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 * Other Builtins:: Other built-in functions.
74 * Target Builtins:: Built-in functions specific to particular targets.
75 * Target Format Checks:: Format checks specific to particular targets.
76 * Pragmas:: Pragmas accepted by GCC.
77 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
78 * Thread-Local:: Per-thread variables.
82 @section Statements and Declarations in Expressions
83 @cindex statements inside expressions
84 @cindex declarations inside expressions
85 @cindex expressions containing statements
86 @cindex macros, statements in expressions
88 @c the above section title wrapped and causes an underfull hbox.. i
89 @c changed it from "within" to "in". --mew 4feb93
90 A compound statement enclosed in parentheses may appear as an expression
91 in GNU C@. This allows you to use loops, switches, and local variables
94 Recall that a compound statement is a sequence of statements surrounded
95 by braces; in this construct, parentheses go around the braces. For
99 (@{ int y = foo (); int z;
106 is a valid (though slightly more complex than necessary) expression
107 for the absolute value of @code{foo ()}.
109 The last thing in the compound statement should be an expression
110 followed by a semicolon; the value of this subexpression serves as the
111 value of the entire construct. (If you use some other kind of statement
112 last within the braces, the construct has type @code{void}, and thus
113 effectively no value.)
115 This feature is especially useful in making macro definitions ``safe'' (so
116 that they evaluate each operand exactly once). For example, the
117 ``maximum'' function is commonly defined as a macro in standard C as
121 #define max(a,b) ((a) > (b) ? (a) : (b))
125 @cindex side effects, macro argument
126 But this definition computes either @var{a} or @var{b} twice, with bad
127 results if the operand has side effects. In GNU C, if you know the
128 type of the operands (here taken as @code{int}), you can define
129 the macro safely as follows:
132 #define maxint(a,b) \
133 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
136 Embedded statements are not allowed in constant expressions, such as
137 the value of an enumeration constant, the width of a bit-field, or
138 the initial value of a static variable.
140 If you don't know the type of the operand, you can still do this, but you
141 must use @code{typeof} (@pxref{Typeof}).
143 In G++, the result value of a statement expression undergoes array and
144 function pointer decay, and is returned by value to the enclosing
145 expression. For instance, if @code{A} is a class, then
154 will construct a temporary @code{A} object to hold the result of the
155 statement expression, and that will be used to invoke @code{Foo}.
156 Therefore the @code{this} pointer observed by @code{Foo} will not be the
159 Any temporaries created within a statement within a statement expression
160 will be destroyed at the statement's end. This makes statement
161 expressions inside macros slightly different from function calls. In
162 the latter case temporaries introduced during argument evaluation will
163 be destroyed at the end of the statement that includes the function
164 call. In the statement expression case they will be destroyed during
165 the statement expression. For instance,
168 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
169 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
179 will have different places where temporaries are destroyed. For the
180 @code{macro} case, the temporary @code{X} will be destroyed just after
181 the initialization of @code{b}. In the @code{function} case that
182 temporary will be destroyed when the function returns.
184 These considerations mean that it is probably a bad idea to use
185 statement-expressions of this form in header files that are designed to
186 work with C++. (Note that some versions of the GNU C Library contained
187 header files using statement-expression that lead to precisely this
190 Jumping into a statement expression with @code{goto} or using a
191 @code{switch} statement outside the statement expression with a
192 @code{case} or @code{default} label inside the statement expression is
193 not permitted. Jumping into a statement expression with a computed
194 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
195 Jumping out of a statement expression is permitted, but if the
196 statement expression is part of a larger expression then it is
197 unspecified which other subexpressions of that expression have been
198 evaluated except where the language definition requires certain
199 subexpressions to be evaluated before or after the statement
200 expression. In any case, as with a function call the evaluation of a
201 statement expression is not interleaved with the evaluation of other
202 parts of the containing expression. For example,
205 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
209 will call @code{foo} and @code{bar1} and will not call @code{baz} but
210 may or may not call @code{bar2}. If @code{bar2} is called, it will be
211 called after @code{foo} and before @code{bar1}
214 @section Locally Declared Labels
216 @cindex macros, local labels
218 GCC allows you to declare @dfn{local labels} in any nested block
219 scope. A local label is just like an ordinary label, but you can
220 only reference it (with a @code{goto} statement, or by taking its
221 address) within the block in which it was declared.
223 A local label declaration looks like this:
226 __label__ @var{label};
233 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
236 Local label declarations must come at the beginning of the block,
237 before any ordinary declarations or statements.
239 The label declaration defines the label @emph{name}, but does not define
240 the label itself. You must do this in the usual way, with
241 @code{@var{label}:}, within the statements of the statement expression.
243 The local label feature is useful for complex macros. If a macro
244 contains nested loops, a @code{goto} can be useful for breaking out of
245 them. However, an ordinary label whose scope is the whole function
246 cannot be used: if the macro can be expanded several times in one
247 function, the label will be multiply defined in that function. A
248 local label avoids this problem. For example:
251 #define SEARCH(value, array, target) \
254 typeof (target) _SEARCH_target = (target); \
255 typeof (*(array)) *_SEARCH_array = (array); \
258 for (i = 0; i < max; i++) \
259 for (j = 0; j < max; j++) \
260 if (_SEARCH_array[i][j] == _SEARCH_target) \
261 @{ (value) = i; goto found; @} \
267 This could also be written using a statement-expression:
270 #define SEARCH(array, target) \
273 typeof (target) _SEARCH_target = (target); \
274 typeof (*(array)) *_SEARCH_array = (array); \
277 for (i = 0; i < max; i++) \
278 for (j = 0; j < max; j++) \
279 if (_SEARCH_array[i][j] == _SEARCH_target) \
280 @{ value = i; goto found; @} \
287 Local label declarations also make the labels they declare visible to
288 nested functions, if there are any. @xref{Nested Functions}, for details.
290 @node Labels as Values
291 @section Labels as Values
292 @cindex labels as values
293 @cindex computed gotos
294 @cindex goto with computed label
295 @cindex address of a label
297 You can get the address of a label defined in the current function
298 (or a containing function) with the unary operator @samp{&&}. The
299 value has type @code{void *}. This value is a constant and can be used
300 wherever a constant of that type is valid. For example:
308 To use these values, you need to be able to jump to one. This is done
309 with the computed goto statement@footnote{The analogous feature in
310 Fortran is called an assigned goto, but that name seems inappropriate in
311 C, where one can do more than simply store label addresses in label
312 variables.}, @code{goto *@var{exp};}. For example,
319 Any expression of type @code{void *} is allowed.
321 One way of using these constants is in initializing a static array that
322 will serve as a jump table:
325 static void *array[] = @{ &&foo, &&bar, &&hack @};
328 Then you can select a label with indexing, like this:
335 Note that this does not check whether the subscript is in bounds---array
336 indexing in C never does that.
338 Such an array of label values serves a purpose much like that of the
339 @code{switch} statement. The @code{switch} statement is cleaner, so
340 use that rather than an array unless the problem does not fit a
341 @code{switch} statement very well.
343 Another use of label values is in an interpreter for threaded code.
344 The labels within the interpreter function can be stored in the
345 threaded code for super-fast dispatching.
347 You may not use this mechanism to jump to code in a different function.
348 If you do that, totally unpredictable things will happen. The best way to
349 avoid this is to store the label address only in automatic variables and
350 never pass it as an argument.
352 An alternate way to write the above example is
355 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
357 goto *(&&foo + array[i]);
361 This is more friendly to code living in shared libraries, as it reduces
362 the number of dynamic relocations that are needed, and by consequence,
363 allows the data to be read-only.
365 @node Nested Functions
366 @section Nested Functions
367 @cindex nested functions
368 @cindex downward funargs
371 A @dfn{nested function} is a function defined inside another function.
372 (Nested functions are not supported for GNU C++.) The nested function's
373 name is local to the block where it is defined. For example, here we
374 define a nested function named @code{square}, and call it twice:
378 foo (double a, double b)
380 double square (double z) @{ return z * z; @}
382 return square (a) + square (b);
387 The nested function can access all the variables of the containing
388 function that are visible at the point of its definition. This is
389 called @dfn{lexical scoping}. For example, here we show a nested
390 function which uses an inherited variable named @code{offset}:
394 bar (int *array, int offset, int size)
396 int access (int *array, int index)
397 @{ return array[index + offset]; @}
400 for (i = 0; i < size; i++)
401 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
406 Nested function definitions are permitted within functions in the places
407 where variable definitions are allowed; that is, in any block, mixed
408 with the other declarations and statements in the block.
410 It is possible to call the nested function from outside the scope of its
411 name by storing its address or passing the address to another function:
414 hack (int *array, int size)
416 void store (int index, int value)
417 @{ array[index] = value; @}
419 intermediate (store, size);
423 Here, the function @code{intermediate} receives the address of
424 @code{store} as an argument. If @code{intermediate} calls @code{store},
425 the arguments given to @code{store} are used to store into @code{array}.
426 But this technique works only so long as the containing function
427 (@code{hack}, in this example) does not exit.
429 If you try to call the nested function through its address after the
430 containing function has exited, all hell will break loose. If you try
431 to call it after a containing scope level has exited, and if it refers
432 to some of the variables that are no longer in scope, you may be lucky,
433 but it's not wise to take the risk. If, however, the nested function
434 does not refer to anything that has gone out of scope, you should be
437 GCC implements taking the address of a nested function using a technique
438 called @dfn{trampolines}. A paper describing them is available as
441 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
443 A nested function can jump to a label inherited from a containing
444 function, provided the label was explicitly declared in the containing
445 function (@pxref{Local Labels}). Such a jump returns instantly to the
446 containing function, exiting the nested function which did the
447 @code{goto} and any intermediate functions as well. Here is an example:
451 bar (int *array, int offset, int size)
454 int access (int *array, int index)
458 return array[index + offset];
462 for (i = 0; i < size; i++)
463 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
467 /* @r{Control comes here from @code{access}
468 if it detects an error.} */
475 A nested function always has no linkage. Declaring one with
476 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
477 before its definition, use @code{auto} (which is otherwise meaningless
478 for function declarations).
481 bar (int *array, int offset, int size)
484 auto int access (int *, int);
486 int access (int *array, int index)
490 return array[index + offset];
496 @node Constructing Calls
497 @section Constructing Function Calls
498 @cindex constructing calls
499 @cindex forwarding calls
501 Using the built-in functions described below, you can record
502 the arguments a function received, and call another function
503 with the same arguments, without knowing the number or types
506 You can also record the return value of that function call,
507 and later return that value, without knowing what data type
508 the function tried to return (as long as your caller expects
511 However, these built-in functions may interact badly with some
512 sophisticated features or other extensions of the language. It
513 is, therefore, not recommended to use them outside very simple
514 functions acting as mere forwarders for their arguments.
516 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
517 This built-in function returns a pointer to data
518 describing how to perform a call with the same arguments as were passed
519 to the current function.
521 The function saves the arg pointer register, structure value address,
522 and all registers that might be used to pass arguments to a function
523 into a block of memory allocated on the stack. Then it returns the
524 address of that block.
527 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
528 This built-in function invokes @var{function}
529 with a copy of the parameters described by @var{arguments}
532 The value of @var{arguments} should be the value returned by
533 @code{__builtin_apply_args}. The argument @var{size} specifies the size
534 of the stack argument data, in bytes.
536 This function returns a pointer to data describing
537 how to return whatever value was returned by @var{function}. The data
538 is saved in a block of memory allocated on the stack.
540 It is not always simple to compute the proper value for @var{size}. The
541 value is used by @code{__builtin_apply} to compute the amount of data
542 that should be pushed on the stack and copied from the incoming argument
546 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
547 This built-in function returns the value described by @var{result} from
548 the containing function. You should specify, for @var{result}, a value
549 returned by @code{__builtin_apply}.
553 @section Referring to a Type with @code{typeof}
556 @cindex macros, types of arguments
558 Another way to refer to the type of an expression is with @code{typeof}.
559 The syntax of using of this keyword looks like @code{sizeof}, but the
560 construct acts semantically like a type name defined with @code{typedef}.
562 There are two ways of writing the argument to @code{typeof}: with an
563 expression or with a type. Here is an example with an expression:
570 This assumes that @code{x} is an array of pointers to functions;
571 the type described is that of the values of the functions.
573 Here is an example with a typename as the argument:
580 Here the type described is that of pointers to @code{int}.
582 If you are writing a header file that must work when included in ISO C
583 programs, write @code{__typeof__} instead of @code{typeof}.
584 @xref{Alternate Keywords}.
586 A @code{typeof}-construct can be used anywhere a typedef name could be
587 used. For example, you can use it in a declaration, in a cast, or inside
588 of @code{sizeof} or @code{typeof}.
590 @code{typeof} is often useful in conjunction with the
591 statements-within-expressions feature. Here is how the two together can
592 be used to define a safe ``maximum'' macro that operates on any
593 arithmetic type and evaluates each of its arguments exactly once:
597 (@{ typeof (a) _a = (a); \
598 typeof (b) _b = (b); \
599 _a > _b ? _a : _b; @})
602 @cindex underscores in variables in macros
603 @cindex @samp{_} in variables in macros
604 @cindex local variables in macros
605 @cindex variables, local, in macros
606 @cindex macros, local variables in
608 The reason for using names that start with underscores for the local
609 variables is to avoid conflicts with variable names that occur within the
610 expressions that are substituted for @code{a} and @code{b}. Eventually we
611 hope to design a new form of declaration syntax that allows you to declare
612 variables whose scopes start only after their initializers; this will be a
613 more reliable way to prevent such conflicts.
616 Some more examples of the use of @code{typeof}:
620 This declares @code{y} with the type of what @code{x} points to.
627 This declares @code{y} as an array of such values.
634 This declares @code{y} as an array of pointers to characters:
637 typeof (typeof (char *)[4]) y;
641 It is equivalent to the following traditional C declaration:
647 To see the meaning of the declaration using @code{typeof}, and why it
648 might be a useful way to write, rewrite it with these macros:
651 #define pointer(T) typeof(T *)
652 #define array(T, N) typeof(T [N])
656 Now the declaration can be rewritten this way:
659 array (pointer (char), 4) y;
663 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
664 pointers to @code{char}.
667 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
668 a more limited extension which permitted one to write
671 typedef @var{T} = @var{expr};
675 with the effect of declaring @var{T} to have the type of the expression
676 @var{expr}. This extension does not work with GCC 3 (versions between
677 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
678 relies on it should be rewritten to use @code{typeof}:
681 typedef typeof(@var{expr}) @var{T};
685 This will work with all versions of GCC@.
688 @section Conditionals with Omitted Operands
689 @cindex conditional expressions, extensions
690 @cindex omitted middle-operands
691 @cindex middle-operands, omitted
692 @cindex extensions, @code{?:}
693 @cindex @code{?:} extensions
695 The middle operand in a conditional expression may be omitted. Then
696 if the first operand is nonzero, its value is the value of the conditional
699 Therefore, the expression
706 has the value of @code{x} if that is nonzero; otherwise, the value of
709 This example is perfectly equivalent to
715 @cindex side effect in ?:
716 @cindex ?: side effect
718 In this simple case, the ability to omit the middle operand is not
719 especially useful. When it becomes useful is when the first operand does,
720 or may (if it is a macro argument), contain a side effect. Then repeating
721 the operand in the middle would perform the side effect twice. Omitting
722 the middle operand uses the value already computed without the undesirable
723 effects of recomputing it.
726 @section Double-Word Integers
727 @cindex @code{long long} data types
728 @cindex double-word arithmetic
729 @cindex multiprecision arithmetic
730 @cindex @code{LL} integer suffix
731 @cindex @code{ULL} integer suffix
733 ISO C99 supports data types for integers that are at least 64 bits wide,
734 and as an extension GCC supports them in C89 mode and in C++.
735 Simply write @code{long long int} for a signed integer, or
736 @code{unsigned long long int} for an unsigned integer. To make an
737 integer constant of type @code{long long int}, add the suffix @samp{LL}
738 to the integer. To make an integer constant of type @code{unsigned long
739 long int}, add the suffix @samp{ULL} to the integer.
741 You can use these types in arithmetic like any other integer types.
742 Addition, subtraction, and bitwise boolean operations on these types
743 are open-coded on all types of machines. Multiplication is open-coded
744 if the machine supports fullword-to-doubleword a widening multiply
745 instruction. Division and shifts are open-coded only on machines that
746 provide special support. The operations that are not open-coded use
747 special library routines that come with GCC@.
749 There may be pitfalls when you use @code{long long} types for function
750 arguments, unless you declare function prototypes. If a function
751 expects type @code{int} for its argument, and you pass a value of type
752 @code{long long int}, confusion will result because the caller and the
753 subroutine will disagree about the number of bytes for the argument.
754 Likewise, if the function expects @code{long long int} and you pass
755 @code{int}. The best way to avoid such problems is to use prototypes.
758 @section Complex Numbers
759 @cindex complex numbers
760 @cindex @code{_Complex} keyword
761 @cindex @code{__complex__} keyword
763 ISO C99 supports complex floating data types, and as an extension GCC
764 supports them in C89 mode and in C++, and supports complex integer data
765 types which are not part of ISO C99. You can declare complex types
766 using the keyword @code{_Complex}. As an extension, the older GNU
767 keyword @code{__complex__} is also supported.
769 For example, @samp{_Complex double x;} declares @code{x} as a
770 variable whose real part and imaginary part are both of type
771 @code{double}. @samp{_Complex short int y;} declares @code{y} to
772 have real and imaginary parts of type @code{short int}; this is not
773 likely to be useful, but it shows that the set of complex types is
776 To write a constant with a complex data type, use the suffix @samp{i} or
777 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
778 has type @code{_Complex float} and @code{3i} has type
779 @code{_Complex int}. Such a constant always has a pure imaginary
780 value, but you can form any complex value you like by adding one to a
781 real constant. This is a GNU extension; if you have an ISO C99
782 conforming C library (such as GNU libc), and want to construct complex
783 constants of floating type, you should include @code{<complex.h>} and
784 use the macros @code{I} or @code{_Complex_I} instead.
786 @cindex @code{__real__} keyword
787 @cindex @code{__imag__} keyword
788 To extract the real part of a complex-valued expression @var{exp}, write
789 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
790 extract the imaginary part. This is a GNU extension; for values of
791 floating type, you should use the ISO C99 functions @code{crealf},
792 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
793 @code{cimagl}, declared in @code{<complex.h>} and also provided as
794 built-in functions by GCC@.
796 @cindex complex conjugation
797 The operator @samp{~} performs complex conjugation when used on a value
798 with a complex type. This is a GNU extension; for values of
799 floating type, you should use the ISO C99 functions @code{conjf},
800 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
801 provided as built-in functions by GCC@.
803 GCC can allocate complex automatic variables in a noncontiguous
804 fashion; it's even possible for the real part to be in a register while
805 the imaginary part is on the stack (or vice-versa). Only the DWARF2
806 debug info format can represent this, so use of DWARF2 is recommended.
807 If you are using the stabs debug info format, GCC describes a noncontiguous
808 complex variable as if it were two separate variables of noncomplex type.
809 If the variable's actual name is @code{foo}, the two fictitious
810 variables are named @code{foo$real} and @code{foo$imag}. You can
811 examine and set these two fictitious variables with your debugger.
817 ISO C99 supports floating-point numbers written not only in the usual
818 decimal notation, such as @code{1.55e1}, but also numbers such as
819 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
820 supports this in C89 mode (except in some cases when strictly
821 conforming) and in C++. In that format the
822 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
823 mandatory. The exponent is a decimal number that indicates the power of
824 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
831 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
832 is the same as @code{1.55e1}.
834 Unlike for floating-point numbers in the decimal notation the exponent
835 is always required in the hexadecimal notation. Otherwise the compiler
836 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
837 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
838 extension for floating-point constants of type @code{float}.
841 @section Arrays of Length Zero
842 @cindex arrays of length zero
843 @cindex zero-length arrays
844 @cindex length-zero arrays
845 @cindex flexible array members
847 Zero-length arrays are allowed in GNU C@. They are very useful as the
848 last element of a structure which is really a header for a variable-length
857 struct line *thisline = (struct line *)
858 malloc (sizeof (struct line) + this_length);
859 thisline->length = this_length;
862 In ISO C90, you would have to give @code{contents} a length of 1, which
863 means either you waste space or complicate the argument to @code{malloc}.
865 In ISO C99, you would use a @dfn{flexible array member}, which is
866 slightly different in syntax and semantics:
870 Flexible array members are written as @code{contents[]} without
874 Flexible array members have incomplete type, and so the @code{sizeof}
875 operator may not be applied. As a quirk of the original implementation
876 of zero-length arrays, @code{sizeof} evaluates to zero.
879 Flexible array members may only appear as the last member of a
880 @code{struct} that is otherwise non-empty.
883 A structure containing a flexible array member, or a union containing
884 such a structure (possibly recursively), may not be a member of a
885 structure or an element of an array. (However, these uses are
886 permitted by GCC as extensions.)
889 GCC versions before 3.0 allowed zero-length arrays to be statically
890 initialized, as if they were flexible arrays. In addition to those
891 cases that were useful, it also allowed initializations in situations
892 that would corrupt later data. Non-empty initialization of zero-length
893 arrays is now treated like any case where there are more initializer
894 elements than the array holds, in that a suitable warning about "excess
895 elements in array" is given, and the excess elements (all of them, in
896 this case) are ignored.
898 Instead GCC allows static initialization of flexible array members.
899 This is equivalent to defining a new structure containing the original
900 structure followed by an array of sufficient size to contain the data.
901 I.e.@: in the following, @code{f1} is constructed as if it were declared
907 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
910 struct f1 f1; int data[3];
911 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
915 The convenience of this extension is that @code{f1} has the desired
916 type, eliminating the need to consistently refer to @code{f2.f1}.
918 This has symmetry with normal static arrays, in that an array of
919 unknown size is also written with @code{[]}.
921 Of course, this extension only makes sense if the extra data comes at
922 the end of a top-level object, as otherwise we would be overwriting
923 data at subsequent offsets. To avoid undue complication and confusion
924 with initialization of deeply nested arrays, we simply disallow any
925 non-empty initialization except when the structure is the top-level
929 struct foo @{ int x; int y[]; @};
930 struct bar @{ struct foo z; @};
932 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
933 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
934 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
935 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
938 @node Empty Structures
939 @section Structures With No Members
940 @cindex empty structures
941 @cindex zero-size structures
943 GCC permits a C structure to have no members:
950 The structure will have size zero. In C++, empty structures are part
951 of the language. G++ treats empty structures as if they had a single
952 member of type @code{char}.
954 @node Variable Length
955 @section Arrays of Variable Length
956 @cindex variable-length arrays
957 @cindex arrays of variable length
960 Variable-length automatic arrays are allowed in ISO C99, and as an
961 extension GCC accepts them in C89 mode and in C++. (However, GCC's
962 implementation of variable-length arrays does not yet conform in detail
963 to the ISO C99 standard.) These arrays are
964 declared like any other automatic arrays, but with a length that is not
965 a constant expression. The storage is allocated at the point of
966 declaration and deallocated when the brace-level is exited. For
971 concat_fopen (char *s1, char *s2, char *mode)
973 char str[strlen (s1) + strlen (s2) + 1];
976 return fopen (str, mode);
980 @cindex scope of a variable length array
981 @cindex variable-length array scope
982 @cindex deallocating variable length arrays
983 Jumping or breaking out of the scope of the array name deallocates the
984 storage. Jumping into the scope is not allowed; you get an error
987 @cindex @code{alloca} vs variable-length arrays
988 You can use the function @code{alloca} to get an effect much like
989 variable-length arrays. The function @code{alloca} is available in
990 many other C implementations (but not in all). On the other hand,
991 variable-length arrays are more elegant.
993 There are other differences between these two methods. Space allocated
994 with @code{alloca} exists until the containing @emph{function} returns.
995 The space for a variable-length array is deallocated as soon as the array
996 name's scope ends. (If you use both variable-length arrays and
997 @code{alloca} in the same function, deallocation of a variable-length array
998 will also deallocate anything more recently allocated with @code{alloca}.)
1000 You can also use variable-length arrays as arguments to functions:
1004 tester (int len, char data[len][len])
1010 The length of an array is computed once when the storage is allocated
1011 and is remembered for the scope of the array in case you access it with
1014 If you want to pass the array first and the length afterward, you can
1015 use a forward declaration in the parameter list---another GNU extension.
1019 tester (int len; char data[len][len], int len)
1025 @cindex parameter forward declaration
1026 The @samp{int len} before the semicolon is a @dfn{parameter forward
1027 declaration}, and it serves the purpose of making the name @code{len}
1028 known when the declaration of @code{data} is parsed.
1030 You can write any number of such parameter forward declarations in the
1031 parameter list. They can be separated by commas or semicolons, but the
1032 last one must end with a semicolon, which is followed by the ``real''
1033 parameter declarations. Each forward declaration must match a ``real''
1034 declaration in parameter name and data type. ISO C99 does not support
1035 parameter forward declarations.
1037 @node Variadic Macros
1038 @section Macros with a Variable Number of Arguments.
1039 @cindex variable number of arguments
1040 @cindex macro with variable arguments
1041 @cindex rest argument (in macro)
1042 @cindex variadic macros
1044 In the ISO C standard of 1999, a macro can be declared to accept a
1045 variable number of arguments much as a function can. The syntax for
1046 defining the macro is similar to that of a function. Here is an
1050 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1053 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1054 such a macro, it represents the zero or more tokens until the closing
1055 parenthesis that ends the invocation, including any commas. This set of
1056 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1057 wherever it appears. See the CPP manual for more information.
1059 GCC has long supported variadic macros, and used a different syntax that
1060 allowed you to give a name to the variable arguments just like any other
1061 argument. Here is an example:
1064 #define debug(format, args...) fprintf (stderr, format, args)
1067 This is in all ways equivalent to the ISO C example above, but arguably
1068 more readable and descriptive.
1070 GNU CPP has two further variadic macro extensions, and permits them to
1071 be used with either of the above forms of macro definition.
1073 In standard C, you are not allowed to leave the variable argument out
1074 entirely; but you are allowed to pass an empty argument. For example,
1075 this invocation is invalid in ISO C, because there is no comma after
1082 GNU CPP permits you to completely omit the variable arguments in this
1083 way. In the above examples, the compiler would complain, though since
1084 the expansion of the macro still has the extra comma after the format
1087 To help solve this problem, CPP behaves specially for variable arguments
1088 used with the token paste operator, @samp{##}. If instead you write
1091 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1094 and if the variable arguments are omitted or empty, the @samp{##}
1095 operator causes the preprocessor to remove the comma before it. If you
1096 do provide some variable arguments in your macro invocation, GNU CPP
1097 does not complain about the paste operation and instead places the
1098 variable arguments after the comma. Just like any other pasted macro
1099 argument, these arguments are not macro expanded.
1101 @node Escaped Newlines
1102 @section Slightly Looser Rules for Escaped Newlines
1103 @cindex escaped newlines
1104 @cindex newlines (escaped)
1106 Recently, the preprocessor has relaxed its treatment of escaped
1107 newlines. Previously, the newline had to immediately follow a
1108 backslash. The current implementation allows whitespace in the form
1109 of spaces, horizontal and vertical tabs, and form feeds between the
1110 backslash and the subsequent newline. The preprocessor issues a
1111 warning, but treats it as a valid escaped newline and combines the two
1112 lines to form a single logical line. This works within comments and
1113 tokens, as well as between tokens. Comments are @emph{not} treated as
1114 whitespace for the purposes of this relaxation, since they have not
1115 yet been replaced with spaces.
1118 @section Non-Lvalue Arrays May Have Subscripts
1119 @cindex subscripting
1120 @cindex arrays, non-lvalue
1122 @cindex subscripting and function values
1123 In ISO C99, arrays that are not lvalues still decay to pointers, and
1124 may be subscripted, although they may not be modified or used after
1125 the next sequence point and the unary @samp{&} operator may not be
1126 applied to them. As an extension, GCC allows such arrays to be
1127 subscripted in C89 mode, though otherwise they do not decay to
1128 pointers outside C99 mode. For example,
1129 this is valid in GNU C though not valid in C89:
1133 struct foo @{int a[4];@};
1139 return f().a[index];
1145 @section Arithmetic on @code{void}- and Function-Pointers
1146 @cindex void pointers, arithmetic
1147 @cindex void, size of pointer to
1148 @cindex function pointers, arithmetic
1149 @cindex function, size of pointer to
1151 In GNU C, addition and subtraction operations are supported on pointers to
1152 @code{void} and on pointers to functions. This is done by treating the
1153 size of a @code{void} or of a function as 1.
1155 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1156 and on function types, and returns 1.
1158 @opindex Wpointer-arith
1159 The option @option{-Wpointer-arith} requests a warning if these extensions
1163 @section Non-Constant Initializers
1164 @cindex initializers, non-constant
1165 @cindex non-constant initializers
1167 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1168 automatic variable are not required to be constant expressions in GNU C@.
1169 Here is an example of an initializer with run-time varying elements:
1172 foo (float f, float g)
1174 float beat_freqs[2] = @{ f-g, f+g @};
1179 @node Compound Literals
1180 @section Compound Literals
1181 @cindex constructor expressions
1182 @cindex initializations in expressions
1183 @cindex structures, constructor expression
1184 @cindex expressions, constructor
1185 @cindex compound literals
1186 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1188 ISO C99 supports compound literals. A compound literal looks like
1189 a cast containing an initializer. Its value is an object of the
1190 type specified in the cast, containing the elements specified in
1191 the initializer; it is an lvalue. As an extension, GCC supports
1192 compound literals in C89 mode and in C++.
1194 Usually, the specified type is a structure. Assume that
1195 @code{struct foo} and @code{structure} are declared as shown:
1198 struct foo @{int a; char b[2];@} structure;
1202 Here is an example of constructing a @code{struct foo} with a compound literal:
1205 structure = ((struct foo) @{x + y, 'a', 0@});
1209 This is equivalent to writing the following:
1213 struct foo temp = @{x + y, 'a', 0@};
1218 You can also construct an array. If all the elements of the compound literal
1219 are (made up of) simple constant expressions, suitable for use in
1220 initializers of objects of static storage duration, then the compound
1221 literal can be coerced to a pointer to its first element and used in
1222 such an initializer, as shown here:
1225 char **foo = (char *[]) @{ "x", "y", "z" @};
1228 Compound literals for scalar types and union types are is
1229 also allowed, but then the compound literal is equivalent
1232 As a GNU extension, GCC allows initialization of objects with static storage
1233 duration by compound literals (which is not possible in ISO C99, because
1234 the initializer is not a constant).
1235 It is handled as if the object was initialized only with the bracket
1236 enclosed list if compound literal's and object types match.
1237 The initializer list of the compound literal must be constant.
1238 If the object being initialized has array type of unknown size, the size is
1239 determined by compound literal size.
1242 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1243 static int y[] = (int []) @{1, 2, 3@};
1244 static int z[] = (int [3]) @{1@};
1248 The above lines are equivalent to the following:
1250 static struct foo x = @{1, 'a', 'b'@};
1251 static int y[] = @{1, 2, 3@};
1252 static int z[] = @{1, 0, 0@};
1255 @node Designated Inits
1256 @section Designated Initializers
1257 @cindex initializers with labeled elements
1258 @cindex labeled elements in initializers
1259 @cindex case labels in initializers
1260 @cindex designated initializers
1262 Standard C89 requires the elements of an initializer to appear in a fixed
1263 order, the same as the order of the elements in the array or structure
1266 In ISO C99 you can give the elements in any order, specifying the array
1267 indices or structure field names they apply to, and GNU C allows this as
1268 an extension in C89 mode as well. This extension is not
1269 implemented in GNU C++.
1271 To specify an array index, write
1272 @samp{[@var{index}] =} before the element value. For example,
1275 int a[6] = @{ [4] = 29, [2] = 15 @};
1282 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1286 The index values must be constant expressions, even if the array being
1287 initialized is automatic.
1289 An alternative syntax for this which has been obsolete since GCC 2.5 but
1290 GCC still accepts is to write @samp{[@var{index}]} before the element
1291 value, with no @samp{=}.
1293 To initialize a range of elements to the same value, write
1294 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1295 extension. For example,
1298 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1302 If the value in it has side-effects, the side-effects will happen only once,
1303 not for each initialized field by the range initializer.
1306 Note that the length of the array is the highest value specified
1309 In a structure initializer, specify the name of a field to initialize
1310 with @samp{.@var{fieldname} =} before the element value. For example,
1311 given the following structure,
1314 struct point @{ int x, y; @};
1318 the following initialization
1321 struct point p = @{ .y = yvalue, .x = xvalue @};
1328 struct point p = @{ xvalue, yvalue @};
1331 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1332 @samp{@var{fieldname}:}, as shown here:
1335 struct point p = @{ y: yvalue, x: xvalue @};
1339 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1340 @dfn{designator}. You can also use a designator (or the obsolete colon
1341 syntax) when initializing a union, to specify which element of the union
1342 should be used. For example,
1345 union foo @{ int i; double d; @};
1347 union foo f = @{ .d = 4 @};
1351 will convert 4 to a @code{double} to store it in the union using
1352 the second element. By contrast, casting 4 to type @code{union foo}
1353 would store it into the union as the integer @code{i}, since it is
1354 an integer. (@xref{Cast to Union}.)
1356 You can combine this technique of naming elements with ordinary C
1357 initialization of successive elements. Each initializer element that
1358 does not have a designator applies to the next consecutive element of the
1359 array or structure. For example,
1362 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1369 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1372 Labeling the elements of an array initializer is especially useful
1373 when the indices are characters or belong to an @code{enum} type.
1378 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1379 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1382 @cindex designator lists
1383 You can also write a series of @samp{.@var{fieldname}} and
1384 @samp{[@var{index}]} designators before an @samp{=} to specify a
1385 nested subobject to initialize; the list is taken relative to the
1386 subobject corresponding to the closest surrounding brace pair. For
1387 example, with the @samp{struct point} declaration above:
1390 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1394 If the same field is initialized multiple times, it will have value from
1395 the last initialization. If any such overridden initialization has
1396 side-effect, it is unspecified whether the side-effect happens or not.
1397 Currently, GCC will discard them and issue a warning.
1400 @section Case Ranges
1402 @cindex ranges in case statements
1404 You can specify a range of consecutive values in a single @code{case} label,
1408 case @var{low} ... @var{high}:
1412 This has the same effect as the proper number of individual @code{case}
1413 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1415 This feature is especially useful for ranges of ASCII character codes:
1421 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1422 it may be parsed wrong when you use it with integer values. For example,
1437 @section Cast to a Union Type
1438 @cindex cast to a union
1439 @cindex union, casting to a
1441 A cast to union type is similar to other casts, except that the type
1442 specified is a union type. You can specify the type either with
1443 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1444 a constructor though, not a cast, and hence does not yield an lvalue like
1445 normal casts. (@xref{Compound Literals}.)
1447 The types that may be cast to the union type are those of the members
1448 of the union. Thus, given the following union and variables:
1451 union foo @{ int i; double d; @};
1457 both @code{x} and @code{y} can be cast to type @code{union foo}.
1459 Using the cast as the right-hand side of an assignment to a variable of
1460 union type is equivalent to storing in a member of the union:
1465 u = (union foo) x @equiv{} u.i = x
1466 u = (union foo) y @equiv{} u.d = y
1469 You can also use the union cast as a function argument:
1472 void hack (union foo);
1474 hack ((union foo) x);
1477 @node Mixed Declarations
1478 @section Mixed Declarations and Code
1479 @cindex mixed declarations and code
1480 @cindex declarations, mixed with code
1481 @cindex code, mixed with declarations
1483 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1484 within compound statements. As an extension, GCC also allows this in
1485 C89 mode. For example, you could do:
1494 Each identifier is visible from where it is declared until the end of
1495 the enclosing block.
1497 @node Function Attributes
1498 @section Declaring Attributes of Functions
1499 @cindex function attributes
1500 @cindex declaring attributes of functions
1501 @cindex functions that never return
1502 @cindex functions that return more than once
1503 @cindex functions that have no side effects
1504 @cindex functions in arbitrary sections
1505 @cindex functions that behave like malloc
1506 @cindex @code{volatile} applied to function
1507 @cindex @code{const} applied to function
1508 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1509 @cindex functions with non-null pointer arguments
1510 @cindex functions that are passed arguments in registers on the 386
1511 @cindex functions that pop the argument stack on the 386
1512 @cindex functions that do not pop the argument stack on the 386
1514 In GNU C, you declare certain things about functions called in your program
1515 which help the compiler optimize function calls and check your code more
1518 The keyword @code{__attribute__} allows you to specify special
1519 attributes when making a declaration. This keyword is followed by an
1520 attribute specification inside double parentheses. The following
1521 attributes are currently defined for functions on all targets:
1522 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1523 @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1524 @code{format}, @code{format_arg}, @code{no_instrument_function},
1525 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1526 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1527 @code{alias}, @code{warn_unused_result} and @code{nonnull}. Several other
1528 attributes are defined for functions on particular target systems. Other
1529 attributes, including @code{section} are supported for variables declarations
1530 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1532 You may also specify attributes with @samp{__} preceding and following
1533 each keyword. This allows you to use them in header files without
1534 being concerned about a possible macro of the same name. For example,
1535 you may use @code{__noreturn__} instead of @code{noreturn}.
1537 @xref{Attribute Syntax}, for details of the exact syntax for using
1541 @c Keep this table alphabetized by attribute name. Treat _ as space.
1543 @item alias ("@var{target}")
1544 @cindex @code{alias} attribute
1545 The @code{alias} attribute causes the declaration to be emitted as an
1546 alias for another symbol, which must be specified. For instance,
1549 void __f () @{ /* @r{Do something.} */; @}
1550 void f () __attribute__ ((weak, alias ("__f")));
1553 declares @samp{f} to be a weak alias for @samp{__f}. In C++, the
1554 mangled name for the target must be used. It is an error if @samp{__f}
1555 is not defined in the same translation unit.
1557 Not all target machines support this attribute.
1560 @cindex @code{always_inline} function attribute
1561 Generally, functions are not inlined unless optimization is specified.
1562 For functions declared inline, this attribute inlines the function even
1563 if no optimization level was specified.
1566 @cindex functions that do pop the argument stack on the 386
1568 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1569 assume that the calling function will pop off the stack space used to
1570 pass arguments. This is
1571 useful to override the effects of the @option{-mrtd} switch.
1574 @cindex @code{const} function attribute
1575 Many functions do not examine any values except their arguments, and
1576 have no effects except the return value. Basically this is just slightly
1577 more strict class than the @code{pure} attribute below, since function is not
1578 allowed to read global memory.
1580 @cindex pointer arguments
1581 Note that a function that has pointer arguments and examines the data
1582 pointed to must @emph{not} be declared @code{const}. Likewise, a
1583 function that calls a non-@code{const} function usually must not be
1584 @code{const}. It does not make sense for a @code{const} function to
1587 The attribute @code{const} is not implemented in GCC versions earlier
1588 than 2.5. An alternative way to declare that a function has no side
1589 effects, which works in the current version and in some older versions,
1593 typedef int intfn ();
1595 extern const intfn square;
1598 This approach does not work in GNU C++ from 2.6.0 on, since the language
1599 specifies that the @samp{const} must be attached to the return value.
1603 @cindex @code{constructor} function attribute
1604 @cindex @code{destructor} function attribute
1605 The @code{constructor} attribute causes the function to be called
1606 automatically before execution enters @code{main ()}. Similarly, the
1607 @code{destructor} attribute causes the function to be called
1608 automatically after @code{main ()} has completed or @code{exit ()} has
1609 been called. Functions with these attributes are useful for
1610 initializing data that will be used implicitly during the execution of
1613 These attributes are not currently implemented for Objective-C@.
1616 @cindex @code{deprecated} attribute.
1617 The @code{deprecated} attribute results in a warning if the function
1618 is used anywhere in the source file. This is useful when identifying
1619 functions that are expected to be removed in a future version of a
1620 program. The warning also includes the location of the declaration
1621 of the deprecated function, to enable users to easily find further
1622 information about why the function is deprecated, or what they should
1623 do instead. Note that the warnings only occurs for uses:
1626 int old_fn () __attribute__ ((deprecated));
1628 int (*fn_ptr)() = old_fn;
1631 results in a warning on line 3 but not line 2.
1633 The @code{deprecated} attribute can also be used for variables and
1634 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1637 @cindex @code{__declspec(dllexport)}
1638 On Microsoft Windows targets and Symbian OS targets the
1639 @code{dllexport} attribute causes the compiler to provide a global
1640 pointer to a pointer in a DLL, so that it can be referenced with the
1641 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1642 name is formed by combining @code{_imp__} and the function or variable
1645 You can use @code{__declspec(dllexport)} as a synonym for
1646 @code{__attribute__ ((dllexport))} for compatibility with other
1649 On systems that support the @code{visibility} attribute, this
1650 attribute also implies ``default'' visibility, unless a
1651 @code{visibility} attribute is explicitly specified. You should avoid
1652 the use of @code{dllexport} with ``hidden'' or ``internal''
1653 visibility; in the future GCC may issue an error for those cases.
1655 Currently, the @code{dllexport} attribute is ignored for inlined
1656 functions, unless the @option{-fkeep-inline-functions} flag has been
1657 used. The attribute is also ignored for undefined symbols.
1659 When applied to C++ classes, the attribute marks defined non-inlined
1660 member functions and static data members as exports. Static consts
1661 initialized in-class are not marked unless they are also defined
1664 For Microsoft Windows targets there are alternative methods for
1665 including the symbol in the DLL's export table such as using a
1666 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1667 the @option{--export-all} linker flag.
1670 @cindex @code{__declspec(dllimport)}
1671 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1672 attribute causes the compiler to reference a function or variable via
1673 a global pointer to a pointer that is set up by the DLL exporting the
1674 symbol. The attribute implies @code{extern} storage. On Microsoft
1675 Windows targets, the pointer name is formed by combining @code{_imp__}
1676 and the function or variable name.
1678 You can use @code{__declspec(dllimport)} as a synonym for
1679 @code{__attribute__ ((dllimport))} for compatibility with other
1682 Currently, the attribute is ignored for inlined functions. If the
1683 attribute is applied to a symbol @emph{definition}, an error is reported.
1684 If a symbol previously declared @code{dllimport} is later defined, the
1685 attribute is ignored in subsequent references, and a warning is emitted.
1686 The attribute is also overridden by a subsequent declaration as
1689 When applied to C++ classes, the attribute marks non-inlined
1690 member functions and static data members as imports. However, the
1691 attribute is ignored for virtual methods to allow creation of vtables
1694 On the SH Symbian OS target the @code{dllimport} attribute also has
1695 another affect---it can cause the vtable and run-time type information
1696 for a class to be exported. This happens when the class has a
1697 dllimport'ed constructor or a non-inline, non-pure virtual function
1698 and, for either of those two conditions, the class also has a inline
1699 constructor or destructor and has a key function that is defined in
1700 the current translation unit.
1702 For Microsoft Windows based targets the use of the @code{dllimport}
1703 attribute on functions is not necessary, but provides a small
1704 performance benefit by eliminating a thunk in the DLL@. The use of the
1705 @code{dllimport} attribute on imported variables was required on older
1706 versions of the GNU linker, but can now be avoided by passing the
1707 @option{--enable-auto-import} switch to the GNU linker. As with
1708 functions, using the attribute for a variable eliminates a thunk in
1711 One drawback to using this attribute is that a pointer to a function
1712 or variable marked as @code{dllimport} cannot be used as a constant
1713 address. On Microsoft Windows targets, the attribute can be disabled
1714 for functions by setting the @option{-mnop-fun-dllimport} flag.
1717 @cindex eight bit data on the H8/300, H8/300H, and H8S
1718 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1719 variable should be placed into the eight bit data section.
1720 The compiler will generate more efficient code for certain operations
1721 on data in the eight bit data area. Note the eight bit data area is limited to
1724 You must use GAS and GLD from GNU binutils version 2.7 or later for
1725 this attribute to work correctly.
1728 @cindex functions which handle memory bank switching
1729 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1730 use a calling convention that takes care of switching memory banks when
1731 entering and leaving a function. This calling convention is also the
1732 default when using the @option{-mlong-calls} option.
1734 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1735 to call and return from a function.
1737 On 68HC11 the compiler will generate a sequence of instructions
1738 to invoke a board-specific routine to switch the memory bank and call the
1739 real function. The board-specific routine simulates a @code{call}.
1740 At the end of a function, it will jump to a board-specific routine
1741 instead of using @code{rts}. The board-specific return routine simulates
1745 @cindex functions that pop the argument stack on the 386
1746 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1747 pass the first two arguments in the registers ECX and EDX@. Subsequent
1748 arguments are passed on the stack. The called function will pop the
1749 arguments off the stack. If the number of arguments is variable all
1750 arguments are pushed on the stack.
1752 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1753 @cindex @code{format} function attribute
1755 The @code{format} attribute specifies that a function takes @code{printf},
1756 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1757 should be type-checked against a format string. For example, the
1762 my_printf (void *my_object, const char *my_format, ...)
1763 __attribute__ ((format (printf, 2, 3)));
1767 causes the compiler to check the arguments in calls to @code{my_printf}
1768 for consistency with the @code{printf} style format string argument
1771 The parameter @var{archetype} determines how the format string is
1772 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1773 or @code{strfmon}. (You can also use @code{__printf__},
1774 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1775 parameter @var{string-index} specifies which argument is the format
1776 string argument (starting from 1), while @var{first-to-check} is the
1777 number of the first argument to check against the format string. For
1778 functions where the arguments are not available to be checked (such as
1779 @code{vprintf}), specify the third parameter as zero. In this case the
1780 compiler only checks the format string for consistency. For
1781 @code{strftime} formats, the third parameter is required to be zero.
1782 Since non-static C++ methods have an implicit @code{this} argument, the
1783 arguments of such methods should be counted from two, not one, when
1784 giving values for @var{string-index} and @var{first-to-check}.
1786 In the example above, the format string (@code{my_format}) is the second
1787 argument of the function @code{my_print}, and the arguments to check
1788 start with the third argument, so the correct parameters for the format
1789 attribute are 2 and 3.
1791 @opindex ffreestanding
1792 @opindex fno-builtin
1793 The @code{format} attribute allows you to identify your own functions
1794 which take format strings as arguments, so that GCC can check the
1795 calls to these functions for errors. The compiler always (unless
1796 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1797 for the standard library functions @code{printf}, @code{fprintf},
1798 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1799 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1800 warnings are requested (using @option{-Wformat}), so there is no need to
1801 modify the header file @file{stdio.h}. In C99 mode, the functions
1802 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1803 @code{vsscanf} are also checked. Except in strictly conforming C
1804 standard modes, the X/Open function @code{strfmon} is also checked as
1805 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1806 @xref{C Dialect Options,,Options Controlling C Dialect}.
1808 The target may provide additional types of format checks.
1809 @xref{Target Format Checks,,Format Checks Specific to Particular
1812 @item format_arg (@var{string-index})
1813 @cindex @code{format_arg} function attribute
1814 @opindex Wformat-nonliteral
1815 The @code{format_arg} attribute specifies that a function takes a format
1816 string for a @code{printf}, @code{scanf}, @code{strftime} or
1817 @code{strfmon} style function and modifies it (for example, to translate
1818 it into another language), so the result can be passed to a
1819 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1820 function (with the remaining arguments to the format function the same
1821 as they would have been for the unmodified string). For example, the
1826 my_dgettext (char *my_domain, const char *my_format)
1827 __attribute__ ((format_arg (2)));
1831 causes the compiler to check the arguments in calls to a @code{printf},
1832 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1833 format string argument is a call to the @code{my_dgettext} function, for
1834 consistency with the format string argument @code{my_format}. If the
1835 @code{format_arg} attribute had not been specified, all the compiler
1836 could tell in such calls to format functions would be that the format
1837 string argument is not constant; this would generate a warning when
1838 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1839 without the attribute.
1841 The parameter @var{string-index} specifies which argument is the format
1842 string argument (starting from one). Since non-static C++ methods have
1843 an implicit @code{this} argument, the arguments of such methods should
1844 be counted from two.
1846 The @code{format-arg} attribute allows you to identify your own
1847 functions which modify format strings, so that GCC can check the
1848 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1849 type function whose operands are a call to one of your own function.
1850 The compiler always treats @code{gettext}, @code{dgettext}, and
1851 @code{dcgettext} in this manner except when strict ISO C support is
1852 requested by @option{-ansi} or an appropriate @option{-std} option, or
1853 @option{-ffreestanding} or @option{-fno-builtin}
1854 is used. @xref{C Dialect Options,,Options
1855 Controlling C Dialect}.
1857 @item function_vector
1858 @cindex calling functions through the function vector on the H8/300 processors
1859 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1860 function should be called through the function vector. Calling a
1861 function through the function vector will reduce code size, however;
1862 the function vector has a limited size (maximum 128 entries on the H8/300
1863 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1865 You must use GAS and GLD from GNU binutils version 2.7 or later for
1866 this attribute to work correctly.
1869 @cindex interrupt handler functions
1870 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
1871 that the specified function is an interrupt handler. The compiler will
1872 generate function entry and exit sequences suitable for use in an
1873 interrupt handler when this attribute is present.
1875 Note, interrupt handlers for the m68k, H8/300, H8/300H, H8S, and SH processors
1876 can be specified via the @code{interrupt_handler} attribute.
1878 Note, on the AVR, interrupts will be enabled inside the function.
1880 Note, for the ARM, you can specify the kind of interrupt to be handled by
1881 adding an optional parameter to the interrupt attribute like this:
1884 void f () __attribute__ ((interrupt ("IRQ")));
1887 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
1889 @item interrupt_handler
1890 @cindex interrupt handler functions on the m68k, H8/300 and SH processors
1891 Use this attribute on the m68k, H8/300, H8/300H, H8S, and SH to indicate that
1892 the specified function is an interrupt handler. The compiler will generate
1893 function entry and exit sequences suitable for use in an interrupt
1894 handler when this attribute is present.
1896 @item long_call/short_call
1897 @cindex indirect calls on ARM
1898 This attribute specifies how a particular function is called on
1899 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
1900 command line switch and @code{#pragma long_calls} settings. The
1901 @code{long_call} attribute causes the compiler to always call the
1902 function by first loading its address into a register and then using the
1903 contents of that register. The @code{short_call} attribute always places
1904 the offset to the function from the call site into the @samp{BL}
1905 instruction directly.
1907 @item longcall/shortcall
1908 @cindex functions called via pointer on the RS/6000 and PowerPC
1909 On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
1910 compiler to always call this function via a pointer, just as it would if
1911 the @option{-mlongcall} option had been specified. The @code{shortcall}
1912 attribute causes the compiler not to do this. These attributes override
1913 both the @option{-mlongcall} switch and the @code{#pragma longcall}
1916 @xref{RS/6000 and PowerPC Options}, for more information on whether long
1917 calls are necessary.
1920 @cindex @code{malloc} attribute
1921 The @code{malloc} attribute is used to tell the compiler that a function
1922 may be treated as if any non-@code{NULL} pointer it returns cannot
1923 alias any other pointer valid when the function returns.
1924 This will often improve optimization.
1925 Standard functions with this property include @code{malloc} and
1926 @code{calloc}. @code{realloc}-like functions have this property as
1927 long as the old pointer is never referred to (including comparing it
1928 to the new pointer) after the function returns a non-@code{NULL}
1931 @item model (@var{model-name})
1932 @cindex function addressability on the M32R/D
1933 @cindex variable addressability on the IA-64
1935 On the M32R/D, use this attribute to set the addressability of an
1936 object, and of the code generated for a function. The identifier
1937 @var{model-name} is one of @code{small}, @code{medium}, or
1938 @code{large}, representing each of the code models.
1940 Small model objects live in the lower 16MB of memory (so that their
1941 addresses can be loaded with the @code{ld24} instruction), and are
1942 callable with the @code{bl} instruction.
1944 Medium model objects may live anywhere in the 32-bit address space (the
1945 compiler will generate @code{seth/add3} instructions to load their addresses),
1946 and are callable with the @code{bl} instruction.
1948 Large model objects may live anywhere in the 32-bit address space (the
1949 compiler will generate @code{seth/add3} instructions to load their addresses),
1950 and may not be reachable with the @code{bl} instruction (the compiler will
1951 generate the much slower @code{seth/add3/jl} instruction sequence).
1953 On IA-64, use this attribute to set the addressability of an object.
1954 At present, the only supported identifier for @var{model-name} is
1955 @code{small}, indicating addressability via ``small'' (22-bit)
1956 addresses (so that their addresses can be loaded with the @code{addl}
1957 instruction). Caveat: such addressing is by definition not position
1958 independent and hence this attribute must not be used for objects
1959 defined by shared libraries.
1962 @cindex function without a prologue/epilogue code
1963 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
1964 specified function does not need prologue/epilogue sequences generated by
1965 the compiler. It is up to the programmer to provide these sequences.
1968 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
1969 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
1970 use the normal calling convention based on @code{jsr} and @code{rts}.
1971 This attribute can be used to cancel the effect of the @option{-mlong-calls}
1974 @item no_instrument_function
1975 @cindex @code{no_instrument_function} function attribute
1976 @opindex finstrument-functions
1977 If @option{-finstrument-functions} is given, profiling function calls will
1978 be generated at entry and exit of most user-compiled functions.
1979 Functions with this attribute will not be so instrumented.
1982 @cindex @code{noinline} function attribute
1983 This function attribute prevents a function from being considered for
1986 @item nonnull (@var{arg-index}, @dots{})
1987 @cindex @code{nonnull} function attribute
1988 The @code{nonnull} attribute specifies that some function parameters should
1989 be non-null pointers. For instance, the declaration:
1993 my_memcpy (void *dest, const void *src, size_t len)
1994 __attribute__((nonnull (1, 2)));
1998 causes the compiler to check that, in calls to @code{my_memcpy},
1999 arguments @var{dest} and @var{src} are non-null. If the compiler
2000 determines that a null pointer is passed in an argument slot marked
2001 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2002 is issued. The compiler may also choose to make optimizations based
2003 on the knowledge that certain function arguments will not be null.
2005 If no argument index list is given to the @code{nonnull} attribute,
2006 all pointer arguments are marked as non-null. To illustrate, the
2007 following declaration is equivalent to the previous example:
2011 my_memcpy (void *dest, const void *src, size_t len)
2012 __attribute__((nonnull));
2016 @cindex @code{noreturn} function attribute
2017 A few standard library functions, such as @code{abort} and @code{exit},
2018 cannot return. GCC knows this automatically. Some programs define
2019 their own functions that never return. You can declare them
2020 @code{noreturn} to tell the compiler this fact. For example,
2024 void fatal () __attribute__ ((noreturn));
2027 fatal (/* @r{@dots{}} */)
2029 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2035 The @code{noreturn} keyword tells the compiler to assume that
2036 @code{fatal} cannot return. It can then optimize without regard to what
2037 would happen if @code{fatal} ever did return. This makes slightly
2038 better code. More importantly, it helps avoid spurious warnings of
2039 uninitialized variables.
2041 The @code{noreturn} keyword does not affect the exceptional path when that
2042 applies: a @code{noreturn}-marked function may still return to the caller
2043 by throwing an exception or calling @code{longjmp}.
2045 Do not assume that registers saved by the calling function are
2046 restored before calling the @code{noreturn} function.
2048 It does not make sense for a @code{noreturn} function to have a return
2049 type other than @code{void}.
2051 The attribute @code{noreturn} is not implemented in GCC versions
2052 earlier than 2.5. An alternative way to declare that a function does
2053 not return, which works in the current version and in some older
2054 versions, is as follows:
2057 typedef void voidfn ();
2059 volatile voidfn fatal;
2062 This approach does not work in GNU C++.
2065 @cindex @code{nothrow} function attribute
2066 The @code{nothrow} attribute is used to inform the compiler that a
2067 function cannot throw an exception. For example, most functions in
2068 the standard C library can be guaranteed not to throw an exception
2069 with the notable exceptions of @code{qsort} and @code{bsearch} that
2070 take function pointer arguments. The @code{nothrow} attribute is not
2071 implemented in GCC versions earlier than 3.3.
2074 @cindex @code{pure} function attribute
2075 Many functions have no effects except the return value and their
2076 return value depends only on the parameters and/or global variables.
2077 Such a function can be subject
2078 to common subexpression elimination and loop optimization just as an
2079 arithmetic operator would be. These functions should be declared
2080 with the attribute @code{pure}. For example,
2083 int square (int) __attribute__ ((pure));
2087 says that the hypothetical function @code{square} is safe to call
2088 fewer times than the program says.
2090 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2091 Interesting non-pure functions are functions with infinite loops or those
2092 depending on volatile memory or other system resource, that may change between
2093 two consecutive calls (such as @code{feof} in a multithreading environment).
2095 The attribute @code{pure} is not implemented in GCC versions earlier
2098 @item regparm (@var{number})
2099 @cindex @code{regparm} attribute
2100 @cindex functions that are passed arguments in registers on the 386
2101 On the Intel 386, the @code{regparm} attribute causes the compiler to
2102 pass up to @var{number} integer arguments in registers EAX,
2103 EDX, and ECX instead of on the stack. Functions that take a
2104 variable number of arguments will continue to be passed all of their
2105 arguments on the stack.
2107 Beware that on some ELF systems this attribute is unsuitable for
2108 global functions in shared libraries with lazy binding (which is the
2109 default). Lazy binding will send the first call via resolving code in
2110 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2111 per the standard calling conventions. Solaris 8 is affected by this.
2112 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2113 safe since the loaders there save all registers. (Lazy binding can be
2114 disabled with the linker or the loader if desired, to avoid the
2118 @cindex @code{returns_twice} attribute
2119 The @code{returns_twice} attribute tells the compiler that a function may
2120 return more than one time. The compiler will ensure that all registers
2121 are dead before calling such a function and will emit a warning about
2122 the variables that may be clobbered after the second return from the
2123 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2124 The @code{longjmp}-like counterpart of such function, if any, might need
2125 to be marked with the @code{noreturn} attribute.
2128 @cindex save all registers on the H8/300, H8/300H, and H8S
2129 Use this attribute on the H8/300, H8/300H, and H8S to indicate that
2130 all registers except the stack pointer should be saved in the prologue
2131 regardless of whether they are used or not.
2133 @item section ("@var{section-name}")
2134 @cindex @code{section} function attribute
2135 Normally, the compiler places the code it generates in the @code{text} section.
2136 Sometimes, however, you need additional sections, or you need certain
2137 particular functions to appear in special sections. The @code{section}
2138 attribute specifies that a function lives in a particular section.
2139 For example, the declaration:
2142 extern void foobar (void) __attribute__ ((section ("bar")));
2146 puts the function @code{foobar} in the @code{bar} section.
2148 Some file formats do not support arbitrary sections so the @code{section}
2149 attribute is not available on all platforms.
2150 If you need to map the entire contents of a module to a particular
2151 section, consider using the facilities of the linker instead.
2154 @cindex @code{sentinel} function attribute
2155 This function attribute ensures that a parameter in a function call is
2156 an explicit @code{NULL}. The attribute is only valid on variadic
2157 functions. By default, the sentinel is located at position zero, the
2158 last parameter of the function call. If an optional integer position
2159 argument P is supplied to the attribute, the sentinel must be located at
2160 position P counting backwards from the end of the argument list.
2163 __attribute__ ((sentinel))
2165 __attribute__ ((sentinel(0)))
2168 The attribute is automatically set with a position of 0 for the built-in
2169 functions @code{execl} and @code{execlp}. The built-in function
2170 @code{execle} has the attribute set with a position of 1.
2172 A valid @code{NULL} in this context is defined as zero with any pointer
2173 type. If your system defines the @code{NULL} macro with an integer type
2174 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2175 with a copy that redefines NULL appropriately.
2177 The warnings for missing or incorrect sentinels are enabled with
2181 See long_call/short_call.
2184 See longcall/shortcall.
2187 @cindex signal handler functions on the AVR processors
2188 Use this attribute on the AVR to indicate that the specified
2189 function is a signal handler. The compiler will generate function
2190 entry and exit sequences suitable for use in a signal handler when this
2191 attribute is present. Interrupts will be disabled inside the function.
2194 Use this attribute on the SH to indicate an @code{interrupt_handler}
2195 function should switch to an alternate stack. It expects a string
2196 argument that names a global variable holding the address of the
2201 void f () __attribute__ ((interrupt_handler,
2202 sp_switch ("alt_stack")));
2206 @cindex functions that pop the argument stack on the 386
2207 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2208 assume that the called function will pop off the stack space used to
2209 pass arguments, unless it takes a variable number of arguments.
2212 @cindex tiny data section on the H8/300H and H8S
2213 Use this attribute on the H8/300H and H8S to indicate that the specified
2214 variable should be placed into the tiny data section.
2215 The compiler will generate more efficient code for loads and stores
2216 on data in the tiny data section. Note the tiny data area is limited to
2217 slightly under 32kbytes of data.
2220 Use this attribute on the SH for an @code{interrupt_handler} to return using
2221 @code{trapa} instead of @code{rte}. This attribute expects an integer
2222 argument specifying the trap number to be used.
2225 @cindex @code{unused} attribute.
2226 This attribute, attached to a function, means that the function is meant
2227 to be possibly unused. GCC will not produce a warning for this
2231 @cindex @code{used} attribute.
2232 This attribute, attached to a function, means that code must be emitted
2233 for the function even if it appears that the function is not referenced.
2234 This is useful, for example, when the function is referenced only in
2237 @item visibility ("@var{visibility_type}")
2238 @cindex @code{visibility} attribute
2239 The @code{visibility} attribute on ELF targets causes the declaration
2240 to be emitted with default, hidden, protected or internal visibility.
2243 void __attribute__ ((visibility ("protected")))
2244 f () @{ /* @r{Do something.} */; @}
2245 int i __attribute__ ((visibility ("hidden")));
2248 See the ELF gABI for complete details, but the short story is:
2251 @c keep this list of visibilities in alphabetical order.
2254 Default visibility is the normal case for ELF@. This value is
2255 available for the visibility attribute to override other options
2256 that may change the assumed visibility of symbols.
2259 Hidden visibility indicates that the symbol will not be placed into
2260 the dynamic symbol table, so no other @dfn{module} (executable or
2261 shared library) can reference it directly.
2264 Internal visibility is like hidden visibility, but with additional
2265 processor specific semantics. Unless otherwise specified by the psABI,
2266 GCC defines internal visibility to mean that the function is @emph{never}
2267 called from another module. Note that hidden symbols, while they cannot
2268 be referenced directly by other modules, can be referenced indirectly via
2269 function pointers. By indicating that a symbol cannot be called from
2270 outside the module, GCC may for instance omit the load of a PIC register
2271 since it is known that the calling function loaded the correct value.
2274 Protected visibility indicates that the symbol will be placed in the
2275 dynamic symbol table, but that references within the defining module
2276 will bind to the local symbol. That is, the symbol cannot be overridden
2281 Not all ELF targets support this attribute.
2283 @item warn_unused_result
2284 @cindex @code{warn_unused_result} attribute
2285 The @code{warn_unused_result} attribute causes a warning to be emitted
2286 if a caller of the function with this attribute does not use its
2287 return value. This is useful for functions where not checking
2288 the result is either a security problem or always a bug, such as
2292 int fn () __attribute__ ((warn_unused_result));
2295 if (fn () < 0) return -1;
2301 results in warning on line 5.
2304 @cindex @code{weak} attribute
2305 The @code{weak} attribute causes the declaration to be emitted as a weak
2306 symbol rather than a global. This is primarily useful in defining
2307 library functions which can be overridden in user code, though it can
2308 also be used with non-function declarations. Weak symbols are supported
2309 for ELF targets, and also for a.out targets when using the GNU assembler
2314 You can specify multiple attributes in a declaration by separating them
2315 by commas within the double parentheses or by immediately following an
2316 attribute declaration with another attribute declaration.
2318 @cindex @code{#pragma}, reason for not using
2319 @cindex pragma, reason for not using
2320 Some people object to the @code{__attribute__} feature, suggesting that
2321 ISO C's @code{#pragma} should be used instead. At the time
2322 @code{__attribute__} was designed, there were two reasons for not doing
2327 It is impossible to generate @code{#pragma} commands from a macro.
2330 There is no telling what the same @code{#pragma} might mean in another
2334 These two reasons applied to almost any application that might have been
2335 proposed for @code{#pragma}. It was basically a mistake to use
2336 @code{#pragma} for @emph{anything}.
2338 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2339 to be generated from macros. In addition, a @code{#pragma GCC}
2340 namespace is now in use for GCC-specific pragmas. However, it has been
2341 found convenient to use @code{__attribute__} to achieve a natural
2342 attachment of attributes to their corresponding declarations, whereas
2343 @code{#pragma GCC} is of use for constructs that do not naturally form
2344 part of the grammar. @xref{Other Directives,,Miscellaneous
2345 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2347 @node Attribute Syntax
2348 @section Attribute Syntax
2349 @cindex attribute syntax
2351 This section describes the syntax with which @code{__attribute__} may be
2352 used, and the constructs to which attribute specifiers bind, for the C
2353 language. Some details may vary for C++ and Objective-C@. Because of
2354 infelicities in the grammar for attributes, some forms described here
2355 may not be successfully parsed in all cases.
2357 There are some problems with the semantics of attributes in C++. For
2358 example, there are no manglings for attributes, although they may affect
2359 code generation, so problems may arise when attributed types are used in
2360 conjunction with templates or overloading. Similarly, @code{typeid}
2361 does not distinguish between types with different attributes. Support
2362 for attributes in C++ may be restricted in future to attributes on
2363 declarations only, but not on nested declarators.
2365 @xref{Function Attributes}, for details of the semantics of attributes
2366 applying to functions. @xref{Variable Attributes}, for details of the
2367 semantics of attributes applying to variables. @xref{Type Attributes},
2368 for details of the semantics of attributes applying to structure, union
2369 and enumerated types.
2371 An @dfn{attribute specifier} is of the form
2372 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2373 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2374 each attribute is one of the following:
2378 Empty. Empty attributes are ignored.
2381 A word (which may be an identifier such as @code{unused}, or a reserved
2382 word such as @code{const}).
2385 A word, followed by, in parentheses, parameters for the attribute.
2386 These parameters take one of the following forms:
2390 An identifier. For example, @code{mode} attributes use this form.
2393 An identifier followed by a comma and a non-empty comma-separated list
2394 of expressions. For example, @code{format} attributes use this form.
2397 A possibly empty comma-separated list of expressions. For example,
2398 @code{format_arg} attributes use this form with the list being a single
2399 integer constant expression, and @code{alias} attributes use this form
2400 with the list being a single string constant.
2404 An @dfn{attribute specifier list} is a sequence of one or more attribute
2405 specifiers, not separated by any other tokens.
2407 In GNU C, an attribute specifier list may appear after the colon following a
2408 label, other than a @code{case} or @code{default} label. The only
2409 attribute it makes sense to use after a label is @code{unused}. This
2410 feature is intended for code generated by programs which contains labels
2411 that may be unused but which is compiled with @option{-Wall}. It would
2412 not normally be appropriate to use in it human-written code, though it
2413 could be useful in cases where the code that jumps to the label is
2414 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2415 such placement of attribute lists, as it is permissible for a
2416 declaration, which could begin with an attribute list, to be labelled in
2417 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2418 does not arise there.
2420 An attribute specifier list may appear as part of a @code{struct},
2421 @code{union} or @code{enum} specifier. It may go either immediately
2422 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2423 the closing brace. It is ignored if the content of the structure, union
2424 or enumerated type is not defined in the specifier in which the
2425 attribute specifier list is used---that is, in usages such as
2426 @code{struct __attribute__((foo)) bar} with no following opening brace.
2427 Where attribute specifiers follow the closing brace, they are considered
2428 to relate to the structure, union or enumerated type defined, not to any
2429 enclosing declaration the type specifier appears in, and the type
2430 defined is not complete until after the attribute specifiers.
2431 @c Otherwise, there would be the following problems: a shift/reduce
2432 @c conflict between attributes binding the struct/union/enum and
2433 @c binding to the list of specifiers/qualifiers; and "aligned"
2434 @c attributes could use sizeof for the structure, but the size could be
2435 @c changed later by "packed" attributes.
2437 Otherwise, an attribute specifier appears as part of a declaration,
2438 counting declarations of unnamed parameters and type names, and relates
2439 to that declaration (which may be nested in another declaration, for
2440 example in the case of a parameter declaration), or to a particular declarator
2441 within a declaration. Where an
2442 attribute specifier is applied to a parameter declared as a function or
2443 an array, it should apply to the function or array rather than the
2444 pointer to which the parameter is implicitly converted, but this is not
2445 yet correctly implemented.
2447 Any list of specifiers and qualifiers at the start of a declaration may
2448 contain attribute specifiers, whether or not such a list may in that
2449 context contain storage class specifiers. (Some attributes, however,
2450 are essentially in the nature of storage class specifiers, and only make
2451 sense where storage class specifiers may be used; for example,
2452 @code{section}.) There is one necessary limitation to this syntax: the
2453 first old-style parameter declaration in a function definition cannot
2454 begin with an attribute specifier, because such an attribute applies to
2455 the function instead by syntax described below (which, however, is not
2456 yet implemented in this case). In some other cases, attribute
2457 specifiers are permitted by this grammar but not yet supported by the
2458 compiler. All attribute specifiers in this place relate to the
2459 declaration as a whole. In the obsolescent usage where a type of
2460 @code{int} is implied by the absence of type specifiers, such a list of
2461 specifiers and qualifiers may be an attribute specifier list with no
2462 other specifiers or qualifiers.
2464 At present, the first parameter in a function prototype must have some
2465 type specifier which is not an attribute specifier; this resolves an
2466 ambiguity in the interpretation of @code{void f(int
2467 (__attribute__((foo)) x))}, but is subject to change. At present, if
2468 the parentheses of a function declarator contain only attributes then
2469 those attributes are ignored, rather than yielding an error or warning
2470 or implying a single parameter of type int, but this is subject to
2473 An attribute specifier list may appear immediately before a declarator
2474 (other than the first) in a comma-separated list of declarators in a
2475 declaration of more than one identifier using a single list of
2476 specifiers and qualifiers. Such attribute specifiers apply
2477 only to the identifier before whose declarator they appear. For
2481 __attribute__((noreturn)) void d0 (void),
2482 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2487 the @code{noreturn} attribute applies to all the functions
2488 declared; the @code{format} attribute only applies to @code{d1}.
2490 An attribute specifier list may appear immediately before the comma,
2491 @code{=} or semicolon terminating the declaration of an identifier other
2492 than a function definition. At present, such attribute specifiers apply
2493 to the declared object or function, but in future they may attach to the
2494 outermost adjacent declarator. In simple cases there is no difference,
2495 but, for example, in
2498 void (****f)(void) __attribute__((noreturn));
2502 at present the @code{noreturn} attribute applies to @code{f}, which
2503 causes a warning since @code{f} is not a function, but in future it may
2504 apply to the function @code{****f}. The precise semantics of what
2505 attributes in such cases will apply to are not yet specified. Where an
2506 assembler name for an object or function is specified (@pxref{Asm
2507 Labels}), at present the attribute must follow the @code{asm}
2508 specification; in future, attributes before the @code{asm} specification
2509 may apply to the adjacent declarator, and those after it to the declared
2512 An attribute specifier list may, in future, be permitted to appear after
2513 the declarator in a function definition (before any old-style parameter
2514 declarations or the function body).
2516 Attribute specifiers may be mixed with type qualifiers appearing inside
2517 the @code{[]} of a parameter array declarator, in the C99 construct by
2518 which such qualifiers are applied to the pointer to which the array is
2519 implicitly converted. Such attribute specifiers apply to the pointer,
2520 not to the array, but at present this is not implemented and they are
2523 An attribute specifier list may appear at the start of a nested
2524 declarator. At present, there are some limitations in this usage: the
2525 attributes correctly apply to the declarator, but for most individual
2526 attributes the semantics this implies are not implemented.
2527 When attribute specifiers follow the @code{*} of a pointer
2528 declarator, they may be mixed with any type qualifiers present.
2529 The following describes the formal semantics of this syntax. It will make the
2530 most sense if you are familiar with the formal specification of
2531 declarators in the ISO C standard.
2533 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2534 D1}, where @code{T} contains declaration specifiers that specify a type
2535 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2536 contains an identifier @var{ident}. The type specified for @var{ident}
2537 for derived declarators whose type does not include an attribute
2538 specifier is as in the ISO C standard.
2540 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2541 and the declaration @code{T D} specifies the type
2542 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2543 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2544 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2546 If @code{D1} has the form @code{*
2547 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2548 declaration @code{T D} specifies the type
2549 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2550 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2551 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2557 void (__attribute__((noreturn)) ****f) (void);
2561 specifies the type ``pointer to pointer to pointer to pointer to
2562 non-returning function returning @code{void}''. As another example,
2565 char *__attribute__((aligned(8))) *f;
2569 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2570 Note again that this does not work with most attributes; for example,
2571 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2572 is not yet supported.
2574 For compatibility with existing code written for compiler versions that
2575 did not implement attributes on nested declarators, some laxity is
2576 allowed in the placing of attributes. If an attribute that only applies
2577 to types is applied to a declaration, it will be treated as applying to
2578 the type of that declaration. If an attribute that only applies to
2579 declarations is applied to the type of a declaration, it will be treated
2580 as applying to that declaration; and, for compatibility with code
2581 placing the attributes immediately before the identifier declared, such
2582 an attribute applied to a function return type will be treated as
2583 applying to the function type, and such an attribute applied to an array
2584 element type will be treated as applying to the array type. If an
2585 attribute that only applies to function types is applied to a
2586 pointer-to-function type, it will be treated as applying to the pointer
2587 target type; if such an attribute is applied to a function return type
2588 that is not a pointer-to-function type, it will be treated as applying
2589 to the function type.
2591 @node Function Prototypes
2592 @section Prototypes and Old-Style Function Definitions
2593 @cindex function prototype declarations
2594 @cindex old-style function definitions
2595 @cindex promotion of formal parameters
2597 GNU C extends ISO C to allow a function prototype to override a later
2598 old-style non-prototype definition. Consider the following example:
2601 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2608 /* @r{Prototype function declaration.} */
2609 int isroot P((uid_t));
2611 /* @r{Old-style function definition.} */
2613 isroot (x) /* @r{??? lossage here ???} */
2620 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2621 not allow this example, because subword arguments in old-style
2622 non-prototype definitions are promoted. Therefore in this example the
2623 function definition's argument is really an @code{int}, which does not
2624 match the prototype argument type of @code{short}.
2626 This restriction of ISO C makes it hard to write code that is portable
2627 to traditional C compilers, because the programmer does not know
2628 whether the @code{uid_t} type is @code{short}, @code{int}, or
2629 @code{long}. Therefore, in cases like these GNU C allows a prototype
2630 to override a later old-style definition. More precisely, in GNU C, a
2631 function prototype argument type overrides the argument type specified
2632 by a later old-style definition if the former type is the same as the
2633 latter type before promotion. Thus in GNU C the above example is
2634 equivalent to the following:
2647 GNU C++ does not support old-style function definitions, so this
2648 extension is irrelevant.
2651 @section C++ Style Comments
2653 @cindex C++ comments
2654 @cindex comments, C++ style
2656 In GNU C, you may use C++ style comments, which start with @samp{//} and
2657 continue until the end of the line. Many other C implementations allow
2658 such comments, and they are included in the 1999 C standard. However,
2659 C++ style comments are not recognized if you specify an @option{-std}
2660 option specifying a version of ISO C before C99, or @option{-ansi}
2661 (equivalent to @option{-std=c89}).
2664 @section Dollar Signs in Identifier Names
2666 @cindex dollar signs in identifier names
2667 @cindex identifier names, dollar signs in
2669 In GNU C, you may normally use dollar signs in identifier names.
2670 This is because many traditional C implementations allow such identifiers.
2671 However, dollar signs in identifiers are not supported on a few target
2672 machines, typically because the target assembler does not allow them.
2674 @node Character Escapes
2675 @section The Character @key{ESC} in Constants
2677 You can use the sequence @samp{\e} in a string or character constant to
2678 stand for the ASCII character @key{ESC}.
2681 @section Inquiring on Alignment of Types or Variables
2683 @cindex type alignment
2684 @cindex variable alignment
2686 The keyword @code{__alignof__} allows you to inquire about how an object
2687 is aligned, or the minimum alignment usually required by a type. Its
2688 syntax is just like @code{sizeof}.
2690 For example, if the target machine requires a @code{double} value to be
2691 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2692 This is true on many RISC machines. On more traditional machine
2693 designs, @code{__alignof__ (double)} is 4 or even 2.
2695 Some machines never actually require alignment; they allow reference to any
2696 data type even at an odd address. For these machines, @code{__alignof__}
2697 reports the @emph{recommended} alignment of a type.
2699 If the operand of @code{__alignof__} is an lvalue rather than a type,
2700 its value is the required alignment for its type, taking into account
2701 any minimum alignment specified with GCC's @code{__attribute__}
2702 extension (@pxref{Variable Attributes}). For example, after this
2706 struct foo @{ int x; char y; @} foo1;
2710 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2711 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2713 It is an error to ask for the alignment of an incomplete type.
2715 @node Variable Attributes
2716 @section Specifying Attributes of Variables
2717 @cindex attribute of variables
2718 @cindex variable attributes
2720 The keyword @code{__attribute__} allows you to specify special
2721 attributes of variables or structure fields. This keyword is followed
2722 by an attribute specification inside double parentheses. Some
2723 attributes are currently defined generically for variables.
2724 Other attributes are defined for variables on particular target
2725 systems. Other attributes are available for functions
2726 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2727 Other front ends might define more attributes
2728 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2730 You may also specify attributes with @samp{__} preceding and following
2731 each keyword. This allows you to use them in header files without
2732 being concerned about a possible macro of the same name. For example,
2733 you may use @code{__aligned__} instead of @code{aligned}.
2735 @xref{Attribute Syntax}, for details of the exact syntax for using
2739 @cindex @code{aligned} attribute
2740 @item aligned (@var{alignment})
2741 This attribute specifies a minimum alignment for the variable or
2742 structure field, measured in bytes. For example, the declaration:
2745 int x __attribute__ ((aligned (16))) = 0;
2749 causes the compiler to allocate the global variable @code{x} on a
2750 16-byte boundary. On a 68040, this could be used in conjunction with
2751 an @code{asm} expression to access the @code{move16} instruction which
2752 requires 16-byte aligned operands.
2754 You can also specify the alignment of structure fields. For example, to
2755 create a double-word aligned @code{int} pair, you could write:
2758 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2762 This is an alternative to creating a union with a @code{double} member
2763 that forces the union to be double-word aligned.
2765 As in the preceding examples, you can explicitly specify the alignment
2766 (in bytes) that you wish the compiler to use for a given variable or
2767 structure field. Alternatively, you can leave out the alignment factor
2768 and just ask the compiler to align a variable or field to the maximum
2769 useful alignment for the target machine you are compiling for. For
2770 example, you could write:
2773 short array[3] __attribute__ ((aligned));
2776 Whenever you leave out the alignment factor in an @code{aligned} attribute
2777 specification, the compiler automatically sets the alignment for the declared
2778 variable or field to the largest alignment which is ever used for any data
2779 type on the target machine you are compiling for. Doing this can often make
2780 copy operations more efficient, because the compiler can use whatever
2781 instructions copy the biggest chunks of memory when performing copies to
2782 or from the variables or fields that you have aligned this way.
2784 The @code{aligned} attribute can only increase the alignment; but you
2785 can decrease it by specifying @code{packed} as well. See below.
2787 Note that the effectiveness of @code{aligned} attributes may be limited
2788 by inherent limitations in your linker. On many systems, the linker is
2789 only able to arrange for variables to be aligned up to a certain maximum
2790 alignment. (For some linkers, the maximum supported alignment may
2791 be very very small.) If your linker is only able to align variables
2792 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
2793 in an @code{__attribute__} will still only provide you with 8 byte
2794 alignment. See your linker documentation for further information.
2796 @item cleanup (@var{cleanup_function})
2797 @cindex @code{cleanup} attribute
2798 The @code{cleanup} attribute runs a function when the variable goes
2799 out of scope. This attribute can only be applied to auto function
2800 scope variables; it may not be applied to parameters or variables
2801 with static storage duration. The function must take one parameter,
2802 a pointer to a type compatible with the variable. The return value
2803 of the function (if any) is ignored.
2805 If @option{-fexceptions} is enabled, then @var{cleanup_function}
2806 will be run during the stack unwinding that happens during the
2807 processing of the exception. Note that the @code{cleanup} attribute
2808 does not allow the exception to be caught, only to perform an action.
2809 It is undefined what happens if @var{cleanup_function} does not
2814 @cindex @code{common} attribute
2815 @cindex @code{nocommon} attribute
2818 The @code{common} attribute requests GCC to place a variable in
2819 ``common'' storage. The @code{nocommon} attribute requests the
2820 opposite---to allocate space for it directly.
2822 These attributes override the default chosen by the
2823 @option{-fno-common} and @option{-fcommon} flags respectively.
2826 @cindex @code{deprecated} attribute
2827 The @code{deprecated} attribute results in a warning if the variable
2828 is used anywhere in the source file. This is useful when identifying
2829 variables that are expected to be removed in a future version of a
2830 program. The warning also includes the location of the declaration
2831 of the deprecated variable, to enable users to easily find further
2832 information about why the variable is deprecated, or what they should
2833 do instead. Note that the warning only occurs for uses:
2836 extern int old_var __attribute__ ((deprecated));
2838 int new_fn () @{ return old_var; @}
2841 results in a warning on line 3 but not line 2.
2843 The @code{deprecated} attribute can also be used for functions and
2844 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
2846 @item mode (@var{mode})
2847 @cindex @code{mode} attribute
2848 This attribute specifies the data type for the declaration---whichever
2849 type corresponds to the mode @var{mode}. This in effect lets you
2850 request an integer or floating point type according to its width.
2852 You may also specify a mode of @samp{byte} or @samp{__byte__} to
2853 indicate the mode corresponding to a one-byte integer, @samp{word} or
2854 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
2855 or @samp{__pointer__} for the mode used to represent pointers.
2858 @cindex @code{packed} attribute
2859 The @code{packed} attribute specifies that a variable or structure field
2860 should have the smallest possible alignment---one byte for a variable,
2861 and one bit for a field, unless you specify a larger value with the
2862 @code{aligned} attribute.
2864 Here is a structure in which the field @code{x} is packed, so that it
2865 immediately follows @code{a}:
2871 int x[2] __attribute__ ((packed));
2875 @item section ("@var{section-name}")
2876 @cindex @code{section} variable attribute
2877 Normally, the compiler places the objects it generates in sections like
2878 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
2879 or you need certain particular variables to appear in special sections,
2880 for example to map to special hardware. The @code{section}
2881 attribute specifies that a variable (or function) lives in a particular
2882 section. For example, this small program uses several specific section names:
2885 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
2886 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
2887 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
2888 int init_data __attribute__ ((section ("INITDATA"))) = 0;
2892 /* @r{Initialize stack pointer} */
2893 init_sp (stack + sizeof (stack));
2895 /* @r{Initialize initialized data} */
2896 memcpy (&init_data, &data, &edata - &data);
2898 /* @r{Turn on the serial ports} */
2905 Use the @code{section} attribute with an @emph{initialized} definition
2906 of a @emph{global} variable, as shown in the example. GCC issues
2907 a warning and otherwise ignores the @code{section} attribute in
2908 uninitialized variable declarations.
2910 You may only use the @code{section} attribute with a fully initialized
2911 global definition because of the way linkers work. The linker requires
2912 each object be defined once, with the exception that uninitialized
2913 variables tentatively go in the @code{common} (or @code{bss}) section
2914 and can be multiply ``defined''. You can force a variable to be
2915 initialized with the @option{-fno-common} flag or the @code{nocommon}
2918 Some file formats do not support arbitrary sections so the @code{section}
2919 attribute is not available on all platforms.
2920 If you need to map the entire contents of a module to a particular
2921 section, consider using the facilities of the linker instead.
2924 @cindex @code{shared} variable attribute
2925 On Microsoft Windows, in addition to putting variable definitions in a named
2926 section, the section can also be shared among all running copies of an
2927 executable or DLL@. For example, this small program defines shared data
2928 by putting it in a named section @code{shared} and marking the section
2932 int foo __attribute__((section ("shared"), shared)) = 0;
2937 /* @r{Read and write foo. All running
2938 copies see the same value.} */
2944 You may only use the @code{shared} attribute along with @code{section}
2945 attribute with a fully initialized global definition because of the way
2946 linkers work. See @code{section} attribute for more information.
2948 The @code{shared} attribute is only available on Microsoft Windows@.
2950 @item tls_model ("@var{tls_model}")
2951 @cindex @code{tls_model} attribute
2952 The @code{tls_model} attribute sets thread-local storage model
2953 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
2954 overriding @option{-ftls-model=} command line switch on a per-variable
2956 The @var{tls_model} argument should be one of @code{global-dynamic},
2957 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
2959 Not all targets support this attribute.
2961 @item transparent_union
2962 This attribute, attached to a function parameter which is a union, means
2963 that the corresponding argument may have the type of any union member,
2964 but the argument is passed as if its type were that of the first union
2965 member. For more details see @xref{Type Attributes}. You can also use
2966 this attribute on a @code{typedef} for a union data type; then it
2967 applies to all function parameters with that type.
2970 This attribute, attached to a variable, means that the variable is meant
2971 to be possibly unused. GCC will not produce a warning for this
2974 @item vector_size (@var{bytes})
2975 This attribute specifies the vector size for the variable, measured in
2976 bytes. For example, the declaration:
2979 int foo __attribute__ ((vector_size (16)));
2983 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
2984 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
2985 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
2987 This attribute is only applicable to integral and float scalars,
2988 although arrays, pointers, and function return values are allowed in
2989 conjunction with this construct.
2991 Aggregates with this attribute are invalid, even if they are of the same
2992 size as a corresponding scalar. For example, the declaration:
2995 struct S @{ int a; @};
2996 struct S __attribute__ ((vector_size (16))) foo;
3000 is invalid even if the size of the structure is the same as the size of
3004 The @code{weak} attribute is described in @xref{Function Attributes}.
3007 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3010 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3014 @subsection M32R/D Variable Attributes
3016 One attribute is currently defined for the M32R/D@.
3019 @item model (@var{model-name})
3020 @cindex variable addressability on the M32R/D
3021 Use this attribute on the M32R/D to set the addressability of an object.
3022 The identifier @var{model-name} is one of @code{small}, @code{medium},
3023 or @code{large}, representing each of the code models.
3025 Small model objects live in the lower 16MB of memory (so that their
3026 addresses can be loaded with the @code{ld24} instruction).
3028 Medium and large model objects may live anywhere in the 32-bit address space
3029 (the compiler will generate @code{seth/add3} instructions to load their
3033 @subsection i386 Variable Attributes
3035 Two attributes are currently defined for i386 configurations:
3036 @code{ms_struct} and @code{gcc_struct}
3041 @cindex @code{ms_struct} attribute
3042 @cindex @code{gcc_struct} attribute
3044 If @code{packed} is used on a structure, or if bit-fields are used
3045 it may be that the Microsoft ABI packs them differently
3046 than GCC would normally pack them. Particularly when moving packed
3047 data between functions compiled with GCC and the native Microsoft compiler
3048 (either via function call or as data in a file), it may be necessary to access
3051 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3052 compilers to match the native Microsoft compiler.
3055 @subsection Xstormy16 Variable Attributes
3057 One attribute is currently defined for xstormy16 configurations:
3062 @cindex @code{below100} attribute
3064 If a variable has the @code{below100} attribute (@code{BELOW100} is
3065 allowed also), GCC will place the variable in the first 0x100 bytes of
3066 memory and use special opcodes to access it. Such variables will be
3067 placed in either the @code{.bss_below100} section or the
3068 @code{.data_below100} section.
3072 @node Type Attributes
3073 @section Specifying Attributes of Types
3074 @cindex attribute of types
3075 @cindex type attributes
3077 The keyword @code{__attribute__} allows you to specify special
3078 attributes of @code{struct} and @code{union} types when you define such
3079 types. This keyword is followed by an attribute specification inside
3080 double parentheses. Six attributes are currently defined for types:
3081 @code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3082 @code{deprecated} and @code{may_alias}. Other attributes are defined for
3083 functions (@pxref{Function Attributes}) and for variables
3084 (@pxref{Variable Attributes}).
3086 You may also specify any one of these attributes with @samp{__}
3087 preceding and following its keyword. This allows you to use these
3088 attributes in header files without being concerned about a possible
3089 macro of the same name. For example, you may use @code{__aligned__}
3090 instead of @code{aligned}.
3092 You may specify the @code{aligned} and @code{transparent_union}
3093 attributes either in a @code{typedef} declaration or just past the
3094 closing curly brace of a complete enum, struct or union type
3095 @emph{definition} and the @code{packed} attribute only past the closing
3096 brace of a definition.
3098 You may also specify attributes between the enum, struct or union
3099 tag and the name of the type rather than after the closing brace.
3101 @xref{Attribute Syntax}, for details of the exact syntax for using
3105 @cindex @code{aligned} attribute
3106 @item aligned (@var{alignment})
3107 This attribute specifies a minimum alignment (in bytes) for variables
3108 of the specified type. For example, the declarations:
3111 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3112 typedef int more_aligned_int __attribute__ ((aligned (8)));
3116 force the compiler to insure (as far as it can) that each variable whose
3117 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3118 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3119 variables of type @code{struct S} aligned to 8-byte boundaries allows
3120 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3121 store) instructions when copying one variable of type @code{struct S} to
3122 another, thus improving run-time efficiency.
3124 Note that the alignment of any given @code{struct} or @code{union} type
3125 is required by the ISO C standard to be at least a perfect multiple of
3126 the lowest common multiple of the alignments of all of the members of
3127 the @code{struct} or @code{union} in question. This means that you @emph{can}
3128 effectively adjust the alignment of a @code{struct} or @code{union}
3129 type by attaching an @code{aligned} attribute to any one of the members
3130 of such a type, but the notation illustrated in the example above is a
3131 more obvious, intuitive, and readable way to request the compiler to
3132 adjust the alignment of an entire @code{struct} or @code{union} type.
3134 As in the preceding example, you can explicitly specify the alignment
3135 (in bytes) that you wish the compiler to use for a given @code{struct}
3136 or @code{union} type. Alternatively, you can leave out the alignment factor
3137 and just ask the compiler to align a type to the maximum
3138 useful alignment for the target machine you are compiling for. For
3139 example, you could write:
3142 struct S @{ short f[3]; @} __attribute__ ((aligned));
3145 Whenever you leave out the alignment factor in an @code{aligned}
3146 attribute specification, the compiler automatically sets the alignment
3147 for the type to the largest alignment which is ever used for any data
3148 type on the target machine you are compiling for. Doing this can often
3149 make copy operations more efficient, because the compiler can use
3150 whatever instructions copy the biggest chunks of memory when performing
3151 copies to or from the variables which have types that you have aligned
3154 In the example above, if the size of each @code{short} is 2 bytes, then
3155 the size of the entire @code{struct S} type is 6 bytes. The smallest
3156 power of two which is greater than or equal to that is 8, so the
3157 compiler sets the alignment for the entire @code{struct S} type to 8
3160 Note that although you can ask the compiler to select a time-efficient
3161 alignment for a given type and then declare only individual stand-alone
3162 objects of that type, the compiler's ability to select a time-efficient
3163 alignment is primarily useful only when you plan to create arrays of
3164 variables having the relevant (efficiently aligned) type. If you
3165 declare or use arrays of variables of an efficiently-aligned type, then
3166 it is likely that your program will also be doing pointer arithmetic (or
3167 subscripting, which amounts to the same thing) on pointers to the
3168 relevant type, and the code that the compiler generates for these
3169 pointer arithmetic operations will often be more efficient for
3170 efficiently-aligned types than for other types.
3172 The @code{aligned} attribute can only increase the alignment; but you
3173 can decrease it by specifying @code{packed} as well. See below.
3175 Note that the effectiveness of @code{aligned} attributes may be limited
3176 by inherent limitations in your linker. On many systems, the linker is
3177 only able to arrange for variables to be aligned up to a certain maximum
3178 alignment. (For some linkers, the maximum supported alignment may
3179 be very very small.) If your linker is only able to align variables
3180 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3181 in an @code{__attribute__} will still only provide you with 8 byte
3182 alignment. See your linker documentation for further information.
3185 This attribute, attached to @code{struct} or @code{union} type
3186 definition, specifies that each member of the structure or union is
3187 placed to minimize the memory required. When attached to an @code{enum}
3188 definition, it indicates that the smallest integral type should be used.
3190 @opindex fshort-enums
3191 Specifying this attribute for @code{struct} and @code{union} types is
3192 equivalent to specifying the @code{packed} attribute on each of the
3193 structure or union members. Specifying the @option{-fshort-enums}
3194 flag on the line is equivalent to specifying the @code{packed}
3195 attribute on all @code{enum} definitions.
3197 In the following example @code{struct my_packed_struct}'s members are
3198 packed closely together, but the internal layout of its @code{s} member
3199 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3203 struct my_unpacked_struct
3209 struct my_packed_struct __attribute__ ((__packed__))
3213 struct my_unpacked_struct s;
3217 You may only specify this attribute on the definition of a @code{enum},
3218 @code{struct} or @code{union}, not on a @code{typedef} which does not
3219 also define the enumerated type, structure or union.
3221 @item transparent_union
3222 This attribute, attached to a @code{union} type definition, indicates
3223 that any function parameter having that union type causes calls to that
3224 function to be treated in a special way.
3226 First, the argument corresponding to a transparent union type can be of
3227 any type in the union; no cast is required. Also, if the union contains
3228 a pointer type, the corresponding argument can be a null pointer
3229 constant or a void pointer expression; and if the union contains a void
3230 pointer type, the corresponding argument can be any pointer expression.
3231 If the union member type is a pointer, qualifiers like @code{const} on
3232 the referenced type must be respected, just as with normal pointer
3235 Second, the argument is passed to the function using the calling
3236 conventions of the first member of the transparent union, not the calling
3237 conventions of the union itself. All members of the union must have the
3238 same machine representation; this is necessary for this argument passing
3241 Transparent unions are designed for library functions that have multiple
3242 interfaces for compatibility reasons. For example, suppose the
3243 @code{wait} function must accept either a value of type @code{int *} to
3244 comply with Posix, or a value of type @code{union wait *} to comply with
3245 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3246 @code{wait} would accept both kinds of arguments, but it would also
3247 accept any other pointer type and this would make argument type checking
3248 less useful. Instead, @code{<sys/wait.h>} might define the interface
3256 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3258 pid_t wait (wait_status_ptr_t);
3261 This interface allows either @code{int *} or @code{union wait *}
3262 arguments to be passed, using the @code{int *} calling convention.
3263 The program can call @code{wait} with arguments of either type:
3266 int w1 () @{ int w; return wait (&w); @}
3267 int w2 () @{ union wait w; return wait (&w); @}
3270 With this interface, @code{wait}'s implementation might look like this:
3273 pid_t wait (wait_status_ptr_t p)
3275 return waitpid (-1, p.__ip, 0);
3280 When attached to a type (including a @code{union} or a @code{struct}),
3281 this attribute means that variables of that type are meant to appear
3282 possibly unused. GCC will not produce a warning for any variables of
3283 that type, even if the variable appears to do nothing. This is often
3284 the case with lock or thread classes, which are usually defined and then
3285 not referenced, but contain constructors and destructors that have
3286 nontrivial bookkeeping functions.
3289 The @code{deprecated} attribute results in a warning if the type
3290 is used anywhere in the source file. This is useful when identifying
3291 types that are expected to be removed in a future version of a program.
3292 If possible, the warning also includes the location of the declaration
3293 of the deprecated type, to enable users to easily find further
3294 information about why the type is deprecated, or what they should do
3295 instead. Note that the warnings only occur for uses and then only
3296 if the type is being applied to an identifier that itself is not being
3297 declared as deprecated.
3300 typedef int T1 __attribute__ ((deprecated));
3304 typedef T1 T3 __attribute__ ((deprecated));
3305 T3 z __attribute__ ((deprecated));
3308 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3309 warning is issued for line 4 because T2 is not explicitly
3310 deprecated. Line 5 has no warning because T3 is explicitly
3311 deprecated. Similarly for line 6.
3313 The @code{deprecated} attribute can also be used for functions and
3314 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3317 Accesses to objects with types with this attribute are not subjected to
3318 type-based alias analysis, but are instead assumed to be able to alias
3319 any other type of objects, just like the @code{char} type. See
3320 @option{-fstrict-aliasing} for more information on aliasing issues.
3325 typedef short __attribute__((__may_alias__)) short_a;
3331 short_a *b = (short_a *) &a;
3335 if (a == 0x12345678)
3342 If you replaced @code{short_a} with @code{short} in the variable
3343 declaration, the above program would abort when compiled with
3344 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3345 above in recent GCC versions.
3347 @subsection ARM Type Attributes
3349 On those ARM targets that support @code{dllimport} (such as Symbian
3350 OS), you can use the @code{notshared} attribute to indicate that the
3351 virtual table and other similar data for a class should not be
3352 exported from a DLL@. For example:
3355 class __declspec(notshared) C @{
3357 __declspec(dllimport) C();
3361 __declspec(dllexport)
3365 In this code, @code{C::C} is exported from the current DLL, but the
3366 virtual table for @code{C} is not exported. (You can use
3367 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3368 most Symbian OS code uses @code{__declspec}.)
3370 @subsection i386 Type Attributes
3372 Two attributes are currently defined for i386 configurations:
3373 @code{ms_struct} and @code{gcc_struct}
3377 @cindex @code{ms_struct}
3378 @cindex @code{gcc_struct}
3380 If @code{packed} is used on a structure, or if bit-fields are used
3381 it may be that the Microsoft ABI packs them differently
3382 than GCC would normally pack them. Particularly when moving packed
3383 data between functions compiled with GCC and the native Microsoft compiler
3384 (either via function call or as data in a file), it may be necessary to access
3387 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3388 compilers to match the native Microsoft compiler.
3391 To specify multiple attributes, separate them by commas within the
3392 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3396 @section An Inline Function is As Fast As a Macro
3397 @cindex inline functions
3398 @cindex integrating function code
3400 @cindex macros, inline alternative
3402 By declaring a function @code{inline}, you can direct GCC to
3403 integrate that function's code into the code for its callers. This
3404 makes execution faster by eliminating the function-call overhead; in
3405 addition, if any of the actual argument values are constant, their known
3406 values may permit simplifications at compile time so that not all of the
3407 inline function's code needs to be included. The effect on code size is
3408 less predictable; object code may be larger or smaller with function
3409 inlining, depending on the particular case. Inlining of functions is an
3410 optimization and it really ``works'' only in optimizing compilation. If
3411 you don't use @option{-O}, no function is really inline.
3413 Inline functions are included in the ISO C99 standard, but there are
3414 currently substantial differences between what GCC implements and what
3415 the ISO C99 standard requires.
3417 To declare a function inline, use the @code{inline} keyword in its
3418 declaration, like this:
3428 (If you are writing a header file to be included in ISO C programs, write
3429 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.)
3430 You can also make all ``simple enough'' functions inline with the option
3431 @option{-finline-functions}.
3434 Note that certain usages in a function definition can make it unsuitable
3435 for inline substitution. Among these usages are: use of varargs, use of
3436 alloca, use of variable sized data types (@pxref{Variable Length}),
3437 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3438 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3439 will warn when a function marked @code{inline} could not be substituted,
3440 and will give the reason for the failure.
3442 Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3443 does not affect the linkage of the function.
3445 @cindex automatic @code{inline} for C++ member fns
3446 @cindex @code{inline} automatic for C++ member fns
3447 @cindex member fns, automatically @code{inline}
3448 @cindex C++ member fns, automatically @code{inline}
3449 @opindex fno-default-inline
3450 GCC automatically inlines member functions defined within the class
3451 body of C++ programs even if they are not explicitly declared
3452 @code{inline}. (You can override this with @option{-fno-default-inline};
3453 @pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3455 @cindex inline functions, omission of
3456 @opindex fkeep-inline-functions
3457 When a function is both inline and @code{static}, if all calls to the
3458 function are integrated into the caller, and the function's address is
3459 never used, then the function's own assembler code is never referenced.
3460 In this case, GCC does not actually output assembler code for the
3461 function, unless you specify the option @option{-fkeep-inline-functions}.
3462 Some calls cannot be integrated for various reasons (in particular,
3463 calls that precede the function's definition cannot be integrated, and
3464 neither can recursive calls within the definition). If there is a
3465 nonintegrated call, then the function is compiled to assembler code as
3466 usual. The function must also be compiled as usual if the program
3467 refers to its address, because that can't be inlined.
3469 @cindex non-static inline function
3470 When an inline function is not @code{static}, then the compiler must assume
3471 that there may be calls from other source files; since a global symbol can
3472 be defined only once in any program, the function must not be defined in
3473 the other source files, so the calls therein cannot be integrated.
3474 Therefore, a non-@code{static} inline function is always compiled on its
3475 own in the usual fashion.
3477 If you specify both @code{inline} and @code{extern} in the function
3478 definition, then the definition is used only for inlining. In no case
3479 is the function compiled on its own, not even if you refer to its
3480 address explicitly. Such an address becomes an external reference, as
3481 if you had only declared the function, and had not defined it.
3483 This combination of @code{inline} and @code{extern} has almost the
3484 effect of a macro. The way to use it is to put a function definition in
3485 a header file with these keywords, and put another copy of the
3486 definition (lacking @code{inline} and @code{extern}) in a library file.
3487 The definition in the header file will cause most calls to the function
3488 to be inlined. If any uses of the function remain, they will refer to
3489 the single copy in the library.
3491 Since GCC eventually will implement ISO C99 semantics for
3492 inline functions, it is best to use @code{static inline} only
3493 to guarantee compatibility. (The
3494 existing semantics will remain available when @option{-std=gnu89} is
3495 specified, but eventually the default will be @option{-std=gnu99} and
3496 that will implement the C99 semantics, though it does not do so yet.)
3498 GCC does not inline any functions when not optimizing unless you specify
3499 the @samp{always_inline} attribute for the function, like this:
3502 /* @r{Prototype.} */
3503 inline void foo (const char) __attribute__((always_inline));
3507 @section Assembler Instructions with C Expression Operands
3508 @cindex extended @code{asm}
3509 @cindex @code{asm} expressions
3510 @cindex assembler instructions
3513 In an assembler instruction using @code{asm}, you can specify the
3514 operands of the instruction using C expressions. This means you need not
3515 guess which registers or memory locations will contain the data you want
3518 You must specify an assembler instruction template much like what
3519 appears in a machine description, plus an operand constraint string for
3522 For example, here is how to use the 68881's @code{fsinx} instruction:
3525 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3529 Here @code{angle} is the C expression for the input operand while
3530 @code{result} is that of the output operand. Each has @samp{"f"} as its
3531 operand constraint, saying that a floating point register is required.
3532 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3533 output operands' constraints must use @samp{=}. The constraints use the
3534 same language used in the machine description (@pxref{Constraints}).
3536 Each operand is described by an operand-constraint string followed by
3537 the C expression in parentheses. A colon separates the assembler
3538 template from the first output operand and another separates the last
3539 output operand from the first input, if any. Commas separate the
3540 operands within each group. The total number of operands is currently
3541 limited to 30; this limitation may be lifted in some future version of
3544 If there are no output operands but there are input operands, you must
3545 place two consecutive colons surrounding the place where the output
3548 As of GCC version 3.1, it is also possible to specify input and output
3549 operands using symbolic names which can be referenced within the
3550 assembler code. These names are specified inside square brackets
3551 preceding the constraint string, and can be referenced inside the
3552 assembler code using @code{%[@var{name}]} instead of a percentage sign
3553 followed by the operand number. Using named operands the above example
3557 asm ("fsinx %[angle],%[output]"
3558 : [output] "=f" (result)
3559 : [angle] "f" (angle));
3563 Note that the symbolic operand names have no relation whatsoever to
3564 other C identifiers. You may use any name you like, even those of
3565 existing C symbols, but you must ensure that no two operands within the same
3566 assembler construct use the same symbolic name.
3568 Output operand expressions must be lvalues; the compiler can check this.
3569 The input operands need not be lvalues. The compiler cannot check
3570 whether the operands have data types that are reasonable for the
3571 instruction being executed. It does not parse the assembler instruction
3572 template and does not know what it means or even whether it is valid
3573 assembler input. The extended @code{asm} feature is most often used for
3574 machine instructions the compiler itself does not know exist. If
3575 the output expression cannot be directly addressed (for example, it is a
3576 bit-field), your constraint must allow a register. In that case, GCC
3577 will use the register as the output of the @code{asm}, and then store
3578 that register into the output.
3580 The ordinary output operands must be write-only; GCC will assume that
3581 the values in these operands before the instruction are dead and need
3582 not be generated. Extended asm supports input-output or read-write
3583 operands. Use the constraint character @samp{+} to indicate such an
3584 operand and list it with the output operands. You should only use
3585 read-write operands when the constraints for the operand (or the
3586 operand in which only some of the bits are to be changed) allow a
3589 You may, as an alternative, logically split its function into two
3590 separate operands, one input operand and one write-only output
3591 operand. The connection between them is expressed by constraints
3592 which say they need to be in the same location when the instruction
3593 executes. You can use the same C expression for both operands, or
3594 different expressions. For example, here we write the (fictitious)
3595 @samp{combine} instruction with @code{bar} as its read-only source
3596 operand and @code{foo} as its read-write destination:
3599 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3603 The constraint @samp{"0"} for operand 1 says that it must occupy the
3604 same location as operand 0. A number in constraint is allowed only in
3605 an input operand and it must refer to an output operand.
3607 Only a number in the constraint can guarantee that one operand will be in
3608 the same place as another. The mere fact that @code{foo} is the value
3609 of both operands is not enough to guarantee that they will be in the
3610 same place in the generated assembler code. The following would not
3614 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3617 Various optimizations or reloading could cause operands 0 and 1 to be in
3618 different registers; GCC knows no reason not to do so. For example, the
3619 compiler might find a copy of the value of @code{foo} in one register and
3620 use it for operand 1, but generate the output operand 0 in a different
3621 register (copying it afterward to @code{foo}'s own address). Of course,
3622 since the register for operand 1 is not even mentioned in the assembler
3623 code, the result will not work, but GCC can't tell that.
3625 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3626 the operand number for a matching constraint. For example:
3629 asm ("cmoveq %1,%2,%[result]"
3630 : [result] "=r"(result)
3631 : "r" (test), "r"(new), "[result]"(old));
3634 Sometimes you need to make an @code{asm} operand be a specific register,
3635 but there's no matching constraint letter for that register @emph{by
3636 itself}. To force the operand into that register, use a local variable
3637 for the operand and specify the register in the variable declaration.
3638 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
3639 register constraint letter that matches the register:
3642 register int *p1 asm ("r0") = @dots{};
3643 register int *p2 asm ("r1") = @dots{};
3644 register int *result asm ("r0");
3645 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
3648 @anchor{Example of asm with clobbered asm reg}
3649 In the above example, beware that a register that is call-clobbered by
3650 the target ABI will be overwritten by any function call in the
3651 assignment, including library calls for arithmetic operators.
3652 Assuming it is a call-clobbered register, this may happen to @code{r0}
3653 above by the assignment to @code{p2}. If you have to use such a
3654 register, use temporary variables for expressions between the register
3659 register int *p1 asm ("r0") = @dots{};
3660 register int *p2 asm ("r1") = t1;
3661 register int *result asm ("r0");
3662 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
3665 Some instructions clobber specific hard registers. To describe this,
3666 write a third colon after the input operands, followed by the names of
3667 the clobbered hard registers (given as strings). Here is a realistic
3668 example for the VAX:
3671 asm volatile ("movc3 %0,%1,%2"
3672 : /* @r{no outputs} */
3673 : "g" (from), "g" (to), "g" (count)
3674 : "r0", "r1", "r2", "r3", "r4", "r5");
3677 You may not write a clobber description in a way that overlaps with an
3678 input or output operand. For example, you may not have an operand
3679 describing a register class with one member if you mention that register
3680 in the clobber list. Variables declared to live in specific registers
3681 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3682 have no part mentioned in the clobber description.
3683 There is no way for you to specify that an input
3684 operand is modified without also specifying it as an output
3685 operand. Note that if all the output operands you specify are for this
3686 purpose (and hence unused), you will then also need to specify
3687 @code{volatile} for the @code{asm} construct, as described below, to
3688 prevent GCC from deleting the @code{asm} statement as unused.
3690 If you refer to a particular hardware register from the assembler code,
3691 you will probably have to list the register after the third colon to
3692 tell the compiler the register's value is modified. In some assemblers,
3693 the register names begin with @samp{%}; to produce one @samp{%} in the
3694 assembler code, you must write @samp{%%} in the input.
3696 If your assembler instruction can alter the condition code register, add
3697 @samp{cc} to the list of clobbered registers. GCC on some machines
3698 represents the condition codes as a specific hardware register;
3699 @samp{cc} serves to name this register. On other machines, the
3700 condition code is handled differently, and specifying @samp{cc} has no
3701 effect. But it is valid no matter what the machine.
3703 If your assembler instructions access memory in an unpredictable
3704 fashion, add @samp{memory} to the list of clobbered registers. This
3705 will cause GCC to not keep memory values cached in registers across the
3706 assembler instruction and not optimize stores or loads to that memory.
3707 You will also want to add the @code{volatile} keyword if the memory
3708 affected is not listed in the inputs or outputs of the @code{asm}, as
3709 the @samp{memory} clobber does not count as a side-effect of the
3710 @code{asm}. If you know how large the accessed memory is, you can add
3711 it as input or output but if this is not known, you should add
3712 @samp{memory}. As an example, if you access ten bytes of a string, you
3713 can use a memory input like:
3716 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
3719 Note that in the following example the memory input is necessary,
3720 otherwise GCC might optimize the store to @code{x} away:
3727 asm ("magic stuff accessing an 'int' pointed to by '%1'"
3728 "=&d" (r) : "a" (y), "m" (*y));
3733 You can put multiple assembler instructions together in a single
3734 @code{asm} template, separated by the characters normally used in assembly
3735 code for the system. A combination that works in most places is a newline
3736 to break the line, plus a tab character to move to the instruction field
3737 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
3738 assembler allows semicolons as a line-breaking character. Note that some
3739 assembler dialects use semicolons to start a comment.
3740 The input operands are guaranteed not to use any of the clobbered
3741 registers, and neither will the output operands' addresses, so you can
3742 read and write the clobbered registers as many times as you like. Here
3743 is an example of multiple instructions in a template; it assumes the
3744 subroutine @code{_foo} accepts arguments in registers 9 and 10:
3747 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
3749 : "g" (from), "g" (to)
3753 Unless an output operand has the @samp{&} constraint modifier, GCC
3754 may allocate it in the same register as an unrelated input operand, on
3755 the assumption the inputs are consumed before the outputs are produced.
3756 This assumption may be false if the assembler code actually consists of
3757 more than one instruction. In such a case, use @samp{&} for each output
3758 operand that may not overlap an input. @xref{Modifiers}.
3760 If you want to test the condition code produced by an assembler
3761 instruction, you must include a branch and a label in the @code{asm}
3762 construct, as follows:
3765 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
3771 This assumes your assembler supports local labels, as the GNU assembler
3772 and most Unix assemblers do.
3774 Speaking of labels, jumps from one @code{asm} to another are not
3775 supported. The compiler's optimizers do not know about these jumps, and
3776 therefore they cannot take account of them when deciding how to
3779 @cindex macros containing @code{asm}
3780 Usually the most convenient way to use these @code{asm} instructions is to
3781 encapsulate them in macros that look like functions. For example,
3785 (@{ double __value, __arg = (x); \
3786 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
3791 Here the variable @code{__arg} is used to make sure that the instruction
3792 operates on a proper @code{double} value, and to accept only those
3793 arguments @code{x} which can convert automatically to a @code{double}.
3795 Another way to make sure the instruction operates on the correct data
3796 type is to use a cast in the @code{asm}. This is different from using a
3797 variable @code{__arg} in that it converts more different types. For
3798 example, if the desired type were @code{int}, casting the argument to
3799 @code{int} would accept a pointer with no complaint, while assigning the
3800 argument to an @code{int} variable named @code{__arg} would warn about
3801 using a pointer unless the caller explicitly casts it.
3803 If an @code{asm} has output operands, GCC assumes for optimization
3804 purposes the instruction has no side effects except to change the output
3805 operands. This does not mean instructions with a side effect cannot be
3806 used, but you must be careful, because the compiler may eliminate them
3807 if the output operands aren't used, or move them out of loops, or
3808 replace two with one if they constitute a common subexpression. Also,
3809 if your instruction does have a side effect on a variable that otherwise
3810 appears not to change, the old value of the variable may be reused later
3811 if it happens to be found in a register.
3813 You can prevent an @code{asm} instruction from being deleted
3814 by writing the keyword @code{volatile} after
3815 the @code{asm}. For example:
3818 #define get_and_set_priority(new) \
3820 asm volatile ("get_and_set_priority %0, %1" \
3821 : "=g" (__old) : "g" (new)); \
3826 The @code{volatile} keyword indicates that the instruction has
3827 important side-effects. GCC will not delete a volatile @code{asm} if
3828 it is reachable. (The instruction can still be deleted if GCC can
3829 prove that control-flow will never reach the location of the
3830 instruction.) Note that even a volatile @code{asm} instruction
3831 can be moved relative to other code, including across jump
3832 instructions. For example, on many targets there is a system
3833 register which can be set to control the rounding mode of
3834 floating point operations. You might try
3835 setting it with a volatile @code{asm}, like this PowerPC example:
3838 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
3843 This will not work reliably, as the compiler may move the addition back
3844 before the volatile @code{asm}. To make it work you need to add an
3845 artificial dependency to the @code{asm} referencing a variable in the code
3846 you don't want moved, for example:
3849 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
3853 Similarly, you can't expect a
3854 sequence of volatile @code{asm} instructions to remain perfectly
3855 consecutive. If you want consecutive output, use a single @code{asm}.
3856 Also, GCC will perform some optimizations across a volatile @code{asm}
3857 instruction; GCC does not ``forget everything'' when it encounters
3858 a volatile @code{asm} instruction the way some other compilers do.
3860 An @code{asm} instruction without any output operands will be treated
3861 identically to a volatile @code{asm} instruction.
3863 It is a natural idea to look for a way to give access to the condition
3864 code left by the assembler instruction. However, when we attempted to
3865 implement this, we found no way to make it work reliably. The problem
3866 is that output operands might need reloading, which would result in
3867 additional following ``store'' instructions. On most machines, these
3868 instructions would alter the condition code before there was time to
3869 test it. This problem doesn't arise for ordinary ``test'' and
3870 ``compare'' instructions because they don't have any output operands.
3872 For reasons similar to those described above, it is not possible to give
3873 an assembler instruction access to the condition code left by previous
3876 If you are writing a header file that should be includable in ISO C
3877 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
3880 @subsection Size of an @code{asm}
3882 Some targets require that GCC track the size of each instruction used in
3883 order to generate correct code. Because the final length of an
3884 @code{asm} is only known by the assembler, GCC must make an estimate as
3885 to how big it will be. The estimate is formed by counting the number of
3886 statements in the pattern of the @code{asm} and multiplying that by the
3887 length of the longest instruction on that processor. Statements in the
3888 @code{asm} are identified by newline characters and whatever statement
3889 separator characters are supported by the assembler; on most processors
3890 this is the `@code{;}' character.
3892 Normally, GCC's estimate is perfectly adequate to ensure that correct
3893 code is generated, but it is possible to confuse the compiler if you use
3894 pseudo instructions or assembler macros that expand into multiple real
3895 instructions or if you use assembler directives that expand to more
3896 space in the object file than would be needed for a single instruction.
3897 If this happens then the assembler will produce a diagnostic saying that
3898 a label is unreachable.
3900 @subsection i386 floating point asm operands
3902 There are several rules on the usage of stack-like regs in
3903 asm_operands insns. These rules apply only to the operands that are
3908 Given a set of input regs that die in an asm_operands, it is
3909 necessary to know which are implicitly popped by the asm, and
3910 which must be explicitly popped by gcc.
3912 An input reg that is implicitly popped by the asm must be
3913 explicitly clobbered, unless it is constrained to match an
3917 For any input reg that is implicitly popped by an asm, it is
3918 necessary to know how to adjust the stack to compensate for the pop.
3919 If any non-popped input is closer to the top of the reg-stack than
3920 the implicitly popped reg, it would not be possible to know what the
3921 stack looked like---it's not clear how the rest of the stack ``slides
3924 All implicitly popped input regs must be closer to the top of
3925 the reg-stack than any input that is not implicitly popped.
3927 It is possible that if an input dies in an insn, reload might
3928 use the input reg for an output reload. Consider this example:
3931 asm ("foo" : "=t" (a) : "f" (b));
3934 This asm says that input B is not popped by the asm, and that
3935 the asm pushes a result onto the reg-stack, i.e., the stack is one
3936 deeper after the asm than it was before. But, it is possible that
3937 reload will think that it can use the same reg for both the input and
3938 the output, if input B dies in this insn.
3940 If any input operand uses the @code{f} constraint, all output reg
3941 constraints must use the @code{&} earlyclobber.
3943 The asm above would be written as
3946 asm ("foo" : "=&t" (a) : "f" (b));
3950 Some operands need to be in particular places on the stack. All
3951 output operands fall in this category---there is no other way to
3952 know which regs the outputs appear in unless the user indicates
3953 this in the constraints.
3955 Output operands must specifically indicate which reg an output
3956 appears in after an asm. @code{=f} is not allowed: the operand
3957 constraints must select a class with a single reg.
3960 Output operands may not be ``inserted'' between existing stack regs.
3961 Since no 387 opcode uses a read/write operand, all output operands
3962 are dead before the asm_operands, and are pushed by the asm_operands.
3963 It makes no sense to push anywhere but the top of the reg-stack.
3965 Output operands must start at the top of the reg-stack: output
3966 operands may not ``skip'' a reg.
3969 Some asm statements may need extra stack space for internal
3970 calculations. This can be guaranteed by clobbering stack registers
3971 unrelated to the inputs and outputs.
3975 Here are a couple of reasonable asms to want to write. This asm
3976 takes one input, which is internally popped, and produces two outputs.
3979 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
3982 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
3983 and replaces them with one output. The user must code the @code{st(1)}
3984 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
3987 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
3993 @section Controlling Names Used in Assembler Code
3994 @cindex assembler names for identifiers
3995 @cindex names used in assembler code
3996 @cindex identifiers, names in assembler code
3998 You can specify the name to be used in the assembler code for a C
3999 function or variable by writing the @code{asm} (or @code{__asm__})
4000 keyword after the declarator as follows:
4003 int foo asm ("myfoo") = 2;
4007 This specifies that the name to be used for the variable @code{foo} in
4008 the assembler code should be @samp{myfoo} rather than the usual
4011 On systems where an underscore is normally prepended to the name of a C
4012 function or variable, this feature allows you to define names for the
4013 linker that do not start with an underscore.
4015 It does not make sense to use this feature with a non-static local
4016 variable since such variables do not have assembler names. If you are
4017 trying to put the variable in a particular register, see @ref{Explicit
4018 Reg Vars}. GCC presently accepts such code with a warning, but will
4019 probably be changed to issue an error, rather than a warning, in the
4022 You cannot use @code{asm} in this way in a function @emph{definition}; but
4023 you can get the same effect by writing a declaration for the function
4024 before its definition and putting @code{asm} there, like this:
4027 extern func () asm ("FUNC");
4034 It is up to you to make sure that the assembler names you choose do not
4035 conflict with any other assembler symbols. Also, you must not use a
4036 register name; that would produce completely invalid assembler code. GCC
4037 does not as yet have the ability to store static variables in registers.
4038 Perhaps that will be added.
4040 @node Explicit Reg Vars
4041 @section Variables in Specified Registers
4042 @cindex explicit register variables
4043 @cindex variables in specified registers
4044 @cindex specified registers
4045 @cindex registers, global allocation
4047 GNU C allows you to put a few global variables into specified hardware
4048 registers. You can also specify the register in which an ordinary
4049 register variable should be allocated.
4053 Global register variables reserve registers throughout the program.
4054 This may be useful in programs such as programming language
4055 interpreters which have a couple of global variables that are accessed
4059 Local register variables in specific registers do not reserve the
4060 registers, except at the point where they are used as input or output
4061 operands in an @code{asm} statement and the @code{asm} statement itself is
4062 not deleted. The compiler's data flow analysis is capable of determining
4063 where the specified registers contain live values, and where they are
4064 available for other uses. Stores into local register variables may be deleted
4065 when they appear to be dead according to dataflow analysis. References
4066 to local register variables may be deleted or moved or simplified.
4068 These local variables are sometimes convenient for use with the extended
4069 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4070 output of the assembler instruction directly into a particular register.
4071 (This will work provided the register you specify fits the constraints
4072 specified for that operand in the @code{asm}.)
4080 @node Global Reg Vars
4081 @subsection Defining Global Register Variables
4082 @cindex global register variables
4083 @cindex registers, global variables in
4085 You can define a global register variable in GNU C like this:
4088 register int *foo asm ("a5");
4092 Here @code{a5} is the name of the register which should be used. Choose a
4093 register which is normally saved and restored by function calls on your
4094 machine, so that library routines will not clobber it.
4096 Naturally the register name is cpu-dependent, so you would need to
4097 conditionalize your program according to cpu type. The register
4098 @code{a5} would be a good choice on a 68000 for a variable of pointer
4099 type. On machines with register windows, be sure to choose a ``global''
4100 register that is not affected magically by the function call mechanism.
4102 In addition, operating systems on one type of cpu may differ in how they
4103 name the registers; then you would need additional conditionals. For
4104 example, some 68000 operating systems call this register @code{%a5}.
4106 Eventually there may be a way of asking the compiler to choose a register
4107 automatically, but first we need to figure out how it should choose and
4108 how to enable you to guide the choice. No solution is evident.
4110 Defining a global register variable in a certain register reserves that
4111 register entirely for this use, at least within the current compilation.
4112 The register will not be allocated for any other purpose in the functions
4113 in the current compilation. The register will not be saved and restored by
4114 these functions. Stores into this register are never deleted even if they
4115 would appear to be dead, but references may be deleted or moved or
4118 It is not safe to access the global register variables from signal
4119 handlers, or from more than one thread of control, because the system
4120 library routines may temporarily use the register for other things (unless
4121 you recompile them specially for the task at hand).
4123 @cindex @code{qsort}, and global register variables
4124 It is not safe for one function that uses a global register variable to
4125 call another such function @code{foo} by way of a third function
4126 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4127 different source file in which the variable wasn't declared). This is
4128 because @code{lose} might save the register and put some other value there.
4129 For example, you can't expect a global register variable to be available in
4130 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4131 might have put something else in that register. (If you are prepared to
4132 recompile @code{qsort} with the same global register variable, you can
4133 solve this problem.)
4135 If you want to recompile @code{qsort} or other source files which do not
4136 actually use your global register variable, so that they will not use that
4137 register for any other purpose, then it suffices to specify the compiler
4138 option @option{-ffixed-@var{reg}}. You need not actually add a global
4139 register declaration to their source code.
4141 A function which can alter the value of a global register variable cannot
4142 safely be called from a function compiled without this variable, because it
4143 could clobber the value the caller expects to find there on return.
4144 Therefore, the function which is the entry point into the part of the
4145 program that uses the global register variable must explicitly save and
4146 restore the value which belongs to its caller.
4148 @cindex register variable after @code{longjmp}
4149 @cindex global register after @code{longjmp}
4150 @cindex value after @code{longjmp}
4153 On most machines, @code{longjmp} will restore to each global register
4154 variable the value it had at the time of the @code{setjmp}. On some
4155 machines, however, @code{longjmp} will not change the value of global
4156 register variables. To be portable, the function that called @code{setjmp}
4157 should make other arrangements to save the values of the global register
4158 variables, and to restore them in a @code{longjmp}. This way, the same
4159 thing will happen regardless of what @code{longjmp} does.
4161 All global register variable declarations must precede all function
4162 definitions. If such a declaration could appear after function
4163 definitions, the declaration would be too late to prevent the register from
4164 being used for other purposes in the preceding functions.
4166 Global register variables may not have initial values, because an
4167 executable file has no means to supply initial contents for a register.
4169 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4170 registers, but certain library functions, such as @code{getwd}, as well
4171 as the subroutines for division and remainder, modify g3 and g4. g1 and
4172 g2 are local temporaries.
4174 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4175 Of course, it will not do to use more than a few of those.
4177 @node Local Reg Vars
4178 @subsection Specifying Registers for Local Variables
4179 @cindex local variables, specifying registers
4180 @cindex specifying registers for local variables
4181 @cindex registers for local variables
4183 You can define a local register variable with a specified register
4187 register int *foo asm ("a5");
4191 Here @code{a5} is the name of the register which should be used. Note
4192 that this is the same syntax used for defining global register
4193 variables, but for a local variable it would appear within a function.
4195 Naturally the register name is cpu-dependent, but this is not a
4196 problem, since specific registers are most often useful with explicit
4197 assembler instructions (@pxref{Extended Asm}). Both of these things
4198 generally require that you conditionalize your program according to
4201 In addition, operating systems on one type of cpu may differ in how they
4202 name the registers; then you would need additional conditionals. For
4203 example, some 68000 operating systems call this register @code{%a5}.
4205 Defining such a register variable does not reserve the register; it
4206 remains available for other uses in places where flow control determines
4207 the variable's value is not live.
4209 This option does not guarantee that GCC will generate code that has
4210 this variable in the register you specify at all times. You may not
4211 code an explicit reference to this register in the @emph{assembler
4212 instruction template} part of an @code{asm} statement and assume it will
4213 always refer to this variable. However, using the variable as an
4214 @code{asm} @emph{operand} guarantees that the specified register is used
4217 Stores into local register variables may be deleted when they appear to be dead
4218 according to dataflow analysis. References to local register variables may
4219 be deleted or moved or simplified.
4221 As for global register variables, it's recommended that you choose a
4222 register which is normally saved and restored by function calls on
4223 your machine, so that library routines will not clobber it. A common
4224 pitfall is to initialize multiple call-clobbered registers with
4225 arbitrary expressions, where a function call or library call for an
4226 arithmetic operator will overwrite a register value from a previous
4227 assignment, for example @code{r0} below:
4229 register int *p1 asm ("r0") = @dots{};
4230 register int *p2 asm ("r1") = @dots{};
4232 In those cases, a solution is to use a temporary variable for
4233 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4235 @node Alternate Keywords
4236 @section Alternate Keywords
4237 @cindex alternate keywords
4238 @cindex keywords, alternate
4240 @option{-ansi} and the various @option{-std} options disable certain
4241 keywords. This causes trouble when you want to use GNU C extensions, or
4242 a general-purpose header file that should be usable by all programs,
4243 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4244 @code{inline} are not available in programs compiled with
4245 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4246 program compiled with @option{-std=c99}). The ISO C99 keyword
4247 @code{restrict} is only available when @option{-std=gnu99} (which will
4248 eventually be the default) or @option{-std=c99} (or the equivalent
4249 @option{-std=iso9899:1999}) is used.
4251 The way to solve these problems is to put @samp{__} at the beginning and
4252 end of each problematical keyword. For example, use @code{__asm__}
4253 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4255 Other C compilers won't accept these alternative keywords; if you want to
4256 compile with another compiler, you can define the alternate keywords as
4257 macros to replace them with the customary keywords. It looks like this:
4265 @findex __extension__
4267 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4269 prevent such warnings within one expression by writing
4270 @code{__extension__} before the expression. @code{__extension__} has no
4271 effect aside from this.
4273 @node Incomplete Enums
4274 @section Incomplete @code{enum} Types
4276 You can define an @code{enum} tag without specifying its possible values.
4277 This results in an incomplete type, much like what you get if you write
4278 @code{struct foo} without describing the elements. A later declaration
4279 which does specify the possible values completes the type.
4281 You can't allocate variables or storage using the type while it is
4282 incomplete. However, you can work with pointers to that type.
4284 This extension may not be very useful, but it makes the handling of
4285 @code{enum} more consistent with the way @code{struct} and @code{union}
4288 This extension is not supported by GNU C++.
4290 @node Function Names
4291 @section Function Names as Strings
4292 @cindex @code{__func__} identifier
4293 @cindex @code{__FUNCTION__} identifier
4294 @cindex @code{__PRETTY_FUNCTION__} identifier
4296 GCC provides three magic variables which hold the name of the current
4297 function, as a string. The first of these is @code{__func__}, which
4298 is part of the C99 standard:
4301 The identifier @code{__func__} is implicitly declared by the translator
4302 as if, immediately following the opening brace of each function
4303 definition, the declaration
4306 static const char __func__[] = "function-name";
4309 appeared, where function-name is the name of the lexically-enclosing
4310 function. This name is the unadorned name of the function.
4313 @code{__FUNCTION__} is another name for @code{__func__}. Older
4314 versions of GCC recognize only this name. However, it is not
4315 standardized. For maximum portability, we recommend you use
4316 @code{__func__}, but provide a fallback definition with the
4320 #if __STDC_VERSION__ < 199901L
4322 # define __func__ __FUNCTION__
4324 # define __func__ "<unknown>"
4329 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4330 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4331 the type signature of the function as well as its bare name. For
4332 example, this program:
4336 extern int printf (char *, ...);
4343 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4344 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4362 __PRETTY_FUNCTION__ = void a::sub(int)
4365 These identifiers are not preprocessor macros. In GCC 3.3 and
4366 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4367 were treated as string literals; they could be used to initialize
4368 @code{char} arrays, and they could be concatenated with other string
4369 literals. GCC 3.4 and later treat them as variables, like
4370 @code{__func__}. In C++, @code{__FUNCTION__} and
4371 @code{__PRETTY_FUNCTION__} have always been variables.
4373 @node Return Address
4374 @section Getting the Return or Frame Address of a Function
4376 These functions may be used to get information about the callers of a
4379 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4380 This function returns the return address of the current function, or of
4381 one of its callers. The @var{level} argument is number of frames to
4382 scan up the call stack. A value of @code{0} yields the return address
4383 of the current function, a value of @code{1} yields the return address
4384 of the caller of the current function, and so forth. When inlining
4385 the expected behavior is that the function will return the address of
4386 the function that will be returned to. To work around this behavior use
4387 the @code{noinline} function attribute.
4389 The @var{level} argument must be a constant integer.
4391 On some machines it may be impossible to determine the return address of
4392 any function other than the current one; in such cases, or when the top
4393 of the stack has been reached, this function will return @code{0} or a
4394 random value. In addition, @code{__builtin_frame_address} may be used
4395 to determine if the top of the stack has been reached.
4397 This function should only be used with a nonzero argument for debugging
4401 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4402 This function is similar to @code{__builtin_return_address}, but it
4403 returns the address of the function frame rather than the return address
4404 of the function. Calling @code{__builtin_frame_address} with a value of
4405 @code{0} yields the frame address of the current function, a value of
4406 @code{1} yields the frame address of the caller of the current function,
4409 The frame is the area on the stack which holds local variables and saved
4410 registers. The frame address is normally the address of the first word
4411 pushed on to the stack by the function. However, the exact definition
4412 depends upon the processor and the calling convention. If the processor
4413 has a dedicated frame pointer register, and the function has a frame,
4414 then @code{__builtin_frame_address} will return the value of the frame
4417 On some machines it may be impossible to determine the frame address of
4418 any function other than the current one; in such cases, or when the top
4419 of the stack has been reached, this function will return @code{0} if
4420 the first frame pointer is properly initialized by the startup code.
4422 This function should only be used with a nonzero argument for debugging
4426 @node Vector Extensions
4427 @section Using vector instructions through built-in functions
4429 On some targets, the instruction set contains SIMD vector instructions that
4430 operate on multiple values contained in one large register at the same time.
4431 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4434 The first step in using these extensions is to provide the necessary data
4435 types. This should be done using an appropriate @code{typedef}:
4438 typedef int v4si __attribute__ ((vector_size (16)));
4441 The @code{int} type specifies the base type, while the attribute specifies
4442 the vector size for the variable, measured in bytes. For example, the
4443 declaration above causes the compiler to set the mode for the @code{v4si}
4444 type to be 16 bytes wide and divided into @code{int} sized units. For
4445 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4446 corresponding mode of @code{foo} will be @acronym{V4SI}.
4448 The @code{vector_size} attribute is only applicable to integral and
4449 float scalars, although arrays, pointers, and function return values
4450 are allowed in conjunction with this construct.
4452 All the basic integer types can be used as base types, both as signed
4453 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4454 @code{long long}. In addition, @code{float} and @code{double} can be
4455 used to build floating-point vector types.
4457 Specifying a combination that is not valid for the current architecture
4458 will cause GCC to synthesize the instructions using a narrower mode.
4459 For example, if you specify a variable of type @code{V4SI} and your
4460 architecture does not allow for this specific SIMD type, GCC will
4461 produce code that uses 4 @code{SIs}.
4463 The types defined in this manner can be used with a subset of normal C
4464 operations. Currently, GCC will allow using the following operators
4465 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4467 The operations behave like C++ @code{valarrays}. Addition is defined as
4468 the addition of the corresponding elements of the operands. For
4469 example, in the code below, each of the 4 elements in @var{a} will be
4470 added to the corresponding 4 elements in @var{b} and the resulting
4471 vector will be stored in @var{c}.
4474 typedef int v4si __attribute__ ((vector_size (16)));
4481 Subtraction, multiplication, division, and the logical operations
4482 operate in a similar manner. Likewise, the result of using the unary
4483 minus or complement operators on a vector type is a vector whose
4484 elements are the negative or complemented values of the corresponding
4485 elements in the operand.
4487 You can declare variables and use them in function calls and returns, as
4488 well as in assignments and some casts. You can specify a vector type as
4489 a return type for a function. Vector types can also be used as function
4490 arguments. It is possible to cast from one vector type to another,
4491 provided they are of the same size (in fact, you can also cast vectors
4492 to and from other datatypes of the same size).
4494 You cannot operate between vectors of different lengths or different
4495 signedness without a cast.
4497 A port that supports hardware vector operations, usually provides a set
4498 of built-in functions that can be used to operate on vectors. For
4499 example, a function to add two vectors and multiply the result by a
4500 third could look like this:
4503 v4si f (v4si a, v4si b, v4si c)
4505 v4si tmp = __builtin_addv4si (a, b);
4506 return __builtin_mulv4si (tmp, c);
4513 @findex __builtin_offsetof
4515 GCC implements for both C and C++ a syntactic extension to implement
4516 the @code{offsetof} macro.
4520 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4522 offsetof_member_designator:
4524 | offsetof_member_designator "." @code{identifier}
4525 | offsetof_member_designator "[" @code{expr} "]"
4528 This extension is sufficient such that
4531 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4534 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4535 may be dependent. In either case, @var{member} may consist of a single
4536 identifier, or a sequence of member accesses and array references.
4538 @node Other Builtins
4539 @section Other built-in functions provided by GCC
4540 @cindex built-in functions
4541 @findex __builtin_isgreater
4542 @findex __builtin_isgreaterequal
4543 @findex __builtin_isless
4544 @findex __builtin_islessequal
4545 @findex __builtin_islessgreater
4546 @findex __builtin_isunordered
4547 @findex __builtin_powi
4548 @findex __builtin_powif
4549 @findex __builtin_powil
4704 @findex fprintf_unlocked
4706 @findex fputs_unlocked
4816 @findex printf_unlocked
4845 @findex significandf
4846 @findex significandl
4913 GCC provides a large number of built-in functions other than the ones
4914 mentioned above. Some of these are for internal use in the processing
4915 of exceptions or variable-length argument lists and will not be
4916 documented here because they may change from time to time; we do not
4917 recommend general use of these functions.
4919 The remaining functions are provided for optimization purposes.
4921 @opindex fno-builtin
4922 GCC includes built-in versions of many of the functions in the standard
4923 C library. The versions prefixed with @code{__builtin_} will always be
4924 treated as having the same meaning as the C library function even if you
4925 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
4926 Many of these functions are only optimized in certain cases; if they are
4927 not optimized in a particular case, a call to the library function will
4932 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
4933 @option{-std=c99}), the functions
4934 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
4935 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
4936 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
4937 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
4938 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
4939 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
4940 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
4941 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
4942 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
4943 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
4944 @code{significandf}, @code{significandl}, @code{significand},
4945 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
4946 @code{strdup}, @code{strfmon}, @code{toascii}, @code{y0f}, @code{y0l},
4947 @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and
4949 may be handled as built-in functions.
4950 All these functions have corresponding versions
4951 prefixed with @code{__builtin_}, which may be used even in strict C89
4954 The ISO C99 functions
4955 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
4956 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
4957 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
4958 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
4959 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
4960 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
4961 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
4962 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
4963 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
4964 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
4965 @code{cimagl}, @code{cimag}, @code{conjf}, @code{conjl}, @code{conj},
4966 @code{copysignf}, @code{copysignl}, @code{copysign}, @code{cpowf},
4967 @code{cpowl}, @code{cpow}, @code{cprojf}, @code{cprojl}, @code{cproj},
4968 @code{crealf}, @code{creall}, @code{creal}, @code{csinf}, @code{csinhf},
4969 @code{csinhl}, @code{csinh}, @code{csinl}, @code{csin}, @code{csqrtf},
4970 @code{csqrtl}, @code{csqrt}, @code{ctanf}, @code{ctanhf}, @code{ctanhl},
4971 @code{ctanh}, @code{ctanl}, @code{ctan}, @code{erfcf}, @code{erfcl},
4972 @code{erfc}, @code{erff}, @code{erfl}, @code{erf}, @code{exp2f},
4973 @code{exp2l}, @code{exp2}, @code{expm1f}, @code{expm1l}, @code{expm1},
4974 @code{fdimf}, @code{fdiml}, @code{fdim}, @code{fmaf}, @code{fmal},
4975 @code{fmaxf}, @code{fmaxl}, @code{fmax}, @code{fma}, @code{fminf},
4976 @code{fminl}, @code{fmin}, @code{hypotf}, @code{hypotl}, @code{hypot},
4977 @code{ilogbf}, @code{ilogbl}, @code{ilogb}, @code{imaxabs},
4978 @code{isblank}, @code{iswblank}, @code{lgammaf}, @code{lgammal},
4979 @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
4980 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
4981 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
4982 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
4983 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
4984 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
4985 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
4986 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
4987 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
4988 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
4989 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
4990 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
4991 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
4992 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
4993 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
4994 are handled as built-in functions
4995 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
4997 There are also built-in versions of the ISO C99 functions
4998 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
4999 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5000 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5001 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5002 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5003 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5004 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5005 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5006 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5007 that are recognized in any mode since ISO C90 reserves these names for
5008 the purpose to which ISO C99 puts them. All these functions have
5009 corresponding versions prefixed with @code{__builtin_}.
5011 The ISO C94 functions
5012 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5013 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5014 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5016 are handled as built-in functions
5017 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5019 The ISO C90 functions
5020 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5021 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5022 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5023 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5024 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5025 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5026 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5027 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5028 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5029 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5030 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5031 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5032 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5033 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5034 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5035 @code{vprintf} and @code{vsprintf}
5036 are all recognized as built-in functions unless
5037 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5038 is specified for an individual function). All of these functions have
5039 corresponding versions prefixed with @code{__builtin_}.
5041 GCC provides built-in versions of the ISO C99 floating point comparison
5042 macros that avoid raising exceptions for unordered operands. They have
5043 the same names as the standard macros ( @code{isgreater},
5044 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5045 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5046 prefixed. We intend for a library implementor to be able to simply
5047 @code{#define} each standard macro to its built-in equivalent.
5049 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5051 You can use the built-in function @code{__builtin_types_compatible_p} to
5052 determine whether two types are the same.
5054 This built-in function returns 1 if the unqualified versions of the
5055 types @var{type1} and @var{type2} (which are types, not expressions) are
5056 compatible, 0 otherwise. The result of this built-in function can be
5057 used in integer constant expressions.
5059 This built-in function ignores top level qualifiers (e.g., @code{const},
5060 @code{volatile}). For example, @code{int} is equivalent to @code{const
5063 The type @code{int[]} and @code{int[5]} are compatible. On the other
5064 hand, @code{int} and @code{char *} are not compatible, even if the size
5065 of their types, on the particular architecture are the same. Also, the
5066 amount of pointer indirection is taken into account when determining
5067 similarity. Consequently, @code{short *} is not similar to
5068 @code{short **}. Furthermore, two types that are typedefed are
5069 considered compatible if their underlying types are compatible.
5071 An @code{enum} type is not considered to be compatible with another
5072 @code{enum} type even if both are compatible with the same integer
5073 type; this is what the C standard specifies.
5074 For example, @code{enum @{foo, bar@}} is not similar to
5075 @code{enum @{hot, dog@}}.
5077 You would typically use this function in code whose execution varies
5078 depending on the arguments' types. For example:
5084 if (__builtin_types_compatible_p (typeof (x), long double)) \
5085 tmp = foo_long_double (tmp); \
5086 else if (__builtin_types_compatible_p (typeof (x), double)) \
5087 tmp = foo_double (tmp); \
5088 else if (__builtin_types_compatible_p (typeof (x), float)) \
5089 tmp = foo_float (tmp); \
5096 @emph{Note:} This construct is only available for C@.
5100 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5102 You can use the built-in function @code{__builtin_choose_expr} to
5103 evaluate code depending on the value of a constant expression. This
5104 built-in function returns @var{exp1} if @var{const_exp}, which is a
5105 constant expression that must be able to be determined at compile time,
5106 is nonzero. Otherwise it returns 0.
5108 This built-in function is analogous to the @samp{? :} operator in C,
5109 except that the expression returned has its type unaltered by promotion
5110 rules. Also, the built-in function does not evaluate the expression
5111 that was not chosen. For example, if @var{const_exp} evaluates to true,
5112 @var{exp2} is not evaluated even if it has side-effects.
5114 This built-in function can return an lvalue if the chosen argument is an
5117 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5118 type. Similarly, if @var{exp2} is returned, its return type is the same
5125 __builtin_choose_expr ( \
5126 __builtin_types_compatible_p (typeof (x), double), \
5128 __builtin_choose_expr ( \
5129 __builtin_types_compatible_p (typeof (x), float), \
5131 /* @r{The void expression results in a compile-time error} \
5132 @r{when assigning the result to something.} */ \
5136 @emph{Note:} This construct is only available for C@. Furthermore, the
5137 unused expression (@var{exp1} or @var{exp2} depending on the value of
5138 @var{const_exp}) may still generate syntax errors. This may change in
5143 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5144 You can use the built-in function @code{__builtin_constant_p} to
5145 determine if a value is known to be constant at compile-time and hence
5146 that GCC can perform constant-folding on expressions involving that
5147 value. The argument of the function is the value to test. The function
5148 returns the integer 1 if the argument is known to be a compile-time
5149 constant and 0 if it is not known to be a compile-time constant. A
5150 return of 0 does not indicate that the value is @emph{not} a constant,
5151 but merely that GCC cannot prove it is a constant with the specified
5152 value of the @option{-O} option.
5154 You would typically use this function in an embedded application where
5155 memory was a critical resource. If you have some complex calculation,
5156 you may want it to be folded if it involves constants, but need to call
5157 a function if it does not. For example:
5160 #define Scale_Value(X) \
5161 (__builtin_constant_p (X) \
5162 ? ((X) * SCALE + OFFSET) : Scale (X))
5165 You may use this built-in function in either a macro or an inline
5166 function. However, if you use it in an inlined function and pass an
5167 argument of the function as the argument to the built-in, GCC will
5168 never return 1 when you call the inline function with a string constant
5169 or compound literal (@pxref{Compound Literals}) and will not return 1
5170 when you pass a constant numeric value to the inline function unless you
5171 specify the @option{-O} option.
5173 You may also use @code{__builtin_constant_p} in initializers for static
5174 data. For instance, you can write
5177 static const int table[] = @{
5178 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5184 This is an acceptable initializer even if @var{EXPRESSION} is not a
5185 constant expression. GCC must be more conservative about evaluating the
5186 built-in in this case, because it has no opportunity to perform
5189 Previous versions of GCC did not accept this built-in in data
5190 initializers. The earliest version where it is completely safe is
5194 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5195 @opindex fprofile-arcs
5196 You may use @code{__builtin_expect} to provide the compiler with
5197 branch prediction information. In general, you should prefer to
5198 use actual profile feedback for this (@option{-fprofile-arcs}), as
5199 programmers are notoriously bad at predicting how their programs
5200 actually perform. However, there are applications in which this
5201 data is hard to collect.
5203 The return value is the value of @var{exp}, which should be an
5204 integral expression. The value of @var{c} must be a compile-time
5205 constant. The semantics of the built-in are that it is expected
5206 that @var{exp} == @var{c}. For example:
5209 if (__builtin_expect (x, 0))
5214 would indicate that we do not expect to call @code{foo}, since
5215 we expect @code{x} to be zero. Since you are limited to integral
5216 expressions for @var{exp}, you should use constructions such as
5219 if (__builtin_expect (ptr != NULL, 1))
5224 when testing pointer or floating-point values.
5227 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5228 This function is used to minimize cache-miss latency by moving data into
5229 a cache before it is accessed.
5230 You can insert calls to @code{__builtin_prefetch} into code for which
5231 you know addresses of data in memory that is likely to be accessed soon.
5232 If the target supports them, data prefetch instructions will be generated.
5233 If the prefetch is done early enough before the access then the data will
5234 be in the cache by the time it is accessed.
5236 The value of @var{addr} is the address of the memory to prefetch.
5237 There are two optional arguments, @var{rw} and @var{locality}.
5238 The value of @var{rw} is a compile-time constant one or zero; one
5239 means that the prefetch is preparing for a write to the memory address
5240 and zero, the default, means that the prefetch is preparing for a read.
5241 The value @var{locality} must be a compile-time constant integer between
5242 zero and three. A value of zero means that the data has no temporal
5243 locality, so it need not be left in the cache after the access. A value
5244 of three means that the data has a high degree of temporal locality and
5245 should be left in all levels of cache possible. Values of one and two
5246 mean, respectively, a low or moderate degree of temporal locality. The
5250 for (i = 0; i < n; i++)
5253 __builtin_prefetch (&a[i+j], 1, 1);
5254 __builtin_prefetch (&b[i+j], 0, 1);
5259 Data prefetch does not generate faults if @var{addr} is invalid, but
5260 the address expression itself must be valid. For example, a prefetch
5261 of @code{p->next} will not fault if @code{p->next} is not a valid
5262 address, but evaluation will fault if @code{p} is not a valid address.
5264 If the target does not support data prefetch, the address expression
5265 is evaluated if it includes side effects but no other code is generated
5266 and GCC does not issue a warning.
5269 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5270 Returns a positive infinity, if supported by the floating-point format,
5271 else @code{DBL_MAX}. This function is suitable for implementing the
5272 ISO C macro @code{HUGE_VAL}.
5275 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5276 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5279 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5280 Similar to @code{__builtin_huge_val}, except the return
5281 type is @code{long double}.
5284 @deftypefn {Built-in Function} double __builtin_inf (void)
5285 Similar to @code{__builtin_huge_val}, except a warning is generated
5286 if the target floating-point format does not support infinities.
5289 @deftypefn {Built-in Function} float __builtin_inff (void)
5290 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5291 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5294 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5295 Similar to @code{__builtin_inf}, except the return
5296 type is @code{long double}.
5299 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
5300 This is an implementation of the ISO C99 function @code{nan}.
5302 Since ISO C99 defines this function in terms of @code{strtod}, which we
5303 do not implement, a description of the parsing is in order. The string
5304 is parsed as by @code{strtol}; that is, the base is recognized by
5305 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
5306 in the significand such that the least significant bit of the number
5307 is at the least significant bit of the significand. The number is
5308 truncated to fit the significand field provided. The significand is
5309 forced to be a quiet NaN@.
5311 This function, if given a string literal, is evaluated early enough
5312 that it is considered a compile-time constant.
5315 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5316 Similar to @code{__builtin_nan}, except the return type is @code{float}.
5319 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5320 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5323 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
5324 Similar to @code{__builtin_nan}, except the significand is forced
5325 to be a signaling NaN@. The @code{nans} function is proposed by
5326 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
5329 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5330 Similar to @code{__builtin_nans}, except the return type is @code{float}.
5333 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5334 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5337 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5338 Returns one plus the index of the least significant 1-bit of @var{x}, or
5339 if @var{x} is zero, returns zero.
5342 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5343 Returns the number of leading 0-bits in @var{x}, starting at the most
5344 significant bit position. If @var{x} is 0, the result is undefined.
5347 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5348 Returns the number of trailing 0-bits in @var{x}, starting at the least
5349 significant bit position. If @var{x} is 0, the result is undefined.
5352 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5353 Returns the number of 1-bits in @var{x}.
5356 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5357 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
5361 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5362 Similar to @code{__builtin_ffs}, except the argument type is
5363 @code{unsigned long}.
5366 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5367 Similar to @code{__builtin_clz}, except the argument type is
5368 @code{unsigned long}.
5371 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5372 Similar to @code{__builtin_ctz}, except the argument type is
5373 @code{unsigned long}.
5376 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5377 Similar to @code{__builtin_popcount}, except the argument type is
5378 @code{unsigned long}.
5381 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5382 Similar to @code{__builtin_parity}, except the argument type is
5383 @code{unsigned long}.
5386 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5387 Similar to @code{__builtin_ffs}, except the argument type is
5388 @code{unsigned long long}.
5391 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5392 Similar to @code{__builtin_clz}, except the argument type is
5393 @code{unsigned long long}.
5396 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5397 Similar to @code{__builtin_ctz}, except the argument type is
5398 @code{unsigned long long}.
5401 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5402 Similar to @code{__builtin_popcount}, except the argument type is
5403 @code{unsigned long long}.
5406 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5407 Similar to @code{__builtin_parity}, except the argument type is
5408 @code{unsigned long long}.
5411 @deftypefn {Built-in Function} double __builtin_powi (double, int)
5412 Returns the first argument raised to the power of the second. Unlike the
5413 @code{pow} function no guarantees about precision and rounding are made.
5416 @deftypefn {Built-in Function} float __builtin_powif (float, int)
5417 Similar to @code{__builtin_powi}, except the argument and return types
5421 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
5422 Similar to @code{__builtin_powi}, except the argument and return types
5423 are @code{long double}.
5427 @node Target Builtins
5428 @section Built-in Functions Specific to Particular Target Machines
5430 On some target machines, GCC supports many built-in functions specific
5431 to those machines. Generally these generate calls to specific machine
5432 instructions, but allow the compiler to schedule those calls.
5435 * Alpha Built-in Functions::
5436 * ARM Built-in Functions::
5437 * FR-V Built-in Functions::
5438 * X86 Built-in Functions::
5439 * MIPS Paired-Single Support::
5440 * PowerPC AltiVec Built-in Functions::
5441 * SPARC VIS Built-in Functions::
5444 @node Alpha Built-in Functions
5445 @subsection Alpha Built-in Functions
5447 These built-in functions are available for the Alpha family of
5448 processors, depending on the command-line switches used.
5450 The following built-in functions are always available. They
5451 all generate the machine instruction that is part of the name.
5454 long __builtin_alpha_implver (void)
5455 long __builtin_alpha_rpcc (void)
5456 long __builtin_alpha_amask (long)
5457 long __builtin_alpha_cmpbge (long, long)
5458 long __builtin_alpha_extbl (long, long)
5459 long __builtin_alpha_extwl (long, long)
5460 long __builtin_alpha_extll (long, long)
5461 long __builtin_alpha_extql (long, long)
5462 long __builtin_alpha_extwh (long, long)
5463 long __builtin_alpha_extlh (long, long)
5464 long __builtin_alpha_extqh (long, long)
5465 long __builtin_alpha_insbl (long, long)
5466 long __builtin_alpha_inswl (long, long)
5467 long __builtin_alpha_insll (long, long)
5468 long __builtin_alpha_insql (long, long)
5469 long __builtin_alpha_inswh (long, long)
5470 long __builtin_alpha_inslh (long, long)
5471 long __builtin_alpha_insqh (long, long)
5472 long __builtin_alpha_mskbl (long, long)
5473 long __builtin_alpha_mskwl (long, long)
5474 long __builtin_alpha_mskll (long, long)
5475 long __builtin_alpha_mskql (long, long)
5476 long __builtin_alpha_mskwh (long, long)
5477 long __builtin_alpha_msklh (long, long)
5478 long __builtin_alpha_mskqh (long, long)
5479 long __builtin_alpha_umulh (long, long)
5480 long __builtin_alpha_zap (long, long)
5481 long __builtin_alpha_zapnot (long, long)
5484 The following built-in functions are always with @option{-mmax}
5485 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5486 later. They all generate the machine instruction that is part
5490 long __builtin_alpha_pklb (long)
5491 long __builtin_alpha_pkwb (long)
5492 long __builtin_alpha_unpkbl (long)
5493 long __builtin_alpha_unpkbw (long)
5494 long __builtin_alpha_minub8 (long, long)
5495 long __builtin_alpha_minsb8 (long, long)
5496 long __builtin_alpha_minuw4 (long, long)
5497 long __builtin_alpha_minsw4 (long, long)
5498 long __builtin_alpha_maxub8 (long, long)
5499 long __builtin_alpha_maxsb8 (long, long)
5500 long __builtin_alpha_maxuw4 (long, long)
5501 long __builtin_alpha_maxsw4 (long, long)
5502 long __builtin_alpha_perr (long, long)
5505 The following built-in functions are always with @option{-mcix}
5506 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5507 later. They all generate the machine instruction that is part
5511 long __builtin_alpha_cttz (long)
5512 long __builtin_alpha_ctlz (long)
5513 long __builtin_alpha_ctpop (long)
5516 The following builtins are available on systems that use the OSF/1
5517 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
5518 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5519 @code{rdval} and @code{wrval}.
5522 void *__builtin_thread_pointer (void)
5523 void __builtin_set_thread_pointer (void *)
5526 @node ARM Built-in Functions
5527 @subsection ARM Built-in Functions
5529 These built-in functions are available for the ARM family of
5530 processors, when the @option{-mcpu=iwmmxt} switch is used:
5533 typedef int v2si __attribute__ ((vector_size (8)));
5534 typedef short v4hi __attribute__ ((vector_size (8)));
5535 typedef char v8qi __attribute__ ((vector_size (8)));
5537 int __builtin_arm_getwcx (int)
5538 void __builtin_arm_setwcx (int, int)
5539 int __builtin_arm_textrmsb (v8qi, int)
5540 int __builtin_arm_textrmsh (v4hi, int)
5541 int __builtin_arm_textrmsw (v2si, int)
5542 int __builtin_arm_textrmub (v8qi, int)
5543 int __builtin_arm_textrmuh (v4hi, int)
5544 int __builtin_arm_textrmuw (v2si, int)
5545 v8qi __builtin_arm_tinsrb (v8qi, int)
5546 v4hi __builtin_arm_tinsrh (v4hi, int)
5547 v2si __builtin_arm_tinsrw (v2si, int)
5548 long long __builtin_arm_tmia (long long, int, int)
5549 long long __builtin_arm_tmiabb (long long, int, int)
5550 long long __builtin_arm_tmiabt (long long, int, int)
5551 long long __builtin_arm_tmiaph (long long, int, int)
5552 long long __builtin_arm_tmiatb (long long, int, int)
5553 long long __builtin_arm_tmiatt (long long, int, int)
5554 int __builtin_arm_tmovmskb (v8qi)
5555 int __builtin_arm_tmovmskh (v4hi)
5556 int __builtin_arm_tmovmskw (v2si)
5557 long long __builtin_arm_waccb (v8qi)
5558 long long __builtin_arm_wacch (v4hi)
5559 long long __builtin_arm_waccw (v2si)
5560 v8qi __builtin_arm_waddb (v8qi, v8qi)
5561 v8qi __builtin_arm_waddbss (v8qi, v8qi)
5562 v8qi __builtin_arm_waddbus (v8qi, v8qi)
5563 v4hi __builtin_arm_waddh (v4hi, v4hi)
5564 v4hi __builtin_arm_waddhss (v4hi, v4hi)
5565 v4hi __builtin_arm_waddhus (v4hi, v4hi)
5566 v2si __builtin_arm_waddw (v2si, v2si)
5567 v2si __builtin_arm_waddwss (v2si, v2si)
5568 v2si __builtin_arm_waddwus (v2si, v2si)
5569 v8qi __builtin_arm_walign (v8qi, v8qi, int)
5570 long long __builtin_arm_wand(long long, long long)
5571 long long __builtin_arm_wandn (long long, long long)
5572 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5573 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5574 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5575 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5576 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5577 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5578 v2si __builtin_arm_wcmpeqw (v2si, v2si)
5579 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5580 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5581 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5582 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5583 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5584 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5585 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5586 long long __builtin_arm_wmacsz (v4hi, v4hi)
5587 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5588 long long __builtin_arm_wmacuz (v4hi, v4hi)
5589 v4hi __builtin_arm_wmadds (v4hi, v4hi)
5590 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5591 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5592 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5593 v2si __builtin_arm_wmaxsw (v2si, v2si)
5594 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5595 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5596 v2si __builtin_arm_wmaxuw (v2si, v2si)
5597 v8qi __builtin_arm_wminsb (v8qi, v8qi)
5598 v4hi __builtin_arm_wminsh (v4hi, v4hi)
5599 v2si __builtin_arm_wminsw (v2si, v2si)
5600 v8qi __builtin_arm_wminub (v8qi, v8qi)
5601 v4hi __builtin_arm_wminuh (v4hi, v4hi)
5602 v2si __builtin_arm_wminuw (v2si, v2si)
5603 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5604 v4hi __builtin_arm_wmulul (v4hi, v4hi)
5605 v4hi __builtin_arm_wmulum (v4hi, v4hi)
5606 long long __builtin_arm_wor (long long, long long)
5607 v2si __builtin_arm_wpackdss (long long, long long)
5608 v2si __builtin_arm_wpackdus (long long, long long)
5609 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5610 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5611 v4hi __builtin_arm_wpackwss (v2si, v2si)
5612 v4hi __builtin_arm_wpackwus (v2si, v2si)
5613 long long __builtin_arm_wrord (long long, long long)
5614 long long __builtin_arm_wrordi (long long, int)
5615 v4hi __builtin_arm_wrorh (v4hi, long long)
5616 v4hi __builtin_arm_wrorhi (v4hi, int)
5617 v2si __builtin_arm_wrorw (v2si, long long)
5618 v2si __builtin_arm_wrorwi (v2si, int)
5619 v2si __builtin_arm_wsadb (v8qi, v8qi)
5620 v2si __builtin_arm_wsadbz (v8qi, v8qi)
5621 v2si __builtin_arm_wsadh (v4hi, v4hi)
5622 v2si __builtin_arm_wsadhz (v4hi, v4hi)
5623 v4hi __builtin_arm_wshufh (v4hi, int)
5624 long long __builtin_arm_wslld (long long, long long)
5625 long long __builtin_arm_wslldi (long long, int)
5626 v4hi __builtin_arm_wsllh (v4hi, long long)
5627 v4hi __builtin_arm_wsllhi (v4hi, int)
5628 v2si __builtin_arm_wsllw (v2si, long long)
5629 v2si __builtin_arm_wsllwi (v2si, int)
5630 long long __builtin_arm_wsrad (long long, long long)
5631 long long __builtin_arm_wsradi (long long, int)
5632 v4hi __builtin_arm_wsrah (v4hi, long long)
5633 v4hi __builtin_arm_wsrahi (v4hi, int)
5634 v2si __builtin_arm_wsraw (v2si, long long)
5635 v2si __builtin_arm_wsrawi (v2si, int)
5636 long long __builtin_arm_wsrld (long long, long long)
5637 long long __builtin_arm_wsrldi (long long, int)
5638 v4hi __builtin_arm_wsrlh (v4hi, long long)
5639 v4hi __builtin_arm_wsrlhi (v4hi, int)
5640 v2si __builtin_arm_wsrlw (v2si, long long)
5641 v2si __builtin_arm_wsrlwi (v2si, int)
5642 v8qi __builtin_arm_wsubb (v8qi, v8qi)
5643 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5644 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5645 v4hi __builtin_arm_wsubh (v4hi, v4hi)
5646 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5647 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5648 v2si __builtin_arm_wsubw (v2si, v2si)
5649 v2si __builtin_arm_wsubwss (v2si, v2si)
5650 v2si __builtin_arm_wsubwus (v2si, v2si)
5651 v4hi __builtin_arm_wunpckehsb (v8qi)
5652 v2si __builtin_arm_wunpckehsh (v4hi)
5653 long long __builtin_arm_wunpckehsw (v2si)
5654 v4hi __builtin_arm_wunpckehub (v8qi)
5655 v2si __builtin_arm_wunpckehuh (v4hi)
5656 long long __builtin_arm_wunpckehuw (v2si)
5657 v4hi __builtin_arm_wunpckelsb (v8qi)
5658 v2si __builtin_arm_wunpckelsh (v4hi)
5659 long long __builtin_arm_wunpckelsw (v2si)
5660 v4hi __builtin_arm_wunpckelub (v8qi)
5661 v2si __builtin_arm_wunpckeluh (v4hi)
5662 long long __builtin_arm_wunpckeluw (v2si)
5663 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5664 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5665 v2si __builtin_arm_wunpckihw (v2si, v2si)
5666 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5667 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5668 v2si __builtin_arm_wunpckilw (v2si, v2si)
5669 long long __builtin_arm_wxor (long long, long long)
5670 long long __builtin_arm_wzero ()
5673 @node FR-V Built-in Functions
5674 @subsection FR-V Built-in Functions
5676 GCC provides many FR-V-specific built-in functions. In general,
5677 these functions are intended to be compatible with those described
5678 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
5679 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
5680 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
5681 pointer rather than by value.
5683 Most of the functions are named after specific FR-V instructions.
5684 Such functions are said to be ``directly mapped'' and are summarized
5685 here in tabular form.
5689 * Directly-mapped Integer Functions::
5690 * Directly-mapped Media Functions::
5691 * Other Built-in Functions::
5694 @node Argument Types
5695 @subsubsection Argument Types
5697 The arguments to the built-in functions can be divided into three groups:
5698 register numbers, compile-time constants and run-time values. In order
5699 to make this classification clear at a glance, the arguments and return
5700 values are given the following pseudo types:
5702 @multitable @columnfractions .20 .30 .15 .35
5703 @item Pseudo type @tab Real C type @tab Constant? @tab Description
5704 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
5705 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
5706 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
5707 @item @code{uw2} @tab @code{unsigned long long} @tab No
5708 @tab an unsigned doubleword
5709 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
5710 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
5711 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
5712 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
5715 These pseudo types are not defined by GCC, they are simply a notational
5716 convenience used in this manual.
5718 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
5719 and @code{sw2} are evaluated at run time. They correspond to
5720 register operands in the underlying FR-V instructions.
5722 @code{const} arguments represent immediate operands in the underlying
5723 FR-V instructions. They must be compile-time constants.
5725 @code{acc} arguments are evaluated at compile time and specify the number
5726 of an accumulator register. For example, an @code{acc} argument of 2
5727 will select the ACC2 register.
5729 @code{iacc} arguments are similar to @code{acc} arguments but specify the
5730 number of an IACC register. See @pxref{Other Built-in Functions}
5733 @node Directly-mapped Integer Functions
5734 @subsubsection Directly-mapped Integer Functions
5736 The functions listed below map directly to FR-V I-type instructions.
5738 @multitable @columnfractions .45 .32 .23
5739 @item Function prototype @tab Example usage @tab Assembly output
5740 @item @code{sw1 __ADDSS (sw1, sw1)}
5741 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
5742 @tab @code{ADDSS @var{a},@var{b},@var{c}}
5743 @item @code{sw1 __SCAN (sw1, sw1)}
5744 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
5745 @tab @code{SCAN @var{a},@var{b},@var{c}}
5746 @item @code{sw1 __SCUTSS (sw1)}
5747 @tab @code{@var{b} = __SCUTSS (@var{a})}
5748 @tab @code{SCUTSS @var{a},@var{b}}
5749 @item @code{sw1 __SLASS (sw1, sw1)}
5750 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
5751 @tab @code{SLASS @var{a},@var{b},@var{c}}
5752 @item @code{void __SMASS (sw1, sw1)}
5753 @tab @code{__SMASS (@var{a}, @var{b})}
5754 @tab @code{SMASS @var{a},@var{b}}
5755 @item @code{void __SMSSS (sw1, sw1)}
5756 @tab @code{__SMSSS (@var{a}, @var{b})}
5757 @tab @code{SMSSS @var{a},@var{b}}
5758 @item @code{void __SMU (sw1, sw1)}
5759 @tab @code{__SMU (@var{a}, @var{b})}
5760 @tab @code{SMU @var{a},@var{b}}
5761 @item @code{sw2 __SMUL (sw1, sw1)}
5762 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
5763 @tab @code{SMUL @var{a},@var{b},@var{c}}
5764 @item @code{sw1 __SUBSS (sw1, sw1)}
5765 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
5766 @tab @code{SUBSS @var{a},@var{b},@var{c}}
5767 @item @code{uw2 __UMUL (uw1, uw1)}
5768 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
5769 @tab @code{UMUL @var{a},@var{b},@var{c}}
5772 @node Directly-mapped Media Functions
5773 @subsubsection Directly-mapped Media Functions
5775 The functions listed below map directly to FR-V M-type instructions.
5777 @multitable @columnfractions .45 .32 .23
5778 @item Function prototype @tab Example usage @tab Assembly output
5779 @item @code{uw1 __MABSHS (sw1)}
5780 @tab @code{@var{b} = __MABSHS (@var{a})}
5781 @tab @code{MABSHS @var{a},@var{b}}
5782 @item @code{void __MADDACCS (acc, acc)}
5783 @tab @code{__MADDACCS (@var{b}, @var{a})}
5784 @tab @code{MADDACCS @var{a},@var{b}}
5785 @item @code{sw1 __MADDHSS (sw1, sw1)}
5786 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
5787 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
5788 @item @code{uw1 __MADDHUS (uw1, uw1)}
5789 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
5790 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
5791 @item @code{uw1 __MAND (uw1, uw1)}
5792 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
5793 @tab @code{MAND @var{a},@var{b},@var{c}}
5794 @item @code{void __MASACCS (acc, acc)}
5795 @tab @code{__MASACCS (@var{b}, @var{a})}
5796 @tab @code{MASACCS @var{a},@var{b}}
5797 @item @code{uw1 __MAVEH (uw1, uw1)}
5798 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
5799 @tab @code{MAVEH @var{a},@var{b},@var{c}}
5800 @item @code{uw2 __MBTOH (uw1)}
5801 @tab @code{@var{b} = __MBTOH (@var{a})}
5802 @tab @code{MBTOH @var{a},@var{b}}
5803 @item @code{void __MBTOHE (uw1 *, uw1)}
5804 @tab @code{__MBTOHE (&@var{b}, @var{a})}
5805 @tab @code{MBTOHE @var{a},@var{b}}
5806 @item @code{void __MCLRACC (acc)}
5807 @tab @code{__MCLRACC (@var{a})}
5808 @tab @code{MCLRACC @var{a}}
5809 @item @code{void __MCLRACCA (void)}
5810 @tab @code{__MCLRACCA ()}
5811 @tab @code{MCLRACCA}
5812 @item @code{uw1 __Mcop1 (uw1, uw1)}
5813 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
5814 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
5815 @item @code{uw1 __Mcop2 (uw1, uw1)}
5816 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
5817 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
5818 @item @code{uw1 __MCPLHI (uw2, const)}
5819 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
5820 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
5821 @item @code{uw1 __MCPLI (uw2, const)}
5822 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
5823 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
5824 @item @code{void __MCPXIS (acc, sw1, sw1)}
5825 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
5826 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
5827 @item @code{void __MCPXIU (acc, uw1, uw1)}
5828 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
5829 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
5830 @item @code{void __MCPXRS (acc, sw1, sw1)}
5831 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
5832 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
5833 @item @code{void __MCPXRU (acc, uw1, uw1)}
5834 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
5835 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
5836 @item @code{uw1 __MCUT (acc, uw1)}
5837 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
5838 @tab @code{MCUT @var{a},@var{b},@var{c}}
5839 @item @code{uw1 __MCUTSS (acc, sw1)}
5840 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
5841 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
5842 @item @code{void __MDADDACCS (acc, acc)}
5843 @tab @code{__MDADDACCS (@var{b}, @var{a})}
5844 @tab @code{MDADDACCS @var{a},@var{b}}
5845 @item @code{void __MDASACCS (acc, acc)}
5846 @tab @code{__MDASACCS (@var{b}, @var{a})}
5847 @tab @code{MDASACCS @var{a},@var{b}}
5848 @item @code{uw2 __MDCUTSSI (acc, const)}
5849 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
5850 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
5851 @item @code{uw2 __MDPACKH (uw2, uw2)}
5852 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
5853 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
5854 @item @code{uw2 __MDROTLI (uw2, const)}
5855 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
5856 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
5857 @item @code{void __MDSUBACCS (acc, acc)}
5858 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
5859 @tab @code{MDSUBACCS @var{a},@var{b}}
5860 @item @code{void __MDUNPACKH (uw1 *, uw2)}
5861 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
5862 @tab @code{MDUNPACKH @var{a},@var{b}}
5863 @item @code{uw2 __MEXPDHD (uw1, const)}
5864 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
5865 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
5866 @item @code{uw1 __MEXPDHW (uw1, const)}
5867 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
5868 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
5869 @item @code{uw1 __MHDSETH (uw1, const)}
5870 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
5871 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
5872 @item @code{sw1 __MHDSETS (const)}
5873 @tab @code{@var{b} = __MHDSETS (@var{a})}
5874 @tab @code{MHDSETS #@var{a},@var{b}}
5875 @item @code{uw1 __MHSETHIH (uw1, const)}
5876 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
5877 @tab @code{MHSETHIH #@var{a},@var{b}}
5878 @item @code{sw1 __MHSETHIS (sw1, const)}
5879 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
5880 @tab @code{MHSETHIS #@var{a},@var{b}}
5881 @item @code{uw1 __MHSETLOH (uw1, const)}
5882 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
5883 @tab @code{MHSETLOH #@var{a},@var{b}}
5884 @item @code{sw1 __MHSETLOS (sw1, const)}
5885 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
5886 @tab @code{MHSETLOS #@var{a},@var{b}}
5887 @item @code{uw1 __MHTOB (uw2)}
5888 @tab @code{@var{b} = __MHTOB (@var{a})}
5889 @tab @code{MHTOB @var{a},@var{b}}
5890 @item @code{void __MMACHS (acc, sw1, sw1)}
5891 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
5892 @tab @code{MMACHS @var{a},@var{b},@var{c}}
5893 @item @code{void __MMACHU (acc, uw1, uw1)}
5894 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
5895 @tab @code{MMACHU @var{a},@var{b},@var{c}}
5896 @item @code{void __MMRDHS (acc, sw1, sw1)}
5897 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
5898 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
5899 @item @code{void __MMRDHU (acc, uw1, uw1)}
5900 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
5901 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
5902 @item @code{void __MMULHS (acc, sw1, sw1)}
5903 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
5904 @tab @code{MMULHS @var{a},@var{b},@var{c}}
5905 @item @code{void __MMULHU (acc, uw1, uw1)}
5906 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
5907 @tab @code{MMULHU @var{a},@var{b},@var{c}}
5908 @item @code{void __MMULXHS (acc, sw1, sw1)}
5909 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
5910 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
5911 @item @code{void __MMULXHU (acc, uw1, uw1)}
5912 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
5913 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
5914 @item @code{uw1 __MNOT (uw1)}
5915 @tab @code{@var{b} = __MNOT (@var{a})}
5916 @tab @code{MNOT @var{a},@var{b}}
5917 @item @code{uw1 __MOR (uw1, uw1)}
5918 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
5919 @tab @code{MOR @var{a},@var{b},@var{c}}
5920 @item @code{uw1 __MPACKH (uh, uh)}
5921 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
5922 @tab @code{MPACKH @var{a},@var{b},@var{c}}
5923 @item @code{sw2 __MQADDHSS (sw2, sw2)}
5924 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
5925 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
5926 @item @code{uw2 __MQADDHUS (uw2, uw2)}
5927 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
5928 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
5929 @item @code{void __MQCPXIS (acc, sw2, sw2)}
5930 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
5931 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
5932 @item @code{void __MQCPXIU (acc, uw2, uw2)}
5933 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
5934 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
5935 @item @code{void __MQCPXRS (acc, sw2, sw2)}
5936 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
5937 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
5938 @item @code{void __MQCPXRU (acc, uw2, uw2)}
5939 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
5940 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
5941 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
5942 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
5943 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
5944 @item @code{sw2 __MQLMTHS (sw2, sw2)}
5945 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
5946 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
5947 @item @code{void __MQMACHS (acc, sw2, sw2)}
5948 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
5949 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
5950 @item @code{void __MQMACHU (acc, uw2, uw2)}
5951 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
5952 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
5953 @item @code{void __MQMACXHS (acc, sw2, sw2)}
5954 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
5955 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
5956 @item @code{void __MQMULHS (acc, sw2, sw2)}
5957 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
5958 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
5959 @item @code{void __MQMULHU (acc, uw2, uw2)}
5960 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
5961 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
5962 @item @code{void __MQMULXHS (acc, sw2, sw2)}
5963 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
5964 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
5965 @item @code{void __MQMULXHU (acc, uw2, uw2)}
5966 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
5967 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
5968 @item @code{sw2 __MQSATHS (sw2, sw2)}
5969 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
5970 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
5971 @item @code{uw2 __MQSLLHI (uw2, int)}
5972 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
5973 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
5974 @item @code{sw2 __MQSRAHI (sw2, int)}
5975 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
5976 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
5977 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
5978 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
5979 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
5980 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
5981 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
5982 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
5983 @item @code{void __MQXMACHS (acc, sw2, sw2)}
5984 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
5985 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
5986 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
5987 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
5988 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
5989 @item @code{uw1 __MRDACC (acc)}
5990 @tab @code{@var{b} = __MRDACC (@var{a})}
5991 @tab @code{MRDACC @var{a},@var{b}}
5992 @item @code{uw1 __MRDACCG (acc)}
5993 @tab @code{@var{b} = __MRDACCG (@var{a})}
5994 @tab @code{MRDACCG @var{a},@var{b}}
5995 @item @code{uw1 __MROTLI (uw1, const)}
5996 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
5997 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
5998 @item @code{uw1 __MROTRI (uw1, const)}
5999 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6000 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6001 @item @code{sw1 __MSATHS (sw1, sw1)}
6002 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6003 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6004 @item @code{uw1 __MSATHU (uw1, uw1)}
6005 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6006 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6007 @item @code{uw1 __MSLLHI (uw1, const)}
6008 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6009 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6010 @item @code{sw1 __MSRAHI (sw1, const)}
6011 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6012 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6013 @item @code{uw1 __MSRLHI (uw1, const)}
6014 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6015 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6016 @item @code{void __MSUBACCS (acc, acc)}
6017 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6018 @tab @code{MSUBACCS @var{a},@var{b}}
6019 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6020 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6021 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6022 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6023 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6024 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6025 @item @code{void __MTRAP (void)}
6026 @tab @code{__MTRAP ()}
6028 @item @code{uw2 __MUNPACKH (uw1)}
6029 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6030 @tab @code{MUNPACKH @var{a},@var{b}}
6031 @item @code{uw1 __MWCUT (uw2, uw1)}
6032 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6033 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6034 @item @code{void __MWTACC (acc, uw1)}
6035 @tab @code{__MWTACC (@var{b}, @var{a})}
6036 @tab @code{MWTACC @var{a},@var{b}}
6037 @item @code{void __MWTACCG (acc, uw1)}
6038 @tab @code{__MWTACCG (@var{b}, @var{a})}
6039 @tab @code{MWTACCG @var{a},@var{b}}
6040 @item @code{uw1 __MXOR (uw1, uw1)}
6041 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6042 @tab @code{MXOR @var{a},@var{b},@var{c}}
6045 @node Other Built-in Functions
6046 @subsubsection Other Built-in Functions
6048 This section describes built-in functions that are not named after
6049 a specific FR-V instruction.
6052 @item sw2 __IACCreadll (iacc @var{reg})
6053 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6054 for future expansion and must be 0.
6056 @item sw1 __IACCreadl (iacc @var{reg})
6057 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6058 Other values of @var{reg} are rejected as invalid.
6060 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6061 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6062 is reserved for future expansion and must be 0.
6064 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6065 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6066 is 1. Other values of @var{reg} are rejected as invalid.
6068 @item void __data_prefetch0 (const void *@var{x})
6069 Use the @code{dcpl} instruction to load the contents of address @var{x}
6070 into the data cache.
6072 @item void __data_prefetch (const void *@var{x})
6073 Use the @code{nldub} instruction to load the contents of address @var{x}
6074 into the data cache. The instruction will be issued in slot I1@.
6077 @node X86 Built-in Functions
6078 @subsection X86 Built-in Functions
6080 These built-in functions are available for the i386 and x86-64 family
6081 of computers, depending on the command-line switches used.
6083 The following machine modes are available for use with MMX built-in functions
6084 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6085 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6086 vector of eight 8-bit integers. Some of the built-in functions operate on
6087 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6089 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6090 of two 32-bit floating point values.
6092 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6093 floating point values. Some instructions use a vector of four 32-bit
6094 integers, these use @code{V4SI}. Finally, some instructions operate on an
6095 entire vector register, interpreting it as a 128-bit integer, these use mode
6098 The following built-in functions are made available by @option{-mmmx}.
6099 All of them generate the machine instruction that is part of the name.
6102 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6103 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6104 v2si __builtin_ia32_paddd (v2si, v2si)
6105 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6106 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6107 v2si __builtin_ia32_psubd (v2si, v2si)
6108 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6109 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6110 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6111 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6112 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6113 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6114 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6115 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6116 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6117 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6118 di __builtin_ia32_pand (di, di)
6119 di __builtin_ia32_pandn (di,di)
6120 di __builtin_ia32_por (di, di)
6121 di __builtin_ia32_pxor (di, di)
6122 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6123 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6124 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6125 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6126 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6127 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6128 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6129 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6130 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6131 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6132 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6133 v2si __builtin_ia32_punpckldq (v2si, v2si)
6134 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6135 v4hi __builtin_ia32_packssdw (v2si, v2si)
6136 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6139 The following built-in functions are made available either with
6140 @option{-msse}, or with a combination of @option{-m3dnow} and
6141 @option{-march=athlon}. All of them generate the machine
6142 instruction that is part of the name.
6145 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6146 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6147 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6148 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6149 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6150 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6151 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6152 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6153 int __builtin_ia32_pextrw (v4hi, int)
6154 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6155 int __builtin_ia32_pmovmskb (v8qi)
6156 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6157 void __builtin_ia32_movntq (di *, di)
6158 void __builtin_ia32_sfence (void)
6161 The following built-in functions are available when @option{-msse} is used.
6162 All of them generate the machine instruction that is part of the name.
6165 int __builtin_ia32_comieq (v4sf, v4sf)
6166 int __builtin_ia32_comineq (v4sf, v4sf)
6167 int __builtin_ia32_comilt (v4sf, v4sf)
6168 int __builtin_ia32_comile (v4sf, v4sf)
6169 int __builtin_ia32_comigt (v4sf, v4sf)
6170 int __builtin_ia32_comige (v4sf, v4sf)
6171 int __builtin_ia32_ucomieq (v4sf, v4sf)
6172 int __builtin_ia32_ucomineq (v4sf, v4sf)
6173 int __builtin_ia32_ucomilt (v4sf, v4sf)
6174 int __builtin_ia32_ucomile (v4sf, v4sf)
6175 int __builtin_ia32_ucomigt (v4sf, v4sf)
6176 int __builtin_ia32_ucomige (v4sf, v4sf)
6177 v4sf __builtin_ia32_addps (v4sf, v4sf)
6178 v4sf __builtin_ia32_subps (v4sf, v4sf)
6179 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6180 v4sf __builtin_ia32_divps (v4sf, v4sf)
6181 v4sf __builtin_ia32_addss (v4sf, v4sf)
6182 v4sf __builtin_ia32_subss (v4sf, v4sf)
6183 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6184 v4sf __builtin_ia32_divss (v4sf, v4sf)
6185 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6186 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6187 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6188 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6189 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6190 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6191 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6192 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6193 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6194 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6195 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6196 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6197 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6198 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6199 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6200 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6201 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6202 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6203 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6204 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6205 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6206 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6207 v4sf __builtin_ia32_minps (v4sf, v4sf)
6208 v4sf __builtin_ia32_minss (v4sf, v4sf)
6209 v4sf __builtin_ia32_andps (v4sf, v4sf)
6210 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6211 v4sf __builtin_ia32_orps (v4sf, v4sf)
6212 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6213 v4sf __builtin_ia32_movss (v4sf, v4sf)
6214 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6215 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6216 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6217 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6218 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6219 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6220 v2si __builtin_ia32_cvtps2pi (v4sf)
6221 int __builtin_ia32_cvtss2si (v4sf)
6222 v2si __builtin_ia32_cvttps2pi (v4sf)
6223 int __builtin_ia32_cvttss2si (v4sf)
6224 v4sf __builtin_ia32_rcpps (v4sf)
6225 v4sf __builtin_ia32_rsqrtps (v4sf)
6226 v4sf __builtin_ia32_sqrtps (v4sf)
6227 v4sf __builtin_ia32_rcpss (v4sf)
6228 v4sf __builtin_ia32_rsqrtss (v4sf)
6229 v4sf __builtin_ia32_sqrtss (v4sf)
6230 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6231 void __builtin_ia32_movntps (float *, v4sf)
6232 int __builtin_ia32_movmskps (v4sf)
6235 The following built-in functions are available when @option{-msse} is used.
6238 @item v4sf __builtin_ia32_loadaps (float *)
6239 Generates the @code{movaps} machine instruction as a load from memory.
6240 @item void __builtin_ia32_storeaps (float *, v4sf)
6241 Generates the @code{movaps} machine instruction as a store to memory.
6242 @item v4sf __builtin_ia32_loadups (float *)
6243 Generates the @code{movups} machine instruction as a load from memory.
6244 @item void __builtin_ia32_storeups (float *, v4sf)
6245 Generates the @code{movups} machine instruction as a store to memory.
6246 @item v4sf __builtin_ia32_loadsss (float *)
6247 Generates the @code{movss} machine instruction as a load from memory.
6248 @item void __builtin_ia32_storess (float *, v4sf)
6249 Generates the @code{movss} machine instruction as a store to memory.
6250 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6251 Generates the @code{movhps} machine instruction as a load from memory.
6252 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6253 Generates the @code{movlps} machine instruction as a load from memory
6254 @item void __builtin_ia32_storehps (v4sf, v2si *)
6255 Generates the @code{movhps} machine instruction as a store to memory.
6256 @item void __builtin_ia32_storelps (v4sf, v2si *)
6257 Generates the @code{movlps} machine instruction as a store to memory.
6260 The following built-in functions are available when @option{-msse3} is used.
6261 All of them generate the machine instruction that is part of the name.
6264 v2df __builtin_ia32_addsubpd (v2df, v2df)
6265 v2df __builtin_ia32_addsubps (v2df, v2df)
6266 v2df __builtin_ia32_haddpd (v2df, v2df)
6267 v2df __builtin_ia32_haddps (v2df, v2df)
6268 v2df __builtin_ia32_hsubpd (v2df, v2df)
6269 v2df __builtin_ia32_hsubps (v2df, v2df)
6270 v16qi __builtin_ia32_lddqu (char const *)
6271 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6272 v2df __builtin_ia32_movddup (v2df)
6273 v4sf __builtin_ia32_movshdup (v4sf)
6274 v4sf __builtin_ia32_movsldup (v4sf)
6275 void __builtin_ia32_mwait (unsigned int, unsigned int)
6278 The following built-in functions are available when @option{-msse3} is used.
6281 @item v2df __builtin_ia32_loadddup (double const *)
6282 Generates the @code{movddup} machine instruction as a load from memory.
6285 The following built-in functions are available when @option{-m3dnow} is used.
6286 All of them generate the machine instruction that is part of the name.
6289 void __builtin_ia32_femms (void)
6290 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6291 v2si __builtin_ia32_pf2id (v2sf)
6292 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6293 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6294 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6295 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6296 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6297 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6298 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6299 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6300 v2sf __builtin_ia32_pfrcp (v2sf)
6301 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6302 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6303 v2sf __builtin_ia32_pfrsqrt (v2sf)
6304 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6305 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6306 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6307 v2sf __builtin_ia32_pi2fd (v2si)
6308 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6311 The following built-in functions are available when both @option{-m3dnow}
6312 and @option{-march=athlon} are used. All of them generate the machine
6313 instruction that is part of the name.
6316 v2si __builtin_ia32_pf2iw (v2sf)
6317 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6318 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6319 v2sf __builtin_ia32_pi2fw (v2si)
6320 v2sf __builtin_ia32_pswapdsf (v2sf)
6321 v2si __builtin_ia32_pswapdsi (v2si)
6324 @node MIPS Paired-Single Support
6325 @subsection MIPS Paired-Single Support
6327 The MIPS64 architecture includes a number of instructions that
6328 operate on pairs of single-precision floating-point values.
6329 Each pair is packed into a 64-bit floating-point register,
6330 with one element being designated the ``upper half'' and
6331 the other being designated the ``lower half''.
6333 GCC supports paired-single operations using both the generic
6334 vector extensions (@pxref{Vector Extensions}) and a collection of
6335 MIPS-specific built-in functions. Both kinds of support are
6336 enabled by the @option{-mpaired-single} command-line option.
6338 The vector type associated with paired-single values is usually
6339 called @code{v2sf}. It can be defined in C as follows:
6342 typedef float v2sf __attribute__ ((vector_size (8)));
6345 @code{v2sf} values are initialized in the same way as aggregates.
6349 v2sf a = @{1.5, 9.1@};
6352 b = (v2sf) @{e, f@};
6355 @emph{Note:} The CPU's endianness determines which value is stored in
6356 the upper half of a register and which value is stored in the lower half.
6357 On little-endian targets, the first value is the lower one and the second
6358 value is the upper one. The opposite order applies to big-endian targets.
6359 For example, the code above will set the lower half of @code{a} to
6360 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
6363 * Paired-Single Arithmetic::
6364 * Paired-Single Built-in Functions::
6365 * MIPS-3D Built-in Functions::
6368 @node Paired-Single Arithmetic
6369 @subsubsection Paired-Single Arithmetic
6371 The table below lists the @code{v2sf} operations for which hardware
6372 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
6373 values and @code{x} is an integral value.
6375 @multitable @columnfractions .50 .50
6376 @item C code @tab MIPS instruction
6377 @item @code{a + b} @tab @code{add.ps}
6378 @item @code{a - b} @tab @code{sub.ps}
6379 @item @code{-a} @tab @code{neg.ps}
6380 @item @code{a * b} @tab @code{mul.ps}
6381 @item @code{a * b + c} @tab @code{madd.ps}
6382 @item @code{a * b - c} @tab @code{msub.ps}
6383 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
6384 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
6385 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
6388 Note that the multiply-accumulate instructions can be disabled
6389 using the command-line option @code{-mno-fused-madd}.
6391 @node Paired-Single Built-in Functions
6392 @subsubsection Paired-Single Built-in Functions
6394 The following paired-single functions map directly to a particular
6395 MIPS instruction. Please refer to the architecture specification
6396 for details on what each instruction does.
6399 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
6400 Pair lower lower (@code{pll.ps}).
6402 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
6403 Pair upper lower (@code{pul.ps}).
6405 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
6406 Pair lower upper (@code{plu.ps}).
6408 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
6409 Pair upper upper (@code{puu.ps}).
6411 @item v2sf __builtin_mips_cvt_ps_s (float, float)
6412 Convert pair to paired single (@code{cvt.ps.s}).
6414 @item float __builtin_mips_cvt_s_pl (v2sf)
6415 Convert pair lower to single (@code{cvt.s.pl}).
6417 @item float __builtin_mips_cvt_s_pu (v2sf)
6418 Convert pair upper to single (@code{cvt.s.pu}).
6420 @item v2sf __builtin_mips_abs_ps (v2sf)
6421 Absolute value (@code{abs.ps}).
6423 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
6424 Align variable (@code{alnv.ps}).
6426 @emph{Note:} The value of the third parameter must be 0 or 4
6427 modulo 8, otherwise the result will be unpredictable. Please read the
6428 instruction description for details.
6431 The following multi-instruction functions are also available.
6432 In each case, @var{cond} can be any of the 16 floating-point conditions:
6433 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
6434 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
6435 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
6438 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6439 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6440 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
6441 @code{movt.ps}/@code{movf.ps}).
6443 The @code{movt} functions return the value @var{x} computed by:
6446 c.@var{cond}.ps @var{cc},@var{a},@var{b}
6447 mov.ps @var{x},@var{c}
6448 movt.ps @var{x},@var{d},@var{cc}
6451 The @code{movf} functions are similar but use @code{movf.ps} instead
6454 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6455 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6456 Comparison of two paired-single values (@code{c.@var{cond}.ps},
6457 @code{bc1t}/@code{bc1f}).
6459 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
6460 and return either the upper or lower half of the result. For example:
6464 if (__builtin_mips_upper_c_eq_ps (a, b))
6465 upper_halves_are_equal ();
6467 upper_halves_are_unequal ();
6469 if (__builtin_mips_lower_c_eq_ps (a, b))
6470 lower_halves_are_equal ();
6472 lower_halves_are_unequal ();
6476 @node MIPS-3D Built-in Functions
6477 @subsubsection MIPS-3D Built-in Functions
6479 The MIPS-3D Application-Specific Extension (ASE) includes additional
6480 paired-single instructions that are designed to improve the performance
6481 of 3D graphics operations. Support for these instructions is controlled
6482 by the @option{-mips3d} command-line option.
6484 The functions listed below map directly to a particular MIPS-3D
6485 instruction. Please refer to the architecture specification for
6486 more details on what each instruction does.
6489 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
6490 Reduction add (@code{addr.ps}).
6492 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
6493 Reduction multiply (@code{mulr.ps}).
6495 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
6496 Convert paired single to paired word (@code{cvt.pw.ps}).
6498 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
6499 Convert paired word to paired single (@code{cvt.ps.pw}).
6501 @item float __builtin_mips_recip1_s (float)
6502 @itemx double __builtin_mips_recip1_d (double)
6503 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
6504 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
6506 @item float __builtin_mips_recip2_s (float, float)
6507 @itemx double __builtin_mips_recip2_d (double, double)
6508 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
6509 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
6511 @item float __builtin_mips_rsqrt1_s (float)
6512 @itemx double __builtin_mips_rsqrt1_d (double)
6513 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
6514 Reduced precision reciprocal square root (sequence step 1)
6515 (@code{rsqrt1.@var{fmt}}).
6517 @item float __builtin_mips_rsqrt2_s (float, float)
6518 @itemx double __builtin_mips_rsqrt2_d (double, double)
6519 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
6520 Reduced precision reciprocal square root (sequence step 2)
6521 (@code{rsqrt2.@var{fmt}}).
6524 The following multi-instruction functions are also available.
6525 In each case, @var{cond} can be any of the 16 floating-point conditions:
6526 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
6527 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
6528 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
6531 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
6532 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
6533 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
6534 @code{bc1t}/@code{bc1f}).
6536 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
6537 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
6542 if (__builtin_mips_cabs_eq_s (a, b))
6548 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6549 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6550 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
6551 @code{bc1t}/@code{bc1f}).
6553 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
6554 and return either the upper or lower half of the result. For example:
6558 if (__builtin_mips_upper_cabs_eq_ps (a, b))
6559 upper_halves_are_equal ();
6561 upper_halves_are_unequal ();
6563 if (__builtin_mips_lower_cabs_eq_ps (a, b))
6564 lower_halves_are_equal ();
6566 lower_halves_are_unequal ();
6569 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6570 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6571 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
6572 @code{movt.ps}/@code{movf.ps}).
6574 The @code{movt} functions return the value @var{x} computed by:
6577 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
6578 mov.ps @var{x},@var{c}
6579 movt.ps @var{x},@var{d},@var{cc}
6582 The @code{movf} functions are similar but use @code{movf.ps} instead
6585 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6586 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6587 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6588 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
6589 Comparison of two paired-single values
6590 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
6591 @code{bc1any2t}/@code{bc1any2f}).
6593 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
6594 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
6595 result is true and the @code{all} forms return true if both results are true.
6600 if (__builtin_mips_any_c_eq_ps (a, b))
6605 if (__builtin_mips_all_c_eq_ps (a, b))
6611 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6612 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6613 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6614 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
6615 Comparison of four paired-single values
6616 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
6617 @code{bc1any4t}/@code{bc1any4f}).
6619 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
6620 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
6621 The @code{any} forms return true if any of the four results are true
6622 and the @code{all} forms return true if all four results are true.
6627 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
6632 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
6639 @node PowerPC AltiVec Built-in Functions
6640 @subsection PowerPC AltiVec Built-in Functions
6642 GCC provides an interface for the PowerPC family of processors to access
6643 the AltiVec operations described in Motorola's AltiVec Programming
6644 Interface Manual. The interface is made available by including
6645 @code{<altivec.h>} and using @option{-maltivec} and
6646 @option{-mabi=altivec}. The interface supports the following vector
6650 vector unsigned char
6654 vector unsigned short
6665 GCC's implementation of the high-level language interface available from
6666 C and C++ code differs from Motorola's documentation in several ways.
6671 A vector constant is a list of constant expressions within curly braces.
6674 A vector initializer requires no cast if the vector constant is of the
6675 same type as the variable it is initializing.
6678 If @code{signed} or @code{unsigned} is omitted, the signedness of the
6679 vector type is the default signedness of the base type. The default
6680 varies depending on the operating system, so a portable program should
6681 always specify the signedness.
6684 Compiling with @option{-maltivec} adds keywords @code{__vector},
6685 @code{__pixel}, and @code{__bool}. Macros @option{vector},
6686 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
6690 GCC allows using a @code{typedef} name as the type specifier for a
6694 For C, overloaded functions are implemented with macros so the following
6698 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6701 Since @code{vec_add} is a macro, the vector constant in the example
6702 is treated as four separate arguments. Wrap the entire argument in
6703 parentheses for this to work.
6706 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
6707 Internally, GCC uses built-in functions to achieve the functionality in
6708 the aforementioned header file, but they are not supported and are
6709 subject to change without notice.
6711 The following interfaces are supported for the generic and specific
6712 AltiVec operations and the AltiVec predicates. In cases where there
6713 is a direct mapping between generic and specific operations, only the
6714 generic names are shown here, although the specific operations can also
6717 Arguments that are documented as @code{const int} require literal
6718 integral values within the range required for that operation.
6721 vector signed char vec_abs (vector signed char);
6722 vector signed short vec_abs (vector signed short);
6723 vector signed int vec_abs (vector signed int);
6724 vector float vec_abs (vector float);
6726 vector signed char vec_abss (vector signed char);
6727 vector signed short vec_abss (vector signed short);
6728 vector signed int vec_abss (vector signed int);
6730 vector signed char vec_add (vector bool char, vector signed char);
6731 vector signed char vec_add (vector signed char, vector bool char);
6732 vector signed char vec_add (vector signed char, vector signed char);
6733 vector unsigned char vec_add (vector bool char, vector unsigned char);
6734 vector unsigned char vec_add (vector unsigned char, vector bool char);
6735 vector unsigned char vec_add (vector unsigned char,
6736 vector unsigned char);
6737 vector signed short vec_add (vector bool short, vector signed short);
6738 vector signed short vec_add (vector signed short, vector bool short);
6739 vector signed short vec_add (vector signed short, vector signed short);
6740 vector unsigned short vec_add (vector bool short,
6741 vector unsigned short);
6742 vector unsigned short vec_add (vector unsigned short,
6744 vector unsigned short vec_add (vector unsigned short,
6745 vector unsigned short);
6746 vector signed int vec_add (vector bool int, vector signed int);
6747 vector signed int vec_add (vector signed int, vector bool int);
6748 vector signed int vec_add (vector signed int, vector signed int);
6749 vector unsigned int vec_add (vector bool int, vector unsigned int);
6750 vector unsigned int vec_add (vector unsigned int, vector bool int);
6751 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6752 vector float vec_add (vector float, vector float);
6754 vector float vec_vaddfp (vector float, vector float);
6756 vector signed int vec_vadduwm (vector bool int, vector signed int);
6757 vector signed int vec_vadduwm (vector signed int, vector bool int);
6758 vector signed int vec_vadduwm (vector signed int, vector signed int);
6759 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
6760 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
6761 vector unsigned int vec_vadduwm (vector unsigned int,
6762 vector unsigned int);
6764 vector signed short vec_vadduhm (vector bool short,
6765 vector signed short);
6766 vector signed short vec_vadduhm (vector signed short,
6768 vector signed short vec_vadduhm (vector signed short,
6769 vector signed short);
6770 vector unsigned short vec_vadduhm (vector bool short,
6771 vector unsigned short);
6772 vector unsigned short vec_vadduhm (vector unsigned short,
6774 vector unsigned short vec_vadduhm (vector unsigned short,
6775 vector unsigned short);
6777 vector signed char vec_vaddubm (vector bool char, vector signed char);
6778 vector signed char vec_vaddubm (vector signed char, vector bool char);
6779 vector signed char vec_vaddubm (vector signed char, vector signed char);
6780 vector unsigned char vec_vaddubm (vector bool char,
6781 vector unsigned char);
6782 vector unsigned char vec_vaddubm (vector unsigned char,
6784 vector unsigned char vec_vaddubm (vector unsigned char,
6785 vector unsigned char);
6787 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6789 vector unsigned char vec_adds (vector bool char, vector unsigned char);
6790 vector unsigned char vec_adds (vector unsigned char, vector bool char);
6791 vector unsigned char vec_adds (vector unsigned char,
6792 vector unsigned char);
6793 vector signed char vec_adds (vector bool char, vector signed char);
6794 vector signed char vec_adds (vector signed char, vector bool char);
6795 vector signed char vec_adds (vector signed char, vector signed char);
6796 vector unsigned short vec_adds (vector bool short,
6797 vector unsigned short);
6798 vector unsigned short vec_adds (vector unsigned short,
6800 vector unsigned short vec_adds (vector unsigned short,
6801 vector unsigned short);
6802 vector signed short vec_adds (vector bool short, vector signed short);
6803 vector signed short vec_adds (vector signed short, vector bool short);
6804 vector signed short vec_adds (vector signed short, vector signed short);
6805 vector unsigned int vec_adds (vector bool int, vector unsigned int);
6806 vector unsigned int vec_adds (vector unsigned int, vector bool int);
6807 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6808 vector signed int vec_adds (vector bool int, vector signed int);
6809 vector signed int vec_adds (vector signed int, vector bool int);
6810 vector signed int vec_adds (vector signed int, vector signed int);
6812 vector signed int vec_vaddsws (vector bool int, vector signed int);
6813 vector signed int vec_vaddsws (vector signed int, vector bool int);
6814 vector signed int vec_vaddsws (vector signed int, vector signed int);
6816 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
6817 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
6818 vector unsigned int vec_vadduws (vector unsigned int,
6819 vector unsigned int);
6821 vector signed short vec_vaddshs (vector bool short,
6822 vector signed short);
6823 vector signed short vec_vaddshs (vector signed short,
6825 vector signed short vec_vaddshs (vector signed short,
6826 vector signed short);
6828 vector unsigned short vec_vadduhs (vector bool short,
6829 vector unsigned short);
6830 vector unsigned short vec_vadduhs (vector unsigned short,
6832 vector unsigned short vec_vadduhs (vector unsigned short,
6833 vector unsigned short);
6835 vector signed char vec_vaddsbs (vector bool char, vector signed char);
6836 vector signed char vec_vaddsbs (vector signed char, vector bool char);
6837 vector signed char vec_vaddsbs (vector signed char, vector signed char);
6839 vector unsigned char vec_vaddubs (vector bool char,
6840 vector unsigned char);
6841 vector unsigned char vec_vaddubs (vector unsigned char,
6843 vector unsigned char vec_vaddubs (vector unsigned char,
6844 vector unsigned char);
6846 vector float vec_and (vector float, vector float);
6847 vector float vec_and (vector float, vector bool int);
6848 vector float vec_and (vector bool int, vector float);
6849 vector bool int vec_and (vector bool int, vector bool int);
6850 vector signed int vec_and (vector bool int, vector signed int);
6851 vector signed int vec_and (vector signed int, vector bool int);
6852 vector signed int vec_and (vector signed int, vector signed int);
6853 vector unsigned int vec_and (vector bool int, vector unsigned int);
6854 vector unsigned int vec_and (vector unsigned int, vector bool int);
6855 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
6856 vector bool short vec_and (vector bool short, vector bool short);
6857 vector signed short vec_and (vector bool short, vector signed short);
6858 vector signed short vec_and (vector signed short, vector bool short);
6859 vector signed short vec_and (vector signed short, vector signed short);
6860 vector unsigned short vec_and (vector bool short,
6861 vector unsigned short);
6862 vector unsigned short vec_and (vector unsigned short,
6864 vector unsigned short vec_and (vector unsigned short,
6865 vector unsigned short);
6866 vector signed char vec_and (vector bool char, vector signed char);
6867 vector bool char vec_and (vector bool char, vector bool char);
6868 vector signed char vec_and (vector signed char, vector bool char);
6869 vector signed char vec_and (vector signed char, vector signed char);
6870 vector unsigned char vec_and (vector bool char, vector unsigned char);
6871 vector unsigned char vec_and (vector unsigned char, vector bool char);
6872 vector unsigned char vec_and (vector unsigned char,
6873 vector unsigned char);
6875 vector float vec_andc (vector float, vector float);
6876 vector float vec_andc (vector float, vector bool int);
6877 vector float vec_andc (vector bool int, vector float);
6878 vector bool int vec_andc (vector bool int, vector bool int);
6879 vector signed int vec_andc (vector bool int, vector signed int);
6880 vector signed int vec_andc (vector signed int, vector bool int);
6881 vector signed int vec_andc (vector signed int, vector signed int);
6882 vector unsigned int vec_andc (vector bool int, vector unsigned int);
6883 vector unsigned int vec_andc (vector unsigned int, vector bool int);
6884 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
6885 vector bool short vec_andc (vector bool short, vector bool short);
6886 vector signed short vec_andc (vector bool short, vector signed short);
6887 vector signed short vec_andc (vector signed short, vector bool short);
6888 vector signed short vec_andc (vector signed short, vector signed short);
6889 vector unsigned short vec_andc (vector bool short,
6890 vector unsigned short);
6891 vector unsigned short vec_andc (vector unsigned short,
6893 vector unsigned short vec_andc (vector unsigned short,
6894 vector unsigned short);
6895 vector signed char vec_andc (vector bool char, vector signed char);
6896 vector bool char vec_andc (vector bool char, vector bool char);
6897 vector signed char vec_andc (vector signed char, vector bool char);
6898 vector signed char vec_andc (vector signed char, vector signed char);
6899 vector unsigned char vec_andc (vector bool char, vector unsigned char);
6900 vector unsigned char vec_andc (vector unsigned char, vector bool char);
6901 vector unsigned char vec_andc (vector unsigned char,
6902 vector unsigned char);
6904 vector unsigned char vec_avg (vector unsigned char,
6905 vector unsigned char);
6906 vector signed char vec_avg (vector signed char, vector signed char);
6907 vector unsigned short vec_avg (vector unsigned short,
6908 vector unsigned short);
6909 vector signed short vec_avg (vector signed short, vector signed short);
6910 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
6911 vector signed int vec_avg (vector signed int, vector signed int);
6913 vector signed int vec_vavgsw (vector signed int, vector signed int);
6915 vector unsigned int vec_vavguw (vector unsigned int,
6916 vector unsigned int);
6918 vector signed short vec_vavgsh (vector signed short,
6919 vector signed short);
6921 vector unsigned short vec_vavguh (vector unsigned short,
6922 vector unsigned short);
6924 vector signed char vec_vavgsb (vector signed char, vector signed char);
6926 vector unsigned char vec_vavgub (vector unsigned char,
6927 vector unsigned char);
6929 vector float vec_ceil (vector float);
6931 vector signed int vec_cmpb (vector float, vector float);
6933 vector bool char vec_cmpeq (vector signed char, vector signed char);
6934 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
6935 vector bool short vec_cmpeq (vector signed short, vector signed short);
6936 vector bool short vec_cmpeq (vector unsigned short,
6937 vector unsigned short);
6938 vector bool int vec_cmpeq (vector signed int, vector signed int);
6939 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
6940 vector bool int vec_cmpeq (vector float, vector float);
6942 vector bool int vec_vcmpeqfp (vector float, vector float);
6944 vector bool int vec_vcmpequw (vector signed int, vector signed int);
6945 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
6947 vector bool short vec_vcmpequh (vector signed short,
6948 vector signed short);
6949 vector bool short vec_vcmpequh (vector unsigned short,
6950 vector unsigned short);
6952 vector bool char vec_vcmpequb (vector signed char, vector signed char);
6953 vector bool char vec_vcmpequb (vector unsigned char,
6954 vector unsigned char);
6956 vector bool int vec_cmpge (vector float, vector float);
6958 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
6959 vector bool char vec_cmpgt (vector signed char, vector signed char);
6960 vector bool short vec_cmpgt (vector unsigned short,
6961 vector unsigned short);
6962 vector bool short vec_cmpgt (vector signed short, vector signed short);
6963 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
6964 vector bool int vec_cmpgt (vector signed int, vector signed int);
6965 vector bool int vec_cmpgt (vector float, vector float);
6967 vector bool int vec_vcmpgtfp (vector float, vector float);
6969 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
6971 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
6973 vector bool short vec_vcmpgtsh (vector signed short,
6974 vector signed short);
6976 vector bool short vec_vcmpgtuh (vector unsigned short,
6977 vector unsigned short);
6979 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
6981 vector bool char vec_vcmpgtub (vector unsigned char,
6982 vector unsigned char);
6984 vector bool int vec_cmple (vector float, vector float);
6986 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
6987 vector bool char vec_cmplt (vector signed char, vector signed char);
6988 vector bool short vec_cmplt (vector unsigned short,
6989 vector unsigned short);
6990 vector bool short vec_cmplt (vector signed short, vector signed short);
6991 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
6992 vector bool int vec_cmplt (vector signed int, vector signed int);
6993 vector bool int vec_cmplt (vector float, vector float);
6995 vector float vec_ctf (vector unsigned int, const int);
6996 vector float vec_ctf (vector signed int, const int);
6998 vector float vec_vcfsx (vector signed int, const int);
7000 vector float vec_vcfux (vector unsigned int, const int);
7002 vector signed int vec_cts (vector float, const int);
7004 vector unsigned int vec_ctu (vector float, const int);
7006 void vec_dss (const int);
7008 void vec_dssall (void);
7010 void vec_dst (const vector unsigned char *, int, const int);
7011 void vec_dst (const vector signed char *, int, const int);
7012 void vec_dst (const vector bool char *, int, const int);
7013 void vec_dst (const vector unsigned short *, int, const int);
7014 void vec_dst (const vector signed short *, int, const int);
7015 void vec_dst (const vector bool short *, int, const int);
7016 void vec_dst (const vector pixel *, int, const int);
7017 void vec_dst (const vector unsigned int *, int, const int);
7018 void vec_dst (const vector signed int *, int, const int);
7019 void vec_dst (const vector bool int *, int, const int);
7020 void vec_dst (const vector float *, int, const int);
7021 void vec_dst (const unsigned char *, int, const int);
7022 void vec_dst (const signed char *, int, const int);
7023 void vec_dst (const unsigned short *, int, const int);
7024 void vec_dst (const short *, int, const int);
7025 void vec_dst (const unsigned int *, int, const int);
7026 void vec_dst (const int *, int, const int);
7027 void vec_dst (const unsigned long *, int, const int);
7028 void vec_dst (const long *, int, const int);
7029 void vec_dst (const float *, int, const int);
7031 void vec_dstst (const vector unsigned char *, int, const int);
7032 void vec_dstst (const vector signed char *, int, const int);
7033 void vec_dstst (const vector bool char *, int, const int);
7034 void vec_dstst (const vector unsigned short *, int, const int);
7035 void vec_dstst (const vector signed short *, int, const int);
7036 void vec_dstst (const vector bool short *, int, const int);
7037 void vec_dstst (const vector pixel *, int, const int);
7038 void vec_dstst (const vector unsigned int *, int, const int);
7039 void vec_dstst (const vector signed int *, int, const int);
7040 void vec_dstst (const vector bool int *, int, const int);
7041 void vec_dstst (const vector float *, int, const int);
7042 void vec_dstst (const unsigned char *, int, const int);
7043 void vec_dstst (const signed char *, int, const int);
7044 void vec_dstst (const unsigned short *, int, const int);
7045 void vec_dstst (const short *, int, const int);
7046 void vec_dstst (const unsigned int *, int, const int);
7047 void vec_dstst (const int *, int, const int);
7048 void vec_dstst (const unsigned long *, int, const int);
7049 void vec_dstst (const long *, int, const int);
7050 void vec_dstst (const float *, int, const int);
7052 void vec_dststt (const vector unsigned char *, int, const int);
7053 void vec_dststt (const vector signed char *, int, const int);
7054 void vec_dststt (const vector bool char *, int, const int);
7055 void vec_dststt (const vector unsigned short *, int, const int);
7056 void vec_dststt (const vector signed short *, int, const int);
7057 void vec_dststt (const vector bool short *, int, const int);
7058 void vec_dststt (const vector pixel *, int, const int);
7059 void vec_dststt (const vector unsigned int *, int, const int);
7060 void vec_dststt (const vector signed int *, int, const int);
7061 void vec_dststt (const vector bool int *, int, const int);
7062 void vec_dststt (const vector float *, int, const int);
7063 void vec_dststt (const unsigned char *, int, const int);
7064 void vec_dststt (const signed char *, int, const int);
7065 void vec_dststt (const unsigned short *, int, const int);
7066 void vec_dststt (const short *, int, const int);
7067 void vec_dststt (const unsigned int *, int, const int);
7068 void vec_dststt (const int *, int, const int);
7069 void vec_dststt (const unsigned long *, int, const int);
7070 void vec_dststt (const long *, int, const int);
7071 void vec_dststt (const float *, int, const int);
7073 void vec_dstt (const vector unsigned char *, int, const int);
7074 void vec_dstt (const vector signed char *, int, const int);
7075 void vec_dstt (const vector bool char *, int, const int);
7076 void vec_dstt (const vector unsigned short *, int, const int);
7077 void vec_dstt (const vector signed short *, int, const int);
7078 void vec_dstt (const vector bool short *, int, const int);
7079 void vec_dstt (const vector pixel *, int, const int);
7080 void vec_dstt (const vector unsigned int *, int, const int);
7081 void vec_dstt (const vector signed int *, int, const int);
7082 void vec_dstt (const vector bool int *, int, const int);
7083 void vec_dstt (const vector float *, int, const int);
7084 void vec_dstt (const unsigned char *, int, const int);
7085 void vec_dstt (const signed char *, int, const int);
7086 void vec_dstt (const unsigned short *, int, const int);
7087 void vec_dstt (const short *, int, const int);
7088 void vec_dstt (const unsigned int *, int, const int);
7089 void vec_dstt (const int *, int, const int);
7090 void vec_dstt (const unsigned long *, int, const int);
7091 void vec_dstt (const long *, int, const int);
7092 void vec_dstt (const float *, int, const int);
7094 vector float vec_expte (vector float);
7096 vector float vec_floor (vector float);
7098 vector float vec_ld (int, const vector float *);
7099 vector float vec_ld (int, const float *);
7100 vector bool int vec_ld (int, const vector bool int *);
7101 vector signed int vec_ld (int, const vector signed int *);
7102 vector signed int vec_ld (int, const int *);
7103 vector signed int vec_ld (int, const long *);
7104 vector unsigned int vec_ld (int, const vector unsigned int *);
7105 vector unsigned int vec_ld (int, const unsigned int *);
7106 vector unsigned int vec_ld (int, const unsigned long *);
7107 vector bool short vec_ld (int, const vector bool short *);
7108 vector pixel vec_ld (int, const vector pixel *);
7109 vector signed short vec_ld (int, const vector signed short *);
7110 vector signed short vec_ld (int, const short *);
7111 vector unsigned short vec_ld (int, const vector unsigned short *);
7112 vector unsigned short vec_ld (int, const unsigned short *);
7113 vector bool char vec_ld (int, const vector bool char *);
7114 vector signed char vec_ld (int, const vector signed char *);
7115 vector signed char vec_ld (int, const signed char *);
7116 vector unsigned char vec_ld (int, const vector unsigned char *);
7117 vector unsigned char vec_ld (int, const unsigned char *);
7119 vector signed char vec_lde (int, const signed char *);
7120 vector unsigned char vec_lde (int, const unsigned char *);
7121 vector signed short vec_lde (int, const short *);
7122 vector unsigned short vec_lde (int, const unsigned short *);
7123 vector float vec_lde (int, const float *);
7124 vector signed int vec_lde (int, const int *);
7125 vector unsigned int vec_lde (int, const unsigned int *);
7126 vector signed int vec_lde (int, const long *);
7127 vector unsigned int vec_lde (int, const unsigned long *);
7129 vector float vec_lvewx (int, float *);
7130 vector signed int vec_lvewx (int, int *);
7131 vector unsigned int vec_lvewx (int, unsigned int *);
7132 vector signed int vec_lvewx (int, long *);
7133 vector unsigned int vec_lvewx (int, unsigned long *);
7135 vector signed short vec_lvehx (int, short *);
7136 vector unsigned short vec_lvehx (int, unsigned short *);
7138 vector signed char vec_lvebx (int, char *);
7139 vector unsigned char vec_lvebx (int, unsigned char *);
7141 vector float vec_ldl (int, const vector float *);
7142 vector float vec_ldl (int, const float *);
7143 vector bool int vec_ldl (int, const vector bool int *);
7144 vector signed int vec_ldl (int, const vector signed int *);
7145 vector signed int vec_ldl (int, const int *);
7146 vector signed int vec_ldl (int, const long *);
7147 vector unsigned int vec_ldl (int, const vector unsigned int *);
7148 vector unsigned int vec_ldl (int, const unsigned int *);
7149 vector unsigned int vec_ldl (int, const unsigned long *);
7150 vector bool short vec_ldl (int, const vector bool short *);
7151 vector pixel vec_ldl (int, const vector pixel *);
7152 vector signed short vec_ldl (int, const vector signed short *);
7153 vector signed short vec_ldl (int, const short *);
7154 vector unsigned short vec_ldl (int, const vector unsigned short *);
7155 vector unsigned short vec_ldl (int, const unsigned short *);
7156 vector bool char vec_ldl (int, const vector bool char *);
7157 vector signed char vec_ldl (int, const vector signed char *);
7158 vector signed char vec_ldl (int, const signed char *);
7159 vector unsigned char vec_ldl (int, const vector unsigned char *);
7160 vector unsigned char vec_ldl (int, const unsigned char *);
7162 vector float vec_loge (vector float);
7164 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
7165 vector unsigned char vec_lvsl (int, const volatile signed char *);
7166 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
7167 vector unsigned char vec_lvsl (int, const volatile short *);
7168 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
7169 vector unsigned char vec_lvsl (int, const volatile int *);
7170 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
7171 vector unsigned char vec_lvsl (int, const volatile long *);
7172 vector unsigned char vec_lvsl (int, const volatile float *);
7174 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
7175 vector unsigned char vec_lvsr (int, const volatile signed char *);
7176 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
7177 vector unsigned char vec_lvsr (int, const volatile short *);
7178 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
7179 vector unsigned char vec_lvsr (int, const volatile int *);
7180 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
7181 vector unsigned char vec_lvsr (int, const volatile long *);
7182 vector unsigned char vec_lvsr (int, const volatile float *);
7184 vector float vec_madd (vector float, vector float, vector float);
7186 vector signed short vec_madds (vector signed short,
7187 vector signed short,
7188 vector signed short);
7190 vector unsigned char vec_max (vector bool char, vector unsigned char);
7191 vector unsigned char vec_max (vector unsigned char, vector bool char);
7192 vector unsigned char vec_max (vector unsigned char,
7193 vector unsigned char);
7194 vector signed char vec_max (vector bool char, vector signed char);
7195 vector signed char vec_max (vector signed char, vector bool char);
7196 vector signed char vec_max (vector signed char, vector signed char);
7197 vector unsigned short vec_max (vector bool short,
7198 vector unsigned short);
7199 vector unsigned short vec_max (vector unsigned short,
7201 vector unsigned short vec_max (vector unsigned short,
7202 vector unsigned short);
7203 vector signed short vec_max (vector bool short, vector signed short);
7204 vector signed short vec_max (vector signed short, vector bool short);
7205 vector signed short vec_max (vector signed short, vector signed short);
7206 vector unsigned int vec_max (vector bool int, vector unsigned int);
7207 vector unsigned int vec_max (vector unsigned int, vector bool int);
7208 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
7209 vector signed int vec_max (vector bool int, vector signed int);
7210 vector signed int vec_max (vector signed int, vector bool int);
7211 vector signed int vec_max (vector signed int, vector signed int);
7212 vector float vec_max (vector float, vector float);
7214 vector float vec_vmaxfp (vector float, vector float);
7216 vector signed int vec_vmaxsw (vector bool int, vector signed int);
7217 vector signed int vec_vmaxsw (vector signed int, vector bool int);
7218 vector signed int vec_vmaxsw (vector signed int, vector signed int);
7220 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
7221 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
7222 vector unsigned int vec_vmaxuw (vector unsigned int,
7223 vector unsigned int);
7225 vector signed short vec_vmaxsh (vector bool short, vector signed short);
7226 vector signed short vec_vmaxsh (vector signed short, vector bool short);
7227 vector signed short vec_vmaxsh (vector signed short,
7228 vector signed short);
7230 vector unsigned short vec_vmaxuh (vector bool short,
7231 vector unsigned short);
7232 vector unsigned short vec_vmaxuh (vector unsigned short,
7234 vector unsigned short vec_vmaxuh (vector unsigned short,
7235 vector unsigned short);
7237 vector signed char vec_vmaxsb (vector bool char, vector signed char);
7238 vector signed char vec_vmaxsb (vector signed char, vector bool char);
7239 vector signed char vec_vmaxsb (vector signed char, vector signed char);
7241 vector unsigned char vec_vmaxub (vector bool char,
7242 vector unsigned char);
7243 vector unsigned char vec_vmaxub (vector unsigned char,
7245 vector unsigned char vec_vmaxub (vector unsigned char,
7246 vector unsigned char);
7248 vector bool char vec_mergeh (vector bool char, vector bool char);
7249 vector signed char vec_mergeh (vector signed char, vector signed char);
7250 vector unsigned char vec_mergeh (vector unsigned char,
7251 vector unsigned char);
7252 vector bool short vec_mergeh (vector bool short, vector bool short);
7253 vector pixel vec_mergeh (vector pixel, vector pixel);
7254 vector signed short vec_mergeh (vector signed short,
7255 vector signed short);
7256 vector unsigned short vec_mergeh (vector unsigned short,
7257 vector unsigned short);
7258 vector float vec_mergeh (vector float, vector float);
7259 vector bool int vec_mergeh (vector bool int, vector bool int);
7260 vector signed int vec_mergeh (vector signed int, vector signed int);
7261 vector unsigned int vec_mergeh (vector unsigned int,
7262 vector unsigned int);
7264 vector float vec_vmrghw (vector float, vector float);
7265 vector bool int vec_vmrghw (vector bool int, vector bool int);
7266 vector signed int vec_vmrghw (vector signed int, vector signed int);
7267 vector unsigned int vec_vmrghw (vector unsigned int,
7268 vector unsigned int);
7270 vector bool short vec_vmrghh (vector bool short, vector bool short);
7271 vector signed short vec_vmrghh (vector signed short,
7272 vector signed short);
7273 vector unsigned short vec_vmrghh (vector unsigned short,
7274 vector unsigned short);
7275 vector pixel vec_vmrghh (vector pixel, vector pixel);
7277 vector bool char vec_vmrghb (vector bool char, vector bool char);
7278 vector signed char vec_vmrghb (vector signed char, vector signed char);
7279 vector unsigned char vec_vmrghb (vector unsigned char,
7280 vector unsigned char);
7282 vector bool char vec_mergel (vector bool char, vector bool char);
7283 vector signed char vec_mergel (vector signed char, vector signed char);
7284 vector unsigned char vec_mergel (vector unsigned char,
7285 vector unsigned char);
7286 vector bool short vec_mergel (vector bool short, vector bool short);
7287 vector pixel vec_mergel (vector pixel, vector pixel);
7288 vector signed short vec_mergel (vector signed short,
7289 vector signed short);
7290 vector unsigned short vec_mergel (vector unsigned short,
7291 vector unsigned short);
7292 vector float vec_mergel (vector float, vector float);
7293 vector bool int vec_mergel (vector bool int, vector bool int);
7294 vector signed int vec_mergel (vector signed int, vector signed int);
7295 vector unsigned int vec_mergel (vector unsigned int,
7296 vector unsigned int);
7298 vector float vec_vmrglw (vector float, vector float);
7299 vector signed int vec_vmrglw (vector signed int, vector signed int);
7300 vector unsigned int vec_vmrglw (vector unsigned int,
7301 vector unsigned int);
7302 vector bool int vec_vmrglw (vector bool int, vector bool int);
7304 vector bool short vec_vmrglh (vector bool short, vector bool short);
7305 vector signed short vec_vmrglh (vector signed short,
7306 vector signed short);
7307 vector unsigned short vec_vmrglh (vector unsigned short,
7308 vector unsigned short);
7309 vector pixel vec_vmrglh (vector pixel, vector pixel);
7311 vector bool char vec_vmrglb (vector bool char, vector bool char);
7312 vector signed char vec_vmrglb (vector signed char, vector signed char);
7313 vector unsigned char vec_vmrglb (vector unsigned char,
7314 vector unsigned char);
7316 vector unsigned short vec_mfvscr (void);
7318 vector unsigned char vec_min (vector bool char, vector unsigned char);
7319 vector unsigned char vec_min (vector unsigned char, vector bool char);
7320 vector unsigned char vec_min (vector unsigned char,
7321 vector unsigned char);
7322 vector signed char vec_min (vector bool char, vector signed char);
7323 vector signed char vec_min (vector signed char, vector bool char);
7324 vector signed char vec_min (vector signed char, vector signed char);
7325 vector unsigned short vec_min (vector bool short,
7326 vector unsigned short);
7327 vector unsigned short vec_min (vector unsigned short,
7329 vector unsigned short vec_min (vector unsigned short,
7330 vector unsigned short);
7331 vector signed short vec_min (vector bool short, vector signed short);
7332 vector signed short vec_min (vector signed short, vector bool short);
7333 vector signed short vec_min (vector signed short, vector signed short);
7334 vector unsigned int vec_min (vector bool int, vector unsigned int);
7335 vector unsigned int vec_min (vector unsigned int, vector bool int);
7336 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
7337 vector signed int vec_min (vector bool int, vector signed int);
7338 vector signed int vec_min (vector signed int, vector bool int);
7339 vector signed int vec_min (vector signed int, vector signed int);
7340 vector float vec_min (vector float, vector float);
7342 vector float vec_vminfp (vector float, vector float);
7344 vector signed int vec_vminsw (vector bool int, vector signed int);
7345 vector signed int vec_vminsw (vector signed int, vector bool int);
7346 vector signed int vec_vminsw (vector signed int, vector signed int);
7348 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
7349 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
7350 vector unsigned int vec_vminuw (vector unsigned int,
7351 vector unsigned int);
7353 vector signed short vec_vminsh (vector bool short, vector signed short);
7354 vector signed short vec_vminsh (vector signed short, vector bool short);
7355 vector signed short vec_vminsh (vector signed short,
7356 vector signed short);
7358 vector unsigned short vec_vminuh (vector bool short,
7359 vector unsigned short);
7360 vector unsigned short vec_vminuh (vector unsigned short,
7362 vector unsigned short vec_vminuh (vector unsigned short,
7363 vector unsigned short);
7365 vector signed char vec_vminsb (vector bool char, vector signed char);
7366 vector signed char vec_vminsb (vector signed char, vector bool char);
7367 vector signed char vec_vminsb (vector signed char, vector signed char);
7369 vector unsigned char vec_vminub (vector bool char,
7370 vector unsigned char);
7371 vector unsigned char vec_vminub (vector unsigned char,
7373 vector unsigned char vec_vminub (vector unsigned char,
7374 vector unsigned char);
7376 vector signed short vec_mladd (vector signed short,
7377 vector signed short,
7378 vector signed short);
7379 vector signed short vec_mladd (vector signed short,
7380 vector unsigned short,
7381 vector unsigned short);
7382 vector signed short vec_mladd (vector unsigned short,
7383 vector signed short,
7384 vector signed short);
7385 vector unsigned short vec_mladd (vector unsigned short,
7386 vector unsigned short,
7387 vector unsigned short);
7389 vector signed short vec_mradds (vector signed short,
7390 vector signed short,
7391 vector signed short);
7393 vector unsigned int vec_msum (vector unsigned char,
7394 vector unsigned char,
7395 vector unsigned int);
7396 vector signed int vec_msum (vector signed char,
7397 vector unsigned char,
7399 vector unsigned int vec_msum (vector unsigned short,
7400 vector unsigned short,
7401 vector unsigned int);
7402 vector signed int vec_msum (vector signed short,
7403 vector signed short,
7406 vector signed int vec_vmsumshm (vector signed short,
7407 vector signed short,
7410 vector unsigned int vec_vmsumuhm (vector unsigned short,
7411 vector unsigned short,
7412 vector unsigned int);
7414 vector signed int vec_vmsummbm (vector signed char,
7415 vector unsigned char,
7418 vector unsigned int vec_vmsumubm (vector unsigned char,
7419 vector unsigned char,
7420 vector unsigned int);
7422 vector unsigned int vec_msums (vector unsigned short,
7423 vector unsigned short,
7424 vector unsigned int);
7425 vector signed int vec_msums (vector signed short,
7426 vector signed short,
7429 vector signed int vec_vmsumshs (vector signed short,
7430 vector signed short,
7433 vector unsigned int vec_vmsumuhs (vector unsigned short,
7434 vector unsigned short,
7435 vector unsigned int);
7437 void vec_mtvscr (vector signed int);
7438 void vec_mtvscr (vector unsigned int);
7439 void vec_mtvscr (vector bool int);
7440 void vec_mtvscr (vector signed short);
7441 void vec_mtvscr (vector unsigned short);
7442 void vec_mtvscr (vector bool short);
7443 void vec_mtvscr (vector pixel);
7444 void vec_mtvscr (vector signed char);
7445 void vec_mtvscr (vector unsigned char);
7446 void vec_mtvscr (vector bool char);
7448 vector unsigned short vec_mule (vector unsigned char,
7449 vector unsigned char);
7450 vector signed short vec_mule (vector signed char,
7451 vector signed char);
7452 vector unsigned int vec_mule (vector unsigned short,
7453 vector unsigned short);
7454 vector signed int vec_mule (vector signed short, vector signed short);
7456 vector signed int vec_vmulesh (vector signed short,
7457 vector signed short);
7459 vector unsigned int vec_vmuleuh (vector unsigned short,
7460 vector unsigned short);
7462 vector signed short vec_vmulesb (vector signed char,
7463 vector signed char);
7465 vector unsigned short vec_vmuleub (vector unsigned char,
7466 vector unsigned char);
7468 vector unsigned short vec_mulo (vector unsigned char,
7469 vector unsigned char);
7470 vector signed short vec_mulo (vector signed char, vector signed char);
7471 vector unsigned int vec_mulo (vector unsigned short,
7472 vector unsigned short);
7473 vector signed int vec_mulo (vector signed short, vector signed short);
7475 vector signed int vec_vmulosh (vector signed short,
7476 vector signed short);
7478 vector unsigned int vec_vmulouh (vector unsigned short,
7479 vector unsigned short);
7481 vector signed short vec_vmulosb (vector signed char,
7482 vector signed char);
7484 vector unsigned short vec_vmuloub (vector unsigned char,
7485 vector unsigned char);
7487 vector float vec_nmsub (vector float, vector float, vector float);
7489 vector float vec_nor (vector float, vector float);
7490 vector signed int vec_nor (vector signed int, vector signed int);
7491 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
7492 vector bool int vec_nor (vector bool int, vector bool int);
7493 vector signed short vec_nor (vector signed short, vector signed short);
7494 vector unsigned short vec_nor (vector unsigned short,
7495 vector unsigned short);
7496 vector bool short vec_nor (vector bool short, vector bool short);
7497 vector signed char vec_nor (vector signed char, vector signed char);
7498 vector unsigned char vec_nor (vector unsigned char,
7499 vector unsigned char);
7500 vector bool char vec_nor (vector bool char, vector bool char);
7502 vector float vec_or (vector float, vector float);
7503 vector float vec_or (vector float, vector bool int);
7504 vector float vec_or (vector bool int, vector float);
7505 vector bool int vec_or (vector bool int, vector bool int);
7506 vector signed int vec_or (vector bool int, vector signed int);
7507 vector signed int vec_or (vector signed int, vector bool int);
7508 vector signed int vec_or (vector signed int, vector signed int);
7509 vector unsigned int vec_or (vector bool int, vector unsigned int);
7510 vector unsigned int vec_or (vector unsigned int, vector bool int);
7511 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
7512 vector bool short vec_or (vector bool short, vector bool short);
7513 vector signed short vec_or (vector bool short, vector signed short);
7514 vector signed short vec_or (vector signed short, vector bool short);
7515 vector signed short vec_or (vector signed short, vector signed short);
7516 vector unsigned short vec_or (vector bool short, vector unsigned short);
7517 vector unsigned short vec_or (vector unsigned short, vector bool short);
7518 vector unsigned short vec_or (vector unsigned short,
7519 vector unsigned short);
7520 vector signed char vec_or (vector bool char, vector signed char);
7521 vector bool char vec_or (vector bool char, vector bool char);
7522 vector signed char vec_or (vector signed char, vector bool char);
7523 vector signed char vec_or (vector signed char, vector signed char);
7524 vector unsigned char vec_or (vector bool char, vector unsigned char);
7525 vector unsigned char vec_or (vector unsigned char, vector bool char);
7526 vector unsigned char vec_or (vector unsigned char,
7527 vector unsigned char);
7529 vector signed char vec_pack (vector signed short, vector signed short);
7530 vector unsigned char vec_pack (vector unsigned short,
7531 vector unsigned short);
7532 vector bool char vec_pack (vector bool short, vector bool short);
7533 vector signed short vec_pack (vector signed int, vector signed int);
7534 vector unsigned short vec_pack (vector unsigned int,
7535 vector unsigned int);
7536 vector bool short vec_pack (vector bool int, vector bool int);
7538 vector bool short vec_vpkuwum (vector bool int, vector bool int);
7539 vector signed short vec_vpkuwum (vector signed int, vector signed int);
7540 vector unsigned short vec_vpkuwum (vector unsigned int,
7541 vector unsigned int);
7543 vector bool char vec_vpkuhum (vector bool short, vector bool short);
7544 vector signed char vec_vpkuhum (vector signed short,
7545 vector signed short);
7546 vector unsigned char vec_vpkuhum (vector unsigned short,
7547 vector unsigned short);
7549 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
7551 vector unsigned char vec_packs (vector unsigned short,
7552 vector unsigned short);
7553 vector signed char vec_packs (vector signed short, vector signed short);
7554 vector unsigned short vec_packs (vector unsigned int,
7555 vector unsigned int);
7556 vector signed short vec_packs (vector signed int, vector signed int);
7558 vector signed short vec_vpkswss (vector signed int, vector signed int);
7560 vector unsigned short vec_vpkuwus (vector unsigned int,
7561 vector unsigned int);
7563 vector signed char vec_vpkshss (vector signed short,
7564 vector signed short);
7566 vector unsigned char vec_vpkuhus (vector unsigned short,
7567 vector unsigned short);
7569 vector unsigned char vec_packsu (vector unsigned short,
7570 vector unsigned short);
7571 vector unsigned char vec_packsu (vector signed short,
7572 vector signed short);
7573 vector unsigned short vec_packsu (vector unsigned int,
7574 vector unsigned int);
7575 vector unsigned short vec_packsu (vector signed int, vector signed int);
7577 vector unsigned short vec_vpkswus (vector signed int,
7580 vector unsigned char vec_vpkshus (vector signed short,
7581 vector signed short);
7583 vector float vec_perm (vector float,
7585 vector unsigned char);
7586 vector signed int vec_perm (vector signed int,
7588 vector unsigned char);
7589 vector unsigned int vec_perm (vector unsigned int,
7590 vector unsigned int,
7591 vector unsigned char);
7592 vector bool int vec_perm (vector bool int,
7594 vector unsigned char);
7595 vector signed short vec_perm (vector signed short,
7596 vector signed short,
7597 vector unsigned char);
7598 vector unsigned short vec_perm (vector unsigned short,
7599 vector unsigned short,
7600 vector unsigned char);
7601 vector bool short vec_perm (vector bool short,
7603 vector unsigned char);
7604 vector pixel vec_perm (vector pixel,
7606 vector unsigned char);
7607 vector signed char vec_perm (vector signed char,
7609 vector unsigned char);
7610 vector unsigned char vec_perm (vector unsigned char,
7611 vector unsigned char,
7612 vector unsigned char);
7613 vector bool char vec_perm (vector bool char,
7615 vector unsigned char);
7617 vector float vec_re (vector float);
7619 vector signed char vec_rl (vector signed char,
7620 vector unsigned char);
7621 vector unsigned char vec_rl (vector unsigned char,
7622 vector unsigned char);
7623 vector signed short vec_rl (vector signed short, vector unsigned short);
7624 vector unsigned short vec_rl (vector unsigned short,
7625 vector unsigned short);
7626 vector signed int vec_rl (vector signed int, vector unsigned int);
7627 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
7629 vector signed int vec_vrlw (vector signed int, vector unsigned int);
7630 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
7632 vector signed short vec_vrlh (vector signed short,
7633 vector unsigned short);
7634 vector unsigned short vec_vrlh (vector unsigned short,
7635 vector unsigned short);
7637 vector signed char vec_vrlb (vector signed char, vector unsigned char);
7638 vector unsigned char vec_vrlb (vector unsigned char,
7639 vector unsigned char);
7641 vector float vec_round (vector float);
7643 vector float vec_rsqrte (vector float);
7645 vector float vec_sel (vector float, vector float, vector bool int);
7646 vector float vec_sel (vector float, vector float, vector unsigned int);
7647 vector signed int vec_sel (vector signed int,
7650 vector signed int vec_sel (vector signed int,
7652 vector unsigned int);
7653 vector unsigned int vec_sel (vector unsigned int,
7654 vector unsigned int,
7656 vector unsigned int vec_sel (vector unsigned int,
7657 vector unsigned int,
7658 vector unsigned int);
7659 vector bool int vec_sel (vector bool int,
7662 vector bool int vec_sel (vector bool int,
7664 vector unsigned int);
7665 vector signed short vec_sel (vector signed short,
7666 vector signed short,
7668 vector signed short vec_sel (vector signed short,
7669 vector signed short,
7670 vector unsigned short);
7671 vector unsigned short vec_sel (vector unsigned short,
7672 vector unsigned short,
7674 vector unsigned short vec_sel (vector unsigned short,
7675 vector unsigned short,
7676 vector unsigned short);
7677 vector bool short vec_sel (vector bool short,
7680 vector bool short vec_sel (vector bool short,
7682 vector unsigned short);
7683 vector signed char vec_sel (vector signed char,
7686 vector signed char vec_sel (vector signed char,
7688 vector unsigned char);
7689 vector unsigned char vec_sel (vector unsigned char,
7690 vector unsigned char,
7692 vector unsigned char vec_sel (vector unsigned char,
7693 vector unsigned char,
7694 vector unsigned char);
7695 vector bool char vec_sel (vector bool char,
7698 vector bool char vec_sel (vector bool char,
7700 vector unsigned char);
7702 vector signed char vec_sl (vector signed char,
7703 vector unsigned char);
7704 vector unsigned char vec_sl (vector unsigned char,
7705 vector unsigned char);
7706 vector signed short vec_sl (vector signed short, vector unsigned short);
7707 vector unsigned short vec_sl (vector unsigned short,
7708 vector unsigned short);
7709 vector signed int vec_sl (vector signed int, vector unsigned int);
7710 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
7712 vector signed int vec_vslw (vector signed int, vector unsigned int);
7713 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
7715 vector signed short vec_vslh (vector signed short,
7716 vector unsigned short);
7717 vector unsigned short vec_vslh (vector unsigned short,
7718 vector unsigned short);
7720 vector signed char vec_vslb (vector signed char, vector unsigned char);
7721 vector unsigned char vec_vslb (vector unsigned char,
7722 vector unsigned char);
7724 vector float vec_sld (vector float, vector float, const int);
7725 vector signed int vec_sld (vector signed int,
7728 vector unsigned int vec_sld (vector unsigned int,
7729 vector unsigned int,
7731 vector bool int vec_sld (vector bool int,
7734 vector signed short vec_sld (vector signed short,
7735 vector signed short,
7737 vector unsigned short vec_sld (vector unsigned short,
7738 vector unsigned short,
7740 vector bool short vec_sld (vector bool short,
7743 vector pixel vec_sld (vector pixel,
7746 vector signed char vec_sld (vector signed char,
7749 vector unsigned char vec_sld (vector unsigned char,
7750 vector unsigned char,
7752 vector bool char vec_sld (vector bool char,
7756 vector signed int vec_sll (vector signed int,
7757 vector unsigned int);
7758 vector signed int vec_sll (vector signed int,
7759 vector unsigned short);
7760 vector signed int vec_sll (vector signed int,
7761 vector unsigned char);
7762 vector unsigned int vec_sll (vector unsigned int,
7763 vector unsigned int);
7764 vector unsigned int vec_sll (vector unsigned int,
7765 vector unsigned short);
7766 vector unsigned int vec_sll (vector unsigned int,
7767 vector unsigned char);
7768 vector bool int vec_sll (vector bool int,
7769 vector unsigned int);
7770 vector bool int vec_sll (vector bool int,
7771 vector unsigned short);
7772 vector bool int vec_sll (vector bool int,
7773 vector unsigned char);
7774 vector signed short vec_sll (vector signed short,
7775 vector unsigned int);
7776 vector signed short vec_sll (vector signed short,
7777 vector unsigned short);
7778 vector signed short vec_sll (vector signed short,
7779 vector unsigned char);
7780 vector unsigned short vec_sll (vector unsigned short,
7781 vector unsigned int);
7782 vector unsigned short vec_sll (vector unsigned short,
7783 vector unsigned short);
7784 vector unsigned short vec_sll (vector unsigned short,
7785 vector unsigned char);
7786 vector bool short vec_sll (vector bool short, vector unsigned int);
7787 vector bool short vec_sll (vector bool short, vector unsigned short);
7788 vector bool short vec_sll (vector bool short, vector unsigned char);
7789 vector pixel vec_sll (vector pixel, vector unsigned int);
7790 vector pixel vec_sll (vector pixel, vector unsigned short);
7791 vector pixel vec_sll (vector pixel, vector unsigned char);
7792 vector signed char vec_sll (vector signed char, vector unsigned int);
7793 vector signed char vec_sll (vector signed char, vector unsigned short);
7794 vector signed char vec_sll (vector signed char, vector unsigned char);
7795 vector unsigned char vec_sll (vector unsigned char,
7796 vector unsigned int);
7797 vector unsigned char vec_sll (vector unsigned char,
7798 vector unsigned short);
7799 vector unsigned char vec_sll (vector unsigned char,
7800 vector unsigned char);
7801 vector bool char vec_sll (vector bool char, vector unsigned int);
7802 vector bool char vec_sll (vector bool char, vector unsigned short);
7803 vector bool char vec_sll (vector bool char, vector unsigned char);
7805 vector float vec_slo (vector float, vector signed char);
7806 vector float vec_slo (vector float, vector unsigned char);
7807 vector signed int vec_slo (vector signed int, vector signed char);
7808 vector signed int vec_slo (vector signed int, vector unsigned char);
7809 vector unsigned int vec_slo (vector unsigned int, vector signed char);
7810 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
7811 vector signed short vec_slo (vector signed short, vector signed char);
7812 vector signed short vec_slo (vector signed short, vector unsigned char);
7813 vector unsigned short vec_slo (vector unsigned short,
7814 vector signed char);
7815 vector unsigned short vec_slo (vector unsigned short,
7816 vector unsigned char);
7817 vector pixel vec_slo (vector pixel, vector signed char);
7818 vector pixel vec_slo (vector pixel, vector unsigned char);
7819 vector signed char vec_slo (vector signed char, vector signed char);
7820 vector signed char vec_slo (vector signed char, vector unsigned char);
7821 vector unsigned char vec_slo (vector unsigned char, vector signed char);
7822 vector unsigned char vec_slo (vector unsigned char,
7823 vector unsigned char);
7825 vector signed char vec_splat (vector signed char, const int);
7826 vector unsigned char vec_splat (vector unsigned char, const int);
7827 vector bool char vec_splat (vector bool char, const int);
7828 vector signed short vec_splat (vector signed short, const int);
7829 vector unsigned short vec_splat (vector unsigned short, const int);
7830 vector bool short vec_splat (vector bool short, const int);
7831 vector pixel vec_splat (vector pixel, const int);
7832 vector float vec_splat (vector float, const int);
7833 vector signed int vec_splat (vector signed int, const int);
7834 vector unsigned int vec_splat (vector unsigned int, const int);
7835 vector bool int vec_splat (vector bool int, const int);
7837 vector float vec_vspltw (vector float, const int);
7838 vector signed int vec_vspltw (vector signed int, const int);
7839 vector unsigned int vec_vspltw (vector unsigned int, const int);
7840 vector bool int vec_vspltw (vector bool int, const int);
7842 vector bool short vec_vsplth (vector bool short, const int);
7843 vector signed short vec_vsplth (vector signed short, const int);
7844 vector unsigned short vec_vsplth (vector unsigned short, const int);
7845 vector pixel vec_vsplth (vector pixel, const int);
7847 vector signed char vec_vspltb (vector signed char, const int);
7848 vector unsigned char vec_vspltb (vector unsigned char, const int);
7849 vector bool char vec_vspltb (vector bool char, const int);
7851 vector signed char vec_splat_s8 (const int);
7853 vector signed short vec_splat_s16 (const int);
7855 vector signed int vec_splat_s32 (const int);
7857 vector unsigned char vec_splat_u8 (const int);
7859 vector unsigned short vec_splat_u16 (const int);
7861 vector unsigned int vec_splat_u32 (const int);
7863 vector signed char vec_sr (vector signed char, vector unsigned char);
7864 vector unsigned char vec_sr (vector unsigned char,
7865 vector unsigned char);
7866 vector signed short vec_sr (vector signed short,
7867 vector unsigned short);
7868 vector unsigned short vec_sr (vector unsigned short,
7869 vector unsigned short);
7870 vector signed int vec_sr (vector signed int, vector unsigned int);
7871 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
7873 vector signed int vec_vsrw (vector signed int, vector unsigned int);
7874 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
7876 vector signed short vec_vsrh (vector signed short,
7877 vector unsigned short);
7878 vector unsigned short vec_vsrh (vector unsigned short,
7879 vector unsigned short);
7881 vector signed char vec_vsrb (vector signed char, vector unsigned char);
7882 vector unsigned char vec_vsrb (vector unsigned char,
7883 vector unsigned char);
7885 vector signed char vec_sra (vector signed char, vector unsigned char);
7886 vector unsigned char vec_sra (vector unsigned char,
7887 vector unsigned char);
7888 vector signed short vec_sra (vector signed short,
7889 vector unsigned short);
7890 vector unsigned short vec_sra (vector unsigned short,
7891 vector unsigned short);
7892 vector signed int vec_sra (vector signed int, vector unsigned int);
7893 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
7895 vector signed int vec_vsraw (vector signed int, vector unsigned int);
7896 vector unsigned int vec_vsraw (vector unsigned int,
7897 vector unsigned int);
7899 vector signed short vec_vsrah (vector signed short,
7900 vector unsigned short);
7901 vector unsigned short vec_vsrah (vector unsigned short,
7902 vector unsigned short);
7904 vector signed char vec_vsrab (vector signed char, vector unsigned char);
7905 vector unsigned char vec_vsrab (vector unsigned char,
7906 vector unsigned char);
7908 vector signed int vec_srl (vector signed int, vector unsigned int);
7909 vector signed int vec_srl (vector signed int, vector unsigned short);
7910 vector signed int vec_srl (vector signed int, vector unsigned char);
7911 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
7912 vector unsigned int vec_srl (vector unsigned int,
7913 vector unsigned short);
7914 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
7915 vector bool int vec_srl (vector bool int, vector unsigned int);
7916 vector bool int vec_srl (vector bool int, vector unsigned short);
7917 vector bool int vec_srl (vector bool int, vector unsigned char);
7918 vector signed short vec_srl (vector signed short, vector unsigned int);
7919 vector signed short vec_srl (vector signed short,
7920 vector unsigned short);
7921 vector signed short vec_srl (vector signed short, vector unsigned char);
7922 vector unsigned short vec_srl (vector unsigned short,
7923 vector unsigned int);
7924 vector unsigned short vec_srl (vector unsigned short,
7925 vector unsigned short);
7926 vector unsigned short vec_srl (vector unsigned short,
7927 vector unsigned char);
7928 vector bool short vec_srl (vector bool short, vector unsigned int);
7929 vector bool short vec_srl (vector bool short, vector unsigned short);
7930 vector bool short vec_srl (vector bool short, vector unsigned char);
7931 vector pixel vec_srl (vector pixel, vector unsigned int);
7932 vector pixel vec_srl (vector pixel, vector unsigned short);
7933 vector pixel vec_srl (vector pixel, vector unsigned char);
7934 vector signed char vec_srl (vector signed char, vector unsigned int);
7935 vector signed char vec_srl (vector signed char, vector unsigned short);
7936 vector signed char vec_srl (vector signed char, vector unsigned char);
7937 vector unsigned char vec_srl (vector unsigned char,
7938 vector unsigned int);
7939 vector unsigned char vec_srl (vector unsigned char,
7940 vector unsigned short);
7941 vector unsigned char vec_srl (vector unsigned char,
7942 vector unsigned char);
7943 vector bool char vec_srl (vector bool char, vector unsigned int);
7944 vector bool char vec_srl (vector bool char, vector unsigned short);
7945 vector bool char vec_srl (vector bool char, vector unsigned char);
7947 vector float vec_sro (vector float, vector signed char);
7948 vector float vec_sro (vector float, vector unsigned char);
7949 vector signed int vec_sro (vector signed int, vector signed char);
7950 vector signed int vec_sro (vector signed int, vector unsigned char);
7951 vector unsigned int vec_sro (vector unsigned int, vector signed char);
7952 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
7953 vector signed short vec_sro (vector signed short, vector signed char);
7954 vector signed short vec_sro (vector signed short, vector unsigned char);
7955 vector unsigned short vec_sro (vector unsigned short,
7956 vector signed char);
7957 vector unsigned short vec_sro (vector unsigned short,
7958 vector unsigned char);
7959 vector pixel vec_sro (vector pixel, vector signed char);
7960 vector pixel vec_sro (vector pixel, vector unsigned char);
7961 vector signed char vec_sro (vector signed char, vector signed char);
7962 vector signed char vec_sro (vector signed char, vector unsigned char);
7963 vector unsigned char vec_sro (vector unsigned char, vector signed char);
7964 vector unsigned char vec_sro (vector unsigned char,
7965 vector unsigned char);
7967 void vec_st (vector float, int, vector float *);
7968 void vec_st (vector float, int, float *);
7969 void vec_st (vector signed int, int, vector signed int *);
7970 void vec_st (vector signed int, int, int *);
7971 void vec_st (vector unsigned int, int, vector unsigned int *);
7972 void vec_st (vector unsigned int, int, unsigned int *);
7973 void vec_st (vector bool int, int, vector bool int *);
7974 void vec_st (vector bool int, int, unsigned int *);
7975 void vec_st (vector bool int, int, int *);
7976 void vec_st (vector signed short, int, vector signed short *);
7977 void vec_st (vector signed short, int, short *);
7978 void vec_st (vector unsigned short, int, vector unsigned short *);
7979 void vec_st (vector unsigned short, int, unsigned short *);
7980 void vec_st (vector bool short, int, vector bool short *);
7981 void vec_st (vector bool short, int, unsigned short *);
7982 void vec_st (vector pixel, int, vector pixel *);
7983 void vec_st (vector pixel, int, unsigned short *);
7984 void vec_st (vector pixel, int, short *);
7985 void vec_st (vector bool short, int, short *);
7986 void vec_st (vector signed char, int, vector signed char *);
7987 void vec_st (vector signed char, int, signed char *);
7988 void vec_st (vector unsigned char, int, vector unsigned char *);
7989 void vec_st (vector unsigned char, int, unsigned char *);
7990 void vec_st (vector bool char, int, vector bool char *);
7991 void vec_st (vector bool char, int, unsigned char *);
7992 void vec_st (vector bool char, int, signed char *);
7994 void vec_ste (vector signed char, int, signed char *);
7995 void vec_ste (vector unsigned char, int, unsigned char *);
7996 void vec_ste (vector bool char, int, signed char *);
7997 void vec_ste (vector bool char, int, unsigned char *);
7998 void vec_ste (vector signed short, int, short *);
7999 void vec_ste (vector unsigned short, int, unsigned short *);
8000 void vec_ste (vector bool short, int, short *);
8001 void vec_ste (vector bool short, int, unsigned short *);
8002 void vec_ste (vector pixel, int, short *);
8003 void vec_ste (vector pixel, int, unsigned short *);
8004 void vec_ste (vector float, int, float *);
8005 void vec_ste (vector signed int, int, int *);
8006 void vec_ste (vector unsigned int, int, unsigned int *);
8007 void vec_ste (vector bool int, int, int *);
8008 void vec_ste (vector bool int, int, unsigned int *);
8010 void vec_stvewx (vector float, int, float *);
8011 void vec_stvewx (vector signed int, int, int *);
8012 void vec_stvewx (vector unsigned int, int, unsigned int *);
8013 void vec_stvewx (vector bool int, int, int *);
8014 void vec_stvewx (vector bool int, int, unsigned int *);
8016 void vec_stvehx (vector signed short, int, short *);
8017 void vec_stvehx (vector unsigned short, int, unsigned short *);
8018 void vec_stvehx (vector bool short, int, short *);
8019 void vec_stvehx (vector bool short, int, unsigned short *);
8020 void vec_stvehx (vector pixel, int, short *);
8021 void vec_stvehx (vector pixel, int, unsigned short *);
8023 void vec_stvebx (vector signed char, int, signed char *);
8024 void vec_stvebx (vector unsigned char, int, unsigned char *);
8025 void vec_stvebx (vector bool char, int, signed char *);
8026 void vec_stvebx (vector bool char, int, unsigned char *);
8028 void vec_stl (vector float, int, vector float *);
8029 void vec_stl (vector float, int, float *);
8030 void vec_stl (vector signed int, int, vector signed int *);
8031 void vec_stl (vector signed int, int, int *);
8032 void vec_stl (vector unsigned int, int, vector unsigned int *);
8033 void vec_stl (vector unsigned int, int, unsigned int *);
8034 void vec_stl (vector bool int, int, vector bool int *);
8035 void vec_stl (vector bool int, int, unsigned int *);
8036 void vec_stl (vector bool int, int, int *);
8037 void vec_stl (vector signed short, int, vector signed short *);
8038 void vec_stl (vector signed short, int, short *);
8039 void vec_stl (vector unsigned short, int, vector unsigned short *);
8040 void vec_stl (vector unsigned short, int, unsigned short *);
8041 void vec_stl (vector bool short, int, vector bool short *);
8042 void vec_stl (vector bool short, int, unsigned short *);
8043 void vec_stl (vector bool short, int, short *);
8044 void vec_stl (vector pixel, int, vector pixel *);
8045 void vec_stl (vector pixel, int, unsigned short *);
8046 void vec_stl (vector pixel, int, short *);
8047 void vec_stl (vector signed char, int, vector signed char *);
8048 void vec_stl (vector signed char, int, signed char *);
8049 void vec_stl (vector unsigned char, int, vector unsigned char *);
8050 void vec_stl (vector unsigned char, int, unsigned char *);
8051 void vec_stl (vector bool char, int, vector bool char *);
8052 void vec_stl (vector bool char, int, unsigned char *);
8053 void vec_stl (vector bool char, int, signed char *);
8055 vector signed char vec_sub (vector bool char, vector signed char);
8056 vector signed char vec_sub (vector signed char, vector bool char);
8057 vector signed char vec_sub (vector signed char, vector signed char);
8058 vector unsigned char vec_sub (vector bool char, vector unsigned char);
8059 vector unsigned char vec_sub (vector unsigned char, vector bool char);
8060 vector unsigned char vec_sub (vector unsigned char,
8061 vector unsigned char);
8062 vector signed short vec_sub (vector bool short, vector signed short);
8063 vector signed short vec_sub (vector signed short, vector bool short);
8064 vector signed short vec_sub (vector signed short, vector signed short);
8065 vector unsigned short vec_sub (vector bool short,
8066 vector unsigned short);
8067 vector unsigned short vec_sub (vector unsigned short,
8069 vector unsigned short vec_sub (vector unsigned short,
8070 vector unsigned short);
8071 vector signed int vec_sub (vector bool int, vector signed int);
8072 vector signed int vec_sub (vector signed int, vector bool int);
8073 vector signed int vec_sub (vector signed int, vector signed int);
8074 vector unsigned int vec_sub (vector bool int, vector unsigned int);
8075 vector unsigned int vec_sub (vector unsigned int, vector bool int);
8076 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
8077 vector float vec_sub (vector float, vector float);
8079 vector float vec_vsubfp (vector float, vector float);
8081 vector signed int vec_vsubuwm (vector bool int, vector signed int);
8082 vector signed int vec_vsubuwm (vector signed int, vector bool int);
8083 vector signed int vec_vsubuwm (vector signed int, vector signed int);
8084 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
8085 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
8086 vector unsigned int vec_vsubuwm (vector unsigned int,
8087 vector unsigned int);
8089 vector signed short vec_vsubuhm (vector bool short,
8090 vector signed short);
8091 vector signed short vec_vsubuhm (vector signed short,
8093 vector signed short vec_vsubuhm (vector signed short,
8094 vector signed short);
8095 vector unsigned short vec_vsubuhm (vector bool short,
8096 vector unsigned short);
8097 vector unsigned short vec_vsubuhm (vector unsigned short,
8099 vector unsigned short vec_vsubuhm (vector unsigned short,
8100 vector unsigned short);
8102 vector signed char vec_vsububm (vector bool char, vector signed char);
8103 vector signed char vec_vsububm (vector signed char, vector bool char);
8104 vector signed char vec_vsububm (vector signed char, vector signed char);
8105 vector unsigned char vec_vsububm (vector bool char,
8106 vector unsigned char);
8107 vector unsigned char vec_vsububm (vector unsigned char,
8109 vector unsigned char vec_vsububm (vector unsigned char,
8110 vector unsigned char);
8112 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
8114 vector unsigned char vec_subs (vector bool char, vector unsigned char);
8115 vector unsigned char vec_subs (vector unsigned char, vector bool char);
8116 vector unsigned char vec_subs (vector unsigned char,
8117 vector unsigned char);
8118 vector signed char vec_subs (vector bool char, vector signed char);
8119 vector signed char vec_subs (vector signed char, vector bool char);
8120 vector signed char vec_subs (vector signed char, vector signed char);
8121 vector unsigned short vec_subs (vector bool short,
8122 vector unsigned short);
8123 vector unsigned short vec_subs (vector unsigned short,
8125 vector unsigned short vec_subs (vector unsigned short,
8126 vector unsigned short);
8127 vector signed short vec_subs (vector bool short, vector signed short);
8128 vector signed short vec_subs (vector signed short, vector bool short);
8129 vector signed short vec_subs (vector signed short, vector signed short);
8130 vector unsigned int vec_subs (vector bool int, vector unsigned int);
8131 vector unsigned int vec_subs (vector unsigned int, vector bool int);
8132 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
8133 vector signed int vec_subs (vector bool int, vector signed int);
8134 vector signed int vec_subs (vector signed int, vector bool int);
8135 vector signed int vec_subs (vector signed int, vector signed int);
8137 vector signed int vec_vsubsws (vector bool int, vector signed int);
8138 vector signed int vec_vsubsws (vector signed int, vector bool int);
8139 vector signed int vec_vsubsws (vector signed int, vector signed int);
8141 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
8142 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
8143 vector unsigned int vec_vsubuws (vector unsigned int,
8144 vector unsigned int);
8146 vector signed short vec_vsubshs (vector bool short,
8147 vector signed short);
8148 vector signed short vec_vsubshs (vector signed short,
8150 vector signed short vec_vsubshs (vector signed short,
8151 vector signed short);
8153 vector unsigned short vec_vsubuhs (vector bool short,
8154 vector unsigned short);
8155 vector unsigned short vec_vsubuhs (vector unsigned short,
8157 vector unsigned short vec_vsubuhs (vector unsigned short,
8158 vector unsigned short);
8160 vector signed char vec_vsubsbs (vector bool char, vector signed char);
8161 vector signed char vec_vsubsbs (vector signed char, vector bool char);
8162 vector signed char vec_vsubsbs (vector signed char, vector signed char);
8164 vector unsigned char vec_vsububs (vector bool char,
8165 vector unsigned char);
8166 vector unsigned char vec_vsububs (vector unsigned char,
8168 vector unsigned char vec_vsububs (vector unsigned char,
8169 vector unsigned char);
8171 vector unsigned int vec_sum4s (vector unsigned char,
8172 vector unsigned int);
8173 vector signed int vec_sum4s (vector signed char, vector signed int);
8174 vector signed int vec_sum4s (vector signed short, vector signed int);
8176 vector signed int vec_vsum4shs (vector signed short, vector signed int);
8178 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
8180 vector unsigned int vec_vsum4ubs (vector unsigned char,
8181 vector unsigned int);
8183 vector signed int vec_sum2s (vector signed int, vector signed int);
8185 vector signed int vec_sums (vector signed int, vector signed int);
8187 vector float vec_trunc (vector float);
8189 vector signed short vec_unpackh (vector signed char);
8190 vector bool short vec_unpackh (vector bool char);
8191 vector signed int vec_unpackh (vector signed short);
8192 vector bool int vec_unpackh (vector bool short);
8193 vector unsigned int vec_unpackh (vector pixel);
8195 vector bool int vec_vupkhsh (vector bool short);
8196 vector signed int vec_vupkhsh (vector signed short);
8198 vector unsigned int vec_vupkhpx (vector pixel);
8200 vector bool short vec_vupkhsb (vector bool char);
8201 vector signed short vec_vupkhsb (vector signed char);
8203 vector signed short vec_unpackl (vector signed char);
8204 vector bool short vec_unpackl (vector bool char);
8205 vector unsigned int vec_unpackl (vector pixel);
8206 vector signed int vec_unpackl (vector signed short);
8207 vector bool int vec_unpackl (vector bool short);
8209 vector unsigned int vec_vupklpx (vector pixel);
8211 vector bool int vec_vupklsh (vector bool short);
8212 vector signed int vec_vupklsh (vector signed short);
8214 vector bool short vec_vupklsb (vector bool char);
8215 vector signed short vec_vupklsb (vector signed char);
8217 vector float vec_xor (vector float, vector float);
8218 vector float vec_xor (vector float, vector bool int);
8219 vector float vec_xor (vector bool int, vector float);
8220 vector bool int vec_xor (vector bool int, vector bool int);
8221 vector signed int vec_xor (vector bool int, vector signed int);
8222 vector signed int vec_xor (vector signed int, vector bool int);
8223 vector signed int vec_xor (vector signed int, vector signed int);
8224 vector unsigned int vec_xor (vector bool int, vector unsigned int);
8225 vector unsigned int vec_xor (vector unsigned int, vector bool int);
8226 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
8227 vector bool short vec_xor (vector bool short, vector bool short);
8228 vector signed short vec_xor (vector bool short, vector signed short);
8229 vector signed short vec_xor (vector signed short, vector bool short);
8230 vector signed short vec_xor (vector signed short, vector signed short);
8231 vector unsigned short vec_xor (vector bool short,
8232 vector unsigned short);
8233 vector unsigned short vec_xor (vector unsigned short,
8235 vector unsigned short vec_xor (vector unsigned short,
8236 vector unsigned short);
8237 vector signed char vec_xor (vector bool char, vector signed char);
8238 vector bool char vec_xor (vector bool char, vector bool char);
8239 vector signed char vec_xor (vector signed char, vector bool char);
8240 vector signed char vec_xor (vector signed char, vector signed char);
8241 vector unsigned char vec_xor (vector bool char, vector unsigned char);
8242 vector unsigned char vec_xor (vector unsigned char, vector bool char);
8243 vector unsigned char vec_xor (vector unsigned char,
8244 vector unsigned char);
8246 int vec_all_eq (vector signed char, vector bool char);
8247 int vec_all_eq (vector signed char, vector signed char);
8248 int vec_all_eq (vector unsigned char, vector bool char);
8249 int vec_all_eq (vector unsigned char, vector unsigned char);
8250 int vec_all_eq (vector bool char, vector bool char);
8251 int vec_all_eq (vector bool char, vector unsigned char);
8252 int vec_all_eq (vector bool char, vector signed char);
8253 int vec_all_eq (vector signed short, vector bool short);
8254 int vec_all_eq (vector signed short, vector signed short);
8255 int vec_all_eq (vector unsigned short, vector bool short);
8256 int vec_all_eq (vector unsigned short, vector unsigned short);
8257 int vec_all_eq (vector bool short, vector bool short);
8258 int vec_all_eq (vector bool short, vector unsigned short);
8259 int vec_all_eq (vector bool short, vector signed short);
8260 int vec_all_eq (vector pixel, vector pixel);
8261 int vec_all_eq (vector signed int, vector bool int);
8262 int vec_all_eq (vector signed int, vector signed int);
8263 int vec_all_eq (vector unsigned int, vector bool int);
8264 int vec_all_eq (vector unsigned int, vector unsigned int);
8265 int vec_all_eq (vector bool int, vector bool int);
8266 int vec_all_eq (vector bool int, vector unsigned int);
8267 int vec_all_eq (vector bool int, vector signed int);
8268 int vec_all_eq (vector float, vector float);
8270 int vec_all_ge (vector bool char, vector unsigned char);
8271 int vec_all_ge (vector unsigned char, vector bool char);
8272 int vec_all_ge (vector unsigned char, vector unsigned char);
8273 int vec_all_ge (vector bool char, vector signed char);
8274 int vec_all_ge (vector signed char, vector bool char);
8275 int vec_all_ge (vector signed char, vector signed char);
8276 int vec_all_ge (vector bool short, vector unsigned short);
8277 int vec_all_ge (vector unsigned short, vector bool short);
8278 int vec_all_ge (vector unsigned short, vector unsigned short);
8279 int vec_all_ge (vector signed short, vector signed short);
8280 int vec_all_ge (vector bool short, vector signed short);
8281 int vec_all_ge (vector signed short, vector bool short);
8282 int vec_all_ge (vector bool int, vector unsigned int);
8283 int vec_all_ge (vector unsigned int, vector bool int);
8284 int vec_all_ge (vector unsigned int, vector unsigned int);
8285 int vec_all_ge (vector bool int, vector signed int);
8286 int vec_all_ge (vector signed int, vector bool int);
8287 int vec_all_ge (vector signed int, vector signed int);
8288 int vec_all_ge (vector float, vector float);
8290 int vec_all_gt (vector bool char, vector unsigned char);
8291 int vec_all_gt (vector unsigned char, vector bool char);
8292 int vec_all_gt (vector unsigned char, vector unsigned char);
8293 int vec_all_gt (vector bool char, vector signed char);
8294 int vec_all_gt (vector signed char, vector bool char);
8295 int vec_all_gt (vector signed char, vector signed char);
8296 int vec_all_gt (vector bool short, vector unsigned short);
8297 int vec_all_gt (vector unsigned short, vector bool short);
8298 int vec_all_gt (vector unsigned short, vector unsigned short);
8299 int vec_all_gt (vector bool short, vector signed short);
8300 int vec_all_gt (vector signed short, vector bool short);
8301 int vec_all_gt (vector signed short, vector signed short);
8302 int vec_all_gt (vector bool int, vector unsigned int);
8303 int vec_all_gt (vector unsigned int, vector bool int);
8304 int vec_all_gt (vector unsigned int, vector unsigned int);
8305 int vec_all_gt (vector bool int, vector signed int);
8306 int vec_all_gt (vector signed int, vector bool int);
8307 int vec_all_gt (vector signed int, vector signed int);
8308 int vec_all_gt (vector float, vector float);
8310 int vec_all_in (vector float, vector float);
8312 int vec_all_le (vector bool char, vector unsigned char);
8313 int vec_all_le (vector unsigned char, vector bool char);
8314 int vec_all_le (vector unsigned char, vector unsigned char);
8315 int vec_all_le (vector bool char, vector signed char);
8316 int vec_all_le (vector signed char, vector bool char);
8317 int vec_all_le (vector signed char, vector signed char);
8318 int vec_all_le (vector bool short, vector unsigned short);
8319 int vec_all_le (vector unsigned short, vector bool short);
8320 int vec_all_le (vector unsigned short, vector unsigned short);
8321 int vec_all_le (vector bool short, vector signed short);
8322 int vec_all_le (vector signed short, vector bool short);
8323 int vec_all_le (vector signed short, vector signed short);
8324 int vec_all_le (vector bool int, vector unsigned int);
8325 int vec_all_le (vector unsigned int, vector bool int);
8326 int vec_all_le (vector unsigned int, vector unsigned int);
8327 int vec_all_le (vector bool int, vector signed int);
8328 int vec_all_le (vector signed int, vector bool int);
8329 int vec_all_le (vector signed int, vector signed int);
8330 int vec_all_le (vector float, vector float);
8332 int vec_all_lt (vector bool char, vector unsigned char);
8333 int vec_all_lt (vector unsigned char, vector bool char);
8334 int vec_all_lt (vector unsigned char, vector unsigned char);
8335 int vec_all_lt (vector bool char, vector signed char);
8336 int vec_all_lt (vector signed char, vector bool char);
8337 int vec_all_lt (vector signed char, vector signed char);
8338 int vec_all_lt (vector bool short, vector unsigned short);
8339 int vec_all_lt (vector unsigned short, vector bool short);
8340 int vec_all_lt (vector unsigned short, vector unsigned short);
8341 int vec_all_lt (vector bool short, vector signed short);
8342 int vec_all_lt (vector signed short, vector bool short);
8343 int vec_all_lt (vector signed short, vector signed short);
8344 int vec_all_lt (vector bool int, vector unsigned int);
8345 int vec_all_lt (vector unsigned int, vector bool int);
8346 int vec_all_lt (vector unsigned int, vector unsigned int);
8347 int vec_all_lt (vector bool int, vector signed int);
8348 int vec_all_lt (vector signed int, vector bool int);
8349 int vec_all_lt (vector signed int, vector signed int);
8350 int vec_all_lt (vector float, vector float);
8352 int vec_all_nan (vector float);
8354 int vec_all_ne (vector signed char, vector bool char);
8355 int vec_all_ne (vector signed char, vector signed char);
8356 int vec_all_ne (vector unsigned char, vector bool char);
8357 int vec_all_ne (vector unsigned char, vector unsigned char);
8358 int vec_all_ne (vector bool char, vector bool char);
8359 int vec_all_ne (vector bool char, vector unsigned char);
8360 int vec_all_ne (vector bool char, vector signed char);
8361 int vec_all_ne (vector signed short, vector bool short);
8362 int vec_all_ne (vector signed short, vector signed short);
8363 int vec_all_ne (vector unsigned short, vector bool short);
8364 int vec_all_ne (vector unsigned short, vector unsigned short);
8365 int vec_all_ne (vector bool short, vector bool short);
8366 int vec_all_ne (vector bool short, vector unsigned short);
8367 int vec_all_ne (vector bool short, vector signed short);
8368 int vec_all_ne (vector pixel, vector pixel);
8369 int vec_all_ne (vector signed int, vector bool int);
8370 int vec_all_ne (vector signed int, vector signed int);
8371 int vec_all_ne (vector unsigned int, vector bool int);
8372 int vec_all_ne (vector unsigned int, vector unsigned int);
8373 int vec_all_ne (vector bool int, vector bool int);
8374 int vec_all_ne (vector bool int, vector unsigned int);
8375 int vec_all_ne (vector bool int, vector signed int);
8376 int vec_all_ne (vector float, vector float);
8378 int vec_all_nge (vector float, vector float);
8380 int vec_all_ngt (vector float, vector float);
8382 int vec_all_nle (vector float, vector float);
8384 int vec_all_nlt (vector float, vector float);
8386 int vec_all_numeric (vector float);
8388 int vec_any_eq (vector signed char, vector bool char);
8389 int vec_any_eq (vector signed char, vector signed char);
8390 int vec_any_eq (vector unsigned char, vector bool char);
8391 int vec_any_eq (vector unsigned char, vector unsigned char);
8392 int vec_any_eq (vector bool char, vector bool char);
8393 int vec_any_eq (vector bool char, vector unsigned char);
8394 int vec_any_eq (vector bool char, vector signed char);
8395 int vec_any_eq (vector signed short, vector bool short);
8396 int vec_any_eq (vector signed short, vector signed short);
8397 int vec_any_eq (vector unsigned short, vector bool short);
8398 int vec_any_eq (vector unsigned short, vector unsigned short);
8399 int vec_any_eq (vector bool short, vector bool short);
8400 int vec_any_eq (vector bool short, vector unsigned short);
8401 int vec_any_eq (vector bool short, vector signed short);
8402 int vec_any_eq (vector pixel, vector pixel);
8403 int vec_any_eq (vector signed int, vector bool int);
8404 int vec_any_eq (vector signed int, vector signed int);
8405 int vec_any_eq (vector unsigned int, vector bool int);
8406 int vec_any_eq (vector unsigned int, vector unsigned int);
8407 int vec_any_eq (vector bool int, vector bool int);
8408 int vec_any_eq (vector bool int, vector unsigned int);
8409 int vec_any_eq (vector bool int, vector signed int);
8410 int vec_any_eq (vector float, vector float);
8412 int vec_any_ge (vector signed char, vector bool char);
8413 int vec_any_ge (vector unsigned char, vector bool char);
8414 int vec_any_ge (vector unsigned char, vector unsigned char);
8415 int vec_any_ge (vector signed char, vector signed char);
8416 int vec_any_ge (vector bool char, vector unsigned char);
8417 int vec_any_ge (vector bool char, vector signed char);
8418 int vec_any_ge (vector unsigned short, vector bool short);
8419 int vec_any_ge (vector unsigned short, vector unsigned short);
8420 int vec_any_ge (vector signed short, vector signed short);
8421 int vec_any_ge (vector signed short, vector bool short);
8422 int vec_any_ge (vector bool short, vector unsigned short);
8423 int vec_any_ge (vector bool short, vector signed short);
8424 int vec_any_ge (vector signed int, vector bool int);
8425 int vec_any_ge (vector unsigned int, vector bool int);
8426 int vec_any_ge (vector unsigned int, vector unsigned int);
8427 int vec_any_ge (vector signed int, vector signed int);
8428 int vec_any_ge (vector bool int, vector unsigned int);
8429 int vec_any_ge (vector bool int, vector signed int);
8430 int vec_any_ge (vector float, vector float);
8432 int vec_any_gt (vector bool char, vector unsigned char);
8433 int vec_any_gt (vector unsigned char, vector bool char);
8434 int vec_any_gt (vector unsigned char, vector unsigned char);
8435 int vec_any_gt (vector bool char, vector signed char);
8436 int vec_any_gt (vector signed char, vector bool char);
8437 int vec_any_gt (vector signed char, vector signed char);
8438 int vec_any_gt (vector bool short, vector unsigned short);
8439 int vec_any_gt (vector unsigned short, vector bool short);
8440 int vec_any_gt (vector unsigned short, vector unsigned short);
8441 int vec_any_gt (vector bool short, vector signed short);
8442 int vec_any_gt (vector signed short, vector bool short);
8443 int vec_any_gt (vector signed short, vector signed short);
8444 int vec_any_gt (vector bool int, vector unsigned int);
8445 int vec_any_gt (vector unsigned int, vector bool int);
8446 int vec_any_gt (vector unsigned int, vector unsigned int);
8447 int vec_any_gt (vector bool int, vector signed int);
8448 int vec_any_gt (vector signed int, vector bool int);
8449 int vec_any_gt (vector signed int, vector signed int);
8450 int vec_any_gt (vector float, vector float);
8452 int vec_any_le (vector bool char, vector unsigned char);
8453 int vec_any_le (vector unsigned char, vector bool char);
8454 int vec_any_le (vector unsigned char, vector unsigned char);
8455 int vec_any_le (vector bool char, vector signed char);
8456 int vec_any_le (vector signed char, vector bool char);
8457 int vec_any_le (vector signed char, vector signed char);
8458 int vec_any_le (vector bool short, vector unsigned short);
8459 int vec_any_le (vector unsigned short, vector bool short);
8460 int vec_any_le (vector unsigned short, vector unsigned short);
8461 int vec_any_le (vector bool short, vector signed short);
8462 int vec_any_le (vector signed short, vector bool short);
8463 int vec_any_le (vector signed short, vector signed short);
8464 int vec_any_le (vector bool int, vector unsigned int);
8465 int vec_any_le (vector unsigned int, vector bool int);
8466 int vec_any_le (vector unsigned int, vector unsigned int);
8467 int vec_any_le (vector bool int, vector signed int);
8468 int vec_any_le (vector signed int, vector bool int);
8469 int vec_any_le (vector signed int, vector signed int);
8470 int vec_any_le (vector float, vector float);
8472 int vec_any_lt (vector bool char, vector unsigned char);
8473 int vec_any_lt (vector unsigned char, vector bool char);
8474 int vec_any_lt (vector unsigned char, vector unsigned char);
8475 int vec_any_lt (vector bool char, vector signed char);
8476 int vec_any_lt (vector signed char, vector bool char);
8477 int vec_any_lt (vector signed char, vector signed char);
8478 int vec_any_lt (vector bool short, vector unsigned short);
8479 int vec_any_lt (vector unsigned short, vector bool short);
8480 int vec_any_lt (vector unsigned short, vector unsigned short);
8481 int vec_any_lt (vector bool short, vector signed short);
8482 int vec_any_lt (vector signed short, vector bool short);
8483 int vec_any_lt (vector signed short, vector signed short);
8484 int vec_any_lt (vector bool int, vector unsigned int);
8485 int vec_any_lt (vector unsigned int, vector bool int);
8486 int vec_any_lt (vector unsigned int, vector unsigned int);
8487 int vec_any_lt (vector bool int, vector signed int);
8488 int vec_any_lt (vector signed int, vector bool int);
8489 int vec_any_lt (vector signed int, vector signed int);
8490 int vec_any_lt (vector float, vector float);
8492 int vec_any_nan (vector float);
8494 int vec_any_ne (vector signed char, vector bool char);
8495 int vec_any_ne (vector signed char, vector signed char);
8496 int vec_any_ne (vector unsigned char, vector bool char);
8497 int vec_any_ne (vector unsigned char, vector unsigned char);
8498 int vec_any_ne (vector bool char, vector bool char);
8499 int vec_any_ne (vector bool char, vector unsigned char);
8500 int vec_any_ne (vector bool char, vector signed char);
8501 int vec_any_ne (vector signed short, vector bool short);
8502 int vec_any_ne (vector signed short, vector signed short);
8503 int vec_any_ne (vector unsigned short, vector bool short);
8504 int vec_any_ne (vector unsigned short, vector unsigned short);
8505 int vec_any_ne (vector bool short, vector bool short);
8506 int vec_any_ne (vector bool short, vector unsigned short);
8507 int vec_any_ne (vector bool short, vector signed short);
8508 int vec_any_ne (vector pixel, vector pixel);
8509 int vec_any_ne (vector signed int, vector bool int);
8510 int vec_any_ne (vector signed int, vector signed int);
8511 int vec_any_ne (vector unsigned int, vector bool int);
8512 int vec_any_ne (vector unsigned int, vector unsigned int);
8513 int vec_any_ne (vector bool int, vector bool int);
8514 int vec_any_ne (vector bool int, vector unsigned int);
8515 int vec_any_ne (vector bool int, vector signed int);
8516 int vec_any_ne (vector float, vector float);
8518 int vec_any_nge (vector float, vector float);
8520 int vec_any_ngt (vector float, vector float);
8522 int vec_any_nle (vector float, vector float);
8524 int vec_any_nlt (vector float, vector float);
8526 int vec_any_numeric (vector float);
8528 int vec_any_out (vector float, vector float);
8531 @node SPARC VIS Built-in Functions
8532 @subsection SPARC VIS Built-in Functions
8534 GCC supports SIMD operations on the SPARC using both the generic vector
8535 extensions (@pxref{Vector Extensions}) as well as built-in functions for
8536 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
8537 switch, the VIS extension is exposed as the following built-in functions:
8540 typedef int v2si __attribute__ ((vector_size (8)));
8541 typedef short v4hi __attribute__ ((vector_size (8)));
8542 typedef short v2hi __attribute__ ((vector_size (4)));
8543 typedef char v8qi __attribute__ ((vector_size (8)));
8544 typedef char v4qi __attribute__ ((vector_size (4)));
8546 void * __builtin_vis_alignaddr (void *, long);
8547 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
8548 v2si __builtin_vis_faligndatav2si (v2si, v2si);
8549 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
8550 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
8552 v4hi __builtin_vis_fexpand (v4qi);
8554 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
8555 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
8556 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
8557 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
8558 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
8559 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
8560 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
8562 v4qi __builtin_vis_fpack16 (v4hi);
8563 v8qi __builtin_vis_fpack32 (v2si, v2si);
8564 v2hi __builtin_vis_fpackfix (v2si);
8565 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
8567 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
8570 @node Target Format Checks
8571 @section Format Checks Specific to Particular Target Machines
8573 For some target machines, GCC supports additional options to the
8575 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
8578 * Solaris Format Checks::
8581 @node Solaris Format Checks
8582 @subsection Solaris Format Checks
8584 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
8585 check. @code{cmn_err} accepts a subset of the standard @code{printf}
8586 conversions, and the two-argument @code{%b} conversion for displaying
8587 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
8590 @section Pragmas Accepted by GCC
8594 GCC supports several types of pragmas, primarily in order to compile
8595 code originally written for other compilers. Note that in general
8596 we do not recommend the use of pragmas; @xref{Function Attributes},
8597 for further explanation.
8601 * RS/6000 and PowerPC Pragmas::
8604 * Symbol-Renaming Pragmas::
8605 * Structure-Packing Pragmas::
8610 @subsection ARM Pragmas
8612 The ARM target defines pragmas for controlling the default addition of
8613 @code{long_call} and @code{short_call} attributes to functions.
8614 @xref{Function Attributes}, for information about the effects of these
8619 @cindex pragma, long_calls
8620 Set all subsequent functions to have the @code{long_call} attribute.
8623 @cindex pragma, no_long_calls
8624 Set all subsequent functions to have the @code{short_call} attribute.
8626 @item long_calls_off
8627 @cindex pragma, long_calls_off
8628 Do not affect the @code{long_call} or @code{short_call} attributes of
8629 subsequent functions.
8632 @node RS/6000 and PowerPC Pragmas
8633 @subsection RS/6000 and PowerPC Pragmas
8635 The RS/6000 and PowerPC targets define one pragma for controlling
8636 whether or not the @code{longcall} attribute is added to function
8637 declarations by default. This pragma overrides the @option{-mlongcall}
8638 option, but not the @code{longcall} and @code{shortcall} attributes.
8639 @xref{RS/6000 and PowerPC Options}, for more information about when long
8640 calls are and are not necessary.
8644 @cindex pragma, longcall
8645 Apply the @code{longcall} attribute to all subsequent function
8649 Do not apply the @code{longcall} attribute to subsequent function
8653 @c Describe c4x pragmas here.
8654 @c Describe h8300 pragmas here.
8655 @c Describe sh pragmas here.
8656 @c Describe v850 pragmas here.
8658 @node Darwin Pragmas
8659 @subsection Darwin Pragmas
8661 The following pragmas are available for all architectures running the
8662 Darwin operating system. These are useful for compatibility with other
8666 @item mark @var{tokens}@dots{}
8667 @cindex pragma, mark
8668 This pragma is accepted, but has no effect.
8670 @item options align=@var{alignment}
8671 @cindex pragma, options align
8672 This pragma sets the alignment of fields in structures. The values of
8673 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
8674 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
8675 properly; to restore the previous setting, use @code{reset} for the
8678 @item segment @var{tokens}@dots{}
8679 @cindex pragma, segment
8680 This pragma is accepted, but has no effect.
8682 @item unused (@var{var} [, @var{var}]@dots{})
8683 @cindex pragma, unused
8684 This pragma declares variables to be possibly unused. GCC will not
8685 produce warnings for the listed variables. The effect is similar to
8686 that of the @code{unused} attribute, except that this pragma may appear
8687 anywhere within the variables' scopes.
8690 @node Solaris Pragmas
8691 @subsection Solaris Pragmas
8693 The Solaris target supports @code{#pragma redefine_extname}
8694 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
8695 @code{#pragma} directives for compatibility with the system compiler.
8698 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
8699 @cindex pragma, align
8701 Increase the minimum alignment of each @var{variable} to @var{alignment}.
8702 This is the same as GCC's @code{aligned} attribute @pxref{Variable
8703 Attributes}). Macro expansion occurs on the arguments to this pragma
8704 when compiling C and Objective-C. It does not currently occur when
8705 compiling C++, but this is a bug which may be fixed in a future
8708 @item fini (@var{function} [, @var{function}]...)
8709 @cindex pragma, fini
8711 This pragma causes each listed @var{function} to be called after
8712 main, or during shared module unloading, by adding a call to the
8713 @code{.fini} section.
8715 @item init (@var{function} [, @var{function}]...)
8716 @cindex pragma, init
8718 This pragma causes each listed @var{function} to be called during
8719 initialization (before @code{main}) or during shared module loading, by
8720 adding a call to the @code{.init} section.
8724 @node Symbol-Renaming Pragmas
8725 @subsection Symbol-Renaming Pragmas
8727 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
8728 supports two @code{#pragma} directives which change the name used in
8729 assembly for a given declaration. These pragmas are only available on
8730 platforms whose system headers need them. To get this effect on all
8731 platforms supported by GCC, use the asm labels extension (@pxref{Asm
8735 @item redefine_extname @var{oldname} @var{newname}
8736 @cindex pragma, redefine_extname
8738 This pragma gives the C function @var{oldname} the assembly symbol
8739 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
8740 will be defined if this pragma is available (currently only on
8743 @item extern_prefix @var{string}
8744 @cindex pragma, extern_prefix
8746 This pragma causes all subsequent external function and variable
8747 declarations to have @var{string} prepended to their assembly symbols.
8748 This effect may be terminated with another @code{extern_prefix} pragma
8749 whose argument is an empty string. The preprocessor macro
8750 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
8751 available (currently only on Tru64 UNIX)@.
8754 These pragmas and the asm labels extension interact in a complicated
8755 manner. Here are some corner cases you may want to be aware of.
8758 @item Both pragmas silently apply only to declarations with external
8759 linkage. Asm labels do not have this restriction.
8761 @item In C++, both pragmas silently apply only to declarations with
8762 ``C'' linkage. Again, asm labels do not have this restriction.
8764 @item If any of the three ways of changing the assembly name of a
8765 declaration is applied to a declaration whose assembly name has
8766 already been determined (either by a previous use of one of these
8767 features, or because the compiler needed the assembly name in order to
8768 generate code), and the new name is different, a warning issues and
8769 the name does not change.
8771 @item The @var{oldname} used by @code{#pragma redefine_extname} is
8772 always the C-language name.
8774 @item If @code{#pragma extern_prefix} is in effect, and a declaration
8775 occurs with an asm label attached, the prefix is silently ignored for
8778 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
8779 apply to the same declaration, whichever triggered first wins, and a
8780 warning issues if they contradict each other. (We would like to have
8781 @code{#pragma redefine_extname} always win, for consistency with asm
8782 labels, but if @code{#pragma extern_prefix} triggers first we have no
8783 way of knowing that that happened.)
8786 @node Structure-Packing Pragmas
8787 @subsection Structure-Packing Pragmas
8789 For compatibility with Win32, GCC supports a set of @code{#pragma}
8790 directives which change the maximum alignment of members of structures,
8791 unions, and classes subsequently defined. The @var{n} value below always
8792 is required to be a small power of two and specifies the new alignment
8796 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
8797 @item @code{#pragma pack()} sets the alignment to the one that was in
8798 effect when compilation started (see also command line option
8799 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
8800 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
8801 setting on an internal stack and then optionally sets the new alignment.
8802 @item @code{#pragma pack(pop)} restores the alignment setting to the one
8803 saved at the top of the internal stack (and removes that stack entry).
8804 Note that @code{#pragma pack([@var{n}])} does not influence this internal
8805 stack; thus it is possible to have @code{#pragma pack(push)} followed by
8806 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
8807 @code{#pragma pack(pop)}.
8811 @subsection Weak Pragmas
8813 For compatibility with SVR4, GCC supports a set of @code{#pragma}
8814 directives for declaring symbols to be weak, and defining weak
8818 @item #pragma weak @var{symbol}
8819 @cindex pragma, weak
8820 This pragma declares @var{symbol} to be weak, as if the declaration
8821 had the attribute of the same name. The pragma may appear before
8822 or after the declaration of @var{symbol}, but must appear before
8823 either its first use or its definition. It is not an error for
8824 @var{symbol} to never be defined at all.
8826 @item #pragma weak @var{symbol1} = @var{symbol2}
8827 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
8828 It is an error if @var{symbol2} is not defined in the current
8832 @node Unnamed Fields
8833 @section Unnamed struct/union fields within structs/unions
8837 For compatibility with other compilers, GCC allows you to define
8838 a structure or union that contains, as fields, structures and unions
8839 without names. For example:
8852 In this example, the user would be able to access members of the unnamed
8853 union with code like @samp{foo.b}. Note that only unnamed structs and
8854 unions are allowed, you may not have, for example, an unnamed
8857 You must never create such structures that cause ambiguous field definitions.
8858 For example, this structure:
8869 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
8870 Such constructs are not supported and must be avoided. In the future,
8871 such constructs may be detected and treated as compilation errors.
8873 @opindex fms-extensions
8874 Unless @option{-fms-extensions} is used, the unnamed field must be a
8875 structure or union definition without a tag (for example, @samp{struct
8876 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
8877 also be a definition with a tag such as @samp{struct foo @{ int a;
8878 @};}, a reference to a previously defined structure or union such as
8879 @samp{struct foo;}, or a reference to a @code{typedef} name for a
8880 previously defined structure or union type.
8883 @section Thread-Local Storage
8884 @cindex Thread-Local Storage
8885 @cindex @acronym{TLS}
8888 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
8889 are allocated such that there is one instance of the variable per extant
8890 thread. The run-time model GCC uses to implement this originates
8891 in the IA-64 processor-specific ABI, but has since been migrated
8892 to other processors as well. It requires significant support from
8893 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
8894 system libraries (@file{libc.so} and @file{libpthread.so}), so it
8895 is not available everywhere.
8897 At the user level, the extension is visible with a new storage
8898 class keyword: @code{__thread}. For example:
8902 extern __thread struct state s;
8903 static __thread char *p;
8906 The @code{__thread} specifier may be used alone, with the @code{extern}
8907 or @code{static} specifiers, but with no other storage class specifier.
8908 When used with @code{extern} or @code{static}, @code{__thread} must appear
8909 immediately after the other storage class specifier.
8911 The @code{__thread} specifier may be applied to any global, file-scoped
8912 static, function-scoped static, or static data member of a class. It may
8913 not be applied to block-scoped automatic or non-static data member.
8915 When the address-of operator is applied to a thread-local variable, it is
8916 evaluated at run-time and returns the address of the current thread's
8917 instance of that variable. An address so obtained may be used by any
8918 thread. When a thread terminates, any pointers to thread-local variables
8919 in that thread become invalid.
8921 No static initialization may refer to the address of a thread-local variable.
8923 In C++, if an initializer is present for a thread-local variable, it must
8924 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
8927 See @uref{http://people.redhat.com/drepper/tls.pdf,
8928 ELF Handling For Thread-Local Storage} for a detailed explanation of
8929 the four thread-local storage addressing models, and how the run-time
8930 is expected to function.
8933 * C99 Thread-Local Edits::
8934 * C++98 Thread-Local Edits::
8937 @node C99 Thread-Local Edits
8938 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
8940 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
8941 that document the exact semantics of the language extension.
8945 @cite{5.1.2 Execution environments}
8947 Add new text after paragraph 1
8950 Within either execution environment, a @dfn{thread} is a flow of
8951 control within a program. It is implementation defined whether
8952 or not there may be more than one thread associated with a program.
8953 It is implementation defined how threads beyond the first are
8954 created, the name and type of the function called at thread
8955 startup, and how threads may be terminated. However, objects
8956 with thread storage duration shall be initialized before thread
8961 @cite{6.2.4 Storage durations of objects}
8963 Add new text before paragraph 3
8966 An object whose identifier is declared with the storage-class
8967 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
8968 Its lifetime is the entire execution of the thread, and its
8969 stored value is initialized only once, prior to thread startup.
8973 @cite{6.4.1 Keywords}
8975 Add @code{__thread}.
8978 @cite{6.7.1 Storage-class specifiers}
8980 Add @code{__thread} to the list of storage class specifiers in
8983 Change paragraph 2 to
8986 With the exception of @code{__thread}, at most one storage-class
8987 specifier may be given [@dots{}]. The @code{__thread} specifier may
8988 be used alone, or immediately following @code{extern} or
8992 Add new text after paragraph 6
8995 The declaration of an identifier for a variable that has
8996 block scope that specifies @code{__thread} shall also
8997 specify either @code{extern} or @code{static}.
8999 The @code{__thread} specifier shall be used only with
9004 @node C++98 Thread-Local Edits
9005 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
9007 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
9008 that document the exact semantics of the language extension.
9012 @b{[intro.execution]}
9014 New text after paragraph 4
9017 A @dfn{thread} is a flow of control within the abstract machine.
9018 It is implementation defined whether or not there may be more than
9022 New text after paragraph 7
9025 It is unspecified whether additional action must be taken to
9026 ensure when and whether side effects are visible to other threads.
9032 Add @code{__thread}.
9035 @b{[basic.start.main]}
9037 Add after paragraph 5
9040 The thread that begins execution at the @code{main} function is called
9041 the @dfn{main thread}. It is implementation defined how functions
9042 beginning threads other than the main thread are designated or typed.
9043 A function so designated, as well as the @code{main} function, is called
9044 a @dfn{thread startup function}. It is implementation defined what
9045 happens if a thread startup function returns. It is implementation
9046 defined what happens to other threads when any thread calls @code{exit}.
9050 @b{[basic.start.init]}
9052 Add after paragraph 4
9055 The storage for an object of thread storage duration shall be
9056 statically initialized before the first statement of the thread startup
9057 function. An object of thread storage duration shall not require
9058 dynamic initialization.
9062 @b{[basic.start.term]}
9064 Add after paragraph 3
9067 The type of an object with thread storage duration shall not have a
9068 non-trivial destructor, nor shall it be an array type whose elements
9069 (directly or indirectly) have non-trivial destructors.
9075 Add ``thread storage duration'' to the list in paragraph 1.
9080 Thread, static, and automatic storage durations are associated with
9081 objects introduced by declarations [@dots{}].
9084 Add @code{__thread} to the list of specifiers in paragraph 3.
9087 @b{[basic.stc.thread]}
9089 New section before @b{[basic.stc.static]}
9092 The keyword @code{__thread} applied to a non-local object gives the
9093 object thread storage duration.
9095 A local variable or class data member declared both @code{static}
9096 and @code{__thread} gives the variable or member thread storage
9101 @b{[basic.stc.static]}
9106 All objects which have neither thread storage duration, dynamic
9107 storage duration nor are local [@dots{}].
9113 Add @code{__thread} to the list in paragraph 1.
9118 With the exception of @code{__thread}, at most one
9119 @var{storage-class-specifier} shall appear in a given
9120 @var{decl-specifier-seq}. The @code{__thread} specifier may
9121 be used alone, or immediately following the @code{extern} or
9122 @code{static} specifiers. [@dots{}]
9125 Add after paragraph 5
9128 The @code{__thread} specifier can be applied only to the names of objects
9129 and to anonymous unions.
9135 Add after paragraph 6
9138 Non-@code{static} members shall not be @code{__thread}.
9142 @node C++ Extensions
9143 @chapter Extensions to the C++ Language
9144 @cindex extensions, C++ language
9145 @cindex C++ language extensions
9147 The GNU compiler provides these extensions to the C++ language (and you
9148 can also use most of the C language extensions in your C++ programs). If you
9149 want to write code that checks whether these features are available, you can
9150 test for the GNU compiler the same way as for C programs: check for a
9151 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
9152 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
9153 Predefined Macros,cpp,The GNU C Preprocessor}).
9156 * Volatiles:: What constitutes an access to a volatile object.
9157 * Restricted Pointers:: C99 restricted pointers and references.
9158 * Vague Linkage:: Where G++ puts inlines, vtables and such.
9159 * C++ Interface:: You can use a single C++ header file for both
9160 declarations and definitions.
9161 * Template Instantiation:: Methods for ensuring that exactly one copy of
9162 each needed template instantiation is emitted.
9163 * Bound member functions:: You can extract a function pointer to the
9164 method denoted by a @samp{->*} or @samp{.*} expression.
9165 * C++ Attributes:: Variable, function, and type attributes for C++ only.
9166 * Strong Using:: Strong using-directives for namespace composition.
9167 * Java Exceptions:: Tweaking exception handling to work with Java.
9168 * Deprecated Features:: Things will disappear from g++.
9169 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
9173 @section When is a Volatile Object Accessed?
9174 @cindex accessing volatiles
9175 @cindex volatile read
9176 @cindex volatile write
9177 @cindex volatile access
9179 Both the C and C++ standard have the concept of volatile objects. These
9180 are normally accessed by pointers and used for accessing hardware. The
9181 standards encourage compilers to refrain from optimizations
9182 concerning accesses to volatile objects that it might perform on
9183 non-volatile objects. The C standard leaves it implementation defined
9184 as to what constitutes a volatile access. The C++ standard omits to
9185 specify this, except to say that C++ should behave in a similar manner
9186 to C with respect to volatiles, where possible. The minimum either
9187 standard specifies is that at a sequence point all previous accesses to
9188 volatile objects have stabilized and no subsequent accesses have
9189 occurred. Thus an implementation is free to reorder and combine
9190 volatile accesses which occur between sequence points, but cannot do so
9191 for accesses across a sequence point. The use of volatiles does not
9192 allow you to violate the restriction on updating objects multiple times
9193 within a sequence point.
9195 In most expressions, it is intuitively obvious what is a read and what is
9196 a write. For instance
9199 volatile int *dst = @var{somevalue};
9200 volatile int *src = @var{someothervalue};
9205 will cause a read of the volatile object pointed to by @var{src} and stores the
9206 value into the volatile object pointed to by @var{dst}. There is no
9207 guarantee that these reads and writes are atomic, especially for objects
9208 larger than @code{int}.
9210 Less obvious expressions are where something which looks like an access
9211 is used in a void context. An example would be,
9214 volatile int *src = @var{somevalue};
9218 With C, such expressions are rvalues, and as rvalues cause a read of
9219 the object, GCC interprets this as a read of the volatile being pointed
9220 to. The C++ standard specifies that such expressions do not undergo
9221 lvalue to rvalue conversion, and that the type of the dereferenced
9222 object may be incomplete. The C++ standard does not specify explicitly
9223 that it is this lvalue to rvalue conversion which is responsible for
9224 causing an access. However, there is reason to believe that it is,
9225 because otherwise certain simple expressions become undefined. However,
9226 because it would surprise most programmers, G++ treats dereferencing a
9227 pointer to volatile object of complete type in a void context as a read
9228 of the object. When the object has incomplete type, G++ issues a
9233 struct T @{int m;@};
9234 volatile S *ptr1 = @var{somevalue};
9235 volatile T *ptr2 = @var{somevalue};
9240 In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
9241 causes a read of the object pointed to. If you wish to force an error on
9242 the first case, you must force a conversion to rvalue with, for instance
9243 a static cast, @code{static_cast<S>(*ptr1)}.
9245 When using a reference to volatile, G++ does not treat equivalent
9246 expressions as accesses to volatiles, but instead issues a warning that
9247 no volatile is accessed. The rationale for this is that otherwise it
9248 becomes difficult to determine where volatile access occur, and not
9249 possible to ignore the return value from functions returning volatile
9250 references. Again, if you wish to force a read, cast the reference to
9253 @node Restricted Pointers
9254 @section Restricting Pointer Aliasing
9255 @cindex restricted pointers
9256 @cindex restricted references
9257 @cindex restricted this pointer
9259 As with the C front end, G++ understands the C99 feature of restricted pointers,
9260 specified with the @code{__restrict__}, or @code{__restrict} type
9261 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
9262 language flag, @code{restrict} is not a keyword in C++.
9264 In addition to allowing restricted pointers, you can specify restricted
9265 references, which indicate that the reference is not aliased in the local
9269 void fn (int *__restrict__ rptr, int &__restrict__ rref)
9276 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
9277 @var{rref} refers to a (different) unaliased integer.
9279 You may also specify whether a member function's @var{this} pointer is
9280 unaliased by using @code{__restrict__} as a member function qualifier.
9283 void T::fn () __restrict__
9290 Within the body of @code{T::fn}, @var{this} will have the effective
9291 definition @code{T *__restrict__ const this}. Notice that the
9292 interpretation of a @code{__restrict__} member function qualifier is
9293 different to that of @code{const} or @code{volatile} qualifier, in that it
9294 is applied to the pointer rather than the object. This is consistent with
9295 other compilers which implement restricted pointers.
9297 As with all outermost parameter qualifiers, @code{__restrict__} is
9298 ignored in function definition matching. This means you only need to
9299 specify @code{__restrict__} in a function definition, rather than
9300 in a function prototype as well.
9303 @section Vague Linkage
9304 @cindex vague linkage
9306 There are several constructs in C++ which require space in the object
9307 file but are not clearly tied to a single translation unit. We say that
9308 these constructs have ``vague linkage''. Typically such constructs are
9309 emitted wherever they are needed, though sometimes we can be more
9313 @item Inline Functions
9314 Inline functions are typically defined in a header file which can be
9315 included in many different compilations. Hopefully they can usually be
9316 inlined, but sometimes an out-of-line copy is necessary, if the address
9317 of the function is taken or if inlining fails. In general, we emit an
9318 out-of-line copy in all translation units where one is needed. As an
9319 exception, we only emit inline virtual functions with the vtable, since
9320 it will always require a copy.
9322 Local static variables and string constants used in an inline function
9323 are also considered to have vague linkage, since they must be shared
9324 between all inlined and out-of-line instances of the function.
9328 C++ virtual functions are implemented in most compilers using a lookup
9329 table, known as a vtable. The vtable contains pointers to the virtual
9330 functions provided by a class, and each object of the class contains a
9331 pointer to its vtable (or vtables, in some multiple-inheritance
9332 situations). If the class declares any non-inline, non-pure virtual
9333 functions, the first one is chosen as the ``key method'' for the class,
9334 and the vtable is only emitted in the translation unit where the key
9337 @emph{Note:} If the chosen key method is later defined as inline, the
9338 vtable will still be emitted in every translation unit which defines it.
9339 Make sure that any inline virtuals are declared inline in the class
9340 body, even if they are not defined there.
9342 @item type_info objects
9345 C++ requires information about types to be written out in order to
9346 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
9347 For polymorphic classes (classes with virtual functions), the type_info
9348 object is written out along with the vtable so that @samp{dynamic_cast}
9349 can determine the dynamic type of a class object at runtime. For all
9350 other types, we write out the type_info object when it is used: when
9351 applying @samp{typeid} to an expression, throwing an object, or
9352 referring to a type in a catch clause or exception specification.
9354 @item Template Instantiations
9355 Most everything in this section also applies to template instantiations,
9356 but there are other options as well.
9357 @xref{Template Instantiation,,Where's the Template?}.
9361 When used with GNU ld version 2.8 or later on an ELF system such as
9362 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
9363 these constructs will be discarded at link time. This is known as
9366 On targets that don't support COMDAT, but do support weak symbols, GCC
9367 will use them. This way one copy will override all the others, but
9368 the unused copies will still take up space in the executable.
9370 For targets which do not support either COMDAT or weak symbols,
9371 most entities with vague linkage will be emitted as local symbols to
9372 avoid duplicate definition errors from the linker. This will not happen
9373 for local statics in inlines, however, as having multiple copies will
9374 almost certainly break things.
9376 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
9377 another way to control placement of these constructs.
9380 @section #pragma interface and implementation
9382 @cindex interface and implementation headers, C++
9383 @cindex C++ interface and implementation headers
9384 @cindex pragmas, interface and implementation
9386 @code{#pragma interface} and @code{#pragma implementation} provide the
9387 user with a way of explicitly directing the compiler to emit entities
9388 with vague linkage (and debugging information) in a particular
9391 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
9392 most cases, because of COMDAT support and the ``key method'' heuristic
9393 mentioned in @ref{Vague Linkage}. Using them can actually cause your
9394 program to grow due to unnecessary out-of-line copies of inline
9395 functions. Currently (3.4) the only benefit of these
9396 @code{#pragma}s is reduced duplication of debugging information, and
9397 that should be addressed soon on DWARF 2 targets with the use of
9401 @item #pragma interface
9402 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
9403 @kindex #pragma interface
9404 Use this directive in @emph{header files} that define object classes, to save
9405 space in most of the object files that use those classes. Normally,
9406 local copies of certain information (backup copies of inline member
9407 functions, debugging information, and the internal tables that implement
9408 virtual functions) must be kept in each object file that includes class
9409 definitions. You can use this pragma to avoid such duplication. When a
9410 header file containing @samp{#pragma interface} is included in a
9411 compilation, this auxiliary information will not be generated (unless
9412 the main input source file itself uses @samp{#pragma implementation}).
9413 Instead, the object files will contain references to be resolved at link
9416 The second form of this directive is useful for the case where you have
9417 multiple headers with the same name in different directories. If you
9418 use this form, you must specify the same string to @samp{#pragma
9421 @item #pragma implementation
9422 @itemx #pragma implementation "@var{objects}.h"
9423 @kindex #pragma implementation
9424 Use this pragma in a @emph{main input file}, when you want full output from
9425 included header files to be generated (and made globally visible). The
9426 included header file, in turn, should use @samp{#pragma interface}.
9427 Backup copies of inline member functions, debugging information, and the
9428 internal tables used to implement virtual functions are all generated in
9429 implementation files.
9431 @cindex implied @code{#pragma implementation}
9432 @cindex @code{#pragma implementation}, implied
9433 @cindex naming convention, implementation headers
9434 If you use @samp{#pragma implementation} with no argument, it applies to
9435 an include file with the same basename@footnote{A file's @dfn{basename}
9436 was the name stripped of all leading path information and of trailing
9437 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
9438 file. For example, in @file{allclass.cc}, giving just
9439 @samp{#pragma implementation}
9440 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
9442 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
9443 an implementation file whenever you would include it from
9444 @file{allclass.cc} even if you never specified @samp{#pragma
9445 implementation}. This was deemed to be more trouble than it was worth,
9446 however, and disabled.
9448 Use the string argument if you want a single implementation file to
9449 include code from multiple header files. (You must also use
9450 @samp{#include} to include the header file; @samp{#pragma
9451 implementation} only specifies how to use the file---it doesn't actually
9454 There is no way to split up the contents of a single header file into
9455 multiple implementation files.
9458 @cindex inlining and C++ pragmas
9459 @cindex C++ pragmas, effect on inlining
9460 @cindex pragmas in C++, effect on inlining
9461 @samp{#pragma implementation} and @samp{#pragma interface} also have an
9462 effect on function inlining.
9464 If you define a class in a header file marked with @samp{#pragma
9465 interface}, the effect on an inline function defined in that class is
9466 similar to an explicit @code{extern} declaration---the compiler emits
9467 no code at all to define an independent version of the function. Its
9468 definition is used only for inlining with its callers.
9470 @opindex fno-implement-inlines
9471 Conversely, when you include the same header file in a main source file
9472 that declares it as @samp{#pragma implementation}, the compiler emits
9473 code for the function itself; this defines a version of the function
9474 that can be found via pointers (or by callers compiled without
9475 inlining). If all calls to the function can be inlined, you can avoid
9476 emitting the function by compiling with @option{-fno-implement-inlines}.
9477 If any calls were not inlined, you will get linker errors.
9479 @node Template Instantiation
9480 @section Where's the Template?
9481 @cindex template instantiation
9483 C++ templates are the first language feature to require more
9484 intelligence from the environment than one usually finds on a UNIX
9485 system. Somehow the compiler and linker have to make sure that each
9486 template instance occurs exactly once in the executable if it is needed,
9487 and not at all otherwise. There are two basic approaches to this
9488 problem, which are referred to as the Borland model and the Cfront model.
9492 Borland C++ solved the template instantiation problem by adding the code
9493 equivalent of common blocks to their linker; the compiler emits template
9494 instances in each translation unit that uses them, and the linker
9495 collapses them together. The advantage of this model is that the linker
9496 only has to consider the object files themselves; there is no external
9497 complexity to worry about. This disadvantage is that compilation time
9498 is increased because the template code is being compiled repeatedly.
9499 Code written for this model tends to include definitions of all
9500 templates in the header file, since they must be seen to be
9504 The AT&T C++ translator, Cfront, solved the template instantiation
9505 problem by creating the notion of a template repository, an
9506 automatically maintained place where template instances are stored. A
9507 more modern version of the repository works as follows: As individual
9508 object files are built, the compiler places any template definitions and
9509 instantiations encountered in the repository. At link time, the link
9510 wrapper adds in the objects in the repository and compiles any needed
9511 instances that were not previously emitted. The advantages of this
9512 model are more optimal compilation speed and the ability to use the
9513 system linker; to implement the Borland model a compiler vendor also
9514 needs to replace the linker. The disadvantages are vastly increased
9515 complexity, and thus potential for error; for some code this can be
9516 just as transparent, but in practice it can been very difficult to build
9517 multiple programs in one directory and one program in multiple
9518 directories. Code written for this model tends to separate definitions
9519 of non-inline member templates into a separate file, which should be
9520 compiled separately.
9523 When used with GNU ld version 2.8 or later on an ELF system such as
9524 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
9525 Borland model. On other systems, G++ implements neither automatic
9528 A future version of G++ will support a hybrid model whereby the compiler
9529 will emit any instantiations for which the template definition is
9530 included in the compile, and store template definitions and
9531 instantiation context information into the object file for the rest.
9532 The link wrapper will extract that information as necessary and invoke
9533 the compiler to produce the remaining instantiations. The linker will
9534 then combine duplicate instantiations.
9536 In the mean time, you have the following options for dealing with
9537 template instantiations:
9542 Compile your template-using code with @option{-frepo}. The compiler will
9543 generate files with the extension @samp{.rpo} listing all of the
9544 template instantiations used in the corresponding object files which
9545 could be instantiated there; the link wrapper, @samp{collect2}, will
9546 then update the @samp{.rpo} files to tell the compiler where to place
9547 those instantiations and rebuild any affected object files. The
9548 link-time overhead is negligible after the first pass, as the compiler
9549 will continue to place the instantiations in the same files.
9551 This is your best option for application code written for the Borland
9552 model, as it will just work. Code written for the Cfront model will
9553 need to be modified so that the template definitions are available at
9554 one or more points of instantiation; usually this is as simple as adding
9555 @code{#include <tmethods.cc>} to the end of each template header.
9557 For library code, if you want the library to provide all of the template
9558 instantiations it needs, just try to link all of its object files
9559 together; the link will fail, but cause the instantiations to be
9560 generated as a side effect. Be warned, however, that this may cause
9561 conflicts if multiple libraries try to provide the same instantiations.
9562 For greater control, use explicit instantiation as described in the next
9566 @opindex fno-implicit-templates
9567 Compile your code with @option{-fno-implicit-templates} to disable the
9568 implicit generation of template instances, and explicitly instantiate
9569 all the ones you use. This approach requires more knowledge of exactly
9570 which instances you need than do the others, but it's less
9571 mysterious and allows greater control. You can scatter the explicit
9572 instantiations throughout your program, perhaps putting them in the
9573 translation units where the instances are used or the translation units
9574 that define the templates themselves; you can put all of the explicit
9575 instantiations you need into one big file; or you can create small files
9582 template class Foo<int>;
9583 template ostream& operator <<
9584 (ostream&, const Foo<int>&);
9587 for each of the instances you need, and create a template instantiation
9590 If you are using Cfront-model code, you can probably get away with not
9591 using @option{-fno-implicit-templates} when compiling files that don't
9592 @samp{#include} the member template definitions.
9594 If you use one big file to do the instantiations, you may want to
9595 compile it without @option{-fno-implicit-templates} so you get all of the
9596 instances required by your explicit instantiations (but not by any
9597 other files) without having to specify them as well.
9599 G++ has extended the template instantiation syntax given in the ISO
9600 standard to allow forward declaration of explicit instantiations
9601 (with @code{extern}), instantiation of the compiler support data for a
9602 template class (i.e.@: the vtable) without instantiating any of its
9603 members (with @code{inline}), and instantiation of only the static data
9604 members of a template class, without the support data or member
9605 functions (with (@code{static}):
9608 extern template int max (int, int);
9609 inline template class Foo<int>;
9610 static template class Foo<int>;
9614 Do nothing. Pretend G++ does implement automatic instantiation
9615 management. Code written for the Borland model will work fine, but
9616 each translation unit will contain instances of each of the templates it
9617 uses. In a large program, this can lead to an unacceptable amount of code
9621 @node Bound member functions
9622 @section Extracting the function pointer from a bound pointer to member function
9624 @cindex pointer to member function
9625 @cindex bound pointer to member function
9627 In C++, pointer to member functions (PMFs) are implemented using a wide
9628 pointer of sorts to handle all the possible call mechanisms; the PMF
9629 needs to store information about how to adjust the @samp{this} pointer,
9630 and if the function pointed to is virtual, where to find the vtable, and
9631 where in the vtable to look for the member function. If you are using
9632 PMFs in an inner loop, you should really reconsider that decision. If
9633 that is not an option, you can extract the pointer to the function that
9634 would be called for a given object/PMF pair and call it directly inside
9635 the inner loop, to save a bit of time.
9637 Note that you will still be paying the penalty for the call through a
9638 function pointer; on most modern architectures, such a call defeats the
9639 branch prediction features of the CPU@. This is also true of normal
9640 virtual function calls.
9642 The syntax for this extension is
9646 extern int (A::*fp)();
9647 typedef int (*fptr)(A *);
9649 fptr p = (fptr)(a.*fp);
9652 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
9653 no object is needed to obtain the address of the function. They can be
9654 converted to function pointers directly:
9657 fptr p1 = (fptr)(&A::foo);
9660 @opindex Wno-pmf-conversions
9661 You must specify @option{-Wno-pmf-conversions} to use this extension.
9663 @node C++ Attributes
9664 @section C++-Specific Variable, Function, and Type Attributes
9666 Some attributes only make sense for C++ programs.
9669 @item init_priority (@var{priority})
9670 @cindex init_priority attribute
9673 In Standard C++, objects defined at namespace scope are guaranteed to be
9674 initialized in an order in strict accordance with that of their definitions
9675 @emph{in a given translation unit}. No guarantee is made for initializations
9676 across translation units. However, GNU C++ allows users to control the
9677 order of initialization of objects defined at namespace scope with the
9678 @code{init_priority} attribute by specifying a relative @var{priority},
9679 a constant integral expression currently bounded between 101 and 65535
9680 inclusive. Lower numbers indicate a higher priority.
9682 In the following example, @code{A} would normally be created before
9683 @code{B}, but the @code{init_priority} attribute has reversed that order:
9686 Some_Class A __attribute__ ((init_priority (2000)));
9687 Some_Class B __attribute__ ((init_priority (543)));
9691 Note that the particular values of @var{priority} do not matter; only their
9694 @item java_interface
9695 @cindex java_interface attribute
9697 This type attribute informs C++ that the class is a Java interface. It may
9698 only be applied to classes declared within an @code{extern "Java"} block.
9699 Calls to methods declared in this interface will be dispatched using GCJ's
9700 interface table mechanism, instead of regular virtual table dispatch.
9704 See also @xref{Strong Using}.
9707 @section Strong Using
9709 @strong{Caution:} The semantics of this extension are not fully
9710 defined. Users should refrain from using this extension as its
9711 semantics may change subtly over time. It is possible that this
9712 extension wil be removed in future versions of G++.
9714 A using-directive with @code{__attribute ((strong))} is stronger
9715 than a normal using-directive in two ways:
9719 Templates from the used namespace can be specialized as though they were members of the using namespace.
9722 The using namespace is considered an associated namespace of all
9723 templates in the used namespace for purposes of argument-dependent
9727 This is useful for composing a namespace transparently from
9728 implementation namespaces. For example:
9733 template <class T> struct A @{ @};
9735 using namespace debug __attribute ((__strong__));
9736 template <> struct A<int> @{ @}; // @r{ok to specialize}
9738 template <class T> void f (A<T>);
9743 f (std::A<float>()); // @r{lookup finds} std::f
9748 @node Java Exceptions
9749 @section Java Exceptions
9751 The Java language uses a slightly different exception handling model
9752 from C++. Normally, GNU C++ will automatically detect when you are
9753 writing C++ code that uses Java exceptions, and handle them
9754 appropriately. However, if C++ code only needs to execute destructors
9755 when Java exceptions are thrown through it, GCC will guess incorrectly.
9756 Sample problematic code is:
9759 struct S @{ ~S(); @};
9760 extern void bar(); // @r{is written in Java, and may throw exceptions}
9769 The usual effect of an incorrect guess is a link failure, complaining of
9770 a missing routine called @samp{__gxx_personality_v0}.
9772 You can inform the compiler that Java exceptions are to be used in a
9773 translation unit, irrespective of what it might think, by writing
9774 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
9775 @samp{#pragma} must appear before any functions that throw or catch
9776 exceptions, or run destructors when exceptions are thrown through them.
9778 You cannot mix Java and C++ exceptions in the same translation unit. It
9779 is believed to be safe to throw a C++ exception from one file through
9780 another file compiled for the Java exception model, or vice versa, but
9781 there may be bugs in this area.
9783 @node Deprecated Features
9784 @section Deprecated Features
9786 In the past, the GNU C++ compiler was extended to experiment with new
9787 features, at a time when the C++ language was still evolving. Now that
9788 the C++ standard is complete, some of those features are superseded by
9789 superior alternatives. Using the old features might cause a warning in
9790 some cases that the feature will be dropped in the future. In other
9791 cases, the feature might be gone already.
9793 While the list below is not exhaustive, it documents some of the options
9794 that are now deprecated:
9797 @item -fexternal-templates
9798 @itemx -falt-external-templates
9799 These are two of the many ways for G++ to implement template
9800 instantiation. @xref{Template Instantiation}. The C++ standard clearly
9801 defines how template definitions have to be organized across
9802 implementation units. G++ has an implicit instantiation mechanism that
9803 should work just fine for standard-conforming code.
9805 @item -fstrict-prototype
9806 @itemx -fno-strict-prototype
9807 Previously it was possible to use an empty prototype parameter list to
9808 indicate an unspecified number of parameters (like C), rather than no
9809 parameters, as C++ demands. This feature has been removed, except where
9810 it is required for backwards compatibility @xref{Backwards Compatibility}.
9813 G++ allows a virtual function returning @samp{void *} to be overridden
9814 by one returning a different pointer type. This extension to the
9815 covariant return type rules is now deprecated and will be removed from a
9818 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
9819 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
9820 and will be removed in a future version. Code using these operators
9821 should be modified to use @code{std::min} and @code{std::max} instead.
9823 The named return value extension has been deprecated, and is now
9826 The use of initializer lists with new expressions has been deprecated,
9827 and is now removed from G++.
9829 Floating and complex non-type template parameters have been deprecated,
9830 and are now removed from G++.
9832 The implicit typename extension has been deprecated and is now
9835 The use of default arguments in function pointers, function typedefs and
9836 and other places where they are not permitted by the standard is
9837 deprecated and will be removed from a future version of G++.
9839 G++ allows floating-point literals to appear in integral constant expressions,
9840 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
9841 This extension is deprecated and will be removed from a future version.
9843 G++ allows static data members of const floating-point type to be declared
9844 with an initializer in a class definition. The standard only allows
9845 initializers for static members of const integral types and const
9846 enumeration types so this extension has been deprecated and will be removed
9847 from a future version.
9849 @node Backwards Compatibility
9850 @section Backwards Compatibility
9851 @cindex Backwards Compatibility
9852 @cindex ARM [Annotated C++ Reference Manual]
9854 Now that there is a definitive ISO standard C++, G++ has a specification
9855 to adhere to. The C++ language evolved over time, and features that
9856 used to be acceptable in previous drafts of the standard, such as the ARM
9857 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
9858 compilation of C++ written to such drafts, G++ contains some backwards
9859 compatibilities. @emph{All such backwards compatibility features are
9860 liable to disappear in future versions of G++.} They should be considered
9861 deprecated @xref{Deprecated Features}.
9865 If a variable is declared at for scope, it used to remain in scope until
9866 the end of the scope which contained the for statement (rather than just
9867 within the for scope). G++ retains this, but issues a warning, if such a
9868 variable is accessed outside the for scope.
9870 @item Implicit C language
9871 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
9872 scope to set the language. On such systems, all header files are
9873 implicitly scoped inside a C language scope. Also, an empty prototype
9874 @code{()} will be treated as an unspecified number of arguments, rather
9875 than no arguments, as C++ demands.