall not have a
9258 non-trivial destructor, nor shall it be an array type whose elements
9259 (directly or indirectly) have non-trivial destructors.
9265 Add ``thread storage duration'' to the list in paragraph 1.
9270 Thread, static, and automatic storage durations are associated with
9271 objects introduced by declarations [@dots{}].
9274 Add @code{__thread} to the list of specifiers in paragraph 3.
9277 @b{[basic.stc.thread]}
9279 New section before @b{[basic.stc.static]}
9282 The keyword @code{__thread} applied to a non-local object gives the
9283 object thread storage duration.
9285 A local variable or class data member declared both @code{static}
9286 and @code{__thread} gives the variable or member thread storage
9291 @b{[basic.stc.static]}
9296 All objects which have neither thread storage duration, dynamic
9297 storage duration nor are local [@dots{}].
9303 Add @code{__thread} to the list in paragraph 1.
9308 With the exception of @code{__thread}, at most one
9309 @var{storage-class-specifier} shall appear in a given
9310 @var{decl-specifier-seq}. The @code{__thread} specifier may
9311 be used alone, or immediately following the @code{extern} or
9312 @code{static} specifiers. [@dots{}]
9315 Add after paragraph 5
9318 The @code{__thread} specifier can be applied only to the names of objects
9319 and to anonymous unions.
9325 Add after paragraph 6
9328 Non-@code{static} members shall not be @code{__thread}.
9332 @node C++ Extensions
9333 @chapter Extensions to the C++ Language
9334 @cindex extensions, C++ language
9335 @cindex C++ language extensions
9337 The GNU compiler provides these extensions to the C++ language (and you
9338 can also use most of the C language extensions in your C++ programs). If you
9339 want to write code that checks whether these features are available, you can
9340 test for the GNU compiler the same way as for C programs: check for a
9341 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
9342 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
9343 Predefined Macros,cpp,The GNU C Preprocessor}).
9346 * Volatiles:: What constitutes an access to a volatile object.
9347 * Restricted Pointers:: C99 restricted pointers and references.
9348 * Vague Linkage:: Where G++ puts inlines, vtables and such.
9349 * C++ Interface:: You can use a single C++ header file for both
9350 declarations and definitions.
9351 * Template Instantiation:: Methods for ensuring that exactly one copy of
9352 each needed template instantiation is emitted.
9353 * Bound member functions:: You can extract a function pointer to the
9354 method denoted by a @samp{->*} or @samp{.*} expression.
9355 * C++ Attributes:: Variable, function, and type attributes for C++ only.
9356 * Strong Using:: Strong using-directives for namespace composition.
9357 * Java Exceptions:: Tweaking exception handling to work with Java.
9358 * Deprecated Features:: Things will disappear from g++.
9359 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
9363 @section When is a Volatile Object Accessed?
9364 @cindex accessing volatiles
9365 @cindex volatile read
9366 @cindex volatile write
9367 @cindex volatile access
9369 Both the C and C++ standard have the concept of volatile objects. These
9370 are normally accessed by pointers and used for accessing hardware. The
9371 standards encourage compilers to refrain from optimizations
9372 concerning accesses to volatile objects that it might perform on
9373 non-volatile objects. The C standard leaves it implementation defined
9374 as to what constitutes a volatile access. The C++ standard omits to
9375 specify this, except to say that C++ should behave in a similar manner
9376 to C with respect to volatiles, where possible. The minimum either
9377 standard specifies is that at a sequence point all previous accesses to
9378 volatile objects have stabilized and no subsequent accesses have
9379 occurred. Thus an implementation is free to reorder and combine
9380 volatile accesses which occur between sequence points, but cannot do so
9381 for accesses across a sequence point. The use of volatiles does not
9382 allow you to violate the restriction on updating objects multiple times
9383 within a sequence point.
9385 In most expressions, it is intuitively obvious what is a read and what is
9386 a write. For instance
9389 volatile int *dst = @var{somevalue};
9390 volatile int *src = @var{someothervalue};
9395 will cause a read of the volatile object pointed to by @var{src} and stores the
9396 value into the volatile object pointed to by @var{dst}. There is no
9397 guarantee that these reads and writes are atomic, especially for objects
9398 larger than @code{int}.
9400 Less obvious expressions are where something which looks like an access
9401 is used in a void context. An example would be,
9404 volatile int *src = @var{somevalue};
9408 With C, such expressions are rvalues, and as rvalues cause a read of
9409 the object, GCC interprets this as a read of the volatile being pointed
9410 to. The C++ standard specifies that such expressions do not undergo
9411 lvalue to rvalue conversion, and that the type of the dereferenced
9412 object may be incomplete. The C++ standard does not specify explicitly
9413 that it is this lvalue to rvalue conversion which is responsible for
9414 causing an access. However, there is reason to believe that it is,
9415 because otherwise certain simple expressions become undefined. However,
9416 because it would surprise most programmers, G++ treats dereferencing a
9417 pointer to volatile object of complete type in a void context as a read
9418 of the object. When the object has incomplete type, G++ issues a
9423 struct T @{int m;@};
9424 volatile S *ptr1 = @var{somevalue};
9425 volatile T *ptr2 = @var{somevalue};
9430 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
9431 causes a read of the object pointed to. If you wish to force an error on
9432 the first case, you must force a conversion to rvalue with, for instance
9433 a static cast, @code{static_cast<S>(*ptr1)}.
9435 When using a reference to volatile, G++ does not treat equivalent
9436 expressions as accesses to volatiles, but instead issues a warning that
9437 no volatile is accessed. The rationale for this is that otherwise it
9438 becomes difficult to determine where volatile access occur, and not
9439 possible to ignore the return value from functions returning volatile
9440 references. Again, if you wish to force a read, cast the reference to
9443 @node Restricted Pointers
9444 @section Restricting Pointer Aliasing
9445 @cindex restricted pointers
9446 @cindex restricted references
9447 @cindex restricted this pointer
9449 As with the C front end, G++ understands the C99 feature of restricted pointers,
9450 specified with the @code{__restrict__}, or @code{__restrict} type
9451 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
9452 language flag, @code{restrict} is not a keyword in C++.
9454 In addition to allowing restricted pointers, you can specify restricted
9455 references, which indicate that the reference is not aliased in the local
9459 void fn (int *__restrict__ rptr, int &__restrict__ rref)
9466 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
9467 @var{rref} refers to a (different) unaliased integer.
9469 You may also specify whether a member function's @var{this} pointer is
9470 unaliased by using @code{__restrict__} as a member function qualifier.
9473 void T::fn () __restrict__
9480 Within the body of @code{T::fn}, @var{this} will have the effective
9481 definition @code{T *__restrict__ const this}. Notice that the
9482 interpretation of a @code{__restrict__} member function qualifier is
9483 different to that of @code{const} or @code{volatile} qualifier, in that it
9484 is applied to the pointer rather than the object. This is consistent with
9485 other compilers which implement restricted pointers.
9487 As with all outermost parameter qualifiers, @code{__restrict__} is
9488 ignored in function definition matching. This means you only need to
9489 specify @code{__restrict__} in a function definition, rather than
9490 in a function prototype as well.
9493 @section Vague Linkage
9494 @cindex vague linkage
9496 There are several constructs in C++ which require space in the object
9497 file but are not clearly tied to a single translation unit. We say that
9498 these constructs have ``vague linkage''. Typically such constructs are
9499 emitted wherever they are needed, though sometimes we can be more
9503 @item Inline Functions
9504 Inline functions are typically defined in a header file which can be
9505 included in many different compilations. Hopefully they can usually be
9506 inlined, but sometimes an out-of-line copy is necessary, if the address
9507 of the function is taken or if inlining fails. In general, we emit an
9508 out-of-line copy in all translation units where one is needed. As an
9509 exception, we only emit inline virtual functions with the vtable, since
9510 it will always require a copy.
9512 Local static variables and string constants used in an inline function
9513 are also considered to have vague linkage, since they must be shared
9514 between all inlined and out-of-line instances of the function.
9518 C++ virtual functions are implemented in most compilers using a lookup
9519 table, known as a vtable. The vtable contains pointers to the virtual
9520 functions provided by a class, and each object of the class contains a
9521 pointer to its vtable (or vtables, in some multiple-inheritance
9522 situations). If the class declares any non-inline, non-pure virtual
9523 functions, the first one is chosen as the ``key method'' for the class,
9524 and the vtable is only emitted in the translation unit where the key
9527 @emph{Note:} If the chosen key method is later defined as inline, the
9528 vtable will still be emitted in every translation unit which defines it.
9529 Make sure that any inline virtuals are declared inline in the class
9530 body, even if they are not defined there.
9532 @item type_info objects
9535 C++ requires information about types to be written out in order to
9536 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
9537 For polymorphic classes (classes with virtual functions), the type_info
9538 object is written out along with the vtable so that @samp{dynamic_cast}
9539 can determine the dynamic type of a class object at runtime. For all
9540 other types, we write out the type_info object when it is used: when
9541 applying @samp{typeid} to an expression, throwing an object, or
9542 referring to a type in a catch clause or exception specification.
9544 @item Template Instantiations
9545 Most everything in this section also applies to template instantiations,
9546 but there are other options as well.
9547 @xref{Template Instantiation,,Where's the Template?}.
9551 When used with GNU ld version 2.8 or later on an ELF system such as
9552 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
9553 these constructs will be discarded at link time. This is known as
9556 On targets that don't support COMDAT, but do support weak symbols, GCC
9557 will use them. This way one copy will override all the others, but
9558 the unused copies will still take up space in the executable.
9560 For targets which do not support either COMDAT or weak symbols,
9561 most entities with vague linkage will be emitted as local symbols to
9562 avoid duplicate definition errors from the linker. This will not happen
9563 for local statics in inlines, however, as having multiple copies will
9564 almost certainly break things.
9566 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
9567 another way to control placement of these constructs.
9570 @section #pragma interface and implementation
9572 @cindex interface and implementation headers, C++
9573 @cindex C++ interface and implementation headers
9574 @cindex pragmas, interface and implementation
9576 @code{#pragma interface} and @code{#pragma implementation} provide the
9577 user with a way of explicitly directing the compiler to emit entities
9578 with vague linkage (and debugging information) in a particular
9581 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
9582 most cases, because of COMDAT support and the ``key method'' heuristic
9583 mentioned in @ref{Vague Linkage}. Using them can actually cause your
9584 program to grow due to unnecessary out-of-line copies of inline
9585 functions. Currently (3.4) the only benefit of these
9586 @code{#pragma}s is reduced duplication of debugging information, and
9587 that should be addressed soon on DWARF 2 targets with the use of
9591 @item #pragma interface
9592 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
9593 @kindex #pragma interface
9594 Use this directive in @emph{header files} that define object classes, to save
9595 space in most of the object files that use those classes. Normally,
9596 local copies of certain information (backup copies of inline member
9597 functions, debugging information, and the internal tables that implement
9598 virtual functions) must be kept in each object file that includes class
9599 definitions. You can use this pragma to avoid such duplication. When a
9600 header file containing @samp{#pragma interface} is included in a
9601 compilation, this auxiliary information will not be generated (unless
9602 the main input source file itself uses @samp{#pragma implementation}).
9603 Instead, the object files will contain references to be resolved at link
9606 The second form of this directive is useful for the case where you have
9607 multiple headers with the same name in different directories. If you
9608 use this form, you must specify the same string to @samp{#pragma
9611 @item #pragma implementation
9612 @itemx #pragma implementation "@var{objects}.h"
9613 @kindex #pragma implementation
9614 Use this pragma in a @emph{main input file}, when you want full output from
9615 included header files to be generated (and made globally visible). The
9616 included header file, in turn, should use @samp{#pragma interface}.
9617 Backup copies of inline member functions, debugging information, and the
9618 internal tables used to implement virtual functions are all generated in
9619 implementation files.
9621 @cindex implied @code{#pragma implementation}
9622 @cindex @code{#pragma implementation}, implied
9623 @cindex naming convention, implementation headers
9624 If you use @samp{#pragma implementation} with no argument, it applies to
9625 an include file with the same basename@footnote{A file's @dfn{basename}
9626 was the name stripped of all leading path information and of trailing
9627 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
9628 file. For example, in @file{allclass.cc}, giving just
9629 @samp{#pragma implementation}
9630 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
9632 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
9633 an implementation file whenever you would include it from
9634 @file{allclass.cc} even if you never specified @samp{#pragma
9635 implementation}. This was deemed to be more trouble than it was worth,
9636 however, and disabled.
9638 Use the string argument if you want a single implementation file to
9639 include code from multiple header files. (You must also use
9640 @samp{#include} to include the header file; @samp{#pragma
9641 implementation} only specifies how to use the file---it doesn't actually
9644 There is no way to split up the contents of a single header file into
9645 multiple implementation files.
9648 @cindex inlining and C++ pragmas
9649 @cindex C++ pragmas, effect on inlining
9650 @cindex pragmas in C++, effect on inlining
9651 @samp{#pragma implementation} and @samp{#pragma interface} also have an
9652 effect on function inlining.
9654 If you define a class in a header file marked with @samp{#pragma
9655 interface}, the effect on an inline function defined in that class is
9656 similar to an explicit @code{extern} declaration---the compiler emits
9657 no code at all to define an independent version of the function. Its
9658 definition is used only for inlining with its callers.
9660 @opindex fno-implement-inlines
9661 Conversely, when you include the same header file in a main source file
9662 that declares it as @samp{#pragma implementation}, the compiler emits
9663 code for the function itself; this defines a version of the function
9664 that can be found via pointers (or by callers compiled without
9665 inlining). If all calls to the function can be inlined, you can avoid
9666 emitting the function by compiling with @option{-fno-implement-inlines}.
9667 If any calls were not inlined, you will get linker errors.
9669 @node Template Instantiation
9670 @section Where's the Template?
9671 @cindex template instantiation
9673 C++ templates are the first language feature to require more
9674 intelligence from the environment than one usually finds on a UNIX
9675 system. Somehow the compiler and linker have to make sure that each
9676 template instance occurs exactly once in the executable if it is needed,
9677 and not at all otherwise. There are two basic approaches to this
9678 problem, which are referred to as the Borland model and the Cfront model.
9682 Borland C++ solved the template instantiation problem by adding the code
9683 equivalent of common blocks to their linker; the compiler emits template
9684 instances in each translation unit that uses them, and the linker
9685 collapses them together. The advantage of this model is that the linker
9686 only has to consider the object files themselves; there is no external
9687 complexity to worry about. This disadvantage is that compilation time
9688 is increased because the template code is being compiled repeatedly.
9689 Code written for this model tends to include definitions of all
9690 templates in the header file, since they must be seen to be
9694 The AT&T C++ translator, Cfront, solved the template instantiation
9695 problem by creating the notion of a template repository, an
9696 automatically maintained place where template instances are stored. A
9697 more modern version of the repository works as follows: As individual
9698 object files are built, the compiler places any template definitions and
9699 instantiations encountered in the repository. At link time, the link
9700 wrapper adds in the objects in the repository and compiles any needed
9701 instances that were not previously emitted. The advantages of this
9702 model are more optimal compilation speed and the ability to use the
9703 system linker; to implement the Borland model a compiler vendor also
9704 needs to replace the linker. The disadvantages are vastly increased
9705 complexity, and thus potential for error; for some code this can be
9706 just as transparent, but in practice it can been very difficult to build
9707 multiple programs in one directory and one program in multiple
9708 directories. Code written for this model tends to separate definitions
9709 of non-inline member templates into a separate file, which should be
9710 compiled separately.
9713 When used with GNU ld version 2.8 or later on an ELF system such as
9714 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
9715 Borland model. On other systems, G++ implements neither automatic
9718 A future version of G++ will support a hybrid model whereby the compiler
9719 will emit any instantiations for which the template definition is
9720 included in the compile, and store template definitions and
9721 instantiation context information into the object file for the rest.
9722 The link wrapper will extract that information as necessary and invoke
9723 the compiler to produce the remaining instantiations. The linker will
9724 then combine duplicate instantiations.
9726 In the mean time, you have the following options for dealing with
9727 template instantiations:
9732 Compile your template-using code with @option{-frepo}. The compiler will
9733 generate files with the extension @samp{.rpo} listing all of the
9734 template instantiations used in the corresponding object files which
9735 could be instantiated there; the link wrapper, @samp{collect2}, will
9736 then update the @samp{.rpo} files to tell the compiler where to place
9737 those instantiations and rebuild any affected object files. The
9738 link-time overhead is negligible after the first pass, as the compiler
9739 will continue to place the instantiations in the same files.
9741 This is your best option for application code written for the Borland
9742 model, as it will just work. Code written for the Cfront model will
9743 need to be modified so that the template definitions are available at
9744 one or more points of instantiation; usually this is as simple as adding
9745 @code{#include <tmethods.cc>} to the end of each template header.
9747 For library code, if you want the library to provide all of the template
9748 instantiations it needs, just try to link all of its object files
9749 together; the link will fail, but cause the instantiations to be
9750 generated as a side effect. Be warned, however, that this may cause
9751 conflicts if multiple libraries try to provide the same instantiations.
9752 For greater control, use explicit instantiation as described in the next
9756 @opindex fno-implicit-templates
9757 Compile your code with @option{-fno-implicit-templates} to disable the
9758 implicit generation of template instances, and explicitly instantiate
9759 all the ones you use. This approach requires more knowledge of exactly
9760 which instances you need than do the others, but it's less
9761 mysterious and allows greater control. You can scatter the explicit
9762 instantiations throughout your program, perhaps putting them in the
9763 translation units where the instances are used or the translation units
9764 that define the templates themselves; you can put all of the explicit
9765 instantiations you need into one big file; or you can create small files
9772 template class Foo<int>;
9773 template ostream& operator <<
9774 (ostream&, const Foo<int>&);
9777 for each of the instances you need, and create a template instantiation
9780 If you are using Cfront-model code, you can probably get away with not
9781 using @option{-fno-implicit-templates} when compiling files that don't
9782 @samp{#include} the member template definitions.
9784 If you use one big file to do the instantiations, you may want to
9785 compile it without @option{-fno-implicit-templates} so you get all of the
9786 instances required by your explicit instantiations (but not by any
9787 other files) without having to specify them as well.
9789 G++ has extended the template instantiation syntax given in the ISO
9790 standard to allow forward declaration of explicit instantiations
9791 (with @code{extern}), instantiation of the compiler support data for a
9792 template class (i.e.@: the vtable) without instantiating any of its
9793 members (with @code{inline}), and instantiation of only the static data
9794 members of a template class, without the support data or member
9795 functions (with (@code{static}):
9798 extern template int max (int, int);
9799 inline template class Foo<int>;
9800 static template class Foo<int>;
9804 Do nothing. Pretend G++ does implement automatic instantiation
9805 management. Code written for the Borland model will work fine, but
9806 each translation unit will contain instances of each of the templates it
9807 uses. In a large program, this can lead to an unacceptable amount of code
9811 @node Bound member functions
9812 @section Extracting the function pointer from a bound pointer to member function
9814 @cindex pointer to member function
9815 @cindex bound pointer to member function
9817 In C++, pointer to member functions (PMFs) are implemented using a wide
9818 pointer of sorts to handle all the possible call mechanisms; the PMF
9819 needs to store information about how to adjust the @samp{this} pointer,
9820 and if the function pointed to is virtual, where to find the vtable, and
9821 where in the vtable to look for the member function. If you are using
9822 PMFs in an inner loop, you should really reconsider that decision. If
9823 that is not an option, you can extract the pointer to the function that
9824 would be called for a given object/PMF pair and call it directly inside
9825 the inner loop, to save a bit of time.
9827 Note that you will still be paying the penalty for the call through a
9828 function pointer; on most modern architectures, such a call defeats the
9829 branch prediction features of the CPU@. This is also true of normal
9830 virtual function calls.
9832 The syntax for this extension is
9836 extern int (A::*fp)();
9837 typedef int (*fptr)(A *);
9839 fptr p = (fptr)(a.*fp);
9842 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
9843 no object is needed to obtain the address of the function. They can be
9844 converted to function pointers directly:
9847 fptr p1 = (fptr)(&A::foo);
9850 @opindex Wno-pmf-conversions
9851 You must specify @option{-Wno-pmf-conversions} to use this extension.
9853 @node C++ Attributes
9854 @section C++-Specific Variable, Function, and Type Attributes
9856 Some attributes only make sense for C++ programs.
9859 @item init_priority (@var{priority})
9860 @cindex init_priority attribute
9863 In Standard C++, objects defined at namespace scope are guaranteed to be
9864 initialized in an order in strict accordance with that of their definitions
9865 @emph{in a given translation unit}. No guarantee is made for initializations
9866 across translation units. However, GNU C++ allows users to control the
9867 order of initialization of objects defined at namespace scope with the
9868 @code{init_priority} attribute by specifying a relative @var{priority},
9869 a constant integral expression currently bounded between 101 and 65535
9870 inclusive. Lower numbers indicate a higher priority.
9872 In the following example, @code{A} would normally be created before
9873 @code{B}, but the @code{init_priority} attribute has reversed that order:
9876 Some_Class A __attribute__ ((init_priority (2000)));
9877 Some_Class B __attribute__ ((init_priority (543)));
9881 Note that the particular values of @var{priority} do not matter; only their
9884 @item java_interface
9885 @cindex java_interface attribute
9887 This type attribute informs C++ that the class is a Java interface. It may
9888 only be applied to classes declared within an @code{extern "Java"} block.
9889 Calls to methods declared in this interface will be dispatched using GCJ's
9890 interface table mechanism, instead of regular virtual table dispatch.
9894 See also @xref{Strong Using}.
9897 @section Strong Using
9899 @strong{Caution:} The semantics of this extension are not fully
9900 defined. Users should refrain from using this extension as its
9901 semantics may change subtly over time. It is possible that this
9902 extension wil be removed in future versions of G++.
9904 A using-directive with @code{__attribute ((strong))} is stronger
9905 than a normal using-directive in two ways:
9909 Templates from the used namespace can be specialized as though they were members of the using namespace.
9912 The using namespace is considered an associated namespace of all
9913 templates in the used namespace for purposes of argument-dependent
9917 This is useful for composing a namespace transparently from
9918 implementation namespaces. For example:
9923 template <class T> struct A @{ @};
9925 using namespace debug __attribute ((__strong__));
9926 template <> struct A<int> @{ @}; // @r{ok to specialize}
9928 template <class T> void f (A<T>);
9933 f (std::A<float>()); // @r{lookup finds} std::f
9938 @node Java Exceptions
9939 @section Java Exceptions
9941 The Java language uses a slightly different exception handling model
9942 from C++. Normally, GNU C++ will automatically detect when you are
9943 writing C++ code that uses Java exceptions, and handle them
9944 appropriately. However, if C++ code only needs to execute destructors
9945 when Java exceptions are thrown through it, GCC will guess incorrectly.
9946 Sample problematic code is:
9949 struct S @{ ~S(); @};
9950 extern void bar(); // @r{is written in Java, and may throw exceptions}
9959 The usual effect of an incorrect guess is a link failure, complaining of
9960 a missing routine called @samp{__gxx_personality_v0}.
9962 You can inform the compiler that Java exceptions are to be used in a
9963 translation unit, irrespective of what it might think, by writing
9964 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
9965 @samp{#pragma} must appear before any functions that throw or catch
9966 exceptions, or run destructors when exceptions are thrown through them.
9968 You cannot mix Java and C++ exceptions in the same translation unit. It
9969 is believed to be safe to throw a C++ exception from one file through
9970 another file compiled for the Java exception model, or vice versa, but
9971 there may be bugs in this area.
9973 @node Deprecated Features
9974 @section Deprecated Features
9976 In the past, the GNU C++ compiler was extended to experiment with new
9977 features, at a time when the C++ language was still evolving. Now that
9978 the C++ standard is complete, some of those features are superseded by
9979 superior alternatives. Using the old features might cause a warning in
9980 some cases that the feature will be dropped in the future. In other
9981 cases, the feature might be gone already.
9983 While the list below is not exhaustive, it documents some of the options
9984 that are now deprecated:
9987 @item -fexternal-templates
9988 @itemx -falt-external-templates
9989 These are two of the many ways for G++ to implement template
9990 instantiation. @xref{Template Instantiation}. The C++ standard clearly
9991 defines how template definitions have to be organized across
9992 implementation units. G++ has an implicit instantiation mechanism that
9993 should work just fine for standard-conforming code.
9995 @item -fstrict-prototype
9996 @itemx -fno-strict-prototype
9997 Previously it was possible to use an empty prototype parameter list to
9998 indicate an unspecified number of parameters (like C), rather than no
9999 parameters, as C++ demands. This feature has been removed, except where
10000 it is required for backwards compatibility @xref{Backwards Compatibility}.
10003 G++ allows a virtual function returning @samp{void *} to be overridden
10004 by one returning a different pointer type. This extension to the
10005 covariant return type rules is now deprecated and will be removed from a
10008 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
10009 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
10010 and will be removed in a future version. Code using these operators
10011 should be modified to use @code{std::min} and @code{std::max} instead.
10013 The named return value extension has been deprecated, and is now
10016 The use of initializer lists with new expressions has been deprecated,
10017 and is now removed from G++.
10019 Floating and complex non-type template parameters have been deprecated,
10020 and are now removed from G++.
10022 The implicit typename extension has been deprecated and is now
10025 The use of default arguments in function pointers, function typedefs and
10026 and other places where they are not permitted by the standard is
10027 deprecated and will be removed from a future version of G++.
10029 G++ allows floating-point literals to appear in integral constant expressions,
10030 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
10031 This extension is deprecated and will be removed from a future version.
10033 G++ allows static data members of const floating-point type to be declared
10034 with an initializer in a class definition. The standard only allows
10035 initializers for static members of const integral types and const
10036 enumeration types so this extension has been deprecated and will be removed
10037 from a future version.
10039 @node Backwards Compatibility
10040 @section Backwards Compatibility
10041 @cindex Backwards Compatibility
10042 @cindex ARM [Annotated C++ Reference Manual]
10044 Now that there is a definitive ISO standard C++, G++ has a specification
10045 to adhere to. The C++ language evolved over time, and features that
10046 used to be acceptable in previous drafts of the standard, such as the ARM
10047 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
10048 compilation of C++ written to such drafts, G++ contains some backwards
10049 compatibilities. @emph{All such backwards compatibility features are
10050 liable to disappear in future versions of G++.} They should be considered
10051 deprecated @xref{Deprecated Features}.
10055 If a variable is declared at for scope, it used to remain in scope until
10056 the end of the scope which contained the for statement (rather than just
10057 within the for scope). G++ retains this, but issues a warning, if such a
10058 variable is accessed outside the for scope.
10060 @item Implicit C language
10061 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
10062 scope to set the language. On such systems, all header files are
10063 implicitly scoped inside a C language scope. Also, an empty prototype
10064 @code{()} will be treated as an unspecified number of arguments, rather
10065 than no arguments, as C++ demands.