1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C and Objective-C@. Most of them are
20 also available in C++. @xref{C++ Extensions,,Extensions to the
21 C++ Language}, for extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Decimal Float:: Decimal Floating Types.
37 * Hex Floats:: Hexadecimal floating-point constants.
38 * Zero Length:: Zero-length arrays.
39 * Variable Length:: Arrays whose length is computed at run time.
40 * Empty Structures:: Structures with no members.
41 * Variadic Macros:: Macros with a variable number of arguments.
42 * Escaped Newlines:: Slightly looser rules for escaped newlines.
43 * Subscripting:: Any array can be subscripted, even if not an lvalue.
44 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
45 * Initializers:: Non-constant initializers.
46 * Compound Literals:: Compound literals give structures, unions
48 * Designated Inits:: Labeling elements of initializers.
49 * Cast to Union:: Casting to union type from any member of the union.
50 * Case Ranges:: `case 1 ... 9' and such.
51 * Mixed Declarations:: Mixing declarations and code.
52 * Function Attributes:: Declaring that functions have no side effects,
53 or that they can never return.
54 * Attribute Syntax:: Formal syntax for attributes.
55 * Function Prototypes:: Prototype declarations and old-style definitions.
56 * C++ Comments:: C++ comments are recognized.
57 * Dollar Signs:: Dollar sign is allowed in identifiers.
58 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Alignment:: Inquiring about the alignment of a type or variable.
62 * Inline:: Defining inline functions (as fast as macros).
63 * Extended Asm:: Assembler instructions with C expressions as operands.
64 (With them you can define ``built-in'' functions.)
65 * Constraints:: Constraints for asm operands
66 * Asm Labels:: Specifying the assembler name to use for a C symbol.
67 * Explicit Reg Vars:: Defining variables residing in specified registers.
68 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
69 * Incomplete Enums:: @code{enum foo;}, with details to follow.
70 * Function Names:: Printable strings which are the name of the current
72 * Return Address:: Getting the return or frame address of a function.
73 * Vector Extensions:: Using vector instructions through built-in functions.
74 * Offsetof:: Special syntax for implementing @code{offsetof}.
75 * Atomic Builtins:: Built-in functions for atomic memory access.
76 * Object Size Checking:: Built-in functions for limited buffer overflow
78 * Other Builtins:: Other built-in functions.
79 * Target Builtins:: Built-in functions specific to particular targets.
80 * Target Format Checks:: Format checks specific to particular targets.
81 * Pragmas:: Pragmas accepted by GCC.
82 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
83 * Thread-Local:: Per-thread variables.
87 @section Statements and Declarations in Expressions
88 @cindex statements inside expressions
89 @cindex declarations inside expressions
90 @cindex expressions containing statements
91 @cindex macros, statements in expressions
93 @c the above section title wrapped and causes an underfull hbox.. i
94 @c changed it from "within" to "in". --mew 4feb93
95 A compound statement enclosed in parentheses may appear as an expression
96 in GNU C@. This allows you to use loops, switches, and local variables
99 Recall that a compound statement is a sequence of statements surrounded
100 by braces; in this construct, parentheses go around the braces. For
104 (@{ int y = foo (); int z;
111 is a valid (though slightly more complex than necessary) expression
112 for the absolute value of @code{foo ()}.
114 The last thing in the compound statement should be an expression
115 followed by a semicolon; the value of this subexpression serves as the
116 value of the entire construct. (If you use some other kind of statement
117 last within the braces, the construct has type @code{void}, and thus
118 effectively no value.)
120 This feature is especially useful in making macro definitions ``safe'' (so
121 that they evaluate each operand exactly once). For example, the
122 ``maximum'' function is commonly defined as a macro in standard C as
126 #define max(a,b) ((a) > (b) ? (a) : (b))
130 @cindex side effects, macro argument
131 But this definition computes either @var{a} or @var{b} twice, with bad
132 results if the operand has side effects. In GNU C, if you know the
133 type of the operands (here taken as @code{int}), you can define
134 the macro safely as follows:
137 #define maxint(a,b) \
138 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
141 Embedded statements are not allowed in constant expressions, such as
142 the value of an enumeration constant, the width of a bit-field, or
143 the initial value of a static variable.
145 If you don't know the type of the operand, you can still do this, but you
146 must use @code{typeof} (@pxref{Typeof}).
148 In G++, the result value of a statement expression undergoes array and
149 function pointer decay, and is returned by value to the enclosing
150 expression. For instance, if @code{A} is a class, then
159 will construct a temporary @code{A} object to hold the result of the
160 statement expression, and that will be used to invoke @code{Foo}.
161 Therefore the @code{this} pointer observed by @code{Foo} will not be the
164 Any temporaries created within a statement within a statement expression
165 will be destroyed at the statement's end. This makes statement
166 expressions inside macros slightly different from function calls. In
167 the latter case temporaries introduced during argument evaluation will
168 be destroyed at the end of the statement that includes the function
169 call. In the statement expression case they will be destroyed during
170 the statement expression. For instance,
173 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
174 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
184 will have different places where temporaries are destroyed. For the
185 @code{macro} case, the temporary @code{X} will be destroyed just after
186 the initialization of @code{b}. In the @code{function} case that
187 temporary will be destroyed when the function returns.
189 These considerations mean that it is probably a bad idea to use
190 statement-expressions of this form in header files that are designed to
191 work with C++. (Note that some versions of the GNU C Library contained
192 header files using statement-expression that lead to precisely this
195 Jumping into a statement expression with @code{goto} or using a
196 @code{switch} statement outside the statement expression with a
197 @code{case} or @code{default} label inside the statement expression is
198 not permitted. Jumping into a statement expression with a computed
199 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
200 Jumping out of a statement expression is permitted, but if the
201 statement expression is part of a larger expression then it is
202 unspecified which other subexpressions of that expression have been
203 evaluated except where the language definition requires certain
204 subexpressions to be evaluated before or after the statement
205 expression. In any case, as with a function call the evaluation of a
206 statement expression is not interleaved with the evaluation of other
207 parts of the containing expression. For example,
210 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
214 will call @code{foo} and @code{bar1} and will not call @code{baz} but
215 may or may not call @code{bar2}. If @code{bar2} is called, it will be
216 called after @code{foo} and before @code{bar1}
219 @section Locally Declared Labels
221 @cindex macros, local labels
223 GCC allows you to declare @dfn{local labels} in any nested block
224 scope. A local label is just like an ordinary label, but you can
225 only reference it (with a @code{goto} statement, or by taking its
226 address) within the block in which it was declared.
228 A local label declaration looks like this:
231 __label__ @var{label};
238 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
241 Local label declarations must come at the beginning of the block,
242 before any ordinary declarations or statements.
244 The label declaration defines the label @emph{name}, but does not define
245 the label itself. You must do this in the usual way, with
246 @code{@var{label}:}, within the statements of the statement expression.
248 The local label feature is useful for complex macros. If a macro
249 contains nested loops, a @code{goto} can be useful for breaking out of
250 them. However, an ordinary label whose scope is the whole function
251 cannot be used: if the macro can be expanded several times in one
252 function, the label will be multiply defined in that function. A
253 local label avoids this problem. For example:
256 #define SEARCH(value, array, target) \
259 typeof (target) _SEARCH_target = (target); \
260 typeof (*(array)) *_SEARCH_array = (array); \
263 for (i = 0; i < max; i++) \
264 for (j = 0; j < max; j++) \
265 if (_SEARCH_array[i][j] == _SEARCH_target) \
266 @{ (value) = i; goto found; @} \
272 This could also be written using a statement-expression:
275 #define SEARCH(array, target) \
278 typeof (target) _SEARCH_target = (target); \
279 typeof (*(array)) *_SEARCH_array = (array); \
282 for (i = 0; i < max; i++) \
283 for (j = 0; j < max; j++) \
284 if (_SEARCH_array[i][j] == _SEARCH_target) \
285 @{ value = i; goto found; @} \
292 Local label declarations also make the labels they declare visible to
293 nested functions, if there are any. @xref{Nested Functions}, for details.
295 @node Labels as Values
296 @section Labels as Values
297 @cindex labels as values
298 @cindex computed gotos
299 @cindex goto with computed label
300 @cindex address of a label
302 You can get the address of a label defined in the current function
303 (or a containing function) with the unary operator @samp{&&}. The
304 value has type @code{void *}. This value is a constant and can be used
305 wherever a constant of that type is valid. For example:
313 To use these values, you need to be able to jump to one. This is done
314 with the computed goto statement@footnote{The analogous feature in
315 Fortran is called an assigned goto, but that name seems inappropriate in
316 C, where one can do more than simply store label addresses in label
317 variables.}, @code{goto *@var{exp};}. For example,
324 Any expression of type @code{void *} is allowed.
326 One way of using these constants is in initializing a static array that
327 will serve as a jump table:
330 static void *array[] = @{ &&foo, &&bar, &&hack @};
333 Then you can select a label with indexing, like this:
340 Note that this does not check whether the subscript is in bounds---array
341 indexing in C never does that.
343 Such an array of label values serves a purpose much like that of the
344 @code{switch} statement. The @code{switch} statement is cleaner, so
345 use that rather than an array unless the problem does not fit a
346 @code{switch} statement very well.
348 Another use of label values is in an interpreter for threaded code.
349 The labels within the interpreter function can be stored in the
350 threaded code for super-fast dispatching.
352 You may not use this mechanism to jump to code in a different function.
353 If you do that, totally unpredictable things will happen. The best way to
354 avoid this is to store the label address only in automatic variables and
355 never pass it as an argument.
357 An alternate way to write the above example is
360 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
362 goto *(&&foo + array[i]);
366 This is more friendly to code living in shared libraries, as it reduces
367 the number of dynamic relocations that are needed, and by consequence,
368 allows the data to be read-only.
370 @node Nested Functions
371 @section Nested Functions
372 @cindex nested functions
373 @cindex downward funargs
376 A @dfn{nested function} is a function defined inside another function.
377 (Nested functions are not supported for GNU C++.) The nested function's
378 name is local to the block where it is defined. For example, here we
379 define a nested function named @code{square}, and call it twice:
383 foo (double a, double b)
385 double square (double z) @{ return z * z; @}
387 return square (a) + square (b);
392 The nested function can access all the variables of the containing
393 function that are visible at the point of its definition. This is
394 called @dfn{lexical scoping}. For example, here we show a nested
395 function which uses an inherited variable named @code{offset}:
399 bar (int *array, int offset, int size)
401 int access (int *array, int index)
402 @{ return array[index + offset]; @}
405 for (i = 0; i < size; i++)
406 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
411 Nested function definitions are permitted within functions in the places
412 where variable definitions are allowed; that is, in any block, mixed
413 with the other declarations and statements in the block.
415 It is possible to call the nested function from outside the scope of its
416 name by storing its address or passing the address to another function:
419 hack (int *array, int size)
421 void store (int index, int value)
422 @{ array[index] = value; @}
424 intermediate (store, size);
428 Here, the function @code{intermediate} receives the address of
429 @code{store} as an argument. If @code{intermediate} calls @code{store},
430 the arguments given to @code{store} are used to store into @code{array}.
431 But this technique works only so long as the containing function
432 (@code{hack}, in this example) does not exit.
434 If you try to call the nested function through its address after the
435 containing function has exited, all hell will break loose. If you try
436 to call it after a containing scope level has exited, and if it refers
437 to some of the variables that are no longer in scope, you may be lucky,
438 but it's not wise to take the risk. If, however, the nested function
439 does not refer to anything that has gone out of scope, you should be
442 GCC implements taking the address of a nested function using a technique
443 called @dfn{trampolines}. A paper describing them is available as
446 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
448 A nested function can jump to a label inherited from a containing
449 function, provided the label was explicitly declared in the containing
450 function (@pxref{Local Labels}). Such a jump returns instantly to the
451 containing function, exiting the nested function which did the
452 @code{goto} and any intermediate functions as well. Here is an example:
456 bar (int *array, int offset, int size)
459 int access (int *array, int index)
463 return array[index + offset];
467 for (i = 0; i < size; i++)
468 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
472 /* @r{Control comes here from @code{access}
473 if it detects an error.} */
480 A nested function always has no linkage. Declaring one with
481 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
482 before its definition, use @code{auto} (which is otherwise meaningless
483 for function declarations).
486 bar (int *array, int offset, int size)
489 auto int access (int *, int);
491 int access (int *array, int index)
495 return array[index + offset];
501 @node Constructing Calls
502 @section Constructing Function Calls
503 @cindex constructing calls
504 @cindex forwarding calls
506 Using the built-in functions described below, you can record
507 the arguments a function received, and call another function
508 with the same arguments, without knowing the number or types
511 You can also record the return value of that function call,
512 and later return that value, without knowing what data type
513 the function tried to return (as long as your caller expects
516 However, these built-in functions may interact badly with some
517 sophisticated features or other extensions of the language. It
518 is, therefore, not recommended to use them outside very simple
519 functions acting as mere forwarders for their arguments.
521 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
522 This built-in function returns a pointer to data
523 describing how to perform a call with the same arguments as were passed
524 to the current function.
526 The function saves the arg pointer register, structure value address,
527 and all registers that might be used to pass arguments to a function
528 into a block of memory allocated on the stack. Then it returns the
529 address of that block.
532 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
533 This built-in function invokes @var{function}
534 with a copy of the parameters described by @var{arguments}
537 The value of @var{arguments} should be the value returned by
538 @code{__builtin_apply_args}. The argument @var{size} specifies the size
539 of the stack argument data, in bytes.
541 This function returns a pointer to data describing
542 how to return whatever value was returned by @var{function}. The data
543 is saved in a block of memory allocated on the stack.
545 It is not always simple to compute the proper value for @var{size}. The
546 value is used by @code{__builtin_apply} to compute the amount of data
547 that should be pushed on the stack and copied from the incoming argument
551 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
552 This built-in function returns the value described by @var{result} from
553 the containing function. You should specify, for @var{result}, a value
554 returned by @code{__builtin_apply}.
558 @section Referring to a Type with @code{typeof}
561 @cindex macros, types of arguments
563 Another way to refer to the type of an expression is with @code{typeof}.
564 The syntax of using of this keyword looks like @code{sizeof}, but the
565 construct acts semantically like a type name defined with @code{typedef}.
567 There are two ways of writing the argument to @code{typeof}: with an
568 expression or with a type. Here is an example with an expression:
575 This assumes that @code{x} is an array of pointers to functions;
576 the type described is that of the values of the functions.
578 Here is an example with a typename as the argument:
585 Here the type described is that of pointers to @code{int}.
587 If you are writing a header file that must work when included in ISO C
588 programs, write @code{__typeof__} instead of @code{typeof}.
589 @xref{Alternate Keywords}.
591 A @code{typeof}-construct can be used anywhere a typedef name could be
592 used. For example, you can use it in a declaration, in a cast, or inside
593 of @code{sizeof} or @code{typeof}.
595 @code{typeof} is often useful in conjunction with the
596 statements-within-expressions feature. Here is how the two together can
597 be used to define a safe ``maximum'' macro that operates on any
598 arithmetic type and evaluates each of its arguments exactly once:
602 (@{ typeof (a) _a = (a); \
603 typeof (b) _b = (b); \
604 _a > _b ? _a : _b; @})
607 @cindex underscores in variables in macros
608 @cindex @samp{_} in variables in macros
609 @cindex local variables in macros
610 @cindex variables, local, in macros
611 @cindex macros, local variables in
613 The reason for using names that start with underscores for the local
614 variables is to avoid conflicts with variable names that occur within the
615 expressions that are substituted for @code{a} and @code{b}. Eventually we
616 hope to design a new form of declaration syntax that allows you to declare
617 variables whose scopes start only after their initializers; this will be a
618 more reliable way to prevent such conflicts.
621 Some more examples of the use of @code{typeof}:
625 This declares @code{y} with the type of what @code{x} points to.
632 This declares @code{y} as an array of such values.
639 This declares @code{y} as an array of pointers to characters:
642 typeof (typeof (char *)[4]) y;
646 It is equivalent to the following traditional C declaration:
652 To see the meaning of the declaration using @code{typeof}, and why it
653 might be a useful way to write, rewrite it with these macros:
656 #define pointer(T) typeof(T *)
657 #define array(T, N) typeof(T [N])
661 Now the declaration can be rewritten this way:
664 array (pointer (char), 4) y;
668 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
669 pointers to @code{char}.
672 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
673 a more limited extension which permitted one to write
676 typedef @var{T} = @var{expr};
680 with the effect of declaring @var{T} to have the type of the expression
681 @var{expr}. This extension does not work with GCC 3 (versions between
682 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
683 relies on it should be rewritten to use @code{typeof}:
686 typedef typeof(@var{expr}) @var{T};
690 This will work with all versions of GCC@.
693 @section Conditionals with Omitted Operands
694 @cindex conditional expressions, extensions
695 @cindex omitted middle-operands
696 @cindex middle-operands, omitted
697 @cindex extensions, @code{?:}
698 @cindex @code{?:} extensions
700 The middle operand in a conditional expression may be omitted. Then
701 if the first operand is nonzero, its value is the value of the conditional
704 Therefore, the expression
711 has the value of @code{x} if that is nonzero; otherwise, the value of
714 This example is perfectly equivalent to
720 @cindex side effect in ?:
721 @cindex ?: side effect
723 In this simple case, the ability to omit the middle operand is not
724 especially useful. When it becomes useful is when the first operand does,
725 or may (if it is a macro argument), contain a side effect. Then repeating
726 the operand in the middle would perform the side effect twice. Omitting
727 the middle operand uses the value already computed without the undesirable
728 effects of recomputing it.
731 @section Double-Word Integers
732 @cindex @code{long long} data types
733 @cindex double-word arithmetic
734 @cindex multiprecision arithmetic
735 @cindex @code{LL} integer suffix
736 @cindex @code{ULL} integer suffix
738 ISO C99 supports data types for integers that are at least 64 bits wide,
739 and as an extension GCC supports them in C89 mode and in C++.
740 Simply write @code{long long int} for a signed integer, or
741 @code{unsigned long long int} for an unsigned integer. To make an
742 integer constant of type @code{long long int}, add the suffix @samp{LL}
743 to the integer. To make an integer constant of type @code{unsigned long
744 long int}, add the suffix @samp{ULL} to the integer.
746 You can use these types in arithmetic like any other integer types.
747 Addition, subtraction, and bitwise boolean operations on these types
748 are open-coded on all types of machines. Multiplication is open-coded
749 if the machine supports fullword-to-doubleword a widening multiply
750 instruction. Division and shifts are open-coded only on machines that
751 provide special support. The operations that are not open-coded use
752 special library routines that come with GCC@.
754 There may be pitfalls when you use @code{long long} types for function
755 arguments, unless you declare function prototypes. If a function
756 expects type @code{int} for its argument, and you pass a value of type
757 @code{long long int}, confusion will result because the caller and the
758 subroutine will disagree about the number of bytes for the argument.
759 Likewise, if the function expects @code{long long int} and you pass
760 @code{int}. The best way to avoid such problems is to use prototypes.
763 @section Complex Numbers
764 @cindex complex numbers
765 @cindex @code{_Complex} keyword
766 @cindex @code{__complex__} keyword
768 ISO C99 supports complex floating data types, and as an extension GCC
769 supports them in C89 mode and in C++, and supports complex integer data
770 types which are not part of ISO C99. You can declare complex types
771 using the keyword @code{_Complex}. As an extension, the older GNU
772 keyword @code{__complex__} is also supported.
774 For example, @samp{_Complex double x;} declares @code{x} as a
775 variable whose real part and imaginary part are both of type
776 @code{double}. @samp{_Complex short int y;} declares @code{y} to
777 have real and imaginary parts of type @code{short int}; this is not
778 likely to be useful, but it shows that the set of complex types is
781 To write a constant with a complex data type, use the suffix @samp{i} or
782 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
783 has type @code{_Complex float} and @code{3i} has type
784 @code{_Complex int}. Such a constant always has a pure imaginary
785 value, but you can form any complex value you like by adding one to a
786 real constant. This is a GNU extension; if you have an ISO C99
787 conforming C library (such as GNU libc), and want to construct complex
788 constants of floating type, you should include @code{<complex.h>} and
789 use the macros @code{I} or @code{_Complex_I} instead.
791 @cindex @code{__real__} keyword
792 @cindex @code{__imag__} keyword
793 To extract the real part of a complex-valued expression @var{exp}, write
794 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
795 extract the imaginary part. This is a GNU extension; for values of
796 floating type, you should use the ISO C99 functions @code{crealf},
797 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
798 @code{cimagl}, declared in @code{<complex.h>} and also provided as
799 built-in functions by GCC@.
801 @cindex complex conjugation
802 The operator @samp{~} performs complex conjugation when used on a value
803 with a complex type. This is a GNU extension; for values of
804 floating type, you should use the ISO C99 functions @code{conjf},
805 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
806 provided as built-in functions by GCC@.
808 GCC can allocate complex automatic variables in a noncontiguous
809 fashion; it's even possible for the real part to be in a register while
810 the imaginary part is on the stack (or vice-versa). Only the DWARF2
811 debug info format can represent this, so use of DWARF2 is recommended.
812 If you are using the stabs debug info format, GCC describes a noncontiguous
813 complex variable as if it were two separate variables of noncomplex type.
814 If the variable's actual name is @code{foo}, the two fictitious
815 variables are named @code{foo$real} and @code{foo$imag}. You can
816 examine and set these two fictitious variables with your debugger.
819 @section Decimal Floating Types
820 @cindex decimal floating types
821 @cindex @code{_Decimal32} data type
822 @cindex @code{_Decimal64} data type
823 @cindex @code{_Decimal128} data type
824 @cindex @code{df} integer suffix
825 @cindex @code{dd} integer suffix
826 @cindex @code{dl} integer suffix
827 @cindex @code{DF} integer suffix
828 @cindex @code{DD} integer suffix
829 @cindex @code{DL} integer suffix
831 As an extension, the GNU C compiler supports decimal floating types as
832 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
833 floating types in GCC will evolve as the draft technical report changes.
834 Calling conventions for any target might also change. Not all targets
835 support decimal floating types.
837 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
838 @code{_Decimal128}. They use a radix of ten, unlike the floating types
839 @code{float}, @code{double}, and @code{long double} whose radix is not
840 specified by the C standard but is usually two.
842 Support for decimal floating types includes the arithmetic operators
843 add, subtract, multiply, divide; unary arithmetic operators;
844 relational operators; equality operators; and conversions to and from
845 integer and other floating types. Use a suffix @samp{df} or
846 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
847 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
850 GCC support of decimal float as specified by the draft technical report
855 Translation time data type (TTDT) is not supported.
858 When the value of a decimal floating type cannot be represented in the
859 integer type to which it is being converted, the result is undefined
860 rather than the result value specified by the draft technical report.
863 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
864 are supported by the DWARF2 debug information format.
870 ISO C99 supports floating-point numbers written not only in the usual
871 decimal notation, such as @code{1.55e1}, but also numbers such as
872 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
873 supports this in C89 mode (except in some cases when strictly
874 conforming) and in C++. In that format the
875 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
876 mandatory. The exponent is a decimal number that indicates the power of
877 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
884 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
885 is the same as @code{1.55e1}.
887 Unlike for floating-point numbers in the decimal notation the exponent
888 is always required in the hexadecimal notation. Otherwise the compiler
889 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
890 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
891 extension for floating-point constants of type @code{float}.
894 @section Arrays of Length Zero
895 @cindex arrays of length zero
896 @cindex zero-length arrays
897 @cindex length-zero arrays
898 @cindex flexible array members
900 Zero-length arrays are allowed in GNU C@. They are very useful as the
901 last element of a structure which is really a header for a variable-length
910 struct line *thisline = (struct line *)
911 malloc (sizeof (struct line) + this_length);
912 thisline->length = this_length;
915 In ISO C90, you would have to give @code{contents} a length of 1, which
916 means either you waste space or complicate the argument to @code{malloc}.
918 In ISO C99, you would use a @dfn{flexible array member}, which is
919 slightly different in syntax and semantics:
923 Flexible array members are written as @code{contents[]} without
927 Flexible array members have incomplete type, and so the @code{sizeof}
928 operator may not be applied. As a quirk of the original implementation
929 of zero-length arrays, @code{sizeof} evaluates to zero.
932 Flexible array members may only appear as the last member of a
933 @code{struct} that is otherwise non-empty.
936 A structure containing a flexible array member, or a union containing
937 such a structure (possibly recursively), may not be a member of a
938 structure or an element of an array. (However, these uses are
939 permitted by GCC as extensions.)
942 GCC versions before 3.0 allowed zero-length arrays to be statically
943 initialized, as if they were flexible arrays. In addition to those
944 cases that were useful, it also allowed initializations in situations
945 that would corrupt later data. Non-empty initialization of zero-length
946 arrays is now treated like any case where there are more initializer
947 elements than the array holds, in that a suitable warning about "excess
948 elements in array" is given, and the excess elements (all of them, in
949 this case) are ignored.
951 Instead GCC allows static initialization of flexible array members.
952 This is equivalent to defining a new structure containing the original
953 structure followed by an array of sufficient size to contain the data.
954 I.e.@: in the following, @code{f1} is constructed as if it were declared
960 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
963 struct f1 f1; int data[3];
964 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
968 The convenience of this extension is that @code{f1} has the desired
969 type, eliminating the need to consistently refer to @code{f2.f1}.
971 This has symmetry with normal static arrays, in that an array of
972 unknown size is also written with @code{[]}.
974 Of course, this extension only makes sense if the extra data comes at
975 the end of a top-level object, as otherwise we would be overwriting
976 data at subsequent offsets. To avoid undue complication and confusion
977 with initialization of deeply nested arrays, we simply disallow any
978 non-empty initialization except when the structure is the top-level
982 struct foo @{ int x; int y[]; @};
983 struct bar @{ struct foo z; @};
985 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
986 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
987 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
988 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
991 @node Empty Structures
992 @section Structures With No Members
993 @cindex empty structures
994 @cindex zero-size structures
996 GCC permits a C structure to have no members:
1003 The structure will have size zero. In C++, empty structures are part
1004 of the language. G++ treats empty structures as if they had a single
1005 member of type @code{char}.
1007 @node Variable Length
1008 @section Arrays of Variable Length
1009 @cindex variable-length arrays
1010 @cindex arrays of variable length
1013 Variable-length automatic arrays are allowed in ISO C99, and as an
1014 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1015 implementation of variable-length arrays does not yet conform in detail
1016 to the ISO C99 standard.) These arrays are
1017 declared like any other automatic arrays, but with a length that is not
1018 a constant expression. The storage is allocated at the point of
1019 declaration and deallocated when the brace-level is exited. For
1024 concat_fopen (char *s1, char *s2, char *mode)
1026 char str[strlen (s1) + strlen (s2) + 1];
1029 return fopen (str, mode);
1033 @cindex scope of a variable length array
1034 @cindex variable-length array scope
1035 @cindex deallocating variable length arrays
1036 Jumping or breaking out of the scope of the array name deallocates the
1037 storage. Jumping into the scope is not allowed; you get an error
1040 @cindex @code{alloca} vs variable-length arrays
1041 You can use the function @code{alloca} to get an effect much like
1042 variable-length arrays. The function @code{alloca} is available in
1043 many other C implementations (but not in all). On the other hand,
1044 variable-length arrays are more elegant.
1046 There are other differences between these two methods. Space allocated
1047 with @code{alloca} exists until the containing @emph{function} returns.
1048 The space for a variable-length array is deallocated as soon as the array
1049 name's scope ends. (If you use both variable-length arrays and
1050 @code{alloca} in the same function, deallocation of a variable-length array
1051 will also deallocate anything more recently allocated with @code{alloca}.)
1053 You can also use variable-length arrays as arguments to functions:
1057 tester (int len, char data[len][len])
1063 The length of an array is computed once when the storage is allocated
1064 and is remembered for the scope of the array in case you access it with
1067 If you want to pass the array first and the length afterward, you can
1068 use a forward declaration in the parameter list---another GNU extension.
1072 tester (int len; char data[len][len], int len)
1078 @cindex parameter forward declaration
1079 The @samp{int len} before the semicolon is a @dfn{parameter forward
1080 declaration}, and it serves the purpose of making the name @code{len}
1081 known when the declaration of @code{data} is parsed.
1083 You can write any number of such parameter forward declarations in the
1084 parameter list. They can be separated by commas or semicolons, but the
1085 last one must end with a semicolon, which is followed by the ``real''
1086 parameter declarations. Each forward declaration must match a ``real''
1087 declaration in parameter name and data type. ISO C99 does not support
1088 parameter forward declarations.
1090 @node Variadic Macros
1091 @section Macros with a Variable Number of Arguments.
1092 @cindex variable number of arguments
1093 @cindex macro with variable arguments
1094 @cindex rest argument (in macro)
1095 @cindex variadic macros
1097 In the ISO C standard of 1999, a macro can be declared to accept a
1098 variable number of arguments much as a function can. The syntax for
1099 defining the macro is similar to that of a function. Here is an
1103 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1106 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1107 such a macro, it represents the zero or more tokens until the closing
1108 parenthesis that ends the invocation, including any commas. This set of
1109 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1110 wherever it appears. See the CPP manual for more information.
1112 GCC has long supported variadic macros, and used a different syntax that
1113 allowed you to give a name to the variable arguments just like any other
1114 argument. Here is an example:
1117 #define debug(format, args...) fprintf (stderr, format, args)
1120 This is in all ways equivalent to the ISO C example above, but arguably
1121 more readable and descriptive.
1123 GNU CPP has two further variadic macro extensions, and permits them to
1124 be used with either of the above forms of macro definition.
1126 In standard C, you are not allowed to leave the variable argument out
1127 entirely; but you are allowed to pass an empty argument. For example,
1128 this invocation is invalid in ISO C, because there is no comma after
1135 GNU CPP permits you to completely omit the variable arguments in this
1136 way. In the above examples, the compiler would complain, though since
1137 the expansion of the macro still has the extra comma after the format
1140 To help solve this problem, CPP behaves specially for variable arguments
1141 used with the token paste operator, @samp{##}. If instead you write
1144 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1147 and if the variable arguments are omitted or empty, the @samp{##}
1148 operator causes the preprocessor to remove the comma before it. If you
1149 do provide some variable arguments in your macro invocation, GNU CPP
1150 does not complain about the paste operation and instead places the
1151 variable arguments after the comma. Just like any other pasted macro
1152 argument, these arguments are not macro expanded.
1154 @node Escaped Newlines
1155 @section Slightly Looser Rules for Escaped Newlines
1156 @cindex escaped newlines
1157 @cindex newlines (escaped)
1159 Recently, the preprocessor has relaxed its treatment of escaped
1160 newlines. Previously, the newline had to immediately follow a
1161 backslash. The current implementation allows whitespace in the form
1162 of spaces, horizontal and vertical tabs, and form feeds between the
1163 backslash and the subsequent newline. The preprocessor issues a
1164 warning, but treats it as a valid escaped newline and combines the two
1165 lines to form a single logical line. This works within comments and
1166 tokens, as well as between tokens. Comments are @emph{not} treated as
1167 whitespace for the purposes of this relaxation, since they have not
1168 yet been replaced with spaces.
1171 @section Non-Lvalue Arrays May Have Subscripts
1172 @cindex subscripting
1173 @cindex arrays, non-lvalue
1175 @cindex subscripting and function values
1176 In ISO C99, arrays that are not lvalues still decay to pointers, and
1177 may be subscripted, although they may not be modified or used after
1178 the next sequence point and the unary @samp{&} operator may not be
1179 applied to them. As an extension, GCC allows such arrays to be
1180 subscripted in C89 mode, though otherwise they do not decay to
1181 pointers outside C99 mode. For example,
1182 this is valid in GNU C though not valid in C89:
1186 struct foo @{int a[4];@};
1192 return f().a[index];
1198 @section Arithmetic on @code{void}- and Function-Pointers
1199 @cindex void pointers, arithmetic
1200 @cindex void, size of pointer to
1201 @cindex function pointers, arithmetic
1202 @cindex function, size of pointer to
1204 In GNU C, addition and subtraction operations are supported on pointers to
1205 @code{void} and on pointers to functions. This is done by treating the
1206 size of a @code{void} or of a function as 1.
1208 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1209 and on function types, and returns 1.
1211 @opindex Wpointer-arith
1212 The option @option{-Wpointer-arith} requests a warning if these extensions
1216 @section Non-Constant Initializers
1217 @cindex initializers, non-constant
1218 @cindex non-constant initializers
1220 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1221 automatic variable are not required to be constant expressions in GNU C@.
1222 Here is an example of an initializer with run-time varying elements:
1225 foo (float f, float g)
1227 float beat_freqs[2] = @{ f-g, f+g @};
1232 @node Compound Literals
1233 @section Compound Literals
1234 @cindex constructor expressions
1235 @cindex initializations in expressions
1236 @cindex structures, constructor expression
1237 @cindex expressions, constructor
1238 @cindex compound literals
1239 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1241 ISO C99 supports compound literals. A compound literal looks like
1242 a cast containing an initializer. Its value is an object of the
1243 type specified in the cast, containing the elements specified in
1244 the initializer; it is an lvalue. As an extension, GCC supports
1245 compound literals in C89 mode and in C++.
1247 Usually, the specified type is a structure. Assume that
1248 @code{struct foo} and @code{structure} are declared as shown:
1251 struct foo @{int a; char b[2];@} structure;
1255 Here is an example of constructing a @code{struct foo} with a compound literal:
1258 structure = ((struct foo) @{x + y, 'a', 0@});
1262 This is equivalent to writing the following:
1266 struct foo temp = @{x + y, 'a', 0@};
1271 You can also construct an array. If all the elements of the compound literal
1272 are (made up of) simple constant expressions, suitable for use in
1273 initializers of objects of static storage duration, then the compound
1274 literal can be coerced to a pointer to its first element and used in
1275 such an initializer, as shown here:
1278 char **foo = (char *[]) @{ "x", "y", "z" @};
1281 Compound literals for scalar types and union types are is
1282 also allowed, but then the compound literal is equivalent
1285 As a GNU extension, GCC allows initialization of objects with static storage
1286 duration by compound literals (which is not possible in ISO C99, because
1287 the initializer is not a constant).
1288 It is handled as if the object was initialized only with the bracket
1289 enclosed list if the types of the compound literal and the object match.
1290 The initializer list of the compound literal must be constant.
1291 If the object being initialized has array type of unknown size, the size is
1292 determined by compound literal size.
1295 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1296 static int y[] = (int []) @{1, 2, 3@};
1297 static int z[] = (int [3]) @{1@};
1301 The above lines are equivalent to the following:
1303 static struct foo x = @{1, 'a', 'b'@};
1304 static int y[] = @{1, 2, 3@};
1305 static int z[] = @{1, 0, 0@};
1308 @node Designated Inits
1309 @section Designated Initializers
1310 @cindex initializers with labeled elements
1311 @cindex labeled elements in initializers
1312 @cindex case labels in initializers
1313 @cindex designated initializers
1315 Standard C89 requires the elements of an initializer to appear in a fixed
1316 order, the same as the order of the elements in the array or structure
1319 In ISO C99 you can give the elements in any order, specifying the array
1320 indices or structure field names they apply to, and GNU C allows this as
1321 an extension in C89 mode as well. This extension is not
1322 implemented in GNU C++.
1324 To specify an array index, write
1325 @samp{[@var{index}] =} before the element value. For example,
1328 int a[6] = @{ [4] = 29, [2] = 15 @};
1335 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1339 The index values must be constant expressions, even if the array being
1340 initialized is automatic.
1342 An alternative syntax for this which has been obsolete since GCC 2.5 but
1343 GCC still accepts is to write @samp{[@var{index}]} before the element
1344 value, with no @samp{=}.
1346 To initialize a range of elements to the same value, write
1347 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1348 extension. For example,
1351 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1355 If the value in it has side-effects, the side-effects will happen only once,
1356 not for each initialized field by the range initializer.
1359 Note that the length of the array is the highest value specified
1362 In a structure initializer, specify the name of a field to initialize
1363 with @samp{.@var{fieldname} =} before the element value. For example,
1364 given the following structure,
1367 struct point @{ int x, y; @};
1371 the following initialization
1374 struct point p = @{ .y = yvalue, .x = xvalue @};
1381 struct point p = @{ xvalue, yvalue @};
1384 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1385 @samp{@var{fieldname}:}, as shown here:
1388 struct point p = @{ y: yvalue, x: xvalue @};
1392 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1393 @dfn{designator}. You can also use a designator (or the obsolete colon
1394 syntax) when initializing a union, to specify which element of the union
1395 should be used. For example,
1398 union foo @{ int i; double d; @};
1400 union foo f = @{ .d = 4 @};
1404 will convert 4 to a @code{double} to store it in the union using
1405 the second element. By contrast, casting 4 to type @code{union foo}
1406 would store it into the union as the integer @code{i}, since it is
1407 an integer. (@xref{Cast to Union}.)
1409 You can combine this technique of naming elements with ordinary C
1410 initialization of successive elements. Each initializer element that
1411 does not have a designator applies to the next consecutive element of the
1412 array or structure. For example,
1415 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1422 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1425 Labeling the elements of an array initializer is especially useful
1426 when the indices are characters or belong to an @code{enum} type.
1431 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1432 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1435 @cindex designator lists
1436 You can also write a series of @samp{.@var{fieldname}} and
1437 @samp{[@var{index}]} designators before an @samp{=} to specify a
1438 nested subobject to initialize; the list is taken relative to the
1439 subobject corresponding to the closest surrounding brace pair. For
1440 example, with the @samp{struct point} declaration above:
1443 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1447 If the same field is initialized multiple times, it will have value from
1448 the last initialization. If any such overridden initialization has
1449 side-effect, it is unspecified whether the side-effect happens or not.
1450 Currently, GCC will discard them and issue a warning.
1453 @section Case Ranges
1455 @cindex ranges in case statements
1457 You can specify a range of consecutive values in a single @code{case} label,
1461 case @var{low} ... @var{high}:
1465 This has the same effect as the proper number of individual @code{case}
1466 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1468 This feature is especially useful for ranges of ASCII character codes:
1474 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1475 it may be parsed wrong when you use it with integer values. For example,
1490 @section Cast to a Union Type
1491 @cindex cast to a union
1492 @cindex union, casting to a
1494 A cast to union type is similar to other casts, except that the type
1495 specified is a union type. You can specify the type either with
1496 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1497 a constructor though, not a cast, and hence does not yield an lvalue like
1498 normal casts. (@xref{Compound Literals}.)
1500 The types that may be cast to the union type are those of the members
1501 of the union. Thus, given the following union and variables:
1504 union foo @{ int i; double d; @};
1510 both @code{x} and @code{y} can be cast to type @code{union foo}.
1512 Using the cast as the right-hand side of an assignment to a variable of
1513 union type is equivalent to storing in a member of the union:
1518 u = (union foo) x @equiv{} u.i = x
1519 u = (union foo) y @equiv{} u.d = y
1522 You can also use the union cast as a function argument:
1525 void hack (union foo);
1527 hack ((union foo) x);
1530 @node Mixed Declarations
1531 @section Mixed Declarations and Code
1532 @cindex mixed declarations and code
1533 @cindex declarations, mixed with code
1534 @cindex code, mixed with declarations
1536 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1537 within compound statements. As an extension, GCC also allows this in
1538 C89 mode. For example, you could do:
1547 Each identifier is visible from where it is declared until the end of
1548 the enclosing block.
1550 @node Function Attributes
1551 @section Declaring Attributes of Functions
1552 @cindex function attributes
1553 @cindex declaring attributes of functions
1554 @cindex functions that never return
1555 @cindex functions that return more than once
1556 @cindex functions that have no side effects
1557 @cindex functions in arbitrary sections
1558 @cindex functions that behave like malloc
1559 @cindex @code{volatile} applied to function
1560 @cindex @code{const} applied to function
1561 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1562 @cindex functions with non-null pointer arguments
1563 @cindex functions that are passed arguments in registers on the 386
1564 @cindex functions that pop the argument stack on the 386
1565 @cindex functions that do not pop the argument stack on the 386
1567 In GNU C, you declare certain things about functions called in your program
1568 which help the compiler optimize function calls and check your code more
1571 The keyword @code{__attribute__} allows you to specify special
1572 attributes when making a declaration. This keyword is followed by an
1573 attribute specification inside double parentheses. The following
1574 attributes are currently defined for functions on all targets:
1575 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1576 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1577 @code{format}, @code{format_arg}, @code{no_instrument_function},
1578 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1579 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1580 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1581 @code{gnu_inline} and @code{externally_visible}. Several other
1582 attributes are defined for functions on particular target systems. Other
1583 attributes, including @code{section} are supported for variables declarations
1584 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1586 You may also specify attributes with @samp{__} preceding and following
1587 each keyword. This allows you to use them in header files without
1588 being concerned about a possible macro of the same name. For example,
1589 you may use @code{__noreturn__} instead of @code{noreturn}.
1591 @xref{Attribute Syntax}, for details of the exact syntax for using
1595 @c Keep this table alphabetized by attribute name. Treat _ as space.
1597 @item alias ("@var{target}")
1598 @cindex @code{alias} attribute
1599 The @code{alias} attribute causes the declaration to be emitted as an
1600 alias for another symbol, which must be specified. For instance,
1603 void __f () @{ /* @r{Do something.} */; @}
1604 void f () __attribute__ ((weak, alias ("__f")));
1607 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1608 mangled name for the target must be used. It is an error if @samp{__f}
1609 is not defined in the same translation unit.
1611 Not all target machines support this attribute.
1614 @cindex @code{always_inline} function attribute
1615 Generally, functions are not inlined unless optimization is specified.
1616 For functions declared inline, this attribute inlines the function even
1617 if no optimization level was specified.
1620 @cindex @code{gnu_inline} function attribute
1621 This attribute should be used with a function which is also declared
1622 with the @code{inline} keyword. It directs GCC to treat the function
1623 as if it were defined in gnu89 mode even when compiling in C99 or
1626 If the function is declared @code{extern}, then this definition of the
1627 function is used only for inlining. In no case is the function
1628 compiled as a standalone function, not even if you take its address
1629 explicitly. Such an address becomes an external reference, as if you
1630 had only declared the function, and had not defined it. This has
1631 almost the effect of a macro. The way to use this is to put a
1632 function definition in a header file with this attribute, and put
1633 another copy of the function, without @code{extern}, in a library
1634 file. The definition in the header file will cause most calls to the
1635 function to be inlined. If any uses of the function remain, they will
1636 refer to the single copy in the library. Note that the two
1637 definitions of the functions need not be precisely the same, although
1638 if they do not have the same effect your program may behave oddly.
1640 If the function is neither @code{extern} nor @code{static}, then the
1641 function is compiled as a standalone function, as well as being
1642 inlined where possible.
1644 This is how GCC traditionally handled functions declared
1645 @code{inline}. Since ISO C99 specifies a different semantics for
1646 @code{inline}, this function attribute is provided as a transition
1647 measure and as a useful feature in its own right. This attribute is
1648 available in GCC 4.1.3 and later. It is available if either of the
1649 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1650 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1651 Function is As Fast As a Macro}.
1653 @cindex @code{flatten} function attribute
1655 Generally, inlining into a function is limited. For a function marked with
1656 this attribute, every call inside this function will be inlined, if possible.
1657 Whether the function itself is considered for inlining depends on its size and
1658 the current inlining parameters. The @code{flatten} attribute only works
1659 reliably in unit-at-a-time mode.
1662 @cindex functions that do pop the argument stack on the 386
1664 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1665 assume that the calling function will pop off the stack space used to
1666 pass arguments. This is
1667 useful to override the effects of the @option{-mrtd} switch.
1670 @cindex @code{const} function attribute
1671 Many functions do not examine any values except their arguments, and
1672 have no effects except the return value. Basically this is just slightly
1673 more strict class than the @code{pure} attribute below, since function is not
1674 allowed to read global memory.
1676 @cindex pointer arguments
1677 Note that a function that has pointer arguments and examines the data
1678 pointed to must @emph{not} be declared @code{const}. Likewise, a
1679 function that calls a non-@code{const} function usually must not be
1680 @code{const}. It does not make sense for a @code{const} function to
1683 The attribute @code{const} is not implemented in GCC versions earlier
1684 than 2.5. An alternative way to declare that a function has no side
1685 effects, which works in the current version and in some older versions,
1689 typedef int intfn ();
1691 extern const intfn square;
1694 This approach does not work in GNU C++ from 2.6.0 on, since the language
1695 specifies that the @samp{const} must be attached to the return value.
1699 @itemx constructor (@var{priority})
1700 @itemx destructor (@var{priority})
1701 @cindex @code{constructor} function attribute
1702 @cindex @code{destructor} function attribute
1703 The @code{constructor} attribute causes the function to be called
1704 automatically before execution enters @code{main ()}. Similarly, the
1705 @code{destructor} attribute causes the function to be called
1706 automatically after @code{main ()} has completed or @code{exit ()} has
1707 been called. Functions with these attributes are useful for
1708 initializing data that will be used implicitly during the execution of
1711 You may provide an optional integer priority to control the order in
1712 which constructor and destructor functions are run. A constructor
1713 with a smaller priority number runs before a constructor with a larger
1714 priority number; the opposite relationship holds for destructors. So,
1715 if you have a constructor that allocates a resource and a destructor
1716 that deallocates the same resource, both functions typically have the
1717 same priority. The priorities for constructor and destructor
1718 functions are the same as those specified for namespace-scope C++
1719 objects (@pxref{C++ Attributes}).
1721 These attributes are not currently implemented for Objective-C@.
1724 @cindex @code{deprecated} attribute.
1725 The @code{deprecated} attribute results in a warning if the function
1726 is used anywhere in the source file. This is useful when identifying
1727 functions that are expected to be removed in a future version of a
1728 program. The warning also includes the location of the declaration
1729 of the deprecated function, to enable users to easily find further
1730 information about why the function is deprecated, or what they should
1731 do instead. Note that the warnings only occurs for uses:
1734 int old_fn () __attribute__ ((deprecated));
1736 int (*fn_ptr)() = old_fn;
1739 results in a warning on line 3 but not line 2.
1741 The @code{deprecated} attribute can also be used for variables and
1742 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1745 @cindex @code{__declspec(dllexport)}
1746 On Microsoft Windows targets and Symbian OS targets the
1747 @code{dllexport} attribute causes the compiler to provide a global
1748 pointer to a pointer in a DLL, so that it can be referenced with the
1749 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1750 name is formed by combining @code{_imp__} and the function or variable
1753 You can use @code{__declspec(dllexport)} as a synonym for
1754 @code{__attribute__ ((dllexport))} for compatibility with other
1757 On systems that support the @code{visibility} attribute, this
1758 attribute also implies ``default'' visibility, unless a
1759 @code{visibility} attribute is explicitly specified. You should avoid
1760 the use of @code{dllexport} with ``hidden'' or ``internal''
1761 visibility; in the future GCC may issue an error for those cases.
1763 Currently, the @code{dllexport} attribute is ignored for inlined
1764 functions, unless the @option{-fkeep-inline-functions} flag has been
1765 used. The attribute is also ignored for undefined symbols.
1767 When applied to C++ classes, the attribute marks defined non-inlined
1768 member functions and static data members as exports. Static consts
1769 initialized in-class are not marked unless they are also defined
1772 For Microsoft Windows targets there are alternative methods for
1773 including the symbol in the DLL's export table such as using a
1774 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1775 the @option{--export-all} linker flag.
1778 @cindex @code{__declspec(dllimport)}
1779 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1780 attribute causes the compiler to reference a function or variable via
1781 a global pointer to a pointer that is set up by the DLL exporting the
1782 symbol. The attribute implies @code{extern} storage. On Microsoft
1783 Windows targets, the pointer name is formed by combining @code{_imp__}
1784 and the function or variable name.
1786 You can use @code{__declspec(dllimport)} as a synonym for
1787 @code{__attribute__ ((dllimport))} for compatibility with other
1790 Currently, the attribute is ignored for inlined functions. If the
1791 attribute is applied to a symbol @emph{definition}, an error is reported.
1792 If a symbol previously declared @code{dllimport} is later defined, the
1793 attribute is ignored in subsequent references, and a warning is emitted.
1794 The attribute is also overridden by a subsequent declaration as
1797 When applied to C++ classes, the attribute marks non-inlined
1798 member functions and static data members as imports. However, the
1799 attribute is ignored for virtual methods to allow creation of vtables
1802 On the SH Symbian OS target the @code{dllimport} attribute also has
1803 another affect---it can cause the vtable and run-time type information
1804 for a class to be exported. This happens when the class has a
1805 dllimport'ed constructor or a non-inline, non-pure virtual function
1806 and, for either of those two conditions, the class also has a inline
1807 constructor or destructor and has a key function that is defined in
1808 the current translation unit.
1810 For Microsoft Windows based targets the use of the @code{dllimport}
1811 attribute on functions is not necessary, but provides a small
1812 performance benefit by eliminating a thunk in the DLL@. The use of the
1813 @code{dllimport} attribute on imported variables was required on older
1814 versions of the GNU linker, but can now be avoided by passing the
1815 @option{--enable-auto-import} switch to the GNU linker. As with
1816 functions, using the attribute for a variable eliminates a thunk in
1819 One drawback to using this attribute is that a pointer to a function
1820 or variable marked as @code{dllimport} cannot be used as a constant
1821 address. On Microsoft Windows targets, the attribute can be disabled
1822 for functions by setting the @option{-mnop-fun-dllimport} flag.
1825 @cindex eight bit data on the H8/300, H8/300H, and H8S
1826 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1827 variable should be placed into the eight bit data section.
1828 The compiler will generate more efficient code for certain operations
1829 on data in the eight bit data area. Note the eight bit data area is limited to
1832 You must use GAS and GLD from GNU binutils version 2.7 or later for
1833 this attribute to work correctly.
1835 @item exception_handler
1836 @cindex exception handler functions on the Blackfin processor
1837 Use this attribute on the Blackfin to indicate that the specified function
1838 is an exception handler. The compiler will generate function entry and
1839 exit sequences suitable for use in an exception handler when this
1840 attribute is present.
1843 @cindex functions which handle memory bank switching
1844 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1845 use a calling convention that takes care of switching memory banks when
1846 entering and leaving a function. This calling convention is also the
1847 default when using the @option{-mlong-calls} option.
1849 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1850 to call and return from a function.
1852 On 68HC11 the compiler will generate a sequence of instructions
1853 to invoke a board-specific routine to switch the memory bank and call the
1854 real function. The board-specific routine simulates a @code{call}.
1855 At the end of a function, it will jump to a board-specific routine
1856 instead of using @code{rts}. The board-specific return routine simulates
1860 @cindex functions that pop the argument stack on the 386
1861 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1862 pass the first argument (if of integral type) in the register ECX and
1863 the second argument (if of integral type) in the register EDX@. Subsequent
1864 and other typed arguments are passed on the stack. The called function will
1865 pop the arguments off the stack. If the number of arguments is variable all
1866 arguments are pushed on the stack.
1868 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1869 @cindex @code{format} function attribute
1871 The @code{format} attribute specifies that a function takes @code{printf},
1872 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1873 should be type-checked against a format string. For example, the
1878 my_printf (void *my_object, const char *my_format, ...)
1879 __attribute__ ((format (printf, 2, 3)));
1883 causes the compiler to check the arguments in calls to @code{my_printf}
1884 for consistency with the @code{printf} style format string argument
1887 The parameter @var{archetype} determines how the format string is
1888 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1889 or @code{strfmon}. (You can also use @code{__printf__},
1890 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1891 parameter @var{string-index} specifies which argument is the format
1892 string argument (starting from 1), while @var{first-to-check} is the
1893 number of the first argument to check against the format string. For
1894 functions where the arguments are not available to be checked (such as
1895 @code{vprintf}), specify the third parameter as zero. In this case the
1896 compiler only checks the format string for consistency. For
1897 @code{strftime} formats, the third parameter is required to be zero.
1898 Since non-static C++ methods have an implicit @code{this} argument, the
1899 arguments of such methods should be counted from two, not one, when
1900 giving values for @var{string-index} and @var{first-to-check}.
1902 In the example above, the format string (@code{my_format}) is the second
1903 argument of the function @code{my_print}, and the arguments to check
1904 start with the third argument, so the correct parameters for the format
1905 attribute are 2 and 3.
1907 @opindex ffreestanding
1908 @opindex fno-builtin
1909 The @code{format} attribute allows you to identify your own functions
1910 which take format strings as arguments, so that GCC can check the
1911 calls to these functions for errors. The compiler always (unless
1912 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1913 for the standard library functions @code{printf}, @code{fprintf},
1914 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1915 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1916 warnings are requested (using @option{-Wformat}), so there is no need to
1917 modify the header file @file{stdio.h}. In C99 mode, the functions
1918 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1919 @code{vsscanf} are also checked. Except in strictly conforming C
1920 standard modes, the X/Open function @code{strfmon} is also checked as
1921 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1922 @xref{C Dialect Options,,Options Controlling C Dialect}.
1924 The target may provide additional types of format checks.
1925 @xref{Target Format Checks,,Format Checks Specific to Particular
1928 @item format_arg (@var{string-index})
1929 @cindex @code{format_arg} function attribute
1930 @opindex Wformat-nonliteral
1931 The @code{format_arg} attribute specifies that a function takes a format
1932 string for a @code{printf}, @code{scanf}, @code{strftime} or
1933 @code{strfmon} style function and modifies it (for example, to translate
1934 it into another language), so the result can be passed to a
1935 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1936 function (with the remaining arguments to the format function the same
1937 as they would have been for the unmodified string). For example, the
1942 my_dgettext (char *my_domain, const char *my_format)
1943 __attribute__ ((format_arg (2)));
1947 causes the compiler to check the arguments in calls to a @code{printf},
1948 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1949 format string argument is a call to the @code{my_dgettext} function, for
1950 consistency with the format string argument @code{my_format}. If the
1951 @code{format_arg} attribute had not been specified, all the compiler
1952 could tell in such calls to format functions would be that the format
1953 string argument is not constant; this would generate a warning when
1954 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1955 without the attribute.
1957 The parameter @var{string-index} specifies which argument is the format
1958 string argument (starting from one). Since non-static C++ methods have
1959 an implicit @code{this} argument, the arguments of such methods should
1960 be counted from two.
1962 The @code{format-arg} attribute allows you to identify your own
1963 functions which modify format strings, so that GCC can check the
1964 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1965 type function whose operands are a call to one of your own function.
1966 The compiler always treats @code{gettext}, @code{dgettext}, and
1967 @code{dcgettext} in this manner except when strict ISO C support is
1968 requested by @option{-ansi} or an appropriate @option{-std} option, or
1969 @option{-ffreestanding} or @option{-fno-builtin}
1970 is used. @xref{C Dialect Options,,Options
1971 Controlling C Dialect}.
1973 @item function_vector
1974 @cindex calling functions through the function vector on the H8/300 processors
1975 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1976 function should be called through the function vector. Calling a
1977 function through the function vector will reduce code size, however;
1978 the function vector has a limited size (maximum 128 entries on the H8/300
1979 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1981 You must use GAS and GLD from GNU binutils version 2.7 or later for
1982 this attribute to work correctly.
1985 @cindex interrupt handler functions
1986 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
1987 ports to indicate that the specified function is an interrupt handler.
1988 The compiler will generate function entry and exit sequences suitable
1989 for use in an interrupt handler when this attribute is present.
1991 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1992 SH processors can be specified via the @code{interrupt_handler} attribute.
1994 Note, on the AVR, interrupts will be enabled inside the function.
1996 Note, for the ARM, you can specify the kind of interrupt to be handled by
1997 adding an optional parameter to the interrupt attribute like this:
2000 void f () __attribute__ ((interrupt ("IRQ")));
2003 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2005 On ARMv7-M the interrupt type is ignored, and the attribute means the function
2006 may be called with a word aligned stack pointer.
2008 @item interrupt_handler
2009 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2010 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2011 indicate that the specified function is an interrupt handler. The compiler
2012 will generate function entry and exit sequences suitable for use in an
2013 interrupt handler when this attribute is present.
2016 @cindex User stack pointer in interrupts on the Blackfin
2017 When used together with @code{interrupt_handler}, @code{exception_handler}
2018 or @code{nmi_handler}, code will be generated to load the stack pointer
2019 from the USP register in the function prologue.
2021 @item long_call/short_call
2022 @cindex indirect calls on ARM
2023 This attribute specifies how a particular function is called on
2024 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2025 command line switch and @code{#pragma long_calls} settings. The
2026 @code{long_call} attribute indicates that the function might be far
2027 away from the call site and require a different (more expensive)
2028 calling sequence. The @code{short_call} attribute always places
2029 the offset to the function from the call site into the @samp{BL}
2030 instruction directly.
2032 @item longcall/shortcall
2033 @cindex functions called via pointer on the RS/6000 and PowerPC
2034 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2035 indicates that the function might be far away from the call site and
2036 require a different (more expensive) calling sequence. The
2037 @code{shortcall} attribute indicates that the function is always close
2038 enough for the shorter calling sequence to be used. These attributes
2039 override both the @option{-mlongcall} switch and, on the RS/6000 and
2040 PowerPC, the @code{#pragma longcall} setting.
2042 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2043 calls are necessary.
2046 @cindex indirect calls on MIPS
2047 This attribute specifies how a particular function is called on MIPS@.
2048 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2049 command line switch. This attribute causes the compiler to always call
2050 the function by first loading its address into a register, and then using
2051 the contents of that register.
2054 @cindex @code{malloc} attribute
2055 The @code{malloc} attribute is used to tell the compiler that a function
2056 may be treated as if any non-@code{NULL} pointer it returns cannot
2057 alias any other pointer valid when the function returns.
2058 This will often improve optimization.
2059 Standard functions with this property include @code{malloc} and
2060 @code{calloc}. @code{realloc}-like functions have this property as
2061 long as the old pointer is never referred to (including comparing it
2062 to the new pointer) after the function returns a non-@code{NULL}
2065 @item model (@var{model-name})
2066 @cindex function addressability on the M32R/D
2067 @cindex variable addressability on the IA-64
2069 On the M32R/D, use this attribute to set the addressability of an
2070 object, and of the code generated for a function. The identifier
2071 @var{model-name} is one of @code{small}, @code{medium}, or
2072 @code{large}, representing each of the code models.
2074 Small model objects live in the lower 16MB of memory (so that their
2075 addresses can be loaded with the @code{ld24} instruction), and are
2076 callable with the @code{bl} instruction.
2078 Medium model objects may live anywhere in the 32-bit address space (the
2079 compiler will generate @code{seth/add3} instructions to load their addresses),
2080 and are callable with the @code{bl} instruction.
2082 Large model objects may live anywhere in the 32-bit address space (the
2083 compiler will generate @code{seth/add3} instructions to load their addresses),
2084 and may not be reachable with the @code{bl} instruction (the compiler will
2085 generate the much slower @code{seth/add3/jl} instruction sequence).
2087 On IA-64, use this attribute to set the addressability of an object.
2088 At present, the only supported identifier for @var{model-name} is
2089 @code{small}, indicating addressability via ``small'' (22-bit)
2090 addresses (so that their addresses can be loaded with the @code{addl}
2091 instruction). Caveat: such addressing is by definition not position
2092 independent and hence this attribute must not be used for objects
2093 defined by shared libraries.
2096 @cindex function without a prologue/epilogue code
2097 Use this attribute on the ARM, AVR, C4x, IP2K and SPU ports to indicate that
2098 the specified function does not need prologue/epilogue sequences generated by
2099 the compiler. It is up to the programmer to provide these sequences.
2102 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2103 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2104 use the normal calling convention based on @code{jsr} and @code{rts}.
2105 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2109 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2110 Use this attribute together with @code{interrupt_handler},
2111 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2112 entry code should enable nested interrupts or exceptions.
2115 @cindex NMI handler functions on the Blackfin processor
2116 Use this attribute on the Blackfin to indicate that the specified function
2117 is an NMI handler. The compiler will generate function entry and
2118 exit sequences suitable for use in an NMI handler when this
2119 attribute is present.
2121 @item no_instrument_function
2122 @cindex @code{no_instrument_function} function attribute
2123 @opindex finstrument-functions
2124 If @option{-finstrument-functions} is given, profiling function calls will
2125 be generated at entry and exit of most user-compiled functions.
2126 Functions with this attribute will not be so instrumented.
2129 @cindex @code{noinline} function attribute
2130 This function attribute prevents a function from being considered for
2133 @item nonnull (@var{arg-index}, @dots{})
2134 @cindex @code{nonnull} function attribute
2135 The @code{nonnull} attribute specifies that some function parameters should
2136 be non-null pointers. For instance, the declaration:
2140 my_memcpy (void *dest, const void *src, size_t len)
2141 __attribute__((nonnull (1, 2)));
2145 causes the compiler to check that, in calls to @code{my_memcpy},
2146 arguments @var{dest} and @var{src} are non-null. If the compiler
2147 determines that a null pointer is passed in an argument slot marked
2148 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2149 is issued. The compiler may also choose to make optimizations based
2150 on the knowledge that certain function arguments will not be null.
2152 If no argument index list is given to the @code{nonnull} attribute,
2153 all pointer arguments are marked as non-null. To illustrate, the
2154 following declaration is equivalent to the previous example:
2158 my_memcpy (void *dest, const void *src, size_t len)
2159 __attribute__((nonnull));
2163 @cindex @code{noreturn} function attribute
2164 A few standard library functions, such as @code{abort} and @code{exit},
2165 cannot return. GCC knows this automatically. Some programs define
2166 their own functions that never return. You can declare them
2167 @code{noreturn} to tell the compiler this fact. For example,
2171 void fatal () __attribute__ ((noreturn));
2174 fatal (/* @r{@dots{}} */)
2176 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2182 The @code{noreturn} keyword tells the compiler to assume that
2183 @code{fatal} cannot return. It can then optimize without regard to what
2184 would happen if @code{fatal} ever did return. This makes slightly
2185 better code. More importantly, it helps avoid spurious warnings of
2186 uninitialized variables.
2188 The @code{noreturn} keyword does not affect the exceptional path when that
2189 applies: a @code{noreturn}-marked function may still return to the caller
2190 by throwing an exception or calling @code{longjmp}.
2192 Do not assume that registers saved by the calling function are
2193 restored before calling the @code{noreturn} function.
2195 It does not make sense for a @code{noreturn} function to have a return
2196 type other than @code{void}.
2198 The attribute @code{noreturn} is not implemented in GCC versions
2199 earlier than 2.5. An alternative way to declare that a function does
2200 not return, which works in the current version and in some older
2201 versions, is as follows:
2204 typedef void voidfn ();
2206 volatile voidfn fatal;
2209 This approach does not work in GNU C++.
2212 @cindex @code{nothrow} function attribute
2213 The @code{nothrow} attribute is used to inform the compiler that a
2214 function cannot throw an exception. For example, most functions in
2215 the standard C library can be guaranteed not to throw an exception
2216 with the notable exceptions of @code{qsort} and @code{bsearch} that
2217 take function pointer arguments. The @code{nothrow} attribute is not
2218 implemented in GCC versions earlier than 3.3.
2221 @cindex @code{pure} function attribute
2222 Many functions have no effects except the return value and their
2223 return value depends only on the parameters and/or global variables.
2224 Such a function can be subject
2225 to common subexpression elimination and loop optimization just as an
2226 arithmetic operator would be. These functions should be declared
2227 with the attribute @code{pure}. For example,
2230 int square (int) __attribute__ ((pure));
2234 says that the hypothetical function @code{square} is safe to call
2235 fewer times than the program says.
2237 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2238 Interesting non-pure functions are functions with infinite loops or those
2239 depending on volatile memory or other system resource, that may change between
2240 two consecutive calls (such as @code{feof} in a multithreading environment).
2242 The attribute @code{pure} is not implemented in GCC versions earlier
2245 @item regparm (@var{number})
2246 @cindex @code{regparm} attribute
2247 @cindex functions that are passed arguments in registers on the 386
2248 On the Intel 386, the @code{regparm} attribute causes the compiler to
2249 pass arguments number one to @var{number} if they are of integral type
2250 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2251 take a variable number of arguments will continue to be passed all of their
2252 arguments on the stack.
2254 Beware that on some ELF systems this attribute is unsuitable for
2255 global functions in shared libraries with lazy binding (which is the
2256 default). Lazy binding will send the first call via resolving code in
2257 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2258 per the standard calling conventions. Solaris 8 is affected by this.
2259 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2260 safe since the loaders there save all registers. (Lazy binding can be
2261 disabled with the linker or the loader if desired, to avoid the
2265 @cindex @code{sseregparm} attribute
2266 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2267 causes the compiler to pass up to 3 floating point arguments in
2268 SSE registers instead of on the stack. Functions that take a
2269 variable number of arguments will continue to pass all of their
2270 floating point arguments on the stack.
2272 @item force_align_arg_pointer
2273 @cindex @code{force_align_arg_pointer} attribute
2274 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2275 applied to individual function definitions, generating an alternate
2276 prologue and epilogue that realigns the runtime stack. This supports
2277 mixing legacy codes that run with a 4-byte aligned stack with modern
2278 codes that keep a 16-byte stack for SSE compatibility. The alternate
2279 prologue and epilogue are slower and bigger than the regular ones, and
2280 the alternate prologue requires a scratch register; this lowers the
2281 number of registers available if used in conjunction with the
2282 @code{regparm} attribute. The @code{force_align_arg_pointer}
2283 attribute is incompatible with nested functions; this is considered a
2287 @cindex @code{returns_twice} attribute
2288 The @code{returns_twice} attribute tells the compiler that a function may
2289 return more than one time. The compiler will ensure that all registers
2290 are dead before calling such a function and will emit a warning about
2291 the variables that may be clobbered after the second return from the
2292 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2293 The @code{longjmp}-like counterpart of such function, if any, might need
2294 to be marked with the @code{noreturn} attribute.
2297 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2298 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2299 all registers except the stack pointer should be saved in the prologue
2300 regardless of whether they are used or not.
2302 @item section ("@var{section-name}")
2303 @cindex @code{section} function attribute
2304 Normally, the compiler places the code it generates in the @code{text} section.
2305 Sometimes, however, you need additional sections, or you need certain
2306 particular functions to appear in special sections. The @code{section}
2307 attribute specifies that a function lives in a particular section.
2308 For example, the declaration:
2311 extern void foobar (void) __attribute__ ((section ("bar")));
2315 puts the function @code{foobar} in the @code{bar} section.
2317 Some file formats do not support arbitrary sections so the @code{section}
2318 attribute is not available on all platforms.
2319 If you need to map the entire contents of a module to a particular
2320 section, consider using the facilities of the linker instead.
2323 @cindex @code{sentinel} function attribute
2324 This function attribute ensures that a parameter in a function call is
2325 an explicit @code{NULL}. The attribute is only valid on variadic
2326 functions. By default, the sentinel is located at position zero, the
2327 last parameter of the function call. If an optional integer position
2328 argument P is supplied to the attribute, the sentinel must be located at
2329 position P counting backwards from the end of the argument list.
2332 __attribute__ ((sentinel))
2334 __attribute__ ((sentinel(0)))
2337 The attribute is automatically set with a position of 0 for the built-in
2338 functions @code{execl} and @code{execlp}. The built-in function
2339 @code{execle} has the attribute set with a position of 1.
2341 A valid @code{NULL} in this context is defined as zero with any pointer
2342 type. If your system defines the @code{NULL} macro with an integer type
2343 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2344 with a copy that redefines NULL appropriately.
2346 The warnings for missing or incorrect sentinels are enabled with
2350 See long_call/short_call.
2353 See longcall/shortcall.
2356 @cindex signal handler functions on the AVR processors
2357 Use this attribute on the AVR to indicate that the specified
2358 function is a signal handler. The compiler will generate function
2359 entry and exit sequences suitable for use in a signal handler when this
2360 attribute is present. Interrupts will be disabled inside the function.
2363 Use this attribute on the SH to indicate an @code{interrupt_handler}
2364 function should switch to an alternate stack. It expects a string
2365 argument that names a global variable holding the address of the
2370 void f () __attribute__ ((interrupt_handler,
2371 sp_switch ("alt_stack")));
2375 @cindex functions that pop the argument stack on the 386
2376 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2377 assume that the called function will pop off the stack space used to
2378 pass arguments, unless it takes a variable number of arguments.
2381 @cindex tiny data section on the H8/300H and H8S
2382 Use this attribute on the H8/300H and H8S to indicate that the specified
2383 variable should be placed into the tiny data section.
2384 The compiler will generate more efficient code for loads and stores
2385 on data in the tiny data section. Note the tiny data area is limited to
2386 slightly under 32kbytes of data.
2389 Use this attribute on the SH for an @code{interrupt_handler} to return using
2390 @code{trapa} instead of @code{rte}. This attribute expects an integer
2391 argument specifying the trap number to be used.
2394 @cindex @code{unused} attribute.
2395 This attribute, attached to a function, means that the function is meant
2396 to be possibly unused. GCC will not produce a warning for this
2400 @cindex @code{used} attribute.
2401 This attribute, attached to a function, means that code must be emitted
2402 for the function even if it appears that the function is not referenced.
2403 This is useful, for example, when the function is referenced only in
2407 @cindex @code{version_id} attribute on IA64 HP-UX
2408 This attribute, attached to a global variable or function, renames a
2409 symbol to contain a version string, thus allowing for function level
2410 versioning. HP-UX system header files may use version level functioning
2411 for some system calls.
2414 extern int foo () __attribute__((version_id ("20040821")));
2417 Calls to @var{foo} will be mapped to calls to @var{foo@{20040821@}}.
2419 @item visibility ("@var{visibility_type}")
2420 @cindex @code{visibility} attribute
2421 This attribute affects the linkage of the declaration to which it is attached.
2422 There are four supported @var{visibility_type} values: default,
2423 hidden, protected or internal visibility.
2426 void __attribute__ ((visibility ("protected")))
2427 f () @{ /* @r{Do something.} */; @}
2428 int i __attribute__ ((visibility ("hidden")));
2431 The possible values of @var{visibility_type} correspond to the
2432 visibility settings in the ELF gABI.
2435 @c keep this list of visibilities in alphabetical order.
2438 Default visibility is the normal case for the object file format.
2439 This value is available for the visibility attribute to override other
2440 options that may change the assumed visibility of entities.
2442 On ELF, default visibility means that the declaration is visible to other
2443 modules and, in shared libraries, means that the declared entity may be
2446 On Darwin, default visibility means that the declaration is visible to
2449 Default visibility corresponds to ``external linkage'' in the language.
2452 Hidden visibility indicates that the entity declared will have a new
2453 form of linkage, which we'll call ``hidden linkage''. Two
2454 declarations of an object with hidden linkage refer to the same object
2455 if they are in the same shared object.
2458 Internal visibility is like hidden visibility, but with additional
2459 processor specific semantics. Unless otherwise specified by the
2460 psABI, GCC defines internal visibility to mean that a function is
2461 @emph{never} called from another module. Compare this with hidden
2462 functions which, while they cannot be referenced directly by other
2463 modules, can be referenced indirectly via function pointers. By
2464 indicating that a function cannot be called from outside the module,
2465 GCC may for instance omit the load of a PIC register since it is known
2466 that the calling function loaded the correct value.
2469 Protected visibility is like default visibility except that it
2470 indicates that references within the defining module will bind to the
2471 definition in that module. That is, the declared entity cannot be
2472 overridden by another module.
2476 All visibilities are supported on many, but not all, ELF targets
2477 (supported when the assembler supports the @samp{.visibility}
2478 pseudo-op). Default visibility is supported everywhere. Hidden
2479 visibility is supported on Darwin targets.
2481 The visibility attribute should be applied only to declarations which
2482 would otherwise have external linkage. The attribute should be applied
2483 consistently, so that the same entity should not be declared with
2484 different settings of the attribute.
2486 In C++, the visibility attribute applies to types as well as functions
2487 and objects, because in C++ types have linkage. A class must not have
2488 greater visibility than its non-static data member types and bases,
2489 and class members default to the visibility of their class. Also, a
2490 declaration without explicit visibility is limited to the visibility
2493 In C++, you can mark member functions and static member variables of a
2494 class with the visibility attribute. This is useful if if you know a
2495 particular method or static member variable should only be used from
2496 one shared object; then you can mark it hidden while the rest of the
2497 class has default visibility. Care must be taken to avoid breaking
2498 the One Definition Rule; for example, it is usually not useful to mark
2499 an inline method as hidden without marking the whole class as hidden.
2501 A C++ namespace declaration can also have the visibility attribute.
2502 This attribute applies only to the particular namespace body, not to
2503 other definitions of the same namespace; it is equivalent to using
2504 @samp{#pragma GCC visibility} before and after the namespace
2505 definition (@pxref{Visibility Pragmas}).
2507 In C++, if a template argument has limited visibility, this
2508 restriction is implicitly propagated to the template instantiation.
2509 Otherwise, template instantiations and specializations default to the
2510 visibility of their template.
2512 If both the template and enclosing class have explicit visibility, the
2513 visibility from the template is used.
2515 @item warn_unused_result
2516 @cindex @code{warn_unused_result} attribute
2517 The @code{warn_unused_result} attribute causes a warning to be emitted
2518 if a caller of the function with this attribute does not use its
2519 return value. This is useful for functions where not checking
2520 the result is either a security problem or always a bug, such as
2524 int fn () __attribute__ ((warn_unused_result));
2527 if (fn () < 0) return -1;
2533 results in warning on line 5.
2536 @cindex @code{weak} attribute
2537 The @code{weak} attribute causes the declaration to be emitted as a weak
2538 symbol rather than a global. This is primarily useful in defining
2539 library functions which can be overridden in user code, though it can
2540 also be used with non-function declarations. Weak symbols are supported
2541 for ELF targets, and also for a.out targets when using the GNU assembler
2545 @itemx weakref ("@var{target}")
2546 @cindex @code{weakref} attribute
2547 The @code{weakref} attribute marks a declaration as a weak reference.
2548 Without arguments, it should be accompanied by an @code{alias} attribute
2549 naming the target symbol. Optionally, the @var{target} may be given as
2550 an argument to @code{weakref} itself. In either case, @code{weakref}
2551 implicitly marks the declaration as @code{weak}. Without a
2552 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2553 @code{weakref} is equivalent to @code{weak}.
2556 static int x() __attribute__ ((weakref ("y")));
2557 /* is equivalent to... */
2558 static int x() __attribute__ ((weak, weakref, alias ("y")));
2560 static int x() __attribute__ ((weakref));
2561 static int x() __attribute__ ((alias ("y")));
2564 A weak reference is an alias that does not by itself require a
2565 definition to be given for the target symbol. If the target symbol is
2566 only referenced through weak references, then the becomes a @code{weak}
2567 undefined symbol. If it is directly referenced, however, then such
2568 strong references prevail, and a definition will be required for the
2569 symbol, not necessarily in the same translation unit.
2571 The effect is equivalent to moving all references to the alias to a
2572 separate translation unit, renaming the alias to the aliased symbol,
2573 declaring it as weak, compiling the two separate translation units and
2574 performing a reloadable link on them.
2576 At present, a declaration to which @code{weakref} is attached can
2577 only be @code{static}.
2579 @item externally_visible
2580 @cindex @code{externally_visible} attribute.
2581 This attribute, attached to a global variable or function nullify
2582 effect of @option{-fwhole-program} command line option, so the object
2583 remain visible outside the current compilation unit
2587 You can specify multiple attributes in a declaration by separating them
2588 by commas within the double parentheses or by immediately following an
2589 attribute declaration with another attribute declaration.
2591 @cindex @code{#pragma}, reason for not using
2592 @cindex pragma, reason for not using
2593 Some people object to the @code{__attribute__} feature, suggesting that
2594 ISO C's @code{#pragma} should be used instead. At the time
2595 @code{__attribute__} was designed, there were two reasons for not doing
2600 It is impossible to generate @code{#pragma} commands from a macro.
2603 There is no telling what the same @code{#pragma} might mean in another
2607 These two reasons applied to almost any application that might have been
2608 proposed for @code{#pragma}. It was basically a mistake to use
2609 @code{#pragma} for @emph{anything}.
2611 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2612 to be generated from macros. In addition, a @code{#pragma GCC}
2613 namespace is now in use for GCC-specific pragmas. However, it has been
2614 found convenient to use @code{__attribute__} to achieve a natural
2615 attachment of attributes to their corresponding declarations, whereas
2616 @code{#pragma GCC} is of use for constructs that do not naturally form
2617 part of the grammar. @xref{Other Directives,,Miscellaneous
2618 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2620 @node Attribute Syntax
2621 @section Attribute Syntax
2622 @cindex attribute syntax
2624 This section describes the syntax with which @code{__attribute__} may be
2625 used, and the constructs to which attribute specifiers bind, for the C
2626 language. Some details may vary for C++ and Objective-C@. Because of
2627 infelicities in the grammar for attributes, some forms described here
2628 may not be successfully parsed in all cases.
2630 There are some problems with the semantics of attributes in C++. For
2631 example, there are no manglings for attributes, although they may affect
2632 code generation, so problems may arise when attributed types are used in
2633 conjunction with templates or overloading. Similarly, @code{typeid}
2634 does not distinguish between types with different attributes. Support
2635 for attributes in C++ may be restricted in future to attributes on
2636 declarations only, but not on nested declarators.
2638 @xref{Function Attributes}, for details of the semantics of attributes
2639 applying to functions. @xref{Variable Attributes}, for details of the
2640 semantics of attributes applying to variables. @xref{Type Attributes},
2641 for details of the semantics of attributes applying to structure, union
2642 and enumerated types.
2644 An @dfn{attribute specifier} is of the form
2645 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2646 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2647 each attribute is one of the following:
2651 Empty. Empty attributes are ignored.
2654 A word (which may be an identifier such as @code{unused}, or a reserved
2655 word such as @code{const}).
2658 A word, followed by, in parentheses, parameters for the attribute.
2659 These parameters take one of the following forms:
2663 An identifier. For example, @code{mode} attributes use this form.
2666 An identifier followed by a comma and a non-empty comma-separated list
2667 of expressions. For example, @code{format} attributes use this form.
2670 A possibly empty comma-separated list of expressions. For example,
2671 @code{format_arg} attributes use this form with the list being a single
2672 integer constant expression, and @code{alias} attributes use this form
2673 with the list being a single string constant.
2677 An @dfn{attribute specifier list} is a sequence of one or more attribute
2678 specifiers, not separated by any other tokens.
2680 In GNU C, an attribute specifier list may appear after the colon following a
2681 label, other than a @code{case} or @code{default} label. The only
2682 attribute it makes sense to use after a label is @code{unused}. This
2683 feature is intended for code generated by programs which contains labels
2684 that may be unused but which is compiled with @option{-Wall}. It would
2685 not normally be appropriate to use in it human-written code, though it
2686 could be useful in cases where the code that jumps to the label is
2687 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2688 such placement of attribute lists, as it is permissible for a
2689 declaration, which could begin with an attribute list, to be labelled in
2690 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2691 does not arise there.
2693 An attribute specifier list may appear as part of a @code{struct},
2694 @code{union} or @code{enum} specifier. It may go either immediately
2695 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2696 the closing brace. The former syntax is preferred.
2697 Where attribute specifiers follow the closing brace, they are considered
2698 to relate to the structure, union or enumerated type defined, not to any
2699 enclosing declaration the type specifier appears in, and the type
2700 defined is not complete until after the attribute specifiers.
2701 @c Otherwise, there would be the following problems: a shift/reduce
2702 @c conflict between attributes binding the struct/union/enum and
2703 @c binding to the list of specifiers/qualifiers; and "aligned"
2704 @c attributes could use sizeof for the structure, but the size could be
2705 @c changed later by "packed" attributes.
2707 Otherwise, an attribute specifier appears as part of a declaration,
2708 counting declarations of unnamed parameters and type names, and relates
2709 to that declaration (which may be nested in another declaration, for
2710 example in the case of a parameter declaration), or to a particular declarator
2711 within a declaration. Where an
2712 attribute specifier is applied to a parameter declared as a function or
2713 an array, it should apply to the function or array rather than the
2714 pointer to which the parameter is implicitly converted, but this is not
2715 yet correctly implemented.
2717 Any list of specifiers and qualifiers at the start of a declaration may
2718 contain attribute specifiers, whether or not such a list may in that
2719 context contain storage class specifiers. (Some attributes, however,
2720 are essentially in the nature of storage class specifiers, and only make
2721 sense where storage class specifiers may be used; for example,
2722 @code{section}.) There is one necessary limitation to this syntax: the
2723 first old-style parameter declaration in a function definition cannot
2724 begin with an attribute specifier, because such an attribute applies to
2725 the function instead by syntax described below (which, however, is not
2726 yet implemented in this case). In some other cases, attribute
2727 specifiers are permitted by this grammar but not yet supported by the
2728 compiler. All attribute specifiers in this place relate to the
2729 declaration as a whole. In the obsolescent usage where a type of
2730 @code{int} is implied by the absence of type specifiers, such a list of
2731 specifiers and qualifiers may be an attribute specifier list with no
2732 other specifiers or qualifiers.
2734 At present, the first parameter in a function prototype must have some
2735 type specifier which is not an attribute specifier; this resolves an
2736 ambiguity in the interpretation of @code{void f(int
2737 (__attribute__((foo)) x))}, but is subject to change. At present, if
2738 the parentheses of a function declarator contain only attributes then
2739 those attributes are ignored, rather than yielding an error or warning
2740 or implying a single parameter of type int, but this is subject to
2743 An attribute specifier list may appear immediately before a declarator
2744 (other than the first) in a comma-separated list of declarators in a
2745 declaration of more than one identifier using a single list of
2746 specifiers and qualifiers. Such attribute specifiers apply
2747 only to the identifier before whose declarator they appear. For
2751 __attribute__((noreturn)) void d0 (void),
2752 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2757 the @code{noreturn} attribute applies to all the functions
2758 declared; the @code{format} attribute only applies to @code{d1}.
2760 An attribute specifier list may appear immediately before the comma,
2761 @code{=} or semicolon terminating the declaration of an identifier other
2762 than a function definition. At present, such attribute specifiers apply
2763 to the declared object or function, but in future they may attach to the
2764 outermost adjacent declarator. In simple cases there is no difference,
2765 but, for example, in
2768 void (****f)(void) __attribute__((noreturn));
2772 at present the @code{noreturn} attribute applies to @code{f}, which
2773 causes a warning since @code{f} is not a function, but in future it may
2774 apply to the function @code{****f}. The precise semantics of what
2775 attributes in such cases will apply to are not yet specified. Where an
2776 assembler name for an object or function is specified (@pxref{Asm
2777 Labels}), at present the attribute must follow the @code{asm}
2778 specification; in future, attributes before the @code{asm} specification
2779 may apply to the adjacent declarator, and those after it to the declared
2782 An attribute specifier list may, in future, be permitted to appear after
2783 the declarator in a function definition (before any old-style parameter
2784 declarations or the function body).
2786 Attribute specifiers may be mixed with type qualifiers appearing inside
2787 the @code{[]} of a parameter array declarator, in the C99 construct by
2788 which such qualifiers are applied to the pointer to which the array is
2789 implicitly converted. Such attribute specifiers apply to the pointer,
2790 not to the array, but at present this is not implemented and they are
2793 An attribute specifier list may appear at the start of a nested
2794 declarator. At present, there are some limitations in this usage: the
2795 attributes correctly apply to the declarator, but for most individual
2796 attributes the semantics this implies are not implemented.
2797 When attribute specifiers follow the @code{*} of a pointer
2798 declarator, they may be mixed with any type qualifiers present.
2799 The following describes the formal semantics of this syntax. It will make the
2800 most sense if you are familiar with the formal specification of
2801 declarators in the ISO C standard.
2803 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2804 D1}, where @code{T} contains declaration specifiers that specify a type
2805 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2806 contains an identifier @var{ident}. The type specified for @var{ident}
2807 for derived declarators whose type does not include an attribute
2808 specifier is as in the ISO C standard.
2810 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2811 and the declaration @code{T D} specifies the type
2812 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2813 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2814 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2816 If @code{D1} has the form @code{*
2817 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2818 declaration @code{T D} specifies the type
2819 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2820 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2821 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2827 void (__attribute__((noreturn)) ****f) (void);
2831 specifies the type ``pointer to pointer to pointer to pointer to
2832 non-returning function returning @code{void}''. As another example,
2835 char *__attribute__((aligned(8))) *f;
2839 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2840 Note again that this does not work with most attributes; for example,
2841 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2842 is not yet supported.
2844 For compatibility with existing code written for compiler versions that
2845 did not implement attributes on nested declarators, some laxity is
2846 allowed in the placing of attributes. If an attribute that only applies
2847 to types is applied to a declaration, it will be treated as applying to
2848 the type of that declaration. If an attribute that only applies to
2849 declarations is applied to the type of a declaration, it will be treated
2850 as applying to that declaration; and, for compatibility with code
2851 placing the attributes immediately before the identifier declared, such
2852 an attribute applied to a function return type will be treated as
2853 applying to the function type, and such an attribute applied to an array
2854 element type will be treated as applying to the array type. If an
2855 attribute that only applies to function types is applied to a
2856 pointer-to-function type, it will be treated as applying to the pointer
2857 target type; if such an attribute is applied to a function return type
2858 that is not a pointer-to-function type, it will be treated as applying
2859 to the function type.
2861 @node Function Prototypes
2862 @section Prototypes and Old-Style Function Definitions
2863 @cindex function prototype declarations
2864 @cindex old-style function definitions
2865 @cindex promotion of formal parameters
2867 GNU C extends ISO C to allow a function prototype to override a later
2868 old-style non-prototype definition. Consider the following example:
2871 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2878 /* @r{Prototype function declaration.} */
2879 int isroot P((uid_t));
2881 /* @r{Old-style function definition.} */
2883 isroot (x) /* @r{??? lossage here ???} */
2890 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2891 not allow this example, because subword arguments in old-style
2892 non-prototype definitions are promoted. Therefore in this example the
2893 function definition's argument is really an @code{int}, which does not
2894 match the prototype argument type of @code{short}.
2896 This restriction of ISO C makes it hard to write code that is portable
2897 to traditional C compilers, because the programmer does not know
2898 whether the @code{uid_t} type is @code{short}, @code{int}, or
2899 @code{long}. Therefore, in cases like these GNU C allows a prototype
2900 to override a later old-style definition. More precisely, in GNU C, a
2901 function prototype argument type overrides the argument type specified
2902 by a later old-style definition if the former type is the same as the
2903 latter type before promotion. Thus in GNU C the above example is
2904 equivalent to the following:
2917 GNU C++ does not support old-style function definitions, so this
2918 extension is irrelevant.
2921 @section C++ Style Comments
2923 @cindex C++ comments
2924 @cindex comments, C++ style
2926 In GNU C, you may use C++ style comments, which start with @samp{//} and
2927 continue until the end of the line. Many other C implementations allow
2928 such comments, and they are included in the 1999 C standard. However,
2929 C++ style comments are not recognized if you specify an @option{-std}
2930 option specifying a version of ISO C before C99, or @option{-ansi}
2931 (equivalent to @option{-std=c89}).
2934 @section Dollar Signs in Identifier Names
2936 @cindex dollar signs in identifier names
2937 @cindex identifier names, dollar signs in
2939 In GNU C, you may normally use dollar signs in identifier names.
2940 This is because many traditional C implementations allow such identifiers.
2941 However, dollar signs in identifiers are not supported on a few target
2942 machines, typically because the target assembler does not allow them.
2944 @node Character Escapes
2945 @section The Character @key{ESC} in Constants
2947 You can use the sequence @samp{\e} in a string or character constant to
2948 stand for the ASCII character @key{ESC}.
2951 @section Inquiring on Alignment of Types or Variables
2953 @cindex type alignment
2954 @cindex variable alignment
2956 The keyword @code{__alignof__} allows you to inquire about how an object
2957 is aligned, or the minimum alignment usually required by a type. Its
2958 syntax is just like @code{sizeof}.
2960 For example, if the target machine requires a @code{double} value to be
2961 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2962 This is true on many RISC machines. On more traditional machine
2963 designs, @code{__alignof__ (double)} is 4 or even 2.
2965 Some machines never actually require alignment; they allow reference to any
2966 data type even at an odd address. For these machines, @code{__alignof__}
2967 reports the @emph{recommended} alignment of a type.
2969 If the operand of @code{__alignof__} is an lvalue rather than a type,
2970 its value is the required alignment for its type, taking into account
2971 any minimum alignment specified with GCC's @code{__attribute__}
2972 extension (@pxref{Variable Attributes}). For example, after this
2976 struct foo @{ int x; char y; @} foo1;
2980 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2981 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2983 It is an error to ask for the alignment of an incomplete type.
2985 @node Variable Attributes
2986 @section Specifying Attributes of Variables
2987 @cindex attribute of variables
2988 @cindex variable attributes
2990 The keyword @code{__attribute__} allows you to specify special
2991 attributes of variables or structure fields. This keyword is followed
2992 by an attribute specification inside double parentheses. Some
2993 attributes are currently defined generically for variables.
2994 Other attributes are defined for variables on particular target
2995 systems. Other attributes are available for functions
2996 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2997 Other front ends might define more attributes
2998 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3000 You may also specify attributes with @samp{__} preceding and following
3001 each keyword. This allows you to use them in header files without
3002 being concerned about a possible macro of the same name. For example,
3003 you may use @code{__aligned__} instead of @code{aligned}.
3005 @xref{Attribute Syntax}, for details of the exact syntax for using
3009 @cindex @code{aligned} attribute
3010 @item aligned (@var{alignment})
3011 This attribute specifies a minimum alignment for the variable or
3012 structure field, measured in bytes. For example, the declaration:
3015 int x __attribute__ ((aligned (16))) = 0;
3019 causes the compiler to allocate the global variable @code{x} on a
3020 16-byte boundary. On a 68040, this could be used in conjunction with
3021 an @code{asm} expression to access the @code{move16} instruction which
3022 requires 16-byte aligned operands.
3024 You can also specify the alignment of structure fields. For example, to
3025 create a double-word aligned @code{int} pair, you could write:
3028 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3032 This is an alternative to creating a union with a @code{double} member
3033 that forces the union to be double-word aligned.
3035 As in the preceding examples, you can explicitly specify the alignment
3036 (in bytes) that you wish the compiler to use for a given variable or
3037 structure field. Alternatively, you can leave out the alignment factor
3038 and just ask the compiler to align a variable or field to the maximum
3039 useful alignment for the target machine you are compiling for. For
3040 example, you could write:
3043 short array[3] __attribute__ ((aligned));
3046 Whenever you leave out the alignment factor in an @code{aligned} attribute
3047 specification, the compiler automatically sets the alignment for the declared
3048 variable or field to the largest alignment which is ever used for any data
3049 type on the target machine you are compiling for. Doing this can often make
3050 copy operations more efficient, because the compiler can use whatever
3051 instructions copy the biggest chunks of memory when performing copies to
3052 or from the variables or fields that you have aligned this way.
3054 The @code{aligned} attribute can only increase the alignment; but you
3055 can decrease it by specifying @code{packed} as well. See below.
3057 Note that the effectiveness of @code{aligned} attributes may be limited
3058 by inherent limitations in your linker. On many systems, the linker is
3059 only able to arrange for variables to be aligned up to a certain maximum
3060 alignment. (For some linkers, the maximum supported alignment may
3061 be very very small.) If your linker is only able to align variables
3062 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3063 in an @code{__attribute__} will still only provide you with 8 byte
3064 alignment. See your linker documentation for further information.
3066 @item cleanup (@var{cleanup_function})
3067 @cindex @code{cleanup} attribute
3068 The @code{cleanup} attribute runs a function when the variable goes
3069 out of scope. This attribute can only be applied to auto function
3070 scope variables; it may not be applied to parameters or variables
3071 with static storage duration. The function must take one parameter,
3072 a pointer to a type compatible with the variable. The return value
3073 of the function (if any) is ignored.
3075 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3076 will be run during the stack unwinding that happens during the
3077 processing of the exception. Note that the @code{cleanup} attribute
3078 does not allow the exception to be caught, only to perform an action.
3079 It is undefined what happens if @var{cleanup_function} does not
3084 @cindex @code{common} attribute
3085 @cindex @code{nocommon} attribute
3088 The @code{common} attribute requests GCC to place a variable in
3089 ``common'' storage. The @code{nocommon} attribute requests the
3090 opposite---to allocate space for it directly.
3092 These attributes override the default chosen by the
3093 @option{-fno-common} and @option{-fcommon} flags respectively.
3096 @cindex @code{deprecated} attribute
3097 The @code{deprecated} attribute results in a warning if the variable
3098 is used anywhere in the source file. This is useful when identifying
3099 variables that are expected to be removed in a future version of a
3100 program. The warning also includes the location of the declaration
3101 of the deprecated variable, to enable users to easily find further
3102 information about why the variable is deprecated, or what they should
3103 do instead. Note that the warning only occurs for uses:
3106 extern int old_var __attribute__ ((deprecated));
3108 int new_fn () @{ return old_var; @}
3111 results in a warning on line 3 but not line 2.
3113 The @code{deprecated} attribute can also be used for functions and
3114 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3116 @item mode (@var{mode})
3117 @cindex @code{mode} attribute
3118 This attribute specifies the data type for the declaration---whichever
3119 type corresponds to the mode @var{mode}. This in effect lets you
3120 request an integer or floating point type according to its width.
3122 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3123 indicate the mode corresponding to a one-byte integer, @samp{word} or
3124 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3125 or @samp{__pointer__} for the mode used to represent pointers.
3128 @cindex @code{packed} attribute
3129 The @code{packed} attribute specifies that a variable or structure field
3130 should have the smallest possible alignment---one byte for a variable,
3131 and one bit for a field, unless you specify a larger value with the
3132 @code{aligned} attribute.
3134 Here is a structure in which the field @code{x} is packed, so that it
3135 immediately follows @code{a}:
3141 int x[2] __attribute__ ((packed));
3145 @item section ("@var{section-name}")
3146 @cindex @code{section} variable attribute
3147 Normally, the compiler places the objects it generates in sections like
3148 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3149 or you need certain particular variables to appear in special sections,
3150 for example to map to special hardware. The @code{section}
3151 attribute specifies that a variable (or function) lives in a particular
3152 section. For example, this small program uses several specific section names:
3155 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3156 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3157 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3158 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3162 /* @r{Initialize stack pointer} */
3163 init_sp (stack + sizeof (stack));
3165 /* @r{Initialize initialized data} */
3166 memcpy (&init_data, &data, &edata - &data);
3168 /* @r{Turn on the serial ports} */
3175 Use the @code{section} attribute with an @emph{initialized} definition
3176 of a @emph{global} variable, as shown in the example. GCC issues
3177 a warning and otherwise ignores the @code{section} attribute in
3178 uninitialized variable declarations.
3180 You may only use the @code{section} attribute with a fully initialized
3181 global definition because of the way linkers work. The linker requires
3182 each object be defined once, with the exception that uninitialized
3183 variables tentatively go in the @code{common} (or @code{bss}) section
3184 and can be multiply ``defined''. You can force a variable to be
3185 initialized with the @option{-fno-common} flag or the @code{nocommon}
3188 Some file formats do not support arbitrary sections so the @code{section}
3189 attribute is not available on all platforms.
3190 If you need to map the entire contents of a module to a particular
3191 section, consider using the facilities of the linker instead.
3194 @cindex @code{shared} variable attribute
3195 On Microsoft Windows, in addition to putting variable definitions in a named
3196 section, the section can also be shared among all running copies of an
3197 executable or DLL@. For example, this small program defines shared data
3198 by putting it in a named section @code{shared} and marking the section
3202 int foo __attribute__((section ("shared"), shared)) = 0;
3207 /* @r{Read and write foo. All running
3208 copies see the same value.} */
3214 You may only use the @code{shared} attribute along with @code{section}
3215 attribute with a fully initialized global definition because of the way
3216 linkers work. See @code{section} attribute for more information.
3218 The @code{shared} attribute is only available on Microsoft Windows@.
3220 @item tls_model ("@var{tls_model}")
3221 @cindex @code{tls_model} attribute
3222 The @code{tls_model} attribute sets thread-local storage model
3223 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3224 overriding @option{-ftls-model=} command line switch on a per-variable
3226 The @var{tls_model} argument should be one of @code{global-dynamic},
3227 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3229 Not all targets support this attribute.
3232 This attribute, attached to a variable, means that the variable is meant
3233 to be possibly unused. GCC will not produce a warning for this
3237 This attribute, attached to a variable, means that the variable must be
3238 emitted even if it appears that the variable is not referenced.
3240 @item vector_size (@var{bytes})
3241 This attribute specifies the vector size for the variable, measured in
3242 bytes. For example, the declaration:
3245 int foo __attribute__ ((vector_size (16)));
3249 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3250 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3251 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3253 This attribute is only applicable to integral and float scalars,
3254 although arrays, pointers, and function return values are allowed in
3255 conjunction with this construct.
3257 Aggregates with this attribute are invalid, even if they are of the same
3258 size as a corresponding scalar. For example, the declaration:
3261 struct S @{ int a; @};
3262 struct S __attribute__ ((vector_size (16))) foo;
3266 is invalid even if the size of the structure is the same as the size of
3270 The @code{selectany} attribute causes an initialized global variable to
3271 have link-once semantics. When multiple definitions of the variable are
3272 encountered by the linker, the first is selected and the remainder are
3273 discarded. Following usage by the Microsoft compiler, the linker is told
3274 @emph{not} to warn about size or content differences of the multiple
3277 Although the primary usage of this attribute is for POD types, the
3278 attribute can also be applied to global C++ objects that are initialized
3279 by a constructor. In this case, the static initialization and destruction
3280 code for the object is emitted in each translation defining the object,
3281 but the calls to the constructor and destructor are protected by a
3282 link-once guard variable.
3284 The @code{selectany} attribute is only available on Microsoft Windows
3285 targets. You can use @code{__declspec (selectany)} as a synonym for
3286 @code{__attribute__ ((selectany))} for compatibility with other
3290 The @code{weak} attribute is described in @xref{Function Attributes}.
3293 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3296 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3300 @subsection M32R/D Variable Attributes
3302 One attribute is currently defined for the M32R/D@.
3305 @item model (@var{model-name})
3306 @cindex variable addressability on the M32R/D
3307 Use this attribute on the M32R/D to set the addressability of an object.
3308 The identifier @var{model-name} is one of @code{small}, @code{medium},
3309 or @code{large}, representing each of the code models.
3311 Small model objects live in the lower 16MB of memory (so that their
3312 addresses can be loaded with the @code{ld24} instruction).
3314 Medium and large model objects may live anywhere in the 32-bit address space
3315 (the compiler will generate @code{seth/add3} instructions to load their
3319 @anchor{i386 Variable Attributes}
3320 @subsection i386 Variable Attributes
3322 Two attributes are currently defined for i386 configurations:
3323 @code{ms_struct} and @code{gcc_struct}
3328 @cindex @code{ms_struct} attribute
3329 @cindex @code{gcc_struct} attribute
3331 If @code{packed} is used on a structure, or if bit-fields are used
3332 it may be that the Microsoft ABI packs them differently
3333 than GCC would normally pack them. Particularly when moving packed
3334 data between functions compiled with GCC and the native Microsoft compiler
3335 (either via function call or as data in a file), it may be necessary to access
3338 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3339 compilers to match the native Microsoft compiler.
3341 The Microsoft structure layout algorithm is fairly simple with the exception
3342 of the bitfield packing:
3344 The padding and alignment of members of structures and whether a bit field
3345 can straddle a storage-unit boundary
3348 @item Structure members are stored sequentially in the order in which they are
3349 declared: the first member has the lowest memory address and the last member
3352 @item Every data object has an alignment-requirement. The alignment-requirement
3353 for all data except structures, unions, and arrays is either the size of the
3354 object or the current packing size (specified with either the aligned attribute
3355 or the pack pragma), whichever is less. For structures, unions, and arrays,
3356 the alignment-requirement is the largest alignment-requirement of its members.
3357 Every object is allocated an offset so that:
3359 offset % alignment-requirement == 0
3361 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3362 unit if the integral types are the same size and if the next bit field fits
3363 into the current allocation unit without crossing the boundary imposed by the
3364 common alignment requirements of the bit fields.
3367 Handling of zero-length bitfields:
3369 MSVC interprets zero-length bitfields in the following ways:
3372 @item If a zero-length bitfield is inserted between two bitfields that would
3373 normally be coalesced, the bitfields will not be coalesced.
3380 unsigned long bf_1 : 12;
3382 unsigned long bf_2 : 12;
3386 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3387 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3389 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3390 alignment of the zero-length bitfield is greater than the member that follows it,
3391 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3411 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3412 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3413 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3416 Taking this into account, it is important to note the following:
3419 @item If a zero-length bitfield follows a normal bitfield, the type of the
3420 zero-length bitfield may affect the alignment of the structure as whole. For
3421 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3422 normal bitfield, and is of type short.
3424 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3425 still affect the alignment of the structure:
3435 Here, @code{t4} will take up 4 bytes.
3438 @item Zero-length bitfields following non-bitfield members are ignored:
3449 Here, @code{t5} will take up 2 bytes.
3453 @subsection PowerPC Variable Attributes
3455 Three attributes currently are defined for PowerPC configurations:
3456 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3458 For full documentation of the struct attributes please see the
3459 documentation in the @xref{i386 Variable Attributes}, section.
3461 For documentation of @code{altivec} attribute please see the
3462 documentation in the @xref{PowerPC Type Attributes}, section.
3464 @subsection SPU Variable Attributes
3466 The SPU supports the @code{spu_vector} attribute for variables. For
3467 documentation of this attribute please see the documentation in the
3468 @xref{SPU Type Attributes}, section.
3470 @subsection Xstormy16 Variable Attributes
3472 One attribute is currently defined for xstormy16 configurations:
3477 @cindex @code{below100} attribute
3479 If a variable has the @code{below100} attribute (@code{BELOW100} is
3480 allowed also), GCC will place the variable in the first 0x100 bytes of
3481 memory and use special opcodes to access it. Such variables will be
3482 placed in either the @code{.bss_below100} section or the
3483 @code{.data_below100} section.
3487 @node Type Attributes
3488 @section Specifying Attributes of Types
3489 @cindex attribute of types
3490 @cindex type attributes
3492 The keyword @code{__attribute__} allows you to specify special
3493 attributes of @code{struct} and @code{union} types when you define
3494 such types. This keyword is followed by an attribute specification
3495 inside double parentheses. Seven attributes are currently defined for
3496 types: @code{aligned}, @code{packed}, @code{transparent_union},
3497 @code{unused}, @code{deprecated}, @code{visibility}, and
3498 @code{may_alias}. Other attributes are defined for functions
3499 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3502 You may also specify any one of these attributes with @samp{__}
3503 preceding and following its keyword. This allows you to use these
3504 attributes in header files without being concerned about a possible
3505 macro of the same name. For example, you may use @code{__aligned__}
3506 instead of @code{aligned}.
3508 You may specify type attributes either in a @code{typedef} declaration
3509 or in an enum, struct or union type declaration or definition.
3511 For an enum, struct or union type, you may specify attributes either
3512 between the enum, struct or union tag and the name of the type, or
3513 just past the closing curly brace of the @emph{definition}. The
3514 former syntax is preferred.
3516 @xref{Attribute Syntax}, for details of the exact syntax for using
3520 @cindex @code{aligned} attribute
3521 @item aligned (@var{alignment})
3522 This attribute specifies a minimum alignment (in bytes) for variables
3523 of the specified type. For example, the declarations:
3526 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3527 typedef int more_aligned_int __attribute__ ((aligned (8)));
3531 force the compiler to insure (as far as it can) that each variable whose
3532 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3533 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3534 variables of type @code{struct S} aligned to 8-byte boundaries allows
3535 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3536 store) instructions when copying one variable of type @code{struct S} to
3537 another, thus improving run-time efficiency.
3539 Note that the alignment of any given @code{struct} or @code{union} type
3540 is required by the ISO C standard to be at least a perfect multiple of
3541 the lowest common multiple of the alignments of all of the members of
3542 the @code{struct} or @code{union} in question. This means that you @emph{can}
3543 effectively adjust the alignment of a @code{struct} or @code{union}
3544 type by attaching an @code{aligned} attribute to any one of the members
3545 of such a type, but the notation illustrated in the example above is a
3546 more obvious, intuitive, and readable way to request the compiler to
3547 adjust the alignment of an entire @code{struct} or @code{union} type.
3549 As in the preceding example, you can explicitly specify the alignment
3550 (in bytes) that you wish the compiler to use for a given @code{struct}
3551 or @code{union} type. Alternatively, you can leave out the alignment factor
3552 and just ask the compiler to align a type to the maximum
3553 useful alignment for the target machine you are compiling for. For
3554 example, you could write:
3557 struct S @{ short f[3]; @} __attribute__ ((aligned));
3560 Whenever you leave out the alignment factor in an @code{aligned}
3561 attribute specification, the compiler automatically sets the alignment
3562 for the type to the largest alignment which is ever used for any data
3563 type on the target machine you are compiling for. Doing this can often
3564 make copy operations more efficient, because the compiler can use
3565 whatever instructions copy the biggest chunks of memory when performing
3566 copies to or from the variables which have types that you have aligned
3569 In the example above, if the size of each @code{short} is 2 bytes, then
3570 the size of the entire @code{struct S} type is 6 bytes. The smallest
3571 power of two which is greater than or equal to that is 8, so the
3572 compiler sets the alignment for the entire @code{struct S} type to 8
3575 Note that although you can ask the compiler to select a time-efficient
3576 alignment for a given type and then declare only individual stand-alone
3577 objects of that type, the compiler's ability to select a time-efficient
3578 alignment is primarily useful only when you plan to create arrays of
3579 variables having the relevant (efficiently aligned) type. If you
3580 declare or use arrays of variables of an efficiently-aligned type, then
3581 it is likely that your program will also be doing pointer arithmetic (or
3582 subscripting, which amounts to the same thing) on pointers to the
3583 relevant type, and the code that the compiler generates for these
3584 pointer arithmetic operations will often be more efficient for
3585 efficiently-aligned types than for other types.
3587 The @code{aligned} attribute can only increase the alignment; but you
3588 can decrease it by specifying @code{packed} as well. See below.
3590 Note that the effectiveness of @code{aligned} attributes may be limited
3591 by inherent limitations in your linker. On many systems, the linker is
3592 only able to arrange for variables to be aligned up to a certain maximum
3593 alignment. (For some linkers, the maximum supported alignment may
3594 be very very small.) If your linker is only able to align variables
3595 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3596 in an @code{__attribute__} will still only provide you with 8 byte
3597 alignment. See your linker documentation for further information.
3600 This attribute, attached to @code{struct} or @code{union} type
3601 definition, specifies that each member (other than zero-width bitfields)
3602 of the structure or union is placed to minimize the memory required. When
3603 attached to an @code{enum} definition, it indicates that the smallest
3604 integral type should be used.
3606 @opindex fshort-enums
3607 Specifying this attribute for @code{struct} and @code{union} types is
3608 equivalent to specifying the @code{packed} attribute on each of the
3609 structure or union members. Specifying the @option{-fshort-enums}
3610 flag on the line is equivalent to specifying the @code{packed}
3611 attribute on all @code{enum} definitions.
3613 In the following example @code{struct my_packed_struct}'s members are
3614 packed closely together, but the internal layout of its @code{s} member
3615 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3619 struct my_unpacked_struct
3625 struct __attribute__ ((__packed__)) my_packed_struct
3629 struct my_unpacked_struct s;
3633 You may only specify this attribute on the definition of a @code{enum},
3634 @code{struct} or @code{union}, not on a @code{typedef} which does not
3635 also define the enumerated type, structure or union.
3637 @item transparent_union
3638 This attribute, attached to a @code{union} type definition, indicates
3639 that any function parameter having that union type causes calls to that
3640 function to be treated in a special way.
3642 First, the argument corresponding to a transparent union type can be of
3643 any type in the union; no cast is required. Also, if the union contains
3644 a pointer type, the corresponding argument can be a null pointer
3645 constant or a void pointer expression; and if the union contains a void
3646 pointer type, the corresponding argument can be any pointer expression.
3647 If the union member type is a pointer, qualifiers like @code{const} on
3648 the referenced type must be respected, just as with normal pointer
3651 Second, the argument is passed to the function using the calling
3652 conventions of the first member of the transparent union, not the calling
3653 conventions of the union itself. All members of the union must have the
3654 same machine representation; this is necessary for this argument passing
3657 Transparent unions are designed for library functions that have multiple
3658 interfaces for compatibility reasons. For example, suppose the
3659 @code{wait} function must accept either a value of type @code{int *} to
3660 comply with Posix, or a value of type @code{union wait *} to comply with
3661 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3662 @code{wait} would accept both kinds of arguments, but it would also
3663 accept any other pointer type and this would make argument type checking
3664 less useful. Instead, @code{<sys/wait.h>} might define the interface
3672 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3674 pid_t wait (wait_status_ptr_t);
3677 This interface allows either @code{int *} or @code{union wait *}
3678 arguments to be passed, using the @code{int *} calling convention.
3679 The program can call @code{wait} with arguments of either type:
3682 int w1 () @{ int w; return wait (&w); @}
3683 int w2 () @{ union wait w; return wait (&w); @}
3686 With this interface, @code{wait}'s implementation might look like this:
3689 pid_t wait (wait_status_ptr_t p)
3691 return waitpid (-1, p.__ip, 0);
3696 When attached to a type (including a @code{union} or a @code{struct}),
3697 this attribute means that variables of that type are meant to appear
3698 possibly unused. GCC will not produce a warning for any variables of
3699 that type, even if the variable appears to do nothing. This is often
3700 the case with lock or thread classes, which are usually defined and then
3701 not referenced, but contain constructors and destructors that have
3702 nontrivial bookkeeping functions.
3705 The @code{deprecated} attribute results in a warning if the type
3706 is used anywhere in the source file. This is useful when identifying
3707 types that are expected to be removed in a future version of a program.
3708 If possible, the warning also includes the location of the declaration
3709 of the deprecated type, to enable users to easily find further
3710 information about why the type is deprecated, or what they should do
3711 instead. Note that the warnings only occur for uses and then only
3712 if the type is being applied to an identifier that itself is not being
3713 declared as deprecated.
3716 typedef int T1 __attribute__ ((deprecated));
3720 typedef T1 T3 __attribute__ ((deprecated));
3721 T3 z __attribute__ ((deprecated));
3724 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3725 warning is issued for line 4 because T2 is not explicitly
3726 deprecated. Line 5 has no warning because T3 is explicitly
3727 deprecated. Similarly for line 6.
3729 The @code{deprecated} attribute can also be used for functions and
3730 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3733 Accesses to objects with types with this attribute are not subjected to
3734 type-based alias analysis, but are instead assumed to be able to alias
3735 any other type of objects, just like the @code{char} type. See
3736 @option{-fstrict-aliasing} for more information on aliasing issues.
3741 typedef short __attribute__((__may_alias__)) short_a;
3747 short_a *b = (short_a *) &a;
3751 if (a == 0x12345678)
3758 If you replaced @code{short_a} with @code{short} in the variable
3759 declaration, the above program would abort when compiled with
3760 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3761 above in recent GCC versions.
3764 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3765 applied to class, struct, union and enum types. Unlike other type
3766 attributes, the attribute must appear between the initial keyword and
3767 the name of the type; it cannot appear after the body of the type.
3769 Note that the type visibility is applied to vague linkage entities
3770 associated with the class (vtable, typeinfo node, etc.). In
3771 particular, if a class is thrown as an exception in one shared object
3772 and caught in another, the class must have default visibility.
3773 Otherwise the two shared objects will be unable to use the same
3774 typeinfo node and exception handling will break.
3776 @subsection ARM Type Attributes
3778 On those ARM targets that support @code{dllimport} (such as Symbian
3779 OS), you can use the @code{notshared} attribute to indicate that the
3780 virtual table and other similar data for a class should not be
3781 exported from a DLL@. For example:
3784 class __declspec(notshared) C @{
3786 __declspec(dllimport) C();
3790 __declspec(dllexport)
3794 In this code, @code{C::C} is exported from the current DLL, but the
3795 virtual table for @code{C} is not exported. (You can use
3796 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3797 most Symbian OS code uses @code{__declspec}.)
3799 @anchor{i386 Type Attributes}
3800 @subsection i386 Type Attributes
3802 Two attributes are currently defined for i386 configurations:
3803 @code{ms_struct} and @code{gcc_struct}
3807 @cindex @code{ms_struct}
3808 @cindex @code{gcc_struct}
3810 If @code{packed} is used on a structure, or if bit-fields are used
3811 it may be that the Microsoft ABI packs them differently
3812 than GCC would normally pack them. Particularly when moving packed
3813 data between functions compiled with GCC and the native Microsoft compiler
3814 (either via function call or as data in a file), it may be necessary to access
3817 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3818 compilers to match the native Microsoft compiler.
3821 To specify multiple attributes, separate them by commas within the
3822 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3825 @anchor{PowerPC Type Attributes}
3826 @subsection PowerPC Type Attributes
3828 Three attributes currently are defined for PowerPC configurations:
3829 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3831 For full documentation of the struct attributes please see the
3832 documentation in the @xref{i386 Type Attributes}, section.
3834 The @code{altivec} attribute allows one to declare AltiVec vector data
3835 types supported by the AltiVec Programming Interface Manual. The
3836 attribute requires an argument to specify one of three vector types:
3837 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3838 and @code{bool__} (always followed by unsigned).
3841 __attribute__((altivec(vector__)))
3842 __attribute__((altivec(pixel__))) unsigned short
3843 __attribute__((altivec(bool__))) unsigned
3846 These attributes mainly are intended to support the @code{__vector},
3847 @code{__pixel}, and @code{__bool} AltiVec keywords.
3849 @anchor{SPU Type Attributes}
3850 @subsection SPU Type Attributes
3852 The SPU supports the @code{spu_vector} attribute for types. This attribute
3853 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
3854 Language Extensions Specification. It is intended to support the
3855 @code{__vector} keyword.
3859 @section An Inline Function is As Fast As a Macro
3860 @cindex inline functions
3861 @cindex integrating function code
3863 @cindex macros, inline alternative
3865 By declaring a function inline, you can direct GCC to make
3866 calls to that function faster. One way GCC can achieve this is to
3867 integrate that function's code into the code for its callers. This
3868 makes execution faster by eliminating the function-call overhead; in
3869 addition, if any of the actual argument values are constant, their
3870 known values may permit simplifications at compile time so that not
3871 all of the inline function's code needs to be included. The effect on
3872 code size is less predictable; object code may be larger or smaller
3873 with function inlining, depending on the particular case. You can
3874 also direct GCC to try to integrate all ``simple enough'' functions
3875 into their callers with the option @option{-finline-functions}.
3877 GCC implements three different semantics of declaring a function
3878 inline. One is available with @option{-std=gnu89} or
3879 @option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
3880 on all inline declarations, another when @option{-std=c99} or
3881 @option{-std=gnu99} (without @option{-fgnu89-inline}), and the third
3882 is used when compiling C++.
3884 To declare a function inline, use the @code{inline} keyword in its
3885 declaration, like this:
3895 If you are writing a header file to be included in ISO C89 programs, write
3896 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3898 The three types of inlining behave similarly in two important cases:
3899 when the @code{inline} keyword is used on a @code{static} function,
3900 like the example above, and when a function is first declared without
3901 using the @code{inline} keyword and then is defined with
3902 @code{inline}, like this:
3905 extern int inc (int *a);
3913 In both of these common cases, the program behaves the same as if you
3914 had not used the @code{inline} keyword, except for its speed.
3916 @cindex inline functions, omission of
3917 @opindex fkeep-inline-functions
3918 When a function is both inline and @code{static}, if all calls to the
3919 function are integrated into the caller, and the function's address is
3920 never used, then the function's own assembler code is never referenced.
3921 In this case, GCC does not actually output assembler code for the
3922 function, unless you specify the option @option{-fkeep-inline-functions}.
3923 Some calls cannot be integrated for various reasons (in particular,
3924 calls that precede the function's definition cannot be integrated, and
3925 neither can recursive calls within the definition). If there is a
3926 nonintegrated call, then the function is compiled to assembler code as
3927 usual. The function must also be compiled as usual if the program
3928 refers to its address, because that can't be inlined.
3931 Note that certain usages in a function definition can make it unsuitable
3932 for inline substitution. Among these usages are: use of varargs, use of
3933 alloca, use of variable sized data types (@pxref{Variable Length}),
3934 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3935 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3936 will warn when a function marked @code{inline} could not be substituted,
3937 and will give the reason for the failure.
3939 @cindex automatic @code{inline} for C++ member fns
3940 @cindex @code{inline} automatic for C++ member fns
3941 @cindex member fns, automatically @code{inline}
3942 @cindex C++ member fns, automatically @code{inline}
3943 @opindex fno-default-inline
3944 As required by ISO C++, GCC considers member functions defined within
3945 the body of a class to be marked inline even if they are
3946 not explicitly declared with the @code{inline} keyword. You can
3947 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
3948 Options,,Options Controlling C++ Dialect}.
3950 GCC does not inline any functions when not optimizing unless you specify
3951 the @samp{always_inline} attribute for the function, like this:
3954 /* @r{Prototype.} */
3955 inline void foo (const char) __attribute__((always_inline));
3958 The remainder of this section is specific to GNU C89 inlining.
3960 @cindex non-static inline function
3961 When an inline function is not @code{static}, then the compiler must assume
3962 that there may be calls from other source files; since a global symbol can
3963 be defined only once in any program, the function must not be defined in
3964 the other source files, so the calls therein cannot be integrated.
3965 Therefore, a non-@code{static} inline function is always compiled on its
3966 own in the usual fashion.
3968 If you specify both @code{inline} and @code{extern} in the function
3969 definition, then the definition is used only for inlining. In no case
3970 is the function compiled on its own, not even if you refer to its
3971 address explicitly. Such an address becomes an external reference, as
3972 if you had only declared the function, and had not defined it.
3974 This combination of @code{inline} and @code{extern} has almost the
3975 effect of a macro. The way to use it is to put a function definition in
3976 a header file with these keywords, and put another copy of the
3977 definition (lacking @code{inline} and @code{extern}) in a library file.
3978 The definition in the header file will cause most calls to the function
3979 to be inlined. If any uses of the function remain, they will refer to
3980 the single copy in the library.
3983 @section Assembler Instructions with C Expression Operands
3984 @cindex extended @code{asm}
3985 @cindex @code{asm} expressions
3986 @cindex assembler instructions
3989 In an assembler instruction using @code{asm}, you can specify the
3990 operands of the instruction using C expressions. This means you need not
3991 guess which registers or memory locations will contain the data you want
3994 You must specify an assembler instruction template much like what
3995 appears in a machine description, plus an operand constraint string for
3998 For example, here is how to use the 68881's @code{fsinx} instruction:
4001 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4005 Here @code{angle} is the C expression for the input operand while
4006 @code{result} is that of the output operand. Each has @samp{"f"} as its
4007 operand constraint, saying that a floating point register is required.
4008 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4009 output operands' constraints must use @samp{=}. The constraints use the
4010 same language used in the machine description (@pxref{Constraints}).
4012 Each operand is described by an operand-constraint string followed by
4013 the C expression in parentheses. A colon separates the assembler
4014 template from the first output operand and another separates the last
4015 output operand from the first input, if any. Commas separate the
4016 operands within each group. The total number of operands is currently
4017 limited to 30; this limitation may be lifted in some future version of
4020 If there are no output operands but there are input operands, you must
4021 place two consecutive colons surrounding the place where the output
4024 As of GCC version 3.1, it is also possible to specify input and output
4025 operands using symbolic names which can be referenced within the
4026 assembler code. These names are specified inside square brackets
4027 preceding the constraint string, and can be referenced inside the
4028 assembler code using @code{%[@var{name}]} instead of a percentage sign
4029 followed by the operand number. Using named operands the above example
4033 asm ("fsinx %[angle],%[output]"
4034 : [output] "=f" (result)
4035 : [angle] "f" (angle));
4039 Note that the symbolic operand names have no relation whatsoever to
4040 other C identifiers. You may use any name you like, even those of
4041 existing C symbols, but you must ensure that no two operands within the same
4042 assembler construct use the same symbolic name.
4044 Output operand expressions must be lvalues; the compiler can check this.
4045 The input operands need not be lvalues. The compiler cannot check
4046 whether the operands have data types that are reasonable for the
4047 instruction being executed. It does not parse the assembler instruction
4048 template and does not know what it means or even whether it is valid
4049 assembler input. The extended @code{asm} feature is most often used for
4050 machine instructions the compiler itself does not know exist. If
4051 the output expression cannot be directly addressed (for example, it is a
4052 bit-field), your constraint must allow a register. In that case, GCC
4053 will use the register as the output of the @code{asm}, and then store
4054 that register into the output.
4056 The ordinary output operands must be write-only; GCC will assume that
4057 the values in these operands before the instruction are dead and need
4058 not be generated. Extended asm supports input-output or read-write
4059 operands. Use the constraint character @samp{+} to indicate such an
4060 operand and list it with the output operands. You should only use
4061 read-write operands when the constraints for the operand (or the
4062 operand in which only some of the bits are to be changed) allow a
4065 You may, as an alternative, logically split its function into two
4066 separate operands, one input operand and one write-only output
4067 operand. The connection between them is expressed by constraints
4068 which say they need to be in the same location when the instruction
4069 executes. You can use the same C expression for both operands, or
4070 different expressions. For example, here we write the (fictitious)
4071 @samp{combine} instruction with @code{bar} as its read-only source
4072 operand and @code{foo} as its read-write destination:
4075 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4079 The constraint @samp{"0"} for operand 1 says that it must occupy the
4080 same location as operand 0. A number in constraint is allowed only in
4081 an input operand and it must refer to an output operand.
4083 Only a number in the constraint can guarantee that one operand will be in
4084 the same place as another. The mere fact that @code{foo} is the value
4085 of both operands is not enough to guarantee that they will be in the
4086 same place in the generated assembler code. The following would not
4090 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4093 Various optimizations or reloading could cause operands 0 and 1 to be in
4094 different registers; GCC knows no reason not to do so. For example, the
4095 compiler might find a copy of the value of @code{foo} in one register and
4096 use it for operand 1, but generate the output operand 0 in a different
4097 register (copying it afterward to @code{foo}'s own address). Of course,
4098 since the register for operand 1 is not even mentioned in the assembler
4099 code, the result will not work, but GCC can't tell that.
4101 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4102 the operand number for a matching constraint. For example:
4105 asm ("cmoveq %1,%2,%[result]"
4106 : [result] "=r"(result)
4107 : "r" (test), "r"(new), "[result]"(old));
4110 Sometimes you need to make an @code{asm} operand be a specific register,
4111 but there's no matching constraint letter for that register @emph{by
4112 itself}. To force the operand into that register, use a local variable
4113 for the operand and specify the register in the variable declaration.
4114 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4115 register constraint letter that matches the register:
4118 register int *p1 asm ("r0") = @dots{};
4119 register int *p2 asm ("r1") = @dots{};
4120 register int *result asm ("r0");
4121 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4124 @anchor{Example of asm with clobbered asm reg}
4125 In the above example, beware that a register that is call-clobbered by
4126 the target ABI will be overwritten by any function call in the
4127 assignment, including library calls for arithmetic operators.
4128 Assuming it is a call-clobbered register, this may happen to @code{r0}
4129 above by the assignment to @code{p2}. If you have to use such a
4130 register, use temporary variables for expressions between the register
4135 register int *p1 asm ("r0") = @dots{};
4136 register int *p2 asm ("r1") = t1;
4137 register int *result asm ("r0");
4138 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4141 Some instructions clobber specific hard registers. To describe this,
4142 write a third colon after the input operands, followed by the names of
4143 the clobbered hard registers (given as strings). Here is a realistic
4144 example for the VAX:
4147 asm volatile ("movc3 %0,%1,%2"
4148 : /* @r{no outputs} */
4149 : "g" (from), "g" (to), "g" (count)
4150 : "r0", "r1", "r2", "r3", "r4", "r5");
4153 You may not write a clobber description in a way that overlaps with an
4154 input or output operand. For example, you may not have an operand
4155 describing a register class with one member if you mention that register
4156 in the clobber list. Variables declared to live in specific registers
4157 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4158 have no part mentioned in the clobber description.
4159 There is no way for you to specify that an input
4160 operand is modified without also specifying it as an output
4161 operand. Note that if all the output operands you specify are for this
4162 purpose (and hence unused), you will then also need to specify
4163 @code{volatile} for the @code{asm} construct, as described below, to
4164 prevent GCC from deleting the @code{asm} statement as unused.
4166 If you refer to a particular hardware register from the assembler code,
4167 you will probably have to list the register after the third colon to
4168 tell the compiler the register's value is modified. In some assemblers,
4169 the register names begin with @samp{%}; to produce one @samp{%} in the
4170 assembler code, you must write @samp{%%} in the input.
4172 If your assembler instruction can alter the condition code register, add
4173 @samp{cc} to the list of clobbered registers. GCC on some machines
4174 represents the condition codes as a specific hardware register;
4175 @samp{cc} serves to name this register. On other machines, the
4176 condition code is handled differently, and specifying @samp{cc} has no
4177 effect. But it is valid no matter what the machine.
4179 If your assembler instructions access memory in an unpredictable
4180 fashion, add @samp{memory} to the list of clobbered registers. This
4181 will cause GCC to not keep memory values cached in registers across the
4182 assembler instruction and not optimize stores or loads to that memory.
4183 You will also want to add the @code{volatile} keyword if the memory
4184 affected is not listed in the inputs or outputs of the @code{asm}, as
4185 the @samp{memory} clobber does not count as a side-effect of the
4186 @code{asm}. If you know how large the accessed memory is, you can add
4187 it as input or output but if this is not known, you should add
4188 @samp{memory}. As an example, if you access ten bytes of a string, you
4189 can use a memory input like:
4192 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4195 Note that in the following example the memory input is necessary,
4196 otherwise GCC might optimize the store to @code{x} away:
4203 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4204 "=&d" (r) : "a" (y), "m" (*y));
4209 You can put multiple assembler instructions together in a single
4210 @code{asm} template, separated by the characters normally used in assembly
4211 code for the system. A combination that works in most places is a newline
4212 to break the line, plus a tab character to move to the instruction field
4213 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4214 assembler allows semicolons as a line-breaking character. Note that some
4215 assembler dialects use semicolons to start a comment.
4216 The input operands are guaranteed not to use any of the clobbered
4217 registers, and neither will the output operands' addresses, so you can
4218 read and write the clobbered registers as many times as you like. Here
4219 is an example of multiple instructions in a template; it assumes the
4220 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4223 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4225 : "g" (from), "g" (to)
4229 Unless an output operand has the @samp{&} constraint modifier, GCC
4230 may allocate it in the same register as an unrelated input operand, on
4231 the assumption the inputs are consumed before the outputs are produced.
4232 This assumption may be false if the assembler code actually consists of
4233 more than one instruction. In such a case, use @samp{&} for each output
4234 operand that may not overlap an input. @xref{Modifiers}.
4236 If you want to test the condition code produced by an assembler
4237 instruction, you must include a branch and a label in the @code{asm}
4238 construct, as follows:
4241 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4247 This assumes your assembler supports local labels, as the GNU assembler
4248 and most Unix assemblers do.
4250 Speaking of labels, jumps from one @code{asm} to another are not
4251 supported. The compiler's optimizers do not know about these jumps, and
4252 therefore they cannot take account of them when deciding how to
4255 @cindex macros containing @code{asm}
4256 Usually the most convenient way to use these @code{asm} instructions is to
4257 encapsulate them in macros that look like functions. For example,
4261 (@{ double __value, __arg = (x); \
4262 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4267 Here the variable @code{__arg} is used to make sure that the instruction
4268 operates on a proper @code{double} value, and to accept only those
4269 arguments @code{x} which can convert automatically to a @code{double}.
4271 Another way to make sure the instruction operates on the correct data
4272 type is to use a cast in the @code{asm}. This is different from using a
4273 variable @code{__arg} in that it converts more different types. For
4274 example, if the desired type were @code{int}, casting the argument to
4275 @code{int} would accept a pointer with no complaint, while assigning the
4276 argument to an @code{int} variable named @code{__arg} would warn about
4277 using a pointer unless the caller explicitly casts it.
4279 If an @code{asm} has output operands, GCC assumes for optimization
4280 purposes the instruction has no side effects except to change the output
4281 operands. This does not mean instructions with a side effect cannot be
4282 used, but you must be careful, because the compiler may eliminate them
4283 if the output operands aren't used, or move them out of loops, or
4284 replace two with one if they constitute a common subexpression. Also,
4285 if your instruction does have a side effect on a variable that otherwise
4286 appears not to change, the old value of the variable may be reused later
4287 if it happens to be found in a register.
4289 You can prevent an @code{asm} instruction from being deleted
4290 by writing the keyword @code{volatile} after
4291 the @code{asm}. For example:
4294 #define get_and_set_priority(new) \
4296 asm volatile ("get_and_set_priority %0, %1" \
4297 : "=g" (__old) : "g" (new)); \
4302 The @code{volatile} keyword indicates that the instruction has
4303 important side-effects. GCC will not delete a volatile @code{asm} if
4304 it is reachable. (The instruction can still be deleted if GCC can
4305 prove that control-flow will never reach the location of the
4306 instruction.) Note that even a volatile @code{asm} instruction
4307 can be moved relative to other code, including across jump
4308 instructions. For example, on many targets there is a system
4309 register which can be set to control the rounding mode of
4310 floating point operations. You might try
4311 setting it with a volatile @code{asm}, like this PowerPC example:
4314 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4319 This will not work reliably, as the compiler may move the addition back
4320 before the volatile @code{asm}. To make it work you need to add an
4321 artificial dependency to the @code{asm} referencing a variable in the code
4322 you don't want moved, for example:
4325 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4329 Similarly, you can't expect a
4330 sequence of volatile @code{asm} instructions to remain perfectly
4331 consecutive. If you want consecutive output, use a single @code{asm}.
4332 Also, GCC will perform some optimizations across a volatile @code{asm}
4333 instruction; GCC does not ``forget everything'' when it encounters
4334 a volatile @code{asm} instruction the way some other compilers do.
4336 An @code{asm} instruction without any output operands will be treated
4337 identically to a volatile @code{asm} instruction.
4339 It is a natural idea to look for a way to give access to the condition
4340 code left by the assembler instruction. However, when we attempted to
4341 implement this, we found no way to make it work reliably. The problem
4342 is that output operands might need reloading, which would result in
4343 additional following ``store'' instructions. On most machines, these
4344 instructions would alter the condition code before there was time to
4345 test it. This problem doesn't arise for ordinary ``test'' and
4346 ``compare'' instructions because they don't have any output operands.
4348 For reasons similar to those described above, it is not possible to give
4349 an assembler instruction access to the condition code left by previous
4352 If you are writing a header file that should be includable in ISO C
4353 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4356 @subsection Size of an @code{asm}
4358 Some targets require that GCC track the size of each instruction used in
4359 order to generate correct code. Because the final length of an
4360 @code{asm} is only known by the assembler, GCC must make an estimate as
4361 to how big it will be. The estimate is formed by counting the number of
4362 statements in the pattern of the @code{asm} and multiplying that by the
4363 length of the longest instruction on that processor. Statements in the
4364 @code{asm} are identified by newline characters and whatever statement
4365 separator characters are supported by the assembler; on most processors
4366 this is the `@code{;}' character.
4368 Normally, GCC's estimate is perfectly adequate to ensure that correct
4369 code is generated, but it is possible to confuse the compiler if you use
4370 pseudo instructions or assembler macros that expand into multiple real
4371 instructions or if you use assembler directives that expand to more
4372 space in the object file than would be needed for a single instruction.
4373 If this happens then the assembler will produce a diagnostic saying that
4374 a label is unreachable.
4376 @subsection i386 floating point asm operands
4378 There are several rules on the usage of stack-like regs in
4379 asm_operands insns. These rules apply only to the operands that are
4384 Given a set of input regs that die in an asm_operands, it is
4385 necessary to know which are implicitly popped by the asm, and
4386 which must be explicitly popped by gcc.
4388 An input reg that is implicitly popped by the asm must be
4389 explicitly clobbered, unless it is constrained to match an
4393 For any input reg that is implicitly popped by an asm, it is
4394 necessary to know how to adjust the stack to compensate for the pop.
4395 If any non-popped input is closer to the top of the reg-stack than
4396 the implicitly popped reg, it would not be possible to know what the
4397 stack looked like---it's not clear how the rest of the stack ``slides
4400 All implicitly popped input regs must be closer to the top of
4401 the reg-stack than any input that is not implicitly popped.
4403 It is possible that if an input dies in an insn, reload might
4404 use the input reg for an output reload. Consider this example:
4407 asm ("foo" : "=t" (a) : "f" (b));
4410 This asm says that input B is not popped by the asm, and that
4411 the asm pushes a result onto the reg-stack, i.e., the stack is one
4412 deeper after the asm than it was before. But, it is possible that
4413 reload will think that it can use the same reg for both the input and
4414 the output, if input B dies in this insn.
4416 If any input operand uses the @code{f} constraint, all output reg
4417 constraints must use the @code{&} earlyclobber.
4419 The asm above would be written as
4422 asm ("foo" : "=&t" (a) : "f" (b));
4426 Some operands need to be in particular places on the stack. All
4427 output operands fall in this category---there is no other way to
4428 know which regs the outputs appear in unless the user indicates
4429 this in the constraints.
4431 Output operands must specifically indicate which reg an output
4432 appears in after an asm. @code{=f} is not allowed: the operand
4433 constraints must select a class with a single reg.
4436 Output operands may not be ``inserted'' between existing stack regs.
4437 Since no 387 opcode uses a read/write operand, all output operands
4438 are dead before the asm_operands, and are pushed by the asm_operands.
4439 It makes no sense to push anywhere but the top of the reg-stack.
4441 Output operands must start at the top of the reg-stack: output
4442 operands may not ``skip'' a reg.
4445 Some asm statements may need extra stack space for internal
4446 calculations. This can be guaranteed by clobbering stack registers
4447 unrelated to the inputs and outputs.
4451 Here are a couple of reasonable asms to want to write. This asm
4452 takes one input, which is internally popped, and produces two outputs.
4455 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4458 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4459 and replaces them with one output. The user must code the @code{st(1)}
4460 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4463 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4469 @section Controlling Names Used in Assembler Code
4470 @cindex assembler names for identifiers
4471 @cindex names used in assembler code
4472 @cindex identifiers, names in assembler code
4474 You can specify the name to be used in the assembler code for a C
4475 function or variable by writing the @code{asm} (or @code{__asm__})
4476 keyword after the declarator as follows:
4479 int foo asm ("myfoo") = 2;
4483 This specifies that the name to be used for the variable @code{foo} in
4484 the assembler code should be @samp{myfoo} rather than the usual
4487 On systems where an underscore is normally prepended to the name of a C
4488 function or variable, this feature allows you to define names for the
4489 linker that do not start with an underscore.
4491 It does not make sense to use this feature with a non-static local
4492 variable since such variables do not have assembler names. If you are
4493 trying to put the variable in a particular register, see @ref{Explicit
4494 Reg Vars}. GCC presently accepts such code with a warning, but will
4495 probably be changed to issue an error, rather than a warning, in the
4498 You cannot use @code{asm} in this way in a function @emph{definition}; but
4499 you can get the same effect by writing a declaration for the function
4500 before its definition and putting @code{asm} there, like this:
4503 extern func () asm ("FUNC");
4510 It is up to you to make sure that the assembler names you choose do not
4511 conflict with any other assembler symbols. Also, you must not use a
4512 register name; that would produce completely invalid assembler code. GCC
4513 does not as yet have the ability to store static variables in registers.
4514 Perhaps that will be added.
4516 @node Explicit Reg Vars
4517 @section Variables in Specified Registers
4518 @cindex explicit register variables
4519 @cindex variables in specified registers
4520 @cindex specified registers
4521 @cindex registers, global allocation
4523 GNU C allows you to put a few global variables into specified hardware
4524 registers. You can also specify the register in which an ordinary
4525 register variable should be allocated.
4529 Global register variables reserve registers throughout the program.
4530 This may be useful in programs such as programming language
4531 interpreters which have a couple of global variables that are accessed
4535 Local register variables in specific registers do not reserve the
4536 registers, except at the point where they are used as input or output
4537 operands in an @code{asm} statement and the @code{asm} statement itself is
4538 not deleted. The compiler's data flow analysis is capable of determining
4539 where the specified registers contain live values, and where they are
4540 available for other uses. Stores into local register variables may be deleted
4541 when they appear to be dead according to dataflow analysis. References
4542 to local register variables may be deleted or moved or simplified.
4544 These local variables are sometimes convenient for use with the extended
4545 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4546 output of the assembler instruction directly into a particular register.
4547 (This will work provided the register you specify fits the constraints
4548 specified for that operand in the @code{asm}.)
4556 @node Global Reg Vars
4557 @subsection Defining Global Register Variables
4558 @cindex global register variables
4559 @cindex registers, global variables in
4561 You can define a global register variable in GNU C like this:
4564 register int *foo asm ("a5");
4568 Here @code{a5} is the name of the register which should be used. Choose a
4569 register which is normally saved and restored by function calls on your
4570 machine, so that library routines will not clobber it.
4572 Naturally the register name is cpu-dependent, so you would need to
4573 conditionalize your program according to cpu type. The register
4574 @code{a5} would be a good choice on a 68000 for a variable of pointer
4575 type. On machines with register windows, be sure to choose a ``global''
4576 register that is not affected magically by the function call mechanism.
4578 In addition, operating systems on one type of cpu may differ in how they
4579 name the registers; then you would need additional conditionals. For
4580 example, some 68000 operating systems call this register @code{%a5}.
4582 Eventually there may be a way of asking the compiler to choose a register
4583 automatically, but first we need to figure out how it should choose and
4584 how to enable you to guide the choice. No solution is evident.
4586 Defining a global register variable in a certain register reserves that
4587 register entirely for this use, at least within the current compilation.
4588 The register will not be allocated for any other purpose in the functions
4589 in the current compilation. The register will not be saved and restored by
4590 these functions. Stores into this register are never deleted even if they
4591 would appear to be dead, but references may be deleted or moved or
4594 It is not safe to access the global register variables from signal
4595 handlers, or from more than one thread of control, because the system
4596 library routines may temporarily use the register for other things (unless
4597 you recompile them specially for the task at hand).
4599 @cindex @code{qsort}, and global register variables
4600 It is not safe for one function that uses a global register variable to
4601 call another such function @code{foo} by way of a third function
4602 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4603 different source file in which the variable wasn't declared). This is
4604 because @code{lose} might save the register and put some other value there.
4605 For example, you can't expect a global register variable to be available in
4606 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4607 might have put something else in that register. (If you are prepared to
4608 recompile @code{qsort} with the same global register variable, you can
4609 solve this problem.)
4611 If you want to recompile @code{qsort} or other source files which do not
4612 actually use your global register variable, so that they will not use that
4613 register for any other purpose, then it suffices to specify the compiler
4614 option @option{-ffixed-@var{reg}}. You need not actually add a global
4615 register declaration to their source code.
4617 A function which can alter the value of a global register variable cannot
4618 safely be called from a function compiled without this variable, because it
4619 could clobber the value the caller expects to find there on return.
4620 Therefore, the function which is the entry point into the part of the
4621 program that uses the global register variable must explicitly save and
4622 restore the value which belongs to its caller.
4624 @cindex register variable after @code{longjmp}
4625 @cindex global register after @code{longjmp}
4626 @cindex value after @code{longjmp}
4629 On most machines, @code{longjmp} will restore to each global register
4630 variable the value it had at the time of the @code{setjmp}. On some
4631 machines, however, @code{longjmp} will not change the value of global
4632 register variables. To be portable, the function that called @code{setjmp}
4633 should make other arrangements to save the values of the global register
4634 variables, and to restore them in a @code{longjmp}. This way, the same
4635 thing will happen regardless of what @code{longjmp} does.
4637 All global register variable declarations must precede all function
4638 definitions. If such a declaration could appear after function
4639 definitions, the declaration would be too late to prevent the register from
4640 being used for other purposes in the preceding functions.
4642 Global register variables may not have initial values, because an
4643 executable file has no means to supply initial contents for a register.
4645 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4646 registers, but certain library functions, such as @code{getwd}, as well
4647 as the subroutines for division and remainder, modify g3 and g4. g1 and
4648 g2 are local temporaries.
4650 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4651 Of course, it will not do to use more than a few of those.
4653 @node Local Reg Vars
4654 @subsection Specifying Registers for Local Variables
4655 @cindex local variables, specifying registers
4656 @cindex specifying registers for local variables
4657 @cindex registers for local variables
4659 You can define a local register variable with a specified register
4663 register int *foo asm ("a5");
4667 Here @code{a5} is the name of the register which should be used. Note
4668 that this is the same syntax used for defining global register
4669 variables, but for a local variable it would appear within a function.
4671 Naturally the register name is cpu-dependent, but this is not a
4672 problem, since specific registers are most often useful with explicit
4673 assembler instructions (@pxref{Extended Asm}). Both of these things
4674 generally require that you conditionalize your program according to
4677 In addition, operating systems on one type of cpu may differ in how they
4678 name the registers; then you would need additional conditionals. For
4679 example, some 68000 operating systems call this register @code{%a5}.
4681 Defining such a register variable does not reserve the register; it
4682 remains available for other uses in places where flow control determines
4683 the variable's value is not live.
4685 This option does not guarantee that GCC will generate code that has
4686 this variable in the register you specify at all times. You may not
4687 code an explicit reference to this register in the @emph{assembler
4688 instruction template} part of an @code{asm} statement and assume it will
4689 always refer to this variable. However, using the variable as an
4690 @code{asm} @emph{operand} guarantees that the specified register is used
4693 Stores into local register variables may be deleted when they appear to be dead
4694 according to dataflow analysis. References to local register variables may
4695 be deleted or moved or simplified.
4697 As for global register variables, it's recommended that you choose a
4698 register which is normally saved and restored by function calls on
4699 your machine, so that library routines will not clobber it. A common
4700 pitfall is to initialize multiple call-clobbered registers with
4701 arbitrary expressions, where a function call or library call for an
4702 arithmetic operator will overwrite a register value from a previous
4703 assignment, for example @code{r0} below:
4705 register int *p1 asm ("r0") = @dots{};
4706 register int *p2 asm ("r1") = @dots{};
4708 In those cases, a solution is to use a temporary variable for
4709 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4711 @node Alternate Keywords
4712 @section Alternate Keywords
4713 @cindex alternate keywords
4714 @cindex keywords, alternate
4716 @option{-ansi} and the various @option{-std} options disable certain
4717 keywords. This causes trouble when you want to use GNU C extensions, or
4718 a general-purpose header file that should be usable by all programs,
4719 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4720 @code{inline} are not available in programs compiled with
4721 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4722 program compiled with @option{-std=c99}). The ISO C99 keyword
4723 @code{restrict} is only available when @option{-std=gnu99} (which will
4724 eventually be the default) or @option{-std=c99} (or the equivalent
4725 @option{-std=iso9899:1999}) is used.
4727 The way to solve these problems is to put @samp{__} at the beginning and
4728 end of each problematical keyword. For example, use @code{__asm__}
4729 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4731 Other C compilers won't accept these alternative keywords; if you want to
4732 compile with another compiler, you can define the alternate keywords as
4733 macros to replace them with the customary keywords. It looks like this:
4741 @findex __extension__
4743 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4745 prevent such warnings within one expression by writing
4746 @code{__extension__} before the expression. @code{__extension__} has no
4747 effect aside from this.
4749 @node Incomplete Enums
4750 @section Incomplete @code{enum} Types
4752 You can define an @code{enum} tag without specifying its possible values.
4753 This results in an incomplete type, much like what you get if you write
4754 @code{struct foo} without describing the elements. A later declaration
4755 which does specify the possible values completes the type.
4757 You can't allocate variables or storage using the type while it is
4758 incomplete. However, you can work with pointers to that type.
4760 This extension may not be very useful, but it makes the handling of
4761 @code{enum} more consistent with the way @code{struct} and @code{union}
4764 This extension is not supported by GNU C++.
4766 @node Function Names
4767 @section Function Names as Strings
4768 @cindex @code{__func__} identifier
4769 @cindex @code{__FUNCTION__} identifier
4770 @cindex @code{__PRETTY_FUNCTION__} identifier
4772 GCC provides three magic variables which hold the name of the current
4773 function, as a string. The first of these is @code{__func__}, which
4774 is part of the C99 standard:
4777 The identifier @code{__func__} is implicitly declared by the translator
4778 as if, immediately following the opening brace of each function
4779 definition, the declaration
4782 static const char __func__[] = "function-name";
4785 appeared, where function-name is the name of the lexically-enclosing
4786 function. This name is the unadorned name of the function.
4789 @code{__FUNCTION__} is another name for @code{__func__}. Older
4790 versions of GCC recognize only this name. However, it is not
4791 standardized. For maximum portability, we recommend you use
4792 @code{__func__}, but provide a fallback definition with the
4796 #if __STDC_VERSION__ < 199901L
4798 # define __func__ __FUNCTION__
4800 # define __func__ "<unknown>"
4805 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4806 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4807 the type signature of the function as well as its bare name. For
4808 example, this program:
4812 extern int printf (char *, ...);
4819 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4820 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4838 __PRETTY_FUNCTION__ = void a::sub(int)
4841 These identifiers are not preprocessor macros. In GCC 3.3 and
4842 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4843 were treated as string literals; they could be used to initialize
4844 @code{char} arrays, and they could be concatenated with other string
4845 literals. GCC 3.4 and later treat them as variables, like
4846 @code{__func__}. In C++, @code{__FUNCTION__} and
4847 @code{__PRETTY_FUNCTION__} have always been variables.
4849 @node Return Address
4850 @section Getting the Return or Frame Address of a Function
4852 These functions may be used to get information about the callers of a
4855 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4856 This function returns the return address of the current function, or of
4857 one of its callers. The @var{level} argument is number of frames to
4858 scan up the call stack. A value of @code{0} yields the return address
4859 of the current function, a value of @code{1} yields the return address
4860 of the caller of the current function, and so forth. When inlining
4861 the expected behavior is that the function will return the address of
4862 the function that will be returned to. To work around this behavior use
4863 the @code{noinline} function attribute.
4865 The @var{level} argument must be a constant integer.
4867 On some machines it may be impossible to determine the return address of
4868 any function other than the current one; in such cases, or when the top
4869 of the stack has been reached, this function will return @code{0} or a
4870 random value. In addition, @code{__builtin_frame_address} may be used
4871 to determine if the top of the stack has been reached.
4873 This function should only be used with a nonzero argument for debugging
4877 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4878 This function is similar to @code{__builtin_return_address}, but it
4879 returns the address of the function frame rather than the return address
4880 of the function. Calling @code{__builtin_frame_address} with a value of
4881 @code{0} yields the frame address of the current function, a value of
4882 @code{1} yields the frame address of the caller of the current function,
4885 The frame is the area on the stack which holds local variables and saved
4886 registers. The frame address is normally the address of the first word
4887 pushed on to the stack by the function. However, the exact definition
4888 depends upon the processor and the calling convention. If the processor
4889 has a dedicated frame pointer register, and the function has a frame,
4890 then @code{__builtin_frame_address} will return the value of the frame
4893 On some machines it may be impossible to determine the frame address of
4894 any function other than the current one; in such cases, or when the top
4895 of the stack has been reached, this function will return @code{0} if
4896 the first frame pointer is properly initialized by the startup code.
4898 This function should only be used with a nonzero argument for debugging
4902 @node Vector Extensions
4903 @section Using vector instructions through built-in functions
4905 On some targets, the instruction set contains SIMD vector instructions that
4906 operate on multiple values contained in one large register at the same time.
4907 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4910 The first step in using these extensions is to provide the necessary data
4911 types. This should be done using an appropriate @code{typedef}:
4914 typedef int v4si __attribute__ ((vector_size (16)));
4917 The @code{int} type specifies the base type, while the attribute specifies
4918 the vector size for the variable, measured in bytes. For example, the
4919 declaration above causes the compiler to set the mode for the @code{v4si}
4920 type to be 16 bytes wide and divided into @code{int} sized units. For
4921 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4922 corresponding mode of @code{foo} will be @acronym{V4SI}.
4924 The @code{vector_size} attribute is only applicable to integral and
4925 float scalars, although arrays, pointers, and function return values
4926 are allowed in conjunction with this construct.
4928 All the basic integer types can be used as base types, both as signed
4929 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4930 @code{long long}. In addition, @code{float} and @code{double} can be
4931 used to build floating-point vector types.
4933 Specifying a combination that is not valid for the current architecture
4934 will cause GCC to synthesize the instructions using a narrower mode.
4935 For example, if you specify a variable of type @code{V4SI} and your
4936 architecture does not allow for this specific SIMD type, GCC will
4937 produce code that uses 4 @code{SIs}.
4939 The types defined in this manner can be used with a subset of normal C
4940 operations. Currently, GCC will allow using the following operators
4941 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4943 The operations behave like C++ @code{valarrays}. Addition is defined as
4944 the addition of the corresponding elements of the operands. For
4945 example, in the code below, each of the 4 elements in @var{a} will be
4946 added to the corresponding 4 elements in @var{b} and the resulting
4947 vector will be stored in @var{c}.
4950 typedef int v4si __attribute__ ((vector_size (16)));
4957 Subtraction, multiplication, division, and the logical operations
4958 operate in a similar manner. Likewise, the result of using the unary
4959 minus or complement operators on a vector type is a vector whose
4960 elements are the negative or complemented values of the corresponding
4961 elements in the operand.
4963 You can declare variables and use them in function calls and returns, as
4964 well as in assignments and some casts. You can specify a vector type as
4965 a return type for a function. Vector types can also be used as function
4966 arguments. It is possible to cast from one vector type to another,
4967 provided they are of the same size (in fact, you can also cast vectors
4968 to and from other datatypes of the same size).
4970 You cannot operate between vectors of different lengths or different
4971 signedness without a cast.
4973 A port that supports hardware vector operations, usually provides a set
4974 of built-in functions that can be used to operate on vectors. For
4975 example, a function to add two vectors and multiply the result by a
4976 third could look like this:
4979 v4si f (v4si a, v4si b, v4si c)
4981 v4si tmp = __builtin_addv4si (a, b);
4982 return __builtin_mulv4si (tmp, c);
4989 @findex __builtin_offsetof
4991 GCC implements for both C and C++ a syntactic extension to implement
4992 the @code{offsetof} macro.
4996 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4998 offsetof_member_designator:
5000 | offsetof_member_designator "." @code{identifier}
5001 | offsetof_member_designator "[" @code{expr} "]"
5004 This extension is sufficient such that
5007 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5010 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5011 may be dependent. In either case, @var{member} may consist of a single
5012 identifier, or a sequence of member accesses and array references.
5014 @node Atomic Builtins
5015 @section Built-in functions for atomic memory access
5017 The following builtins are intended to be compatible with those described
5018 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5019 section 7.4. As such, they depart from the normal GCC practice of using
5020 the ``__builtin_'' prefix, and further that they are overloaded such that
5021 they work on multiple types.
5023 The definition given in the Intel documentation allows only for the use of
5024 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5025 counterparts. GCC will allow any integral scalar or pointer type that is
5026 1, 2, 4 or 8 bytes in length.
5028 Not all operations are supported by all target processors. If a particular
5029 operation cannot be implemented on the target processor, a warning will be
5030 generated and a call an external function will be generated. The external
5031 function will carry the same name as the builtin, with an additional suffix
5032 @samp{_@var{n}} where @var{n} is the size of the data type.
5034 @c ??? Should we have a mechanism to suppress this warning? This is almost
5035 @c useful for implementing the operation under the control of an external
5038 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5039 no memory operand will be moved across the operation, either forward or
5040 backward. Further, instructions will be issued as necessary to prevent the
5041 processor from speculating loads across the operation and from queuing stores
5042 after the operation.
5044 All of the routines are are described in the Intel documentation to take
5045 ``an optional list of variables protected by the memory barrier''. It's
5046 not clear what is meant by that; it could mean that @emph{only} the
5047 following variables are protected, or it could mean that these variables
5048 should in addition be protected. At present GCC ignores this list and
5049 protects all variables which are globally accessible. If in the future
5050 we make some use of this list, an empty list will continue to mean all
5051 globally accessible variables.
5054 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5055 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5056 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5057 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5058 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5059 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5060 @findex __sync_fetch_and_add
5061 @findex __sync_fetch_and_sub
5062 @findex __sync_fetch_and_or
5063 @findex __sync_fetch_and_and
5064 @findex __sync_fetch_and_xor
5065 @findex __sync_fetch_and_nand
5066 These builtins perform the operation suggested by the name, and
5067 returns the value that had previously been in memory. That is,
5070 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5071 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5074 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5075 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5076 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5077 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5078 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5079 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5080 @findex __sync_add_and_fetch
5081 @findex __sync_sub_and_fetch
5082 @findex __sync_or_and_fetch
5083 @findex __sync_and_and_fetch
5084 @findex __sync_xor_and_fetch
5085 @findex __sync_nand_and_fetch
5086 These builtins perform the operation suggested by the name, and
5087 return the new value. That is,
5090 @{ *ptr @var{op}= value; return *ptr; @}
5091 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5094 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5095 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5096 @findex __sync_bool_compare_and_swap
5097 @findex __sync_val_compare_and_swap
5098 These builtins perform an atomic compare and swap. That is, if the current
5099 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5102 The ``bool'' version returns true if the comparison is successful and
5103 @var{newval} was written. The ``val'' version returns the contents
5104 of @code{*@var{ptr}} before the operation.
5106 @item __sync_synchronize (...)
5107 @findex __sync_synchronize
5108 This builtin issues a full memory barrier.
5110 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5111 @findex __sync_lock_test_and_set
5112 This builtin, as described by Intel, is not a traditional test-and-set
5113 operation, but rather an atomic exchange operation. It writes @var{value}
5114 into @code{*@var{ptr}}, and returns the previous contents of
5117 Many targets have only minimal support for such locks, and do not support
5118 a full exchange operation. In this case, a target may support reduced
5119 functionality here by which the @emph{only} valid value to store is the
5120 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5121 is implementation defined.
5123 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5124 This means that references after the builtin cannot move to (or be
5125 speculated to) before the builtin, but previous memory stores may not
5126 be globally visible yet, and previous memory loads may not yet be
5129 @item void __sync_lock_release (@var{type} *ptr, ...)
5130 @findex __sync_lock_release
5131 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5132 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5134 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5135 This means that all previous memory stores are globally visible, and all
5136 previous memory loads have been satisfied, but following memory reads
5137 are not prevented from being speculated to before the barrier.
5140 @node Object Size Checking
5141 @section Object Size Checking Builtins
5142 @findex __builtin_object_size
5143 @findex __builtin___memcpy_chk
5144 @findex __builtin___mempcpy_chk
5145 @findex __builtin___memmove_chk
5146 @findex __builtin___memset_chk
5147 @findex __builtin___strcpy_chk
5148 @findex __builtin___stpcpy_chk
5149 @findex __builtin___strncpy_chk
5150 @findex __builtin___strcat_chk
5151 @findex __builtin___strncat_chk
5152 @findex __builtin___sprintf_chk
5153 @findex __builtin___snprintf_chk
5154 @findex __builtin___vsprintf_chk
5155 @findex __builtin___vsnprintf_chk
5156 @findex __builtin___printf_chk
5157 @findex __builtin___vprintf_chk
5158 @findex __builtin___fprintf_chk
5159 @findex __builtin___vfprintf_chk
5161 GCC implements a limited buffer overflow protection mechanism
5162 that can prevent some buffer overflow attacks.
5164 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5165 is a built-in construct that returns a constant number of bytes from
5166 @var{ptr} to the end of the object @var{ptr} pointer points to
5167 (if known at compile time). @code{__builtin_object_size} never evaluates
5168 its arguments for side-effects. If there are any side-effects in them, it
5169 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5170 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5171 point to and all of them are known at compile time, the returned number
5172 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5173 0 and minimum if nonzero. If it is not possible to determine which objects
5174 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5175 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5176 for @var{type} 2 or 3.
5178 @var{type} is an integer constant from 0 to 3. If the least significant
5179 bit is clear, objects are whole variables, if it is set, a closest
5180 surrounding subobject is considered the object a pointer points to.
5181 The second bit determines if maximum or minimum of remaining bytes
5185 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5186 char *p = &var.buf1[1], *q = &var.b;
5188 /* Here the object p points to is var. */
5189 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5190 /* The subobject p points to is var.buf1. */
5191 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5192 /* The object q points to is var. */
5193 assert (__builtin_object_size (q, 0)
5194 == (char *) (&var + 1) - (char *) &var.b);
5195 /* The subobject q points to is var.b. */
5196 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5200 There are built-in functions added for many common string operation
5201 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5202 built-in is provided. This built-in has an additional last argument,
5203 which is the number of bytes remaining in object the @var{dest}
5204 argument points to or @code{(size_t) -1} if the size is not known.
5206 The built-in functions are optimized into the normal string functions
5207 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5208 it is known at compile time that the destination object will not
5209 be overflown. If the compiler can determine at compile time the
5210 object will be always overflown, it issues a warning.
5212 The intended use can be e.g.
5216 #define bos0(dest) __builtin_object_size (dest, 0)
5217 #define memcpy(dest, src, n) \
5218 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5222 /* It is unknown what object p points to, so this is optimized
5223 into plain memcpy - no checking is possible. */
5224 memcpy (p, "abcde", n);
5225 /* Destination is known and length too. It is known at compile
5226 time there will be no overflow. */
5227 memcpy (&buf[5], "abcde", 5);
5228 /* Destination is known, but the length is not known at compile time.
5229 This will result in __memcpy_chk call that can check for overflow
5231 memcpy (&buf[5], "abcde", n);
5232 /* Destination is known and it is known at compile time there will
5233 be overflow. There will be a warning and __memcpy_chk call that
5234 will abort the program at runtime. */
5235 memcpy (&buf[6], "abcde", 5);
5238 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5239 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5240 @code{strcat} and @code{strncat}.
5242 There are also checking built-in functions for formatted output functions.
5244 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5245 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5246 const char *fmt, ...);
5247 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5249 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5250 const char *fmt, va_list ap);
5253 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5254 etc. functions and can contain implementation specific flags on what
5255 additional security measures the checking function might take, such as
5256 handling @code{%n} differently.
5258 The @var{os} argument is the object size @var{s} points to, like in the
5259 other built-in functions. There is a small difference in the behavior
5260 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5261 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5262 the checking function is called with @var{os} argument set to
5265 In addition to this, there are checking built-in functions
5266 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5267 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5268 These have just one additional argument, @var{flag}, right before
5269 format string @var{fmt}. If the compiler is able to optimize them to
5270 @code{fputc} etc. functions, it will, otherwise the checking function
5271 should be called and the @var{flag} argument passed to it.
5273 @node Other Builtins
5274 @section Other built-in functions provided by GCC
5275 @cindex built-in functions
5276 @findex __builtin_isgreater
5277 @findex __builtin_isgreaterequal
5278 @findex __builtin_isless
5279 @findex __builtin_islessequal
5280 @findex __builtin_islessgreater
5281 @findex __builtin_isunordered
5282 @findex __builtin_powi
5283 @findex __builtin_powif
5284 @findex __builtin_powil
5442 @findex fprintf_unlocked
5444 @findex fputs_unlocked
5554 @findex printf_unlocked
5583 @findex significandf
5584 @findex significandl
5655 GCC provides a large number of built-in functions other than the ones
5656 mentioned above. Some of these are for internal use in the processing
5657 of exceptions or variable-length argument lists and will not be
5658 documented here because they may change from time to time; we do not
5659 recommend general use of these functions.
5661 The remaining functions are provided for optimization purposes.
5663 @opindex fno-builtin
5664 GCC includes built-in versions of many of the functions in the standard
5665 C library. The versions prefixed with @code{__builtin_} will always be
5666 treated as having the same meaning as the C library function even if you
5667 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5668 Many of these functions are only optimized in certain cases; if they are
5669 not optimized in a particular case, a call to the library function will
5674 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5675 @option{-std=c99}), the functions
5676 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5677 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5678 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5679 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5680 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5681 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5682 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5683 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5684 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5685 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5686 @code{significandf}, @code{significandl}, @code{significand},
5687 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5688 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5689 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5690 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5691 @code{ynl} and @code{yn}
5692 may be handled as built-in functions.
5693 All these functions have corresponding versions
5694 prefixed with @code{__builtin_}, which may be used even in strict C89
5697 The ISO C99 functions
5698 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5699 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5700 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5701 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5702 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5703 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5704 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5705 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5706 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5707 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5708 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5709 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5710 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5711 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5712 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5713 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5714 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5715 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5716 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5717 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5718 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5719 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5720 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5721 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5722 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5723 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5724 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5725 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5726 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5727 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5728 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5729 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5730 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5731 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5732 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5733 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5734 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5735 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5736 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5737 are handled as built-in functions
5738 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5740 There are also built-in versions of the ISO C99 functions
5741 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5742 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5743 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5744 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5745 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5746 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5747 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5748 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5749 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5750 that are recognized in any mode since ISO C90 reserves these names for
5751 the purpose to which ISO C99 puts them. All these functions have
5752 corresponding versions prefixed with @code{__builtin_}.
5754 The ISO C94 functions
5755 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5756 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5757 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5759 are handled as built-in functions
5760 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5762 The ISO C90 functions
5763 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5764 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5765 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5766 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5767 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5768 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5769 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5770 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5771 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5772 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5773 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5774 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5775 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5776 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5777 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5778 @code{vprintf} and @code{vsprintf}
5779 are all recognized as built-in functions unless
5780 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5781 is specified for an individual function). All of these functions have
5782 corresponding versions prefixed with @code{__builtin_}.
5784 GCC provides built-in versions of the ISO C99 floating point comparison
5785 macros that avoid raising exceptions for unordered operands. They have
5786 the same names as the standard macros ( @code{isgreater},
5787 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5788 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5789 prefixed. We intend for a library implementor to be able to simply
5790 @code{#define} each standard macro to its built-in equivalent.
5792 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5794 You can use the built-in function @code{__builtin_types_compatible_p} to
5795 determine whether two types are the same.
5797 This built-in function returns 1 if the unqualified versions of the
5798 types @var{type1} and @var{type2} (which are types, not expressions) are
5799 compatible, 0 otherwise. The result of this built-in function can be
5800 used in integer constant expressions.
5802 This built-in function ignores top level qualifiers (e.g., @code{const},
5803 @code{volatile}). For example, @code{int} is equivalent to @code{const
5806 The type @code{int[]} and @code{int[5]} are compatible. On the other
5807 hand, @code{int} and @code{char *} are not compatible, even if the size
5808 of their types, on the particular architecture are the same. Also, the
5809 amount of pointer indirection is taken into account when determining
5810 similarity. Consequently, @code{short *} is not similar to
5811 @code{short **}. Furthermore, two types that are typedefed are
5812 considered compatible if their underlying types are compatible.
5814 An @code{enum} type is not considered to be compatible with another
5815 @code{enum} type even if both are compatible with the same integer
5816 type; this is what the C standard specifies.
5817 For example, @code{enum @{foo, bar@}} is not similar to
5818 @code{enum @{hot, dog@}}.
5820 You would typically use this function in code whose execution varies
5821 depending on the arguments' types. For example:
5826 typeof (x) tmp = (x); \
5827 if (__builtin_types_compatible_p (typeof (x), long double)) \
5828 tmp = foo_long_double (tmp); \
5829 else if (__builtin_types_compatible_p (typeof (x), double)) \
5830 tmp = foo_double (tmp); \
5831 else if (__builtin_types_compatible_p (typeof (x), float)) \
5832 tmp = foo_float (tmp); \
5839 @emph{Note:} This construct is only available for C@.
5843 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5845 You can use the built-in function @code{__builtin_choose_expr} to
5846 evaluate code depending on the value of a constant expression. This
5847 built-in function returns @var{exp1} if @var{const_exp}, which is a
5848 constant expression that must be able to be determined at compile time,
5849 is nonzero. Otherwise it returns 0.
5851 This built-in function is analogous to the @samp{? :} operator in C,
5852 except that the expression returned has its type unaltered by promotion
5853 rules. Also, the built-in function does not evaluate the expression
5854 that was not chosen. For example, if @var{const_exp} evaluates to true,
5855 @var{exp2} is not evaluated even if it has side-effects.
5857 This built-in function can return an lvalue if the chosen argument is an
5860 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5861 type. Similarly, if @var{exp2} is returned, its return type is the same
5868 __builtin_choose_expr ( \
5869 __builtin_types_compatible_p (typeof (x), double), \
5871 __builtin_choose_expr ( \
5872 __builtin_types_compatible_p (typeof (x), float), \
5874 /* @r{The void expression results in a compile-time error} \
5875 @r{when assigning the result to something.} */ \
5879 @emph{Note:} This construct is only available for C@. Furthermore, the
5880 unused expression (@var{exp1} or @var{exp2} depending on the value of
5881 @var{const_exp}) may still generate syntax errors. This may change in
5886 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5887 You can use the built-in function @code{__builtin_constant_p} to
5888 determine if a value is known to be constant at compile-time and hence
5889 that GCC can perform constant-folding on expressions involving that
5890 value. The argument of the function is the value to test. The function
5891 returns the integer 1 if the argument is known to be a compile-time
5892 constant and 0 if it is not known to be a compile-time constant. A
5893 return of 0 does not indicate that the value is @emph{not} a constant,
5894 but merely that GCC cannot prove it is a constant with the specified
5895 value of the @option{-O} option.
5897 You would typically use this function in an embedded application where
5898 memory was a critical resource. If you have some complex calculation,
5899 you may want it to be folded if it involves constants, but need to call
5900 a function if it does not. For example:
5903 #define Scale_Value(X) \
5904 (__builtin_constant_p (X) \
5905 ? ((X) * SCALE + OFFSET) : Scale (X))
5908 You may use this built-in function in either a macro or an inline
5909 function. However, if you use it in an inlined function and pass an
5910 argument of the function as the argument to the built-in, GCC will
5911 never return 1 when you call the inline function with a string constant
5912 or compound literal (@pxref{Compound Literals}) and will not return 1
5913 when you pass a constant numeric value to the inline function unless you
5914 specify the @option{-O} option.
5916 You may also use @code{__builtin_constant_p} in initializers for static
5917 data. For instance, you can write
5920 static const int table[] = @{
5921 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5927 This is an acceptable initializer even if @var{EXPRESSION} is not a
5928 constant expression. GCC must be more conservative about evaluating the
5929 built-in in this case, because it has no opportunity to perform
5932 Previous versions of GCC did not accept this built-in in data
5933 initializers. The earliest version where it is completely safe is
5937 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5938 @opindex fprofile-arcs
5939 You may use @code{__builtin_expect} to provide the compiler with
5940 branch prediction information. In general, you should prefer to
5941 use actual profile feedback for this (@option{-fprofile-arcs}), as
5942 programmers are notoriously bad at predicting how their programs
5943 actually perform. However, there are applications in which this
5944 data is hard to collect.
5946 The return value is the value of @var{exp}, which should be an integral
5947 expression. The semantics of the built-in are that it is expected that
5948 @var{exp} == @var{c}. For example:
5951 if (__builtin_expect (x, 0))
5956 would indicate that we do not expect to call @code{foo}, since
5957 we expect @code{x} to be zero. Since you are limited to integral
5958 expressions for @var{exp}, you should use constructions such as
5961 if (__builtin_expect (ptr != NULL, 1))
5966 when testing pointer or floating-point values.
5969 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5970 This function is used to minimize cache-miss latency by moving data into
5971 a cache before it is accessed.
5972 You can insert calls to @code{__builtin_prefetch} into code for which
5973 you know addresses of data in memory that is likely to be accessed soon.
5974 If the target supports them, data prefetch instructions will be generated.
5975 If the prefetch is done early enough before the access then the data will
5976 be in the cache by the time it is accessed.
5978 The value of @var{addr} is the address of the memory to prefetch.
5979 There are two optional arguments, @var{rw} and @var{locality}.
5980 The value of @var{rw} is a compile-time constant one or zero; one
5981 means that the prefetch is preparing for a write to the memory address
5982 and zero, the default, means that the prefetch is preparing for a read.
5983 The value @var{locality} must be a compile-time constant integer between
5984 zero and three. A value of zero means that the data has no temporal
5985 locality, so it need not be left in the cache after the access. A value
5986 of three means that the data has a high degree of temporal locality and
5987 should be left in all levels of cache possible. Values of one and two
5988 mean, respectively, a low or moderate degree of temporal locality. The
5992 for (i = 0; i < n; i++)
5995 __builtin_prefetch (&a[i+j], 1, 1);
5996 __builtin_prefetch (&b[i+j], 0, 1);
6001 Data prefetch does not generate faults if @var{addr} is invalid, but
6002 the address expression itself must be valid. For example, a prefetch
6003 of @code{p->next} will not fault if @code{p->next} is not a valid
6004 address, but evaluation will fault if @code{p} is not a valid address.
6006 If the target does not support data prefetch, the address expression
6007 is evaluated if it includes side effects but no other code is generated
6008 and GCC does not issue a warning.
6011 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6012 Returns a positive infinity, if supported by the floating-point format,
6013 else @code{DBL_MAX}. This function is suitable for implementing the
6014 ISO C macro @code{HUGE_VAL}.
6017 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6018 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6021 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6022 Similar to @code{__builtin_huge_val}, except the return
6023 type is @code{long double}.
6026 @deftypefn {Built-in Function} double __builtin_inf (void)
6027 Similar to @code{__builtin_huge_val}, except a warning is generated
6028 if the target floating-point format does not support infinities.
6031 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6032 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6035 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6036 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6039 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6040 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6043 @deftypefn {Built-in Function} float __builtin_inff (void)
6044 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6045 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6048 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6049 Similar to @code{__builtin_inf}, except the return
6050 type is @code{long double}.
6053 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6054 This is an implementation of the ISO C99 function @code{nan}.
6056 Since ISO C99 defines this function in terms of @code{strtod}, which we
6057 do not implement, a description of the parsing is in order. The string
6058 is parsed as by @code{strtol}; that is, the base is recognized by
6059 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6060 in the significand such that the least significant bit of the number
6061 is at the least significant bit of the significand. The number is
6062 truncated to fit the significand field provided. The significand is
6063 forced to be a quiet NaN@.
6065 This function, if given a string literal all of which would have been
6066 consumed by strtol, is evaluated early enough that it is considered a
6067 compile-time constant.
6070 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6071 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6074 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6075 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6078 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6079 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6082 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6083 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6086 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6087 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6090 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6091 Similar to @code{__builtin_nan}, except the significand is forced
6092 to be a signaling NaN@. The @code{nans} function is proposed by
6093 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6096 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6097 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6100 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6101 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6104 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6105 Returns one plus the index of the least significant 1-bit of @var{x}, or
6106 if @var{x} is zero, returns zero.
6109 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6110 Returns the number of leading 0-bits in @var{x}, starting at the most
6111 significant bit position. If @var{x} is 0, the result is undefined.
6114 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6115 Returns the number of trailing 0-bits in @var{x}, starting at the least
6116 significant bit position. If @var{x} is 0, the result is undefined.
6119 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6120 Returns the number of 1-bits in @var{x}.
6123 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6124 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6128 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6129 Similar to @code{__builtin_ffs}, except the argument type is
6130 @code{unsigned long}.
6133 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6134 Similar to @code{__builtin_clz}, except the argument type is
6135 @code{unsigned long}.
6138 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6139 Similar to @code{__builtin_ctz}, except the argument type is
6140 @code{unsigned long}.
6143 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6144 Similar to @code{__builtin_popcount}, except the argument type is
6145 @code{unsigned long}.
6148 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6149 Similar to @code{__builtin_parity}, except the argument type is
6150 @code{unsigned long}.
6153 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6154 Similar to @code{__builtin_ffs}, except the argument type is
6155 @code{unsigned long long}.
6158 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6159 Similar to @code{__builtin_clz}, except the argument type is
6160 @code{unsigned long long}.
6163 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6164 Similar to @code{__builtin_ctz}, except the argument type is
6165 @code{unsigned long long}.
6168 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6169 Similar to @code{__builtin_popcount}, except the argument type is
6170 @code{unsigned long long}.
6173 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6174 Similar to @code{__builtin_parity}, except the argument type is
6175 @code{unsigned long long}.
6178 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6179 Returns the first argument raised to the power of the second. Unlike the
6180 @code{pow} function no guarantees about precision and rounding are made.
6183 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6184 Similar to @code{__builtin_powi}, except the argument and return types
6188 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6189 Similar to @code{__builtin_powi}, except the argument and return types
6190 are @code{long double}.
6193 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6194 Returns @var{x} with the order of the bytes reversed; for example,
6195 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6199 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6200 Similar to @code{__builtin_bswap32}, except the argument and return types
6204 @node Target Builtins
6205 @section Built-in Functions Specific to Particular Target Machines
6207 On some target machines, GCC supports many built-in functions specific
6208 to those machines. Generally these generate calls to specific machine
6209 instructions, but allow the compiler to schedule those calls.
6212 * Alpha Built-in Functions::
6213 * ARM Built-in Functions::
6214 * Blackfin Built-in Functions::
6215 * FR-V Built-in Functions::
6216 * X86 Built-in Functions::
6217 * MIPS DSP Built-in Functions::
6218 * MIPS Paired-Single Support::
6219 * PowerPC AltiVec Built-in Functions::
6220 * SPARC VIS Built-in Functions::
6221 * SPU Built-in Functions::
6224 @node Alpha Built-in Functions
6225 @subsection Alpha Built-in Functions
6227 These built-in functions are available for the Alpha family of
6228 processors, depending on the command-line switches used.
6230 The following built-in functions are always available. They
6231 all generate the machine instruction that is part of the name.
6234 long __builtin_alpha_implver (void)
6235 long __builtin_alpha_rpcc (void)
6236 long __builtin_alpha_amask (long)
6237 long __builtin_alpha_cmpbge (long, long)
6238 long __builtin_alpha_extbl (long, long)
6239 long __builtin_alpha_extwl (long, long)
6240 long __builtin_alpha_extll (long, long)
6241 long __builtin_alpha_extql (long, long)
6242 long __builtin_alpha_extwh (long, long)
6243 long __builtin_alpha_extlh (long, long)
6244 long __builtin_alpha_extqh (long, long)
6245 long __builtin_alpha_insbl (long, long)
6246 long __builtin_alpha_inswl (long, long)
6247 long __builtin_alpha_insll (long, long)
6248 long __builtin_alpha_insql (long, long)
6249 long __builtin_alpha_inswh (long, long)
6250 long __builtin_alpha_inslh (long, long)
6251 long __builtin_alpha_insqh (long, long)
6252 long __builtin_alpha_mskbl (long, long)
6253 long __builtin_alpha_mskwl (long, long)
6254 long __builtin_alpha_mskll (long, long)
6255 long __builtin_alpha_mskql (long, long)
6256 long __builtin_alpha_mskwh (long, long)
6257 long __builtin_alpha_msklh (long, long)
6258 long __builtin_alpha_mskqh (long, long)
6259 long __builtin_alpha_umulh (long, long)
6260 long __builtin_alpha_zap (long, long)
6261 long __builtin_alpha_zapnot (long, long)
6264 The following built-in functions are always with @option{-mmax}
6265 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6266 later. They all generate the machine instruction that is part
6270 long __builtin_alpha_pklb (long)
6271 long __builtin_alpha_pkwb (long)
6272 long __builtin_alpha_unpkbl (long)
6273 long __builtin_alpha_unpkbw (long)
6274 long __builtin_alpha_minub8 (long, long)
6275 long __builtin_alpha_minsb8 (long, long)
6276 long __builtin_alpha_minuw4 (long, long)
6277 long __builtin_alpha_minsw4 (long, long)
6278 long __builtin_alpha_maxub8 (long, long)
6279 long __builtin_alpha_maxsb8 (long, long)
6280 long __builtin_alpha_maxuw4 (long, long)
6281 long __builtin_alpha_maxsw4 (long, long)
6282 long __builtin_alpha_perr (long, long)
6285 The following built-in functions are always with @option{-mcix}
6286 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6287 later. They all generate the machine instruction that is part
6291 long __builtin_alpha_cttz (long)
6292 long __builtin_alpha_ctlz (long)
6293 long __builtin_alpha_ctpop (long)
6296 The following builtins are available on systems that use the OSF/1
6297 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6298 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6299 @code{rdval} and @code{wrval}.
6302 void *__builtin_thread_pointer (void)
6303 void __builtin_set_thread_pointer (void *)
6306 @node ARM Built-in Functions
6307 @subsection ARM Built-in Functions
6309 These built-in functions are available for the ARM family of
6310 processors, when the @option{-mcpu=iwmmxt} switch is used:
6313 typedef int v2si __attribute__ ((vector_size (8)));
6314 typedef short v4hi __attribute__ ((vector_size (8)));
6315 typedef char v8qi __attribute__ ((vector_size (8)));
6317 int __builtin_arm_getwcx (int)
6318 void __builtin_arm_setwcx (int, int)
6319 int __builtin_arm_textrmsb (v8qi, int)
6320 int __builtin_arm_textrmsh (v4hi, int)
6321 int __builtin_arm_textrmsw (v2si, int)
6322 int __builtin_arm_textrmub (v8qi, int)
6323 int __builtin_arm_textrmuh (v4hi, int)
6324 int __builtin_arm_textrmuw (v2si, int)
6325 v8qi __builtin_arm_tinsrb (v8qi, int)
6326 v4hi __builtin_arm_tinsrh (v4hi, int)
6327 v2si __builtin_arm_tinsrw (v2si, int)
6328 long long __builtin_arm_tmia (long long, int, int)
6329 long long __builtin_arm_tmiabb (long long, int, int)
6330 long long __builtin_arm_tmiabt (long long, int, int)
6331 long long __builtin_arm_tmiaph (long long, int, int)
6332 long long __builtin_arm_tmiatb (long long, int, int)
6333 long long __builtin_arm_tmiatt (long long, int, int)
6334 int __builtin_arm_tmovmskb (v8qi)
6335 int __builtin_arm_tmovmskh (v4hi)
6336 int __builtin_arm_tmovmskw (v2si)
6337 long long __builtin_arm_waccb (v8qi)
6338 long long __builtin_arm_wacch (v4hi)
6339 long long __builtin_arm_waccw (v2si)
6340 v8qi __builtin_arm_waddb (v8qi, v8qi)
6341 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6342 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6343 v4hi __builtin_arm_waddh (v4hi, v4hi)
6344 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6345 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6346 v2si __builtin_arm_waddw (v2si, v2si)
6347 v2si __builtin_arm_waddwss (v2si, v2si)
6348 v2si __builtin_arm_waddwus (v2si, v2si)
6349 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6350 long long __builtin_arm_wand(long long, long long)
6351 long long __builtin_arm_wandn (long long, long long)
6352 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6353 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6354 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6355 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6356 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6357 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6358 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6359 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6360 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6361 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6362 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6363 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6364 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6365 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6366 long long __builtin_arm_wmacsz (v4hi, v4hi)
6367 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6368 long long __builtin_arm_wmacuz (v4hi, v4hi)
6369 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6370 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6371 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6372 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6373 v2si __builtin_arm_wmaxsw (v2si, v2si)
6374 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6375 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6376 v2si __builtin_arm_wmaxuw (v2si, v2si)
6377 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6378 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6379 v2si __builtin_arm_wminsw (v2si, v2si)
6380 v8qi __builtin_arm_wminub (v8qi, v8qi)
6381 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6382 v2si __builtin_arm_wminuw (v2si, v2si)
6383 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6384 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6385 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6386 long long __builtin_arm_wor (long long, long long)
6387 v2si __builtin_arm_wpackdss (long long, long long)
6388 v2si __builtin_arm_wpackdus (long long, long long)
6389 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6390 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6391 v4hi __builtin_arm_wpackwss (v2si, v2si)
6392 v4hi __builtin_arm_wpackwus (v2si, v2si)
6393 long long __builtin_arm_wrord (long long, long long)
6394 long long __builtin_arm_wrordi (long long, int)
6395 v4hi __builtin_arm_wrorh (v4hi, long long)
6396 v4hi __builtin_arm_wrorhi (v4hi, int)
6397 v2si __builtin_arm_wrorw (v2si, long long)
6398 v2si __builtin_arm_wrorwi (v2si, int)
6399 v2si __builtin_arm_wsadb (v8qi, v8qi)
6400 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6401 v2si __builtin_arm_wsadh (v4hi, v4hi)
6402 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6403 v4hi __builtin_arm_wshufh (v4hi, int)
6404 long long __builtin_arm_wslld (long long, long long)
6405 long long __builtin_arm_wslldi (long long, int)
6406 v4hi __builtin_arm_wsllh (v4hi, long long)
6407 v4hi __builtin_arm_wsllhi (v4hi, int)
6408 v2si __builtin_arm_wsllw (v2si, long long)
6409 v2si __builtin_arm_wsllwi (v2si, int)
6410 long long __builtin_arm_wsrad (long long, long long)
6411 long long __builtin_arm_wsradi (long long, int)
6412 v4hi __builtin_arm_wsrah (v4hi, long long)
6413 v4hi __builtin_arm_wsrahi (v4hi, int)
6414 v2si __builtin_arm_wsraw (v2si, long long)
6415 v2si __builtin_arm_wsrawi (v2si, int)
6416 long long __builtin_arm_wsrld (long long, long long)
6417 long long __builtin_arm_wsrldi (long long, int)
6418 v4hi __builtin_arm_wsrlh (v4hi, long long)
6419 v4hi __builtin_arm_wsrlhi (v4hi, int)
6420 v2si __builtin_arm_wsrlw (v2si, long long)
6421 v2si __builtin_arm_wsrlwi (v2si, int)
6422 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6423 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6424 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6425 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6426 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6427 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6428 v2si __builtin_arm_wsubw (v2si, v2si)
6429 v2si __builtin_arm_wsubwss (v2si, v2si)
6430 v2si __builtin_arm_wsubwus (v2si, v2si)
6431 v4hi __builtin_arm_wunpckehsb (v8qi)
6432 v2si __builtin_arm_wunpckehsh (v4hi)
6433 long long __builtin_arm_wunpckehsw (v2si)
6434 v4hi __builtin_arm_wunpckehub (v8qi)
6435 v2si __builtin_arm_wunpckehuh (v4hi)
6436 long long __builtin_arm_wunpckehuw (v2si)
6437 v4hi __builtin_arm_wunpckelsb (v8qi)
6438 v2si __builtin_arm_wunpckelsh (v4hi)
6439 long long __builtin_arm_wunpckelsw (v2si)
6440 v4hi __builtin_arm_wunpckelub (v8qi)
6441 v2si __builtin_arm_wunpckeluh (v4hi)
6442 long long __builtin_arm_wunpckeluw (v2si)
6443 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6444 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6445 v2si __builtin_arm_wunpckihw (v2si, v2si)
6446 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6447 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6448 v2si __builtin_arm_wunpckilw (v2si, v2si)
6449 long long __builtin_arm_wxor (long long, long long)
6450 long long __builtin_arm_wzero ()
6453 @node Blackfin Built-in Functions
6454 @subsection Blackfin Built-in Functions
6456 Currently, there are two Blackfin-specific built-in functions. These are
6457 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6458 using inline assembly; by using these built-in functions the compiler can
6459 automatically add workarounds for hardware errata involving these
6460 instructions. These functions are named as follows:
6463 void __builtin_bfin_csync (void)
6464 void __builtin_bfin_ssync (void)
6467 @node FR-V Built-in Functions
6468 @subsection FR-V Built-in Functions
6470 GCC provides many FR-V-specific built-in functions. In general,
6471 these functions are intended to be compatible with those described
6472 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6473 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6474 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6475 pointer rather than by value.
6477 Most of the functions are named after specific FR-V instructions.
6478 Such functions are said to be ``directly mapped'' and are summarized
6479 here in tabular form.
6483 * Directly-mapped Integer Functions::
6484 * Directly-mapped Media Functions::
6485 * Raw read/write Functions::
6486 * Other Built-in Functions::
6489 @node Argument Types
6490 @subsubsection Argument Types
6492 The arguments to the built-in functions can be divided into three groups:
6493 register numbers, compile-time constants and run-time values. In order
6494 to make this classification clear at a glance, the arguments and return
6495 values are given the following pseudo types:
6497 @multitable @columnfractions .20 .30 .15 .35
6498 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6499 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6500 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6501 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6502 @item @code{uw2} @tab @code{unsigned long long} @tab No
6503 @tab an unsigned doubleword
6504 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6505 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6506 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6507 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6510 These pseudo types are not defined by GCC, they are simply a notational
6511 convenience used in this manual.
6513 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6514 and @code{sw2} are evaluated at run time. They correspond to
6515 register operands in the underlying FR-V instructions.
6517 @code{const} arguments represent immediate operands in the underlying
6518 FR-V instructions. They must be compile-time constants.
6520 @code{acc} arguments are evaluated at compile time and specify the number
6521 of an accumulator register. For example, an @code{acc} argument of 2
6522 will select the ACC2 register.
6524 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6525 number of an IACC register. See @pxref{Other Built-in Functions}
6528 @node Directly-mapped Integer Functions
6529 @subsubsection Directly-mapped Integer Functions
6531 The functions listed below map directly to FR-V I-type instructions.
6533 @multitable @columnfractions .45 .32 .23
6534 @item Function prototype @tab Example usage @tab Assembly output
6535 @item @code{sw1 __ADDSS (sw1, sw1)}
6536 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6537 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6538 @item @code{sw1 __SCAN (sw1, sw1)}
6539 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6540 @tab @code{SCAN @var{a},@var{b},@var{c}}
6541 @item @code{sw1 __SCUTSS (sw1)}
6542 @tab @code{@var{b} = __SCUTSS (@var{a})}
6543 @tab @code{SCUTSS @var{a},@var{b}}
6544 @item @code{sw1 __SLASS (sw1, sw1)}
6545 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6546 @tab @code{SLASS @var{a},@var{b},@var{c}}
6547 @item @code{void __SMASS (sw1, sw1)}
6548 @tab @code{__SMASS (@var{a}, @var{b})}
6549 @tab @code{SMASS @var{a},@var{b}}
6550 @item @code{void __SMSSS (sw1, sw1)}
6551 @tab @code{__SMSSS (@var{a}, @var{b})}
6552 @tab @code{SMSSS @var{a},@var{b}}
6553 @item @code{void __SMU (sw1, sw1)}
6554 @tab @code{__SMU (@var{a}, @var{b})}
6555 @tab @code{SMU @var{a},@var{b}}
6556 @item @code{sw2 __SMUL (sw1, sw1)}
6557 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6558 @tab @code{SMUL @var{a},@var{b},@var{c}}
6559 @item @code{sw1 __SUBSS (sw1, sw1)}
6560 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6561 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6562 @item @code{uw2 __UMUL (uw1, uw1)}
6563 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6564 @tab @code{UMUL @var{a},@var{b},@var{c}}
6567 @node Directly-mapped Media Functions
6568 @subsubsection Directly-mapped Media Functions
6570 The functions listed below map directly to FR-V M-type instructions.
6572 @multitable @columnfractions .45 .32 .23
6573 @item Function prototype @tab Example usage @tab Assembly output
6574 @item @code{uw1 __MABSHS (sw1)}
6575 @tab @code{@var{b} = __MABSHS (@var{a})}
6576 @tab @code{MABSHS @var{a},@var{b}}
6577 @item @code{void __MADDACCS (acc, acc)}
6578 @tab @code{__MADDACCS (@var{b}, @var{a})}
6579 @tab @code{MADDACCS @var{a},@var{b}}
6580 @item @code{sw1 __MADDHSS (sw1, sw1)}
6581 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6582 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6583 @item @code{uw1 __MADDHUS (uw1, uw1)}
6584 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6585 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6586 @item @code{uw1 __MAND (uw1, uw1)}
6587 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6588 @tab @code{MAND @var{a},@var{b},@var{c}}
6589 @item @code{void __MASACCS (acc, acc)}
6590 @tab @code{__MASACCS (@var{b}, @var{a})}
6591 @tab @code{MASACCS @var{a},@var{b}}
6592 @item @code{uw1 __MAVEH (uw1, uw1)}
6593 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6594 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6595 @item @code{uw2 __MBTOH (uw1)}
6596 @tab @code{@var{b} = __MBTOH (@var{a})}
6597 @tab @code{MBTOH @var{a},@var{b}}
6598 @item @code{void __MBTOHE (uw1 *, uw1)}
6599 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6600 @tab @code{MBTOHE @var{a},@var{b}}
6601 @item @code{void __MCLRACC (acc)}
6602 @tab @code{__MCLRACC (@var{a})}
6603 @tab @code{MCLRACC @var{a}}
6604 @item @code{void __MCLRACCA (void)}
6605 @tab @code{__MCLRACCA ()}
6606 @tab @code{MCLRACCA}
6607 @item @code{uw1 __Mcop1 (uw1, uw1)}
6608 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6609 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6610 @item @code{uw1 __Mcop2 (uw1, uw1)}
6611 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6612 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6613 @item @code{uw1 __MCPLHI (uw2, const)}
6614 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6615 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6616 @item @code{uw1 __MCPLI (uw2, const)}
6617 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6618 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6619 @item @code{void __MCPXIS (acc, sw1, sw1)}
6620 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6621 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6622 @item @code{void __MCPXIU (acc, uw1, uw1)}
6623 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6624 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6625 @item @code{void __MCPXRS (acc, sw1, sw1)}
6626 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6627 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6628 @item @code{void __MCPXRU (acc, uw1, uw1)}
6629 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6630 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6631 @item @code{uw1 __MCUT (acc, uw1)}
6632 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6633 @tab @code{MCUT @var{a},@var{b},@var{c}}
6634 @item @code{uw1 __MCUTSS (acc, sw1)}
6635 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6636 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6637 @item @code{void __MDADDACCS (acc, acc)}
6638 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6639 @tab @code{MDADDACCS @var{a},@var{b}}
6640 @item @code{void __MDASACCS (acc, acc)}
6641 @tab @code{__MDASACCS (@var{b}, @var{a})}
6642 @tab @code{MDASACCS @var{a},@var{b}}
6643 @item @code{uw2 __MDCUTSSI (acc, const)}
6644 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6645 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6646 @item @code{uw2 __MDPACKH (uw2, uw2)}
6647 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6648 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6649 @item @code{uw2 __MDROTLI (uw2, const)}
6650 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6651 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6652 @item @code{void __MDSUBACCS (acc, acc)}
6653 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6654 @tab @code{MDSUBACCS @var{a},@var{b}}
6655 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6656 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6657 @tab @code{MDUNPACKH @var{a},@var{b}}
6658 @item @code{uw2 __MEXPDHD (uw1, const)}
6659 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6660 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6661 @item @code{uw1 __MEXPDHW (uw1, const)}
6662 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6663 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6664 @item @code{uw1 __MHDSETH (uw1, const)}
6665 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6666 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6667 @item @code{sw1 __MHDSETS (const)}
6668 @tab @code{@var{b} = __MHDSETS (@var{a})}
6669 @tab @code{MHDSETS #@var{a},@var{b}}
6670 @item @code{uw1 __MHSETHIH (uw1, const)}
6671 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6672 @tab @code{MHSETHIH #@var{a},@var{b}}
6673 @item @code{sw1 __MHSETHIS (sw1, const)}
6674 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6675 @tab @code{MHSETHIS #@var{a},@var{b}}
6676 @item @code{uw1 __MHSETLOH (uw1, const)}
6677 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6678 @tab @code{MHSETLOH #@var{a},@var{b}}
6679 @item @code{sw1 __MHSETLOS (sw1, const)}
6680 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6681 @tab @code{MHSETLOS #@var{a},@var{b}}
6682 @item @code{uw1 __MHTOB (uw2)}
6683 @tab @code{@var{b} = __MHTOB (@var{a})}
6684 @tab @code{MHTOB @var{a},@var{b}}
6685 @item @code{void __MMACHS (acc, sw1, sw1)}
6686 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6687 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6688 @item @code{void __MMACHU (acc, uw1, uw1)}
6689 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6690 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6691 @item @code{void __MMRDHS (acc, sw1, sw1)}
6692 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6693 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6694 @item @code{void __MMRDHU (acc, uw1, uw1)}
6695 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6696 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6697 @item @code{void __MMULHS (acc, sw1, sw1)}
6698 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6699 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6700 @item @code{void __MMULHU (acc, uw1, uw1)}
6701 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6702 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6703 @item @code{void __MMULXHS (acc, sw1, sw1)}
6704 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6705 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6706 @item @code{void __MMULXHU (acc, uw1, uw1)}
6707 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6708 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6709 @item @code{uw1 __MNOT (uw1)}
6710 @tab @code{@var{b} = __MNOT (@var{a})}
6711 @tab @code{MNOT @var{a},@var{b}}
6712 @item @code{uw1 __MOR (uw1, uw1)}
6713 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6714 @tab @code{MOR @var{a},@var{b},@var{c}}
6715 @item @code{uw1 __MPACKH (uh, uh)}
6716 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6717 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6718 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6719 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6720 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6721 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6722 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6723 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6724 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6725 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6726 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6727 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6728 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6729 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6730 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6731 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6732 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6733 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6734 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6735 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6736 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6737 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6738 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6739 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6740 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6741 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6742 @item @code{void __MQMACHS (acc, sw2, sw2)}
6743 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6744 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6745 @item @code{void __MQMACHU (acc, uw2, uw2)}
6746 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6747 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6748 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6749 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6750 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6751 @item @code{void __MQMULHS (acc, sw2, sw2)}
6752 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6753 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6754 @item @code{void __MQMULHU (acc, uw2, uw2)}
6755 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6756 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6757 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6758 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6759 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6760 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6761 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6762 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6763 @item @code{sw2 __MQSATHS (sw2, sw2)}
6764 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6765 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6766 @item @code{uw2 __MQSLLHI (uw2, int)}
6767 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6768 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6769 @item @code{sw2 __MQSRAHI (sw2, int)}
6770 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6771 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6772 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6773 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6774 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6775 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6776 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6777 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6778 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6779 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6780 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6781 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6782 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6783 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6784 @item @code{uw1 __MRDACC (acc)}
6785 @tab @code{@var{b} = __MRDACC (@var{a})}
6786 @tab @code{MRDACC @var{a},@var{b}}
6787 @item @code{uw1 __MRDACCG (acc)}
6788 @tab @code{@var{b} = __MRDACCG (@var{a})}
6789 @tab @code{MRDACCG @var{a},@var{b}}
6790 @item @code{uw1 __MROTLI (uw1, const)}
6791 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6792 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6793 @item @code{uw1 __MROTRI (uw1, const)}
6794 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6795 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6796 @item @code{sw1 __MSATHS (sw1, sw1)}
6797 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6798 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6799 @item @code{uw1 __MSATHU (uw1, uw1)}
6800 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6801 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6802 @item @code{uw1 __MSLLHI (uw1, const)}
6803 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6804 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6805 @item @code{sw1 __MSRAHI (sw1, const)}
6806 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6807 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6808 @item @code{uw1 __MSRLHI (uw1, const)}
6809 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6810 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6811 @item @code{void __MSUBACCS (acc, acc)}
6812 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6813 @tab @code{MSUBACCS @var{a},@var{b}}
6814 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6815 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6816 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6817 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6818 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6819 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6820 @item @code{void __MTRAP (void)}
6821 @tab @code{__MTRAP ()}
6823 @item @code{uw2 __MUNPACKH (uw1)}
6824 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6825 @tab @code{MUNPACKH @var{a},@var{b}}
6826 @item @code{uw1 __MWCUT (uw2, uw1)}
6827 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6828 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6829 @item @code{void __MWTACC (acc, uw1)}
6830 @tab @code{__MWTACC (@var{b}, @var{a})}
6831 @tab @code{MWTACC @var{a},@var{b}}
6832 @item @code{void __MWTACCG (acc, uw1)}
6833 @tab @code{__MWTACCG (@var{b}, @var{a})}
6834 @tab @code{MWTACCG @var{a},@var{b}}
6835 @item @code{uw1 __MXOR (uw1, uw1)}
6836 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6837 @tab @code{MXOR @var{a},@var{b},@var{c}}
6840 @node Raw read/write Functions
6841 @subsubsection Raw read/write Functions
6843 This sections describes built-in functions related to read and write
6844 instructions to access memory. These functions generate
6845 @code{membar} instructions to flush the I/O load and stores where
6846 appropriate, as described in Fujitsu's manual described above.
6850 @item unsigned char __builtin_read8 (void *@var{data})
6851 @item unsigned short __builtin_read16 (void *@var{data})
6852 @item unsigned long __builtin_read32 (void *@var{data})
6853 @item unsigned long long __builtin_read64 (void *@var{data})
6855 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6856 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6857 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6858 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6861 @node Other Built-in Functions
6862 @subsubsection Other Built-in Functions
6864 This section describes built-in functions that are not named after
6865 a specific FR-V instruction.
6868 @item sw2 __IACCreadll (iacc @var{reg})
6869 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6870 for future expansion and must be 0.
6872 @item sw1 __IACCreadl (iacc @var{reg})
6873 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6874 Other values of @var{reg} are rejected as invalid.
6876 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6877 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6878 is reserved for future expansion and must be 0.
6880 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6881 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6882 is 1. Other values of @var{reg} are rejected as invalid.
6884 @item void __data_prefetch0 (const void *@var{x})
6885 Use the @code{dcpl} instruction to load the contents of address @var{x}
6886 into the data cache.
6888 @item void __data_prefetch (const void *@var{x})
6889 Use the @code{nldub} instruction to load the contents of address @var{x}
6890 into the data cache. The instruction will be issued in slot I1@.
6893 @node X86 Built-in Functions
6894 @subsection X86 Built-in Functions
6896 These built-in functions are available for the i386 and x86-64 family
6897 of computers, depending on the command-line switches used.
6899 Note that, if you specify command-line switches such as @option{-msse},
6900 the compiler could use the extended instruction sets even if the built-ins
6901 are not used explicitly in the program. For this reason, applications
6902 which perform runtime CPU detection must compile separate files for each
6903 supported architecture, using the appropriate flags. In particular,
6904 the file containing the CPU detection code should be compiled without
6907 The following machine modes are available for use with MMX built-in functions
6908 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6909 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6910 vector of eight 8-bit integers. Some of the built-in functions operate on
6911 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6913 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6914 of two 32-bit floating point values.
6916 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6917 floating point values. Some instructions use a vector of four 32-bit
6918 integers, these use @code{V4SI}. Finally, some instructions operate on an
6919 entire vector register, interpreting it as a 128-bit integer, these use mode
6922 The following built-in functions are made available by @option{-mmmx}.
6923 All of them generate the machine instruction that is part of the name.
6926 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6927 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6928 v2si __builtin_ia32_paddd (v2si, v2si)
6929 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6930 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6931 v2si __builtin_ia32_psubd (v2si, v2si)
6932 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6933 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6934 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6935 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6936 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6937 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6938 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6939 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6940 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6941 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6942 di __builtin_ia32_pand (di, di)
6943 di __builtin_ia32_pandn (di,di)
6944 di __builtin_ia32_por (di, di)
6945 di __builtin_ia32_pxor (di, di)
6946 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6947 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6948 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6949 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6950 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6951 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6952 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6953 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6954 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6955 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6956 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6957 v2si __builtin_ia32_punpckldq (v2si, v2si)
6958 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6959 v4hi __builtin_ia32_packssdw (v2si, v2si)
6960 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6963 The following built-in functions are made available either with
6964 @option{-msse}, or with a combination of @option{-m3dnow} and
6965 @option{-march=athlon}. All of them generate the machine
6966 instruction that is part of the name.
6969 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6970 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6971 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6972 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6973 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6974 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6975 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6976 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6977 int __builtin_ia32_pextrw (v4hi, int)
6978 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6979 int __builtin_ia32_pmovmskb (v8qi)
6980 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6981 void __builtin_ia32_movntq (di *, di)
6982 void __builtin_ia32_sfence (void)
6985 The following built-in functions are available when @option{-msse} is used.
6986 All of them generate the machine instruction that is part of the name.
6989 int __builtin_ia32_comieq (v4sf, v4sf)
6990 int __builtin_ia32_comineq (v4sf, v4sf)
6991 int __builtin_ia32_comilt (v4sf, v4sf)
6992 int __builtin_ia32_comile (v4sf, v4sf)
6993 int __builtin_ia32_comigt (v4sf, v4sf)
6994 int __builtin_ia32_comige (v4sf, v4sf)
6995 int __builtin_ia32_ucomieq (v4sf, v4sf)
6996 int __builtin_ia32_ucomineq (v4sf, v4sf)
6997 int __builtin_ia32_ucomilt (v4sf, v4sf)
6998 int __builtin_ia32_ucomile (v4sf, v4sf)
6999 int __builtin_ia32_ucomigt (v4sf, v4sf)
7000 int __builtin_ia32_ucomige (v4sf, v4sf)
7001 v4sf __builtin_ia32_addps (v4sf, v4sf)
7002 v4sf __builtin_ia32_subps (v4sf, v4sf)
7003 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7004 v4sf __builtin_ia32_divps (v4sf, v4sf)
7005 v4sf __builtin_ia32_addss (v4sf, v4sf)
7006 v4sf __builtin_ia32_subss (v4sf, v4sf)
7007 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7008 v4sf __builtin_ia32_divss (v4sf, v4sf)
7009 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7010 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7011 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7012 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7013 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7014 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7015 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7016 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7017 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7018 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7019 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7020 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7021 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7022 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7023 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7024 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7025 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7026 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7027 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7028 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7029 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7030 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7031 v4sf __builtin_ia32_minps (v4sf, v4sf)
7032 v4sf __builtin_ia32_minss (v4sf, v4sf)
7033 v4sf __builtin_ia32_andps (v4sf, v4sf)
7034 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7035 v4sf __builtin_ia32_orps (v4sf, v4sf)
7036 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7037 v4sf __builtin_ia32_movss (v4sf, v4sf)
7038 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7039 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7040 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7041 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7042 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7043 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7044 v2si __builtin_ia32_cvtps2pi (v4sf)
7045 int __builtin_ia32_cvtss2si (v4sf)
7046 v2si __builtin_ia32_cvttps2pi (v4sf)
7047 int __builtin_ia32_cvttss2si (v4sf)
7048 v4sf __builtin_ia32_rcpps (v4sf)
7049 v4sf __builtin_ia32_rsqrtps (v4sf)
7050 v4sf __builtin_ia32_sqrtps (v4sf)
7051 v4sf __builtin_ia32_rcpss (v4sf)
7052 v4sf __builtin_ia32_rsqrtss (v4sf)
7053 v4sf __builtin_ia32_sqrtss (v4sf)
7054 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7055 void __builtin_ia32_movntps (float *, v4sf)
7056 int __builtin_ia32_movmskps (v4sf)
7059 The following built-in functions are available when @option{-msse} is used.
7062 @item v4sf __builtin_ia32_loadaps (float *)
7063 Generates the @code{movaps} machine instruction as a load from memory.
7064 @item void __builtin_ia32_storeaps (float *, v4sf)
7065 Generates the @code{movaps} machine instruction as a store to memory.
7066 @item v4sf __builtin_ia32_loadups (float *)
7067 Generates the @code{movups} machine instruction as a load from memory.
7068 @item void __builtin_ia32_storeups (float *, v4sf)
7069 Generates the @code{movups} machine instruction as a store to memory.
7070 @item v4sf __builtin_ia32_loadsss (float *)
7071 Generates the @code{movss} machine instruction as a load from memory.
7072 @item void __builtin_ia32_storess (float *, v4sf)
7073 Generates the @code{movss} machine instruction as a store to memory.
7074 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7075 Generates the @code{movhps} machine instruction as a load from memory.
7076 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7077 Generates the @code{movlps} machine instruction as a load from memory
7078 @item void __builtin_ia32_storehps (v4sf, v2si *)
7079 Generates the @code{movhps} machine instruction as a store to memory.
7080 @item void __builtin_ia32_storelps (v4sf, v2si *)
7081 Generates the @code{movlps} machine instruction as a store to memory.
7084 The following built-in functions are available when @option{-msse2} is used.
7085 All of them generate the machine instruction that is part of the name.
7088 int __builtin_ia32_comisdeq (v2df, v2df)
7089 int __builtin_ia32_comisdlt (v2df, v2df)
7090 int __builtin_ia32_comisdle (v2df, v2df)
7091 int __builtin_ia32_comisdgt (v2df, v2df)
7092 int __builtin_ia32_comisdge (v2df, v2df)
7093 int __builtin_ia32_comisdneq (v2df, v2df)
7094 int __builtin_ia32_ucomisdeq (v2df, v2df)
7095 int __builtin_ia32_ucomisdlt (v2df, v2df)
7096 int __builtin_ia32_ucomisdle (v2df, v2df)
7097 int __builtin_ia32_ucomisdgt (v2df, v2df)
7098 int __builtin_ia32_ucomisdge (v2df, v2df)
7099 int __builtin_ia32_ucomisdneq (v2df, v2df)
7100 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7101 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7102 v2df __builtin_ia32_cmplepd (v2df, v2df)
7103 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7104 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7105 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7106 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7107 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7108 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7109 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7110 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7111 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7112 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7113 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7114 v2df __builtin_ia32_cmplesd (v2df, v2df)
7115 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7116 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7117 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7118 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7119 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7120 v2di __builtin_ia32_paddq (v2di, v2di)
7121 v2di __builtin_ia32_psubq (v2di, v2di)
7122 v2df __builtin_ia32_addpd (v2df, v2df)
7123 v2df __builtin_ia32_subpd (v2df, v2df)
7124 v2df __builtin_ia32_mulpd (v2df, v2df)
7125 v2df __builtin_ia32_divpd (v2df, v2df)
7126 v2df __builtin_ia32_addsd (v2df, v2df)
7127 v2df __builtin_ia32_subsd (v2df, v2df)
7128 v2df __builtin_ia32_mulsd (v2df, v2df)
7129 v2df __builtin_ia32_divsd (v2df, v2df)
7130 v2df __builtin_ia32_minpd (v2df, v2df)
7131 v2df __builtin_ia32_maxpd (v2df, v2df)
7132 v2df __builtin_ia32_minsd (v2df, v2df)
7133 v2df __builtin_ia32_maxsd (v2df, v2df)
7134 v2df __builtin_ia32_andpd (v2df, v2df)
7135 v2df __builtin_ia32_andnpd (v2df, v2df)
7136 v2df __builtin_ia32_orpd (v2df, v2df)
7137 v2df __builtin_ia32_xorpd (v2df, v2df)
7138 v2df __builtin_ia32_movsd (v2df, v2df)
7139 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7140 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7141 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7142 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7143 v4si __builtin_ia32_paddd128 (v4si, v4si)
7144 v2di __builtin_ia32_paddq128 (v2di, v2di)
7145 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7146 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7147 v4si __builtin_ia32_psubd128 (v4si, v4si)
7148 v2di __builtin_ia32_psubq128 (v2di, v2di)
7149 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7150 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7151 v2di __builtin_ia32_pand128 (v2di, v2di)
7152 v2di __builtin_ia32_pandn128 (v2di, v2di)
7153 v2di __builtin_ia32_por128 (v2di, v2di)
7154 v2di __builtin_ia32_pxor128 (v2di, v2di)
7155 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7156 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7157 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7158 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7159 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7160 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7161 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7162 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7163 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7164 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7165 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7166 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7167 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7168 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7169 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7170 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7171 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7172 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7173 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7174 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7175 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7176 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7177 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7178 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7179 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7180 v2df __builtin_ia32_loadupd (double *)
7181 void __builtin_ia32_storeupd (double *, v2df)
7182 v2df __builtin_ia32_loadhpd (v2df, double *)
7183 v2df __builtin_ia32_loadlpd (v2df, double *)
7184 int __builtin_ia32_movmskpd (v2df)
7185 int __builtin_ia32_pmovmskb128 (v16qi)
7186 void __builtin_ia32_movnti (int *, int)
7187 void __builtin_ia32_movntpd (double *, v2df)
7188 void __builtin_ia32_movntdq (v2df *, v2df)
7189 v4si __builtin_ia32_pshufd (v4si, int)
7190 v8hi __builtin_ia32_pshuflw (v8hi, int)
7191 v8hi __builtin_ia32_pshufhw (v8hi, int)
7192 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7193 v2df __builtin_ia32_sqrtpd (v2df)
7194 v2df __builtin_ia32_sqrtsd (v2df)
7195 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7196 v2df __builtin_ia32_cvtdq2pd (v4si)
7197 v4sf __builtin_ia32_cvtdq2ps (v4si)
7198 v4si __builtin_ia32_cvtpd2dq (v2df)
7199 v2si __builtin_ia32_cvtpd2pi (v2df)
7200 v4sf __builtin_ia32_cvtpd2ps (v2df)
7201 v4si __builtin_ia32_cvttpd2dq (v2df)
7202 v2si __builtin_ia32_cvttpd2pi (v2df)
7203 v2df __builtin_ia32_cvtpi2pd (v2si)
7204 int __builtin_ia32_cvtsd2si (v2df)
7205 int __builtin_ia32_cvttsd2si (v2df)
7206 long long __builtin_ia32_cvtsd2si64 (v2df)
7207 long long __builtin_ia32_cvttsd2si64 (v2df)
7208 v4si __builtin_ia32_cvtps2dq (v4sf)
7209 v2df __builtin_ia32_cvtps2pd (v4sf)
7210 v4si __builtin_ia32_cvttps2dq (v4sf)
7211 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7212 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7213 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7214 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7215 void __builtin_ia32_clflush (const void *)
7216 void __builtin_ia32_lfence (void)
7217 void __builtin_ia32_mfence (void)
7218 v16qi __builtin_ia32_loaddqu (const char *)
7219 void __builtin_ia32_storedqu (char *, v16qi)
7220 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7221 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7222 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7223 v4si __builtin_ia32_pslld128 (v4si, v2di)
7224 v2di __builtin_ia32_psllq128 (v4si, v2di)
7225 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7226 v4si __builtin_ia32_psrld128 (v4si, v2di)
7227 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7228 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7229 v4si __builtin_ia32_psrad128 (v4si, v2di)
7230 v2di __builtin_ia32_pslldqi128 (v2di, int)
7231 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7232 v4si __builtin_ia32_pslldi128 (v4si, int)
7233 v2di __builtin_ia32_psllqi128 (v2di, int)
7234 v2di __builtin_ia32_psrldqi128 (v2di, int)
7235 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7236 v4si __builtin_ia32_psrldi128 (v4si, int)
7237 v2di __builtin_ia32_psrlqi128 (v2di, int)
7238 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7239 v4si __builtin_ia32_psradi128 (v4si, int)
7240 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7243 The following built-in functions are available when @option{-msse3} is used.
7244 All of them generate the machine instruction that is part of the name.
7247 v2df __builtin_ia32_addsubpd (v2df, v2df)
7248 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7249 v2df __builtin_ia32_haddpd (v2df, v2df)
7250 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7251 v2df __builtin_ia32_hsubpd (v2df, v2df)
7252 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7253 v16qi __builtin_ia32_lddqu (char const *)
7254 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7255 v2df __builtin_ia32_movddup (v2df)
7256 v4sf __builtin_ia32_movshdup (v4sf)
7257 v4sf __builtin_ia32_movsldup (v4sf)
7258 void __builtin_ia32_mwait (unsigned int, unsigned int)
7261 The following built-in functions are available when @option{-msse3} is used.
7264 @item v2df __builtin_ia32_loadddup (double const *)
7265 Generates the @code{movddup} machine instruction as a load from memory.
7268 The following built-in functions are available when @option{-mssse3} is used.
7269 All of them generate the machine instruction that is part of the name
7273 v2si __builtin_ia32_phaddd (v2si, v2si)
7274 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7275 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7276 v2si __builtin_ia32_phsubd (v2si, v2si)
7277 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7278 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7279 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7280 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7281 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7282 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7283 v2si __builtin_ia32_psignd (v2si, v2si)
7284 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7285 long long __builtin_ia32_palignr (long long, long long, int)
7286 v8qi __builtin_ia32_pabsb (v8qi)
7287 v2si __builtin_ia32_pabsd (v2si)
7288 v4hi __builtin_ia32_pabsw (v4hi)
7291 The following built-in functions are available when @option{-mssse3} is used.
7292 All of them generate the machine instruction that is part of the name
7296 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7297 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7298 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7299 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7300 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7301 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7302 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7303 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7304 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7305 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7306 v4si __builtin_ia32_psignd128 (v4si, v4si)
7307 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7308 v2di __builtin_ia32_palignr (v2di, v2di, int)
7309 v16qi __builtin_ia32_pabsb128 (v16qi)
7310 v4si __builtin_ia32_pabsd128 (v4si)
7311 v8hi __builtin_ia32_pabsw128 (v8hi)
7314 The following built-in functions are available when @option{-msse4a} is used.
7317 void _mm_stream_sd (double*,__m128d);
7318 Generates the @code{movntsd} machine instruction.
7319 void _mm_stream_ss (float*,__m128);
7320 Generates the @code{movntss} machine instruction.
7321 __m128i _mm_extract_si64 (__m128i, __m128i);
7322 Generates the @code{extrq} machine instruction with only SSE register operands.
7323 __m128i _mm_extracti_si64 (__m128i, int, int);
7324 Generates the @code{extrq} machine instruction with SSE register and immediate operands.
7325 __m128i _mm_insert_si64 (__m128i, __m128i);
7326 Generates the @code{insertq} machine instruction with only SSE register operands.
7327 __m128i _mm_inserti_si64 (__m128i, __m128i, int, int);
7328 Generates the @code{insertq} machine instruction with SSE register and immediate operands.
7331 The following built-in functions are available when @option{-m3dnow} is used.
7332 All of them generate the machine instruction that is part of the name.
7335 void __builtin_ia32_femms (void)
7336 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7337 v2si __builtin_ia32_pf2id (v2sf)
7338 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7339 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7340 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7341 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7342 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7343 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7344 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7345 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7346 v2sf __builtin_ia32_pfrcp (v2sf)
7347 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7348 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7349 v2sf __builtin_ia32_pfrsqrt (v2sf)
7350 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7351 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7352 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7353 v2sf __builtin_ia32_pi2fd (v2si)
7354 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7357 The following built-in functions are available when both @option{-m3dnow}
7358 and @option{-march=athlon} are used. All of them generate the machine
7359 instruction that is part of the name.
7362 v2si __builtin_ia32_pf2iw (v2sf)
7363 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7364 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7365 v2sf __builtin_ia32_pi2fw (v2si)
7366 v2sf __builtin_ia32_pswapdsf (v2sf)
7367 v2si __builtin_ia32_pswapdsi (v2si)
7370 @node MIPS DSP Built-in Functions
7371 @subsection MIPS DSP Built-in Functions
7373 The MIPS DSP Application-Specific Extension (ASE) includes new
7374 instructions that are designed to improve the performance of DSP and
7375 media applications. It provides instructions that operate on packed
7376 8-bit integer data, Q15 fractional data and Q31 fractional data.
7378 GCC supports MIPS DSP operations using both the generic
7379 vector extensions (@pxref{Vector Extensions}) and a collection of
7380 MIPS-specific built-in functions. Both kinds of support are
7381 enabled by the @option{-mdsp} command-line option.
7383 At present, GCC only provides support for operations on 32-bit
7384 vectors. The vector type associated with 8-bit integer data is
7385 usually called @code{v4i8} and the vector type associated with Q15 is
7386 usually called @code{v2q15}. They can be defined in C as follows:
7389 typedef char v4i8 __attribute__ ((vector_size(4)));
7390 typedef short v2q15 __attribute__ ((vector_size(4)));
7393 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7394 aggregates. For example:
7397 v4i8 a = @{1, 2, 3, 4@};
7399 b = (v4i8) @{5, 6, 7, 8@};
7401 v2q15 c = @{0x0fcb, 0x3a75@};
7403 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7406 @emph{Note:} The CPU's endianness determines the order in which values
7407 are packed. On little-endian targets, the first value is the least
7408 significant and the last value is the most significant. The opposite
7409 order applies to big-endian targets. For example, the code above will
7410 set the lowest byte of @code{a} to @code{1} on little-endian targets
7411 and @code{4} on big-endian targets.
7413 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7414 representation. As shown in this example, the integer representation
7415 of a Q15 value can be obtained by multiplying the fractional value by
7416 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7419 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7420 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7421 and @code{c} and @code{d} are @code{v2q15} values.
7423 @multitable @columnfractions .50 .50
7424 @item C code @tab MIPS instruction
7425 @item @code{a + b} @tab @code{addu.qb}
7426 @item @code{c + d} @tab @code{addq.ph}
7427 @item @code{a - b} @tab @code{subu.qb}
7428 @item @code{c - d} @tab @code{subq.ph}
7431 It is easier to describe the DSP built-in functions if we first define
7432 the following types:
7437 typedef long long a64;
7440 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7441 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7442 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7443 @code{long long}, but we use @code{a64} to indicate values that will
7444 be placed in one of the four DSP accumulators (@code{$ac0},
7445 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7447 Also, some built-in functions prefer or require immediate numbers as
7448 parameters, because the corresponding DSP instructions accept both immediate
7449 numbers and register operands, or accept immediate numbers only. The
7450 immediate parameters are listed as follows.
7458 imm_n32_31: -32 to 31.
7459 imm_n512_511: -512 to 511.
7462 The following built-in functions map directly to a particular MIPS DSP
7463 instruction. Please refer to the architecture specification
7464 for details on what each instruction does.
7467 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7468 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7469 q31 __builtin_mips_addq_s_w (q31, q31)
7470 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7471 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7472 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7473 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7474 q31 __builtin_mips_subq_s_w (q31, q31)
7475 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7476 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7477 i32 __builtin_mips_addsc (i32, i32)
7478 i32 __builtin_mips_addwc (i32, i32)
7479 i32 __builtin_mips_modsub (i32, i32)
7480 i32 __builtin_mips_raddu_w_qb (v4i8)
7481 v2q15 __builtin_mips_absq_s_ph (v2q15)
7482 q31 __builtin_mips_absq_s_w (q31)
7483 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7484 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7485 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7486 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7487 q31 __builtin_mips_preceq_w_phl (v2q15)
7488 q31 __builtin_mips_preceq_w_phr (v2q15)
7489 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7490 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7491 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7492 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7493 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7494 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7495 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7496 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7497 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7498 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7499 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7500 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7501 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7502 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7503 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7504 q31 __builtin_mips_shll_s_w (q31, i32)
7505 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7506 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7507 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7508 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7509 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7510 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7511 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7512 q31 __builtin_mips_shra_r_w (q31, i32)
7513 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7514 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7515 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7516 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7517 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7518 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7519 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7520 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7521 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7522 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7523 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7524 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7525 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7526 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7527 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7528 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7529 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7530 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7531 i32 __builtin_mips_bitrev (i32)
7532 i32 __builtin_mips_insv (i32, i32)
7533 v4i8 __builtin_mips_repl_qb (imm0_255)
7534 v4i8 __builtin_mips_repl_qb (i32)
7535 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7536 v2q15 __builtin_mips_repl_ph (i32)
7537 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7538 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7539 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7540 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7541 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7542 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7543 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7544 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7545 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7546 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7547 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7548 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7549 i32 __builtin_mips_extr_w (a64, imm0_31)
7550 i32 __builtin_mips_extr_w (a64, i32)
7551 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7552 i32 __builtin_mips_extr_s_h (a64, i32)
7553 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7554 i32 __builtin_mips_extr_rs_w (a64, i32)
7555 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7556 i32 __builtin_mips_extr_r_w (a64, i32)
7557 i32 __builtin_mips_extp (a64, imm0_31)
7558 i32 __builtin_mips_extp (a64, i32)
7559 i32 __builtin_mips_extpdp (a64, imm0_31)
7560 i32 __builtin_mips_extpdp (a64, i32)
7561 a64 __builtin_mips_shilo (a64, imm_n32_31)
7562 a64 __builtin_mips_shilo (a64, i32)
7563 a64 __builtin_mips_mthlip (a64, i32)
7564 void __builtin_mips_wrdsp (i32, imm0_63)
7565 i32 __builtin_mips_rddsp (imm0_63)
7566 i32 __builtin_mips_lbux (void *, i32)
7567 i32 __builtin_mips_lhx (void *, i32)
7568 i32 __builtin_mips_lwx (void *, i32)
7569 i32 __builtin_mips_bposge32 (void)
7572 @node MIPS Paired-Single Support
7573 @subsection MIPS Paired-Single Support
7575 The MIPS64 architecture includes a number of instructions that
7576 operate on pairs of single-precision floating-point values.
7577 Each pair is packed into a 64-bit floating-point register,
7578 with one element being designated the ``upper half'' and
7579 the other being designated the ``lower half''.
7581 GCC supports paired-single operations using both the generic
7582 vector extensions (@pxref{Vector Extensions}) and a collection of
7583 MIPS-specific built-in functions. Both kinds of support are
7584 enabled by the @option{-mpaired-single} command-line option.
7586 The vector type associated with paired-single values is usually
7587 called @code{v2sf}. It can be defined in C as follows:
7590 typedef float v2sf __attribute__ ((vector_size (8)));
7593 @code{v2sf} values are initialized in the same way as aggregates.
7597 v2sf a = @{1.5, 9.1@};
7600 b = (v2sf) @{e, f@};
7603 @emph{Note:} The CPU's endianness determines which value is stored in
7604 the upper half of a register and which value is stored in the lower half.
7605 On little-endian targets, the first value is the lower one and the second
7606 value is the upper one. The opposite order applies to big-endian targets.
7607 For example, the code above will set the lower half of @code{a} to
7608 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7611 * Paired-Single Arithmetic::
7612 * Paired-Single Built-in Functions::
7613 * MIPS-3D Built-in Functions::
7616 @node Paired-Single Arithmetic
7617 @subsubsection Paired-Single Arithmetic
7619 The table below lists the @code{v2sf} operations for which hardware
7620 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7621 values and @code{x} is an integral value.
7623 @multitable @columnfractions .50 .50
7624 @item C code @tab MIPS instruction
7625 @item @code{a + b} @tab @code{add.ps}
7626 @item @code{a - b} @tab @code{sub.ps}
7627 @item @code{-a} @tab @code{neg.ps}
7628 @item @code{a * b} @tab @code{mul.ps}
7629 @item @code{a * b + c} @tab @code{madd.ps}
7630 @item @code{a * b - c} @tab @code{msub.ps}
7631 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7632 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7633 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7636 Note that the multiply-accumulate instructions can be disabled
7637 using the command-line option @code{-mno-fused-madd}.
7639 @node Paired-Single Built-in Functions
7640 @subsubsection Paired-Single Built-in Functions
7642 The following paired-single functions map directly to a particular
7643 MIPS instruction. Please refer to the architecture specification
7644 for details on what each instruction does.
7647 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7648 Pair lower lower (@code{pll.ps}).
7650 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7651 Pair upper lower (@code{pul.ps}).
7653 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7654 Pair lower upper (@code{plu.ps}).
7656 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7657 Pair upper upper (@code{puu.ps}).
7659 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7660 Convert pair to paired single (@code{cvt.ps.s}).
7662 @item float __builtin_mips_cvt_s_pl (v2sf)
7663 Convert pair lower to single (@code{cvt.s.pl}).
7665 @item float __builtin_mips_cvt_s_pu (v2sf)
7666 Convert pair upper to single (@code{cvt.s.pu}).
7668 @item v2sf __builtin_mips_abs_ps (v2sf)
7669 Absolute value (@code{abs.ps}).
7671 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7672 Align variable (@code{alnv.ps}).
7674 @emph{Note:} The value of the third parameter must be 0 or 4
7675 modulo 8, otherwise the result will be unpredictable. Please read the
7676 instruction description for details.
7679 The following multi-instruction functions are also available.
7680 In each case, @var{cond} can be any of the 16 floating-point conditions:
7681 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7682 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7683 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7686 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7687 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7688 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7689 @code{movt.ps}/@code{movf.ps}).
7691 The @code{movt} functions return the value @var{x} computed by:
7694 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7695 mov.ps @var{x},@var{c}
7696 movt.ps @var{x},@var{d},@var{cc}
7699 The @code{movf} functions are similar but use @code{movf.ps} instead
7702 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7703 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7704 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7705 @code{bc1t}/@code{bc1f}).
7707 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7708 and return either the upper or lower half of the result. For example:
7712 if (__builtin_mips_upper_c_eq_ps (a, b))
7713 upper_halves_are_equal ();
7715 upper_halves_are_unequal ();
7717 if (__builtin_mips_lower_c_eq_ps (a, b))
7718 lower_halves_are_equal ();
7720 lower_halves_are_unequal ();
7724 @node MIPS-3D Built-in Functions
7725 @subsubsection MIPS-3D Built-in Functions
7727 The MIPS-3D Application-Specific Extension (ASE) includes additional
7728 paired-single instructions that are designed to improve the performance
7729 of 3D graphics operations. Support for these instructions is controlled
7730 by the @option{-mips3d} command-line option.
7732 The functions listed below map directly to a particular MIPS-3D
7733 instruction. Please refer to the architecture specification for
7734 more details on what each instruction does.
7737 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7738 Reduction add (@code{addr.ps}).
7740 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7741 Reduction multiply (@code{mulr.ps}).
7743 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7744 Convert paired single to paired word (@code{cvt.pw.ps}).
7746 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7747 Convert paired word to paired single (@code{cvt.ps.pw}).
7749 @item float __builtin_mips_recip1_s (float)
7750 @itemx double __builtin_mips_recip1_d (double)
7751 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7752 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7754 @item float __builtin_mips_recip2_s (float, float)
7755 @itemx double __builtin_mips_recip2_d (double, double)
7756 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7757 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7759 @item float __builtin_mips_rsqrt1_s (float)
7760 @itemx double __builtin_mips_rsqrt1_d (double)
7761 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7762 Reduced precision reciprocal square root (sequence step 1)
7763 (@code{rsqrt1.@var{fmt}}).
7765 @item float __builtin_mips_rsqrt2_s (float, float)
7766 @itemx double __builtin_mips_rsqrt2_d (double, double)
7767 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7768 Reduced precision reciprocal square root (sequence step 2)
7769 (@code{rsqrt2.@var{fmt}}).
7772 The following multi-instruction functions are also available.
7773 In each case, @var{cond} can be any of the 16 floating-point conditions:
7774 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7775 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7776 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7779 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7780 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7781 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7782 @code{bc1t}/@code{bc1f}).
7784 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7785 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7790 if (__builtin_mips_cabs_eq_s (a, b))
7796 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7797 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7798 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7799 @code{bc1t}/@code{bc1f}).
7801 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7802 and return either the upper or lower half of the result. For example:
7806 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7807 upper_halves_are_equal ();
7809 upper_halves_are_unequal ();
7811 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7812 lower_halves_are_equal ();
7814 lower_halves_are_unequal ();
7817 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7818 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7819 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7820 @code{movt.ps}/@code{movf.ps}).
7822 The @code{movt} functions return the value @var{x} computed by:
7825 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7826 mov.ps @var{x},@var{c}
7827 movt.ps @var{x},@var{d},@var{cc}
7830 The @code{movf} functions are similar but use @code{movf.ps} instead
7833 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7834 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7835 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7836 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7837 Comparison of two paired-single values
7838 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7839 @code{bc1any2t}/@code{bc1any2f}).
7841 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7842 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7843 result is true and the @code{all} forms return true if both results are true.
7848 if (__builtin_mips_any_c_eq_ps (a, b))
7853 if (__builtin_mips_all_c_eq_ps (a, b))
7859 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7860 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7861 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7862 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7863 Comparison of four paired-single values
7864 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7865 @code{bc1any4t}/@code{bc1any4f}).
7867 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7868 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7869 The @code{any} forms return true if any of the four results are true
7870 and the @code{all} forms return true if all four results are true.
7875 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7880 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7887 @node PowerPC AltiVec Built-in Functions
7888 @subsection PowerPC AltiVec Built-in Functions
7890 GCC provides an interface for the PowerPC family of processors to access
7891 the AltiVec operations described in Motorola's AltiVec Programming
7892 Interface Manual. The interface is made available by including
7893 @code{<altivec.h>} and using @option{-maltivec} and
7894 @option{-mabi=altivec}. The interface supports the following vector
7898 vector unsigned char
7902 vector unsigned short
7913 GCC's implementation of the high-level language interface available from
7914 C and C++ code differs from Motorola's documentation in several ways.
7919 A vector constant is a list of constant expressions within curly braces.
7922 A vector initializer requires no cast if the vector constant is of the
7923 same type as the variable it is initializing.
7926 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7927 vector type is the default signedness of the base type. The default
7928 varies depending on the operating system, so a portable program should
7929 always specify the signedness.
7932 Compiling with @option{-maltivec} adds keywords @code{__vector},
7933 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7934 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7938 GCC allows using a @code{typedef} name as the type specifier for a
7942 For C, overloaded functions are implemented with macros so the following
7946 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7949 Since @code{vec_add} is a macro, the vector constant in the example
7950 is treated as four separate arguments. Wrap the entire argument in
7951 parentheses for this to work.
7954 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
7955 Internally, GCC uses built-in functions to achieve the functionality in
7956 the aforementioned header file, but they are not supported and are
7957 subject to change without notice.
7959 The following interfaces are supported for the generic and specific
7960 AltiVec operations and the AltiVec predicates. In cases where there
7961 is a direct mapping between generic and specific operations, only the
7962 generic names are shown here, although the specific operations can also
7965 Arguments that are documented as @code{const int} require literal
7966 integral values within the range required for that operation.
7969 vector signed char vec_abs (vector signed char);
7970 vector signed short vec_abs (vector signed short);
7971 vector signed int vec_abs (vector signed int);
7972 vector float vec_abs (vector float);
7974 vector signed char vec_abss (vector signed char);
7975 vector signed short vec_abss (vector signed short);
7976 vector signed int vec_abss (vector signed int);
7978 vector signed char vec_add (vector bool char, vector signed char);
7979 vector signed char vec_add (vector signed char, vector bool char);
7980 vector signed char vec_add (vector signed char, vector signed char);
7981 vector unsigned char vec_add (vector bool char, vector unsigned char);
7982 vector unsigned char vec_add (vector unsigned char, vector bool char);
7983 vector unsigned char vec_add (vector unsigned char,
7984 vector unsigned char);
7985 vector signed short vec_add (vector bool short, vector signed short);
7986 vector signed short vec_add (vector signed short, vector bool short);
7987 vector signed short vec_add (vector signed short, vector signed short);
7988 vector unsigned short vec_add (vector bool short,
7989 vector unsigned short);
7990 vector unsigned short vec_add (vector unsigned short,
7992 vector unsigned short vec_add (vector unsigned short,
7993 vector unsigned short);
7994 vector signed int vec_add (vector bool int, vector signed int);
7995 vector signed int vec_add (vector signed int, vector bool int);
7996 vector signed int vec_add (vector signed int, vector signed int);
7997 vector unsigned int vec_add (vector bool int, vector unsigned int);
7998 vector unsigned int vec_add (vector unsigned int, vector bool int);
7999 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
8000 vector float vec_add (vector float, vector float);
8002 vector float vec_vaddfp (vector float, vector float);
8004 vector signed int vec_vadduwm (vector bool int, vector signed int);
8005 vector signed int vec_vadduwm (vector signed int, vector bool int);
8006 vector signed int vec_vadduwm (vector signed int, vector signed int);
8007 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
8008 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
8009 vector unsigned int vec_vadduwm (vector unsigned int,
8010 vector unsigned int);
8012 vector signed short vec_vadduhm (vector bool short,
8013 vector signed short);
8014 vector signed short vec_vadduhm (vector signed short,
8016 vector signed short vec_vadduhm (vector signed short,
8017 vector signed short);
8018 vector unsigned short vec_vadduhm (vector bool short,
8019 vector unsigned short);
8020 vector unsigned short vec_vadduhm (vector unsigned short,
8022 vector unsigned short vec_vadduhm (vector unsigned short,
8023 vector unsigned short);
8025 vector signed char vec_vaddubm (vector bool char, vector signed char);
8026 vector signed char vec_vaddubm (vector signed char, vector bool char);
8027 vector signed char vec_vaddubm (vector signed char, vector signed char);
8028 vector unsigned char vec_vaddubm (vector bool char,
8029 vector unsigned char);
8030 vector unsigned char vec_vaddubm (vector unsigned char,
8032 vector unsigned char vec_vaddubm (vector unsigned char,
8033 vector unsigned char);
8035 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
8037 vector unsigned char vec_adds (vector bool char, vector unsigned char);
8038 vector unsigned char vec_adds (vector unsigned char, vector bool char);
8039 vector unsigned char vec_adds (vector unsigned char,
8040 vector unsigned char);
8041 vector signed char vec_adds (vector bool char, vector signed char);
8042 vector signed char vec_adds (vector signed char, vector bool char);
8043 vector signed char vec_adds (vector signed char, vector signed char);
8044 vector unsigned short vec_adds (vector bool short,
8045 vector unsigned short);
8046 vector unsigned short vec_adds (vector unsigned short,
8048 vector unsigned short vec_adds (vector unsigned short,
8049 vector unsigned short);
8050 vector signed short vec_adds (vector bool short, vector signed short);
8051 vector signed short vec_adds (vector signed short, vector bool short);
8052 vector signed short vec_adds (vector signed short, vector signed short);
8053 vector unsigned int vec_adds (vector bool int, vector unsigned int);
8054 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8055 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8056 vector signed int vec_adds (vector bool int, vector signed int);
8057 vector signed int vec_adds (vector signed int, vector bool int);
8058 vector signed int vec_adds (vector signed int, vector signed int);
8060 vector signed int vec_vaddsws (vector bool int, vector signed int);
8061 vector signed int vec_vaddsws (vector signed int, vector bool int);
8062 vector signed int vec_vaddsws (vector signed int, vector signed int);
8064 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8065 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8066 vector unsigned int vec_vadduws (vector unsigned int,
8067 vector unsigned int);
8069 vector signed short vec_vaddshs (vector bool short,
8070 vector signed short);
8071 vector signed short vec_vaddshs (vector signed short,
8073 vector signed short vec_vaddshs (vector signed short,
8074 vector signed short);
8076 vector unsigned short vec_vadduhs (vector bool short,
8077 vector unsigned short);
8078 vector unsigned short vec_vadduhs (vector unsigned short,
8080 vector unsigned short vec_vadduhs (vector unsigned short,
8081 vector unsigned short);
8083 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8084 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8085 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8087 vector unsigned char vec_vaddubs (vector bool char,
8088 vector unsigned char);
8089 vector unsigned char vec_vaddubs (vector unsigned char,
8091 vector unsigned char vec_vaddubs (vector unsigned char,
8092 vector unsigned char);
8094 vector float vec_and (vector float, vector float);
8095 vector float vec_and (vector float, vector bool int);
8096 vector float vec_and (vector bool int, vector float);
8097 vector bool int vec_and (vector bool int, vector bool int);
8098 vector signed int vec_and (vector bool int, vector signed int);
8099 vector signed int vec_and (vector signed int, vector bool int);
8100 vector signed int vec_and (vector signed int, vector signed int);
8101 vector unsigned int vec_and (vector bool int, vector unsigned int);
8102 vector unsigned int vec_and (vector unsigned int, vector bool int);
8103 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8104 vector bool short vec_and (vector bool short, vector bool short);
8105 vector signed short vec_and (vector bool short, vector signed short);
8106 vector signed short vec_and (vector signed short, vector bool short);
8107 vector signed short vec_and (vector signed short, vector signed short);
8108 vector unsigned short vec_and (vector bool short,
8109 vector unsigned short);
8110 vector unsigned short vec_and (vector unsigned short,
8112 vector unsigned short vec_and (vector unsigned short,
8113 vector unsigned short);
8114 vector signed char vec_and (vector bool char, vector signed char);
8115 vector bool char vec_and (vector bool char, vector bool char);
8116 vector signed char vec_and (vector signed char, vector bool char);
8117 vector signed char vec_and (vector signed char, vector signed char);
8118 vector unsigned char vec_and (vector bool char, vector unsigned char);
8119 vector unsigned char vec_and (vector unsigned char, vector bool char);
8120 vector unsigned char vec_and (vector unsigned char,
8121 vector unsigned char);
8123 vector float vec_andc (vector float, vector float);
8124 vector float vec_andc (vector float, vector bool int);
8125 vector float vec_andc (vector bool int, vector float);
8126 vector bool int vec_andc (vector bool int, vector bool int);
8127 vector signed int vec_andc (vector bool int, vector signed int);
8128 vector signed int vec_andc (vector signed int, vector bool int);
8129 vector signed int vec_andc (vector signed int, vector signed int);
8130 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8131 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8132 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8133 vector bool short vec_andc (vector bool short, vector bool short);
8134 vector signed short vec_andc (vector bool short, vector signed short);
8135 vector signed short vec_andc (vector signed short, vector bool short);
8136 vector signed short vec_andc (vector signed short, vector signed short);
8137 vector unsigned short vec_andc (vector bool short,
8138 vector unsigned short);
8139 vector unsigned short vec_andc (vector unsigned short,
8141 vector unsigned short vec_andc (vector unsigned short,
8142 vector unsigned short);
8143 vector signed char vec_andc (vector bool char, vector signed char);
8144 vector bool char vec_andc (vector bool char, vector bool char);
8145 vector signed char vec_andc (vector signed char, vector bool char);
8146 vector signed char vec_andc (vector signed char, vector signed char);
8147 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8148 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8149 vector unsigned char vec_andc (vector unsigned char,
8150 vector unsigned char);
8152 vector unsigned char vec_avg (vector unsigned char,
8153 vector unsigned char);
8154 vector signed char vec_avg (vector signed char, vector signed char);
8155 vector unsigned short vec_avg (vector unsigned short,
8156 vector unsigned short);
8157 vector signed short vec_avg (vector signed short, vector signed short);
8158 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8159 vector signed int vec_avg (vector signed int, vector signed int);
8161 vector signed int vec_vavgsw (vector signed int, vector signed int);
8163 vector unsigned int vec_vavguw (vector unsigned int,
8164 vector unsigned int);
8166 vector signed short vec_vavgsh (vector signed short,
8167 vector signed short);
8169 vector unsigned short vec_vavguh (vector unsigned short,
8170 vector unsigned short);
8172 vector signed char vec_vavgsb (vector signed char, vector signed char);
8174 vector unsigned char vec_vavgub (vector unsigned char,
8175 vector unsigned char);
8177 vector float vec_ceil (vector float);
8179 vector signed int vec_cmpb (vector float, vector float);
8181 vector bool char vec_cmpeq (vector signed char, vector signed char);
8182 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8183 vector bool short vec_cmpeq (vector signed short, vector signed short);
8184 vector bool short vec_cmpeq (vector unsigned short,
8185 vector unsigned short);
8186 vector bool int vec_cmpeq (vector signed int, vector signed int);
8187 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8188 vector bool int vec_cmpeq (vector float, vector float);
8190 vector bool int vec_vcmpeqfp (vector float, vector float);
8192 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8193 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8195 vector bool short vec_vcmpequh (vector signed short,
8196 vector signed short);
8197 vector bool short vec_vcmpequh (vector unsigned short,
8198 vector unsigned short);
8200 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8201 vector bool char vec_vcmpequb (vector unsigned char,
8202 vector unsigned char);
8204 vector bool int vec_cmpge (vector float, vector float);
8206 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8207 vector bool char vec_cmpgt (vector signed char, vector signed char);
8208 vector bool short vec_cmpgt (vector unsigned short,
8209 vector unsigned short);
8210 vector bool short vec_cmpgt (vector signed short, vector signed short);
8211 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8212 vector bool int vec_cmpgt (vector signed int, vector signed int);
8213 vector bool int vec_cmpgt (vector float, vector float);
8215 vector bool int vec_vcmpgtfp (vector float, vector float);
8217 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8219 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8221 vector bool short vec_vcmpgtsh (vector signed short,
8222 vector signed short);
8224 vector bool short vec_vcmpgtuh (vector unsigned short,
8225 vector unsigned short);
8227 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8229 vector bool char vec_vcmpgtub (vector unsigned char,
8230 vector unsigned char);
8232 vector bool int vec_cmple (vector float, vector float);
8234 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8235 vector bool char vec_cmplt (vector signed char, vector signed char);
8236 vector bool short vec_cmplt (vector unsigned short,
8237 vector unsigned short);
8238 vector bool short vec_cmplt (vector signed short, vector signed short);
8239 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8240 vector bool int vec_cmplt (vector signed int, vector signed int);
8241 vector bool int vec_cmplt (vector float, vector float);
8243 vector float vec_ctf (vector unsigned int, const int);
8244 vector float vec_ctf (vector signed int, const int);
8246 vector float vec_vcfsx (vector signed int, const int);
8248 vector float vec_vcfux (vector unsigned int, const int);
8250 vector signed int vec_cts (vector float, const int);
8252 vector unsigned int vec_ctu (vector float, const int);
8254 void vec_dss (const int);
8256 void vec_dssall (void);
8258 void vec_dst (const vector unsigned char *, int, const int);
8259 void vec_dst (const vector signed char *, int, const int);
8260 void vec_dst (const vector bool char *, int, const int);
8261 void vec_dst (const vector unsigned short *, int, const int);
8262 void vec_dst (const vector signed short *, int, const int);
8263 void vec_dst (const vector bool short *, int, const int);
8264 void vec_dst (const vector pixel *, int, const int);
8265 void vec_dst (const vector unsigned int *, int, const int);
8266 void vec_dst (const vector signed int *, int, const int);
8267 void vec_dst (const vector bool int *, int, const int);
8268 void vec_dst (const vector float *, int, const int);
8269 void vec_dst (const unsigned char *, int, const int);
8270 void vec_dst (const signed char *, int, const int);
8271 void vec_dst (const unsigned short *, int, const int);
8272 void vec_dst (const short *, int, const int);
8273 void vec_dst (const unsigned int *, int, const int);
8274 void vec_dst (const int *, int, const int);
8275 void vec_dst (const unsigned long *, int, const int);
8276 void vec_dst (const long *, int, const int);
8277 void vec_dst (const float *, int, const int);
8279 void vec_dstst (const vector unsigned char *, int, const int);
8280 void vec_dstst (const vector signed char *, int, const int);
8281 void vec_dstst (const vector bool char *, int, const int);
8282 void vec_dstst (const vector unsigned short *, int, const int);
8283 void vec_dstst (const vector signed short *, int, const int);
8284 void vec_dstst (const vector bool short *, int, const int);
8285 void vec_dstst (const vector pixel *, int, const int);
8286 void vec_dstst (const vector unsigned int *, int, const int);
8287 void vec_dstst (const vector signed int *, int, const int);
8288 void vec_dstst (const vector bool int *, int, const int);
8289 void vec_dstst (const vector float *, int, const int);
8290 void vec_dstst (const unsigned char *, int, const int);
8291 void vec_dstst (const signed char *, int, const int);
8292 void vec_dstst (const unsigned short *, int, const int);
8293 void vec_dstst (const short *, int, const int);
8294 void vec_dstst (const unsigned int *, int, const int);
8295 void vec_dstst (const int *, int, const int);
8296 void vec_dstst (const unsigned long *, int, const int);
8297 void vec_dstst (const long *, int, const int);
8298 void vec_dstst (const float *, int, const int);
8300 void vec_dststt (const vector unsigned char *, int, const int);
8301 void vec_dststt (const vector signed char *, int, const int);
8302 void vec_dststt (const vector bool char *, int, const int);
8303 void vec_dststt (const vector unsigned short *, int, const int);
8304 void vec_dststt (const vector signed short *, int, const int);
8305 void vec_dststt (const vector bool short *, int, const int);
8306 void vec_dststt (const vector pixel *, int, const int);
8307 void vec_dststt (const vector unsigned int *, int, const int);
8308 void vec_dststt (const vector signed int *, int, const int);
8309 void vec_dststt (const vector bool int *, int, const int);
8310 void vec_dststt (const vector float *, int, const int);
8311 void vec_dststt (const unsigned char *, int, const int);
8312 void vec_dststt (const signed char *, int, const int);
8313 void vec_dststt (const unsigned short *, int, const int);
8314 void vec_dststt (const short *, int, const int);
8315 void vec_dststt (const unsigned int *, int, const int);
8316 void vec_dststt (const int *, int, const int);
8317 void vec_dststt (const unsigned long *, int, const int);
8318 void vec_dststt (const long *, int, const int);
8319 void vec_dststt (const float *, int, const int);
8321 void vec_dstt (const vector unsigned char *, int, const int);
8322 void vec_dstt (const vector signed char *, int, const int);
8323 void vec_dstt (const vector bool char *, int, const int);
8324 void vec_dstt (const vector unsigned short *, int, const int);
8325 void vec_dstt (const vector signed short *, int, const int);
8326 void vec_dstt (const vector bool short *, int, const int);
8327 void vec_dstt (const vector pixel *, int, const int);
8328 void vec_dstt (const vector unsigned int *, int, const int);
8329 void vec_dstt (const vector signed int *, int, const int);
8330 void vec_dstt (const vector bool int *, int, const int);
8331 void vec_dstt (const vector float *, int, const int);
8332 void vec_dstt (const unsigned char *, int, const int);
8333 void vec_dstt (const signed char *, int, const int);
8334 void vec_dstt (const unsigned short *, int, const int);
8335 void vec_dstt (const short *, int, const int);
8336 void vec_dstt (const unsigned int *, int, const int);
8337 void vec_dstt (const int *, int, const int);
8338 void vec_dstt (const unsigned long *, int, const int);
8339 void vec_dstt (const long *, int, const int);
8340 void vec_dstt (const float *, int, const int);
8342 vector float vec_expte (vector float);
8344 vector float vec_floor (vector float);
8346 vector float vec_ld (int, const vector float *);
8347 vector float vec_ld (int, const float *);
8348 vector bool int vec_ld (int, const vector bool int *);
8349 vector signed int vec_ld (int, const vector signed int *);
8350 vector signed int vec_ld (int, const int *);
8351 vector signed int vec_ld (int, const long *);
8352 vector unsigned int vec_ld (int, const vector unsigned int *);
8353 vector unsigned int vec_ld (int, const unsigned int *);
8354 vector unsigned int vec_ld (int, const unsigned long *);
8355 vector bool short vec_ld (int, const vector bool short *);
8356 vector pixel vec_ld (int, const vector pixel *);
8357 vector signed short vec_ld (int, const vector signed short *);
8358 vector signed short vec_ld (int, const short *);
8359 vector unsigned short vec_ld (int, const vector unsigned short *);
8360 vector unsigned short vec_ld (int, const unsigned short *);
8361 vector bool char vec_ld (int, const vector bool char *);
8362 vector signed char vec_ld (int, const vector signed char *);
8363 vector signed char vec_ld (int, const signed char *);
8364 vector unsigned char vec_ld (int, const vector unsigned char *);
8365 vector unsigned char vec_ld (int, const unsigned char *);
8367 vector signed char vec_lde (int, const signed char *);
8368 vector unsigned char vec_lde (int, const unsigned char *);
8369 vector signed short vec_lde (int, const short *);
8370 vector unsigned short vec_lde (int, const unsigned short *);
8371 vector float vec_lde (int, const float *);
8372 vector signed int vec_lde (int, const int *);
8373 vector unsigned int vec_lde (int, const unsigned int *);
8374 vector signed int vec_lde (int, const long *);
8375 vector unsigned int vec_lde (int, const unsigned long *);
8377 vector float vec_lvewx (int, float *);
8378 vector signed int vec_lvewx (int, int *);
8379 vector unsigned int vec_lvewx (int, unsigned int *);
8380 vector signed int vec_lvewx (int, long *);
8381 vector unsigned int vec_lvewx (int, unsigned long *);
8383 vector signed short vec_lvehx (int, short *);
8384 vector unsigned short vec_lvehx (int, unsigned short *);
8386 vector signed char vec_lvebx (int, char *);
8387 vector unsigned char vec_lvebx (int, unsigned char *);
8389 vector float vec_ldl (int, const vector float *);
8390 vector float vec_ldl (int, const float *);
8391 vector bool int vec_ldl (int, const vector bool int *);
8392 vector signed int vec_ldl (int, const vector signed int *);
8393 vector signed int vec_ldl (int, const int *);
8394 vector signed int vec_ldl (int, const long *);
8395 vector unsigned int vec_ldl (int, const vector unsigned int *);
8396 vector unsigned int vec_ldl (int, const unsigned int *);
8397 vector unsigned int vec_ldl (int, const unsigned long *);
8398 vector bool short vec_ldl (int, const vector bool short *);
8399 vector pixel vec_ldl (int, const vector pixel *);
8400 vector signed short vec_ldl (int, const vector signed short *);
8401 vector signed short vec_ldl (int, const short *);
8402 vector unsigned short vec_ldl (int, const vector unsigned short *);
8403 vector unsigned short vec_ldl (int, const unsigned short *);
8404 vector bool char vec_ldl (int, const vector bool char *);
8405 vector signed char vec_ldl (int, const vector signed char *);
8406 vector signed char vec_ldl (int, const signed char *);
8407 vector unsigned char vec_ldl (int, const vector unsigned char *);
8408 vector unsigned char vec_ldl (int, const unsigned char *);
8410 vector float vec_loge (vector float);
8412 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8413 vector unsigned char vec_lvsl (int, const volatile signed char *);
8414 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8415 vector unsigned char vec_lvsl (int, const volatile short *);
8416 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8417 vector unsigned char vec_lvsl (int, const volatile int *);
8418 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8419 vector unsigned char vec_lvsl (int, const volatile long *);
8420 vector unsigned char vec_lvsl (int, const volatile float *);
8422 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8423 vector unsigned char vec_lvsr (int, const volatile signed char *);
8424 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8425 vector unsigned char vec_lvsr (int, const volatile short *);
8426 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8427 vector unsigned char vec_lvsr (int, const volatile int *);
8428 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8429 vector unsigned char vec_lvsr (int, const volatile long *);
8430 vector unsigned char vec_lvsr (int, const volatile float *);
8432 vector float vec_madd (vector float, vector float, vector float);
8434 vector signed short vec_madds (vector signed short,
8435 vector signed short,
8436 vector signed short);
8438 vector unsigned char vec_max (vector bool char, vector unsigned char);
8439 vector unsigned char vec_max (vector unsigned char, vector bool char);
8440 vector unsigned char vec_max (vector unsigned char,
8441 vector unsigned char);
8442 vector signed char vec_max (vector bool char, vector signed char);
8443 vector signed char vec_max (vector signed char, vector bool char);
8444 vector signed char vec_max (vector signed char, vector signed char);
8445 vector unsigned short vec_max (vector bool short,
8446 vector unsigned short);
8447 vector unsigned short vec_max (vector unsigned short,
8449 vector unsigned short vec_max (vector unsigned short,
8450 vector unsigned short);
8451 vector signed short vec_max (vector bool short, vector signed short);
8452 vector signed short vec_max (vector signed short, vector bool short);
8453 vector signed short vec_max (vector signed short, vector signed short);
8454 vector unsigned int vec_max (vector bool int, vector unsigned int);
8455 vector unsigned int vec_max (vector unsigned int, vector bool int);
8456 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8457 vector signed int vec_max (vector bool int, vector signed int);
8458 vector signed int vec_max (vector signed int, vector bool int);
8459 vector signed int vec_max (vector signed int, vector signed int);
8460 vector float vec_max (vector float, vector float);
8462 vector float vec_vmaxfp (vector float, vector float);
8464 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8465 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8466 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8468 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8469 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8470 vector unsigned int vec_vmaxuw (vector unsigned int,
8471 vector unsigned int);
8473 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8474 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8475 vector signed short vec_vmaxsh (vector signed short,
8476 vector signed short);
8478 vector unsigned short vec_vmaxuh (vector bool short,
8479 vector unsigned short);
8480 vector unsigned short vec_vmaxuh (vector unsigned short,
8482 vector unsigned short vec_vmaxuh (vector unsigned short,
8483 vector unsigned short);
8485 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8486 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8487 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8489 vector unsigned char vec_vmaxub (vector bool char,
8490 vector unsigned char);
8491 vector unsigned char vec_vmaxub (vector unsigned char,
8493 vector unsigned char vec_vmaxub (vector unsigned char,
8494 vector unsigned char);
8496 vector bool char vec_mergeh (vector bool char, vector bool char);
8497 vector signed char vec_mergeh (vector signed char, vector signed char);
8498 vector unsigned char vec_mergeh (vector unsigned char,
8499 vector unsigned char);
8500 vector bool short vec_mergeh (vector bool short, vector bool short);
8501 vector pixel vec_mergeh (vector pixel, vector pixel);
8502 vector signed short vec_mergeh (vector signed short,
8503 vector signed short);
8504 vector unsigned short vec_mergeh (vector unsigned short,
8505 vector unsigned short);
8506 vector float vec_mergeh (vector float, vector float);
8507 vector bool int vec_mergeh (vector bool int, vector bool int);
8508 vector signed int vec_mergeh (vector signed int, vector signed int);
8509 vector unsigned int vec_mergeh (vector unsigned int,
8510 vector unsigned int);
8512 vector float vec_vmrghw (vector float, vector float);
8513 vector bool int vec_vmrghw (vector bool int, vector bool int);
8514 vector signed int vec_vmrghw (vector signed int, vector signed int);
8515 vector unsigned int vec_vmrghw (vector unsigned int,
8516 vector unsigned int);
8518 vector bool short vec_vmrghh (vector bool short, vector bool short);
8519 vector signed short vec_vmrghh (vector signed short,
8520 vector signed short);
8521 vector unsigned short vec_vmrghh (vector unsigned short,
8522 vector unsigned short);
8523 vector pixel vec_vmrghh (vector pixel, vector pixel);
8525 vector bool char vec_vmrghb (vector bool char, vector bool char);
8526 vector signed char vec_vmrghb (vector signed char, vector signed char);
8527 vector unsigned char vec_vmrghb (vector unsigned char,
8528 vector unsigned char);
8530 vector bool char vec_mergel (vector bool char, vector bool char);
8531 vector signed char vec_mergel (vector signed char, vector signed char);
8532 vector unsigned char vec_mergel (vector unsigned char,
8533 vector unsigned char);
8534 vector bool short vec_mergel (vector bool short, vector bool short);
8535 vector pixel vec_mergel (vector pixel, vector pixel);
8536 vector signed short vec_mergel (vector signed short,
8537 vector signed short);
8538 vector unsigned short vec_mergel (vector unsigned short,
8539 vector unsigned short);
8540 vector float vec_mergel (vector float, vector float);
8541 vector bool int vec_mergel (vector bool int, vector bool int);
8542 vector signed int vec_mergel (vector signed int, vector signed int);
8543 vector unsigned int vec_mergel (vector unsigned int,
8544 vector unsigned int);
8546 vector float vec_vmrglw (vector float, vector float);
8547 vector signed int vec_vmrglw (vector signed int, vector signed int);
8548 vector unsigned int vec_vmrglw (vector unsigned int,
8549 vector unsigned int);
8550 vector bool int vec_vmrglw (vector bool int, vector bool int);
8552 vector bool short vec_vmrglh (vector bool short, vector bool short);
8553 vector signed short vec_vmrglh (vector signed short,
8554 vector signed short);
8555 vector unsigned short vec_vmrglh (vector unsigned short,
8556 vector unsigned short);
8557 vector pixel vec_vmrglh (vector pixel, vector pixel);
8559 vector bool char vec_vmrglb (vector bool char, vector bool char);
8560 vector signed char vec_vmrglb (vector signed char, vector signed char);
8561 vector unsigned char vec_vmrglb (vector unsigned char,
8562 vector unsigned char);
8564 vector unsigned short vec_mfvscr (void);
8566 vector unsigned char vec_min (vector bool char, vector unsigned char);
8567 vector unsigned char vec_min (vector unsigned char, vector bool char);
8568 vector unsigned char vec_min (vector unsigned char,
8569 vector unsigned char);
8570 vector signed char vec_min (vector bool char, vector signed char);
8571 vector signed char vec_min (vector signed char, vector bool char);
8572 vector signed char vec_min (vector signed char, vector signed char);
8573 vector unsigned short vec_min (vector bool short,
8574 vector unsigned short);
8575 vector unsigned short vec_min (vector unsigned short,
8577 vector unsigned short vec_min (vector unsigned short,
8578 vector unsigned short);
8579 vector signed short vec_min (vector bool short, vector signed short);
8580 vector signed short vec_min (vector signed short, vector bool short);
8581 vector signed short vec_min (vector signed short, vector signed short);
8582 vector unsigned int vec_min (vector bool int, vector unsigned int);
8583 vector unsigned int vec_min (vector unsigned int, vector bool int);
8584 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8585 vector signed int vec_min (vector bool int, vector signed int);
8586 vector signed int vec_min (vector signed int, vector bool int);
8587 vector signed int vec_min (vector signed int, vector signed int);
8588 vector float vec_min (vector float, vector float);
8590 vector float vec_vminfp (vector float, vector float);
8592 vector signed int vec_vminsw (vector bool int, vector signed int);
8593 vector signed int vec_vminsw (vector signed int, vector bool int);
8594 vector signed int vec_vminsw (vector signed int, vector signed int);
8596 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8597 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8598 vector unsigned int vec_vminuw (vector unsigned int,
8599 vector unsigned int);
8601 vector signed short vec_vminsh (vector bool short, vector signed short);
8602 vector signed short vec_vminsh (vector signed short, vector bool short);
8603 vector signed short vec_vminsh (vector signed short,
8604 vector signed short);
8606 vector unsigned short vec_vminuh (vector bool short,
8607 vector unsigned short);
8608 vector unsigned short vec_vminuh (vector unsigned short,
8610 vector unsigned short vec_vminuh (vector unsigned short,
8611 vector unsigned short);
8613 vector signed char vec_vminsb (vector bool char, vector signed char);
8614 vector signed char vec_vminsb (vector signed char, vector bool char);
8615 vector signed char vec_vminsb (vector signed char, vector signed char);
8617 vector unsigned char vec_vminub (vector bool char,
8618 vector unsigned char);
8619 vector unsigned char vec_vminub (vector unsigned char,
8621 vector unsigned char vec_vminub (vector unsigned char,
8622 vector unsigned char);
8624 vector signed short vec_mladd (vector signed short,
8625 vector signed short,
8626 vector signed short);
8627 vector signed short vec_mladd (vector signed short,
8628 vector unsigned short,
8629 vector unsigned short);
8630 vector signed short vec_mladd (vector unsigned short,
8631 vector signed short,
8632 vector signed short);
8633 vector unsigned short vec_mladd (vector unsigned short,
8634 vector unsigned short,
8635 vector unsigned short);
8637 vector signed short vec_mradds (vector signed short,
8638 vector signed short,
8639 vector signed short);
8641 vector unsigned int vec_msum (vector unsigned char,
8642 vector unsigned char,
8643 vector unsigned int);
8644 vector signed int vec_msum (vector signed char,
8645 vector unsigned char,
8647 vector unsigned int vec_msum (vector unsigned short,
8648 vector unsigned short,
8649 vector unsigned int);
8650 vector signed int vec_msum (vector signed short,
8651 vector signed short,
8654 vector signed int vec_vmsumshm (vector signed short,
8655 vector signed short,
8658 vector unsigned int vec_vmsumuhm (vector unsigned short,
8659 vector unsigned short,
8660 vector unsigned int);
8662 vector signed int vec_vmsummbm (vector signed char,
8663 vector unsigned char,
8666 vector unsigned int vec_vmsumubm (vector unsigned char,
8667 vector unsigned char,
8668 vector unsigned int);
8670 vector unsigned int vec_msums (vector unsigned short,
8671 vector unsigned short,
8672 vector unsigned int);
8673 vector signed int vec_msums (vector signed short,
8674 vector signed short,
8677 vector signed int vec_vmsumshs (vector signed short,
8678 vector signed short,
8681 vector unsigned int vec_vmsumuhs (vector unsigned short,
8682 vector unsigned short,
8683 vector unsigned int);
8685 void vec_mtvscr (vector signed int);
8686 void vec_mtvscr (vector unsigned int);
8687 void vec_mtvscr (vector bool int);
8688 void vec_mtvscr (vector signed short);
8689 void vec_mtvscr (vector unsigned short);
8690 void vec_mtvscr (vector bool short);
8691 void vec_mtvscr (vector pixel);
8692 void vec_mtvscr (vector signed char);
8693 void vec_mtvscr (vector unsigned char);
8694 void vec_mtvscr (vector bool char);
8696 vector unsigned short vec_mule (vector unsigned char,
8697 vector unsigned char);
8698 vector signed short vec_mule (vector signed char,
8699 vector signed char);
8700 vector unsigned int vec_mule (vector unsigned short,
8701 vector unsigned short);
8702 vector signed int vec_mule (vector signed short, vector signed short);
8704 vector signed int vec_vmulesh (vector signed short,
8705 vector signed short);
8707 vector unsigned int vec_vmuleuh (vector unsigned short,
8708 vector unsigned short);
8710 vector signed short vec_vmulesb (vector signed char,
8711 vector signed char);
8713 vector unsigned short vec_vmuleub (vector unsigned char,
8714 vector unsigned char);
8716 vector unsigned short vec_mulo (vector unsigned char,
8717 vector unsigned char);
8718 vector signed short vec_mulo (vector signed char, vector signed char);
8719 vector unsigned int vec_mulo (vector unsigned short,
8720 vector unsigned short);
8721 vector signed int vec_mulo (vector signed short, vector signed short);
8723 vector signed int vec_vmulosh (vector signed short,
8724 vector signed short);
8726 vector unsigned int vec_vmulouh (vector unsigned short,
8727 vector unsigned short);
8729 vector signed short vec_vmulosb (vector signed char,
8730 vector signed char);
8732 vector unsigned short vec_vmuloub (vector unsigned char,
8733 vector unsigned char);
8735 vector float vec_nmsub (vector float, vector float, vector float);
8737 vector float vec_nor (vector float, vector float);
8738 vector signed int vec_nor (vector signed int, vector signed int);
8739 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8740 vector bool int vec_nor (vector bool int, vector bool int);
8741 vector signed short vec_nor (vector signed short, vector signed short);
8742 vector unsigned short vec_nor (vector unsigned short,
8743 vector unsigned short);
8744 vector bool short vec_nor (vector bool short, vector bool short);
8745 vector signed char vec_nor (vector signed char, vector signed char);
8746 vector unsigned char vec_nor (vector unsigned char,
8747 vector unsigned char);
8748 vector bool char vec_nor (vector bool char, vector bool char);
8750 vector float vec_or (vector float, vector float);
8751 vector float vec_or (vector float, vector bool int);
8752 vector float vec_or (vector bool int, vector float);
8753 vector bool int vec_or (vector bool int, vector bool int);
8754 vector signed int vec_or (vector bool int, vector signed int);
8755 vector signed int vec_or (vector signed int, vector bool int);
8756 vector signed int vec_or (vector signed int, vector signed int);
8757 vector unsigned int vec_or (vector bool int, vector unsigned int);
8758 vector unsigned int vec_or (vector unsigned int, vector bool int);
8759 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8760 vector bool short vec_or (vector bool short, vector bool short);
8761 vector signed short vec_or (vector bool short, vector signed short);
8762 vector signed short vec_or (vector signed short, vector bool short);
8763 vector signed short vec_or (vector signed short, vector signed short);
8764 vector unsigned short vec_or (vector bool short, vector unsigned short);
8765 vector unsigned short vec_or (vector unsigned short, vector bool short);
8766 vector unsigned short vec_or (vector unsigned short,
8767 vector unsigned short);
8768 vector signed char vec_or (vector bool char, vector signed char);
8769 vector bool char vec_or (vector bool char, vector bool char);
8770 vector signed char vec_or (vector signed char, vector bool char);
8771 vector signed char vec_or (vector signed char, vector signed char);
8772 vector unsigned char vec_or (vector bool char, vector unsigned char);
8773 vector unsigned char vec_or (vector unsigned char, vector bool char);
8774 vector unsigned char vec_or (vector unsigned char,
8775 vector unsigned char);
8777 vector signed char vec_pack (vector signed short, vector signed short);
8778 vector unsigned char vec_pack (vector unsigned short,
8779 vector unsigned short);
8780 vector bool char vec_pack (vector bool short, vector bool short);
8781 vector signed short vec_pack (vector signed int, vector signed int);
8782 vector unsigned short vec_pack (vector unsigned int,
8783 vector unsigned int);
8784 vector bool short vec_pack (vector bool int, vector bool int);
8786 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8787 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8788 vector unsigned short vec_vpkuwum (vector unsigned int,
8789 vector unsigned int);
8791 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8792 vector signed char vec_vpkuhum (vector signed short,
8793 vector signed short);
8794 vector unsigned char vec_vpkuhum (vector unsigned short,
8795 vector unsigned short);
8797 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8799 vector unsigned char vec_packs (vector unsigned short,
8800 vector unsigned short);
8801 vector signed char vec_packs (vector signed short, vector signed short);
8802 vector unsigned short vec_packs (vector unsigned int,
8803 vector unsigned int);
8804 vector signed short vec_packs (vector signed int, vector signed int);
8806 vector signed short vec_vpkswss (vector signed int, vector signed int);
8808 vector unsigned short vec_vpkuwus (vector unsigned int,
8809 vector unsigned int);
8811 vector signed char vec_vpkshss (vector signed short,
8812 vector signed short);
8814 vector unsigned char vec_vpkuhus (vector unsigned short,
8815 vector unsigned short);
8817 vector unsigned char vec_packsu (vector unsigned short,
8818 vector unsigned short);
8819 vector unsigned char vec_packsu (vector signed short,
8820 vector signed short);
8821 vector unsigned short vec_packsu (vector unsigned int,
8822 vector unsigned int);
8823 vector unsigned short vec_packsu (vector signed int, vector signed int);
8825 vector unsigned short vec_vpkswus (vector signed int,
8828 vector unsigned char vec_vpkshus (vector signed short,
8829 vector signed short);
8831 vector float vec_perm (vector float,
8833 vector unsigned char);
8834 vector signed int vec_perm (vector signed int,
8836 vector unsigned char);
8837 vector unsigned int vec_perm (vector unsigned int,
8838 vector unsigned int,
8839 vector unsigned char);
8840 vector bool int vec_perm (vector bool int,
8842 vector unsigned char);
8843 vector signed short vec_perm (vector signed short,
8844 vector signed short,
8845 vector unsigned char);
8846 vector unsigned short vec_perm (vector unsigned short,
8847 vector unsigned short,
8848 vector unsigned char);
8849 vector bool short vec_perm (vector bool short,
8851 vector unsigned char);
8852 vector pixel vec_perm (vector pixel,
8854 vector unsigned char);
8855 vector signed char vec_perm (vector signed char,
8857 vector unsigned char);
8858 vector unsigned char vec_perm (vector unsigned char,
8859 vector unsigned char,
8860 vector unsigned char);
8861 vector bool char vec_perm (vector bool char,
8863 vector unsigned char);
8865 vector float vec_re (vector float);
8867 vector signed char vec_rl (vector signed char,
8868 vector unsigned char);
8869 vector unsigned char vec_rl (vector unsigned char,
8870 vector unsigned char);
8871 vector signed short vec_rl (vector signed short, vector unsigned short);
8872 vector unsigned short vec_rl (vector unsigned short,
8873 vector unsigned short);
8874 vector signed int vec_rl (vector signed int, vector unsigned int);
8875 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8877 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8878 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8880 vector signed short vec_vrlh (vector signed short,
8881 vector unsigned short);
8882 vector unsigned short vec_vrlh (vector unsigned short,
8883 vector unsigned short);
8885 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8886 vector unsigned char vec_vrlb (vector unsigned char,
8887 vector unsigned char);
8889 vector float vec_round (vector float);
8891 vector float vec_rsqrte (vector float);
8893 vector float vec_sel (vector float, vector float, vector bool int);
8894 vector float vec_sel (vector float, vector float, vector unsigned int);
8895 vector signed int vec_sel (vector signed int,
8898 vector signed int vec_sel (vector signed int,
8900 vector unsigned int);
8901 vector unsigned int vec_sel (vector unsigned int,
8902 vector unsigned int,
8904 vector unsigned int vec_sel (vector unsigned int,
8905 vector unsigned int,
8906 vector unsigned int);
8907 vector bool int vec_sel (vector bool int,
8910 vector bool int vec_sel (vector bool int,
8912 vector unsigned int);
8913 vector signed short vec_sel (vector signed short,
8914 vector signed short,
8916 vector signed short vec_sel (vector signed short,
8917 vector signed short,
8918 vector unsigned short);
8919 vector unsigned short vec_sel (vector unsigned short,
8920 vector unsigned short,
8922 vector unsigned short vec_sel (vector unsigned short,
8923 vector unsigned short,
8924 vector unsigned short);
8925 vector bool short vec_sel (vector bool short,
8928 vector bool short vec_sel (vector bool short,
8930 vector unsigned short);
8931 vector signed char vec_sel (vector signed char,
8934 vector signed char vec_sel (vector signed char,
8936 vector unsigned char);
8937 vector unsigned char vec_sel (vector unsigned char,
8938 vector unsigned char,
8940 vector unsigned char vec_sel (vector unsigned char,
8941 vector unsigned char,
8942 vector unsigned char);
8943 vector bool char vec_sel (vector bool char,
8946 vector bool char vec_sel (vector bool char,
8948 vector unsigned char);
8950 vector signed char vec_sl (vector signed char,
8951 vector unsigned char);
8952 vector unsigned char vec_sl (vector unsigned char,
8953 vector unsigned char);
8954 vector signed short vec_sl (vector signed short, vector unsigned short);
8955 vector unsigned short vec_sl (vector unsigned short,
8956 vector unsigned short);
8957 vector signed int vec_sl (vector signed int, vector unsigned int);
8958 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
8960 vector signed int vec_vslw (vector signed int, vector unsigned int);
8961 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
8963 vector signed short vec_vslh (vector signed short,
8964 vector unsigned short);
8965 vector unsigned short vec_vslh (vector unsigned short,
8966 vector unsigned short);
8968 vector signed char vec_vslb (vector signed char, vector unsigned char);
8969 vector unsigned char vec_vslb (vector unsigned char,
8970 vector unsigned char);
8972 vector float vec_sld (vector float, vector float, const int);
8973 vector signed int vec_sld (vector signed int,
8976 vector unsigned int vec_sld (vector unsigned int,
8977 vector unsigned int,
8979 vector bool int vec_sld (vector bool int,
8982 vector signed short vec_sld (vector signed short,
8983 vector signed short,
8985 vector unsigned short vec_sld (vector unsigned short,
8986 vector unsigned short,
8988 vector bool short vec_sld (vector bool short,
8991 vector pixel vec_sld (vector pixel,
8994 vector signed char vec_sld (vector signed char,
8997 vector unsigned char vec_sld (vector unsigned char,
8998 vector unsigned char,
9000 vector bool char vec_sld (vector bool char,
9004 vector signed int vec_sll (vector signed int,
9005 vector unsigned int);
9006 vector signed int vec_sll (vector signed int,
9007 vector unsigned short);
9008 vector signed int vec_sll (vector signed int,
9009 vector unsigned char);
9010 vector unsigned int vec_sll (vector unsigned int,
9011 vector unsigned int);
9012 vector unsigned int vec_sll (vector unsigned int,
9013 vector unsigned short);
9014 vector unsigned int vec_sll (vector unsigned int,
9015 vector unsigned char);
9016 vector bool int vec_sll (vector bool int,
9017 vector unsigned int);
9018 vector bool int vec_sll (vector bool int,
9019 vector unsigned short);
9020 vector bool int vec_sll (vector bool int,
9021 vector unsigned char);
9022 vector signed short vec_sll (vector signed short,
9023 vector unsigned int);
9024 vector signed short vec_sll (vector signed short,
9025 vector unsigned short);
9026 vector signed short vec_sll (vector signed short,
9027 vector unsigned char);
9028 vector unsigned short vec_sll (vector unsigned short,
9029 vector unsigned int);
9030 vector unsigned short vec_sll (vector unsigned short,
9031 vector unsigned short);
9032 vector unsigned short vec_sll (vector unsigned short,
9033 vector unsigned char);
9034 vector bool short vec_sll (vector bool short, vector unsigned int);
9035 vector bool short vec_sll (vector bool short, vector unsigned short);
9036 vector bool short vec_sll (vector bool short, vector unsigned char);
9037 vector pixel vec_sll (vector pixel, vector unsigned int);
9038 vector pixel vec_sll (vector pixel, vector unsigned short);
9039 vector pixel vec_sll (vector pixel, vector unsigned char);
9040 vector signed char vec_sll (vector signed char, vector unsigned int);
9041 vector signed char vec_sll (vector signed char, vector unsigned short);
9042 vector signed char vec_sll (vector signed char, vector unsigned char);
9043 vector unsigned char vec_sll (vector unsigned char,
9044 vector unsigned int);
9045 vector unsigned char vec_sll (vector unsigned char,
9046 vector unsigned short);
9047 vector unsigned char vec_sll (vector unsigned char,
9048 vector unsigned char);
9049 vector bool char vec_sll (vector bool char, vector unsigned int);
9050 vector bool char vec_sll (vector bool char, vector unsigned short);
9051 vector bool char vec_sll (vector bool char, vector unsigned char);
9053 vector float vec_slo (vector float, vector signed char);
9054 vector float vec_slo (vector float, vector unsigned char);
9055 vector signed int vec_slo (vector signed int, vector signed char);
9056 vector signed int vec_slo (vector signed int, vector unsigned char);
9057 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9058 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9059 vector signed short vec_slo (vector signed short, vector signed char);
9060 vector signed short vec_slo (vector signed short, vector unsigned char);
9061 vector unsigned short vec_slo (vector unsigned short,
9062 vector signed char);
9063 vector unsigned short vec_slo (vector unsigned short,
9064 vector unsigned char);
9065 vector pixel vec_slo (vector pixel, vector signed char);
9066 vector pixel vec_slo (vector pixel, vector unsigned char);
9067 vector signed char vec_slo (vector signed char, vector signed char);
9068 vector signed char vec_slo (vector signed char, vector unsigned char);
9069 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9070 vector unsigned char vec_slo (vector unsigned char,
9071 vector unsigned char);
9073 vector signed char vec_splat (vector signed char, const int);
9074 vector unsigned char vec_splat (vector unsigned char, const int);
9075 vector bool char vec_splat (vector bool char, const int);
9076 vector signed short vec_splat (vector signed short, const int);
9077 vector unsigned short vec_splat (vector unsigned short, const int);
9078 vector bool short vec_splat (vector bool short, const int);
9079 vector pixel vec_splat (vector pixel, const int);
9080 vector float vec_splat (vector float, const int);
9081 vector signed int vec_splat (vector signed int, const int);
9082 vector unsigned int vec_splat (vector unsigned int, const int);
9083 vector bool int vec_splat (vector bool int, const int);
9085 vector float vec_vspltw (vector float, const int);
9086 vector signed int vec_vspltw (vector signed int, const int);
9087 vector unsigned int vec_vspltw (vector unsigned int, const int);
9088 vector bool int vec_vspltw (vector bool int, const int);
9090 vector bool short vec_vsplth (vector bool short, const int);
9091 vector signed short vec_vsplth (vector signed short, const int);
9092 vector unsigned short vec_vsplth (vector unsigned short, const int);
9093 vector pixel vec_vsplth (vector pixel, const int);
9095 vector signed char vec_vspltb (vector signed char, const int);
9096 vector unsigned char vec_vspltb (vector unsigned char, const int);
9097 vector bool char vec_vspltb (vector bool char, const int);
9099 vector signed char vec_splat_s8 (const int);
9101 vector signed short vec_splat_s16 (const int);
9103 vector signed int vec_splat_s32 (const int);
9105 vector unsigned char vec_splat_u8 (const int);
9107 vector unsigned short vec_splat_u16 (const int);
9109 vector unsigned int vec_splat_u32 (const int);
9111 vector signed char vec_sr (vector signed char, vector unsigned char);
9112 vector unsigned char vec_sr (vector unsigned char,
9113 vector unsigned char);
9114 vector signed short vec_sr (vector signed short,
9115 vector unsigned short);
9116 vector unsigned short vec_sr (vector unsigned short,
9117 vector unsigned short);
9118 vector signed int vec_sr (vector signed int, vector unsigned int);
9119 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9121 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9122 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9124 vector signed short vec_vsrh (vector signed short,
9125 vector unsigned short);
9126 vector unsigned short vec_vsrh (vector unsigned short,
9127 vector unsigned short);
9129 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9130 vector unsigned char vec_vsrb (vector unsigned char,
9131 vector unsigned char);
9133 vector signed char vec_sra (vector signed char, vector unsigned char);
9134 vector unsigned char vec_sra (vector unsigned char,
9135 vector unsigned char);
9136 vector signed short vec_sra (vector signed short,
9137 vector unsigned short);
9138 vector unsigned short vec_sra (vector unsigned short,
9139 vector unsigned short);
9140 vector signed int vec_sra (vector signed int, vector unsigned int);
9141 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9143 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9144 vector unsigned int vec_vsraw (vector unsigned int,
9145 vector unsigned int);
9147 vector signed short vec_vsrah (vector signed short,
9148 vector unsigned short);
9149 vector unsigned short vec_vsrah (vector unsigned short,
9150 vector unsigned short);
9152 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9153 vector unsigned char vec_vsrab (vector unsigned char,
9154 vector unsigned char);
9156 vector signed int vec_srl (vector signed int, vector unsigned int);
9157 vector signed int vec_srl (vector signed int, vector unsigned short);
9158 vector signed int vec_srl (vector signed int, vector unsigned char);
9159 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9160 vector unsigned int vec_srl (vector unsigned int,
9161 vector unsigned short);
9162 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9163 vector bool int vec_srl (vector bool int, vector unsigned int);
9164 vector bool int vec_srl (vector bool int, vector unsigned short);
9165 vector bool int vec_srl (vector bool int, vector unsigned char);
9166 vector signed short vec_srl (vector signed short, vector unsigned int);
9167 vector signed short vec_srl (vector signed short,
9168 vector unsigned short);
9169 vector signed short vec_srl (vector signed short, vector unsigned char);
9170 vector unsigned short vec_srl (vector unsigned short,
9171 vector unsigned int);
9172 vector unsigned short vec_srl (vector unsigned short,
9173 vector unsigned short);
9174 vector unsigned short vec_srl (vector unsigned short,
9175 vector unsigned char);
9176 vector bool short vec_srl (vector bool short, vector unsigned int);
9177 vector bool short vec_srl (vector bool short, vector unsigned short);
9178 vector bool short vec_srl (vector bool short, vector unsigned char);
9179 vector pixel vec_srl (vector pixel, vector unsigned int);
9180 vector pixel vec_srl (vector pixel, vector unsigned short);
9181 vector pixel vec_srl (vector pixel, vector unsigned char);
9182 vector signed char vec_srl (vector signed char, vector unsigned int);
9183 vector signed char vec_srl (vector signed char, vector unsigned short);
9184 vector signed char vec_srl (vector signed char, vector unsigned char);
9185 vector unsigned char vec_srl (vector unsigned char,
9186 vector unsigned int);
9187 vector unsigned char vec_srl (vector unsigned char,
9188 vector unsigned short);
9189 vector unsigned char vec_srl (vector unsigned char,
9190 vector unsigned char);
9191 vector bool char vec_srl (vector bool char, vector unsigned int);
9192 vector bool char vec_srl (vector bool char, vector unsigned short);
9193 vector bool char vec_srl (vector bool char, vector unsigned char);
9195 vector float vec_sro (vector float, vector signed char);
9196 vector float vec_sro (vector float, vector unsigned char);
9197 vector signed int vec_sro (vector signed int, vector signed char);
9198 vector signed int vec_sro (vector signed int, vector unsigned char);
9199 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9200 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9201 vector signed short vec_sro (vector signed short, vector signed char);
9202 vector signed short vec_sro (vector signed short, vector unsigned char);
9203 vector unsigned short vec_sro (vector unsigned short,
9204 vector signed char);
9205 vector unsigned short vec_sro (vector unsigned short,
9206 vector unsigned char);
9207 vector pixel vec_sro (vector pixel, vector signed char);
9208 vector pixel vec_sro (vector pixel, vector unsigned char);
9209 vector signed char vec_sro (vector signed char, vector signed char);
9210 vector signed char vec_sro (vector signed char, vector unsigned char);
9211 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9212 vector unsigned char vec_sro (vector unsigned char,
9213 vector unsigned char);
9215 void vec_st (vector float, int, vector float *);
9216 void vec_st (vector float, int, float *);
9217 void vec_st (vector signed int, int, vector signed int *);
9218 void vec_st (vector signed int, int, int *);
9219 void vec_st (vector unsigned int, int, vector unsigned int *);
9220 void vec_st (vector unsigned int, int, unsigned int *);
9221 void vec_st (vector bool int, int, vector bool int *);
9222 void vec_st (vector bool int, int, unsigned int *);
9223 void vec_st (vector bool int, int, int *);
9224 void vec_st (vector signed short, int, vector signed short *);
9225 void vec_st (vector signed short, int, short *);
9226 void vec_st (vector unsigned short, int, vector unsigned short *);
9227 void vec_st (vector unsigned short, int, unsigned short *);
9228 void vec_st (vector bool short, int, vector bool short *);
9229 void vec_st (vector bool short, int, unsigned short *);
9230 void vec_st (vector pixel, int, vector pixel *);
9231 void vec_st (vector pixel, int, unsigned short *);
9232 void vec_st (vector pixel, int, short *);
9233 void vec_st (vector bool short, int, short *);
9234 void vec_st (vector signed char, int, vector signed char *);
9235 void vec_st (vector signed char, int, signed char *);
9236 void vec_st (vector unsigned char, int, vector unsigned char *);
9237 void vec_st (vector unsigned char, int, unsigned char *);
9238 void vec_st (vector bool char, int, vector bool char *);
9239 void vec_st (vector bool char, int, unsigned char *);
9240 void vec_st (vector bool char, int, signed char *);
9242 void vec_ste (vector signed char, int, signed char *);
9243 void vec_ste (vector unsigned char, int, unsigned char *);
9244 void vec_ste (vector bool char, int, signed char *);
9245 void vec_ste (vector bool char, int, unsigned char *);
9246 void vec_ste (vector signed short, int, short *);
9247 void vec_ste (vector unsigned short, int, unsigned short *);
9248 void vec_ste (vector bool short, int, short *);
9249 void vec_ste (vector bool short, int, unsigned short *);
9250 void vec_ste (vector pixel, int, short *);
9251 void vec_ste (vector pixel, int, unsigned short *);
9252 void vec_ste (vector float, int, float *);
9253 void vec_ste (vector signed int, int, int *);
9254 void vec_ste (vector unsigned int, int, unsigned int *);
9255 void vec_ste (vector bool int, int, int *);
9256 void vec_ste (vector bool int, int, unsigned int *);
9258 void vec_stvewx (vector float, int, float *);
9259 void vec_stvewx (vector signed int, int, int *);
9260 void vec_stvewx (vector unsigned int, int, unsigned int *);
9261 void vec_stvewx (vector bool int, int, int *);
9262 void vec_stvewx (vector bool int, int, unsigned int *);
9264 void vec_stvehx (vector signed short, int, short *);
9265 void vec_stvehx (vector unsigned short, int, unsigned short *);
9266 void vec_stvehx (vector bool short, int, short *);
9267 void vec_stvehx (vector bool short, int, unsigned short *);
9268 void vec_stvehx (vector pixel, int, short *);
9269 void vec_stvehx (vector pixel, int, unsigned short *);
9271 void vec_stvebx (vector signed char, int, signed char *);
9272 void vec_stvebx (vector unsigned char, int, unsigned char *);
9273 void vec_stvebx (vector bool char, int, signed char *);
9274 void vec_stvebx (vector bool char, int, unsigned char *);
9276 void vec_stl (vector float, int, vector float *);
9277 void vec_stl (vector float, int, float *);
9278 void vec_stl (vector signed int, int, vector signed int *);
9279 void vec_stl (vector signed int, int, int *);
9280 void vec_stl (vector unsigned int, int, vector unsigned int *);
9281 void vec_stl (vector unsigned int, int, unsigned int *);
9282 void vec_stl (vector bool int, int, vector bool int *);
9283 void vec_stl (vector bool int, int, unsigned int *);
9284 void vec_stl (vector bool int, int, int *);
9285 void vec_stl (vector signed short, int, vector signed short *);
9286 void vec_stl (vector signed short, int, short *);
9287 void vec_stl (vector unsigned short, int, vector unsigned short *);
9288 void vec_stl (vector unsigned short, int, unsigned short *);
9289 void vec_stl (vector bool short, int, vector bool short *);
9290 void vec_stl (vector bool short, int, unsigned short *);
9291 void vec_stl (vector bool short, int, short *);
9292 void vec_stl (vector pixel, int, vector pixel *);
9293 void vec_stl (vector pixel, int, unsigned short *);
9294 void vec_stl (vector pixel, int, short *);
9295 void vec_stl (vector signed char, int, vector signed char *);
9296 void vec_stl (vector signed char, int, signed char *);
9297 void vec_stl (vector unsigned char, int, vector unsigned char *);
9298 void vec_stl (vector unsigned char, int, unsigned char *);
9299 void vec_stl (vector bool char, int, vector bool char *);
9300 void vec_stl (vector bool char, int, unsigned char *);
9301 void vec_stl (vector bool char, int, signed char *);
9303 vector signed char vec_sub (vector bool char, vector signed char);
9304 vector signed char vec_sub (vector signed char, vector bool char);
9305 vector signed char vec_sub (vector signed char, vector signed char);
9306 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9307 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9308 vector unsigned char vec_sub (vector unsigned char,
9309 vector unsigned char);
9310 vector signed short vec_sub (vector bool short, vector signed short);
9311 vector signed short vec_sub (vector signed short, vector bool short);
9312 vector signed short vec_sub (vector signed short, vector signed short);
9313 vector unsigned short vec_sub (vector bool short,
9314 vector unsigned short);
9315 vector unsigned short vec_sub (vector unsigned short,
9317 vector unsigned short vec_sub (vector unsigned short,
9318 vector unsigned short);
9319 vector signed int vec_sub (vector bool int, vector signed int);
9320 vector signed int vec_sub (vector signed int, vector bool int);
9321 vector signed int vec_sub (vector signed int, vector signed int);
9322 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9323 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9324 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9325 vector float vec_sub (vector float, vector float);
9327 vector float vec_vsubfp (vector float, vector float);
9329 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9330 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9331 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9332 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9333 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9334 vector unsigned int vec_vsubuwm (vector unsigned int,
9335 vector unsigned int);
9337 vector signed short vec_vsubuhm (vector bool short,
9338 vector signed short);
9339 vector signed short vec_vsubuhm (vector signed short,
9341 vector signed short vec_vsubuhm (vector signed short,
9342 vector signed short);
9343 vector unsigned short vec_vsubuhm (vector bool short,
9344 vector unsigned short);
9345 vector unsigned short vec_vsubuhm (vector unsigned short,
9347 vector unsigned short vec_vsubuhm (vector unsigned short,
9348 vector unsigned short);
9350 vector signed char vec_vsububm (vector bool char, vector signed char);
9351 vector signed char vec_vsububm (vector signed char, vector bool char);
9352 vector signed char vec_vsububm (vector signed char, vector signed char);
9353 vector unsigned char vec_vsububm (vector bool char,
9354 vector unsigned char);
9355 vector unsigned char vec_vsububm (vector unsigned char,
9357 vector unsigned char vec_vsububm (vector unsigned char,
9358 vector unsigned char);
9360 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9362 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9363 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9364 vector unsigned char vec_subs (vector unsigned char,
9365 vector unsigned char);
9366 vector signed char vec_subs (vector bool char, vector signed char);
9367 vector signed char vec_subs (vector signed char, vector bool char);
9368 vector signed char vec_subs (vector signed char, vector signed char);
9369 vector unsigned short vec_subs (vector bool short,
9370 vector unsigned short);
9371 vector unsigned short vec_subs (vector unsigned short,
9373 vector unsigned short vec_subs (vector unsigned short,
9374 vector unsigned short);
9375 vector signed short vec_subs (vector bool short, vector signed short);
9376 vector signed short vec_subs (vector signed short, vector bool short);
9377 vector signed short vec_subs (vector signed short, vector signed short);
9378 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9379 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9380 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9381 vector signed int vec_subs (vector bool int, vector signed int);
9382 vector signed int vec_subs (vector signed int, vector bool int);
9383 vector signed int vec_subs (vector signed int, vector signed int);
9385 vector signed int vec_vsubsws (vector bool int, vector signed int);
9386 vector signed int vec_vsubsws (vector signed int, vector bool int);
9387 vector signed int vec_vsubsws (vector signed int, vector signed int);
9389 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9390 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9391 vector unsigned int vec_vsubuws (vector unsigned int,
9392 vector unsigned int);
9394 vector signed short vec_vsubshs (vector bool short,
9395 vector signed short);
9396 vector signed short vec_vsubshs (vector signed short,
9398 vector signed short vec_vsubshs (vector signed short,
9399 vector signed short);
9401 vector unsigned short vec_vsubuhs (vector bool short,
9402 vector unsigned short);
9403 vector unsigned short vec_vsubuhs (vector unsigned short,
9405 vector unsigned short vec_vsubuhs (vector unsigned short,
9406 vector unsigned short);
9408 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9409 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9410 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9412 vector unsigned char vec_vsububs (vector bool char,
9413 vector unsigned char);
9414 vector unsigned char vec_vsububs (vector unsigned char,
9416 vector unsigned char vec_vsububs (vector unsigned char,
9417 vector unsigned char);
9419 vector unsigned int vec_sum4s (vector unsigned char,
9420 vector unsigned int);
9421 vector signed int vec_sum4s (vector signed char, vector signed int);
9422 vector signed int vec_sum4s (vector signed short, vector signed int);
9424 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9426 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9428 vector unsigned int vec_vsum4ubs (vector unsigned char,
9429 vector unsigned int);
9431 vector signed int vec_sum2s (vector signed int, vector signed int);
9433 vector signed int vec_sums (vector signed int, vector signed int);
9435 vector float vec_trunc (vector float);
9437 vector signed short vec_unpackh (vector signed char);
9438 vector bool short vec_unpackh (vector bool char);
9439 vector signed int vec_unpackh (vector signed short);
9440 vector bool int vec_unpackh (vector bool short);
9441 vector unsigned int vec_unpackh (vector pixel);
9443 vector bool int vec_vupkhsh (vector bool short);
9444 vector signed int vec_vupkhsh (vector signed short);
9446 vector unsigned int vec_vupkhpx (vector pixel);
9448 vector bool short vec_vupkhsb (vector bool char);
9449 vector signed short vec_vupkhsb (vector signed char);
9451 vector signed short vec_unpackl (vector signed char);
9452 vector bool short vec_unpackl (vector bool char);
9453 vector unsigned int vec_unpackl (vector pixel);
9454 vector signed int vec_unpackl (vector signed short);
9455 vector bool int vec_unpackl (vector bool short);
9457 vector unsigned int vec_vupklpx (vector pixel);
9459 vector bool int vec_vupklsh (vector bool short);
9460 vector signed int vec_vupklsh (vector signed short);
9462 vector bool short vec_vupklsb (vector bool char);
9463 vector signed short vec_vupklsb (vector signed char);
9465 vector float vec_xor (vector float, vector float);
9466 vector float vec_xor (vector float, vector bool int);
9467 vector float vec_xor (vector bool int, vector float);
9468 vector bool int vec_xor (vector bool int, vector bool int);
9469 vector signed int vec_xor (vector bool int, vector signed int);
9470 vector signed int vec_xor (vector signed int, vector bool int);
9471 vector signed int vec_xor (vector signed int, vector signed int);
9472 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9473 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9474 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9475 vector bool short vec_xor (vector bool short, vector bool short);
9476 vector signed short vec_xor (vector bool short, vector signed short);
9477 vector signed short vec_xor (vector signed short, vector bool short);
9478 vector signed short vec_xor (vector signed short, vector signed short);
9479 vector unsigned short vec_xor (vector bool short,
9480 vector unsigned short);
9481 vector unsigned short vec_xor (vector unsigned short,
9483 vector unsigned short vec_xor (vector unsigned short,
9484 vector unsigned short);
9485 vector signed char vec_xor (vector bool char, vector signed char);
9486 vector bool char vec_xor (vector bool char, vector bool char);
9487 vector signed char vec_xor (vector signed char, vector bool char);
9488 vector signed char vec_xor (vector signed char, vector signed char);
9489 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9490 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9491 vector unsigned char vec_xor (vector unsigned char,
9492 vector unsigned char);
9494 int vec_all_eq (vector signed char, vector bool char);
9495 int vec_all_eq (vector signed char, vector signed char);
9496 int vec_all_eq (vector unsigned char, vector bool char);
9497 int vec_all_eq (vector unsigned char, vector unsigned char);
9498 int vec_all_eq (vector bool char, vector bool char);
9499 int vec_all_eq (vector bool char, vector unsigned char);
9500 int vec_all_eq (vector bool char, vector signed char);
9501 int vec_all_eq (vector signed short, vector bool short);
9502 int vec_all_eq (vector signed short, vector signed short);
9503 int vec_all_eq (vector unsigned short, vector bool short);
9504 int vec_all_eq (vector unsigned short, vector unsigned short);
9505 int vec_all_eq (vector bool short, vector bool short);
9506 int vec_all_eq (vector bool short, vector unsigned short);
9507 int vec_all_eq (vector bool short, vector signed short);
9508 int vec_all_eq (vector pixel, vector pixel);
9509 int vec_all_eq (vector signed int, vector bool int);
9510 int vec_all_eq (vector signed int, vector signed int);
9511 int vec_all_eq (vector unsigned int, vector bool int);
9512 int vec_all_eq (vector unsigned int, vector unsigned int);
9513 int vec_all_eq (vector bool int, vector bool int);
9514 int vec_all_eq (vector bool int, vector unsigned int);
9515 int vec_all_eq (vector bool int, vector signed int);
9516 int vec_all_eq (vector float, vector float);
9518 int vec_all_ge (vector bool char, vector unsigned char);
9519 int vec_all_ge (vector unsigned char, vector bool char);
9520 int vec_all_ge (vector unsigned char, vector unsigned char);
9521 int vec_all_ge (vector bool char, vector signed char);
9522 int vec_all_ge (vector signed char, vector bool char);
9523 int vec_all_ge (vector signed char, vector signed char);
9524 int vec_all_ge (vector bool short, vector unsigned short);
9525 int vec_all_ge (vector unsigned short, vector bool short);
9526 int vec_all_ge (vector unsigned short, vector unsigned short);
9527 int vec_all_ge (vector signed short, vector signed short);
9528 int vec_all_ge (vector bool short, vector signed short);
9529 int vec_all_ge (vector signed short, vector bool short);
9530 int vec_all_ge (vector bool int, vector unsigned int);
9531 int vec_all_ge (vector unsigned int, vector bool int);
9532 int vec_all_ge (vector unsigned int, vector unsigned int);
9533 int vec_all_ge (vector bool int, vector signed int);
9534 int vec_all_ge (vector signed int, vector bool int);
9535 int vec_all_ge (vector signed int, vector signed int);
9536 int vec_all_ge (vector float, vector float);
9538 int vec_all_gt (vector bool char, vector unsigned char);
9539 int vec_all_gt (vector unsigned char, vector bool char);
9540 int vec_all_gt (vector unsigned char, vector unsigned char);
9541 int vec_all_gt (vector bool char, vector signed char);
9542 int vec_all_gt (vector signed char, vector bool char);
9543 int vec_all_gt (vector signed char, vector signed char);
9544 int vec_all_gt (vector bool short, vector unsigned short);
9545 int vec_all_gt (vector unsigned short, vector bool short);
9546 int vec_all_gt (vector unsigned short, vector unsigned short);
9547 int vec_all_gt (vector bool short, vector signed short);
9548 int vec_all_gt (vector signed short, vector bool short);
9549 int vec_all_gt (vector signed short, vector signed short);
9550 int vec_all_gt (vector bool int, vector unsigned int);
9551 int vec_all_gt (vector unsigned int, vector bool int);
9552 int vec_all_gt (vector unsigned int, vector unsigned int);
9553 int vec_all_gt (vector bool int, vector signed int);
9554 int vec_all_gt (vector signed int, vector bool int);
9555 int vec_all_gt (vector signed int, vector signed int);
9556 int vec_all_gt (vector float, vector float);
9558 int vec_all_in (vector float, vector float);
9560 int vec_all_le (vector bool char, vector unsigned char);
9561 int vec_all_le (vector unsigned char, vector bool char);
9562 int vec_all_le (vector unsigned char, vector unsigned char);
9563 int vec_all_le (vector bool char, vector signed char);
9564 int vec_all_le (vector signed char, vector bool char);
9565 int vec_all_le (vector signed char, vector signed char);
9566 int vec_all_le (vector bool short, vector unsigned short);
9567 int vec_all_le (vector unsigned short, vector bool short);
9568 int vec_all_le (vector unsigned short, vector unsigned short);
9569 int vec_all_le (vector bool short, vector signed short);
9570 int vec_all_le (vector signed short, vector bool short);
9571 int vec_all_le (vector signed short, vector signed short);
9572 int vec_all_le (vector bool int, vector unsigned int);
9573 int vec_all_le (vector unsigned int, vector bool int);
9574 int vec_all_le (vector unsigned int, vector unsigned int);
9575 int vec_all_le (vector bool int, vector signed int);
9576 int vec_all_le (vector signed int, vector bool int);
9577 int vec_all_le (vector signed int, vector signed int);
9578 int vec_all_le (vector float, vector float);
9580 int vec_all_lt (vector bool char, vector unsigned char);
9581 int vec_all_lt (vector unsigned char, vector bool char);
9582 int vec_all_lt (vector unsigned char, vector unsigned char);
9583 int vec_all_lt (vector bool char, vector signed char);
9584 int vec_all_lt (vector signed char, vector bool char);
9585 int vec_all_lt (vector signed char, vector signed char);
9586 int vec_all_lt (vector bool short, vector unsigned short);
9587 int vec_all_lt (vector unsigned short, vector bool short);
9588 int vec_all_lt (vector unsigned short, vector unsigned short);
9589 int vec_all_lt (vector bool short, vector signed short);
9590 int vec_all_lt (vector signed short, vector bool short);
9591 int vec_all_lt (vector signed short, vector signed short);
9592 int vec_all_lt (vector bool int, vector unsigned int);
9593 int vec_all_lt (vector unsigned int, vector bool int);
9594 int vec_all_lt (vector unsigned int, vector unsigned int);
9595 int vec_all_lt (vector bool int, vector signed int);
9596 int vec_all_lt (vector signed int, vector bool int);
9597 int vec_all_lt (vector signed int, vector signed int);
9598 int vec_all_lt (vector float, vector float);
9600 int vec_all_nan (vector float);
9602 int vec_all_ne (vector signed char, vector bool char);
9603 int vec_all_ne (vector signed char, vector signed char);
9604 int vec_all_ne (vector unsigned char, vector bool char);
9605 int vec_all_ne (vector unsigned char, vector unsigned char);
9606 int vec_all_ne (vector bool char, vector bool char);
9607 int vec_all_ne (vector bool char, vector unsigned char);
9608 int vec_all_ne (vector bool char, vector signed char);
9609 int vec_all_ne (vector signed short, vector bool short);
9610 int vec_all_ne (vector signed short, vector signed short);
9611 int vec_all_ne (vector unsigned short, vector bool short);
9612 int vec_all_ne (vector unsigned short, vector unsigned short);
9613 int vec_all_ne (vector bool short, vector bool short);
9614 int vec_all_ne (vector bool short, vector unsigned short);
9615 int vec_all_ne (vector bool short, vector signed short);
9616 int vec_all_ne (vector pixel, vector pixel);
9617 int vec_all_ne (vector signed int, vector bool int);
9618 int vec_all_ne (vector signed int, vector signed int);
9619 int vec_all_ne (vector unsigned int, vector bool int);
9620 int vec_all_ne (vector unsigned int, vector unsigned int);
9621 int vec_all_ne (vector bool int, vector bool int);
9622 int vec_all_ne (vector bool int, vector unsigned int);
9623 int vec_all_ne (vector bool int, vector signed int);
9624 int vec_all_ne (vector float, vector float);
9626 int vec_all_nge (vector float, vector float);
9628 int vec_all_ngt (vector float, vector float);
9630 int vec_all_nle (vector float, vector float);
9632 int vec_all_nlt (vector float, vector float);
9634 int vec_all_numeric (vector float);
9636 int vec_any_eq (vector signed char, vector bool char);
9637 int vec_any_eq (vector signed char, vector signed char);
9638 int vec_any_eq (vector unsigned char, vector bool char);
9639 int vec_any_eq (vector unsigned char, vector unsigned char);
9640 int vec_any_eq (vector bool char, vector bool char);
9641 int vec_any_eq (vector bool char, vector unsigned char);
9642 int vec_any_eq (vector bool char, vector signed char);
9643 int vec_any_eq (vector signed short, vector bool short);
9644 int vec_any_eq (vector signed short, vector signed short);
9645 int vec_any_eq (vector unsigned short, vector bool short);
9646 int vec_any_eq (vector unsigned short, vector unsigned short);
9647 int vec_any_eq (vector bool short, vector bool short);
9648 int vec_any_eq (vector bool short, vector unsigned short);
9649 int vec_any_eq (vector bool short, vector signed short);
9650 int vec_any_eq (vector pixel, vector pixel);
9651 int vec_any_eq (vector signed int, vector bool int);
9652 int vec_any_eq (vector signed int, vector signed int);
9653 int vec_any_eq (vector unsigned int, vector bool int);
9654 int vec_any_eq (vector unsigned int, vector unsigned int);
9655 int vec_any_eq (vector bool int, vector bool int);
9656 int vec_any_eq (vector bool int, vector unsigned int);
9657 int vec_any_eq (vector bool int, vector signed int);
9658 int vec_any_eq (vector float, vector float);
9660 int vec_any_ge (vector signed char, vector bool char);
9661 int vec_any_ge (vector unsigned char, vector bool char);
9662 int vec_any_ge (vector unsigned char, vector unsigned char);
9663 int vec_any_ge (vector signed char, vector signed char);
9664 int vec_any_ge (vector bool char, vector unsigned char);
9665 int vec_any_ge (vector bool char, vector signed char);
9666 int vec_any_ge (vector unsigned short, vector bool short);
9667 int vec_any_ge (vector unsigned short, vector unsigned short);
9668 int vec_any_ge (vector signed short, vector signed short);
9669 int vec_any_ge (vector signed short, vector bool short);
9670 int vec_any_ge (vector bool short, vector unsigned short);
9671 int vec_any_ge (vector bool short, vector signed short);
9672 int vec_any_ge (vector signed int, vector bool int);
9673 int vec_any_ge (vector unsigned int, vector bool int);
9674 int vec_any_ge (vector unsigned int, vector unsigned int);
9675 int vec_any_ge (vector signed int, vector signed int);
9676 int vec_any_ge (vector bool int, vector unsigned int);
9677 int vec_any_ge (vector bool int, vector signed int);
9678 int vec_any_ge (vector float, vector float);
9680 int vec_any_gt (vector bool char, vector unsigned char);
9681 int vec_any_gt (vector unsigned char, vector bool char);
9682 int vec_any_gt (vector unsigned char, vector unsigned char);
9683 int vec_any_gt (vector bool char, vector signed char);
9684 int vec_any_gt (vector signed char, vector bool char);
9685 int vec_any_gt (vector signed char, vector signed char);
9686 int vec_any_gt (vector bool short, vector unsigned short);
9687 int vec_any_gt (vector unsigned short, vector bool short);
9688 int vec_any_gt (vector unsigned short, vector unsigned short);
9689 int vec_any_gt (vector bool short, vector signed short);
9690 int vec_any_gt (vector signed short, vector bool short);
9691 int vec_any_gt (vector signed short, vector signed short);
9692 int vec_any_gt (vector bool int, vector unsigned int);
9693 int vec_any_gt (vector unsigned int, vector bool int);
9694 int vec_any_gt (vector unsigned int, vector unsigned int);
9695 int vec_any_gt (vector bool int, vector signed int);
9696 int vec_any_gt (vector signed int, vector bool int);
9697 int vec_any_gt (vector signed int, vector signed int);
9698 int vec_any_gt (vector float, vector float);
9700 int vec_any_le (vector bool char, vector unsigned char);
9701 int vec_any_le (vector unsigned char, vector bool char);
9702 int vec_any_le (vector unsigned char, vector unsigned char);
9703 int vec_any_le (vector bool char, vector signed char);
9704 int vec_any_le (vector signed char, vector bool char);
9705 int vec_any_le (vector signed char, vector signed char);
9706 int vec_any_le (vector bool short, vector unsigned short);
9707 int vec_any_le (vector unsigned short, vector bool short);
9708 int vec_any_le (vector unsigned short, vector unsigned short);
9709 int vec_any_le (vector bool short, vector signed short);
9710 int vec_any_le (vector signed short, vector bool short);
9711 int vec_any_le (vector signed short, vector signed short);
9712 int vec_any_le (vector bool int, vector unsigned int);
9713 int vec_any_le (vector unsigned int, vector bool int);
9714 int vec_any_le (vector unsigned int, vector unsigned int);
9715 int vec_any_le (vector bool int, vector signed int);
9716 int vec_any_le (vector signed int, vector bool int);
9717 int vec_any_le (vector signed int, vector signed int);
9718 int vec_any_le (vector float, vector float);
9720 int vec_any_lt (vector bool char, vector unsigned char);
9721 int vec_any_lt (vector unsigned char, vector bool char);
9722 int vec_any_lt (vector unsigned char, vector unsigned char);
9723 int vec_any_lt (vector bool char, vector signed char);
9724 int vec_any_lt (vector signed char, vector bool char);
9725 int vec_any_lt (vector signed char, vector signed char);
9726 int vec_any_lt (vector bool short, vector unsigned short);
9727 int vec_any_lt (vector unsigned short, vector bool short);
9728 int vec_any_lt (vector unsigned short, vector unsigned short);
9729 int vec_any_lt (vector bool short, vector signed short);
9730 int vec_any_lt (vector signed short, vector bool short);
9731 int vec_any_lt (vector signed short, vector signed short);
9732 int vec_any_lt (vector bool int, vector unsigned int);
9733 int vec_any_lt (vector unsigned int, vector bool int);
9734 int vec_any_lt (vector unsigned int, vector unsigned int);
9735 int vec_any_lt (vector bool int, vector signed int);
9736 int vec_any_lt (vector signed int, vector bool int);
9737 int vec_any_lt (vector signed int, vector signed int);
9738 int vec_any_lt (vector float, vector float);
9740 int vec_any_nan (vector float);
9742 int vec_any_ne (vector signed char, vector bool char);
9743 int vec_any_ne (vector signed char, vector signed char);
9744 int vec_any_ne (vector unsigned char, vector bool char);
9745 int vec_any_ne (vector unsigned char, vector unsigned char);
9746 int vec_any_ne (vector bool char, vector bool char);
9747 int vec_any_ne (vector bool char, vector unsigned char);
9748 int vec_any_ne (vector bool char, vector signed char);
9749 int vec_any_ne (vector signed short, vector bool short);
9750 int vec_any_ne (vector signed short, vector signed short);
9751 int vec_any_ne (vector unsigned short, vector bool short);
9752 int vec_any_ne (vector unsigned short, vector unsigned short);
9753 int vec_any_ne (vector bool short, vector bool short);
9754 int vec_any_ne (vector bool short, vector unsigned short);
9755 int vec_any_ne (vector bool short, vector signed short);
9756 int vec_any_ne (vector pixel, vector pixel);
9757 int vec_any_ne (vector signed int, vector bool int);
9758 int vec_any_ne (vector signed int, vector signed int);
9759 int vec_any_ne (vector unsigned int, vector bool int);
9760 int vec_any_ne (vector unsigned int, vector unsigned int);
9761 int vec_any_ne (vector bool int, vector bool int);
9762 int vec_any_ne (vector bool int, vector unsigned int);
9763 int vec_any_ne (vector bool int, vector signed int);
9764 int vec_any_ne (vector float, vector float);
9766 int vec_any_nge (vector float, vector float);
9768 int vec_any_ngt (vector float, vector float);
9770 int vec_any_nle (vector float, vector float);
9772 int vec_any_nlt (vector float, vector float);
9774 int vec_any_numeric (vector float);
9776 int vec_any_out (vector float, vector float);
9779 @node SPARC VIS Built-in Functions
9780 @subsection SPARC VIS Built-in Functions
9782 GCC supports SIMD operations on the SPARC using both the generic vector
9783 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9784 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9785 switch, the VIS extension is exposed as the following built-in functions:
9788 typedef int v2si __attribute__ ((vector_size (8)));
9789 typedef short v4hi __attribute__ ((vector_size (8)));
9790 typedef short v2hi __attribute__ ((vector_size (4)));
9791 typedef char v8qi __attribute__ ((vector_size (8)));
9792 typedef char v4qi __attribute__ ((vector_size (4)));
9794 void * __builtin_vis_alignaddr (void *, long);
9795 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9796 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9797 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9798 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9800 v4hi __builtin_vis_fexpand (v4qi);
9802 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9803 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9804 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9805 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9806 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9807 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9808 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9810 v4qi __builtin_vis_fpack16 (v4hi);
9811 v8qi __builtin_vis_fpack32 (v2si, v2si);
9812 v2hi __builtin_vis_fpackfix (v2si);
9813 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9815 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9818 @node SPU Built-in Functions
9819 @subsection SPU Built-in Functions
9821 GCC provides extensions for the SPU processor as described in the
9822 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
9823 found at @uref{http://cell.scei.co.jp/} or
9824 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
9825 implementation differs in several ways.
9830 The optional extension of specifying vector constants in parentheses is
9834 A vector initializer requires no cast if the vector constant is of the
9835 same type as the variable it is initializing.
9838 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9839 vector type is the default signedness of the base type. The default
9840 varies depending on the operating system, so a portable program should
9841 always specify the signedness.
9844 By default, the keyword @code{__vector} is added. The macro
9845 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
9849 GCC allows using a @code{typedef} name as the type specifier for a
9853 For C, overloaded functions are implemented with macros so the following
9857 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
9860 Since @code{spu_add} is a macro, the vector constant in the example
9861 is treated as four separate arguments. Wrap the entire argument in
9862 parentheses for this to work.
9865 The extended version of @code{__builtin_expect} is not supported.
9869 @emph{Note:} Only the interface described in the aforementioned
9870 specification is supported. Internally, GCC uses built-in functions to
9871 implement the required functionality, but these are not supported and
9872 are subject to change without notice.
9874 @node Target Format Checks
9875 @section Format Checks Specific to Particular Target Machines
9877 For some target machines, GCC supports additional options to the
9879 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9882 * Solaris Format Checks::
9885 @node Solaris Format Checks
9886 @subsection Solaris Format Checks
9888 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9889 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9890 conversions, and the two-argument @code{%b} conversion for displaying
9891 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9894 @section Pragmas Accepted by GCC
9898 GCC supports several types of pragmas, primarily in order to compile
9899 code originally written for other compilers. Note that in general
9900 we do not recommend the use of pragmas; @xref{Function Attributes},
9901 for further explanation.
9906 * RS/6000 and PowerPC Pragmas::
9909 * Symbol-Renaming Pragmas::
9910 * Structure-Packing Pragmas::
9912 * Diagnostic Pragmas::
9913 * Visibility Pragmas::
9917 @subsection ARM Pragmas
9919 The ARM target defines pragmas for controlling the default addition of
9920 @code{long_call} and @code{short_call} attributes to functions.
9921 @xref{Function Attributes}, for information about the effects of these
9926 @cindex pragma, long_calls
9927 Set all subsequent functions to have the @code{long_call} attribute.
9930 @cindex pragma, no_long_calls
9931 Set all subsequent functions to have the @code{short_call} attribute.
9933 @item long_calls_off
9934 @cindex pragma, long_calls_off
9935 Do not affect the @code{long_call} or @code{short_call} attributes of
9936 subsequent functions.
9940 @subsection M32C Pragmas
9943 @item memregs @var{number}
9944 @cindex pragma, memregs
9945 Overrides the command line option @code{-memregs=} for the current
9946 file. Use with care! This pragma must be before any function in the
9947 file, and mixing different memregs values in different objects may
9948 make them incompatible. This pragma is useful when a
9949 performance-critical function uses a memreg for temporary values,
9950 as it may allow you to reduce the number of memregs used.
9954 @node RS/6000 and PowerPC Pragmas
9955 @subsection RS/6000 and PowerPC Pragmas
9957 The RS/6000 and PowerPC targets define one pragma for controlling
9958 whether or not the @code{longcall} attribute is added to function
9959 declarations by default. This pragma overrides the @option{-mlongcall}
9960 option, but not the @code{longcall} and @code{shortcall} attributes.
9961 @xref{RS/6000 and PowerPC Options}, for more information about when long
9962 calls are and are not necessary.
9966 @cindex pragma, longcall
9967 Apply the @code{longcall} attribute to all subsequent function
9971 Do not apply the @code{longcall} attribute to subsequent function
9975 @c Describe c4x pragmas here.
9976 @c Describe h8300 pragmas here.
9977 @c Describe sh pragmas here.
9978 @c Describe v850 pragmas here.
9980 @node Darwin Pragmas
9981 @subsection Darwin Pragmas
9983 The following pragmas are available for all architectures running the
9984 Darwin operating system. These are useful for compatibility with other
9988 @item mark @var{tokens}@dots{}
9989 @cindex pragma, mark
9990 This pragma is accepted, but has no effect.
9992 @item options align=@var{alignment}
9993 @cindex pragma, options align
9994 This pragma sets the alignment of fields in structures. The values of
9995 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9996 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9997 properly; to restore the previous setting, use @code{reset} for the
10000 @item segment @var{tokens}@dots{}
10001 @cindex pragma, segment
10002 This pragma is accepted, but has no effect.
10004 @item unused (@var{var} [, @var{var}]@dots{})
10005 @cindex pragma, unused
10006 This pragma declares variables to be possibly unused. GCC will not
10007 produce warnings for the listed variables. The effect is similar to
10008 that of the @code{unused} attribute, except that this pragma may appear
10009 anywhere within the variables' scopes.
10012 @node Solaris Pragmas
10013 @subsection Solaris Pragmas
10015 The Solaris target supports @code{#pragma redefine_extname}
10016 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
10017 @code{#pragma} directives for compatibility with the system compiler.
10020 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
10021 @cindex pragma, align
10023 Increase the minimum alignment of each @var{variable} to @var{alignment}.
10024 This is the same as GCC's @code{aligned} attribute @pxref{Variable
10025 Attributes}). Macro expansion occurs on the arguments to this pragma
10026 when compiling C and Objective-C. It does not currently occur when
10027 compiling C++, but this is a bug which may be fixed in a future
10030 @item fini (@var{function} [, @var{function}]...)
10031 @cindex pragma, fini
10033 This pragma causes each listed @var{function} to be called after
10034 main, or during shared module unloading, by adding a call to the
10035 @code{.fini} section.
10037 @item init (@var{function} [, @var{function}]...)
10038 @cindex pragma, init
10040 This pragma causes each listed @var{function} to be called during
10041 initialization (before @code{main}) or during shared module loading, by
10042 adding a call to the @code{.init} section.
10046 @node Symbol-Renaming Pragmas
10047 @subsection Symbol-Renaming Pragmas
10049 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
10050 supports two @code{#pragma} directives which change the name used in
10051 assembly for a given declaration. These pragmas are only available on
10052 platforms whose system headers need them. To get this effect on all
10053 platforms supported by GCC, use the asm labels extension (@pxref{Asm
10057 @item redefine_extname @var{oldname} @var{newname}
10058 @cindex pragma, redefine_extname
10060 This pragma gives the C function @var{oldname} the assembly symbol
10061 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
10062 will be defined if this pragma is available (currently only on
10065 @item extern_prefix @var{string}
10066 @cindex pragma, extern_prefix
10068 This pragma causes all subsequent external function and variable
10069 declarations to have @var{string} prepended to their assembly symbols.
10070 This effect may be terminated with another @code{extern_prefix} pragma
10071 whose argument is an empty string. The preprocessor macro
10072 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10073 available (currently only on Tru64 UNIX)@.
10076 These pragmas and the asm labels extension interact in a complicated
10077 manner. Here are some corner cases you may want to be aware of.
10080 @item Both pragmas silently apply only to declarations with external
10081 linkage. Asm labels do not have this restriction.
10083 @item In C++, both pragmas silently apply only to declarations with
10084 ``C'' linkage. Again, asm labels do not have this restriction.
10086 @item If any of the three ways of changing the assembly name of a
10087 declaration is applied to a declaration whose assembly name has
10088 already been determined (either by a previous use of one of these
10089 features, or because the compiler needed the assembly name in order to
10090 generate code), and the new name is different, a warning issues and
10091 the name does not change.
10093 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10094 always the C-language name.
10096 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10097 occurs with an asm label attached, the prefix is silently ignored for
10100 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10101 apply to the same declaration, whichever triggered first wins, and a
10102 warning issues if they contradict each other. (We would like to have
10103 @code{#pragma redefine_extname} always win, for consistency with asm
10104 labels, but if @code{#pragma extern_prefix} triggers first we have no
10105 way of knowing that that happened.)
10108 @node Structure-Packing Pragmas
10109 @subsection Structure-Packing Pragmas
10111 For compatibility with Win32, GCC supports a set of @code{#pragma}
10112 directives which change the maximum alignment of members of structures
10113 (other than zero-width bitfields), unions, and classes subsequently
10114 defined. The @var{n} value below always is required to be a small power
10115 of two and specifies the new alignment in bytes.
10118 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10119 @item @code{#pragma pack()} sets the alignment to the one that was in
10120 effect when compilation started (see also command line option
10121 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10122 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10123 setting on an internal stack and then optionally sets the new alignment.
10124 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10125 saved at the top of the internal stack (and removes that stack entry).
10126 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10127 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10128 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10129 @code{#pragma pack(pop)}.
10132 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10133 @code{#pragma} which lays out a structure as the documented
10134 @code{__attribute__ ((ms_struct))}.
10136 @item @code{#pragma ms_struct on} turns on the layout for structures
10138 @item @code{#pragma ms_struct off} turns off the layout for structures
10140 @item @code{#pragma ms_struct reset} goes back to the default layout.
10144 @subsection Weak Pragmas
10146 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10147 directives for declaring symbols to be weak, and defining weak
10151 @item #pragma weak @var{symbol}
10152 @cindex pragma, weak
10153 This pragma declares @var{symbol} to be weak, as if the declaration
10154 had the attribute of the same name. The pragma may appear before
10155 or after the declaration of @var{symbol}, but must appear before
10156 either its first use or its definition. It is not an error for
10157 @var{symbol} to never be defined at all.
10159 @item #pragma weak @var{symbol1} = @var{symbol2}
10160 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10161 It is an error if @var{symbol2} is not defined in the current
10165 @node Diagnostic Pragmas
10166 @subsection Diagnostic Pragmas
10168 GCC allows the user to selectively enable or disable certain types of
10169 diagnostics, and change the kind of the diagnostic. For example, a
10170 project's policy might require that all sources compile with
10171 @option{-Werror} but certain files might have exceptions allowing
10172 specific types of warnings. Or, a project might selectively enable
10173 diagnostics and treat them as errors depending on which preprocessor
10174 macros are defined.
10177 @item #pragma GCC diagnostic @var{kind} @var{option}
10178 @cindex pragma, diagnostic
10180 Modifies the disposition of a diagnostic. Note that not all
10181 diagnostics are modifiable; at the moment only warnings (normally
10182 controlled by @samp{-W...}) can be controlled, and not all of them.
10183 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10184 are controllable and which option controls them.
10186 @var{kind} is @samp{error} to treat this diagnostic as an error,
10187 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10188 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10189 @var{option} is a double quoted string which matches the command line
10193 #pragma GCC diagnostic warning "-Wformat"
10194 #pragma GCC diagnostic error "-Wformat"
10195 #pragma GCC diagnostic ignored "-Wformat"
10198 Note that these pragmas override any command line options. Also,
10199 while it is syntactically valid to put these pragmas anywhere in your
10200 sources, the only supported location for them is before any data or
10201 functions are defined. Doing otherwise may result in unpredictable
10202 results depending on how the optimizer manages your sources. If the
10203 same option is listed multiple times, the last one specified is the
10204 one that is in effect. This pragma is not intended to be a general
10205 purpose replacement for command line options, but for implementing
10206 strict control over project policies.
10210 @node Visibility Pragmas
10211 @subsection Visibility Pragmas
10214 @item #pragma GCC visibility push(@var{visibility})
10215 @itemx #pragma GCC visibility pop
10216 @cindex pragma, visibility
10218 This pragma allows the user to set the visibility for multiple
10219 declarations without having to give each a visibility attribute
10220 @xref{Function Attributes}, for more information about visibility and
10221 the attribute syntax.
10223 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10224 declarations. Class members and template specializations are not
10225 affected; if you want to override the visibility for a particular
10226 member or instantiation, you must use an attribute.
10230 @node Unnamed Fields
10231 @section Unnamed struct/union fields within structs/unions
10235 For compatibility with other compilers, GCC allows you to define
10236 a structure or union that contains, as fields, structures and unions
10237 without names. For example:
10250 In this example, the user would be able to access members of the unnamed
10251 union with code like @samp{foo.b}. Note that only unnamed structs and
10252 unions are allowed, you may not have, for example, an unnamed
10255 You must never create such structures that cause ambiguous field definitions.
10256 For example, this structure:
10267 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10268 Such constructs are not supported and must be avoided. In the future,
10269 such constructs may be detected and treated as compilation errors.
10271 @opindex fms-extensions
10272 Unless @option{-fms-extensions} is used, the unnamed field must be a
10273 structure or union definition without a tag (for example, @samp{struct
10274 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10275 also be a definition with a tag such as @samp{struct foo @{ int a;
10276 @};}, a reference to a previously defined structure or union such as
10277 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10278 previously defined structure or union type.
10281 @section Thread-Local Storage
10282 @cindex Thread-Local Storage
10283 @cindex @acronym{TLS}
10286 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10287 are allocated such that there is one instance of the variable per extant
10288 thread. The run-time model GCC uses to implement this originates
10289 in the IA-64 processor-specific ABI, but has since been migrated
10290 to other processors as well. It requires significant support from
10291 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10292 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10293 is not available everywhere.
10295 At the user level, the extension is visible with a new storage
10296 class keyword: @code{__thread}. For example:
10300 extern __thread struct state s;
10301 static __thread char *p;
10304 The @code{__thread} specifier may be used alone, with the @code{extern}
10305 or @code{static} specifiers, but with no other storage class specifier.
10306 When used with @code{extern} or @code{static}, @code{__thread} must appear
10307 immediately after the other storage class specifier.
10309 The @code{__thread} specifier may be applied to any global, file-scoped
10310 static, function-scoped static, or static data member of a class. It may
10311 not be applied to block-scoped automatic or non-static data member.
10313 When the address-of operator is applied to a thread-local variable, it is
10314 evaluated at run-time and returns the address of the current thread's
10315 instance of that variable. An address so obtained may be used by any
10316 thread. When a thread terminates, any pointers to thread-local variables
10317 in that thread become invalid.
10319 No static initialization may refer to the address of a thread-local variable.
10321 In C++, if an initializer is present for a thread-local variable, it must
10322 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10325 See @uref{http://people.redhat.com/drepper/tls.pdf,
10326 ELF Handling For Thread-Local Storage} for a detailed explanation of
10327 the four thread-local storage addressing models, and how the run-time
10328 is expected to function.
10331 * C99 Thread-Local Edits::
10332 * C++98 Thread-Local Edits::
10335 @node C99 Thread-Local Edits
10336 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10338 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10339 that document the exact semantics of the language extension.
10343 @cite{5.1.2 Execution environments}
10345 Add new text after paragraph 1
10348 Within either execution environment, a @dfn{thread} is a flow of
10349 control within a program. It is implementation defined whether
10350 or not there may be more than one thread associated with a program.
10351 It is implementation defined how threads beyond the first are
10352 created, the name and type of the function called at thread
10353 startup, and how threads may be terminated. However, objects
10354 with thread storage duration shall be initialized before thread
10359 @cite{6.2.4 Storage durations of objects}
10361 Add new text before paragraph 3
10364 An object whose identifier is declared with the storage-class
10365 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10366 Its lifetime is the entire execution of the thread, and its
10367 stored value is initialized only once, prior to thread startup.
10371 @cite{6.4.1 Keywords}
10373 Add @code{__thread}.
10376 @cite{6.7.1 Storage-class specifiers}
10378 Add @code{__thread} to the list of storage class specifiers in
10381 Change paragraph 2 to
10384 With the exception of @code{__thread}, at most one storage-class
10385 specifier may be given [@dots{}]. The @code{__thread} specifier may
10386 be used alone, or immediately following @code{extern} or
10390 Add new text after paragraph 6
10393 The declaration of an identifier for a variable that has
10394 block scope that specifies @code{__thread} shall also
10395 specify either @code{extern} or @code{static}.
10397 The @code{__thread} specifier shall be used only with
10402 @node C++98 Thread-Local Edits
10403 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10405 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10406 that document the exact semantics of the language extension.
10410 @b{[intro.execution]}
10412 New text after paragraph 4
10415 A @dfn{thread} is a flow of control within the abstract machine.
10416 It is implementation defined whether or not there may be more than
10420 New text after paragraph 7
10423 It is unspecified whether additional action must be taken to
10424 ensure when and whether side effects are visible to other threads.
10430 Add @code{__thread}.
10433 @b{[basic.start.main]}
10435 Add after paragraph 5
10438 The thread that begins execution at the @code{main} function is called
10439 the @dfn{main thread}. It is implementation defined how functions
10440 beginning threads other than the main thread are designated or typed.
10441 A function so designated, as well as the @code{main} function, is called
10442 a @dfn{thread startup function}. It is implementation defined what
10443 happens if a thread startup function returns. It is implementation
10444 defined what happens to other threads when any thread calls @code{exit}.
10448 @b{[basic.start.init]}
10450 Add after paragraph 4
10453 The storage for an object of thread storage duration shall be
10454 statically initialized before the first statement of the thread startup
10455 function. An object of thread storage duration shall not require
10456 dynamic initialization.
10460 @b{[basic.start.term]}
10462 Add after paragraph 3
10465 The type of an object with thread storage duration shall not have a
10466 non-trivial destructor, nor shall it be an array type whose elements
10467 (directly or indirectly) have non-trivial destructors.
10473 Add ``thread storage duration'' to the list in paragraph 1.
10478 Thread, static, and automatic storage durations are associated with
10479 objects introduced by declarations [@dots{}].
10482 Add @code{__thread} to the list of specifiers in paragraph 3.
10485 @b{[basic.stc.thread]}
10487 New section before @b{[basic.stc.static]}
10490 The keyword @code{__thread} applied to a non-local object gives the
10491 object thread storage duration.
10493 A local variable or class data member declared both @code{static}
10494 and @code{__thread} gives the variable or member thread storage
10499 @b{[basic.stc.static]}
10504 All objects which have neither thread storage duration, dynamic
10505 storage duration nor are local [@dots{}].
10511 Add @code{__thread} to the list in paragraph 1.
10516 With the exception of @code{__thread}, at most one
10517 @var{storage-class-specifier} shall appear in a given
10518 @var{decl-specifier-seq}. The @code{__thread} specifier may
10519 be used alone, or immediately following the @code{extern} or
10520 @code{static} specifiers. [@dots{}]
10523 Add after paragraph 5
10526 The @code{__thread} specifier can be applied only to the names of objects
10527 and to anonymous unions.
10533 Add after paragraph 6
10536 Non-@code{static} members shall not be @code{__thread}.
10540 @node C++ Extensions
10541 @chapter Extensions to the C++ Language
10542 @cindex extensions, C++ language
10543 @cindex C++ language extensions
10545 The GNU compiler provides these extensions to the C++ language (and you
10546 can also use most of the C language extensions in your C++ programs). If you
10547 want to write code that checks whether these features are available, you can
10548 test for the GNU compiler the same way as for C programs: check for a
10549 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10550 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10551 Predefined Macros,cpp,The GNU C Preprocessor}).
10554 * Volatiles:: What constitutes an access to a volatile object.
10555 * Restricted Pointers:: C99 restricted pointers and references.
10556 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10557 * C++ Interface:: You can use a single C++ header file for both
10558 declarations and definitions.
10559 * Template Instantiation:: Methods for ensuring that exactly one copy of
10560 each needed template instantiation is emitted.
10561 * Bound member functions:: You can extract a function pointer to the
10562 method denoted by a @samp{->*} or @samp{.*} expression.
10563 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10564 * Namespace Association:: Strong using-directives for namespace association.
10565 * Java Exceptions:: Tweaking exception handling to work with Java.
10566 * Deprecated Features:: Things will disappear from g++.
10567 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10571 @section When is a Volatile Object Accessed?
10572 @cindex accessing volatiles
10573 @cindex volatile read
10574 @cindex volatile write
10575 @cindex volatile access
10577 Both the C and C++ standard have the concept of volatile objects. These
10578 are normally accessed by pointers and used for accessing hardware. The
10579 standards encourage compilers to refrain from optimizations concerning
10580 accesses to volatile objects. The C standard leaves it implementation
10581 defined as to what constitutes a volatile access. The C++ standard omits
10582 to specify this, except to say that C++ should behave in a similar manner
10583 to C with respect to volatiles, where possible. The minimum either
10584 standard specifies is that at a sequence point all previous accesses to
10585 volatile objects have stabilized and no subsequent accesses have
10586 occurred. Thus an implementation is free to reorder and combine
10587 volatile accesses which occur between sequence points, but cannot do so
10588 for accesses across a sequence point. The use of volatiles does not
10589 allow you to violate the restriction on updating objects multiple times
10590 within a sequence point.
10592 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10594 The behavior differs slightly between C and C++ in the non-obvious cases:
10597 volatile int *src = @var{somevalue};
10601 With C, such expressions are rvalues, and GCC interprets this either as a
10602 read of the volatile object being pointed to or only as request to evaluate
10603 the side-effects. The C++ standard specifies that such expressions do not
10604 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10605 object may be incomplete. The C++ standard does not specify explicitly
10606 that it is this lvalue to rvalue conversion which may be responsible for
10607 causing an access. However, there is reason to believe that it is,
10608 because otherwise certain simple expressions become undefined. However,
10609 because it would surprise most programmers, G++ treats dereferencing a
10610 pointer to volatile object of complete type when the value is unused as
10611 GCC would do for an equivalent type in C. When the object has incomplete
10612 type, G++ issues a warning; if you wish to force an error, you must
10613 force a conversion to rvalue with, for instance, a static cast.
10615 When using a reference to volatile, G++ does not treat equivalent
10616 expressions as accesses to volatiles, but instead issues a warning that
10617 no volatile is accessed. The rationale for this is that otherwise it
10618 becomes difficult to determine where volatile access occur, and not
10619 possible to ignore the return value from functions returning volatile
10620 references. Again, if you wish to force a read, cast the reference to
10623 @node Restricted Pointers
10624 @section Restricting Pointer Aliasing
10625 @cindex restricted pointers
10626 @cindex restricted references
10627 @cindex restricted this pointer
10629 As with the C front end, G++ understands the C99 feature of restricted pointers,
10630 specified with the @code{__restrict__}, or @code{__restrict} type
10631 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10632 language flag, @code{restrict} is not a keyword in C++.
10634 In addition to allowing restricted pointers, you can specify restricted
10635 references, which indicate that the reference is not aliased in the local
10639 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10646 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10647 @var{rref} refers to a (different) unaliased integer.
10649 You may also specify whether a member function's @var{this} pointer is
10650 unaliased by using @code{__restrict__} as a member function qualifier.
10653 void T::fn () __restrict__
10660 Within the body of @code{T::fn}, @var{this} will have the effective
10661 definition @code{T *__restrict__ const this}. Notice that the
10662 interpretation of a @code{__restrict__} member function qualifier is
10663 different to that of @code{const} or @code{volatile} qualifier, in that it
10664 is applied to the pointer rather than the object. This is consistent with
10665 other compilers which implement restricted pointers.
10667 As with all outermost parameter qualifiers, @code{__restrict__} is
10668 ignored in function definition matching. This means you only need to
10669 specify @code{__restrict__} in a function definition, rather than
10670 in a function prototype as well.
10672 @node Vague Linkage
10673 @section Vague Linkage
10674 @cindex vague linkage
10676 There are several constructs in C++ which require space in the object
10677 file but are not clearly tied to a single translation unit. We say that
10678 these constructs have ``vague linkage''. Typically such constructs are
10679 emitted wherever they are needed, though sometimes we can be more
10683 @item Inline Functions
10684 Inline functions are typically defined in a header file which can be
10685 included in many different compilations. Hopefully they can usually be
10686 inlined, but sometimes an out-of-line copy is necessary, if the address
10687 of the function is taken or if inlining fails. In general, we emit an
10688 out-of-line copy in all translation units where one is needed. As an
10689 exception, we only emit inline virtual functions with the vtable, since
10690 it will always require a copy.
10692 Local static variables and string constants used in an inline function
10693 are also considered to have vague linkage, since they must be shared
10694 between all inlined and out-of-line instances of the function.
10698 C++ virtual functions are implemented in most compilers using a lookup
10699 table, known as a vtable. The vtable contains pointers to the virtual
10700 functions provided by a class, and each object of the class contains a
10701 pointer to its vtable (or vtables, in some multiple-inheritance
10702 situations). If the class declares any non-inline, non-pure virtual
10703 functions, the first one is chosen as the ``key method'' for the class,
10704 and the vtable is only emitted in the translation unit where the key
10707 @emph{Note:} If the chosen key method is later defined as inline, the
10708 vtable will still be emitted in every translation unit which defines it.
10709 Make sure that any inline virtuals are declared inline in the class
10710 body, even if they are not defined there.
10712 @item type_info objects
10715 C++ requires information about types to be written out in order to
10716 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10717 For polymorphic classes (classes with virtual functions), the type_info
10718 object is written out along with the vtable so that @samp{dynamic_cast}
10719 can determine the dynamic type of a class object at runtime. For all
10720 other types, we write out the type_info object when it is used: when
10721 applying @samp{typeid} to an expression, throwing an object, or
10722 referring to a type in a catch clause or exception specification.
10724 @item Template Instantiations
10725 Most everything in this section also applies to template instantiations,
10726 but there are other options as well.
10727 @xref{Template Instantiation,,Where's the Template?}.
10731 When used with GNU ld version 2.8 or later on an ELF system such as
10732 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10733 these constructs will be discarded at link time. This is known as
10736 On targets that don't support COMDAT, but do support weak symbols, GCC
10737 will use them. This way one copy will override all the others, but
10738 the unused copies will still take up space in the executable.
10740 For targets which do not support either COMDAT or weak symbols,
10741 most entities with vague linkage will be emitted as local symbols to
10742 avoid duplicate definition errors from the linker. This will not happen
10743 for local statics in inlines, however, as having multiple copies will
10744 almost certainly break things.
10746 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10747 another way to control placement of these constructs.
10749 @node C++ Interface
10750 @section #pragma interface and implementation
10752 @cindex interface and implementation headers, C++
10753 @cindex C++ interface and implementation headers
10754 @cindex pragmas, interface and implementation
10756 @code{#pragma interface} and @code{#pragma implementation} provide the
10757 user with a way of explicitly directing the compiler to emit entities
10758 with vague linkage (and debugging information) in a particular
10761 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10762 most cases, because of COMDAT support and the ``key method'' heuristic
10763 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10764 program to grow due to unnecessary out-of-line copies of inline
10765 functions. Currently (3.4) the only benefit of these
10766 @code{#pragma}s is reduced duplication of debugging information, and
10767 that should be addressed soon on DWARF 2 targets with the use of
10771 @item #pragma interface
10772 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10773 @kindex #pragma interface
10774 Use this directive in @emph{header files} that define object classes, to save
10775 space in most of the object files that use those classes. Normally,
10776 local copies of certain information (backup copies of inline member
10777 functions, debugging information, and the internal tables that implement
10778 virtual functions) must be kept in each object file that includes class
10779 definitions. You can use this pragma to avoid such duplication. When a
10780 header file containing @samp{#pragma interface} is included in a
10781 compilation, this auxiliary information will not be generated (unless
10782 the main input source file itself uses @samp{#pragma implementation}).
10783 Instead, the object files will contain references to be resolved at link
10786 The second form of this directive is useful for the case where you have
10787 multiple headers with the same name in different directories. If you
10788 use this form, you must specify the same string to @samp{#pragma
10791 @item #pragma implementation
10792 @itemx #pragma implementation "@var{objects}.h"
10793 @kindex #pragma implementation
10794 Use this pragma in a @emph{main input file}, when you want full output from
10795 included header files to be generated (and made globally visible). The
10796 included header file, in turn, should use @samp{#pragma interface}.
10797 Backup copies of inline member functions, debugging information, and the
10798 internal tables used to implement virtual functions are all generated in
10799 implementation files.
10801 @cindex implied @code{#pragma implementation}
10802 @cindex @code{#pragma implementation}, implied
10803 @cindex naming convention, implementation headers
10804 If you use @samp{#pragma implementation} with no argument, it applies to
10805 an include file with the same basename@footnote{A file's @dfn{basename}
10806 was the name stripped of all leading path information and of trailing
10807 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10808 file. For example, in @file{allclass.cc}, giving just
10809 @samp{#pragma implementation}
10810 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10812 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10813 an implementation file whenever you would include it from
10814 @file{allclass.cc} even if you never specified @samp{#pragma
10815 implementation}. This was deemed to be more trouble than it was worth,
10816 however, and disabled.
10818 Use the string argument if you want a single implementation file to
10819 include code from multiple header files. (You must also use
10820 @samp{#include} to include the header file; @samp{#pragma
10821 implementation} only specifies how to use the file---it doesn't actually
10824 There is no way to split up the contents of a single header file into
10825 multiple implementation files.
10828 @cindex inlining and C++ pragmas
10829 @cindex C++ pragmas, effect on inlining
10830 @cindex pragmas in C++, effect on inlining
10831 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10832 effect on function inlining.
10834 If you define a class in a header file marked with @samp{#pragma
10835 interface}, the effect on an inline function defined in that class is
10836 similar to an explicit @code{extern} declaration---the compiler emits
10837 no code at all to define an independent version of the function. Its
10838 definition is used only for inlining with its callers.
10840 @opindex fno-implement-inlines
10841 Conversely, when you include the same header file in a main source file
10842 that declares it as @samp{#pragma implementation}, the compiler emits
10843 code for the function itself; this defines a version of the function
10844 that can be found via pointers (or by callers compiled without
10845 inlining). If all calls to the function can be inlined, you can avoid
10846 emitting the function by compiling with @option{-fno-implement-inlines}.
10847 If any calls were not inlined, you will get linker errors.
10849 @node Template Instantiation
10850 @section Where's the Template?
10851 @cindex template instantiation
10853 C++ templates are the first language feature to require more
10854 intelligence from the environment than one usually finds on a UNIX
10855 system. Somehow the compiler and linker have to make sure that each
10856 template instance occurs exactly once in the executable if it is needed,
10857 and not at all otherwise. There are two basic approaches to this
10858 problem, which are referred to as the Borland model and the Cfront model.
10861 @item Borland model
10862 Borland C++ solved the template instantiation problem by adding the code
10863 equivalent of common blocks to their linker; the compiler emits template
10864 instances in each translation unit that uses them, and the linker
10865 collapses them together. The advantage of this model is that the linker
10866 only has to consider the object files themselves; there is no external
10867 complexity to worry about. This disadvantage is that compilation time
10868 is increased because the template code is being compiled repeatedly.
10869 Code written for this model tends to include definitions of all
10870 templates in the header file, since they must be seen to be
10874 The AT&T C++ translator, Cfront, solved the template instantiation
10875 problem by creating the notion of a template repository, an
10876 automatically maintained place where template instances are stored. A
10877 more modern version of the repository works as follows: As individual
10878 object files are built, the compiler places any template definitions and
10879 instantiations encountered in the repository. At link time, the link
10880 wrapper adds in the objects in the repository and compiles any needed
10881 instances that were not previously emitted. The advantages of this
10882 model are more optimal compilation speed and the ability to use the
10883 system linker; to implement the Borland model a compiler vendor also
10884 needs to replace the linker. The disadvantages are vastly increased
10885 complexity, and thus potential for error; for some code this can be
10886 just as transparent, but in practice it can been very difficult to build
10887 multiple programs in one directory and one program in multiple
10888 directories. Code written for this model tends to separate definitions
10889 of non-inline member templates into a separate file, which should be
10890 compiled separately.
10893 When used with GNU ld version 2.8 or later on an ELF system such as
10894 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10895 Borland model. On other systems, G++ implements neither automatic
10898 A future version of G++ will support a hybrid model whereby the compiler
10899 will emit any instantiations for which the template definition is
10900 included in the compile, and store template definitions and
10901 instantiation context information into the object file for the rest.
10902 The link wrapper will extract that information as necessary and invoke
10903 the compiler to produce the remaining instantiations. The linker will
10904 then combine duplicate instantiations.
10906 In the mean time, you have the following options for dealing with
10907 template instantiations:
10912 Compile your template-using code with @option{-frepo}. The compiler will
10913 generate files with the extension @samp{.rpo} listing all of the
10914 template instantiations used in the corresponding object files which
10915 could be instantiated there; the link wrapper, @samp{collect2}, will
10916 then update the @samp{.rpo} files to tell the compiler where to place
10917 those instantiations and rebuild any affected object files. The
10918 link-time overhead is negligible after the first pass, as the compiler
10919 will continue to place the instantiations in the same files.
10921 This is your best option for application code written for the Borland
10922 model, as it will just work. Code written for the Cfront model will
10923 need to be modified so that the template definitions are available at
10924 one or more points of instantiation; usually this is as simple as adding
10925 @code{#include <tmethods.cc>} to the end of each template header.
10927 For library code, if you want the library to provide all of the template
10928 instantiations it needs, just try to link all of its object files
10929 together; the link will fail, but cause the instantiations to be
10930 generated as a side effect. Be warned, however, that this may cause
10931 conflicts if multiple libraries try to provide the same instantiations.
10932 For greater control, use explicit instantiation as described in the next
10936 @opindex fno-implicit-templates
10937 Compile your code with @option{-fno-implicit-templates} to disable the
10938 implicit generation of template instances, and explicitly instantiate
10939 all the ones you use. This approach requires more knowledge of exactly
10940 which instances you need than do the others, but it's less
10941 mysterious and allows greater control. You can scatter the explicit
10942 instantiations throughout your program, perhaps putting them in the
10943 translation units where the instances are used or the translation units
10944 that define the templates themselves; you can put all of the explicit
10945 instantiations you need into one big file; or you can create small files
10952 template class Foo<int>;
10953 template ostream& operator <<
10954 (ostream&, const Foo<int>&);
10957 for each of the instances you need, and create a template instantiation
10958 library from those.
10960 If you are using Cfront-model code, you can probably get away with not
10961 using @option{-fno-implicit-templates} when compiling files that don't
10962 @samp{#include} the member template definitions.
10964 If you use one big file to do the instantiations, you may want to
10965 compile it without @option{-fno-implicit-templates} so you get all of the
10966 instances required by your explicit instantiations (but not by any
10967 other files) without having to specify them as well.
10969 G++ has extended the template instantiation syntax given in the ISO
10970 standard to allow forward declaration of explicit instantiations
10971 (with @code{extern}), instantiation of the compiler support data for a
10972 template class (i.e.@: the vtable) without instantiating any of its
10973 members (with @code{inline}), and instantiation of only the static data
10974 members of a template class, without the support data or member
10975 functions (with (@code{static}):
10978 extern template int max (int, int);
10979 inline template class Foo<int>;
10980 static template class Foo<int>;
10984 Do nothing. Pretend G++ does implement automatic instantiation
10985 management. Code written for the Borland model will work fine, but
10986 each translation unit will contain instances of each of the templates it
10987 uses. In a large program, this can lead to an unacceptable amount of code
10991 @node Bound member functions
10992 @section Extracting the function pointer from a bound pointer to member function
10994 @cindex pointer to member function
10995 @cindex bound pointer to member function
10997 In C++, pointer to member functions (PMFs) are implemented using a wide
10998 pointer of sorts to handle all the possible call mechanisms; the PMF
10999 needs to store information about how to adjust the @samp{this} pointer,
11000 and if the function pointed to is virtual, where to find the vtable, and
11001 where in the vtable to look for the member function. If you are using
11002 PMFs in an inner loop, you should really reconsider that decision. If
11003 that is not an option, you can extract the pointer to the function that
11004 would be called for a given object/PMF pair and call it directly inside
11005 the inner loop, to save a bit of time.
11007 Note that you will still be paying the penalty for the call through a
11008 function pointer; on most modern architectures, such a call defeats the
11009 branch prediction features of the CPU@. This is also true of normal
11010 virtual function calls.
11012 The syntax for this extension is
11016 extern int (A::*fp)();
11017 typedef int (*fptr)(A *);
11019 fptr p = (fptr)(a.*fp);
11022 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
11023 no object is needed to obtain the address of the function. They can be
11024 converted to function pointers directly:
11027 fptr p1 = (fptr)(&A::foo);
11030 @opindex Wno-pmf-conversions
11031 You must specify @option{-Wno-pmf-conversions} to use this extension.
11033 @node C++ Attributes
11034 @section C++-Specific Variable, Function, and Type Attributes
11036 Some attributes only make sense for C++ programs.
11039 @item init_priority (@var{priority})
11040 @cindex init_priority attribute
11043 In Standard C++, objects defined at namespace scope are guaranteed to be
11044 initialized in an order in strict accordance with that of their definitions
11045 @emph{in a given translation unit}. No guarantee is made for initializations
11046 across translation units. However, GNU C++ allows users to control the
11047 order of initialization of objects defined at namespace scope with the
11048 @code{init_priority} attribute by specifying a relative @var{priority},
11049 a constant integral expression currently bounded between 101 and 65535
11050 inclusive. Lower numbers indicate a higher priority.
11052 In the following example, @code{A} would normally be created before
11053 @code{B}, but the @code{init_priority} attribute has reversed that order:
11056 Some_Class A __attribute__ ((init_priority (2000)));
11057 Some_Class B __attribute__ ((init_priority (543)));
11061 Note that the particular values of @var{priority} do not matter; only their
11064 @item java_interface
11065 @cindex java_interface attribute
11067 This type attribute informs C++ that the class is a Java interface. It may
11068 only be applied to classes declared within an @code{extern "Java"} block.
11069 Calls to methods declared in this interface will be dispatched using GCJ's
11070 interface table mechanism, instead of regular virtual table dispatch.
11074 See also @xref{Namespace Association}.
11076 @node Namespace Association
11077 @section Namespace Association
11079 @strong{Caution:} The semantics of this extension are not fully
11080 defined. Users should refrain from using this extension as its
11081 semantics may change subtly over time. It is possible that this
11082 extension will be removed in future versions of G++.
11084 A using-directive with @code{__attribute ((strong))} is stronger
11085 than a normal using-directive in two ways:
11089 Templates from the used namespace can be specialized and explicitly
11090 instantiated as though they were members of the using namespace.
11093 The using namespace is considered an associated namespace of all
11094 templates in the used namespace for purposes of argument-dependent
11098 The used namespace must be nested within the using namespace so that
11099 normal unqualified lookup works properly.
11101 This is useful for composing a namespace transparently from
11102 implementation namespaces. For example:
11107 template <class T> struct A @{ @};
11109 using namespace debug __attribute ((__strong__));
11110 template <> struct A<int> @{ @}; // @r{ok to specialize}
11112 template <class T> void f (A<T>);
11117 f (std::A<float>()); // @r{lookup finds} std::f
11122 @node Java Exceptions
11123 @section Java Exceptions
11125 The Java language uses a slightly different exception handling model
11126 from C++. Normally, GNU C++ will automatically detect when you are
11127 writing C++ code that uses Java exceptions, and handle them
11128 appropriately. However, if C++ code only needs to execute destructors
11129 when Java exceptions are thrown through it, GCC will guess incorrectly.
11130 Sample problematic code is:
11133 struct S @{ ~S(); @};
11134 extern void bar(); // @r{is written in Java, and may throw exceptions}
11143 The usual effect of an incorrect guess is a link failure, complaining of
11144 a missing routine called @samp{__gxx_personality_v0}.
11146 You can inform the compiler that Java exceptions are to be used in a
11147 translation unit, irrespective of what it might think, by writing
11148 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11149 @samp{#pragma} must appear before any functions that throw or catch
11150 exceptions, or run destructors when exceptions are thrown through them.
11152 You cannot mix Java and C++ exceptions in the same translation unit. It
11153 is believed to be safe to throw a C++ exception from one file through
11154 another file compiled for the Java exception model, or vice versa, but
11155 there may be bugs in this area.
11157 @node Deprecated Features
11158 @section Deprecated Features
11160 In the past, the GNU C++ compiler was extended to experiment with new
11161 features, at a time when the C++ language was still evolving. Now that
11162 the C++ standard is complete, some of those features are superseded by
11163 superior alternatives. Using the old features might cause a warning in
11164 some cases that the feature will be dropped in the future. In other
11165 cases, the feature might be gone already.
11167 While the list below is not exhaustive, it documents some of the options
11168 that are now deprecated:
11171 @item -fexternal-templates
11172 @itemx -falt-external-templates
11173 These are two of the many ways for G++ to implement template
11174 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11175 defines how template definitions have to be organized across
11176 implementation units. G++ has an implicit instantiation mechanism that
11177 should work just fine for standard-conforming code.
11179 @item -fstrict-prototype
11180 @itemx -fno-strict-prototype
11181 Previously it was possible to use an empty prototype parameter list to
11182 indicate an unspecified number of parameters (like C), rather than no
11183 parameters, as C++ demands. This feature has been removed, except where
11184 it is required for backwards compatibility @xref{Backwards Compatibility}.
11187 G++ allows a virtual function returning @samp{void *} to be overridden
11188 by one returning a different pointer type. This extension to the
11189 covariant return type rules is now deprecated and will be removed from a
11192 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11193 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11194 and will be removed in a future version. Code using these operators
11195 should be modified to use @code{std::min} and @code{std::max} instead.
11197 The named return value extension has been deprecated, and is now
11200 The use of initializer lists with new expressions has been deprecated,
11201 and is now removed from G++.
11203 Floating and complex non-type template parameters have been deprecated,
11204 and are now removed from G++.
11206 The implicit typename extension has been deprecated and is now
11209 The use of default arguments in function pointers, function typedefs
11210 and other places where they are not permitted by the standard is
11211 deprecated and will be removed from a future version of G++.
11213 G++ allows floating-point literals to appear in integral constant expressions,
11214 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11215 This extension is deprecated and will be removed from a future version.
11217 G++ allows static data members of const floating-point type to be declared
11218 with an initializer in a class definition. The standard only allows
11219 initializers for static members of const integral types and const
11220 enumeration types so this extension has been deprecated and will be removed
11221 from a future version.
11223 @node Backwards Compatibility
11224 @section Backwards Compatibility
11225 @cindex Backwards Compatibility
11226 @cindex ARM [Annotated C++ Reference Manual]
11228 Now that there is a definitive ISO standard C++, G++ has a specification
11229 to adhere to. The C++ language evolved over time, and features that
11230 used to be acceptable in previous drafts of the standard, such as the ARM
11231 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11232 compilation of C++ written to such drafts, G++ contains some backwards
11233 compatibilities. @emph{All such backwards compatibility features are
11234 liable to disappear in future versions of G++.} They should be considered
11235 deprecated @xref{Deprecated Features}.
11239 If a variable is declared at for scope, it used to remain in scope until
11240 the end of the scope which contained the for statement (rather than just
11241 within the for scope). G++ retains this, but issues a warning, if such a
11242 variable is accessed outside the for scope.
11244 @item Implicit C language
11245 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11246 scope to set the language. On such systems, all header files are
11247 implicitly scoped inside a C language scope. Also, an empty prototype
11248 @code{()} will be treated as an unspecified number of arguments, rather
11249 than no arguments, as C++ demands.