1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C. Most of them are also available
20 in C++. @xref{C++ Extensions,,Extensions to the C++ Language}, for
21 extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Decimal Float:: Decimal Floating Types.
37 * Hex Floats:: Hexadecimal floating-point constants.
38 * Zero Length:: Zero-length arrays.
39 * Variable Length:: Arrays whose length is computed at run time.
40 * Empty Structures:: Structures with no members.
41 * Variadic Macros:: Macros with a variable number of arguments.
42 * Escaped Newlines:: Slightly looser rules for escaped newlines.
43 * Subscripting:: Any array can be subscripted, even if not an lvalue.
44 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
45 * Initializers:: Non-constant initializers.
46 * Compound Literals:: Compound literals give structures, unions
48 * Designated Inits:: Labeling elements of initializers.
49 * Cast to Union:: Casting to union type from any member of the union.
50 * Case Ranges:: `case 1 ... 9' and such.
51 * Mixed Declarations:: Mixing declarations and code.
52 * Function Attributes:: Declaring that functions have no side effects,
53 or that they can never return.
54 * Attribute Syntax:: Formal syntax for attributes.
55 * Function Prototypes:: Prototype declarations and old-style definitions.
56 * C++ Comments:: C++ comments are recognized.
57 * Dollar Signs:: Dollar sign is allowed in identifiers.
58 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Alignment:: Inquiring about the alignment of a type or variable.
62 * Inline:: Defining inline functions (as fast as macros).
63 * Extended Asm:: Assembler instructions with C expressions as operands.
64 (With them you can define ``built-in'' functions.)
65 * Constraints:: Constraints for asm operands
66 * Asm Labels:: Specifying the assembler name to use for a C symbol.
67 * Explicit Reg Vars:: Defining variables residing in specified registers.
68 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
69 * Incomplete Enums:: @code{enum foo;}, with details to follow.
70 * Function Names:: Printable strings which are the name of the current
72 * Return Address:: Getting the return or frame address of a function.
73 * Vector Extensions:: Using vector instructions through built-in functions.
74 * Offsetof:: Special syntax for implementing @code{offsetof}.
75 * Atomic Builtins:: Built-in functions for atomic memory access.
76 * Object Size Checking:: Built-in functions for limited buffer overflow
78 * Other Builtins:: Other built-in functions.
79 * Target Builtins:: Built-in functions specific to particular targets.
80 * Target Format Checks:: Format checks specific to particular targets.
81 * Pragmas:: Pragmas accepted by GCC.
82 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
83 * Thread-Local:: Per-thread variables.
87 @section Statements and Declarations in Expressions
88 @cindex statements inside expressions
89 @cindex declarations inside expressions
90 @cindex expressions containing statements
91 @cindex macros, statements in expressions
93 @c the above section title wrapped and causes an underfull hbox.. i
94 @c changed it from "within" to "in". --mew 4feb93
95 A compound statement enclosed in parentheses may appear as an expression
96 in GNU C@. This allows you to use loops, switches, and local variables
99 Recall that a compound statement is a sequence of statements surrounded
100 by braces; in this construct, parentheses go around the braces. For
104 (@{ int y = foo (); int z;
111 is a valid (though slightly more complex than necessary) expression
112 for the absolute value of @code{foo ()}.
114 The last thing in the compound statement should be an expression
115 followed by a semicolon; the value of this subexpression serves as the
116 value of the entire construct. (If you use some other kind of statement
117 last within the braces, the construct has type @code{void}, and thus
118 effectively no value.)
120 This feature is especially useful in making macro definitions ``safe'' (so
121 that they evaluate each operand exactly once). For example, the
122 ``maximum'' function is commonly defined as a macro in standard C as
126 #define max(a,b) ((a) > (b) ? (a) : (b))
130 @cindex side effects, macro argument
131 But this definition computes either @var{a} or @var{b} twice, with bad
132 results if the operand has side effects. In GNU C, if you know the
133 type of the operands (here taken as @code{int}), you can define
134 the macro safely as follows:
137 #define maxint(a,b) \
138 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
141 Embedded statements are not allowed in constant expressions, such as
142 the value of an enumeration constant, the width of a bit-field, or
143 the initial value of a static variable.
145 If you don't know the type of the operand, you can still do this, but you
146 must use @code{typeof} (@pxref{Typeof}).
148 In G++, the result value of a statement expression undergoes array and
149 function pointer decay, and is returned by value to the enclosing
150 expression. For instance, if @code{A} is a class, then
159 will construct a temporary @code{A} object to hold the result of the
160 statement expression, and that will be used to invoke @code{Foo}.
161 Therefore the @code{this} pointer observed by @code{Foo} will not be the
164 Any temporaries created within a statement within a statement expression
165 will be destroyed at the statement's end. This makes statement
166 expressions inside macros slightly different from function calls. In
167 the latter case temporaries introduced during argument evaluation will
168 be destroyed at the end of the statement that includes the function
169 call. In the statement expression case they will be destroyed during
170 the statement expression. For instance,
173 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
174 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
184 will have different places where temporaries are destroyed. For the
185 @code{macro} case, the temporary @code{X} will be destroyed just after
186 the initialization of @code{b}. In the @code{function} case that
187 temporary will be destroyed when the function returns.
189 These considerations mean that it is probably a bad idea to use
190 statement-expressions of this form in header files that are designed to
191 work with C++. (Note that some versions of the GNU C Library contained
192 header files using statement-expression that lead to precisely this
195 Jumping into a statement expression with @code{goto} or using a
196 @code{switch} statement outside the statement expression with a
197 @code{case} or @code{default} label inside the statement expression is
198 not permitted. Jumping into a statement expression with a computed
199 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
200 Jumping out of a statement expression is permitted, but if the
201 statement expression is part of a larger expression then it is
202 unspecified which other subexpressions of that expression have been
203 evaluated except where the language definition requires certain
204 subexpressions to be evaluated before or after the statement
205 expression. In any case, as with a function call the evaluation of a
206 statement expression is not interleaved with the evaluation of other
207 parts of the containing expression. For example,
210 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
214 will call @code{foo} and @code{bar1} and will not call @code{baz} but
215 may or may not call @code{bar2}. If @code{bar2} is called, it will be
216 called after @code{foo} and before @code{bar1}
219 @section Locally Declared Labels
221 @cindex macros, local labels
223 GCC allows you to declare @dfn{local labels} in any nested block
224 scope. A local label is just like an ordinary label, but you can
225 only reference it (with a @code{goto} statement, or by taking its
226 address) within the block in which it was declared.
228 A local label declaration looks like this:
231 __label__ @var{label};
238 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
241 Local label declarations must come at the beginning of the block,
242 before any ordinary declarations or statements.
244 The label declaration defines the label @emph{name}, but does not define
245 the label itself. You must do this in the usual way, with
246 @code{@var{label}:}, within the statements of the statement expression.
248 The local label feature is useful for complex macros. If a macro
249 contains nested loops, a @code{goto} can be useful for breaking out of
250 them. However, an ordinary label whose scope is the whole function
251 cannot be used: if the macro can be expanded several times in one
252 function, the label will be multiply defined in that function. A
253 local label avoids this problem. For example:
256 #define SEARCH(value, array, target) \
259 typeof (target) _SEARCH_target = (target); \
260 typeof (*(array)) *_SEARCH_array = (array); \
263 for (i = 0; i < max; i++) \
264 for (j = 0; j < max; j++) \
265 if (_SEARCH_array[i][j] == _SEARCH_target) \
266 @{ (value) = i; goto found; @} \
272 This could also be written using a statement-expression:
275 #define SEARCH(array, target) \
278 typeof (target) _SEARCH_target = (target); \
279 typeof (*(array)) *_SEARCH_array = (array); \
282 for (i = 0; i < max; i++) \
283 for (j = 0; j < max; j++) \
284 if (_SEARCH_array[i][j] == _SEARCH_target) \
285 @{ value = i; goto found; @} \
292 Local label declarations also make the labels they declare visible to
293 nested functions, if there are any. @xref{Nested Functions}, for details.
295 @node Labels as Values
296 @section Labels as Values
297 @cindex labels as values
298 @cindex computed gotos
299 @cindex goto with computed label
300 @cindex address of a label
302 You can get the address of a label defined in the current function
303 (or a containing function) with the unary operator @samp{&&}. The
304 value has type @code{void *}. This value is a constant and can be used
305 wherever a constant of that type is valid. For example:
313 To use these values, you need to be able to jump to one. This is done
314 with the computed goto statement@footnote{The analogous feature in
315 Fortran is called an assigned goto, but that name seems inappropriate in
316 C, where one can do more than simply store label addresses in label
317 variables.}, @code{goto *@var{exp};}. For example,
324 Any expression of type @code{void *} is allowed.
326 One way of using these constants is in initializing a static array that
327 will serve as a jump table:
330 static void *array[] = @{ &&foo, &&bar, &&hack @};
333 Then you can select a label with indexing, like this:
340 Note that this does not check whether the subscript is in bounds---array
341 indexing in C never does that.
343 Such an array of label values serves a purpose much like that of the
344 @code{switch} statement. The @code{switch} statement is cleaner, so
345 use that rather than an array unless the problem does not fit a
346 @code{switch} statement very well.
348 Another use of label values is in an interpreter for threaded code.
349 The labels within the interpreter function can be stored in the
350 threaded code for super-fast dispatching.
352 You may not use this mechanism to jump to code in a different function.
353 If you do that, totally unpredictable things will happen. The best way to
354 avoid this is to store the label address only in automatic variables and
355 never pass it as an argument.
357 An alternate way to write the above example is
360 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
362 goto *(&&foo + array[i]);
366 This is more friendly to code living in shared libraries, as it reduces
367 the number of dynamic relocations that are needed, and by consequence,
368 allows the data to be read-only.
370 @node Nested Functions
371 @section Nested Functions
372 @cindex nested functions
373 @cindex downward funargs
376 A @dfn{nested function} is a function defined inside another function.
377 (Nested functions are not supported for GNU C++.) The nested function's
378 name is local to the block where it is defined. For example, here we
379 define a nested function named @code{square}, and call it twice:
383 foo (double a, double b)
385 double square (double z) @{ return z * z; @}
387 return square (a) + square (b);
392 The nested function can access all the variables of the containing
393 function that are visible at the point of its definition. This is
394 called @dfn{lexical scoping}. For example, here we show a nested
395 function which uses an inherited variable named @code{offset}:
399 bar (int *array, int offset, int size)
401 int access (int *array, int index)
402 @{ return array[index + offset]; @}
405 for (i = 0; i < size; i++)
406 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
411 Nested function definitions are permitted within functions in the places
412 where variable definitions are allowed; that is, in any block, mixed
413 with the other declarations and statements in the block.
415 It is possible to call the nested function from outside the scope of its
416 name by storing its address or passing the address to another function:
419 hack (int *array, int size)
421 void store (int index, int value)
422 @{ array[index] = value; @}
424 intermediate (store, size);
428 Here, the function @code{intermediate} receives the address of
429 @code{store} as an argument. If @code{intermediate} calls @code{store},
430 the arguments given to @code{store} are used to store into @code{array}.
431 But this technique works only so long as the containing function
432 (@code{hack}, in this example) does not exit.
434 If you try to call the nested function through its address after the
435 containing function has exited, all hell will break loose. If you try
436 to call it after a containing scope level has exited, and if it refers
437 to some of the variables that are no longer in scope, you may be lucky,
438 but it's not wise to take the risk. If, however, the nested function
439 does not refer to anything that has gone out of scope, you should be
442 GCC implements taking the address of a nested function using a technique
443 called @dfn{trampolines}. A paper describing them is available as
446 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
448 A nested function can jump to a label inherited from a containing
449 function, provided the label was explicitly declared in the containing
450 function (@pxref{Local Labels}). Such a jump returns instantly to the
451 containing function, exiting the nested function which did the
452 @code{goto} and any intermediate functions as well. Here is an example:
456 bar (int *array, int offset, int size)
459 int access (int *array, int index)
463 return array[index + offset];
467 for (i = 0; i < size; i++)
468 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
472 /* @r{Control comes here from @code{access}
473 if it detects an error.} */
480 A nested function always has no linkage. Declaring one with
481 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
482 before its definition, use @code{auto} (which is otherwise meaningless
483 for function declarations).
486 bar (int *array, int offset, int size)
489 auto int access (int *, int);
491 int access (int *array, int index)
495 return array[index + offset];
501 @node Constructing Calls
502 @section Constructing Function Calls
503 @cindex constructing calls
504 @cindex forwarding calls
506 Using the built-in functions described below, you can record
507 the arguments a function received, and call another function
508 with the same arguments, without knowing the number or types
511 You can also record the return value of that function call,
512 and later return that value, without knowing what data type
513 the function tried to return (as long as your caller expects
516 However, these built-in functions may interact badly with some
517 sophisticated features or other extensions of the language. It
518 is, therefore, not recommended to use them outside very simple
519 functions acting as mere forwarders for their arguments.
521 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
522 This built-in function returns a pointer to data
523 describing how to perform a call with the same arguments as were passed
524 to the current function.
526 The function saves the arg pointer register, structure value address,
527 and all registers that might be used to pass arguments to a function
528 into a block of memory allocated on the stack. Then it returns the
529 address of that block.
532 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
533 This built-in function invokes @var{function}
534 with a copy of the parameters described by @var{arguments}
537 The value of @var{arguments} should be the value returned by
538 @code{__builtin_apply_args}. The argument @var{size} specifies the size
539 of the stack argument data, in bytes.
541 This function returns a pointer to data describing
542 how to return whatever value was returned by @var{function}. The data
543 is saved in a block of memory allocated on the stack.
545 It is not always simple to compute the proper value for @var{size}. The
546 value is used by @code{__builtin_apply} to compute the amount of data
547 that should be pushed on the stack and copied from the incoming argument
551 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
552 This built-in function returns the value described by @var{result} from
553 the containing function. You should specify, for @var{result}, a value
554 returned by @code{__builtin_apply}.
558 @section Referring to a Type with @code{typeof}
561 @cindex macros, types of arguments
563 Another way to refer to the type of an expression is with @code{typeof}.
564 The syntax of using of this keyword looks like @code{sizeof}, but the
565 construct acts semantically like a type name defined with @code{typedef}.
567 There are two ways of writing the argument to @code{typeof}: with an
568 expression or with a type. Here is an example with an expression:
575 This assumes that @code{x} is an array of pointers to functions;
576 the type described is that of the values of the functions.
578 Here is an example with a typename as the argument:
585 Here the type described is that of pointers to @code{int}.
587 If you are writing a header file that must work when included in ISO C
588 programs, write @code{__typeof__} instead of @code{typeof}.
589 @xref{Alternate Keywords}.
591 A @code{typeof}-construct can be used anywhere a typedef name could be
592 used. For example, you can use it in a declaration, in a cast, or inside
593 of @code{sizeof} or @code{typeof}.
595 @code{typeof} is often useful in conjunction with the
596 statements-within-expressions feature. Here is how the two together can
597 be used to define a safe ``maximum'' macro that operates on any
598 arithmetic type and evaluates each of its arguments exactly once:
602 (@{ typeof (a) _a = (a); \
603 typeof (b) _b = (b); \
604 _a > _b ? _a : _b; @})
607 @cindex underscores in variables in macros
608 @cindex @samp{_} in variables in macros
609 @cindex local variables in macros
610 @cindex variables, local, in macros
611 @cindex macros, local variables in
613 The reason for using names that start with underscores for the local
614 variables is to avoid conflicts with variable names that occur within the
615 expressions that are substituted for @code{a} and @code{b}. Eventually we
616 hope to design a new form of declaration syntax that allows you to declare
617 variables whose scopes start only after their initializers; this will be a
618 more reliable way to prevent such conflicts.
621 Some more examples of the use of @code{typeof}:
625 This declares @code{y} with the type of what @code{x} points to.
632 This declares @code{y} as an array of such values.
639 This declares @code{y} as an array of pointers to characters:
642 typeof (typeof (char *)[4]) y;
646 It is equivalent to the following traditional C declaration:
652 To see the meaning of the declaration using @code{typeof}, and why it
653 might be a useful way to write, rewrite it with these macros:
656 #define pointer(T) typeof(T *)
657 #define array(T, N) typeof(T [N])
661 Now the declaration can be rewritten this way:
664 array (pointer (char), 4) y;
668 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
669 pointers to @code{char}.
672 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
673 a more limited extension which permitted one to write
676 typedef @var{T} = @var{expr};
680 with the effect of declaring @var{T} to have the type of the expression
681 @var{expr}. This extension does not work with GCC 3 (versions between
682 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
683 relies on it should be rewritten to use @code{typeof}:
686 typedef typeof(@var{expr}) @var{T};
690 This will work with all versions of GCC@.
693 @section Conditionals with Omitted Operands
694 @cindex conditional expressions, extensions
695 @cindex omitted middle-operands
696 @cindex middle-operands, omitted
697 @cindex extensions, @code{?:}
698 @cindex @code{?:} extensions
700 The middle operand in a conditional expression may be omitted. Then
701 if the first operand is nonzero, its value is the value of the conditional
704 Therefore, the expression
711 has the value of @code{x} if that is nonzero; otherwise, the value of
714 This example is perfectly equivalent to
720 @cindex side effect in ?:
721 @cindex ?: side effect
723 In this simple case, the ability to omit the middle operand is not
724 especially useful. When it becomes useful is when the first operand does,
725 or may (if it is a macro argument), contain a side effect. Then repeating
726 the operand in the middle would perform the side effect twice. Omitting
727 the middle operand uses the value already computed without the undesirable
728 effects of recomputing it.
731 @section Double-Word Integers
732 @cindex @code{long long} data types
733 @cindex double-word arithmetic
734 @cindex multiprecision arithmetic
735 @cindex @code{LL} integer suffix
736 @cindex @code{ULL} integer suffix
738 ISO C99 supports data types for integers that are at least 64 bits wide,
739 and as an extension GCC supports them in C89 mode and in C++.
740 Simply write @code{long long int} for a signed integer, or
741 @code{unsigned long long int} for an unsigned integer. To make an
742 integer constant of type @code{long long int}, add the suffix @samp{LL}
743 to the integer. To make an integer constant of type @code{unsigned long
744 long int}, add the suffix @samp{ULL} to the integer.
746 You can use these types in arithmetic like any other integer types.
747 Addition, subtraction, and bitwise boolean operations on these types
748 are open-coded on all types of machines. Multiplication is open-coded
749 if the machine supports fullword-to-doubleword a widening multiply
750 instruction. Division and shifts are open-coded only on machines that
751 provide special support. The operations that are not open-coded use
752 special library routines that come with GCC@.
754 There may be pitfalls when you use @code{long long} types for function
755 arguments, unless you declare function prototypes. If a function
756 expects type @code{int} for its argument, and you pass a value of type
757 @code{long long int}, confusion will result because the caller and the
758 subroutine will disagree about the number of bytes for the argument.
759 Likewise, if the function expects @code{long long int} and you pass
760 @code{int}. The best way to avoid such problems is to use prototypes.
763 @section Complex Numbers
764 @cindex complex numbers
765 @cindex @code{_Complex} keyword
766 @cindex @code{__complex__} keyword
768 ISO C99 supports complex floating data types, and as an extension GCC
769 supports them in C89 mode and in C++, and supports complex integer data
770 types which are not part of ISO C99. You can declare complex types
771 using the keyword @code{_Complex}. As an extension, the older GNU
772 keyword @code{__complex__} is also supported.
774 For example, @samp{_Complex double x;} declares @code{x} as a
775 variable whose real part and imaginary part are both of type
776 @code{double}. @samp{_Complex short int y;} declares @code{y} to
777 have real and imaginary parts of type @code{short int}; this is not
778 likely to be useful, but it shows that the set of complex types is
781 To write a constant with a complex data type, use the suffix @samp{i} or
782 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
783 has type @code{_Complex float} and @code{3i} has type
784 @code{_Complex int}. Such a constant always has a pure imaginary
785 value, but you can form any complex value you like by adding one to a
786 real constant. This is a GNU extension; if you have an ISO C99
787 conforming C library (such as GNU libc), and want to construct complex
788 constants of floating type, you should include @code{<complex.h>} and
789 use the macros @code{I} or @code{_Complex_I} instead.
791 @cindex @code{__real__} keyword
792 @cindex @code{__imag__} keyword
793 To extract the real part of a complex-valued expression @var{exp}, write
794 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
795 extract the imaginary part. This is a GNU extension; for values of
796 floating type, you should use the ISO C99 functions @code{crealf},
797 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
798 @code{cimagl}, declared in @code{<complex.h>} and also provided as
799 built-in functions by GCC@.
801 @cindex complex conjugation
802 The operator @samp{~} performs complex conjugation when used on a value
803 with a complex type. This is a GNU extension; for values of
804 floating type, you should use the ISO C99 functions @code{conjf},
805 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
806 provided as built-in functions by GCC@.
808 GCC can allocate complex automatic variables in a noncontiguous
809 fashion; it's even possible for the real part to be in a register while
810 the imaginary part is on the stack (or vice-versa). Only the DWARF2
811 debug info format can represent this, so use of DWARF2 is recommended.
812 If you are using the stabs debug info format, GCC describes a noncontiguous
813 complex variable as if it were two separate variables of noncomplex type.
814 If the variable's actual name is @code{foo}, the two fictitious
815 variables are named @code{foo$real} and @code{foo$imag}. You can
816 examine and set these two fictitious variables with your debugger.
819 @section Decimal Floating Types
820 @cindex decimal floating types
821 @cindex @code{_Decimal32} data type
822 @cindex @code{_Decimal64} data type
823 @cindex @code{_Decimal128} data type
824 @cindex @code{df} integer suffix
825 @cindex @code{dd} integer suffix
826 @cindex @code{dl} integer suffix
827 @cindex @code{DF} integer suffix
828 @cindex @code{DD} integer suffix
829 @cindex @code{DL} integer suffix
831 As an extension, the GNU C compiler supports decimal floating types as
832 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
833 floating types in GCC will evolve as the draft technical report changes.
834 Calling conventions for any target might also change. Not all targets
835 support decimal floating types.
837 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
838 @code{_Decimal128}. They use a radix of ten, unlike the floating types
839 @code{float}, @code{double}, and @code{long double} whose radix is not
840 specified by the C standard but is usually two.
842 Support for decimal floating types includes the arithmetic operators
843 add, subtract, multiply, divide; unary arithmetic operators;
844 relational operators; equality operators; and conversions to and from
845 integer and other floating types. Use a suffix @samp{df} or
846 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
847 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
850 GCC support of decimal float as specified by the draft technical report
855 Translation time data type (TTDT) is not supported.
858 Characteristics of decimal floating types are defined in header file
859 @file{decfloat.h} rather than @file{float.h}.
862 When the value of a decimal floating type cannot be represented in the
863 integer type to which it is being converted, the result is undefined
864 rather than the result value specified by the draft technical report.
867 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
868 are supported by the DWARF2 debug information format.
874 ISO C99 supports floating-point numbers written not only in the usual
875 decimal notation, such as @code{1.55e1}, but also numbers such as
876 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
877 supports this in C89 mode (except in some cases when strictly
878 conforming) and in C++. In that format the
879 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
880 mandatory. The exponent is a decimal number that indicates the power of
881 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
888 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
889 is the same as @code{1.55e1}.
891 Unlike for floating-point numbers in the decimal notation the exponent
892 is always required in the hexadecimal notation. Otherwise the compiler
893 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
894 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
895 extension for floating-point constants of type @code{float}.
898 @section Arrays of Length Zero
899 @cindex arrays of length zero
900 @cindex zero-length arrays
901 @cindex length-zero arrays
902 @cindex flexible array members
904 Zero-length arrays are allowed in GNU C@. They are very useful as the
905 last element of a structure which is really a header for a variable-length
914 struct line *thisline = (struct line *)
915 malloc (sizeof (struct line) + this_length);
916 thisline->length = this_length;
919 In ISO C90, you would have to give @code{contents} a length of 1, which
920 means either you waste space or complicate the argument to @code{malloc}.
922 In ISO C99, you would use a @dfn{flexible array member}, which is
923 slightly different in syntax and semantics:
927 Flexible array members are written as @code{contents[]} without
931 Flexible array members have incomplete type, and so the @code{sizeof}
932 operator may not be applied. As a quirk of the original implementation
933 of zero-length arrays, @code{sizeof} evaluates to zero.
936 Flexible array members may only appear as the last member of a
937 @code{struct} that is otherwise non-empty.
940 A structure containing a flexible array member, or a union containing
941 such a structure (possibly recursively), may not be a member of a
942 structure or an element of an array. (However, these uses are
943 permitted by GCC as extensions.)
946 GCC versions before 3.0 allowed zero-length arrays to be statically
947 initialized, as if they were flexible arrays. In addition to those
948 cases that were useful, it also allowed initializations in situations
949 that would corrupt later data. Non-empty initialization of zero-length
950 arrays is now treated like any case where there are more initializer
951 elements than the array holds, in that a suitable warning about "excess
952 elements in array" is given, and the excess elements (all of them, in
953 this case) are ignored.
955 Instead GCC allows static initialization of flexible array members.
956 This is equivalent to defining a new structure containing the original
957 structure followed by an array of sufficient size to contain the data.
958 I.e.@: in the following, @code{f1} is constructed as if it were declared
964 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
967 struct f1 f1; int data[3];
968 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
972 The convenience of this extension is that @code{f1} has the desired
973 type, eliminating the need to consistently refer to @code{f2.f1}.
975 This has symmetry with normal static arrays, in that an array of
976 unknown size is also written with @code{[]}.
978 Of course, this extension only makes sense if the extra data comes at
979 the end of a top-level object, as otherwise we would be overwriting
980 data at subsequent offsets. To avoid undue complication and confusion
981 with initialization of deeply nested arrays, we simply disallow any
982 non-empty initialization except when the structure is the top-level
986 struct foo @{ int x; int y[]; @};
987 struct bar @{ struct foo z; @};
989 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
990 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
991 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
992 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
995 @node Empty Structures
996 @section Structures With No Members
997 @cindex empty structures
998 @cindex zero-size structures
1000 GCC permits a C structure to have no members:
1007 The structure will have size zero. In C++, empty structures are part
1008 of the language. G++ treats empty structures as if they had a single
1009 member of type @code{char}.
1011 @node Variable Length
1012 @section Arrays of Variable Length
1013 @cindex variable-length arrays
1014 @cindex arrays of variable length
1017 Variable-length automatic arrays are allowed in ISO C99, and as an
1018 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1019 implementation of variable-length arrays does not yet conform in detail
1020 to the ISO C99 standard.) These arrays are
1021 declared like any other automatic arrays, but with a length that is not
1022 a constant expression. The storage is allocated at the point of
1023 declaration and deallocated when the brace-level is exited. For
1028 concat_fopen (char *s1, char *s2, char *mode)
1030 char str[strlen (s1) + strlen (s2) + 1];
1033 return fopen (str, mode);
1037 @cindex scope of a variable length array
1038 @cindex variable-length array scope
1039 @cindex deallocating variable length arrays
1040 Jumping or breaking out of the scope of the array name deallocates the
1041 storage. Jumping into the scope is not allowed; you get an error
1044 @cindex @code{alloca} vs variable-length arrays
1045 You can use the function @code{alloca} to get an effect much like
1046 variable-length arrays. The function @code{alloca} is available in
1047 many other C implementations (but not in all). On the other hand,
1048 variable-length arrays are more elegant.
1050 There are other differences between these two methods. Space allocated
1051 with @code{alloca} exists until the containing @emph{function} returns.
1052 The space for a variable-length array is deallocated as soon as the array
1053 name's scope ends. (If you use both variable-length arrays and
1054 @code{alloca} in the same function, deallocation of a variable-length array
1055 will also deallocate anything more recently allocated with @code{alloca}.)
1057 You can also use variable-length arrays as arguments to functions:
1061 tester (int len, char data[len][len])
1067 The length of an array is computed once when the storage is allocated
1068 and is remembered for the scope of the array in case you access it with
1071 If you want to pass the array first and the length afterward, you can
1072 use a forward declaration in the parameter list---another GNU extension.
1076 tester (int len; char data[len][len], int len)
1082 @cindex parameter forward declaration
1083 The @samp{int len} before the semicolon is a @dfn{parameter forward
1084 declaration}, and it serves the purpose of making the name @code{len}
1085 known when the declaration of @code{data} is parsed.
1087 You can write any number of such parameter forward declarations in the
1088 parameter list. They can be separated by commas or semicolons, but the
1089 last one must end with a semicolon, which is followed by the ``real''
1090 parameter declarations. Each forward declaration must match a ``real''
1091 declaration in parameter name and data type. ISO C99 does not support
1092 parameter forward declarations.
1094 @node Variadic Macros
1095 @section Macros with a Variable Number of Arguments.
1096 @cindex variable number of arguments
1097 @cindex macro with variable arguments
1098 @cindex rest argument (in macro)
1099 @cindex variadic macros
1101 In the ISO C standard of 1999, a macro can be declared to accept a
1102 variable number of arguments much as a function can. The syntax for
1103 defining the macro is similar to that of a function. Here is an
1107 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1110 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1111 such a macro, it represents the zero or more tokens until the closing
1112 parenthesis that ends the invocation, including any commas. This set of
1113 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1114 wherever it appears. See the CPP manual for more information.
1116 GCC has long supported variadic macros, and used a different syntax that
1117 allowed you to give a name to the variable arguments just like any other
1118 argument. Here is an example:
1121 #define debug(format, args...) fprintf (stderr, format, args)
1124 This is in all ways equivalent to the ISO C example above, but arguably
1125 more readable and descriptive.
1127 GNU CPP has two further variadic macro extensions, and permits them to
1128 be used with either of the above forms of macro definition.
1130 In standard C, you are not allowed to leave the variable argument out
1131 entirely; but you are allowed to pass an empty argument. For example,
1132 this invocation is invalid in ISO C, because there is no comma after
1139 GNU CPP permits you to completely omit the variable arguments in this
1140 way. In the above examples, the compiler would complain, though since
1141 the expansion of the macro still has the extra comma after the format
1144 To help solve this problem, CPP behaves specially for variable arguments
1145 used with the token paste operator, @samp{##}. If instead you write
1148 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1151 and if the variable arguments are omitted or empty, the @samp{##}
1152 operator causes the preprocessor to remove the comma before it. If you
1153 do provide some variable arguments in your macro invocation, GNU CPP
1154 does not complain about the paste operation and instead places the
1155 variable arguments after the comma. Just like any other pasted macro
1156 argument, these arguments are not macro expanded.
1158 @node Escaped Newlines
1159 @section Slightly Looser Rules for Escaped Newlines
1160 @cindex escaped newlines
1161 @cindex newlines (escaped)
1163 Recently, the preprocessor has relaxed its treatment of escaped
1164 newlines. Previously, the newline had to immediately follow a
1165 backslash. The current implementation allows whitespace in the form
1166 of spaces, horizontal and vertical tabs, and form feeds between the
1167 backslash and the subsequent newline. The preprocessor issues a
1168 warning, but treats it as a valid escaped newline and combines the two
1169 lines to form a single logical line. This works within comments and
1170 tokens, as well as between tokens. Comments are @emph{not} treated as
1171 whitespace for the purposes of this relaxation, since they have not
1172 yet been replaced with spaces.
1175 @section Non-Lvalue Arrays May Have Subscripts
1176 @cindex subscripting
1177 @cindex arrays, non-lvalue
1179 @cindex subscripting and function values
1180 In ISO C99, arrays that are not lvalues still decay to pointers, and
1181 may be subscripted, although they may not be modified or used after
1182 the next sequence point and the unary @samp{&} operator may not be
1183 applied to them. As an extension, GCC allows such arrays to be
1184 subscripted in C89 mode, though otherwise they do not decay to
1185 pointers outside C99 mode. For example,
1186 this is valid in GNU C though not valid in C89:
1190 struct foo @{int a[4];@};
1196 return f().a[index];
1202 @section Arithmetic on @code{void}- and Function-Pointers
1203 @cindex void pointers, arithmetic
1204 @cindex void, size of pointer to
1205 @cindex function pointers, arithmetic
1206 @cindex function, size of pointer to
1208 In GNU C, addition and subtraction operations are supported on pointers to
1209 @code{void} and on pointers to functions. This is done by treating the
1210 size of a @code{void} or of a function as 1.
1212 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1213 and on function types, and returns 1.
1215 @opindex Wpointer-arith
1216 The option @option{-Wpointer-arith} requests a warning if these extensions
1220 @section Non-Constant Initializers
1221 @cindex initializers, non-constant
1222 @cindex non-constant initializers
1224 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1225 automatic variable are not required to be constant expressions in GNU C@.
1226 Here is an example of an initializer with run-time varying elements:
1229 foo (float f, float g)
1231 float beat_freqs[2] = @{ f-g, f+g @};
1236 @node Compound Literals
1237 @section Compound Literals
1238 @cindex constructor expressions
1239 @cindex initializations in expressions
1240 @cindex structures, constructor expression
1241 @cindex expressions, constructor
1242 @cindex compound literals
1243 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1245 ISO C99 supports compound literals. A compound literal looks like
1246 a cast containing an initializer. Its value is an object of the
1247 type specified in the cast, containing the elements specified in
1248 the initializer; it is an lvalue. As an extension, GCC supports
1249 compound literals in C89 mode and in C++.
1251 Usually, the specified type is a structure. Assume that
1252 @code{struct foo} and @code{structure} are declared as shown:
1255 struct foo @{int a; char b[2];@} structure;
1259 Here is an example of constructing a @code{struct foo} with a compound literal:
1262 structure = ((struct foo) @{x + y, 'a', 0@});
1266 This is equivalent to writing the following:
1270 struct foo temp = @{x + y, 'a', 0@};
1275 You can also construct an array. If all the elements of the compound literal
1276 are (made up of) simple constant expressions, suitable for use in
1277 initializers of objects of static storage duration, then the compound
1278 literal can be coerced to a pointer to its first element and used in
1279 such an initializer, as shown here:
1282 char **foo = (char *[]) @{ "x", "y", "z" @};
1285 Compound literals for scalar types and union types are is
1286 also allowed, but then the compound literal is equivalent
1289 As a GNU extension, GCC allows initialization of objects with static storage
1290 duration by compound literals (which is not possible in ISO C99, because
1291 the initializer is not a constant).
1292 It is handled as if the object was initialized only with the bracket
1293 enclosed list if the types of the compound literal and the object match.
1294 The initializer list of the compound literal must be constant.
1295 If the object being initialized has array type of unknown size, the size is
1296 determined by compound literal size.
1299 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1300 static int y[] = (int []) @{1, 2, 3@};
1301 static int z[] = (int [3]) @{1@};
1305 The above lines are equivalent to the following:
1307 static struct foo x = @{1, 'a', 'b'@};
1308 static int y[] = @{1, 2, 3@};
1309 static int z[] = @{1, 0, 0@};
1312 @node Designated Inits
1313 @section Designated Initializers
1314 @cindex initializers with labeled elements
1315 @cindex labeled elements in initializers
1316 @cindex case labels in initializers
1317 @cindex designated initializers
1319 Standard C89 requires the elements of an initializer to appear in a fixed
1320 order, the same as the order of the elements in the array or structure
1323 In ISO C99 you can give the elements in any order, specifying the array
1324 indices or structure field names they apply to, and GNU C allows this as
1325 an extension in C89 mode as well. This extension is not
1326 implemented in GNU C++.
1328 To specify an array index, write
1329 @samp{[@var{index}] =} before the element value. For example,
1332 int a[6] = @{ [4] = 29, [2] = 15 @};
1339 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1343 The index values must be constant expressions, even if the array being
1344 initialized is automatic.
1346 An alternative syntax for this which has been obsolete since GCC 2.5 but
1347 GCC still accepts is to write @samp{[@var{index}]} before the element
1348 value, with no @samp{=}.
1350 To initialize a range of elements to the same value, write
1351 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1352 extension. For example,
1355 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1359 If the value in it has side-effects, the side-effects will happen only once,
1360 not for each initialized field by the range initializer.
1363 Note that the length of the array is the highest value specified
1366 In a structure initializer, specify the name of a field to initialize
1367 with @samp{.@var{fieldname} =} before the element value. For example,
1368 given the following structure,
1371 struct point @{ int x, y; @};
1375 the following initialization
1378 struct point p = @{ .y = yvalue, .x = xvalue @};
1385 struct point p = @{ xvalue, yvalue @};
1388 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1389 @samp{@var{fieldname}:}, as shown here:
1392 struct point p = @{ y: yvalue, x: xvalue @};
1396 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1397 @dfn{designator}. You can also use a designator (or the obsolete colon
1398 syntax) when initializing a union, to specify which element of the union
1399 should be used. For example,
1402 union foo @{ int i; double d; @};
1404 union foo f = @{ .d = 4 @};
1408 will convert 4 to a @code{double} to store it in the union using
1409 the second element. By contrast, casting 4 to type @code{union foo}
1410 would store it into the union as the integer @code{i}, since it is
1411 an integer. (@xref{Cast to Union}.)
1413 You can combine this technique of naming elements with ordinary C
1414 initialization of successive elements. Each initializer element that
1415 does not have a designator applies to the next consecutive element of the
1416 array or structure. For example,
1419 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1426 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1429 Labeling the elements of an array initializer is especially useful
1430 when the indices are characters or belong to an @code{enum} type.
1435 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1436 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1439 @cindex designator lists
1440 You can also write a series of @samp{.@var{fieldname}} and
1441 @samp{[@var{index}]} designators before an @samp{=} to specify a
1442 nested subobject to initialize; the list is taken relative to the
1443 subobject corresponding to the closest surrounding brace pair. For
1444 example, with the @samp{struct point} declaration above:
1447 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1451 If the same field is initialized multiple times, it will have value from
1452 the last initialization. If any such overridden initialization has
1453 side-effect, it is unspecified whether the side-effect happens or not.
1454 Currently, GCC will discard them and issue a warning.
1457 @section Case Ranges
1459 @cindex ranges in case statements
1461 You can specify a range of consecutive values in a single @code{case} label,
1465 case @var{low} ... @var{high}:
1469 This has the same effect as the proper number of individual @code{case}
1470 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1472 This feature is especially useful for ranges of ASCII character codes:
1478 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1479 it may be parsed wrong when you use it with integer values. For example,
1494 @section Cast to a Union Type
1495 @cindex cast to a union
1496 @cindex union, casting to a
1498 A cast to union type is similar to other casts, except that the type
1499 specified is a union type. You can specify the type either with
1500 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1501 a constructor though, not a cast, and hence does not yield an lvalue like
1502 normal casts. (@xref{Compound Literals}.)
1504 The types that may be cast to the union type are those of the members
1505 of the union. Thus, given the following union and variables:
1508 union foo @{ int i; double d; @};
1514 both @code{x} and @code{y} can be cast to type @code{union foo}.
1516 Using the cast as the right-hand side of an assignment to a variable of
1517 union type is equivalent to storing in a member of the union:
1522 u = (union foo) x @equiv{} u.i = x
1523 u = (union foo) y @equiv{} u.d = y
1526 You can also use the union cast as a function argument:
1529 void hack (union foo);
1531 hack ((union foo) x);
1534 @node Mixed Declarations
1535 @section Mixed Declarations and Code
1536 @cindex mixed declarations and code
1537 @cindex declarations, mixed with code
1538 @cindex code, mixed with declarations
1540 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1541 within compound statements. As an extension, GCC also allows this in
1542 C89 mode. For example, you could do:
1551 Each identifier is visible from where it is declared until the end of
1552 the enclosing block.
1554 @node Function Attributes
1555 @section Declaring Attributes of Functions
1556 @cindex function attributes
1557 @cindex declaring attributes of functions
1558 @cindex functions that never return
1559 @cindex functions that return more than once
1560 @cindex functions that have no side effects
1561 @cindex functions in arbitrary sections
1562 @cindex functions that behave like malloc
1563 @cindex @code{volatile} applied to function
1564 @cindex @code{const} applied to function
1565 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1566 @cindex functions with non-null pointer arguments
1567 @cindex functions that are passed arguments in registers on the 386
1568 @cindex functions that pop the argument stack on the 386
1569 @cindex functions that do not pop the argument stack on the 386
1571 In GNU C, you declare certain things about functions called in your program
1572 which help the compiler optimize function calls and check your code more
1575 The keyword @code{__attribute__} allows you to specify special
1576 attributes when making a declaration. This keyword is followed by an
1577 attribute specification inside double parentheses. The following
1578 attributes are currently defined for functions on all targets:
1579 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1580 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1581 @code{format}, @code{format_arg}, @code{no_instrument_function},
1582 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1583 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1584 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1585 @code{gnu_inline} and @code{externally_visible}. Several other
1586 attributes are defined for functions on particular target systems. Other
1587 attributes, including @code{section} are supported for variables declarations
1588 (@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1590 You may also specify attributes with @samp{__} preceding and following
1591 each keyword. This allows you to use them in header files without
1592 being concerned about a possible macro of the same name. For example,
1593 you may use @code{__noreturn__} instead of @code{noreturn}.
1595 @xref{Attribute Syntax}, for details of the exact syntax for using
1599 @c Keep this table alphabetized by attribute name. Treat _ as space.
1601 @item alias ("@var{target}")
1602 @cindex @code{alias} attribute
1603 The @code{alias} attribute causes the declaration to be emitted as an
1604 alias for another symbol, which must be specified. For instance,
1607 void __f () @{ /* @r{Do something.} */; @}
1608 void f () __attribute__ ((weak, alias ("__f")));
1611 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1612 mangled name for the target must be used. It is an error if @samp{__f}
1613 is not defined in the same translation unit.
1615 Not all target machines support this attribute.
1618 @cindex @code{always_inline} function attribute
1619 Generally, functions are not inlined unless optimization is specified.
1620 For functions declared inline, this attribute inlines the function even
1621 if no optimization level was specified.
1624 @cindex @code{gnu_inline} function attribute
1625 This attribute should be used with a function which is also declared
1626 with the @code{inline} keyword. It directs GCC to treat the function
1627 as if it were defined in gnu89 mode even when compiling in C99 or
1630 If the function is declared @code{extern}, then this definition of the
1631 function is used only for inlining. In no case is the function
1632 compiled as a standalone function, not even if you take its address
1633 explicitly. Such an address becomes an external reference, as if you
1634 had only declared the function, and had not defined it. This has
1635 almost the effect of a macro. The way to use this is to put a
1636 function definition in a header file with this attribute, and put
1637 another copy of the function, without @code{extern}, in a library
1638 file. The definition in the header file will cause most calls to the
1639 function to be inlined. If any uses of the function remain, they will
1640 refer to the single copy in the library. Note that the two
1641 definitions of the functions need not be precisely the same, although
1642 if they do not have the same effect your program may behave oddly.
1644 If the function is neither @code{extern} nor @code{static}, then the
1645 function is compiled as a standalone function, as well as being
1646 inlined where possible.
1648 This is how GCC traditionally handled functions declared
1649 @code{inline}. Since ISO C99 specifies a different semantics for
1650 @code{inline}, this function attribute is provided as a transition
1651 measure and as a useful feature in its own right. This attribute is
1652 available in GCC 4.1.3 and later. It is available if either of the
1653 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1654 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1655 Function is As Fast As a Macro}.
1657 Note that since the first version of GCC to support C99 inline semantics
1658 is 4.3, earlier versions of GCC which accept this attribute effectively
1659 assume that it is always present, whether or not it is given explicitly.
1660 In versions prior to 4.3, the only effect of explicitly including it is
1661 to disable warnings about using inline functions in C99 mode.
1663 @cindex @code{flatten} function attribute
1665 Generally, inlining into a function is limited. For a function marked with
1666 this attribute, every call inside this function will be inlined, if possible.
1667 Whether the function itself is considered for inlining depends on its size and
1668 the current inlining parameters. The @code{flatten} attribute only works
1669 reliably in unit-at-a-time mode.
1672 @cindex functions that do pop the argument stack on the 386
1674 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1675 assume that the calling function will pop off the stack space used to
1676 pass arguments. This is
1677 useful to override the effects of the @option{-mrtd} switch.
1680 @cindex @code{const} function attribute
1681 Many functions do not examine any values except their arguments, and
1682 have no effects except the return value. Basically this is just slightly
1683 more strict class than the @code{pure} attribute below, since function is not
1684 allowed to read global memory.
1686 @cindex pointer arguments
1687 Note that a function that has pointer arguments and examines the data
1688 pointed to must @emph{not} be declared @code{const}. Likewise, a
1689 function that calls a non-@code{const} function usually must not be
1690 @code{const}. It does not make sense for a @code{const} function to
1693 The attribute @code{const} is not implemented in GCC versions earlier
1694 than 2.5. An alternative way to declare that a function has no side
1695 effects, which works in the current version and in some older versions,
1699 typedef int intfn ();
1701 extern const intfn square;
1704 This approach does not work in GNU C++ from 2.6.0 on, since the language
1705 specifies that the @samp{const} must be attached to the return value.
1709 @cindex @code{constructor} function attribute
1710 @cindex @code{destructor} function attribute
1711 The @code{constructor} attribute causes the function to be called
1712 automatically before execution enters @code{main ()}. Similarly, the
1713 @code{destructor} attribute causes the function to be called
1714 automatically after @code{main ()} has completed or @code{exit ()} has
1715 been called. Functions with these attributes are useful for
1716 initializing data that will be used implicitly during the execution of
1720 @cindex @code{deprecated} attribute.
1721 The @code{deprecated} attribute results in a warning if the function
1722 is used anywhere in the source file. This is useful when identifying
1723 functions that are expected to be removed in a future version of a
1724 program. The warning also includes the location of the declaration
1725 of the deprecated function, to enable users to easily find further
1726 information about why the function is deprecated, or what they should
1727 do instead. Note that the warnings only occurs for uses:
1730 int old_fn () __attribute__ ((deprecated));
1732 int (*fn_ptr)() = old_fn;
1735 results in a warning on line 3 but not line 2.
1737 The @code{deprecated} attribute can also be used for variables and
1738 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1741 @cindex @code{__declspec(dllexport)}
1742 On Microsoft Windows targets and Symbian OS targets the
1743 @code{dllexport} attribute causes the compiler to provide a global
1744 pointer to a pointer in a DLL, so that it can be referenced with the
1745 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1746 name is formed by combining @code{_imp__} and the function or variable
1749 You can use @code{__declspec(dllexport)} as a synonym for
1750 @code{__attribute__ ((dllexport))} for compatibility with other
1753 On systems that support the @code{visibility} attribute, this
1754 attribute also implies ``default'' visibility, unless a
1755 @code{visibility} attribute is explicitly specified. You should avoid
1756 the use of @code{dllexport} with ``hidden'' or ``internal''
1757 visibility; in the future GCC may issue an error for those cases.
1759 Currently, the @code{dllexport} attribute is ignored for inlined
1760 functions, unless the @option{-fkeep-inline-functions} flag has been
1761 used. The attribute is also ignored for undefined symbols.
1763 When applied to C++ classes, the attribute marks defined non-inlined
1764 member functions and static data members as exports. Static consts
1765 initialized in-class are not marked unless they are also defined
1768 For Microsoft Windows targets there are alternative methods for
1769 including the symbol in the DLL's export table such as using a
1770 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1771 the @option{--export-all} linker flag.
1774 @cindex @code{__declspec(dllimport)}
1775 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1776 attribute causes the compiler to reference a function or variable via
1777 a global pointer to a pointer that is set up by the DLL exporting the
1778 symbol. The attribute implies @code{extern} storage. On Microsoft
1779 Windows targets, the pointer name is formed by combining @code{_imp__}
1780 and the function or variable name.
1782 You can use @code{__declspec(dllimport)} as a synonym for
1783 @code{__attribute__ ((dllimport))} for compatibility with other
1786 Currently, the attribute is ignored for inlined functions. If the
1787 attribute is applied to a symbol @emph{definition}, an error is reported.
1788 If a symbol previously declared @code{dllimport} is later defined, the
1789 attribute is ignored in subsequent references, and a warning is emitted.
1790 The attribute is also overridden by a subsequent declaration as
1793 When applied to C++ classes, the attribute marks non-inlined
1794 member functions and static data members as imports. However, the
1795 attribute is ignored for virtual methods to allow creation of vtables
1798 On the SH Symbian OS target the @code{dllimport} attribute also has
1799 another affect---it can cause the vtable and run-time type information
1800 for a class to be exported. This happens when the class has a
1801 dllimport'ed constructor or a non-inline, non-pure virtual function
1802 and, for either of those two conditions, the class also has a inline
1803 constructor or destructor and has a key function that is defined in
1804 the current translation unit.
1806 For Microsoft Windows based targets the use of the @code{dllimport}
1807 attribute on functions is not necessary, but provides a small
1808 performance benefit by eliminating a thunk in the DLL@. The use of the
1809 @code{dllimport} attribute on imported variables was required on older
1810 versions of the GNU linker, but can now be avoided by passing the
1811 @option{--enable-auto-import} switch to the GNU linker. As with
1812 functions, using the attribute for a variable eliminates a thunk in
1815 One drawback to using this attribute is that a pointer to a function
1816 or variable marked as @code{dllimport} cannot be used as a constant
1817 address. On Microsoft Windows targets, the attribute can be disabled
1818 for functions by setting the @option{-mnop-fun-dllimport} flag.
1821 @cindex eight bit data on the H8/300, H8/300H, and H8S
1822 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1823 variable should be placed into the eight bit data section.
1824 The compiler will generate more efficient code for certain operations
1825 on data in the eight bit data area. Note the eight bit data area is limited to
1828 You must use GAS and GLD from GNU binutils version 2.7 or later for
1829 this attribute to work correctly.
1831 @item exception_handler
1832 @cindex exception handler functions on the Blackfin processor
1833 Use this attribute on the Blackfin to indicate that the specified function
1834 is an exception handler. The compiler will generate function entry and
1835 exit sequences suitable for use in an exception handler when this
1836 attribute is present.
1839 @cindex functions which handle memory bank switching
1840 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1841 use a calling convention that takes care of switching memory banks when
1842 entering and leaving a function. This calling convention is also the
1843 default when using the @option{-mlong-calls} option.
1845 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1846 to call and return from a function.
1848 On 68HC11 the compiler will generate a sequence of instructions
1849 to invoke a board-specific routine to switch the memory bank and call the
1850 real function. The board-specific routine simulates a @code{call}.
1851 At the end of a function, it will jump to a board-specific routine
1852 instead of using @code{rts}. The board-specific return routine simulates
1856 @cindex functions that pop the argument stack on the 386
1857 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1858 pass the first argument (if of integral type) in the register ECX and
1859 the second argument (if of integral type) in the register EDX@. Subsequent
1860 and other typed arguments are passed on the stack. The called function will
1861 pop the arguments off the stack. If the number of arguments is variable all
1862 arguments are pushed on the stack.
1864 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1865 @cindex @code{format} function attribute
1867 The @code{format} attribute specifies that a function takes @code{printf},
1868 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1869 should be type-checked against a format string. For example, the
1874 my_printf (void *my_object, const char *my_format, ...)
1875 __attribute__ ((format (printf, 2, 3)));
1879 causes the compiler to check the arguments in calls to @code{my_printf}
1880 for consistency with the @code{printf} style format string argument
1883 The parameter @var{archetype} determines how the format string is
1884 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1885 or @code{strfmon}. (You can also use @code{__printf__},
1886 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1887 parameter @var{string-index} specifies which argument is the format
1888 string argument (starting from 1), while @var{first-to-check} is the
1889 number of the first argument to check against the format string. For
1890 functions where the arguments are not available to be checked (such as
1891 @code{vprintf}), specify the third parameter as zero. In this case the
1892 compiler only checks the format string for consistency. For
1893 @code{strftime} formats, the third parameter is required to be zero.
1894 Since non-static C++ methods have an implicit @code{this} argument, the
1895 arguments of such methods should be counted from two, not one, when
1896 giving values for @var{string-index} and @var{first-to-check}.
1898 In the example above, the format string (@code{my_format}) is the second
1899 argument of the function @code{my_print}, and the arguments to check
1900 start with the third argument, so the correct parameters for the format
1901 attribute are 2 and 3.
1903 @opindex ffreestanding
1904 @opindex fno-builtin
1905 The @code{format} attribute allows you to identify your own functions
1906 which take format strings as arguments, so that GCC can check the
1907 calls to these functions for errors. The compiler always (unless
1908 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1909 for the standard library functions @code{printf}, @code{fprintf},
1910 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1911 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1912 warnings are requested (using @option{-Wformat}), so there is no need to
1913 modify the header file @file{stdio.h}. In C99 mode, the functions
1914 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1915 @code{vsscanf} are also checked. Except in strictly conforming C
1916 standard modes, the X/Open function @code{strfmon} is also checked as
1917 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1918 @xref{C Dialect Options,,Options Controlling C Dialect}.
1920 The target may provide additional types of format checks.
1921 @xref{Target Format Checks,,Format Checks Specific to Particular
1924 @item format_arg (@var{string-index})
1925 @cindex @code{format_arg} function attribute
1926 @opindex Wformat-nonliteral
1927 The @code{format_arg} attribute specifies that a function takes a format
1928 string for a @code{printf}, @code{scanf}, @code{strftime} or
1929 @code{strfmon} style function and modifies it (for example, to translate
1930 it into another language), so the result can be passed to a
1931 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1932 function (with the remaining arguments to the format function the same
1933 as they would have been for the unmodified string). For example, the
1938 my_dgettext (char *my_domain, const char *my_format)
1939 __attribute__ ((format_arg (2)));
1943 causes the compiler to check the arguments in calls to a @code{printf},
1944 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1945 format string argument is a call to the @code{my_dgettext} function, for
1946 consistency with the format string argument @code{my_format}. If the
1947 @code{format_arg} attribute had not been specified, all the compiler
1948 could tell in such calls to format functions would be that the format
1949 string argument is not constant; this would generate a warning when
1950 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1951 without the attribute.
1953 The parameter @var{string-index} specifies which argument is the format
1954 string argument (starting from one). Since non-static C++ methods have
1955 an implicit @code{this} argument, the arguments of such methods should
1956 be counted from two.
1958 The @code{format-arg} attribute allows you to identify your own
1959 functions which modify format strings, so that GCC can check the
1960 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1961 type function whose operands are a call to one of your own function.
1962 The compiler always treats @code{gettext}, @code{dgettext}, and
1963 @code{dcgettext} in this manner except when strict ISO C support is
1964 requested by @option{-ansi} or an appropriate @option{-std} option, or
1965 @option{-ffreestanding} or @option{-fno-builtin}
1966 is used. @xref{C Dialect Options,,Options
1967 Controlling C Dialect}.
1969 @item function_vector
1970 @cindex calling functions through the function vector on the H8/300 processors
1971 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1972 function should be called through the function vector. Calling a
1973 function through the function vector will reduce code size, however;
1974 the function vector has a limited size (maximum 128 entries on the H8/300
1975 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1977 You must use GAS and GLD from GNU binutils version 2.7 or later for
1978 this attribute to work correctly.
1981 @cindex interrupt handler functions
1982 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
1983 ports to indicate that the specified function is an interrupt handler.
1984 The compiler will generate function entry and exit sequences suitable
1985 for use in an interrupt handler when this attribute is present.
1987 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1988 SH processors can be specified via the @code{interrupt_handler} attribute.
1990 Note, on the AVR, interrupts will be enabled inside the function.
1992 Note, for the ARM, you can specify the kind of interrupt to be handled by
1993 adding an optional parameter to the interrupt attribute like this:
1996 void f () __attribute__ ((interrupt ("IRQ")));
1999 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2001 @item interrupt_handler
2002 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2003 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2004 indicate that the specified function is an interrupt handler. The compiler
2005 will generate function entry and exit sequences suitable for use in an
2006 interrupt handler when this attribute is present.
2009 @cindex User stack pointer in interrupts on the Blackfin
2010 When used together with @code{interrupt_handler}, @code{exception_handler}
2011 or @code{nmi_handler}, code will be generated to load the stack pointer
2012 from the USP register in the function prologue.
2014 @item long_call/short_call
2015 @cindex indirect calls on ARM
2016 This attribute specifies how a particular function is called on
2017 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2018 command line switch and @code{#pragma long_calls} settings. The
2019 @code{long_call} attribute indicates that the function might be far
2020 away from the call site and require a different (more expensive)
2021 calling sequence. The @code{short_call} attribute always places
2022 the offset to the function from the call site into the @samp{BL}
2023 instruction directly.
2025 @item longcall/shortcall
2026 @cindex functions called via pointer on the RS/6000 and PowerPC
2027 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2028 indicates that the function might be far away from the call site and
2029 require a different (more expensive) calling sequence. The
2030 @code{shortcall} attribute indicates that the function is always close
2031 enough for the shorter calling sequence to be used. These attributes
2032 override both the @option{-mlongcall} switch and, on the RS/6000 and
2033 PowerPC, the @code{#pragma longcall} setting.
2035 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2036 calls are necessary.
2039 @cindex indirect calls on MIPS
2040 This attribute specifies how a particular function is called on MIPS@.
2041 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2042 command line switch. This attribute causes the compiler to always call
2043 the function by first loading its address into a register, and then using
2044 the contents of that register.
2047 @cindex @code{malloc} attribute
2048 The @code{malloc} attribute is used to tell the compiler that a function
2049 may be treated as if any non-@code{NULL} pointer it returns cannot
2050 alias any other pointer valid when the function returns.
2051 This will often improve optimization.
2052 Standard functions with this property include @code{malloc} and
2053 @code{calloc}. @code{realloc}-like functions have this property as
2054 long as the old pointer is never referred to (including comparing it
2055 to the new pointer) after the function returns a non-@code{NULL}
2058 @item model (@var{model-name})
2059 @cindex function addressability on the M32R/D
2060 @cindex variable addressability on the IA-64
2062 On the M32R/D, use this attribute to set the addressability of an
2063 object, and of the code generated for a function. The identifier
2064 @var{model-name} is one of @code{small}, @code{medium}, or
2065 @code{large}, representing each of the code models.
2067 Small model objects live in the lower 16MB of memory (so that their
2068 addresses can be loaded with the @code{ld24} instruction), and are
2069 callable with the @code{bl} instruction.
2071 Medium model objects may live anywhere in the 32-bit address space (the
2072 compiler will generate @code{seth/add3} instructions to load their addresses),
2073 and are callable with the @code{bl} instruction.
2075 Large model objects may live anywhere in the 32-bit address space (the
2076 compiler will generate @code{seth/add3} instructions to load their addresses),
2077 and may not be reachable with the @code{bl} instruction (the compiler will
2078 generate the much slower @code{seth/add3/jl} instruction sequence).
2080 On IA-64, use this attribute to set the addressability of an object.
2081 At present, the only supported identifier for @var{model-name} is
2082 @code{small}, indicating addressability via ``small'' (22-bit)
2083 addresses (so that their addresses can be loaded with the @code{addl}
2084 instruction). Caveat: such addressing is by definition not position
2085 independent and hence this attribute must not be used for objects
2086 defined by shared libraries.
2089 @cindex function without a prologue/epilogue code
2090 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2091 specified function does not need prologue/epilogue sequences generated by
2092 the compiler. It is up to the programmer to provide these sequences.
2095 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2096 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2097 use the normal calling convention based on @code{jsr} and @code{rts}.
2098 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2102 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2103 Use this attribute together with @code{interrupt_handler},
2104 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2105 entry code should enable nested interrupts or exceptions.
2108 @cindex NMI handler functions on the Blackfin processor
2109 Use this attribute on the Blackfin to indicate that the specified function
2110 is an NMI handler. The compiler will generate function entry and
2111 exit sequences suitable for use in an NMI handler when this
2112 attribute is present.
2114 @item no_instrument_function
2115 @cindex @code{no_instrument_function} function attribute
2116 @opindex finstrument-functions
2117 If @option{-finstrument-functions} is given, profiling function calls will
2118 be generated at entry and exit of most user-compiled functions.
2119 Functions with this attribute will not be so instrumented.
2122 @cindex @code{noinline} function attribute
2123 This function attribute prevents a function from being considered for
2126 @item nonnull (@var{arg-index}, @dots{})
2127 @cindex @code{nonnull} function attribute
2128 The @code{nonnull} attribute specifies that some function parameters should
2129 be non-null pointers. For instance, the declaration:
2133 my_memcpy (void *dest, const void *src, size_t len)
2134 __attribute__((nonnull (1, 2)));
2138 causes the compiler to check that, in calls to @code{my_memcpy},
2139 arguments @var{dest} and @var{src} are non-null. If the compiler
2140 determines that a null pointer is passed in an argument slot marked
2141 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2142 is issued. The compiler may also choose to make optimizations based
2143 on the knowledge that certain function arguments will not be null.
2145 If no argument index list is given to the @code{nonnull} attribute,
2146 all pointer arguments are marked as non-null. To illustrate, the
2147 following declaration is equivalent to the previous example:
2151 my_memcpy (void *dest, const void *src, size_t len)
2152 __attribute__((nonnull));
2156 @cindex @code{noreturn} function attribute
2157 A few standard library functions, such as @code{abort} and @code{exit},
2158 cannot return. GCC knows this automatically. Some programs define
2159 their own functions that never return. You can declare them
2160 @code{noreturn} to tell the compiler this fact. For example,
2164 void fatal () __attribute__ ((noreturn));
2167 fatal (/* @r{@dots{}} */)
2169 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2175 The @code{noreturn} keyword tells the compiler to assume that
2176 @code{fatal} cannot return. It can then optimize without regard to what
2177 would happen if @code{fatal} ever did return. This makes slightly
2178 better code. More importantly, it helps avoid spurious warnings of
2179 uninitialized variables.
2181 The @code{noreturn} keyword does not affect the exceptional path when that
2182 applies: a @code{noreturn}-marked function may still return to the caller
2183 by throwing an exception or calling @code{longjmp}.
2185 Do not assume that registers saved by the calling function are
2186 restored before calling the @code{noreturn} function.
2188 It does not make sense for a @code{noreturn} function to have a return
2189 type other than @code{void}.
2191 The attribute @code{noreturn} is not implemented in GCC versions
2192 earlier than 2.5. An alternative way to declare that a function does
2193 not return, which works in the current version and in some older
2194 versions, is as follows:
2197 typedef void voidfn ();
2199 volatile voidfn fatal;
2202 This approach does not work in GNU C++.
2205 @cindex @code{nothrow} function attribute
2206 The @code{nothrow} attribute is used to inform the compiler that a
2207 function cannot throw an exception. For example, most functions in
2208 the standard C library can be guaranteed not to throw an exception
2209 with the notable exceptions of @code{qsort} and @code{bsearch} that
2210 take function pointer arguments. The @code{nothrow} attribute is not
2211 implemented in GCC versions earlier than 3.3.
2214 @cindex @code{pure} function attribute
2215 Many functions have no effects except the return value and their
2216 return value depends only on the parameters and/or global variables.
2217 Such a function can be subject
2218 to common subexpression elimination and loop optimization just as an
2219 arithmetic operator would be. These functions should be declared
2220 with the attribute @code{pure}. For example,
2223 int square (int) __attribute__ ((pure));
2227 says that the hypothetical function @code{square} is safe to call
2228 fewer times than the program says.
2230 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2231 Interesting non-pure functions are functions with infinite loops or those
2232 depending on volatile memory or other system resource, that may change between
2233 two consecutive calls (such as @code{feof} in a multithreading environment).
2235 The attribute @code{pure} is not implemented in GCC versions earlier
2238 @item regparm (@var{number})
2239 @cindex @code{regparm} attribute
2240 @cindex functions that are passed arguments in registers on the 386
2241 On the Intel 386, the @code{regparm} attribute causes the compiler to
2242 pass arguments number one to @var{number} if they are of integral type
2243 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2244 take a variable number of arguments will continue to be passed all of their
2245 arguments on the stack.
2247 Beware that on some ELF systems this attribute is unsuitable for
2248 global functions in shared libraries with lazy binding (which is the
2249 default). Lazy binding will send the first call via resolving code in
2250 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2251 per the standard calling conventions. Solaris 8 is affected by this.
2252 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2253 safe since the loaders there save all registers. (Lazy binding can be
2254 disabled with the linker or the loader if desired, to avoid the
2258 @cindex @code{sseregparm} attribute
2259 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2260 causes the compiler to pass up to 3 floating point arguments in
2261 SSE registers instead of on the stack. Functions that take a
2262 variable number of arguments will continue to pass all of their
2263 floating point arguments on the stack.
2265 @item force_align_arg_pointer
2266 @cindex @code{force_align_arg_pointer} attribute
2267 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2268 applied to individual function definitions, generating an alternate
2269 prologue and epilogue that realigns the runtime stack. This supports
2270 mixing legacy codes that run with a 4-byte aligned stack with modern
2271 codes that keep a 16-byte stack for SSE compatibility. The alternate
2272 prologue and epilogue are slower and bigger than the regular ones, and
2273 the alternate prologue requires a scratch register; this lowers the
2274 number of registers available if used in conjunction with the
2275 @code{regparm} attribute. The @code{force_align_arg_pointer}
2276 attribute is incompatible with nested functions; this is considered a
2280 @cindex @code{returns_twice} attribute
2281 The @code{returns_twice} attribute tells the compiler that a function may
2282 return more than one time. The compiler will ensure that all registers
2283 are dead before calling such a function and will emit a warning about
2284 the variables that may be clobbered after the second return from the
2285 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2286 The @code{longjmp}-like counterpart of such function, if any, might need
2287 to be marked with the @code{noreturn} attribute.
2290 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2291 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2292 all registers except the stack pointer should be saved in the prologue
2293 regardless of whether they are used or not.
2295 @item section ("@var{section-name}")
2296 @cindex @code{section} function attribute
2297 Normally, the compiler places the code it generates in the @code{text} section.
2298 Sometimes, however, you need additional sections, or you need certain
2299 particular functions to appear in special sections. The @code{section}
2300 attribute specifies that a function lives in a particular section.
2301 For example, the declaration:
2304 extern void foobar (void) __attribute__ ((section ("bar")));
2308 puts the function @code{foobar} in the @code{bar} section.
2310 Some file formats do not support arbitrary sections so the @code{section}
2311 attribute is not available on all platforms.
2312 If you need to map the entire contents of a module to a particular
2313 section, consider using the facilities of the linker instead.
2316 @cindex @code{sentinel} function attribute
2317 This function attribute ensures that a parameter in a function call is
2318 an explicit @code{NULL}. The attribute is only valid on variadic
2319 functions. By default, the sentinel is located at position zero, the
2320 last parameter of the function call. If an optional integer position
2321 argument P is supplied to the attribute, the sentinel must be located at
2322 position P counting backwards from the end of the argument list.
2325 __attribute__ ((sentinel))
2327 __attribute__ ((sentinel(0)))
2330 The attribute is automatically set with a position of 0 for the built-in
2331 functions @code{execl} and @code{execlp}. The built-in function
2332 @code{execle} has the attribute set with a position of 1.
2334 A valid @code{NULL} in this context is defined as zero with any pointer
2335 type. If your system defines the @code{NULL} macro with an integer type
2336 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2337 with a copy that redefines NULL appropriately.
2339 The warnings for missing or incorrect sentinels are enabled with
2343 See long_call/short_call.
2346 See longcall/shortcall.
2349 @cindex signal handler functions on the AVR processors
2350 Use this attribute on the AVR to indicate that the specified
2351 function is a signal handler. The compiler will generate function
2352 entry and exit sequences suitable for use in a signal handler when this
2353 attribute is present. Interrupts will be disabled inside the function.
2356 Use this attribute on the SH to indicate an @code{interrupt_handler}
2357 function should switch to an alternate stack. It expects a string
2358 argument that names a global variable holding the address of the
2363 void f () __attribute__ ((interrupt_handler,
2364 sp_switch ("alt_stack")));
2368 @cindex functions that pop the argument stack on the 386
2369 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2370 assume that the called function will pop off the stack space used to
2371 pass arguments, unless it takes a variable number of arguments.
2374 @cindex tiny data section on the H8/300H and H8S
2375 Use this attribute on the H8/300H and H8S to indicate that the specified
2376 variable should be placed into the tiny data section.
2377 The compiler will generate more efficient code for loads and stores
2378 on data in the tiny data section. Note the tiny data area is limited to
2379 slightly under 32kbytes of data.
2382 Use this attribute on the SH for an @code{interrupt_handler} to return using
2383 @code{trapa} instead of @code{rte}. This attribute expects an integer
2384 argument specifying the trap number to be used.
2387 @cindex @code{unused} attribute.
2388 This attribute, attached to a function, means that the function is meant
2389 to be possibly unused. GCC will not produce a warning for this
2393 @cindex @code{used} attribute.
2394 This attribute, attached to a function, means that code must be emitted
2395 for the function even if it appears that the function is not referenced.
2396 This is useful, for example, when the function is referenced only in
2399 @item visibility ("@var{visibility_type}")
2400 @cindex @code{visibility} attribute
2401 This attribute affects the linkage of the declaration to which it is attached.
2402 There are four supported @var{visibility_type} values: default,
2403 hidden, protected or internal visibility.
2406 void __attribute__ ((visibility ("protected")))
2407 f () @{ /* @r{Do something.} */; @}
2408 int i __attribute__ ((visibility ("hidden")));
2411 The possible values of @var{visibility_type} correspond to the
2412 visibility settings in the ELF gABI.
2415 @c keep this list of visibilities in alphabetical order.
2418 Default visibility is the normal case for the object file format.
2419 This value is available for the visibility attribute to override other
2420 options that may change the assumed visibility of entities.
2422 On ELF, default visibility means that the declaration is visible to other
2423 modules and, in shared libraries, means that the declared entity may be
2426 On Darwin, default visibility means that the declaration is visible to
2429 Default visibility corresponds to ``external linkage'' in the language.
2432 Hidden visibility indicates that the entity declared will have a new
2433 form of linkage, which we'll call ``hidden linkage''. Two
2434 declarations of an object with hidden linkage refer to the same object
2435 if they are in the same shared object.
2438 Internal visibility is like hidden visibility, but with additional
2439 processor specific semantics. Unless otherwise specified by the
2440 psABI, GCC defines internal visibility to mean that a function is
2441 @emph{never} called from another module. Compare this with hidden
2442 functions which, while they cannot be referenced directly by other
2443 modules, can be referenced indirectly via function pointers. By
2444 indicating that a function cannot be called from outside the module,
2445 GCC may for instance omit the load of a PIC register since it is known
2446 that the calling function loaded the correct value.
2449 Protected visibility is like default visibility except that it
2450 indicates that references within the defining module will bind to the
2451 definition in that module. That is, the declared entity cannot be
2452 overridden by another module.
2456 All visibilities are supported on many, but not all, ELF targets
2457 (supported when the assembler supports the @samp{.visibility}
2458 pseudo-op). Default visibility is supported everywhere. Hidden
2459 visibility is supported on Darwin targets.
2461 The visibility attribute should be applied only to declarations which
2462 would otherwise have external linkage. The attribute should be applied
2463 consistently, so that the same entity should not be declared with
2464 different settings of the attribute.
2466 In C++, the visibility attribute applies to types as well as functions
2467 and objects, because in C++ types have linkage. A class must not have
2468 greater visibility than its non-static data member types and bases,
2469 and class members default to the visibility of their class. Also, a
2470 declaration without explicit visibility is limited to the visibility
2473 In C++, you can mark member functions and static member variables of a
2474 class with the visibility attribute. This is useful if if you know a
2475 particular method or static member variable should only be used from
2476 one shared object; then you can mark it hidden while the rest of the
2477 class has default visibility. Care must be taken to avoid breaking
2478 the One Definition Rule; for example, it is usually not useful to mark
2479 an inline method as hidden without marking the whole class as hidden.
2481 A C++ namespace declaration can also have the visibility attribute.
2482 This attribute applies only to the particular namespace body, not to
2483 other definitions of the same namespace; it is equivalent to using
2484 @samp{#pragma GCC visibility} before and after the namespace
2485 definition (@pxref{Visibility Pragmas}).
2487 In C++, if a template argument has limited visibility, this
2488 restriction is implicitly propagated to the template instantiation.
2489 Otherwise, template instantiations and specializations default to the
2490 visibility of their template.
2492 If both the template and enclosing class have explicit visibility, the
2493 visibility from the template is used.
2495 @item warn_unused_result
2496 @cindex @code{warn_unused_result} attribute
2497 The @code{warn_unused_result} attribute causes a warning to be emitted
2498 if a caller of the function with this attribute does not use its
2499 return value. This is useful for functions where not checking
2500 the result is either a security problem or always a bug, such as
2504 int fn () __attribute__ ((warn_unused_result));
2507 if (fn () < 0) return -1;
2513 results in warning on line 5.
2516 @cindex @code{weak} attribute
2517 The @code{weak} attribute causes the declaration to be emitted as a weak
2518 symbol rather than a global. This is primarily useful in defining
2519 library functions which can be overridden in user code, though it can
2520 also be used with non-function declarations. Weak symbols are supported
2521 for ELF targets, and also for a.out targets when using the GNU assembler
2525 @itemx weakref ("@var{target}")
2526 @cindex @code{weakref} attribute
2527 The @code{weakref} attribute marks a declaration as a weak reference.
2528 Without arguments, it should be accompanied by an @code{alias} attribute
2529 naming the target symbol. Optionally, the @var{target} may be given as
2530 an argument to @code{weakref} itself. In either case, @code{weakref}
2531 implicitly marks the declaration as @code{weak}. Without a
2532 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2533 @code{weakref} is equivalent to @code{weak}.
2536 static int x() __attribute__ ((weakref ("y")));
2537 /* is equivalent to... */
2538 static int x() __attribute__ ((weak, weakref, alias ("y")));
2540 static int x() __attribute__ ((weakref));
2541 static int x() __attribute__ ((alias ("y")));
2544 A weak reference is an alias that does not by itself require a
2545 definition to be given for the target symbol. If the target symbol is
2546 only referenced through weak references, then the becomes a @code{weak}
2547 undefined symbol. If it is directly referenced, however, then such
2548 strong references prevail, and a definition will be required for the
2549 symbol, not necessarily in the same translation unit.
2551 The effect is equivalent to moving all references to the alias to a
2552 separate translation unit, renaming the alias to the aliased symbol,
2553 declaring it as weak, compiling the two separate translation units and
2554 performing a reloadable link on them.
2556 At present, a declaration to which @code{weakref} is attached can
2557 only be @code{static}.
2559 @item externally_visible
2560 @cindex @code{externally_visible} attribute.
2561 This attribute, attached to a global variable or function nullify
2562 effect of @option{-fwhole-program} command line option, so the object
2563 remain visible outside the current compilation unit
2567 You can specify multiple attributes in a declaration by separating them
2568 by commas within the double parentheses or by immediately following an
2569 attribute declaration with another attribute declaration.
2571 @cindex @code{#pragma}, reason for not using
2572 @cindex pragma, reason for not using
2573 Some people object to the @code{__attribute__} feature, suggesting that
2574 ISO C's @code{#pragma} should be used instead. At the time
2575 @code{__attribute__} was designed, there were two reasons for not doing
2580 It is impossible to generate @code{#pragma} commands from a macro.
2583 There is no telling what the same @code{#pragma} might mean in another
2587 These two reasons applied to almost any application that might have been
2588 proposed for @code{#pragma}. It was basically a mistake to use
2589 @code{#pragma} for @emph{anything}.
2591 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2592 to be generated from macros. In addition, a @code{#pragma GCC}
2593 namespace is now in use for GCC-specific pragmas. However, it has been
2594 found convenient to use @code{__attribute__} to achieve a natural
2595 attachment of attributes to their corresponding declarations, whereas
2596 @code{#pragma GCC} is of use for constructs that do not naturally form
2597 part of the grammar. @xref{Other Directives,,Miscellaneous
2598 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2600 @node Attribute Syntax
2601 @section Attribute Syntax
2602 @cindex attribute syntax
2604 This section describes the syntax with which @code{__attribute__} may be
2605 used, and the constructs to which attribute specifiers bind, for the C
2606 language. Some details may vary for C++. Because of infelicities in
2607 the grammar for attributes, some forms described here may not be
2608 successfully parsed in all cases.
2610 There are some problems with the semantics of attributes in C++. For
2611 example, there are no manglings for attributes, although they may affect
2612 code generation, so problems may arise when attributed types are used in
2613 conjunction with templates or overloading. Similarly, @code{typeid}
2614 does not distinguish between types with different attributes. Support
2615 for attributes in C++ may be restricted in future to attributes on
2616 declarations only, but not on nested declarators.
2618 @xref{Function Attributes}, for details of the semantics of attributes
2619 applying to functions. @xref{Variable Attributes}, for details of the
2620 semantics of attributes applying to variables. @xref{Type Attributes},
2621 for details of the semantics of attributes applying to structure, union
2622 and enumerated types.
2624 An @dfn{attribute specifier} is of the form
2625 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2626 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2627 each attribute is one of the following:
2631 Empty. Empty attributes are ignored.
2634 A word (which may be an identifier such as @code{unused}, or a reserved
2635 word such as @code{const}).
2638 A word, followed by, in parentheses, parameters for the attribute.
2639 These parameters take one of the following forms:
2643 An identifier. For example, @code{mode} attributes use this form.
2646 An identifier followed by a comma and a non-empty comma-separated list
2647 of expressions. For example, @code{format} attributes use this form.
2650 A possibly empty comma-separated list of expressions. For example,
2651 @code{format_arg} attributes use this form with the list being a single
2652 integer constant expression, and @code{alias} attributes use this form
2653 with the list being a single string constant.
2657 An @dfn{attribute specifier list} is a sequence of one or more attribute
2658 specifiers, not separated by any other tokens.
2660 In GNU C, an attribute specifier list may appear after the colon following a
2661 label, other than a @code{case} or @code{default} label. The only
2662 attribute it makes sense to use after a label is @code{unused}. This
2663 feature is intended for code generated by programs which contains labels
2664 that may be unused but which is compiled with @option{-Wall}. It would
2665 not normally be appropriate to use in it human-written code, though it
2666 could be useful in cases where the code that jumps to the label is
2667 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2668 such placement of attribute lists, as it is permissible for a
2669 declaration, which could begin with an attribute list, to be labelled in
2670 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2671 does not arise there.
2673 An attribute specifier list may appear as part of a @code{struct},
2674 @code{union} or @code{enum} specifier. It may go either immediately
2675 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2676 the closing brace. The former syntax is preferred.
2677 Where attribute specifiers follow the closing brace, they are considered
2678 to relate to the structure, union or enumerated type defined, not to any
2679 enclosing declaration the type specifier appears in, and the type
2680 defined is not complete until after the attribute specifiers.
2681 @c Otherwise, there would be the following problems: a shift/reduce
2682 @c conflict between attributes binding the struct/union/enum and
2683 @c binding to the list of specifiers/qualifiers; and "aligned"
2684 @c attributes could use sizeof for the structure, but the size could be
2685 @c changed later by "packed" attributes.
2687 Otherwise, an attribute specifier appears as part of a declaration,
2688 counting declarations of unnamed parameters and type names, and relates
2689 to that declaration (which may be nested in another declaration, for
2690 example in the case of a parameter declaration), or to a particular declarator
2691 within a declaration. Where an
2692 attribute specifier is applied to a parameter declared as a function or
2693 an array, it should apply to the function or array rather than the
2694 pointer to which the parameter is implicitly converted, but this is not
2695 yet correctly implemented.
2697 Any list of specifiers and qualifiers at the start of a declaration may
2698 contain attribute specifiers, whether or not such a list may in that
2699 context contain storage class specifiers. (Some attributes, however,
2700 are essentially in the nature of storage class specifiers, and only make
2701 sense where storage class specifiers may be used; for example,
2702 @code{section}.) There is one necessary limitation to this syntax: the
2703 first old-style parameter declaration in a function definition cannot
2704 begin with an attribute specifier, because such an attribute applies to
2705 the function instead by syntax described below (which, however, is not
2706 yet implemented in this case). In some other cases, attribute
2707 specifiers are permitted by this grammar but not yet supported by the
2708 compiler. All attribute specifiers in this place relate to the
2709 declaration as a whole. In the obsolescent usage where a type of
2710 @code{int} is implied by the absence of type specifiers, such a list of
2711 specifiers and qualifiers may be an attribute specifier list with no
2712 other specifiers or qualifiers.
2714 At present, the first parameter in a function prototype must have some
2715 type specifier which is not an attribute specifier; this resolves an
2716 ambiguity in the interpretation of @code{void f(int
2717 (__attribute__((foo)) x))}, but is subject to change. At present, if
2718 the parentheses of a function declarator contain only attributes then
2719 those attributes are ignored, rather than yielding an error or warning
2720 or implying a single parameter of type int, but this is subject to
2723 An attribute specifier list may appear immediately before a declarator
2724 (other than the first) in a comma-separated list of declarators in a
2725 declaration of more than one identifier using a single list of
2726 specifiers and qualifiers. Such attribute specifiers apply
2727 only to the identifier before whose declarator they appear. For
2731 __attribute__((noreturn)) void d0 (void),
2732 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2737 the @code{noreturn} attribute applies to all the functions
2738 declared; the @code{format} attribute only applies to @code{d1}.
2740 An attribute specifier list may appear immediately before the comma,
2741 @code{=} or semicolon terminating the declaration of an identifier other
2742 than a function definition. At present, such attribute specifiers apply
2743 to the declared object or function, but in future they may attach to the
2744 outermost adjacent declarator. In simple cases there is no difference,
2745 but, for example, in
2748 void (****f)(void) __attribute__((noreturn));
2752 at present the @code{noreturn} attribute applies to @code{f}, which
2753 causes a warning since @code{f} is not a function, but in future it may
2754 apply to the function @code{****f}. The precise semantics of what
2755 attributes in such cases will apply to are not yet specified. Where an
2756 assembler name for an object or function is specified (@pxref{Asm
2757 Labels}), at present the attribute must follow the @code{asm}
2758 specification; in future, attributes before the @code{asm} specification
2759 may apply to the adjacent declarator, and those after it to the declared
2762 An attribute specifier list may, in future, be permitted to appear after
2763 the declarator in a function definition (before any old-style parameter
2764 declarations or the function body).
2766 Attribute specifiers may be mixed with type qualifiers appearing inside
2767 the @code{[]} of a parameter array declarator, in the C99 construct by
2768 which such qualifiers are applied to the pointer to which the array is
2769 implicitly converted. Such attribute specifiers apply to the pointer,
2770 not to the array, but at present this is not implemented and they are
2773 An attribute specifier list may appear at the start of a nested
2774 declarator. At present, there are some limitations in this usage: the
2775 attributes correctly apply to the declarator, but for most individual
2776 attributes the semantics this implies are not implemented.
2777 When attribute specifiers follow the @code{*} of a pointer
2778 declarator, they may be mixed with any type qualifiers present.
2779 The following describes the formal semantics of this syntax. It will make the
2780 most sense if you are familiar with the formal specification of
2781 declarators in the ISO C standard.
2783 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2784 D1}, where @code{T} contains declaration specifiers that specify a type
2785 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2786 contains an identifier @var{ident}. The type specified for @var{ident}
2787 for derived declarators whose type does not include an attribute
2788 specifier is as in the ISO C standard.
2790 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2791 and the declaration @code{T D} specifies the type
2792 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2793 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2794 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2796 If @code{D1} has the form @code{*
2797 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2798 declaration @code{T D} specifies the type
2799 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2800 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2801 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2807 void (__attribute__((noreturn)) ****f) (void);
2811 specifies the type ``pointer to pointer to pointer to pointer to
2812 non-returning function returning @code{void}''. As another example,
2815 char *__attribute__((aligned(8))) *f;
2819 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2820 Note again that this does not work with most attributes; for example,
2821 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2822 is not yet supported.
2824 For compatibility with existing code written for compiler versions that
2825 did not implement attributes on nested declarators, some laxity is
2826 allowed in the placing of attributes. If an attribute that only applies
2827 to types is applied to a declaration, it will be treated as applying to
2828 the type of that declaration. If an attribute that only applies to
2829 declarations is applied to the type of a declaration, it will be treated
2830 as applying to that declaration; and, for compatibility with code
2831 placing the attributes immediately before the identifier declared, such
2832 an attribute applied to a function return type will be treated as
2833 applying to the function type, and such an attribute applied to an array
2834 element type will be treated as applying to the array type. If an
2835 attribute that only applies to function types is applied to a
2836 pointer-to-function type, it will be treated as applying to the pointer
2837 target type; if such an attribute is applied to a function return type
2838 that is not a pointer-to-function type, it will be treated as applying
2839 to the function type.
2841 @node Function Prototypes
2842 @section Prototypes and Old-Style Function Definitions
2843 @cindex function prototype declarations
2844 @cindex old-style function definitions
2845 @cindex promotion of formal parameters
2847 GNU C extends ISO C to allow a function prototype to override a later
2848 old-style non-prototype definition. Consider the following example:
2851 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2858 /* @r{Prototype function declaration.} */
2859 int isroot P((uid_t));
2861 /* @r{Old-style function definition.} */
2863 isroot (x) /* @r{??? lossage here ???} */
2870 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2871 not allow this example, because subword arguments in old-style
2872 non-prototype definitions are promoted. Therefore in this example the
2873 function definition's argument is really an @code{int}, which does not
2874 match the prototype argument type of @code{short}.
2876 This restriction of ISO C makes it hard to write code that is portable
2877 to traditional C compilers, because the programmer does not know
2878 whether the @code{uid_t} type is @code{short}, @code{int}, or
2879 @code{long}. Therefore, in cases like these GNU C allows a prototype
2880 to override a later old-style definition. More precisely, in GNU C, a
2881 function prototype argument type overrides the argument type specified
2882 by a later old-style definition if the former type is the same as the
2883 latter type before promotion. Thus in GNU C the above example is
2884 equivalent to the following:
2897 GNU C++ does not support old-style function definitions, so this
2898 extension is irrelevant.
2901 @section C++ Style Comments
2903 @cindex C++ comments
2904 @cindex comments, C++ style
2906 In GNU C, you may use C++ style comments, which start with @samp{//} and
2907 continue until the end of the line. Many other C implementations allow
2908 such comments, and they are included in the 1999 C standard. However,
2909 C++ style comments are not recognized if you specify an @option{-std}
2910 option specifying a version of ISO C before C99, or @option{-ansi}
2911 (equivalent to @option{-std=c89}).
2914 @section Dollar Signs in Identifier Names
2916 @cindex dollar signs in identifier names
2917 @cindex identifier names, dollar signs in
2919 In GNU C, you may normally use dollar signs in identifier names.
2920 This is because many traditional C implementations allow such identifiers.
2921 However, dollar signs in identifiers are not supported on a few target
2922 machines, typically because the target assembler does not allow them.
2924 @node Character Escapes
2925 @section The Character @key{ESC} in Constants
2927 You can use the sequence @samp{\e} in a string or character constant to
2928 stand for the ASCII character @key{ESC}.
2931 @section Inquiring on Alignment of Types or Variables
2933 @cindex type alignment
2934 @cindex variable alignment
2936 The keyword @code{__alignof__} allows you to inquire about how an object
2937 is aligned, or the minimum alignment usually required by a type. Its
2938 syntax is just like @code{sizeof}.
2940 For example, if the target machine requires a @code{double} value to be
2941 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2942 This is true on many RISC machines. On more traditional machine
2943 designs, @code{__alignof__ (double)} is 4 or even 2.
2945 Some machines never actually require alignment; they allow reference to any
2946 data type even at an odd address. For these machines, @code{__alignof__}
2947 reports the @emph{recommended} alignment of a type.
2949 If the operand of @code{__alignof__} is an lvalue rather than a type,
2950 its value is the required alignment for its type, taking into account
2951 any minimum alignment specified with GCC's @code{__attribute__}
2952 extension (@pxref{Variable Attributes}). For example, after this
2956 struct foo @{ int x; char y; @} foo1;
2960 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2961 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2963 It is an error to ask for the alignment of an incomplete type.
2965 @node Variable Attributes
2966 @section Specifying Attributes of Variables
2967 @cindex attribute of variables
2968 @cindex variable attributes
2970 The keyword @code{__attribute__} allows you to specify special
2971 attributes of variables or structure fields. This keyword is followed
2972 by an attribute specification inside double parentheses. Some
2973 attributes are currently defined generically for variables.
2974 Other attributes are defined for variables on particular target
2975 systems. Other attributes are available for functions
2976 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2977 Other front ends might define more attributes
2978 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2980 You may also specify attributes with @samp{__} preceding and following
2981 each keyword. This allows you to use them in header files without
2982 being concerned about a possible macro of the same name. For example,
2983 you may use @code{__aligned__} instead of @code{aligned}.
2985 @xref{Attribute Syntax}, for details of the exact syntax for using
2989 @cindex @code{aligned} attribute
2990 @item aligned (@var{alignment})
2991 This attribute specifies a minimum alignment for the variable or
2992 structure field, measured in bytes. For example, the declaration:
2995 int x __attribute__ ((aligned (16))) = 0;
2999 causes the compiler to allocate the global variable @code{x} on a
3000 16-byte boundary. On a 68040, this could be used in conjunction with
3001 an @code{asm} expression to access the @code{move16} instruction which
3002 requires 16-byte aligned operands.
3004 You can also specify the alignment of structure fields. For example, to
3005 create a double-word aligned @code{int} pair, you could write:
3008 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3012 This is an alternative to creating a union with a @code{double} member
3013 that forces the union to be double-word aligned.
3015 As in the preceding examples, you can explicitly specify the alignment
3016 (in bytes) that you wish the compiler to use for a given variable or
3017 structure field. Alternatively, you can leave out the alignment factor
3018 and just ask the compiler to align a variable or field to the maximum
3019 useful alignment for the target machine you are compiling for. For
3020 example, you could write:
3023 short array[3] __attribute__ ((aligned));
3026 Whenever you leave out the alignment factor in an @code{aligned} attribute
3027 specification, the compiler automatically sets the alignment for the declared
3028 variable or field to the largest alignment which is ever used for any data
3029 type on the target machine you are compiling for. Doing this can often make
3030 copy operations more efficient, because the compiler can use whatever
3031 instructions copy the biggest chunks of memory when performing copies to
3032 or from the variables or fields that you have aligned this way.
3034 The @code{aligned} attribute can only increase the alignment; but you
3035 can decrease it by specifying @code{packed} as well. See below.
3037 Note that the effectiveness of @code{aligned} attributes may be limited
3038 by inherent limitations in your linker. On many systems, the linker is
3039 only able to arrange for variables to be aligned up to a certain maximum
3040 alignment. (For some linkers, the maximum supported alignment may
3041 be very very small.) If your linker is only able to align variables
3042 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3043 in an @code{__attribute__} will still only provide you with 8 byte
3044 alignment. See your linker documentation for further information.
3046 @item cleanup (@var{cleanup_function})
3047 @cindex @code{cleanup} attribute
3048 The @code{cleanup} attribute runs a function when the variable goes
3049 out of scope. This attribute can only be applied to auto function
3050 scope variables; it may not be applied to parameters or variables
3051 with static storage duration. The function must take one parameter,
3052 a pointer to a type compatible with the variable. The return value
3053 of the function (if any) is ignored.
3055 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3056 will be run during the stack unwinding that happens during the
3057 processing of the exception. Note that the @code{cleanup} attribute
3058 does not allow the exception to be caught, only to perform an action.
3059 It is undefined what happens if @var{cleanup_function} does not
3064 @cindex @code{common} attribute
3065 @cindex @code{nocommon} attribute
3068 The @code{common} attribute requests GCC to place a variable in
3069 ``common'' storage. The @code{nocommon} attribute requests the
3070 opposite---to allocate space for it directly.
3072 These attributes override the default chosen by the
3073 @option{-fno-common} and @option{-fcommon} flags respectively.
3076 @cindex @code{deprecated} attribute
3077 The @code{deprecated} attribute results in a warning if the variable
3078 is used anywhere in the source file. This is useful when identifying
3079 variables that are expected to be removed in a future version of a
3080 program. The warning also includes the location of the declaration
3081 of the deprecated variable, to enable users to easily find further
3082 information about why the variable is deprecated, or what they should
3083 do instead. Note that the warning only occurs for uses:
3086 extern int old_var __attribute__ ((deprecated));
3088 int new_fn () @{ return old_var; @}
3091 results in a warning on line 3 but not line 2.
3093 The @code{deprecated} attribute can also be used for functions and
3094 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3096 @item mode (@var{mode})
3097 @cindex @code{mode} attribute
3098 This attribute specifies the data type for the declaration---whichever
3099 type corresponds to the mode @var{mode}. This in effect lets you
3100 request an integer or floating point type according to its width.
3102 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3103 indicate the mode corresponding to a one-byte integer, @samp{word} or
3104 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3105 or @samp{__pointer__} for the mode used to represent pointers.
3108 @cindex @code{packed} attribute
3109 The @code{packed} attribute specifies that a variable or structure field
3110 should have the smallest possible alignment---one byte for a variable,
3111 and one bit for a field, unless you specify a larger value with the
3112 @code{aligned} attribute.
3114 Here is a structure in which the field @code{x} is packed, so that it
3115 immediately follows @code{a}:
3121 int x[2] __attribute__ ((packed));
3125 @item section ("@var{section-name}")
3126 @cindex @code{section} variable attribute
3127 Normally, the compiler places the objects it generates in sections like
3128 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3129 or you need certain particular variables to appear in special sections,
3130 for example to map to special hardware. The @code{section}
3131 attribute specifies that a variable (or function) lives in a particular
3132 section. For example, this small program uses several specific section names:
3135 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3136 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3137 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3138 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3142 /* @r{Initialize stack pointer} */
3143 init_sp (stack + sizeof (stack));
3145 /* @r{Initialize initialized data} */
3146 memcpy (&init_data, &data, &edata - &data);
3148 /* @r{Turn on the serial ports} */
3155 Use the @code{section} attribute with an @emph{initialized} definition
3156 of a @emph{global} variable, as shown in the example. GCC issues
3157 a warning and otherwise ignores the @code{section} attribute in
3158 uninitialized variable declarations.
3160 You may only use the @code{section} attribute with a fully initialized
3161 global definition because of the way linkers work. The linker requires
3162 each object be defined once, with the exception that uninitialized
3163 variables tentatively go in the @code{common} (or @code{bss}) section
3164 and can be multiply ``defined''. You can force a variable to be
3165 initialized with the @option{-fno-common} flag or the @code{nocommon}
3168 Some file formats do not support arbitrary sections so the @code{section}
3169 attribute is not available on all platforms.
3170 If you need to map the entire contents of a module to a particular
3171 section, consider using the facilities of the linker instead.
3174 @cindex @code{shared} variable attribute
3175 On Microsoft Windows, in addition to putting variable definitions in a named
3176 section, the section can also be shared among all running copies of an
3177 executable or DLL@. For example, this small program defines shared data
3178 by putting it in a named section @code{shared} and marking the section
3182 int foo __attribute__((section ("shared"), shared)) = 0;
3187 /* @r{Read and write foo. All running
3188 copies see the same value.} */
3194 You may only use the @code{shared} attribute along with @code{section}
3195 attribute with a fully initialized global definition because of the way
3196 linkers work. See @code{section} attribute for more information.
3198 The @code{shared} attribute is only available on Microsoft Windows@.
3200 @item tls_model ("@var{tls_model}")
3201 @cindex @code{tls_model} attribute
3202 The @code{tls_model} attribute sets thread-local storage model
3203 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3204 overriding @option{-ftls-model=} command line switch on a per-variable
3206 The @var{tls_model} argument should be one of @code{global-dynamic},
3207 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3209 Not all targets support this attribute.
3212 This attribute, attached to a variable, means that the variable is meant
3213 to be possibly unused. GCC will not produce a warning for this
3217 This attribute, attached to a variable, means that the variable must be
3218 emitted even if it appears that the variable is not referenced.
3220 @item vector_size (@var{bytes})
3221 This attribute specifies the vector size for the variable, measured in
3222 bytes. For example, the declaration:
3225 int foo __attribute__ ((vector_size (16)));
3229 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3230 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3231 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3233 This attribute is only applicable to integral and float scalars,
3234 although arrays, pointers, and function return values are allowed in
3235 conjunction with this construct.
3237 Aggregates with this attribute are invalid, even if they are of the same
3238 size as a corresponding scalar. For example, the declaration:
3241 struct S @{ int a; @};
3242 struct S __attribute__ ((vector_size (16))) foo;
3246 is invalid even if the size of the structure is the same as the size of
3250 The @code{selectany} attribute causes an initialized global variable to
3251 have link-once semantics. When multiple definitions of the variable are
3252 encountered by the linker, the first is selected and the remainder are
3253 discarded. Following usage by the Microsoft compiler, the linker is told
3254 @emph{not} to warn about size or content differences of the multiple
3257 Although the primary usage of this attribute is for POD types, the
3258 attribute can also be applied to global C++ objects that are initialized
3259 by a constructor. In this case, the static initialization and destruction
3260 code for the object is emitted in each translation defining the object,
3261 but the calls to the constructor and destructor are protected by a
3262 link-once guard variable.
3264 The @code{selectany} attribute is only available on Microsoft Windows
3265 targets. You can use @code{__declspec (selectany)} as a synonym for
3266 @code{__attribute__ ((selectany))} for compatibility with other
3270 The @code{weak} attribute is described in @xref{Function Attributes}.
3273 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3276 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3280 @subsection M32R/D Variable Attributes
3282 One attribute is currently defined for the M32R/D@.
3285 @item model (@var{model-name})
3286 @cindex variable addressability on the M32R/D
3287 Use this attribute on the M32R/D to set the addressability of an object.
3288 The identifier @var{model-name} is one of @code{small}, @code{medium},
3289 or @code{large}, representing each of the code models.
3291 Small model objects live in the lower 16MB of memory (so that their
3292 addresses can be loaded with the @code{ld24} instruction).
3294 Medium and large model objects may live anywhere in the 32-bit address space
3295 (the compiler will generate @code{seth/add3} instructions to load their
3299 @anchor{i386 Variable Attributes}
3300 @subsection i386 Variable Attributes
3302 Two attributes are currently defined for i386 configurations:
3303 @code{ms_struct} and @code{gcc_struct}
3308 @cindex @code{ms_struct} attribute
3309 @cindex @code{gcc_struct} attribute
3311 If @code{packed} is used on a structure, or if bit-fields are used
3312 it may be that the Microsoft ABI packs them differently
3313 than GCC would normally pack them. Particularly when moving packed
3314 data between functions compiled with GCC and the native Microsoft compiler
3315 (either via function call or as data in a file), it may be necessary to access
3318 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3319 compilers to match the native Microsoft compiler.
3321 The Microsoft structure layout algorithm is fairly simple with the exception
3322 of the bitfield packing:
3324 The padding and alignment of members of structures and whether a bit field
3325 can straddle a storage-unit boundary
3328 @item Structure members are stored sequentially in the order in which they are
3329 declared: the first member has the lowest memory address and the last member
3332 @item Every data object has an alignment-requirement. The alignment-requirement
3333 for all data except structures, unions, and arrays is either the size of the
3334 object or the current packing size (specified with either the aligned attribute
3335 or the pack pragma), whichever is less. For structures, unions, and arrays,
3336 the alignment-requirement is the largest alignment-requirement of its members.
3337 Every object is allocated an offset so that:
3339 offset % alignment-requirement == 0
3341 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3342 unit if the integral types are the same size and if the next bit field fits
3343 into the current allocation unit without crossing the boundary imposed by the
3344 common alignment requirements of the bit fields.
3347 Handling of zero-length bitfields:
3349 MSVC interprets zero-length bitfields in the following ways:
3352 @item If a zero-length bitfield is inserted between two bitfields that would
3353 normally be coalesced, the bitfields will not be coalesced.
3360 unsigned long bf_1 : 12;
3362 unsigned long bf_2 : 12;
3366 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3367 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3369 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3370 alignment of the zero-length bitfield is greater than the member that follows it,
3371 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3391 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3392 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3393 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3396 Taking this into account, it is important to note the following:
3399 @item If a zero-length bitfield follows a normal bitfield, the type of the
3400 zero-length bitfield may affect the alignment of the structure as whole. For
3401 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3402 normal bitfield, and is of type short.
3404 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3405 still affect the alignment of the structure:
3415 Here, @code{t4} will take up 4 bytes.
3418 @item Zero-length bitfields following non-bitfield members are ignored:
3429 Here, @code{t5} will take up 2 bytes.
3433 @subsection PowerPC Variable Attributes
3435 Three attributes currently are defined for PowerPC configurations:
3436 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3438 For full documentation of the struct attributes please see the
3439 documentation in the @xref{i386 Variable Attributes}, section.
3441 For documentation of @code{altivec} attribute please see the
3442 documentation in the @xref{PowerPC Type Attributes}, section.
3444 @subsection Xstormy16 Variable Attributes
3446 One attribute is currently defined for xstormy16 configurations:
3451 @cindex @code{below100} attribute
3453 If a variable has the @code{below100} attribute (@code{BELOW100} is
3454 allowed also), GCC will place the variable in the first 0x100 bytes of
3455 memory and use special opcodes to access it. Such variables will be
3456 placed in either the @code{.bss_below100} section or the
3457 @code{.data_below100} section.
3461 @node Type Attributes
3462 @section Specifying Attributes of Types
3463 @cindex attribute of types
3464 @cindex type attributes
3466 The keyword @code{__attribute__} allows you to specify special
3467 attributes of @code{struct} and @code{union} types when you define
3468 such types. This keyword is followed by an attribute specification
3469 inside double parentheses. Seven attributes are currently defined for
3470 types: @code{aligned}, @code{packed}, @code{transparent_union},
3471 @code{unused}, @code{deprecated}, @code{visibility}, and
3472 @code{may_alias}. Other attributes are defined for functions
3473 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3476 You may also specify any one of these attributes with @samp{__}
3477 preceding and following its keyword. This allows you to use these
3478 attributes in header files without being concerned about a possible
3479 macro of the same name. For example, you may use @code{__aligned__}
3480 instead of @code{aligned}.
3482 You may specify type attributes either in a @code{typedef} declaration
3483 or in an enum, struct or union type declaration or definition.
3485 For an enum, struct or union type, you may specify attributes either
3486 between the enum, struct or union tag and the name of the type, or
3487 just past the closing curly brace of the @emph{definition}. The
3488 former syntax is preferred.
3490 @xref{Attribute Syntax}, for details of the exact syntax for using
3494 @cindex @code{aligned} attribute
3495 @item aligned (@var{alignment})
3496 This attribute specifies a minimum alignment (in bytes) for variables
3497 of the specified type. For example, the declarations:
3500 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3501 typedef int more_aligned_int __attribute__ ((aligned (8)));
3505 force the compiler to insure (as far as it can) that each variable whose
3506 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3507 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3508 variables of type @code{struct S} aligned to 8-byte boundaries allows
3509 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3510 store) instructions when copying one variable of type @code{struct S} to
3511 another, thus improving run-time efficiency.
3513 Note that the alignment of any given @code{struct} or @code{union} type
3514 is required by the ISO C standard to be at least a perfect multiple of
3515 the lowest common multiple of the alignments of all of the members of
3516 the @code{struct} or @code{union} in question. This means that you @emph{can}
3517 effectively adjust the alignment of a @code{struct} or @code{union}
3518 type by attaching an @code{aligned} attribute to any one of the members
3519 of such a type, but the notation illustrated in the example above is a
3520 more obvious, intuitive, and readable way to request the compiler to
3521 adjust the alignment of an entire @code{struct} or @code{union} type.
3523 As in the preceding example, you can explicitly specify the alignment
3524 (in bytes) that you wish the compiler to use for a given @code{struct}
3525 or @code{union} type. Alternatively, you can leave out the alignment factor
3526 and just ask the compiler to align a type to the maximum
3527 useful alignment for the target machine you are compiling for. For
3528 example, you could write:
3531 struct S @{ short f[3]; @} __attribute__ ((aligned));
3534 Whenever you leave out the alignment factor in an @code{aligned}
3535 attribute specification, the compiler automatically sets the alignment
3536 for the type to the largest alignment which is ever used for any data
3537 type on the target machine you are compiling for. Doing this can often
3538 make copy operations more efficient, because the compiler can use
3539 whatever instructions copy the biggest chunks of memory when performing
3540 copies to or from the variables which have types that you have aligned
3543 In the example above, if the size of each @code{short} is 2 bytes, then
3544 the size of the entire @code{struct S} type is 6 bytes. The smallest
3545 power of two which is greater than or equal to that is 8, so the
3546 compiler sets the alignment for the entire @code{struct S} type to 8
3549 Note that although you can ask the compiler to select a time-efficient
3550 alignment for a given type and then declare only individual stand-alone
3551 objects of that type, the compiler's ability to select a time-efficient
3552 alignment is primarily useful only when you plan to create arrays of
3553 variables having the relevant (efficiently aligned) type. If you
3554 declare or use arrays of variables of an efficiently-aligned type, then
3555 it is likely that your program will also be doing pointer arithmetic (or
3556 subscripting, which amounts to the same thing) on pointers to the
3557 relevant type, and the code that the compiler generates for these
3558 pointer arithmetic operations will often be more efficient for
3559 efficiently-aligned types than for other types.
3561 The @code{aligned} attribute can only increase the alignment; but you
3562 can decrease it by specifying @code{packed} as well. See below.
3564 Note that the effectiveness of @code{aligned} attributes may be limited
3565 by inherent limitations in your linker. On many systems, the linker is
3566 only able to arrange for variables to be aligned up to a certain maximum
3567 alignment. (For some linkers, the maximum supported alignment may
3568 be very very small.) If your linker is only able to align variables
3569 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3570 in an @code{__attribute__} will still only provide you with 8 byte
3571 alignment. See your linker documentation for further information.
3574 This attribute, attached to @code{struct} or @code{union} type
3575 definition, specifies that each member (other than zero-width bitfields)
3576 of the structure or union is placed to minimize the memory required. When
3577 attached to an @code{enum} definition, it indicates that the smallest
3578 integral type should be used.
3580 @opindex fshort-enums
3581 Specifying this attribute for @code{struct} and @code{union} types is
3582 equivalent to specifying the @code{packed} attribute on each of the
3583 structure or union members. Specifying the @option{-fshort-enums}
3584 flag on the line is equivalent to specifying the @code{packed}
3585 attribute on all @code{enum} definitions.
3587 In the following example @code{struct my_packed_struct}'s members are
3588 packed closely together, but the internal layout of its @code{s} member
3589 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3593 struct my_unpacked_struct
3599 struct __attribute__ ((__packed__)) my_packed_struct
3603 struct my_unpacked_struct s;
3607 You may only specify this attribute on the definition of a @code{enum},
3608 @code{struct} or @code{union}, not on a @code{typedef} which does not
3609 also define the enumerated type, structure or union.
3611 @item transparent_union
3612 This attribute, attached to a @code{union} type definition, indicates
3613 that any function parameter having that union type causes calls to that
3614 function to be treated in a special way.
3616 First, the argument corresponding to a transparent union type can be of
3617 any type in the union; no cast is required. Also, if the union contains
3618 a pointer type, the corresponding argument can be a null pointer
3619 constant or a void pointer expression; and if the union contains a void
3620 pointer type, the corresponding argument can be any pointer expression.
3621 If the union member type is a pointer, qualifiers like @code{const} on
3622 the referenced type must be respected, just as with normal pointer
3625 Second, the argument is passed to the function using the calling
3626 conventions of the first member of the transparent union, not the calling
3627 conventions of the union itself. All members of the union must have the
3628 same machine representation; this is necessary for this argument passing
3631 Transparent unions are designed for library functions that have multiple
3632 interfaces for compatibility reasons. For example, suppose the
3633 @code{wait} function must accept either a value of type @code{int *} to
3634 comply with Posix, or a value of type @code{union wait *} to comply with
3635 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3636 @code{wait} would accept both kinds of arguments, but it would also
3637 accept any other pointer type and this would make argument type checking
3638 less useful. Instead, @code{<sys/wait.h>} might define the interface
3646 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3648 pid_t wait (wait_status_ptr_t);
3651 This interface allows either @code{int *} or @code{union wait *}
3652 arguments to be passed, using the @code{int *} calling convention.
3653 The program can call @code{wait} with arguments of either type:
3656 int w1 () @{ int w; return wait (&w); @}
3657 int w2 () @{ union wait w; return wait (&w); @}
3660 With this interface, @code{wait}'s implementation might look like this:
3663 pid_t wait (wait_status_ptr_t p)
3665 return waitpid (-1, p.__ip, 0);
3670 When attached to a type (including a @code{union} or a @code{struct}),
3671 this attribute means that variables of that type are meant to appear
3672 possibly unused. GCC will not produce a warning for any variables of
3673 that type, even if the variable appears to do nothing. This is often
3674 the case with lock or thread classes, which are usually defined and then
3675 not referenced, but contain constructors and destructors that have
3676 nontrivial bookkeeping functions.
3679 The @code{deprecated} attribute results in a warning if the type
3680 is used anywhere in the source file. This is useful when identifying
3681 types that are expected to be removed in a future version of a program.
3682 If possible, the warning also includes the location of the declaration
3683 of the deprecated type, to enable users to easily find further
3684 information about why the type is deprecated, or what they should do
3685 instead. Note that the warnings only occur for uses and then only
3686 if the type is being applied to an identifier that itself is not being
3687 declared as deprecated.
3690 typedef int T1 __attribute__ ((deprecated));
3694 typedef T1 T3 __attribute__ ((deprecated));
3695 T3 z __attribute__ ((deprecated));
3698 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3699 warning is issued for line 4 because T2 is not explicitly
3700 deprecated. Line 5 has no warning because T3 is explicitly
3701 deprecated. Similarly for line 6.
3703 The @code{deprecated} attribute can also be used for functions and
3704 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3707 Accesses to objects with types with this attribute are not subjected to
3708 type-based alias analysis, but are instead assumed to be able to alias
3709 any other type of objects, just like the @code{char} type. See
3710 @option{-fstrict-aliasing} for more information on aliasing issues.
3715 typedef short __attribute__((__may_alias__)) short_a;
3721 short_a *b = (short_a *) &a;
3725 if (a == 0x12345678)
3732 If you replaced @code{short_a} with @code{short} in the variable
3733 declaration, the above program would abort when compiled with
3734 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3735 above in recent GCC versions.
3738 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3739 applied to class, struct, union and enum types. Unlike other type
3740 attributes, the attribute must appear between the initial keyword and
3741 the name of the type; it cannot appear after the body of the type.
3743 Note that the type visibility is applied to vague linkage entities
3744 associated with the class (vtable, typeinfo node, etc.). In
3745 particular, if a class is thrown as an exception in one shared object
3746 and caught in another, the class must have default visibility.
3747 Otherwise the two shared objects will be unable to use the same
3748 typeinfo node and exception handling will break.
3750 @subsection ARM Type Attributes
3752 On those ARM targets that support @code{dllimport} (such as Symbian
3753 OS), you can use the @code{notshared} attribute to indicate that the
3754 virtual table and other similar data for a class should not be
3755 exported from a DLL@. For example:
3758 class __declspec(notshared) C @{
3760 __declspec(dllimport) C();
3764 __declspec(dllexport)
3768 In this code, @code{C::C} is exported from the current DLL, but the
3769 virtual table for @code{C} is not exported. (You can use
3770 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3771 most Symbian OS code uses @code{__declspec}.)
3773 @anchor{i386 Type Attributes}
3774 @subsection i386 Type Attributes
3776 Two attributes are currently defined for i386 configurations:
3777 @code{ms_struct} and @code{gcc_struct}
3781 @cindex @code{ms_struct}
3782 @cindex @code{gcc_struct}
3784 If @code{packed} is used on a structure, or if bit-fields are used
3785 it may be that the Microsoft ABI packs them differently
3786 than GCC would normally pack them. Particularly when moving packed
3787 data between functions compiled with GCC and the native Microsoft compiler
3788 (either via function call or as data in a file), it may be necessary to access
3791 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3792 compilers to match the native Microsoft compiler.
3795 To specify multiple attributes, separate them by commas within the
3796 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3799 @anchor{PowerPC Type Attributes}
3800 @subsection PowerPC Type Attributes
3802 Three attributes currently are defined for PowerPC configurations:
3803 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3805 For full documentation of the struct attributes please see the
3806 documentation in the @xref{i386 Type Attributes}, section.
3808 The @code{altivec} attribute allows one to declare AltiVec vector data
3809 types supported by the AltiVec Programming Interface Manual. The
3810 attribute requires an argument to specify one of three vector types:
3811 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3812 and @code{bool__} (always followed by unsigned).
3815 __attribute__((altivec(vector__)))
3816 __attribute__((altivec(pixel__))) unsigned short
3817 __attribute__((altivec(bool__))) unsigned
3820 These attributes mainly are intended to support the @code{__vector},
3821 @code{__pixel}, and @code{__bool} AltiVec keywords.
3824 @section An Inline Function is As Fast As a Macro
3825 @cindex inline functions
3826 @cindex integrating function code
3828 @cindex macros, inline alternative
3830 By declaring a function inline, you can direct GCC to make
3831 calls to that function faster. One way GCC can achieve this is to
3832 integrate that function's code into the code for its callers. This
3833 makes execution faster by eliminating the function-call overhead; in
3834 addition, if any of the actual argument values are constant, their
3835 known values may permit simplifications at compile time so that not
3836 all of the inline function's code needs to be included. The effect on
3837 code size is less predictable; object code may be larger or smaller
3838 with function inlining, depending on the particular case. You can
3839 also direct GCC to try to integrate all ``simple enough'' functions
3840 into their callers with the option @option{-finline-functions}.
3842 GCC implements three different semantics of declaring a function
3843 inline. One is available with @option{-std=gnu89}, another when
3844 @option{-std=c99} or @option{-std=gnu99}, and the third is used when
3847 To declare a function inline, use the @code{inline} keyword in its
3848 declaration, like this:
3858 If you are writing a header file to be included in ISO C89 programs, write
3859 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3861 The three types of inlining behave similarly in two important cases:
3862 when the @code{inline} keyword is used on a @code{static} function,
3863 like the example above, and when a function is first declared without
3864 using the @code{inline} keyword and then is defined with
3865 @code{inline}, like this:
3868 extern int inc (int *a);
3876 In both of these common cases, the program behaves the same as if you
3877 had not used the @code{inline} keyword, except for its speed.
3879 @cindex inline functions, omission of
3880 @opindex fkeep-inline-functions
3881 When a function is both inline and @code{static}, if all calls to the
3882 function are integrated into the caller, and the function's address is
3883 never used, then the function's own assembler code is never referenced.
3884 In this case, GCC does not actually output assembler code for the
3885 function, unless you specify the option @option{-fkeep-inline-functions}.
3886 Some calls cannot be integrated for various reasons (in particular,
3887 calls that precede the function's definition cannot be integrated, and
3888 neither can recursive calls within the definition). If there is a
3889 nonintegrated call, then the function is compiled to assembler code as
3890 usual. The function must also be compiled as usual if the program
3891 refers to its address, because that can't be inlined.
3893 @cindex automatic @code{inline} for C++ member fns
3894 @cindex @code{inline} automatic for C++ member fns
3895 @cindex member fns, automatically @code{inline}
3896 @cindex C++ member fns, automatically @code{inline}
3897 @opindex fno-default-inline
3898 As required by ISO C++, GCC considers member functions defined within
3899 the body of a class to be marked inline even if they are
3900 not explicitly declared with the @code{inline} keyword. You can
3901 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
3902 Options,,Options Controlling C++ Dialect}.
3904 GCC does not inline any functions when not optimizing unless you specify
3905 the @samp{always_inline} attribute for the function, like this:
3908 /* @r{Prototype.} */
3909 inline void foo (const char) __attribute__((always_inline));
3912 The remainder of this section is specific to GNU C89 inlining.
3914 @cindex non-static inline function
3915 When an inline function is not @code{static}, then the compiler must assume
3916 that there may be calls from other source files; since a global symbol can
3917 be defined only once in any program, the function must not be defined in
3918 the other source files, so the calls therein cannot be integrated.
3919 Therefore, a non-@code{static} inline function is always compiled on its
3920 own in the usual fashion.
3922 If you specify both @code{inline} and @code{extern} in the function
3923 definition, then the definition is used only for inlining. In no case
3924 is the function compiled on its own, not even if you refer to its
3925 address explicitly. Such an address becomes an external reference, as
3926 if you had only declared the function, and had not defined it.
3928 This combination of @code{inline} and @code{extern} has almost the
3929 effect of a macro. The way to use it is to put a function definition in
3930 a header file with these keywords, and put another copy of the
3931 definition (lacking @code{inline} and @code{extern}) in a library file.
3932 The definition in the header file will cause most calls to the function
3933 to be inlined. If any uses of the function remain, they will refer to
3934 the single copy in the library.
3937 @section Assembler Instructions with C Expression Operands
3938 @cindex extended @code{asm}
3939 @cindex @code{asm} expressions
3940 @cindex assembler instructions
3943 In an assembler instruction using @code{asm}, you can specify the
3944 operands of the instruction using C expressions. This means you need not
3945 guess which registers or memory locations will contain the data you want
3948 You must specify an assembler instruction template much like what
3949 appears in a machine description, plus an operand constraint string for
3952 For example, here is how to use the 68881's @code{fsinx} instruction:
3955 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3959 Here @code{angle} is the C expression for the input operand while
3960 @code{result} is that of the output operand. Each has @samp{"f"} as its
3961 operand constraint, saying that a floating point register is required.
3962 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3963 output operands' constraints must use @samp{=}. The constraints use the
3964 same language used in the machine description (@pxref{Constraints}).
3966 Each operand is described by an operand-constraint string followed by
3967 the C expression in parentheses. A colon separates the assembler
3968 template from the first output operand and another separates the last
3969 output operand from the first input, if any. Commas separate the
3970 operands within each group. The total number of operands is currently
3971 limited to 30; this limitation may be lifted in some future version of
3974 If there are no output operands but there are input operands, you must
3975 place two consecutive colons surrounding the place where the output
3978 As of GCC version 3.1, it is also possible to specify input and output
3979 operands using symbolic names which can be referenced within the
3980 assembler code. These names are specified inside square brackets
3981 preceding the constraint string, and can be referenced inside the
3982 assembler code using @code{%[@var{name}]} instead of a percentage sign
3983 followed by the operand number. Using named operands the above example
3987 asm ("fsinx %[angle],%[output]"
3988 : [output] "=f" (result)
3989 : [angle] "f" (angle));
3993 Note that the symbolic operand names have no relation whatsoever to
3994 other C identifiers. You may use any name you like, even those of
3995 existing C symbols, but you must ensure that no two operands within the same
3996 assembler construct use the same symbolic name.
3998 Output operand expressions must be lvalues; the compiler can check this.
3999 The input operands need not be lvalues. The compiler cannot check
4000 whether the operands have data types that are reasonable for the
4001 instruction being executed. It does not parse the assembler instruction
4002 template and does not know what it means or even whether it is valid
4003 assembler input. The extended @code{asm} feature is most often used for
4004 machine instructions the compiler itself does not know exist. If
4005 the output expression cannot be directly addressed (for example, it is a
4006 bit-field), your constraint must allow a register. In that case, GCC
4007 will use the register as the output of the @code{asm}, and then store
4008 that register into the output.
4010 The ordinary output operands must be write-only; GCC will assume that
4011 the values in these operands before the instruction are dead and need
4012 not be generated. Extended asm supports input-output or read-write
4013 operands. Use the constraint character @samp{+} to indicate such an
4014 operand and list it with the output operands. You should only use
4015 read-write operands when the constraints for the operand (or the
4016 operand in which only some of the bits are to be changed) allow a
4019 You may, as an alternative, logically split its function into two
4020 separate operands, one input operand and one write-only output
4021 operand. The connection between them is expressed by constraints
4022 which say they need to be in the same location when the instruction
4023 executes. You can use the same C expression for both operands, or
4024 different expressions. For example, here we write the (fictitious)
4025 @samp{combine} instruction with @code{bar} as its read-only source
4026 operand and @code{foo} as its read-write destination:
4029 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4033 The constraint @samp{"0"} for operand 1 says that it must occupy the
4034 same location as operand 0. A number in constraint is allowed only in
4035 an input operand and it must refer to an output operand.
4037 Only a number in the constraint can guarantee that one operand will be in
4038 the same place as another. The mere fact that @code{foo} is the value
4039 of both operands is not enough to guarantee that they will be in the
4040 same place in the generated assembler code. The following would not
4044 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4047 Various optimizations or reloading could cause operands 0 and 1 to be in
4048 different registers; GCC knows no reason not to do so. For example, the
4049 compiler might find a copy of the value of @code{foo} in one register and
4050 use it for operand 1, but generate the output operand 0 in a different
4051 register (copying it afterward to @code{foo}'s own address). Of course,
4052 since the register for operand 1 is not even mentioned in the assembler
4053 code, the result will not work, but GCC can't tell that.
4055 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4056 the operand number for a matching constraint. For example:
4059 asm ("cmoveq %1,%2,%[result]"
4060 : [result] "=r"(result)
4061 : "r" (test), "r"(new), "[result]"(old));
4064 Sometimes you need to make an @code{asm} operand be a specific register,
4065 but there's no matching constraint letter for that register @emph{by
4066 itself}. To force the operand into that register, use a local variable
4067 for the operand and specify the register in the variable declaration.
4068 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4069 register constraint letter that matches the register:
4072 register int *p1 asm ("r0") = @dots{};
4073 register int *p2 asm ("r1") = @dots{};
4074 register int *result asm ("r0");
4075 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4078 @anchor{Example of asm with clobbered asm reg}
4079 In the above example, beware that a register that is call-clobbered by
4080 the target ABI will be overwritten by any function call in the
4081 assignment, including library calls for arithmetic operators.
4082 Assuming it is a call-clobbered register, this may happen to @code{r0}
4083 above by the assignment to @code{p2}. If you have to use such a
4084 register, use temporary variables for expressions between the register
4089 register int *p1 asm ("r0") = @dots{};
4090 register int *p2 asm ("r1") = t1;
4091 register int *result asm ("r0");
4092 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4095 Some instructions clobber specific hard registers. To describe this,
4096 write a third colon after the input operands, followed by the names of
4097 the clobbered hard registers (given as strings). Here is a realistic
4098 example for the VAX:
4101 asm volatile ("movc3 %0,%1,%2"
4102 : /* @r{no outputs} */
4103 : "g" (from), "g" (to), "g" (count)
4104 : "r0", "r1", "r2", "r3", "r4", "r5");
4107 You may not write a clobber description in a way that overlaps with an
4108 input or output operand. For example, you may not have an operand
4109 describing a register class with one member if you mention that register
4110 in the clobber list. Variables declared to live in specific registers
4111 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4112 have no part mentioned in the clobber description.
4113 There is no way for you to specify that an input
4114 operand is modified without also specifying it as an output
4115 operand. Note that if all the output operands you specify are for this
4116 purpose (and hence unused), you will then also need to specify
4117 @code{volatile} for the @code{asm} construct, as described below, to
4118 prevent GCC from deleting the @code{asm} statement as unused.
4120 If you refer to a particular hardware register from the assembler code,
4121 you will probably have to list the register after the third colon to
4122 tell the compiler the register's value is modified. In some assemblers,
4123 the register names begin with @samp{%}; to produce one @samp{%} in the
4124 assembler code, you must write @samp{%%} in the input.
4126 If your assembler instruction can alter the condition code register, add
4127 @samp{cc} to the list of clobbered registers. GCC on some machines
4128 represents the condition codes as a specific hardware register;
4129 @samp{cc} serves to name this register. On other machines, the
4130 condition code is handled differently, and specifying @samp{cc} has no
4131 effect. But it is valid no matter what the machine.
4133 If your assembler instructions access memory in an unpredictable
4134 fashion, add @samp{memory} to the list of clobbered registers. This
4135 will cause GCC to not keep memory values cached in registers across the
4136 assembler instruction and not optimize stores or loads to that memory.
4137 You will also want to add the @code{volatile} keyword if the memory
4138 affected is not listed in the inputs or outputs of the @code{asm}, as
4139 the @samp{memory} clobber does not count as a side-effect of the
4140 @code{asm}. If you know how large the accessed memory is, you can add
4141 it as input or output but if this is not known, you should add
4142 @samp{memory}. As an example, if you access ten bytes of a string, you
4143 can use a memory input like:
4146 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4149 Note that in the following example the memory input is necessary,
4150 otherwise GCC might optimize the store to @code{x} away:
4157 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4158 "=&d" (r) : "a" (y), "m" (*y));
4163 You can put multiple assembler instructions together in a single
4164 @code{asm} template, separated by the characters normally used in assembly
4165 code for the system. A combination that works in most places is a newline
4166 to break the line, plus a tab character to move to the instruction field
4167 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4168 assembler allows semicolons as a line-breaking character. Note that some
4169 assembler dialects use semicolons to start a comment.
4170 The input operands are guaranteed not to use any of the clobbered
4171 registers, and neither will the output operands' addresses, so you can
4172 read and write the clobbered registers as many times as you like. Here
4173 is an example of multiple instructions in a template; it assumes the
4174 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4177 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4179 : "g" (from), "g" (to)
4183 Unless an output operand has the @samp{&} constraint modifier, GCC
4184 may allocate it in the same register as an unrelated input operand, on
4185 the assumption the inputs are consumed before the outputs are produced.
4186 This assumption may be false if the assembler code actually consists of
4187 more than one instruction. In such a case, use @samp{&} for each output
4188 operand that may not overlap an input. @xref{Modifiers}.
4190 If you want to test the condition code produced by an assembler
4191 instruction, you must include a branch and a label in the @code{asm}
4192 construct, as follows:
4195 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4201 This assumes your assembler supports local labels, as the GNU assembler
4202 and most Unix assemblers do.
4204 Speaking of labels, jumps from one @code{asm} to another are not
4205 supported. The compiler's optimizers do not know about these jumps, and
4206 therefore they cannot take account of them when deciding how to
4209 @cindex macros containing @code{asm}
4210 Usually the most convenient way to use these @code{asm} instructions is to
4211 encapsulate them in macros that look like functions. For example,
4215 (@{ double __value, __arg = (x); \
4216 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4221 Here the variable @code{__arg} is used to make sure that the instruction
4222 operates on a proper @code{double} value, and to accept only those
4223 arguments @code{x} which can convert automatically to a @code{double}.
4225 Another way to make sure the instruction operates on the correct data
4226 type is to use a cast in the @code{asm}. This is different from using a
4227 variable @code{__arg} in that it converts more different types. For
4228 example, if the desired type were @code{int}, casting the argument to
4229 @code{int} would accept a pointer with no complaint, while assigning the
4230 argument to an @code{int} variable named @code{__arg} would warn about
4231 using a pointer unless the caller explicitly casts it.
4233 If an @code{asm} has output operands, GCC assumes for optimization
4234 purposes the instruction has no side effects except to change the output
4235 operands. This does not mean instructions with a side effect cannot be
4236 used, but you must be careful, because the compiler may eliminate them
4237 if the output operands aren't used, or move them out of loops, or
4238 replace two with one if they constitute a common subexpression. Also,
4239 if your instruction does have a side effect on a variable that otherwise
4240 appears not to change, the old value of the variable may be reused later
4241 if it happens to be found in a register.
4243 You can prevent an @code{asm} instruction from being deleted
4244 by writing the keyword @code{volatile} after
4245 the @code{asm}. For example:
4248 #define get_and_set_priority(new) \
4250 asm volatile ("get_and_set_priority %0, %1" \
4251 : "=g" (__old) : "g" (new)); \
4256 The @code{volatile} keyword indicates that the instruction has
4257 important side-effects. GCC will not delete a volatile @code{asm} if
4258 it is reachable. (The instruction can still be deleted if GCC can
4259 prove that control-flow will never reach the location of the
4260 instruction.) Note that even a volatile @code{asm} instruction
4261 can be moved relative to other code, including across jump
4262 instructions. For example, on many targets there is a system
4263 register which can be set to control the rounding mode of
4264 floating point operations. You might try
4265 setting it with a volatile @code{asm}, like this PowerPC example:
4268 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4273 This will not work reliably, as the compiler may move the addition back
4274 before the volatile @code{asm}. To make it work you need to add an
4275 artificial dependency to the @code{asm} referencing a variable in the code
4276 you don't want moved, for example:
4279 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4283 Similarly, you can't expect a
4284 sequence of volatile @code{asm} instructions to remain perfectly
4285 consecutive. If you want consecutive output, use a single @code{asm}.
4286 Also, GCC will perform some optimizations across a volatile @code{asm}
4287 instruction; GCC does not ``forget everything'' when it encounters
4288 a volatile @code{asm} instruction the way some other compilers do.
4290 An @code{asm} instruction without any output operands will be treated
4291 identically to a volatile @code{asm} instruction.
4293 It is a natural idea to look for a way to give access to the condition
4294 code left by the assembler instruction. However, when we attempted to
4295 implement this, we found no way to make it work reliably. The problem
4296 is that output operands might need reloading, which would result in
4297 additional following ``store'' instructions. On most machines, these
4298 instructions would alter the condition code before there was time to
4299 test it. This problem doesn't arise for ordinary ``test'' and
4300 ``compare'' instructions because they don't have any output operands.
4302 For reasons similar to those described above, it is not possible to give
4303 an assembler instruction access to the condition code left by previous
4306 If you are writing a header file that should be includable in ISO C
4307 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4310 @subsection Size of an @code{asm}
4312 Some targets require that GCC track the size of each instruction used in
4313 order to generate correct code. Because the final length of an
4314 @code{asm} is only known by the assembler, GCC must make an estimate as
4315 to how big it will be. The estimate is formed by counting the number of
4316 statements in the pattern of the @code{asm} and multiplying that by the
4317 length of the longest instruction on that processor. Statements in the
4318 @code{asm} are identified by newline characters and whatever statement
4319 separator characters are supported by the assembler; on most processors
4320 this is the `@code{;}' character.
4322 Normally, GCC's estimate is perfectly adequate to ensure that correct
4323 code is generated, but it is possible to confuse the compiler if you use
4324 pseudo instructions or assembler macros that expand into multiple real
4325 instructions or if you use assembler directives that expand to more
4326 space in the object file than would be needed for a single instruction.
4327 If this happens then the assembler will produce a diagnostic saying that
4328 a label is unreachable.
4330 @subsection i386 floating point asm operands
4332 There are several rules on the usage of stack-like regs in
4333 asm_operands insns. These rules apply only to the operands that are
4338 Given a set of input regs that die in an asm_operands, it is
4339 necessary to know which are implicitly popped by the asm, and
4340 which must be explicitly popped by gcc.
4342 An input reg that is implicitly popped by the asm must be
4343 explicitly clobbered, unless it is constrained to match an
4347 For any input reg that is implicitly popped by an asm, it is
4348 necessary to know how to adjust the stack to compensate for the pop.
4349 If any non-popped input is closer to the top of the reg-stack than
4350 the implicitly popped reg, it would not be possible to know what the
4351 stack looked like---it's not clear how the rest of the stack ``slides
4354 All implicitly popped input regs must be closer to the top of
4355 the reg-stack than any input that is not implicitly popped.
4357 It is possible that if an input dies in an insn, reload might
4358 use the input reg for an output reload. Consider this example:
4361 asm ("foo" : "=t" (a) : "f" (b));
4364 This asm says that input B is not popped by the asm, and that
4365 the asm pushes a result onto the reg-stack, i.e., the stack is one
4366 deeper after the asm than it was before. But, it is possible that
4367 reload will think that it can use the same reg for both the input and
4368 the output, if input B dies in this insn.
4370 If any input operand uses the @code{f} constraint, all output reg
4371 constraints must use the @code{&} earlyclobber.
4373 The asm above would be written as
4376 asm ("foo" : "=&t" (a) : "f" (b));
4380 Some operands need to be in particular places on the stack. All
4381 output operands fall in this category---there is no other way to
4382 know which regs the outputs appear in unless the user indicates
4383 this in the constraints.
4385 Output operands must specifically indicate which reg an output
4386 appears in after an asm. @code{=f} is not allowed: the operand
4387 constraints must select a class with a single reg.
4390 Output operands may not be ``inserted'' between existing stack regs.
4391 Since no 387 opcode uses a read/write operand, all output operands
4392 are dead before the asm_operands, and are pushed by the asm_operands.
4393 It makes no sense to push anywhere but the top of the reg-stack.
4395 Output operands must start at the top of the reg-stack: output
4396 operands may not ``skip'' a reg.
4399 Some asm statements may need extra stack space for internal
4400 calculations. This can be guaranteed by clobbering stack registers
4401 unrelated to the inputs and outputs.
4405 Here are a couple of reasonable asms to want to write. This asm
4406 takes one input, which is internally popped, and produces two outputs.
4409 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4412 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4413 and replaces them with one output. The user must code the @code{st(1)}
4414 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4417 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4423 @section Controlling Names Used in Assembler Code
4424 @cindex assembler names for identifiers
4425 @cindex names used in assembler code
4426 @cindex identifiers, names in assembler code
4428 You can specify the name to be used in the assembler code for a C
4429 function or variable by writing the @code{asm} (or @code{__asm__})
4430 keyword after the declarator as follows:
4433 int foo asm ("myfoo") = 2;
4437 This specifies that the name to be used for the variable @code{foo} in
4438 the assembler code should be @samp{myfoo} rather than the usual
4441 On systems where an underscore is normally prepended to the name of a C
4442 function or variable, this feature allows you to define names for the
4443 linker that do not start with an underscore.
4445 It does not make sense to use this feature with a non-static local
4446 variable since such variables do not have assembler names. If you are
4447 trying to put the variable in a particular register, see @ref{Explicit
4448 Reg Vars}. GCC presently accepts such code with a warning, but will
4449 probably be changed to issue an error, rather than a warning, in the
4452 You cannot use @code{asm} in this way in a function @emph{definition}; but
4453 you can get the same effect by writing a declaration for the function
4454 before its definition and putting @code{asm} there, like this:
4457 extern func () asm ("FUNC");
4464 It is up to you to make sure that the assembler names you choose do not
4465 conflict with any other assembler symbols. Also, you must not use a
4466 register name; that would produce completely invalid assembler code. GCC
4467 does not as yet have the ability to store static variables in registers.
4468 Perhaps that will be added.
4470 @node Explicit Reg Vars
4471 @section Variables in Specified Registers
4472 @cindex explicit register variables
4473 @cindex variables in specified registers
4474 @cindex specified registers
4475 @cindex registers, global allocation
4477 GNU C allows you to put a few global variables into specified hardware
4478 registers. You can also specify the register in which an ordinary
4479 register variable should be allocated.
4483 Global register variables reserve registers throughout the program.
4484 This may be useful in programs such as programming language
4485 interpreters which have a couple of global variables that are accessed
4489 Local register variables in specific registers do not reserve the
4490 registers, except at the point where they are used as input or output
4491 operands in an @code{asm} statement and the @code{asm} statement itself is
4492 not deleted. The compiler's data flow analysis is capable of determining
4493 where the specified registers contain live values, and where they are
4494 available for other uses. Stores into local register variables may be deleted
4495 when they appear to be dead according to dataflow analysis. References
4496 to local register variables may be deleted or moved or simplified.
4498 These local variables are sometimes convenient for use with the extended
4499 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4500 output of the assembler instruction directly into a particular register.
4501 (This will work provided the register you specify fits the constraints
4502 specified for that operand in the @code{asm}.)
4510 @node Global Reg Vars
4511 @subsection Defining Global Register Variables
4512 @cindex global register variables
4513 @cindex registers, global variables in
4515 You can define a global register variable in GNU C like this:
4518 register int *foo asm ("a5");
4522 Here @code{a5} is the name of the register which should be used. Choose a
4523 register which is normally saved and restored by function calls on your
4524 machine, so that library routines will not clobber it.
4526 Naturally the register name is cpu-dependent, so you would need to
4527 conditionalize your program according to cpu type. The register
4528 @code{a5} would be a good choice on a 68000 for a variable of pointer
4529 type. On machines with register windows, be sure to choose a ``global''
4530 register that is not affected magically by the function call mechanism.
4532 In addition, operating systems on one type of cpu may differ in how they
4533 name the registers; then you would need additional conditionals. For
4534 example, some 68000 operating systems call this register @code{%a5}.
4536 Eventually there may be a way of asking the compiler to choose a register
4537 automatically, but first we need to figure out how it should choose and
4538 how to enable you to guide the choice. No solution is evident.
4540 Defining a global register variable in a certain register reserves that
4541 register entirely for this use, at least within the current compilation.
4542 The register will not be allocated for any other purpose in the functions
4543 in the current compilation. The register will not be saved and restored by
4544 these functions. Stores into this register are never deleted even if they
4545 would appear to be dead, but references may be deleted or moved or
4548 It is not safe to access the global register variables from signal
4549 handlers, or from more than one thread of control, because the system
4550 library routines may temporarily use the register for other things (unless
4551 you recompile them specially for the task at hand).
4553 @cindex @code{qsort}, and global register variables
4554 It is not safe for one function that uses a global register variable to
4555 call another such function @code{foo} by way of a third function
4556 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4557 different source file in which the variable wasn't declared). This is
4558 because @code{lose} might save the register and put some other value there.
4559 For example, you can't expect a global register variable to be available in
4560 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4561 might have put something else in that register. (If you are prepared to
4562 recompile @code{qsort} with the same global register variable, you can
4563 solve this problem.)
4565 If you want to recompile @code{qsort} or other source files which do not
4566 actually use your global register variable, so that they will not use that
4567 register for any other purpose, then it suffices to specify the compiler
4568 option @option{-ffixed-@var{reg}}. You need not actually add a global
4569 register declaration to their source code.
4571 A function which can alter the value of a global register variable cannot
4572 safely be called from a function compiled without this variable, because it
4573 could clobber the value the caller expects to find there on return.
4574 Therefore, the function which is the entry point into the part of the
4575 program that uses the global register variable must explicitly save and
4576 restore the value which belongs to its caller.
4578 @cindex register variable after @code{longjmp}
4579 @cindex global register after @code{longjmp}
4580 @cindex value after @code{longjmp}
4583 On most machines, @code{longjmp} will restore to each global register
4584 variable the value it had at the time of the @code{setjmp}. On some
4585 machines, however, @code{longjmp} will not change the value of global
4586 register variables. To be portable, the function that called @code{setjmp}
4587 should make other arrangements to save the values of the global register
4588 variables, and to restore them in a @code{longjmp}. This way, the same
4589 thing will happen regardless of what @code{longjmp} does.
4591 All global register variable declarations must precede all function
4592 definitions. If such a declaration could appear after function
4593 definitions, the declaration would be too late to prevent the register from
4594 being used for other purposes in the preceding functions.
4596 Global register variables may not have initial values, because an
4597 executable file has no means to supply initial contents for a register.
4599 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4600 registers, but certain library functions, such as @code{getwd}, as well
4601 as the subroutines for division and remainder, modify g3 and g4. g1 and
4602 g2 are local temporaries.
4604 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4605 Of course, it will not do to use more than a few of those.
4607 @node Local Reg Vars
4608 @subsection Specifying Registers for Local Variables
4609 @cindex local variables, specifying registers
4610 @cindex specifying registers for local variables
4611 @cindex registers for local variables
4613 You can define a local register variable with a specified register
4617 register int *foo asm ("a5");
4621 Here @code{a5} is the name of the register which should be used. Note
4622 that this is the same syntax used for defining global register
4623 variables, but for a local variable it would appear within a function.
4625 Naturally the register name is cpu-dependent, but this is not a
4626 problem, since specific registers are most often useful with explicit
4627 assembler instructions (@pxref{Extended Asm}). Both of these things
4628 generally require that you conditionalize your program according to
4631 In addition, operating systems on one type of cpu may differ in how they
4632 name the registers; then you would need additional conditionals. For
4633 example, some 68000 operating systems call this register @code{%a5}.
4635 Defining such a register variable does not reserve the register; it
4636 remains available for other uses in places where flow control determines
4637 the variable's value is not live.
4639 This option does not guarantee that GCC will generate code that has
4640 this variable in the register you specify at all times. You may not
4641 code an explicit reference to this register in the @emph{assembler
4642 instruction template} part of an @code{asm} statement and assume it will
4643 always refer to this variable. However, using the variable as an
4644 @code{asm} @emph{operand} guarantees that the specified register is used
4647 Stores into local register variables may be deleted when they appear to be dead
4648 according to dataflow analysis. References to local register variables may
4649 be deleted or moved or simplified.
4651 As for global register variables, it's recommended that you choose a
4652 register which is normally saved and restored by function calls on
4653 your machine, so that library routines will not clobber it. A common
4654 pitfall is to initialize multiple call-clobbered registers with
4655 arbitrary expressions, where a function call or library call for an
4656 arithmetic operator will overwrite a register value from a previous
4657 assignment, for example @code{r0} below:
4659 register int *p1 asm ("r0") = @dots{};
4660 register int *p2 asm ("r1") = @dots{};
4662 In those cases, a solution is to use a temporary variable for
4663 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4665 @node Alternate Keywords
4666 @section Alternate Keywords
4667 @cindex alternate keywords
4668 @cindex keywords, alternate
4670 @option{-ansi} and the various @option{-std} options disable certain
4671 keywords. This causes trouble when you want to use GNU C extensions, or
4672 a general-purpose header file that should be usable by all programs,
4673 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4674 @code{inline} are not available in programs compiled with
4675 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4676 program compiled with @option{-std=c99}). The ISO C99 keyword
4677 @code{restrict} is only available when @option{-std=gnu99} (which will
4678 eventually be the default) or @option{-std=c99} (or the equivalent
4679 @option{-std=iso9899:1999}) is used.
4681 The way to solve these problems is to put @samp{__} at the beginning and
4682 end of each problematical keyword. For example, use @code{__asm__}
4683 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4685 Other C compilers won't accept these alternative keywords; if you want to
4686 compile with another compiler, you can define the alternate keywords as
4687 macros to replace them with the customary keywords. It looks like this:
4695 @findex __extension__
4697 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4699 prevent such warnings within one expression by writing
4700 @code{__extension__} before the expression. @code{__extension__} has no
4701 effect aside from this.
4703 @node Incomplete Enums
4704 @section Incomplete @code{enum} Types
4706 You can define an @code{enum} tag without specifying its possible values.
4707 This results in an incomplete type, much like what you get if you write
4708 @code{struct foo} without describing the elements. A later declaration
4709 which does specify the possible values completes the type.
4711 You can't allocate variables or storage using the type while it is
4712 incomplete. However, you can work with pointers to that type.
4714 This extension may not be very useful, but it makes the handling of
4715 @code{enum} more consistent with the way @code{struct} and @code{union}
4718 This extension is not supported by GNU C++.
4720 @node Function Names
4721 @section Function Names as Strings
4722 @cindex @code{__func__} identifier
4723 @cindex @code{__FUNCTION__} identifier
4724 @cindex @code{__PRETTY_FUNCTION__} identifier
4726 GCC provides three magic variables which hold the name of the current
4727 function, as a string. The first of these is @code{__func__}, which
4728 is part of the C99 standard:
4731 The identifier @code{__func__} is implicitly declared by the translator
4732 as if, immediately following the opening brace of each function
4733 definition, the declaration
4736 static const char __func__[] = "function-name";
4739 appeared, where function-name is the name of the lexically-enclosing
4740 function. This name is the unadorned name of the function.
4743 @code{__FUNCTION__} is another name for @code{__func__}. Older
4744 versions of GCC recognize only this name. However, it is not
4745 standardized. For maximum portability, we recommend you use
4746 @code{__func__}, but provide a fallback definition with the
4750 #if __STDC_VERSION__ < 199901L
4752 # define __func__ __FUNCTION__
4754 # define __func__ "<unknown>"
4759 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4760 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4761 the type signature of the function as well as its bare name. For
4762 example, this program:
4766 extern int printf (char *, ...);
4773 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4774 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4792 __PRETTY_FUNCTION__ = void a::sub(int)
4795 These identifiers are not preprocessor macros. In GCC 3.3 and
4796 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4797 were treated as string literals; they could be used to initialize
4798 @code{char} arrays, and they could be concatenated with other string
4799 literals. GCC 3.4 and later treat them as variables, like
4800 @code{__func__}. In C++, @code{__FUNCTION__} and
4801 @code{__PRETTY_FUNCTION__} have always been variables.
4803 @node Return Address
4804 @section Getting the Return or Frame Address of a Function
4806 These functions may be used to get information about the callers of a
4809 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4810 This function returns the return address of the current function, or of
4811 one of its callers. The @var{level} argument is number of frames to
4812 scan up the call stack. A value of @code{0} yields the return address
4813 of the current function, a value of @code{1} yields the return address
4814 of the caller of the current function, and so forth. When inlining
4815 the expected behavior is that the function will return the address of
4816 the function that will be returned to. To work around this behavior use
4817 the @code{noinline} function attribute.
4819 The @var{level} argument must be a constant integer.
4821 On some machines it may be impossible to determine the return address of
4822 any function other than the current one; in such cases, or when the top
4823 of the stack has been reached, this function will return @code{0} or a
4824 random value. In addition, @code{__builtin_frame_address} may be used
4825 to determine if the top of the stack has been reached.
4827 This function should only be used with a nonzero argument for debugging
4831 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4832 This function is similar to @code{__builtin_return_address}, but it
4833 returns the address of the function frame rather than the return address
4834 of the function. Calling @code{__builtin_frame_address} with a value of
4835 @code{0} yields the frame address of the current function, a value of
4836 @code{1} yields the frame address of the caller of the current function,
4839 The frame is the area on the stack which holds local variables and saved
4840 registers. The frame address is normally the address of the first word
4841 pushed on to the stack by the function. However, the exact definition
4842 depends upon the processor and the calling convention. If the processor
4843 has a dedicated frame pointer register, and the function has a frame,
4844 then @code{__builtin_frame_address} will return the value of the frame
4847 On some machines it may be impossible to determine the frame address of
4848 any function other than the current one; in such cases, or when the top
4849 of the stack has been reached, this function will return @code{0} if
4850 the first frame pointer is properly initialized by the startup code.
4852 This function should only be used with a nonzero argument for debugging
4856 @node Vector Extensions
4857 @section Using vector instructions through built-in functions
4859 On some targets, the instruction set contains SIMD vector instructions that
4860 operate on multiple values contained in one large register at the same time.
4861 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4864 The first step in using these extensions is to provide the necessary data
4865 types. This should be done using an appropriate @code{typedef}:
4868 typedef int v4si __attribute__ ((vector_size (16)));
4871 The @code{int} type specifies the base type, while the attribute specifies
4872 the vector size for the variable, measured in bytes. For example, the
4873 declaration above causes the compiler to set the mode for the @code{v4si}
4874 type to be 16 bytes wide and divided into @code{int} sized units. For
4875 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4876 corresponding mode of @code{foo} will be @acronym{V4SI}.
4878 The @code{vector_size} attribute is only applicable to integral and
4879 float scalars, although arrays, pointers, and function return values
4880 are allowed in conjunction with this construct.
4882 All the basic integer types can be used as base types, both as signed
4883 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4884 @code{long long}. In addition, @code{float} and @code{double} can be
4885 used to build floating-point vector types.
4887 Specifying a combination that is not valid for the current architecture
4888 will cause GCC to synthesize the instructions using a narrower mode.
4889 For example, if you specify a variable of type @code{V4SI} and your
4890 architecture does not allow for this specific SIMD type, GCC will
4891 produce code that uses 4 @code{SIs}.
4893 The types defined in this manner can be used with a subset of normal C
4894 operations. Currently, GCC will allow using the following operators
4895 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4897 The operations behave like C++ @code{valarrays}. Addition is defined as
4898 the addition of the corresponding elements of the operands. For
4899 example, in the code below, each of the 4 elements in @var{a} will be
4900 added to the corresponding 4 elements in @var{b} and the resulting
4901 vector will be stored in @var{c}.
4904 typedef int v4si __attribute__ ((vector_size (16)));
4911 Subtraction, multiplication, division, and the logical operations
4912 operate in a similar manner. Likewise, the result of using the unary
4913 minus or complement operators on a vector type is a vector whose
4914 elements are the negative or complemented values of the corresponding
4915 elements in the operand.
4917 You can declare variables and use them in function calls and returns, as
4918 well as in assignments and some casts. You can specify a vector type as
4919 a return type for a function. Vector types can also be used as function
4920 arguments. It is possible to cast from one vector type to another,
4921 provided they are of the same size (in fact, you can also cast vectors
4922 to and from other datatypes of the same size).
4924 You cannot operate between vectors of different lengths or different
4925 signedness without a cast.
4927 A port that supports hardware vector operations, usually provides a set
4928 of built-in functions that can be used to operate on vectors. For
4929 example, a function to add two vectors and multiply the result by a
4930 third could look like this:
4933 v4si f (v4si a, v4si b, v4si c)
4935 v4si tmp = __builtin_addv4si (a, b);
4936 return __builtin_mulv4si (tmp, c);
4943 @findex __builtin_offsetof
4945 GCC implements for both C and C++ a syntactic extension to implement
4946 the @code{offsetof} macro.
4950 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4952 offsetof_member_designator:
4954 | offsetof_member_designator "." @code{identifier}
4955 | offsetof_member_designator "[" @code{expr} "]"
4958 This extension is sufficient such that
4961 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4964 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4965 may be dependent. In either case, @var{member} may consist of a single
4966 identifier, or a sequence of member accesses and array references.
4968 @node Atomic Builtins
4969 @section Built-in functions for atomic memory access
4971 The following builtins are intended to be compatible with those described
4972 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
4973 section 7.4. As such, they depart from the normal GCC practice of using
4974 the ``__builtin_'' prefix, and further that they are overloaded such that
4975 they work on multiple types.
4977 The definition given in the Intel documentation allows only for the use of
4978 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
4979 counterparts. GCC will allow any integral scalar or pointer type that is
4980 1, 2, 4 or 8 bytes in length.
4982 Not all operations are supported by all target processors. If a particular
4983 operation cannot be implemented on the target processor, a warning will be
4984 generated and a call an external function will be generated. The external
4985 function will carry the same name as the builtin, with an additional suffix
4986 @samp{_@var{n}} where @var{n} is the size of the data type.
4988 @c ??? Should we have a mechanism to suppress this warning? This is almost
4989 @c useful for implementing the operation under the control of an external
4992 In most cases, these builtins are considered a @dfn{full barrier}. That is,
4993 no memory operand will be moved across the operation, either forward or
4994 backward. Further, instructions will be issued as necessary to prevent the
4995 processor from speculating loads across the operation and from queuing stores
4996 after the operation.
4998 All of the routines are are described in the Intel documentation to take
4999 ``an optional list of variables protected by the memory barrier''. It's
5000 not clear what is meant by that; it could mean that @emph{only} the
5001 following variables are protected, or it could mean that these variables
5002 should in addition be protected. At present GCC ignores this list and
5003 protects all variables which are globally accessible. If in the future
5004 we make some use of this list, an empty list will continue to mean all
5005 globally accessible variables.
5008 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5009 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5010 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5011 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5012 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5013 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5014 @findex __sync_fetch_and_add
5015 @findex __sync_fetch_and_sub
5016 @findex __sync_fetch_and_or
5017 @findex __sync_fetch_and_and
5018 @findex __sync_fetch_and_xor
5019 @findex __sync_fetch_and_nand
5020 These builtins perform the operation suggested by the name, and
5021 returns the value that had previously been in memory. That is,
5024 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5025 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5028 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5029 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5030 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5031 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5032 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5033 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5034 @findex __sync_add_and_fetch
5035 @findex __sync_sub_and_fetch
5036 @findex __sync_or_and_fetch
5037 @findex __sync_and_and_fetch
5038 @findex __sync_xor_and_fetch
5039 @findex __sync_nand_and_fetch
5040 These builtins perform the operation suggested by the name, and
5041 return the new value. That is,
5044 @{ *ptr @var{op}= value; return *ptr; @}
5045 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5048 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5049 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5050 @findex __sync_bool_compare_and_swap
5051 @findex __sync_val_compare_and_swap
5052 These builtins perform an atomic compare and swap. That is, if the current
5053 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5056 The ``bool'' version returns true if the comparison is successful and
5057 @var{newval} was written. The ``val'' version returns the contents
5058 of @code{*@var{ptr}} before the operation.
5060 @item __sync_synchronize (...)
5061 @findex __sync_synchronize
5062 This builtin issues a full memory barrier.
5064 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5065 @findex __sync_lock_test_and_set
5066 This builtin, as described by Intel, is not a traditional test-and-set
5067 operation, but rather an atomic exchange operation. It writes @var{value}
5068 into @code{*@var{ptr}}, and returns the previous contents of
5071 Many targets have only minimal support for such locks, and do not support
5072 a full exchange operation. In this case, a target may support reduced
5073 functionality here by which the @emph{only} valid value to store is the
5074 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5075 is implementation defined.
5077 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5078 This means that references after the builtin cannot move to (or be
5079 speculated to) before the builtin, but previous memory stores may not
5080 be globally visible yet, and previous memory loads may not yet be
5083 @item void __sync_lock_release (@var{type} *ptr, ...)
5084 @findex __sync_lock_release
5085 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5086 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5088 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5089 This means that all previous memory stores are globally visible, and all
5090 previous memory loads have been satisfied, but following memory reads
5091 are not prevented from being speculated to before the barrier.
5094 @node Object Size Checking
5095 @section Object Size Checking Builtins
5096 @findex __builtin_object_size
5097 @findex __builtin___memcpy_chk
5098 @findex __builtin___mempcpy_chk
5099 @findex __builtin___memmove_chk
5100 @findex __builtin___memset_chk
5101 @findex __builtin___strcpy_chk
5102 @findex __builtin___stpcpy_chk
5103 @findex __builtin___strncpy_chk
5104 @findex __builtin___strcat_chk
5105 @findex __builtin___strncat_chk
5106 @findex __builtin___sprintf_chk
5107 @findex __builtin___snprintf_chk
5108 @findex __builtin___vsprintf_chk
5109 @findex __builtin___vsnprintf_chk
5110 @findex __builtin___printf_chk
5111 @findex __builtin___vprintf_chk
5112 @findex __builtin___fprintf_chk
5113 @findex __builtin___vfprintf_chk
5115 GCC implements a limited buffer overflow protection mechanism
5116 that can prevent some buffer overflow attacks.
5118 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5119 is a built-in construct that returns a constant number of bytes from
5120 @var{ptr} to the end of the object @var{ptr} pointer points to
5121 (if known at compile time). @code{__builtin_object_size} never evaluates
5122 its arguments for side-effects. If there are any side-effects in them, it
5123 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5124 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5125 point to and all of them are known at compile time, the returned number
5126 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5127 0 and minimum if nonzero. If it is not possible to determine which objects
5128 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5129 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5130 for @var{type} 2 or 3.
5132 @var{type} is an integer constant from 0 to 3. If the least significant
5133 bit is clear, objects are whole variables, if it is set, a closest
5134 surrounding subobject is considered the object a pointer points to.
5135 The second bit determines if maximum or minimum of remaining bytes
5139 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5140 char *p = &var.buf1[1], *q = &var.b;
5142 /* Here the object p points to is var. */
5143 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5144 /* The subobject p points to is var.buf1. */
5145 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5146 /* The object q points to is var. */
5147 assert (__builtin_object_size (q, 0)
5148 == (char *) (&var + 1) - (char *) &var.b);
5149 /* The subobject q points to is var.b. */
5150 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5154 There are built-in functions added for many common string operation
5155 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5156 built-in is provided. This built-in has an additional last argument,
5157 which is the number of bytes remaining in object the @var{dest}
5158 argument points to or @code{(size_t) -1} if the size is not known.
5160 The built-in functions are optimized into the normal string functions
5161 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5162 it is known at compile time that the destination object will not
5163 be overflown. If the compiler can determine at compile time the
5164 object will be always overflown, it issues a warning.
5166 The intended use can be e.g.
5170 #define bos0(dest) __builtin_object_size (dest, 0)
5171 #define memcpy(dest, src, n) \
5172 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5176 /* It is unknown what object p points to, so this is optimized
5177 into plain memcpy - no checking is possible. */
5178 memcpy (p, "abcde", n);
5179 /* Destination is known and length too. It is known at compile
5180 time there will be no overflow. */
5181 memcpy (&buf[5], "abcde", 5);
5182 /* Destination is known, but the length is not known at compile time.
5183 This will result in __memcpy_chk call that can check for overflow
5185 memcpy (&buf[5], "abcde", n);
5186 /* Destination is known and it is known at compile time there will
5187 be overflow. There will be a warning and __memcpy_chk call that
5188 will abort the program at runtime. */
5189 memcpy (&buf[6], "abcde", 5);
5192 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5193 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5194 @code{strcat} and @code{strncat}.
5196 There are also checking built-in functions for formatted output functions.
5198 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5199 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5200 const char *fmt, ...);
5201 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5203 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5204 const char *fmt, va_list ap);
5207 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5208 etc. functions and can contain implementation specific flags on what
5209 additional security measures the checking function might take, such as
5210 handling @code{%n} differently.
5212 The @var{os} argument is the object size @var{s} points to, like in the
5213 other built-in functions. There is a small difference in the behavior
5214 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5215 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5216 the checking function is called with @var{os} argument set to
5219 In addition to this, there are checking built-in functions
5220 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5221 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5222 These have just one additional argument, @var{flag}, right before
5223 format string @var{fmt}. If the compiler is able to optimize them to
5224 @code{fputc} etc. functions, it will, otherwise the checking function
5225 should be called and the @var{flag} argument passed to it.
5227 @node Other Builtins
5228 @section Other built-in functions provided by GCC
5229 @cindex built-in functions
5230 @findex __builtin_isgreater
5231 @findex __builtin_isgreaterequal
5232 @findex __builtin_isless
5233 @findex __builtin_islessequal
5234 @findex __builtin_islessgreater
5235 @findex __builtin_isunordered
5236 @findex __builtin_powi
5237 @findex __builtin_powif
5238 @findex __builtin_powil
5396 @findex fprintf_unlocked
5398 @findex fputs_unlocked
5508 @findex printf_unlocked
5537 @findex significandf
5538 @findex significandl
5609 GCC provides a large number of built-in functions other than the ones
5610 mentioned above. Some of these are for internal use in the processing
5611 of exceptions or variable-length argument lists and will not be
5612 documented here because they may change from time to time; we do not
5613 recommend general use of these functions.
5615 The remaining functions are provided for optimization purposes.
5617 @opindex fno-builtin
5618 GCC includes built-in versions of many of the functions in the standard
5619 C library. The versions prefixed with @code{__builtin_} will always be
5620 treated as having the same meaning as the C library function even if you
5621 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5622 Many of these functions are only optimized in certain cases; if they are
5623 not optimized in a particular case, a call to the library function will
5628 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5629 @option{-std=c99}), the functions
5630 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5631 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5632 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5633 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5634 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5635 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5636 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5637 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5638 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5639 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5640 @code{significandf}, @code{significandl}, @code{significand},
5641 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5642 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5643 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5644 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5645 @code{ynl} and @code{yn}
5646 may be handled as built-in functions.
5647 All these functions have corresponding versions
5648 prefixed with @code{__builtin_}, which may be used even in strict C89
5651 The ISO C99 functions
5652 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5653 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5654 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5655 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5656 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5657 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5658 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5659 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5660 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5661 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5662 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5663 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5664 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5665 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5666 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5667 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5668 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5669 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5670 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5671 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5672 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5673 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5674 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5675 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5676 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5677 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5678 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5679 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5680 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5681 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5682 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5683 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5684 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5685 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5686 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5687 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5688 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5689 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5690 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5691 are handled as built-in functions
5692 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5694 There are also built-in versions of the ISO C99 functions
5695 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5696 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5697 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5698 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5699 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5700 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5701 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5702 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5703 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5704 that are recognized in any mode since ISO C90 reserves these names for
5705 the purpose to which ISO C99 puts them. All these functions have
5706 corresponding versions prefixed with @code{__builtin_}.
5708 The ISO C94 functions
5709 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5710 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5711 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5713 are handled as built-in functions
5714 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5716 The ISO C90 functions
5717 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5718 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5719 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5720 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5721 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5722 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5723 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5724 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5725 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5726 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5727 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5728 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5729 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5730 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5731 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5732 @code{vprintf} and @code{vsprintf}
5733 are all recognized as built-in functions unless
5734 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5735 is specified for an individual function). All of these functions have
5736 corresponding versions prefixed with @code{__builtin_}.
5738 GCC provides built-in versions of the ISO C99 floating point comparison
5739 macros that avoid raising exceptions for unordered operands. They have
5740 the same names as the standard macros ( @code{isgreater},
5741 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5742 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5743 prefixed. We intend for a library implementor to be able to simply
5744 @code{#define} each standard macro to its built-in equivalent.
5746 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5748 You can use the built-in function @code{__builtin_types_compatible_p} to
5749 determine whether two types are the same.
5751 This built-in function returns 1 if the unqualified versions of the
5752 types @var{type1} and @var{type2} (which are types, not expressions) are
5753 compatible, 0 otherwise. The result of this built-in function can be
5754 used in integer constant expressions.
5756 This built-in function ignores top level qualifiers (e.g., @code{const},
5757 @code{volatile}). For example, @code{int} is equivalent to @code{const
5760 The type @code{int[]} and @code{int[5]} are compatible. On the other
5761 hand, @code{int} and @code{char *} are not compatible, even if the size
5762 of their types, on the particular architecture are the same. Also, the
5763 amount of pointer indirection is taken into account when determining
5764 similarity. Consequently, @code{short *} is not similar to
5765 @code{short **}. Furthermore, two types that are typedefed are
5766 considered compatible if their underlying types are compatible.
5768 An @code{enum} type is not considered to be compatible with another
5769 @code{enum} type even if both are compatible with the same integer
5770 type; this is what the C standard specifies.
5771 For example, @code{enum @{foo, bar@}} is not similar to
5772 @code{enum @{hot, dog@}}.
5774 You would typically use this function in code whose execution varies
5775 depending on the arguments' types. For example:
5780 typeof (x) tmp = (x); \
5781 if (__builtin_types_compatible_p (typeof (x), long double)) \
5782 tmp = foo_long_double (tmp); \
5783 else if (__builtin_types_compatible_p (typeof (x), double)) \
5784 tmp = foo_double (tmp); \
5785 else if (__builtin_types_compatible_p (typeof (x), float)) \
5786 tmp = foo_float (tmp); \
5793 @emph{Note:} This construct is only available for C@.
5797 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5799 You can use the built-in function @code{__builtin_choose_expr} to
5800 evaluate code depending on the value of a constant expression. This
5801 built-in function returns @var{exp1} if @var{const_exp}, which is a
5802 constant expression that must be able to be determined at compile time,
5803 is nonzero. Otherwise it returns 0.
5805 This built-in function is analogous to the @samp{? :} operator in C,
5806 except that the expression returned has its type unaltered by promotion
5807 rules. Also, the built-in function does not evaluate the expression
5808 that was not chosen. For example, if @var{const_exp} evaluates to true,
5809 @var{exp2} is not evaluated even if it has side-effects.
5811 This built-in function can return an lvalue if the chosen argument is an
5814 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5815 type. Similarly, if @var{exp2} is returned, its return type is the same
5822 __builtin_choose_expr ( \
5823 __builtin_types_compatible_p (typeof (x), double), \
5825 __builtin_choose_expr ( \
5826 __builtin_types_compatible_p (typeof (x), float), \
5828 /* @r{The void expression results in a compile-time error} \
5829 @r{when assigning the result to something.} */ \
5833 @emph{Note:} This construct is only available for C@. Furthermore, the
5834 unused expression (@var{exp1} or @var{exp2} depending on the value of
5835 @var{const_exp}) may still generate syntax errors. This may change in
5840 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5841 You can use the built-in function @code{__builtin_constant_p} to
5842 determine if a value is known to be constant at compile-time and hence
5843 that GCC can perform constant-folding on expressions involving that
5844 value. The argument of the function is the value to test. The function
5845 returns the integer 1 if the argument is known to be a compile-time
5846 constant and 0 if it is not known to be a compile-time constant. A
5847 return of 0 does not indicate that the value is @emph{not} a constant,
5848 but merely that GCC cannot prove it is a constant with the specified
5849 value of the @option{-O} option.
5851 You would typically use this function in an embedded application where
5852 memory was a critical resource. If you have some complex calculation,
5853 you may want it to be folded if it involves constants, but need to call
5854 a function if it does not. For example:
5857 #define Scale_Value(X) \
5858 (__builtin_constant_p (X) \
5859 ? ((X) * SCALE + OFFSET) : Scale (X))
5862 You may use this built-in function in either a macro or an inline
5863 function. However, if you use it in an inlined function and pass an
5864 argument of the function as the argument to the built-in, GCC will
5865 never return 1 when you call the inline function with a string constant
5866 or compound literal (@pxref{Compound Literals}) and will not return 1
5867 when you pass a constant numeric value to the inline function unless you
5868 specify the @option{-O} option.
5870 You may also use @code{__builtin_constant_p} in initializers for static
5871 data. For instance, you can write
5874 static const int table[] = @{
5875 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5881 This is an acceptable initializer even if @var{EXPRESSION} is not a
5882 constant expression. GCC must be more conservative about evaluating the
5883 built-in in this case, because it has no opportunity to perform
5886 Previous versions of GCC did not accept this built-in in data
5887 initializers. The earliest version where it is completely safe is
5891 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5892 @opindex fprofile-arcs
5893 You may use @code{__builtin_expect} to provide the compiler with
5894 branch prediction information. In general, you should prefer to
5895 use actual profile feedback for this (@option{-fprofile-arcs}), as
5896 programmers are notoriously bad at predicting how their programs
5897 actually perform. However, there are applications in which this
5898 data is hard to collect.
5900 The return value is the value of @var{exp}, which should be an
5901 integral expression. The value of @var{c} must be a compile-time
5902 constant. The semantics of the built-in are that it is expected
5903 that @var{exp} == @var{c}. For example:
5906 if (__builtin_expect (x, 0))
5911 would indicate that we do not expect to call @code{foo}, since
5912 we expect @code{x} to be zero. Since you are limited to integral
5913 expressions for @var{exp}, you should use constructions such as
5916 if (__builtin_expect (ptr != NULL, 1))
5921 when testing pointer or floating-point values.
5924 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5925 This function is used to minimize cache-miss latency by moving data into
5926 a cache before it is accessed.
5927 You can insert calls to @code{__builtin_prefetch} into code for which
5928 you know addresses of data in memory that is likely to be accessed soon.
5929 If the target supports them, data prefetch instructions will be generated.
5930 If the prefetch is done early enough before the access then the data will
5931 be in the cache by the time it is accessed.
5933 The value of @var{addr} is the address of the memory to prefetch.
5934 There are two optional arguments, @var{rw} and @var{locality}.
5935 The value of @var{rw} is a compile-time constant one or zero; one
5936 means that the prefetch is preparing for a write to the memory address
5937 and zero, the default, means that the prefetch is preparing for a read.
5938 The value @var{locality} must be a compile-time constant integer between
5939 zero and three. A value of zero means that the data has no temporal
5940 locality, so it need not be left in the cache after the access. A value
5941 of three means that the data has a high degree of temporal locality and
5942 should be left in all levels of cache possible. Values of one and two
5943 mean, respectively, a low or moderate degree of temporal locality. The
5947 for (i = 0; i < n; i++)
5950 __builtin_prefetch (&a[i+j], 1, 1);
5951 __builtin_prefetch (&b[i+j], 0, 1);
5956 Data prefetch does not generate faults if @var{addr} is invalid, but
5957 the address expression itself must be valid. For example, a prefetch
5958 of @code{p->next} will not fault if @code{p->next} is not a valid
5959 address, but evaluation will fault if @code{p} is not a valid address.
5961 If the target does not support data prefetch, the address expression
5962 is evaluated if it includes side effects but no other code is generated
5963 and GCC does not issue a warning.
5966 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5967 Returns a positive infinity, if supported by the floating-point format,
5968 else @code{DBL_MAX}. This function is suitable for implementing the
5969 ISO C macro @code{HUGE_VAL}.
5972 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5973 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5976 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5977 Similar to @code{__builtin_huge_val}, except the return
5978 type is @code{long double}.
5981 @deftypefn {Built-in Function} double __builtin_inf (void)
5982 Similar to @code{__builtin_huge_val}, except a warning is generated
5983 if the target floating-point format does not support infinities.
5986 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
5987 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
5990 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
5991 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
5994 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
5995 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
5998 @deftypefn {Built-in Function} float __builtin_inff (void)
5999 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6000 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6003 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6004 Similar to @code{__builtin_inf}, except the return
6005 type is @code{long double}.
6008 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6009 This is an implementation of the ISO C99 function @code{nan}.
6011 Since ISO C99 defines this function in terms of @code{strtod}, which we
6012 do not implement, a description of the parsing is in order. The string
6013 is parsed as by @code{strtol}; that is, the base is recognized by
6014 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6015 in the significand such that the least significant bit of the number
6016 is at the least significant bit of the significand. The number is
6017 truncated to fit the significand field provided. The significand is
6018 forced to be a quiet NaN@.
6020 This function, if given a string literal all of which would have been
6021 consumed by strtol, is evaluated early enough that it is considered a
6022 compile-time constant.
6025 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6026 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6029 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6030 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6033 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6034 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6037 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6038 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6041 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6042 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6045 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6046 Similar to @code{__builtin_nan}, except the significand is forced
6047 to be a signaling NaN@. The @code{nans} function is proposed by
6048 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6051 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6052 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6055 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6056 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6059 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6060 Returns one plus the index of the least significant 1-bit of @var{x}, or
6061 if @var{x} is zero, returns zero.
6064 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6065 Returns the number of leading 0-bits in @var{x}, starting at the most
6066 significant bit position. If @var{x} is 0, the result is undefined.
6069 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6070 Returns the number of trailing 0-bits in @var{x}, starting at the least
6071 significant bit position. If @var{x} is 0, the result is undefined.
6074 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6075 Returns the number of 1-bits in @var{x}.
6078 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6079 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6083 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6084 Similar to @code{__builtin_ffs}, except the argument type is
6085 @code{unsigned long}.
6088 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6089 Similar to @code{__builtin_clz}, except the argument type is
6090 @code{unsigned long}.
6093 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6094 Similar to @code{__builtin_ctz}, except the argument type is
6095 @code{unsigned long}.
6098 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6099 Similar to @code{__builtin_popcount}, except the argument type is
6100 @code{unsigned long}.
6103 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6104 Similar to @code{__builtin_parity}, except the argument type is
6105 @code{unsigned long}.
6108 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6109 Similar to @code{__builtin_ffs}, except the argument type is
6110 @code{unsigned long long}.
6113 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6114 Similar to @code{__builtin_clz}, except the argument type is
6115 @code{unsigned long long}.
6118 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6119 Similar to @code{__builtin_ctz}, except the argument type is
6120 @code{unsigned long long}.
6123 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6124 Similar to @code{__builtin_popcount}, except the argument type is
6125 @code{unsigned long long}.
6128 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6129 Similar to @code{__builtin_parity}, except the argument type is
6130 @code{unsigned long long}.
6133 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6134 Returns the first argument raised to the power of the second. Unlike the
6135 @code{pow} function no guarantees about precision and rounding are made.
6138 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6139 Similar to @code{__builtin_powi}, except the argument and return types
6143 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6144 Similar to @code{__builtin_powi}, except the argument and return types
6145 are @code{long double}.
6149 @node Target Builtins
6150 @section Built-in Functions Specific to Particular Target Machines
6152 On some target machines, GCC supports many built-in functions specific
6153 to those machines. Generally these generate calls to specific machine
6154 instructions, but allow the compiler to schedule those calls.
6157 * Alpha Built-in Functions::
6158 * ARM Built-in Functions::
6159 * Blackfin Built-in Functions::
6160 * FR-V Built-in Functions::
6161 * X86 Built-in Functions::
6162 * MIPS DSP Built-in Functions::
6163 * MIPS Paired-Single Support::
6164 * PowerPC AltiVec Built-in Functions::
6165 * SPARC VIS Built-in Functions::
6168 @node Alpha Built-in Functions
6169 @subsection Alpha Built-in Functions
6171 These built-in functions are available for the Alpha family of
6172 processors, depending on the command-line switches used.
6174 The following built-in functions are always available. They
6175 all generate the machine instruction that is part of the name.
6178 long __builtin_alpha_implver (void)
6179 long __builtin_alpha_rpcc (void)
6180 long __builtin_alpha_amask (long)
6181 long __builtin_alpha_cmpbge (long, long)
6182 long __builtin_alpha_extbl (long, long)
6183 long __builtin_alpha_extwl (long, long)
6184 long __builtin_alpha_extll (long, long)
6185 long __builtin_alpha_extql (long, long)
6186 long __builtin_alpha_extwh (long, long)
6187 long __builtin_alpha_extlh (long, long)
6188 long __builtin_alpha_extqh (long, long)
6189 long __builtin_alpha_insbl (long, long)
6190 long __builtin_alpha_inswl (long, long)
6191 long __builtin_alpha_insll (long, long)
6192 long __builtin_alpha_insql (long, long)
6193 long __builtin_alpha_inswh (long, long)
6194 long __builtin_alpha_inslh (long, long)
6195 long __builtin_alpha_insqh (long, long)
6196 long __builtin_alpha_mskbl (long, long)
6197 long __builtin_alpha_mskwl (long, long)
6198 long __builtin_alpha_mskll (long, long)
6199 long __builtin_alpha_mskql (long, long)
6200 long __builtin_alpha_mskwh (long, long)
6201 long __builtin_alpha_msklh (long, long)
6202 long __builtin_alpha_mskqh (long, long)
6203 long __builtin_alpha_umulh (long, long)
6204 long __builtin_alpha_zap (long, long)
6205 long __builtin_alpha_zapnot (long, long)
6208 The following built-in functions are always with @option{-mmax}
6209 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6210 later. They all generate the machine instruction that is part
6214 long __builtin_alpha_pklb (long)
6215 long __builtin_alpha_pkwb (long)
6216 long __builtin_alpha_unpkbl (long)
6217 long __builtin_alpha_unpkbw (long)
6218 long __builtin_alpha_minub8 (long, long)
6219 long __builtin_alpha_minsb8 (long, long)
6220 long __builtin_alpha_minuw4 (long, long)
6221 long __builtin_alpha_minsw4 (long, long)
6222 long __builtin_alpha_maxub8 (long, long)
6223 long __builtin_alpha_maxsb8 (long, long)
6224 long __builtin_alpha_maxuw4 (long, long)
6225 long __builtin_alpha_maxsw4 (long, long)
6226 long __builtin_alpha_perr (long, long)
6229 The following built-in functions are always with @option{-mcix}
6230 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6231 later. They all generate the machine instruction that is part
6235 long __builtin_alpha_cttz (long)
6236 long __builtin_alpha_ctlz (long)
6237 long __builtin_alpha_ctpop (long)
6240 The following builtins are available on systems that use the OSF/1
6241 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6242 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6243 @code{rdval} and @code{wrval}.
6246 void *__builtin_thread_pointer (void)
6247 void __builtin_set_thread_pointer (void *)
6250 @node ARM Built-in Functions
6251 @subsection ARM Built-in Functions
6253 These built-in functions are available for the ARM family of
6254 processors, when the @option{-mcpu=iwmmxt} switch is used:
6257 typedef int v2si __attribute__ ((vector_size (8)));
6258 typedef short v4hi __attribute__ ((vector_size (8)));
6259 typedef char v8qi __attribute__ ((vector_size (8)));
6261 int __builtin_arm_getwcx (int)
6262 void __builtin_arm_setwcx (int, int)
6263 int __builtin_arm_textrmsb (v8qi, int)
6264 int __builtin_arm_textrmsh (v4hi, int)
6265 int __builtin_arm_textrmsw (v2si, int)
6266 int __builtin_arm_textrmub (v8qi, int)
6267 int __builtin_arm_textrmuh (v4hi, int)
6268 int __builtin_arm_textrmuw (v2si, int)
6269 v8qi __builtin_arm_tinsrb (v8qi, int)
6270 v4hi __builtin_arm_tinsrh (v4hi, int)
6271 v2si __builtin_arm_tinsrw (v2si, int)
6272 long long __builtin_arm_tmia (long long, int, int)
6273 long long __builtin_arm_tmiabb (long long, int, int)
6274 long long __builtin_arm_tmiabt (long long, int, int)
6275 long long __builtin_arm_tmiaph (long long, int, int)
6276 long long __builtin_arm_tmiatb (long long, int, int)
6277 long long __builtin_arm_tmiatt (long long, int, int)
6278 int __builtin_arm_tmovmskb (v8qi)
6279 int __builtin_arm_tmovmskh (v4hi)
6280 int __builtin_arm_tmovmskw (v2si)
6281 long long __builtin_arm_waccb (v8qi)
6282 long long __builtin_arm_wacch (v4hi)
6283 long long __builtin_arm_waccw (v2si)
6284 v8qi __builtin_arm_waddb (v8qi, v8qi)
6285 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6286 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6287 v4hi __builtin_arm_waddh (v4hi, v4hi)
6288 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6289 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6290 v2si __builtin_arm_waddw (v2si, v2si)
6291 v2si __builtin_arm_waddwss (v2si, v2si)
6292 v2si __builtin_arm_waddwus (v2si, v2si)
6293 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6294 long long __builtin_arm_wand(long long, long long)
6295 long long __builtin_arm_wandn (long long, long long)
6296 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6297 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6298 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6299 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6300 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6301 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6302 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6303 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6304 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6305 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6306 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6307 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6308 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6309 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6310 long long __builtin_arm_wmacsz (v4hi, v4hi)
6311 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6312 long long __builtin_arm_wmacuz (v4hi, v4hi)
6313 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6314 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6315 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6316 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6317 v2si __builtin_arm_wmaxsw (v2si, v2si)
6318 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6319 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6320 v2si __builtin_arm_wmaxuw (v2si, v2si)
6321 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6322 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6323 v2si __builtin_arm_wminsw (v2si, v2si)
6324 v8qi __builtin_arm_wminub (v8qi, v8qi)
6325 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6326 v2si __builtin_arm_wminuw (v2si, v2si)
6327 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6328 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6329 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6330 long long __builtin_arm_wor (long long, long long)
6331 v2si __builtin_arm_wpackdss (long long, long long)
6332 v2si __builtin_arm_wpackdus (long long, long long)
6333 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6334 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6335 v4hi __builtin_arm_wpackwss (v2si, v2si)
6336 v4hi __builtin_arm_wpackwus (v2si, v2si)
6337 long long __builtin_arm_wrord (long long, long long)
6338 long long __builtin_arm_wrordi (long long, int)
6339 v4hi __builtin_arm_wrorh (v4hi, long long)
6340 v4hi __builtin_arm_wrorhi (v4hi, int)
6341 v2si __builtin_arm_wrorw (v2si, long long)
6342 v2si __builtin_arm_wrorwi (v2si, int)
6343 v2si __builtin_arm_wsadb (v8qi, v8qi)
6344 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6345 v2si __builtin_arm_wsadh (v4hi, v4hi)
6346 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6347 v4hi __builtin_arm_wshufh (v4hi, int)
6348 long long __builtin_arm_wslld (long long, long long)
6349 long long __builtin_arm_wslldi (long long, int)
6350 v4hi __builtin_arm_wsllh (v4hi, long long)
6351 v4hi __builtin_arm_wsllhi (v4hi, int)
6352 v2si __builtin_arm_wsllw (v2si, long long)
6353 v2si __builtin_arm_wsllwi (v2si, int)
6354 long long __builtin_arm_wsrad (long long, long long)
6355 long long __builtin_arm_wsradi (long long, int)
6356 v4hi __builtin_arm_wsrah (v4hi, long long)
6357 v4hi __builtin_arm_wsrahi (v4hi, int)
6358 v2si __builtin_arm_wsraw (v2si, long long)
6359 v2si __builtin_arm_wsrawi (v2si, int)
6360 long long __builtin_arm_wsrld (long long, long long)
6361 long long __builtin_arm_wsrldi (long long, int)
6362 v4hi __builtin_arm_wsrlh (v4hi, long long)
6363 v4hi __builtin_arm_wsrlhi (v4hi, int)
6364 v2si __builtin_arm_wsrlw (v2si, long long)
6365 v2si __builtin_arm_wsrlwi (v2si, int)
6366 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6367 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6368 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6369 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6370 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6371 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6372 v2si __builtin_arm_wsubw (v2si, v2si)
6373 v2si __builtin_arm_wsubwss (v2si, v2si)
6374 v2si __builtin_arm_wsubwus (v2si, v2si)
6375 v4hi __builtin_arm_wunpckehsb (v8qi)
6376 v2si __builtin_arm_wunpckehsh (v4hi)
6377 long long __builtin_arm_wunpckehsw (v2si)
6378 v4hi __builtin_arm_wunpckehub (v8qi)
6379 v2si __builtin_arm_wunpckehuh (v4hi)
6380 long long __builtin_arm_wunpckehuw (v2si)
6381 v4hi __builtin_arm_wunpckelsb (v8qi)
6382 v2si __builtin_arm_wunpckelsh (v4hi)
6383 long long __builtin_arm_wunpckelsw (v2si)
6384 v4hi __builtin_arm_wunpckelub (v8qi)
6385 v2si __builtin_arm_wunpckeluh (v4hi)
6386 long long __builtin_arm_wunpckeluw (v2si)
6387 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6388 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6389 v2si __builtin_arm_wunpckihw (v2si, v2si)
6390 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6391 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6392 v2si __builtin_arm_wunpckilw (v2si, v2si)
6393 long long __builtin_arm_wxor (long long, long long)
6394 long long __builtin_arm_wzero ()
6397 @node Blackfin Built-in Functions
6398 @subsection Blackfin Built-in Functions
6400 Currently, there are two Blackfin-specific built-in functions. These are
6401 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6402 using inline assembly; by using these built-in functions the compiler can
6403 automatically add workarounds for hardware errata involving these
6404 instructions. These functions are named as follows:
6407 void __builtin_bfin_csync (void)
6408 void __builtin_bfin_ssync (void)
6411 @node FR-V Built-in Functions
6412 @subsection FR-V Built-in Functions
6414 GCC provides many FR-V-specific built-in functions. In general,
6415 these functions are intended to be compatible with those described
6416 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6417 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6418 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6419 pointer rather than by value.
6421 Most of the functions are named after specific FR-V instructions.
6422 Such functions are said to be ``directly mapped'' and are summarized
6423 here in tabular form.
6427 * Directly-mapped Integer Functions::
6428 * Directly-mapped Media Functions::
6429 * Raw read/write Functions::
6430 * Other Built-in Functions::
6433 @node Argument Types
6434 @subsubsection Argument Types
6436 The arguments to the built-in functions can be divided into three groups:
6437 register numbers, compile-time constants and run-time values. In order
6438 to make this classification clear at a glance, the arguments and return
6439 values are given the following pseudo types:
6441 @multitable @columnfractions .20 .30 .15 .35
6442 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6443 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6444 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6445 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6446 @item @code{uw2} @tab @code{unsigned long long} @tab No
6447 @tab an unsigned doubleword
6448 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6449 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6450 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6451 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6454 These pseudo types are not defined by GCC, they are simply a notational
6455 convenience used in this manual.
6457 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6458 and @code{sw2} are evaluated at run time. They correspond to
6459 register operands in the underlying FR-V instructions.
6461 @code{const} arguments represent immediate operands in the underlying
6462 FR-V instructions. They must be compile-time constants.
6464 @code{acc} arguments are evaluated at compile time and specify the number
6465 of an accumulator register. For example, an @code{acc} argument of 2
6466 will select the ACC2 register.
6468 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6469 number of an IACC register. See @pxref{Other Built-in Functions}
6472 @node Directly-mapped Integer Functions
6473 @subsubsection Directly-mapped Integer Functions
6475 The functions listed below map directly to FR-V I-type instructions.
6477 @multitable @columnfractions .45 .32 .23
6478 @item Function prototype @tab Example usage @tab Assembly output
6479 @item @code{sw1 __ADDSS (sw1, sw1)}
6480 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6481 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6482 @item @code{sw1 __SCAN (sw1, sw1)}
6483 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6484 @tab @code{SCAN @var{a},@var{b},@var{c}}
6485 @item @code{sw1 __SCUTSS (sw1)}
6486 @tab @code{@var{b} = __SCUTSS (@var{a})}
6487 @tab @code{SCUTSS @var{a},@var{b}}
6488 @item @code{sw1 __SLASS (sw1, sw1)}
6489 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6490 @tab @code{SLASS @var{a},@var{b},@var{c}}
6491 @item @code{void __SMASS (sw1, sw1)}
6492 @tab @code{__SMASS (@var{a}, @var{b})}
6493 @tab @code{SMASS @var{a},@var{b}}
6494 @item @code{void __SMSSS (sw1, sw1)}
6495 @tab @code{__SMSSS (@var{a}, @var{b})}
6496 @tab @code{SMSSS @var{a},@var{b}}
6497 @item @code{void __SMU (sw1, sw1)}
6498 @tab @code{__SMU (@var{a}, @var{b})}
6499 @tab @code{SMU @var{a},@var{b}}
6500 @item @code{sw2 __SMUL (sw1, sw1)}
6501 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6502 @tab @code{SMUL @var{a},@var{b},@var{c}}
6503 @item @code{sw1 __SUBSS (sw1, sw1)}
6504 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6505 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6506 @item @code{uw2 __UMUL (uw1, uw1)}
6507 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6508 @tab @code{UMUL @var{a},@var{b},@var{c}}
6511 @node Directly-mapped Media Functions
6512 @subsubsection Directly-mapped Media Functions
6514 The functions listed below map directly to FR-V M-type instructions.
6516 @multitable @columnfractions .45 .32 .23
6517 @item Function prototype @tab Example usage @tab Assembly output
6518 @item @code{uw1 __MABSHS (sw1)}
6519 @tab @code{@var{b} = __MABSHS (@var{a})}
6520 @tab @code{MABSHS @var{a},@var{b}}
6521 @item @code{void __MADDACCS (acc, acc)}
6522 @tab @code{__MADDACCS (@var{b}, @var{a})}
6523 @tab @code{MADDACCS @var{a},@var{b}}
6524 @item @code{sw1 __MADDHSS (sw1, sw1)}
6525 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6526 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6527 @item @code{uw1 __MADDHUS (uw1, uw1)}
6528 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6529 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6530 @item @code{uw1 __MAND (uw1, uw1)}
6531 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6532 @tab @code{MAND @var{a},@var{b},@var{c}}
6533 @item @code{void __MASACCS (acc, acc)}
6534 @tab @code{__MASACCS (@var{b}, @var{a})}
6535 @tab @code{MASACCS @var{a},@var{b}}
6536 @item @code{uw1 __MAVEH (uw1, uw1)}
6537 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6538 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6539 @item @code{uw2 __MBTOH (uw1)}
6540 @tab @code{@var{b} = __MBTOH (@var{a})}
6541 @tab @code{MBTOH @var{a},@var{b}}
6542 @item @code{void __MBTOHE (uw1 *, uw1)}
6543 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6544 @tab @code{MBTOHE @var{a},@var{b}}
6545 @item @code{void __MCLRACC (acc)}
6546 @tab @code{__MCLRACC (@var{a})}
6547 @tab @code{MCLRACC @var{a}}
6548 @item @code{void __MCLRACCA (void)}
6549 @tab @code{__MCLRACCA ()}
6550 @tab @code{MCLRACCA}
6551 @item @code{uw1 __Mcop1 (uw1, uw1)}
6552 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6553 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6554 @item @code{uw1 __Mcop2 (uw1, uw1)}
6555 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6556 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6557 @item @code{uw1 __MCPLHI (uw2, const)}
6558 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6559 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6560 @item @code{uw1 __MCPLI (uw2, const)}
6561 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6562 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6563 @item @code{void __MCPXIS (acc, sw1, sw1)}
6564 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6565 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6566 @item @code{void __MCPXIU (acc, uw1, uw1)}
6567 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6568 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6569 @item @code{void __MCPXRS (acc, sw1, sw1)}
6570 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6571 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6572 @item @code{void __MCPXRU (acc, uw1, uw1)}
6573 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6574 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6575 @item @code{uw1 __MCUT (acc, uw1)}
6576 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6577 @tab @code{MCUT @var{a},@var{b},@var{c}}
6578 @item @code{uw1 __MCUTSS (acc, sw1)}
6579 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6580 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6581 @item @code{void __MDADDACCS (acc, acc)}
6582 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6583 @tab @code{MDADDACCS @var{a},@var{b}}
6584 @item @code{void __MDASACCS (acc, acc)}
6585 @tab @code{__MDASACCS (@var{b}, @var{a})}
6586 @tab @code{MDASACCS @var{a},@var{b}}
6587 @item @code{uw2 __MDCUTSSI (acc, const)}
6588 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6589 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6590 @item @code{uw2 __MDPACKH (uw2, uw2)}
6591 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6592 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6593 @item @code{uw2 __MDROTLI (uw2, const)}
6594 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6595 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6596 @item @code{void __MDSUBACCS (acc, acc)}
6597 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6598 @tab @code{MDSUBACCS @var{a},@var{b}}
6599 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6600 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6601 @tab @code{MDUNPACKH @var{a},@var{b}}
6602 @item @code{uw2 __MEXPDHD (uw1, const)}
6603 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6604 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6605 @item @code{uw1 __MEXPDHW (uw1, const)}
6606 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6607 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6608 @item @code{uw1 __MHDSETH (uw1, const)}
6609 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6610 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6611 @item @code{sw1 __MHDSETS (const)}
6612 @tab @code{@var{b} = __MHDSETS (@var{a})}
6613 @tab @code{MHDSETS #@var{a},@var{b}}
6614 @item @code{uw1 __MHSETHIH (uw1, const)}
6615 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6616 @tab @code{MHSETHIH #@var{a},@var{b}}
6617 @item @code{sw1 __MHSETHIS (sw1, const)}
6618 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6619 @tab @code{MHSETHIS #@var{a},@var{b}}
6620 @item @code{uw1 __MHSETLOH (uw1, const)}
6621 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6622 @tab @code{MHSETLOH #@var{a},@var{b}}
6623 @item @code{sw1 __MHSETLOS (sw1, const)}
6624 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6625 @tab @code{MHSETLOS #@var{a},@var{b}}
6626 @item @code{uw1 __MHTOB (uw2)}
6627 @tab @code{@var{b} = __MHTOB (@var{a})}
6628 @tab @code{MHTOB @var{a},@var{b}}
6629 @item @code{void __MMACHS (acc, sw1, sw1)}
6630 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6631 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6632 @item @code{void __MMACHU (acc, uw1, uw1)}
6633 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6634 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6635 @item @code{void __MMRDHS (acc, sw1, sw1)}
6636 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6637 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6638 @item @code{void __MMRDHU (acc, uw1, uw1)}
6639 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6640 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6641 @item @code{void __MMULHS (acc, sw1, sw1)}
6642 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6643 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6644 @item @code{void __MMULHU (acc, uw1, uw1)}
6645 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6646 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6647 @item @code{void __MMULXHS (acc, sw1, sw1)}
6648 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6649 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6650 @item @code{void __MMULXHU (acc, uw1, uw1)}
6651 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6652 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6653 @item @code{uw1 __MNOT (uw1)}
6654 @tab @code{@var{b} = __MNOT (@var{a})}
6655 @tab @code{MNOT @var{a},@var{b}}
6656 @item @code{uw1 __MOR (uw1, uw1)}
6657 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6658 @tab @code{MOR @var{a},@var{b},@var{c}}
6659 @item @code{uw1 __MPACKH (uh, uh)}
6660 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6661 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6662 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6663 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6664 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6665 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6666 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6667 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6668 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6669 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6670 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6671 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6672 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6673 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6674 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6675 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6676 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6677 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6678 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6679 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6680 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6681 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6682 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6683 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6684 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6685 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6686 @item @code{void __MQMACHS (acc, sw2, sw2)}
6687 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6688 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6689 @item @code{void __MQMACHU (acc, uw2, uw2)}
6690 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6691 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6692 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6693 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6694 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6695 @item @code{void __MQMULHS (acc, sw2, sw2)}
6696 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6697 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6698 @item @code{void __MQMULHU (acc, uw2, uw2)}
6699 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6700 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6701 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6702 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6703 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6704 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6705 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6706 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6707 @item @code{sw2 __MQSATHS (sw2, sw2)}
6708 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6709 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6710 @item @code{uw2 __MQSLLHI (uw2, int)}
6711 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6712 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6713 @item @code{sw2 __MQSRAHI (sw2, int)}
6714 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6715 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6716 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6717 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6718 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6719 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6720 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6721 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6722 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6723 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6724 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6725 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6726 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6727 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6728 @item @code{uw1 __MRDACC (acc)}
6729 @tab @code{@var{b} = __MRDACC (@var{a})}
6730 @tab @code{MRDACC @var{a},@var{b}}
6731 @item @code{uw1 __MRDACCG (acc)}
6732 @tab @code{@var{b} = __MRDACCG (@var{a})}
6733 @tab @code{MRDACCG @var{a},@var{b}}
6734 @item @code{uw1 __MROTLI (uw1, const)}
6735 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6736 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6737 @item @code{uw1 __MROTRI (uw1, const)}
6738 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6739 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6740 @item @code{sw1 __MSATHS (sw1, sw1)}
6741 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6742 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6743 @item @code{uw1 __MSATHU (uw1, uw1)}
6744 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6745 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6746 @item @code{uw1 __MSLLHI (uw1, const)}
6747 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6748 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6749 @item @code{sw1 __MSRAHI (sw1, const)}
6750 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6751 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6752 @item @code{uw1 __MSRLHI (uw1, const)}
6753 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6754 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6755 @item @code{void __MSUBACCS (acc, acc)}
6756 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6757 @tab @code{MSUBACCS @var{a},@var{b}}
6758 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6759 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6760 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6761 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6762 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6763 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6764 @item @code{void __MTRAP (void)}
6765 @tab @code{__MTRAP ()}
6767 @item @code{uw2 __MUNPACKH (uw1)}
6768 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6769 @tab @code{MUNPACKH @var{a},@var{b}}
6770 @item @code{uw1 __MWCUT (uw2, uw1)}
6771 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6772 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6773 @item @code{void __MWTACC (acc, uw1)}
6774 @tab @code{__MWTACC (@var{b}, @var{a})}
6775 @tab @code{MWTACC @var{a},@var{b}}
6776 @item @code{void __MWTACCG (acc, uw1)}
6777 @tab @code{__MWTACCG (@var{b}, @var{a})}
6778 @tab @code{MWTACCG @var{a},@var{b}}
6779 @item @code{uw1 __MXOR (uw1, uw1)}
6780 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6781 @tab @code{MXOR @var{a},@var{b},@var{c}}
6784 @node Raw read/write Functions
6785 @subsubsection Raw read/write Functions
6787 This sections describes built-in functions related to read and write
6788 instructions to access memory. These functions generate
6789 @code{membar} instructions to flush the I/O load and stores where
6790 appropriate, as described in Fujitsu's manual described above.
6794 @item unsigned char __builtin_read8 (void *@var{data})
6795 @item unsigned short __builtin_read16 (void *@var{data})
6796 @item unsigned long __builtin_read32 (void *@var{data})
6797 @item unsigned long long __builtin_read64 (void *@var{data})
6799 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6800 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6801 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6802 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6805 @node Other Built-in Functions
6806 @subsubsection Other Built-in Functions
6808 This section describes built-in functions that are not named after
6809 a specific FR-V instruction.
6812 @item sw2 __IACCreadll (iacc @var{reg})
6813 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6814 for future expansion and must be 0.
6816 @item sw1 __IACCreadl (iacc @var{reg})
6817 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6818 Other values of @var{reg} are rejected as invalid.
6820 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6821 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6822 is reserved for future expansion and must be 0.
6824 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6825 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6826 is 1. Other values of @var{reg} are rejected as invalid.
6828 @item void __data_prefetch0 (const void *@var{x})
6829 Use the @code{dcpl} instruction to load the contents of address @var{x}
6830 into the data cache.
6832 @item void __data_prefetch (const void *@var{x})
6833 Use the @code{nldub} instruction to load the contents of address @var{x}
6834 into the data cache. The instruction will be issued in slot I1@.
6837 @node X86 Built-in Functions
6838 @subsection X86 Built-in Functions
6840 These built-in functions are available for the i386 and x86-64 family
6841 of computers, depending on the command-line switches used.
6843 Note that, if you specify command-line switches such as @option{-msse},
6844 the compiler could use the extended instruction sets even if the built-ins
6845 are not used explicitly in the program. For this reason, applications
6846 which perform runtime CPU detection must compile separate files for each
6847 supported architecture, using the appropriate flags. In particular,
6848 the file containing the CPU detection code should be compiled without
6851 The following machine modes are available for use with MMX built-in functions
6852 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6853 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6854 vector of eight 8-bit integers. Some of the built-in functions operate on
6855 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6857 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6858 of two 32-bit floating point values.
6860 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6861 floating point values. Some instructions use a vector of four 32-bit
6862 integers, these use @code{V4SI}. Finally, some instructions operate on an
6863 entire vector register, interpreting it as a 128-bit integer, these use mode
6866 The following built-in functions are made available by @option{-mmmx}.
6867 All of them generate the machine instruction that is part of the name.
6870 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6871 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6872 v2si __builtin_ia32_paddd (v2si, v2si)
6873 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6874 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6875 v2si __builtin_ia32_psubd (v2si, v2si)
6876 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6877 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6878 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6879 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6880 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6881 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6882 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6883 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6884 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6885 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6886 di __builtin_ia32_pand (di, di)
6887 di __builtin_ia32_pandn (di,di)
6888 di __builtin_ia32_por (di, di)
6889 di __builtin_ia32_pxor (di, di)
6890 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6891 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6892 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6893 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6894 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6895 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6896 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6897 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6898 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6899 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6900 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6901 v2si __builtin_ia32_punpckldq (v2si, v2si)
6902 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6903 v4hi __builtin_ia32_packssdw (v2si, v2si)
6904 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6907 The following built-in functions are made available either with
6908 @option{-msse}, or with a combination of @option{-m3dnow} and
6909 @option{-march=athlon}. All of them generate the machine
6910 instruction that is part of the name.
6913 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6914 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6915 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6916 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6917 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6918 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6919 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6920 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6921 int __builtin_ia32_pextrw (v4hi, int)
6922 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6923 int __builtin_ia32_pmovmskb (v8qi)
6924 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6925 void __builtin_ia32_movntq (di *, di)
6926 void __builtin_ia32_sfence (void)
6929 The following built-in functions are available when @option{-msse} is used.
6930 All of them generate the machine instruction that is part of the name.
6933 int __builtin_ia32_comieq (v4sf, v4sf)
6934 int __builtin_ia32_comineq (v4sf, v4sf)
6935 int __builtin_ia32_comilt (v4sf, v4sf)
6936 int __builtin_ia32_comile (v4sf, v4sf)
6937 int __builtin_ia32_comigt (v4sf, v4sf)
6938 int __builtin_ia32_comige (v4sf, v4sf)
6939 int __builtin_ia32_ucomieq (v4sf, v4sf)
6940 int __builtin_ia32_ucomineq (v4sf, v4sf)
6941 int __builtin_ia32_ucomilt (v4sf, v4sf)
6942 int __builtin_ia32_ucomile (v4sf, v4sf)
6943 int __builtin_ia32_ucomigt (v4sf, v4sf)
6944 int __builtin_ia32_ucomige (v4sf, v4sf)
6945 v4sf __builtin_ia32_addps (v4sf, v4sf)
6946 v4sf __builtin_ia32_subps (v4sf, v4sf)
6947 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6948 v4sf __builtin_ia32_divps (v4sf, v4sf)
6949 v4sf __builtin_ia32_addss (v4sf, v4sf)
6950 v4sf __builtin_ia32_subss (v4sf, v4sf)
6951 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6952 v4sf __builtin_ia32_divss (v4sf, v4sf)
6953 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6954 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6955 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6956 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6957 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6958 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6959 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6960 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6961 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6962 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6963 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6964 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6965 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6966 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6967 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6968 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6969 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6970 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6971 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6972 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6973 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6974 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6975 v4sf __builtin_ia32_minps (v4sf, v4sf)
6976 v4sf __builtin_ia32_minss (v4sf, v4sf)
6977 v4sf __builtin_ia32_andps (v4sf, v4sf)
6978 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6979 v4sf __builtin_ia32_orps (v4sf, v4sf)
6980 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6981 v4sf __builtin_ia32_movss (v4sf, v4sf)
6982 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6983 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6984 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6985 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6986 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6987 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6988 v2si __builtin_ia32_cvtps2pi (v4sf)
6989 int __builtin_ia32_cvtss2si (v4sf)
6990 v2si __builtin_ia32_cvttps2pi (v4sf)
6991 int __builtin_ia32_cvttss2si (v4sf)
6992 v4sf __builtin_ia32_rcpps (v4sf)
6993 v4sf __builtin_ia32_rsqrtps (v4sf)
6994 v4sf __builtin_ia32_sqrtps (v4sf)
6995 v4sf __builtin_ia32_rcpss (v4sf)
6996 v4sf __builtin_ia32_rsqrtss (v4sf)
6997 v4sf __builtin_ia32_sqrtss (v4sf)
6998 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6999 void __builtin_ia32_movntps (float *, v4sf)
7000 int __builtin_ia32_movmskps (v4sf)
7003 The following built-in functions are available when @option{-msse} is used.
7006 @item v4sf __builtin_ia32_loadaps (float *)
7007 Generates the @code{movaps} machine instruction as a load from memory.
7008 @item void __builtin_ia32_storeaps (float *, v4sf)
7009 Generates the @code{movaps} machine instruction as a store to memory.
7010 @item v4sf __builtin_ia32_loadups (float *)
7011 Generates the @code{movups} machine instruction as a load from memory.
7012 @item void __builtin_ia32_storeups (float *, v4sf)
7013 Generates the @code{movups} machine instruction as a store to memory.
7014 @item v4sf __builtin_ia32_loadsss (float *)
7015 Generates the @code{movss} machine instruction as a load from memory.
7016 @item void __builtin_ia32_storess (float *, v4sf)
7017 Generates the @code{movss} machine instruction as a store to memory.
7018 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7019 Generates the @code{movhps} machine instruction as a load from memory.
7020 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7021 Generates the @code{movlps} machine instruction as a load from memory
7022 @item void __builtin_ia32_storehps (v4sf, v2si *)
7023 Generates the @code{movhps} machine instruction as a store to memory.
7024 @item void __builtin_ia32_storelps (v4sf, v2si *)
7025 Generates the @code{movlps} machine instruction as a store to memory.
7028 The following built-in functions are available when @option{-msse2} is used.
7029 All of them generate the machine instruction that is part of the name.
7032 int __builtin_ia32_comisdeq (v2df, v2df)
7033 int __builtin_ia32_comisdlt (v2df, v2df)
7034 int __builtin_ia32_comisdle (v2df, v2df)
7035 int __builtin_ia32_comisdgt (v2df, v2df)
7036 int __builtin_ia32_comisdge (v2df, v2df)
7037 int __builtin_ia32_comisdneq (v2df, v2df)
7038 int __builtin_ia32_ucomisdeq (v2df, v2df)
7039 int __builtin_ia32_ucomisdlt (v2df, v2df)
7040 int __builtin_ia32_ucomisdle (v2df, v2df)
7041 int __builtin_ia32_ucomisdgt (v2df, v2df)
7042 int __builtin_ia32_ucomisdge (v2df, v2df)
7043 int __builtin_ia32_ucomisdneq (v2df, v2df)
7044 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7045 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7046 v2df __builtin_ia32_cmplepd (v2df, v2df)
7047 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7048 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7049 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7050 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7051 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7052 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7053 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7054 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7055 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7056 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7057 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7058 v2df __builtin_ia32_cmplesd (v2df, v2df)
7059 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7060 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7061 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7062 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7063 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7064 v2di __builtin_ia32_paddq (v2di, v2di)
7065 v2di __builtin_ia32_psubq (v2di, v2di)
7066 v2df __builtin_ia32_addpd (v2df, v2df)
7067 v2df __builtin_ia32_subpd (v2df, v2df)
7068 v2df __builtin_ia32_mulpd (v2df, v2df)
7069 v2df __builtin_ia32_divpd (v2df, v2df)
7070 v2df __builtin_ia32_addsd (v2df, v2df)
7071 v2df __builtin_ia32_subsd (v2df, v2df)
7072 v2df __builtin_ia32_mulsd (v2df, v2df)
7073 v2df __builtin_ia32_divsd (v2df, v2df)
7074 v2df __builtin_ia32_minpd (v2df, v2df)
7075 v2df __builtin_ia32_maxpd (v2df, v2df)
7076 v2df __builtin_ia32_minsd (v2df, v2df)
7077 v2df __builtin_ia32_maxsd (v2df, v2df)
7078 v2df __builtin_ia32_andpd (v2df, v2df)
7079 v2df __builtin_ia32_andnpd (v2df, v2df)
7080 v2df __builtin_ia32_orpd (v2df, v2df)
7081 v2df __builtin_ia32_xorpd (v2df, v2df)
7082 v2df __builtin_ia32_movsd (v2df, v2df)
7083 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7084 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7085 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7086 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7087 v4si __builtin_ia32_paddd128 (v4si, v4si)
7088 v2di __builtin_ia32_paddq128 (v2di, v2di)
7089 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7090 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7091 v4si __builtin_ia32_psubd128 (v4si, v4si)
7092 v2di __builtin_ia32_psubq128 (v2di, v2di)
7093 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7094 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7095 v2di __builtin_ia32_pand128 (v2di, v2di)
7096 v2di __builtin_ia32_pandn128 (v2di, v2di)
7097 v2di __builtin_ia32_por128 (v2di, v2di)
7098 v2di __builtin_ia32_pxor128 (v2di, v2di)
7099 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7100 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7101 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7102 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7103 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7104 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7105 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7106 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7107 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7108 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7109 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7110 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7111 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7112 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7113 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7114 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7115 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7116 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7117 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7118 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7119 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7120 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7121 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7122 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7123 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7124 v2df __builtin_ia32_loadupd (double *)
7125 void __builtin_ia32_storeupd (double *, v2df)
7126 v2df __builtin_ia32_loadhpd (v2df, double *)
7127 v2df __builtin_ia32_loadlpd (v2df, double *)
7128 int __builtin_ia32_movmskpd (v2df)
7129 int __builtin_ia32_pmovmskb128 (v16qi)
7130 void __builtin_ia32_movnti (int *, int)
7131 void __builtin_ia32_movntpd (double *, v2df)
7132 void __builtin_ia32_movntdq (v2df *, v2df)
7133 v4si __builtin_ia32_pshufd (v4si, int)
7134 v8hi __builtin_ia32_pshuflw (v8hi, int)
7135 v8hi __builtin_ia32_pshufhw (v8hi, int)
7136 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7137 v2df __builtin_ia32_sqrtpd (v2df)
7138 v2df __builtin_ia32_sqrtsd (v2df)
7139 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7140 v2df __builtin_ia32_cvtdq2pd (v4si)
7141 v4sf __builtin_ia32_cvtdq2ps (v4si)
7142 v4si __builtin_ia32_cvtpd2dq (v2df)
7143 v2si __builtin_ia32_cvtpd2pi (v2df)
7144 v4sf __builtin_ia32_cvtpd2ps (v2df)
7145 v4si __builtin_ia32_cvttpd2dq (v2df)
7146 v2si __builtin_ia32_cvttpd2pi (v2df)
7147 v2df __builtin_ia32_cvtpi2pd (v2si)
7148 int __builtin_ia32_cvtsd2si (v2df)
7149 int __builtin_ia32_cvttsd2si (v2df)
7150 long long __builtin_ia32_cvtsd2si64 (v2df)
7151 long long __builtin_ia32_cvttsd2si64 (v2df)
7152 v4si __builtin_ia32_cvtps2dq (v4sf)
7153 v2df __builtin_ia32_cvtps2pd (v4sf)
7154 v4si __builtin_ia32_cvttps2dq (v4sf)
7155 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7156 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7157 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7158 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7159 void __builtin_ia32_clflush (const void *)
7160 void __builtin_ia32_lfence (void)
7161 void __builtin_ia32_mfence (void)
7162 v16qi __builtin_ia32_loaddqu (const char *)
7163 void __builtin_ia32_storedqu (char *, v16qi)
7164 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7165 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7166 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7167 v4si __builtin_ia32_pslld128 (v4si, v2di)
7168 v2di __builtin_ia32_psllq128 (v4si, v2di)
7169 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7170 v4si __builtin_ia32_psrld128 (v4si, v2di)
7171 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7172 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7173 v4si __builtin_ia32_psrad128 (v4si, v2di)
7174 v2di __builtin_ia32_pslldqi128 (v2di, int)
7175 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7176 v4si __builtin_ia32_pslldi128 (v4si, int)
7177 v2di __builtin_ia32_psllqi128 (v2di, int)
7178 v2di __builtin_ia32_psrldqi128 (v2di, int)
7179 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7180 v4si __builtin_ia32_psrldi128 (v4si, int)
7181 v2di __builtin_ia32_psrlqi128 (v2di, int)
7182 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7183 v4si __builtin_ia32_psradi128 (v4si, int)
7184 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7187 The following built-in functions are available when @option{-msse3} is used.
7188 All of them generate the machine instruction that is part of the name.
7191 v2df __builtin_ia32_addsubpd (v2df, v2df)
7192 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7193 v2df __builtin_ia32_haddpd (v2df, v2df)
7194 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7195 v2df __builtin_ia32_hsubpd (v2df, v2df)
7196 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7197 v16qi __builtin_ia32_lddqu (char const *)
7198 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7199 v2df __builtin_ia32_movddup (v2df)
7200 v4sf __builtin_ia32_movshdup (v4sf)
7201 v4sf __builtin_ia32_movsldup (v4sf)
7202 void __builtin_ia32_mwait (unsigned int, unsigned int)
7205 The following built-in functions are available when @option{-msse3} is used.
7208 @item v2df __builtin_ia32_loadddup (double const *)
7209 Generates the @code{movddup} machine instruction as a load from memory.
7212 The following built-in functions are available when @option{-mssse3} is used.
7213 All of them generate the machine instruction that is part of the name
7217 v2si __builtin_ia32_phaddd (v2si, v2si)
7218 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7219 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7220 v2si __builtin_ia32_phsubd (v2si, v2si)
7221 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7222 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7223 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7224 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7225 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7226 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7227 v2si __builtin_ia32_psignd (v2si, v2si)
7228 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7229 long long __builtin_ia32_palignr (long long, long long, int)
7230 v8qi __builtin_ia32_pabsb (v8qi)
7231 v2si __builtin_ia32_pabsd (v2si)
7232 v4hi __builtin_ia32_pabsw (v4hi)
7235 The following built-in functions are available when @option{-mssse3} is used.
7236 All of them generate the machine instruction that is part of the name
7240 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7241 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7242 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7243 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7244 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7245 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7246 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7247 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7248 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7249 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7250 v4si __builtin_ia32_psignd128 (v4si, v4si)
7251 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7252 v2di __builtin_ia32_palignr (v2di, v2di, int)
7253 v16qi __builtin_ia32_pabsb128 (v16qi)
7254 v4si __builtin_ia32_pabsd128 (v4si)
7255 v8hi __builtin_ia32_pabsw128 (v8hi)
7258 The following built-in functions are available when @option{-msse4a} is used.
7261 void _mm_stream_sd (double*,__m128d);
7262 Generates the @code{movntsd} machine instruction.
7263 void _mm_stream_ss (float*,__m128);
7264 Generates the @code{movntss} machine instruction.
7265 __m128i _mm_extract_si64 (__m128i, __m128i);
7266 Generates the @code{extrq} machine instruction with only SSE register operands.
7267 __m128i _mm_extracti_si64 (__m128i, int, int);
7268 Generates the @code{extrq} machine instruction with SSE register and immediate operands.
7269 __m128i _mm_insert_si64 (__m128i, __m128i);
7270 Generates the @code{insertq} machine instruction with only SSE register operands.
7271 __m128i _mm_inserti_si64 (__m128i, __m128i, int, int);
7272 Generates the @code{insertq} machine instruction with SSE register and immediate operands.
7275 The following built-in functions are available when @option{-m3dnow} is used.
7276 All of them generate the machine instruction that is part of the name.
7279 void __builtin_ia32_femms (void)
7280 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7281 v2si __builtin_ia32_pf2id (v2sf)
7282 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7283 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7284 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7285 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7286 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7287 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7288 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7289 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7290 v2sf __builtin_ia32_pfrcp (v2sf)
7291 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7292 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7293 v2sf __builtin_ia32_pfrsqrt (v2sf)
7294 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7295 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7296 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7297 v2sf __builtin_ia32_pi2fd (v2si)
7298 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7301 The following built-in functions are available when both @option{-m3dnow}
7302 and @option{-march=athlon} are used. All of them generate the machine
7303 instruction that is part of the name.
7306 v2si __builtin_ia32_pf2iw (v2sf)
7307 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7308 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7309 v2sf __builtin_ia32_pi2fw (v2si)
7310 v2sf __builtin_ia32_pswapdsf (v2sf)
7311 v2si __builtin_ia32_pswapdsi (v2si)
7314 @node MIPS DSP Built-in Functions
7315 @subsection MIPS DSP Built-in Functions
7317 The MIPS DSP Application-Specific Extension (ASE) includes new
7318 instructions that are designed to improve the performance of DSP and
7319 media applications. It provides instructions that operate on packed
7320 8-bit integer data, Q15 fractional data and Q31 fractional data.
7322 GCC supports MIPS DSP operations using both the generic
7323 vector extensions (@pxref{Vector Extensions}) and a collection of
7324 MIPS-specific built-in functions. Both kinds of support are
7325 enabled by the @option{-mdsp} command-line option.
7327 At present, GCC only provides support for operations on 32-bit
7328 vectors. The vector type associated with 8-bit integer data is
7329 usually called @code{v4i8} and the vector type associated with Q15 is
7330 usually called @code{v2q15}. They can be defined in C as follows:
7333 typedef char v4i8 __attribute__ ((vector_size(4)));
7334 typedef short v2q15 __attribute__ ((vector_size(4)));
7337 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7338 aggregates. For example:
7341 v4i8 a = @{1, 2, 3, 4@};
7343 b = (v4i8) @{5, 6, 7, 8@};
7345 v2q15 c = @{0x0fcb, 0x3a75@};
7347 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7350 @emph{Note:} The CPU's endianness determines the order in which values
7351 are packed. On little-endian targets, the first value is the least
7352 significant and the last value is the most significant. The opposite
7353 order applies to big-endian targets. For example, the code above will
7354 set the lowest byte of @code{a} to @code{1} on little-endian targets
7355 and @code{4} on big-endian targets.
7357 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7358 representation. As shown in this example, the integer representation
7359 of a Q15 value can be obtained by multiplying the fractional value by
7360 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7363 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7364 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7365 and @code{c} and @code{d} are @code{v2q15} values.
7367 @multitable @columnfractions .50 .50
7368 @item C code @tab MIPS instruction
7369 @item @code{a + b} @tab @code{addu.qb}
7370 @item @code{c + d} @tab @code{addq.ph}
7371 @item @code{a - b} @tab @code{subu.qb}
7372 @item @code{c - d} @tab @code{subq.ph}
7375 It is easier to describe the DSP built-in functions if we first define
7376 the following types:
7381 typedef long long a64;
7384 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7385 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7386 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7387 @code{long long}, but we use @code{a64} to indicate values that will
7388 be placed in one of the four DSP accumulators (@code{$ac0},
7389 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7391 Also, some built-in functions prefer or require immediate numbers as
7392 parameters, because the corresponding DSP instructions accept both immediate
7393 numbers and register operands, or accept immediate numbers only. The
7394 immediate parameters are listed as follows.
7402 imm_n32_31: -32 to 31.
7403 imm_n512_511: -512 to 511.
7406 The following built-in functions map directly to a particular MIPS DSP
7407 instruction. Please refer to the architecture specification
7408 for details on what each instruction does.
7411 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7412 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7413 q31 __builtin_mips_addq_s_w (q31, q31)
7414 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7415 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7416 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7417 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7418 q31 __builtin_mips_subq_s_w (q31, q31)
7419 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7420 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7421 i32 __builtin_mips_addsc (i32, i32)
7422 i32 __builtin_mips_addwc (i32, i32)
7423 i32 __builtin_mips_modsub (i32, i32)
7424 i32 __builtin_mips_raddu_w_qb (v4i8)
7425 v2q15 __builtin_mips_absq_s_ph (v2q15)
7426 q31 __builtin_mips_absq_s_w (q31)
7427 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7428 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7429 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7430 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7431 q31 __builtin_mips_preceq_w_phl (v2q15)
7432 q31 __builtin_mips_preceq_w_phr (v2q15)
7433 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7434 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7435 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7436 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7437 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7438 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7439 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7440 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7441 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7442 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7443 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7444 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7445 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7446 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7447 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7448 q31 __builtin_mips_shll_s_w (q31, i32)
7449 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7450 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7451 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7452 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7453 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7454 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7455 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7456 q31 __builtin_mips_shra_r_w (q31, i32)
7457 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7458 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7459 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7460 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7461 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7462 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7463 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7464 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7465 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7466 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7467 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7468 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7469 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7470 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7471 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7472 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7473 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7474 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7475 i32 __builtin_mips_bitrev (i32)
7476 i32 __builtin_mips_insv (i32, i32)
7477 v4i8 __builtin_mips_repl_qb (imm0_255)
7478 v4i8 __builtin_mips_repl_qb (i32)
7479 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7480 v2q15 __builtin_mips_repl_ph (i32)
7481 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7482 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7483 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7484 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7485 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7486 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7487 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7488 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7489 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7490 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7491 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7492 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7493 i32 __builtin_mips_extr_w (a64, imm0_31)
7494 i32 __builtin_mips_extr_w (a64, i32)
7495 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7496 i32 __builtin_mips_extr_s_h (a64, i32)
7497 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7498 i32 __builtin_mips_extr_rs_w (a64, i32)
7499 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7500 i32 __builtin_mips_extr_r_w (a64, i32)
7501 i32 __builtin_mips_extp (a64, imm0_31)
7502 i32 __builtin_mips_extp (a64, i32)
7503 i32 __builtin_mips_extpdp (a64, imm0_31)
7504 i32 __builtin_mips_extpdp (a64, i32)
7505 a64 __builtin_mips_shilo (a64, imm_n32_31)
7506 a64 __builtin_mips_shilo (a64, i32)
7507 a64 __builtin_mips_mthlip (a64, i32)
7508 void __builtin_mips_wrdsp (i32, imm0_63)
7509 i32 __builtin_mips_rddsp (imm0_63)
7510 i32 __builtin_mips_lbux (void *, i32)
7511 i32 __builtin_mips_lhx (void *, i32)
7512 i32 __builtin_mips_lwx (void *, i32)
7513 i32 __builtin_mips_bposge32 (void)
7516 @node MIPS Paired-Single Support
7517 @subsection MIPS Paired-Single Support
7519 The MIPS64 architecture includes a number of instructions that
7520 operate on pairs of single-precision floating-point values.
7521 Each pair is packed into a 64-bit floating-point register,
7522 with one element being designated the ``upper half'' and
7523 the other being designated the ``lower half''.
7525 GCC supports paired-single operations using both the generic
7526 vector extensions (@pxref{Vector Extensions}) and a collection of
7527 MIPS-specific built-in functions. Both kinds of support are
7528 enabled by the @option{-mpaired-single} command-line option.
7530 The vector type associated with paired-single values is usually
7531 called @code{v2sf}. It can be defined in C as follows:
7534 typedef float v2sf __attribute__ ((vector_size (8)));
7537 @code{v2sf} values are initialized in the same way as aggregates.
7541 v2sf a = @{1.5, 9.1@};
7544 b = (v2sf) @{e, f@};
7547 @emph{Note:} The CPU's endianness determines which value is stored in
7548 the upper half of a register and which value is stored in the lower half.
7549 On little-endian targets, the first value is the lower one and the second
7550 value is the upper one. The opposite order applies to big-endian targets.
7551 For example, the code above will set the lower half of @code{a} to
7552 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7555 * Paired-Single Arithmetic::
7556 * Paired-Single Built-in Functions::
7557 * MIPS-3D Built-in Functions::
7560 @node Paired-Single Arithmetic
7561 @subsubsection Paired-Single Arithmetic
7563 The table below lists the @code{v2sf} operations for which hardware
7564 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7565 values and @code{x} is an integral value.
7567 @multitable @columnfractions .50 .50
7568 @item C code @tab MIPS instruction
7569 @item @code{a + b} @tab @code{add.ps}
7570 @item @code{a - b} @tab @code{sub.ps}
7571 @item @code{-a} @tab @code{neg.ps}
7572 @item @code{a * b} @tab @code{mul.ps}
7573 @item @code{a * b + c} @tab @code{madd.ps}
7574 @item @code{a * b - c} @tab @code{msub.ps}
7575 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7576 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7577 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7580 Note that the multiply-accumulate instructions can be disabled
7581 using the command-line option @code{-mno-fused-madd}.
7583 @node Paired-Single Built-in Functions
7584 @subsubsection Paired-Single Built-in Functions
7586 The following paired-single functions map directly to a particular
7587 MIPS instruction. Please refer to the architecture specification
7588 for details on what each instruction does.
7591 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7592 Pair lower lower (@code{pll.ps}).
7594 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7595 Pair upper lower (@code{pul.ps}).
7597 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7598 Pair lower upper (@code{plu.ps}).
7600 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7601 Pair upper upper (@code{puu.ps}).
7603 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7604 Convert pair to paired single (@code{cvt.ps.s}).
7606 @item float __builtin_mips_cvt_s_pl (v2sf)
7607 Convert pair lower to single (@code{cvt.s.pl}).
7609 @item float __builtin_mips_cvt_s_pu (v2sf)
7610 Convert pair upper to single (@code{cvt.s.pu}).
7612 @item v2sf __builtin_mips_abs_ps (v2sf)
7613 Absolute value (@code{abs.ps}).
7615 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7616 Align variable (@code{alnv.ps}).
7618 @emph{Note:} The value of the third parameter must be 0 or 4
7619 modulo 8, otherwise the result will be unpredictable. Please read the
7620 instruction description for details.
7623 The following multi-instruction functions are also available.
7624 In each case, @var{cond} can be any of the 16 floating-point conditions:
7625 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7626 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7627 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7630 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7631 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7632 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7633 @code{movt.ps}/@code{movf.ps}).
7635 The @code{movt} functions return the value @var{x} computed by:
7638 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7639 mov.ps @var{x},@var{c}
7640 movt.ps @var{x},@var{d},@var{cc}
7643 The @code{movf} functions are similar but use @code{movf.ps} instead
7646 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7647 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7648 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7649 @code{bc1t}/@code{bc1f}).
7651 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7652 and return either the upper or lower half of the result. For example:
7656 if (__builtin_mips_upper_c_eq_ps (a, b))
7657 upper_halves_are_equal ();
7659 upper_halves_are_unequal ();
7661 if (__builtin_mips_lower_c_eq_ps (a, b))
7662 lower_halves_are_equal ();
7664 lower_halves_are_unequal ();
7668 @node MIPS-3D Built-in Functions
7669 @subsubsection MIPS-3D Built-in Functions
7671 The MIPS-3D Application-Specific Extension (ASE) includes additional
7672 paired-single instructions that are designed to improve the performance
7673 of 3D graphics operations. Support for these instructions is controlled
7674 by the @option{-mips3d} command-line option.
7676 The functions listed below map directly to a particular MIPS-3D
7677 instruction. Please refer to the architecture specification for
7678 more details on what each instruction does.
7681 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7682 Reduction add (@code{addr.ps}).
7684 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7685 Reduction multiply (@code{mulr.ps}).
7687 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7688 Convert paired single to paired word (@code{cvt.pw.ps}).
7690 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7691 Convert paired word to paired single (@code{cvt.ps.pw}).
7693 @item float __builtin_mips_recip1_s (float)
7694 @itemx double __builtin_mips_recip1_d (double)
7695 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7696 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7698 @item float __builtin_mips_recip2_s (float, float)
7699 @itemx double __builtin_mips_recip2_d (double, double)
7700 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7701 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7703 @item float __builtin_mips_rsqrt1_s (float)
7704 @itemx double __builtin_mips_rsqrt1_d (double)
7705 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7706 Reduced precision reciprocal square root (sequence step 1)
7707 (@code{rsqrt1.@var{fmt}}).
7709 @item float __builtin_mips_rsqrt2_s (float, float)
7710 @itemx double __builtin_mips_rsqrt2_d (double, double)
7711 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7712 Reduced precision reciprocal square root (sequence step 2)
7713 (@code{rsqrt2.@var{fmt}}).
7716 The following multi-instruction functions are also available.
7717 In each case, @var{cond} can be any of the 16 floating-point conditions:
7718 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7719 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7720 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7723 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7724 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7725 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7726 @code{bc1t}/@code{bc1f}).
7728 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7729 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7734 if (__builtin_mips_cabs_eq_s (a, b))
7740 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7741 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7742 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7743 @code{bc1t}/@code{bc1f}).
7745 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7746 and return either the upper or lower half of the result. For example:
7750 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7751 upper_halves_are_equal ();
7753 upper_halves_are_unequal ();
7755 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7756 lower_halves_are_equal ();
7758 lower_halves_are_unequal ();
7761 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7762 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7763 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7764 @code{movt.ps}/@code{movf.ps}).
7766 The @code{movt} functions return the value @var{x} computed by:
7769 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7770 mov.ps @var{x},@var{c}
7771 movt.ps @var{x},@var{d},@var{cc}
7774 The @code{movf} functions are similar but use @code{movf.ps} instead
7777 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7778 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7779 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7780 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7781 Comparison of two paired-single values
7782 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7783 @code{bc1any2t}/@code{bc1any2f}).
7785 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7786 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7787 result is true and the @code{all} forms return true if both results are true.
7792 if (__builtin_mips_any_c_eq_ps (a, b))
7797 if (__builtin_mips_all_c_eq_ps (a, b))
7803 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7804 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7805 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7806 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7807 Comparison of four paired-single values
7808 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7809 @code{bc1any4t}/@code{bc1any4f}).
7811 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7812 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7813 The @code{any} forms return true if any of the four results are true
7814 and the @code{all} forms return true if all four results are true.
7819 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7824 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7831 @node PowerPC AltiVec Built-in Functions
7832 @subsection PowerPC AltiVec Built-in Functions
7834 GCC provides an interface for the PowerPC family of processors to access
7835 the AltiVec operations described in Motorola's AltiVec Programming
7836 Interface Manual. The interface is made available by including
7837 @code{<altivec.h>} and using @option{-maltivec} and
7838 @option{-mabi=altivec}. The interface supports the following vector
7842 vector unsigned char
7846 vector unsigned short
7857 GCC's implementation of the high-level language interface available from
7858 C and C++ code differs from Motorola's documentation in several ways.
7863 A vector constant is a list of constant expressions within curly braces.
7866 A vector initializer requires no cast if the vector constant is of the
7867 same type as the variable it is initializing.
7870 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7871 vector type is the default signedness of the base type. The default
7872 varies depending on the operating system, so a portable program should
7873 always specify the signedness.
7876 Compiling with @option{-maltivec} adds keywords @code{__vector},
7877 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7878 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7882 GCC allows using a @code{typedef} name as the type specifier for a
7886 For C, overloaded functions are implemented with macros so the following
7890 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7893 Since @code{vec_add} is a macro, the vector constant in the example
7894 is treated as four separate arguments. Wrap the entire argument in
7895 parentheses for this to work.
7898 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
7899 Internally, GCC uses built-in functions to achieve the functionality in
7900 the aforementioned header file, but they are not supported and are
7901 subject to change without notice.
7903 The following interfaces are supported for the generic and specific
7904 AltiVec operations and the AltiVec predicates. In cases where there
7905 is a direct mapping between generic and specific operations, only the
7906 generic names are shown here, although the specific operations can also
7909 Arguments that are documented as @code{const int} require literal
7910 integral values within the range required for that operation.
7913 vector signed char vec_abs (vector signed char);
7914 vector signed short vec_abs (vector signed short);
7915 vector signed int vec_abs (vector signed int);
7916 vector float vec_abs (vector float);
7918 vector signed char vec_abss (vector signed char);
7919 vector signed short vec_abss (vector signed short);
7920 vector signed int vec_abss (vector signed int);
7922 vector signed char vec_add (vector bool char, vector signed char);
7923 vector signed char vec_add (vector signed char, vector bool char);
7924 vector signed char vec_add (vector signed char, vector signed char);
7925 vector unsigned char vec_add (vector bool char, vector unsigned char);
7926 vector unsigned char vec_add (vector unsigned char, vector bool char);
7927 vector unsigned char vec_add (vector unsigned char,
7928 vector unsigned char);
7929 vector signed short vec_add (vector bool short, vector signed short);
7930 vector signed short vec_add (vector signed short, vector bool short);
7931 vector signed short vec_add (vector signed short, vector signed short);
7932 vector unsigned short vec_add (vector bool short,
7933 vector unsigned short);
7934 vector unsigned short vec_add (vector unsigned short,
7936 vector unsigned short vec_add (vector unsigned short,
7937 vector unsigned short);
7938 vector signed int vec_add (vector bool int, vector signed int);
7939 vector signed int vec_add (vector signed int, vector bool int);
7940 vector signed int vec_add (vector signed int, vector signed int);
7941 vector unsigned int vec_add (vector bool int, vector unsigned int);
7942 vector unsigned int vec_add (vector unsigned int, vector bool int);
7943 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
7944 vector float vec_add (vector float, vector float);
7946 vector float vec_vaddfp (vector float, vector float);
7948 vector signed int vec_vadduwm (vector bool int, vector signed int);
7949 vector signed int vec_vadduwm (vector signed int, vector bool int);
7950 vector signed int vec_vadduwm (vector signed int, vector signed int);
7951 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
7952 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
7953 vector unsigned int vec_vadduwm (vector unsigned int,
7954 vector unsigned int);
7956 vector signed short vec_vadduhm (vector bool short,
7957 vector signed short);
7958 vector signed short vec_vadduhm (vector signed short,
7960 vector signed short vec_vadduhm (vector signed short,
7961 vector signed short);
7962 vector unsigned short vec_vadduhm (vector bool short,
7963 vector unsigned short);
7964 vector unsigned short vec_vadduhm (vector unsigned short,
7966 vector unsigned short vec_vadduhm (vector unsigned short,
7967 vector unsigned short);
7969 vector signed char vec_vaddubm (vector bool char, vector signed char);
7970 vector signed char vec_vaddubm (vector signed char, vector bool char);
7971 vector signed char vec_vaddubm (vector signed char, vector signed char);
7972 vector unsigned char vec_vaddubm (vector bool char,
7973 vector unsigned char);
7974 vector unsigned char vec_vaddubm (vector unsigned char,
7976 vector unsigned char vec_vaddubm (vector unsigned char,
7977 vector unsigned char);
7979 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
7981 vector unsigned char vec_adds (vector bool char, vector unsigned char);
7982 vector unsigned char vec_adds (vector unsigned char, vector bool char);
7983 vector unsigned char vec_adds (vector unsigned char,
7984 vector unsigned char);
7985 vector signed char vec_adds (vector bool char, vector signed char);
7986 vector signed char vec_adds (vector signed char, vector bool char);
7987 vector signed char vec_adds (vector signed char, vector signed char);
7988 vector unsigned short vec_adds (vector bool short,
7989 vector unsigned short);
7990 vector unsigned short vec_adds (vector unsigned short,
7992 vector unsigned short vec_adds (vector unsigned short,
7993 vector unsigned short);
7994 vector signed short vec_adds (vector bool short, vector signed short);
7995 vector signed short vec_adds (vector signed short, vector bool short);
7996 vector signed short vec_adds (vector signed short, vector signed short);
7997 vector unsigned int vec_adds (vector bool int, vector unsigned int);
7998 vector unsigned int vec_adds (vector unsigned int, vector bool int);
7999 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8000 vector signed int vec_adds (vector bool int, vector signed int);
8001 vector signed int vec_adds (vector signed int, vector bool int);
8002 vector signed int vec_adds (vector signed int, vector signed int);
8004 vector signed int vec_vaddsws (vector bool int, vector signed int);
8005 vector signed int vec_vaddsws (vector signed int, vector bool int);
8006 vector signed int vec_vaddsws (vector signed int, vector signed int);
8008 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8009 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8010 vector unsigned int vec_vadduws (vector unsigned int,
8011 vector unsigned int);
8013 vector signed short vec_vaddshs (vector bool short,
8014 vector signed short);
8015 vector signed short vec_vaddshs (vector signed short,
8017 vector signed short vec_vaddshs (vector signed short,
8018 vector signed short);
8020 vector unsigned short vec_vadduhs (vector bool short,
8021 vector unsigned short);
8022 vector unsigned short vec_vadduhs (vector unsigned short,
8024 vector unsigned short vec_vadduhs (vector unsigned short,
8025 vector unsigned short);
8027 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8028 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8029 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8031 vector unsigned char vec_vaddubs (vector bool char,
8032 vector unsigned char);
8033 vector unsigned char vec_vaddubs (vector unsigned char,
8035 vector unsigned char vec_vaddubs (vector unsigned char,
8036 vector unsigned char);
8038 vector float vec_and (vector float, vector float);
8039 vector float vec_and (vector float, vector bool int);
8040 vector float vec_and (vector bool int, vector float);
8041 vector bool int vec_and (vector bool int, vector bool int);
8042 vector signed int vec_and (vector bool int, vector signed int);
8043 vector signed int vec_and (vector signed int, vector bool int);
8044 vector signed int vec_and (vector signed int, vector signed int);
8045 vector unsigned int vec_and (vector bool int, vector unsigned int);
8046 vector unsigned int vec_and (vector unsigned int, vector bool int);
8047 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8048 vector bool short vec_and (vector bool short, vector bool short);
8049 vector signed short vec_and (vector bool short, vector signed short);
8050 vector signed short vec_and (vector signed short, vector bool short);
8051 vector signed short vec_and (vector signed short, vector signed short);
8052 vector unsigned short vec_and (vector bool short,
8053 vector unsigned short);
8054 vector unsigned short vec_and (vector unsigned short,
8056 vector unsigned short vec_and (vector unsigned short,
8057 vector unsigned short);
8058 vector signed char vec_and (vector bool char, vector signed char);
8059 vector bool char vec_and (vector bool char, vector bool char);
8060 vector signed char vec_and (vector signed char, vector bool char);
8061 vector signed char vec_and (vector signed char, vector signed char);
8062 vector unsigned char vec_and (vector bool char, vector unsigned char);
8063 vector unsigned char vec_and (vector unsigned char, vector bool char);
8064 vector unsigned char vec_and (vector unsigned char,
8065 vector unsigned char);
8067 vector float vec_andc (vector float, vector float);
8068 vector float vec_andc (vector float, vector bool int);
8069 vector float vec_andc (vector bool int, vector float);
8070 vector bool int vec_andc (vector bool int, vector bool int);
8071 vector signed int vec_andc (vector bool int, vector signed int);
8072 vector signed int vec_andc (vector signed int, vector bool int);
8073 vector signed int vec_andc (vector signed int, vector signed int);
8074 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8075 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8076 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8077 vector bool short vec_andc (vector bool short, vector bool short);
8078 vector signed short vec_andc (vector bool short, vector signed short);
8079 vector signed short vec_andc (vector signed short, vector bool short);
8080 vector signed short vec_andc (vector signed short, vector signed short);
8081 vector unsigned short vec_andc (vector bool short,
8082 vector unsigned short);
8083 vector unsigned short vec_andc (vector unsigned short,
8085 vector unsigned short vec_andc (vector unsigned short,
8086 vector unsigned short);
8087 vector signed char vec_andc (vector bool char, vector signed char);
8088 vector bool char vec_andc (vector bool char, vector bool char);
8089 vector signed char vec_andc (vector signed char, vector bool char);
8090 vector signed char vec_andc (vector signed char, vector signed char);
8091 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8092 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8093 vector unsigned char vec_andc (vector unsigned char,
8094 vector unsigned char);
8096 vector unsigned char vec_avg (vector unsigned char,
8097 vector unsigned char);
8098 vector signed char vec_avg (vector signed char, vector signed char);
8099 vector unsigned short vec_avg (vector unsigned short,
8100 vector unsigned short);
8101 vector signed short vec_avg (vector signed short, vector signed short);
8102 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8103 vector signed int vec_avg (vector signed int, vector signed int);
8105 vector signed int vec_vavgsw (vector signed int, vector signed int);
8107 vector unsigned int vec_vavguw (vector unsigned int,
8108 vector unsigned int);
8110 vector signed short vec_vavgsh (vector signed short,
8111 vector signed short);
8113 vector unsigned short vec_vavguh (vector unsigned short,
8114 vector unsigned short);
8116 vector signed char vec_vavgsb (vector signed char, vector signed char);
8118 vector unsigned char vec_vavgub (vector unsigned char,
8119 vector unsigned char);
8121 vector float vec_ceil (vector float);
8123 vector signed int vec_cmpb (vector float, vector float);
8125 vector bool char vec_cmpeq (vector signed char, vector signed char);
8126 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8127 vector bool short vec_cmpeq (vector signed short, vector signed short);
8128 vector bool short vec_cmpeq (vector unsigned short,
8129 vector unsigned short);
8130 vector bool int vec_cmpeq (vector signed int, vector signed int);
8131 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8132 vector bool int vec_cmpeq (vector float, vector float);
8134 vector bool int vec_vcmpeqfp (vector float, vector float);
8136 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8137 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8139 vector bool short vec_vcmpequh (vector signed short,
8140 vector signed short);
8141 vector bool short vec_vcmpequh (vector unsigned short,
8142 vector unsigned short);
8144 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8145 vector bool char vec_vcmpequb (vector unsigned char,
8146 vector unsigned char);
8148 vector bool int vec_cmpge (vector float, vector float);
8150 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8151 vector bool char vec_cmpgt (vector signed char, vector signed char);
8152 vector bool short vec_cmpgt (vector unsigned short,
8153 vector unsigned short);
8154 vector bool short vec_cmpgt (vector signed short, vector signed short);
8155 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8156 vector bool int vec_cmpgt (vector signed int, vector signed int);
8157 vector bool int vec_cmpgt (vector float, vector float);
8159 vector bool int vec_vcmpgtfp (vector float, vector float);
8161 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8163 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8165 vector bool short vec_vcmpgtsh (vector signed short,
8166 vector signed short);
8168 vector bool short vec_vcmpgtuh (vector unsigned short,
8169 vector unsigned short);
8171 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8173 vector bool char vec_vcmpgtub (vector unsigned char,
8174 vector unsigned char);
8176 vector bool int vec_cmple (vector float, vector float);
8178 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8179 vector bool char vec_cmplt (vector signed char, vector signed char);
8180 vector bool short vec_cmplt (vector unsigned short,
8181 vector unsigned short);
8182 vector bool short vec_cmplt (vector signed short, vector signed short);
8183 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8184 vector bool int vec_cmplt (vector signed int, vector signed int);
8185 vector bool int vec_cmplt (vector float, vector float);
8187 vector float vec_ctf (vector unsigned int, const int);
8188 vector float vec_ctf (vector signed int, const int);
8190 vector float vec_vcfsx (vector signed int, const int);
8192 vector float vec_vcfux (vector unsigned int, const int);
8194 vector signed int vec_cts (vector float, const int);
8196 vector unsigned int vec_ctu (vector float, const int);
8198 void vec_dss (const int);
8200 void vec_dssall (void);
8202 void vec_dst (const vector unsigned char *, int, const int);
8203 void vec_dst (const vector signed char *, int, const int);
8204 void vec_dst (const vector bool char *, int, const int);
8205 void vec_dst (const vector unsigned short *, int, const int);
8206 void vec_dst (const vector signed short *, int, const int);
8207 void vec_dst (const vector bool short *, int, const int);
8208 void vec_dst (const vector pixel *, int, const int);
8209 void vec_dst (const vector unsigned int *, int, const int);
8210 void vec_dst (const vector signed int *, int, const int);
8211 void vec_dst (const vector bool int *, int, const int);
8212 void vec_dst (const vector float *, int, const int);
8213 void vec_dst (const unsigned char *, int, const int);
8214 void vec_dst (const signed char *, int, const int);
8215 void vec_dst (const unsigned short *, int, const int);
8216 void vec_dst (const short *, int, const int);
8217 void vec_dst (const unsigned int *, int, const int);
8218 void vec_dst (const int *, int, const int);
8219 void vec_dst (const unsigned long *, int, const int);
8220 void vec_dst (const long *, int, const int);
8221 void vec_dst (const float *, int, const int);
8223 void vec_dstst (const vector unsigned char *, int, const int);
8224 void vec_dstst (const vector signed char *, int, const int);
8225 void vec_dstst (const vector bool char *, int, const int);
8226 void vec_dstst (const vector unsigned short *, int, const int);
8227 void vec_dstst (const vector signed short *, int, const int);
8228 void vec_dstst (const vector bool short *, int, const int);
8229 void vec_dstst (const vector pixel *, int, const int);
8230 void vec_dstst (const vector unsigned int *, int, const int);
8231 void vec_dstst (const vector signed int *, int, const int);
8232 void vec_dstst (const vector bool int *, int, const int);
8233 void vec_dstst (const vector float *, int, const int);
8234 void vec_dstst (const unsigned char *, int, const int);
8235 void vec_dstst (const signed char *, int, const int);
8236 void vec_dstst (const unsigned short *, int, const int);
8237 void vec_dstst (const short *, int, const int);
8238 void vec_dstst (const unsigned int *, int, const int);
8239 void vec_dstst (const int *, int, const int);
8240 void vec_dstst (const unsigned long *, int, const int);
8241 void vec_dstst (const long *, int, const int);
8242 void vec_dstst (const float *, int, const int);
8244 void vec_dststt (const vector unsigned char *, int, const int);
8245 void vec_dststt (const vector signed char *, int, const int);
8246 void vec_dststt (const vector bool char *, int, const int);
8247 void vec_dststt (const vector unsigned short *, int, const int);
8248 void vec_dststt (const vector signed short *, int, const int);
8249 void vec_dststt (const vector bool short *, int, const int);
8250 void vec_dststt (const vector pixel *, int, const int);
8251 void vec_dststt (const vector unsigned int *, int, const int);
8252 void vec_dststt (const vector signed int *, int, const int);
8253 void vec_dststt (const vector bool int *, int, const int);
8254 void vec_dststt (const vector float *, int, const int);
8255 void vec_dststt (const unsigned char *, int, const int);
8256 void vec_dststt (const signed char *, int, const int);
8257 void vec_dststt (const unsigned short *, int, const int);
8258 void vec_dststt (const short *, int, const int);
8259 void vec_dststt (const unsigned int *, int, const int);
8260 void vec_dststt (const int *, int, const int);
8261 void vec_dststt (const unsigned long *, int, const int);
8262 void vec_dststt (const long *, int, const int);
8263 void vec_dststt (const float *, int, const int);
8265 void vec_dstt (const vector unsigned char *, int, const int);
8266 void vec_dstt (const vector signed char *, int, const int);
8267 void vec_dstt (const vector bool char *, int, const int);
8268 void vec_dstt (const vector unsigned short *, int, const int);
8269 void vec_dstt (const vector signed short *, int, const int);
8270 void vec_dstt (const vector bool short *, int, const int);
8271 void vec_dstt (const vector pixel *, int, const int);
8272 void vec_dstt (const vector unsigned int *, int, const int);
8273 void vec_dstt (const vector signed int *, int, const int);
8274 void vec_dstt (const vector bool int *, int, const int);
8275 void vec_dstt (const vector float *, int, const int);
8276 void vec_dstt (const unsigned char *, int, const int);
8277 void vec_dstt (const signed char *, int, const int);
8278 void vec_dstt (const unsigned short *, int, const int);
8279 void vec_dstt (const short *, int, const int);
8280 void vec_dstt (const unsigned int *, int, const int);
8281 void vec_dstt (const int *, int, const int);
8282 void vec_dstt (const unsigned long *, int, const int);
8283 void vec_dstt (const long *, int, const int);
8284 void vec_dstt (const float *, int, const int);
8286 vector float vec_expte (vector float);
8288 vector float vec_floor (vector float);
8290 vector float vec_ld (int, const vector float *);
8291 vector float vec_ld (int, const float *);
8292 vector bool int vec_ld (int, const vector bool int *);
8293 vector signed int vec_ld (int, const vector signed int *);
8294 vector signed int vec_ld (int, const int *);
8295 vector signed int vec_ld (int, const long *);
8296 vector unsigned int vec_ld (int, const vector unsigned int *);
8297 vector unsigned int vec_ld (int, const unsigned int *);
8298 vector unsigned int vec_ld (int, const unsigned long *);
8299 vector bool short vec_ld (int, const vector bool short *);
8300 vector pixel vec_ld (int, const vector pixel *);
8301 vector signed short vec_ld (int, const vector signed short *);
8302 vector signed short vec_ld (int, const short *);
8303 vector unsigned short vec_ld (int, const vector unsigned short *);
8304 vector unsigned short vec_ld (int, const unsigned short *);
8305 vector bool char vec_ld (int, const vector bool char *);
8306 vector signed char vec_ld (int, const vector signed char *);
8307 vector signed char vec_ld (int, const signed char *);
8308 vector unsigned char vec_ld (int, const vector unsigned char *);
8309 vector unsigned char vec_ld (int, const unsigned char *);
8311 vector signed char vec_lde (int, const signed char *);
8312 vector unsigned char vec_lde (int, const unsigned char *);
8313 vector signed short vec_lde (int, const short *);
8314 vector unsigned short vec_lde (int, const unsigned short *);
8315 vector float vec_lde (int, const float *);
8316 vector signed int vec_lde (int, const int *);
8317 vector unsigned int vec_lde (int, const unsigned int *);
8318 vector signed int vec_lde (int, const long *);
8319 vector unsigned int vec_lde (int, const unsigned long *);
8321 vector float vec_lvewx (int, float *);
8322 vector signed int vec_lvewx (int, int *);
8323 vector unsigned int vec_lvewx (int, unsigned int *);
8324 vector signed int vec_lvewx (int, long *);
8325 vector unsigned int vec_lvewx (int, unsigned long *);
8327 vector signed short vec_lvehx (int, short *);
8328 vector unsigned short vec_lvehx (int, unsigned short *);
8330 vector signed char vec_lvebx (int, char *);
8331 vector unsigned char vec_lvebx (int, unsigned char *);
8333 vector float vec_ldl (int, const vector float *);
8334 vector float vec_ldl (int, const float *);
8335 vector bool int vec_ldl (int, const vector bool int *);
8336 vector signed int vec_ldl (int, const vector signed int *);
8337 vector signed int vec_ldl (int, const int *);
8338 vector signed int vec_ldl (int, const long *);
8339 vector unsigned int vec_ldl (int, const vector unsigned int *);
8340 vector unsigned int vec_ldl (int, const unsigned int *);
8341 vector unsigned int vec_ldl (int, const unsigned long *);
8342 vector bool short vec_ldl (int, const vector bool short *);
8343 vector pixel vec_ldl (int, const vector pixel *);
8344 vector signed short vec_ldl (int, const vector signed short *);
8345 vector signed short vec_ldl (int, const short *);
8346 vector unsigned short vec_ldl (int, const vector unsigned short *);
8347 vector unsigned short vec_ldl (int, const unsigned short *);
8348 vector bool char vec_ldl (int, const vector bool char *);
8349 vector signed char vec_ldl (int, const vector signed char *);
8350 vector signed char vec_ldl (int, const signed char *);
8351 vector unsigned char vec_ldl (int, const vector unsigned char *);
8352 vector unsigned char vec_ldl (int, const unsigned char *);
8354 vector float vec_loge (vector float);
8356 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8357 vector unsigned char vec_lvsl (int, const volatile signed char *);
8358 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8359 vector unsigned char vec_lvsl (int, const volatile short *);
8360 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8361 vector unsigned char vec_lvsl (int, const volatile int *);
8362 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8363 vector unsigned char vec_lvsl (int, const volatile long *);
8364 vector unsigned char vec_lvsl (int, const volatile float *);
8366 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8367 vector unsigned char vec_lvsr (int, const volatile signed char *);
8368 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8369 vector unsigned char vec_lvsr (int, const volatile short *);
8370 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8371 vector unsigned char vec_lvsr (int, const volatile int *);
8372 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8373 vector unsigned char vec_lvsr (int, const volatile long *);
8374 vector unsigned char vec_lvsr (int, const volatile float *);
8376 vector float vec_madd (vector float, vector float, vector float);
8378 vector signed short vec_madds (vector signed short,
8379 vector signed short,
8380 vector signed short);
8382 vector unsigned char vec_max (vector bool char, vector unsigned char);
8383 vector unsigned char vec_max (vector unsigned char, vector bool char);
8384 vector unsigned char vec_max (vector unsigned char,
8385 vector unsigned char);
8386 vector signed char vec_max (vector bool char, vector signed char);
8387 vector signed char vec_max (vector signed char, vector bool char);
8388 vector signed char vec_max (vector signed char, vector signed char);
8389 vector unsigned short vec_max (vector bool short,
8390 vector unsigned short);
8391 vector unsigned short vec_max (vector unsigned short,
8393 vector unsigned short vec_max (vector unsigned short,
8394 vector unsigned short);
8395 vector signed short vec_max (vector bool short, vector signed short);
8396 vector signed short vec_max (vector signed short, vector bool short);
8397 vector signed short vec_max (vector signed short, vector signed short);
8398 vector unsigned int vec_max (vector bool int, vector unsigned int);
8399 vector unsigned int vec_max (vector unsigned int, vector bool int);
8400 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8401 vector signed int vec_max (vector bool int, vector signed int);
8402 vector signed int vec_max (vector signed int, vector bool int);
8403 vector signed int vec_max (vector signed int, vector signed int);
8404 vector float vec_max (vector float, vector float);
8406 vector float vec_vmaxfp (vector float, vector float);
8408 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8409 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8410 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8412 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8413 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8414 vector unsigned int vec_vmaxuw (vector unsigned int,
8415 vector unsigned int);
8417 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8418 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8419 vector signed short vec_vmaxsh (vector signed short,
8420 vector signed short);
8422 vector unsigned short vec_vmaxuh (vector bool short,
8423 vector unsigned short);
8424 vector unsigned short vec_vmaxuh (vector unsigned short,
8426 vector unsigned short vec_vmaxuh (vector unsigned short,
8427 vector unsigned short);
8429 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8430 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8431 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8433 vector unsigned char vec_vmaxub (vector bool char,
8434 vector unsigned char);
8435 vector unsigned char vec_vmaxub (vector unsigned char,
8437 vector unsigned char vec_vmaxub (vector unsigned char,
8438 vector unsigned char);
8440 vector bool char vec_mergeh (vector bool char, vector bool char);
8441 vector signed char vec_mergeh (vector signed char, vector signed char);
8442 vector unsigned char vec_mergeh (vector unsigned char,
8443 vector unsigned char);
8444 vector bool short vec_mergeh (vector bool short, vector bool short);
8445 vector pixel vec_mergeh (vector pixel, vector pixel);
8446 vector signed short vec_mergeh (vector signed short,
8447 vector signed short);
8448 vector unsigned short vec_mergeh (vector unsigned short,
8449 vector unsigned short);
8450 vector float vec_mergeh (vector float, vector float);
8451 vector bool int vec_mergeh (vector bool int, vector bool int);
8452 vector signed int vec_mergeh (vector signed int, vector signed int);
8453 vector unsigned int vec_mergeh (vector unsigned int,
8454 vector unsigned int);
8456 vector float vec_vmrghw (vector float, vector float);
8457 vector bool int vec_vmrghw (vector bool int, vector bool int);
8458 vector signed int vec_vmrghw (vector signed int, vector signed int);
8459 vector unsigned int vec_vmrghw (vector unsigned int,
8460 vector unsigned int);
8462 vector bool short vec_vmrghh (vector bool short, vector bool short);
8463 vector signed short vec_vmrghh (vector signed short,
8464 vector signed short);
8465 vector unsigned short vec_vmrghh (vector unsigned short,
8466 vector unsigned short);
8467 vector pixel vec_vmrghh (vector pixel, vector pixel);
8469 vector bool char vec_vmrghb (vector bool char, vector bool char);
8470 vector signed char vec_vmrghb (vector signed char, vector signed char);
8471 vector unsigned char vec_vmrghb (vector unsigned char,
8472 vector unsigned char);
8474 vector bool char vec_mergel (vector bool char, vector bool char);
8475 vector signed char vec_mergel (vector signed char, vector signed char);
8476 vector unsigned char vec_mergel (vector unsigned char,
8477 vector unsigned char);
8478 vector bool short vec_mergel (vector bool short, vector bool short);
8479 vector pixel vec_mergel (vector pixel, vector pixel);
8480 vector signed short vec_mergel (vector signed short,
8481 vector signed short);
8482 vector unsigned short vec_mergel (vector unsigned short,
8483 vector unsigned short);
8484 vector float vec_mergel (vector float, vector float);
8485 vector bool int vec_mergel (vector bool int, vector bool int);
8486 vector signed int vec_mergel (vector signed int, vector signed int);
8487 vector unsigned int vec_mergel (vector unsigned int,
8488 vector unsigned int);
8490 vector float vec_vmrglw (vector float, vector float);
8491 vector signed int vec_vmrglw (vector signed int, vector signed int);
8492 vector unsigned int vec_vmrglw (vector unsigned int,
8493 vector unsigned int);
8494 vector bool int vec_vmrglw (vector bool int, vector bool int);
8496 vector bool short vec_vmrglh (vector bool short, vector bool short);
8497 vector signed short vec_vmrglh (vector signed short,
8498 vector signed short);
8499 vector unsigned short vec_vmrglh (vector unsigned short,
8500 vector unsigned short);
8501 vector pixel vec_vmrglh (vector pixel, vector pixel);
8503 vector bool char vec_vmrglb (vector bool char, vector bool char);
8504 vector signed char vec_vmrglb (vector signed char, vector signed char);
8505 vector unsigned char vec_vmrglb (vector unsigned char,
8506 vector unsigned char);
8508 vector unsigned short vec_mfvscr (void);
8510 vector unsigned char vec_min (vector bool char, vector unsigned char);
8511 vector unsigned char vec_min (vector unsigned char, vector bool char);
8512 vector unsigned char vec_min (vector unsigned char,
8513 vector unsigned char);
8514 vector signed char vec_min (vector bool char, vector signed char);
8515 vector signed char vec_min (vector signed char, vector bool char);
8516 vector signed char vec_min (vector signed char, vector signed char);
8517 vector unsigned short vec_min (vector bool short,
8518 vector unsigned short);
8519 vector unsigned short vec_min (vector unsigned short,
8521 vector unsigned short vec_min (vector unsigned short,
8522 vector unsigned short);
8523 vector signed short vec_min (vector bool short, vector signed short);
8524 vector signed short vec_min (vector signed short, vector bool short);
8525 vector signed short vec_min (vector signed short, vector signed short);
8526 vector unsigned int vec_min (vector bool int, vector unsigned int);
8527 vector unsigned int vec_min (vector unsigned int, vector bool int);
8528 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8529 vector signed int vec_min (vector bool int, vector signed int);
8530 vector signed int vec_min (vector signed int, vector bool int);
8531 vector signed int vec_min (vector signed int, vector signed int);
8532 vector float vec_min (vector float, vector float);
8534 vector float vec_vminfp (vector float, vector float);
8536 vector signed int vec_vminsw (vector bool int, vector signed int);
8537 vector signed int vec_vminsw (vector signed int, vector bool int);
8538 vector signed int vec_vminsw (vector signed int, vector signed int);
8540 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8541 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8542 vector unsigned int vec_vminuw (vector unsigned int,
8543 vector unsigned int);
8545 vector signed short vec_vminsh (vector bool short, vector signed short);
8546 vector signed short vec_vminsh (vector signed short, vector bool short);
8547 vector signed short vec_vminsh (vector signed short,
8548 vector signed short);
8550 vector unsigned short vec_vminuh (vector bool short,
8551 vector unsigned short);
8552 vector unsigned short vec_vminuh (vector unsigned short,
8554 vector unsigned short vec_vminuh (vector unsigned short,
8555 vector unsigned short);
8557 vector signed char vec_vminsb (vector bool char, vector signed char);
8558 vector signed char vec_vminsb (vector signed char, vector bool char);
8559 vector signed char vec_vminsb (vector signed char, vector signed char);
8561 vector unsigned char vec_vminub (vector bool char,
8562 vector unsigned char);
8563 vector unsigned char vec_vminub (vector unsigned char,
8565 vector unsigned char vec_vminub (vector unsigned char,
8566 vector unsigned char);
8568 vector signed short vec_mladd (vector signed short,
8569 vector signed short,
8570 vector signed short);
8571 vector signed short vec_mladd (vector signed short,
8572 vector unsigned short,
8573 vector unsigned short);
8574 vector signed short vec_mladd (vector unsigned short,
8575 vector signed short,
8576 vector signed short);
8577 vector unsigned short vec_mladd (vector unsigned short,
8578 vector unsigned short,
8579 vector unsigned short);
8581 vector signed short vec_mradds (vector signed short,
8582 vector signed short,
8583 vector signed short);
8585 vector unsigned int vec_msum (vector unsigned char,
8586 vector unsigned char,
8587 vector unsigned int);
8588 vector signed int vec_msum (vector signed char,
8589 vector unsigned char,
8591 vector unsigned int vec_msum (vector unsigned short,
8592 vector unsigned short,
8593 vector unsigned int);
8594 vector signed int vec_msum (vector signed short,
8595 vector signed short,
8598 vector signed int vec_vmsumshm (vector signed short,
8599 vector signed short,
8602 vector unsigned int vec_vmsumuhm (vector unsigned short,
8603 vector unsigned short,
8604 vector unsigned int);
8606 vector signed int vec_vmsummbm (vector signed char,
8607 vector unsigned char,
8610 vector unsigned int vec_vmsumubm (vector unsigned char,
8611 vector unsigned char,
8612 vector unsigned int);
8614 vector unsigned int vec_msums (vector unsigned short,
8615 vector unsigned short,
8616 vector unsigned int);
8617 vector signed int vec_msums (vector signed short,
8618 vector signed short,
8621 vector signed int vec_vmsumshs (vector signed short,
8622 vector signed short,
8625 vector unsigned int vec_vmsumuhs (vector unsigned short,
8626 vector unsigned short,
8627 vector unsigned int);
8629 void vec_mtvscr (vector signed int);
8630 void vec_mtvscr (vector unsigned int);
8631 void vec_mtvscr (vector bool int);
8632 void vec_mtvscr (vector signed short);
8633 void vec_mtvscr (vector unsigned short);
8634 void vec_mtvscr (vector bool short);
8635 void vec_mtvscr (vector pixel);
8636 void vec_mtvscr (vector signed char);
8637 void vec_mtvscr (vector unsigned char);
8638 void vec_mtvscr (vector bool char);
8640 vector unsigned short vec_mule (vector unsigned char,
8641 vector unsigned char);
8642 vector signed short vec_mule (vector signed char,
8643 vector signed char);
8644 vector unsigned int vec_mule (vector unsigned short,
8645 vector unsigned short);
8646 vector signed int vec_mule (vector signed short, vector signed short);
8648 vector signed int vec_vmulesh (vector signed short,
8649 vector signed short);
8651 vector unsigned int vec_vmuleuh (vector unsigned short,
8652 vector unsigned short);
8654 vector signed short vec_vmulesb (vector signed char,
8655 vector signed char);
8657 vector unsigned short vec_vmuleub (vector unsigned char,
8658 vector unsigned char);
8660 vector unsigned short vec_mulo (vector unsigned char,
8661 vector unsigned char);
8662 vector signed short vec_mulo (vector signed char, vector signed char);
8663 vector unsigned int vec_mulo (vector unsigned short,
8664 vector unsigned short);
8665 vector signed int vec_mulo (vector signed short, vector signed short);
8667 vector signed int vec_vmulosh (vector signed short,
8668 vector signed short);
8670 vector unsigned int vec_vmulouh (vector unsigned short,
8671 vector unsigned short);
8673 vector signed short vec_vmulosb (vector signed char,
8674 vector signed char);
8676 vector unsigned short vec_vmuloub (vector unsigned char,
8677 vector unsigned char);
8679 vector float vec_nmsub (vector float, vector float, vector float);
8681 vector float vec_nor (vector float, vector float);
8682 vector signed int vec_nor (vector signed int, vector signed int);
8683 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8684 vector bool int vec_nor (vector bool int, vector bool int);
8685 vector signed short vec_nor (vector signed short, vector signed short);
8686 vector unsigned short vec_nor (vector unsigned short,
8687 vector unsigned short);
8688 vector bool short vec_nor (vector bool short, vector bool short);
8689 vector signed char vec_nor (vector signed char, vector signed char);
8690 vector unsigned char vec_nor (vector unsigned char,
8691 vector unsigned char);
8692 vector bool char vec_nor (vector bool char, vector bool char);
8694 vector float vec_or (vector float, vector float);
8695 vector float vec_or (vector float, vector bool int);
8696 vector float vec_or (vector bool int, vector float);
8697 vector bool int vec_or (vector bool int, vector bool int);
8698 vector signed int vec_or (vector bool int, vector signed int);
8699 vector signed int vec_or (vector signed int, vector bool int);
8700 vector signed int vec_or (vector signed int, vector signed int);
8701 vector unsigned int vec_or (vector bool int, vector unsigned int);
8702 vector unsigned int vec_or (vector unsigned int, vector bool int);
8703 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8704 vector bool short vec_or (vector bool short, vector bool short);
8705 vector signed short vec_or (vector bool short, vector signed short);
8706 vector signed short vec_or (vector signed short, vector bool short);
8707 vector signed short vec_or (vector signed short, vector signed short);
8708 vector unsigned short vec_or (vector bool short, vector unsigned short);
8709 vector unsigned short vec_or (vector unsigned short, vector bool short);
8710 vector unsigned short vec_or (vector unsigned short,
8711 vector unsigned short);
8712 vector signed char vec_or (vector bool char, vector signed char);
8713 vector bool char vec_or (vector bool char, vector bool char);
8714 vector signed char vec_or (vector signed char, vector bool char);
8715 vector signed char vec_or (vector signed char, vector signed char);
8716 vector unsigned char vec_or (vector bool char, vector unsigned char);
8717 vector unsigned char vec_or (vector unsigned char, vector bool char);
8718 vector unsigned char vec_or (vector unsigned char,
8719 vector unsigned char);
8721 vector signed char vec_pack (vector signed short, vector signed short);
8722 vector unsigned char vec_pack (vector unsigned short,
8723 vector unsigned short);
8724 vector bool char vec_pack (vector bool short, vector bool short);
8725 vector signed short vec_pack (vector signed int, vector signed int);
8726 vector unsigned short vec_pack (vector unsigned int,
8727 vector unsigned int);
8728 vector bool short vec_pack (vector bool int, vector bool int);
8730 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8731 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8732 vector unsigned short vec_vpkuwum (vector unsigned int,
8733 vector unsigned int);
8735 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8736 vector signed char vec_vpkuhum (vector signed short,
8737 vector signed short);
8738 vector unsigned char vec_vpkuhum (vector unsigned short,
8739 vector unsigned short);
8741 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8743 vector unsigned char vec_packs (vector unsigned short,
8744 vector unsigned short);
8745 vector signed char vec_packs (vector signed short, vector signed short);
8746 vector unsigned short vec_packs (vector unsigned int,
8747 vector unsigned int);
8748 vector signed short vec_packs (vector signed int, vector signed int);
8750 vector signed short vec_vpkswss (vector signed int, vector signed int);
8752 vector unsigned short vec_vpkuwus (vector unsigned int,
8753 vector unsigned int);
8755 vector signed char vec_vpkshss (vector signed short,
8756 vector signed short);
8758 vector unsigned char vec_vpkuhus (vector unsigned short,
8759 vector unsigned short);
8761 vector unsigned char vec_packsu (vector unsigned short,
8762 vector unsigned short);
8763 vector unsigned char vec_packsu (vector signed short,
8764 vector signed short);
8765 vector unsigned short vec_packsu (vector unsigned int,
8766 vector unsigned int);
8767 vector unsigned short vec_packsu (vector signed int, vector signed int);
8769 vector unsigned short vec_vpkswus (vector signed int,
8772 vector unsigned char vec_vpkshus (vector signed short,
8773 vector signed short);
8775 vector float vec_perm (vector float,
8777 vector unsigned char);
8778 vector signed int vec_perm (vector signed int,
8780 vector unsigned char);
8781 vector unsigned int vec_perm (vector unsigned int,
8782 vector unsigned int,
8783 vector unsigned char);
8784 vector bool int vec_perm (vector bool int,
8786 vector unsigned char);
8787 vector signed short vec_perm (vector signed short,
8788 vector signed short,
8789 vector unsigned char);
8790 vector unsigned short vec_perm (vector unsigned short,
8791 vector unsigned short,
8792 vector unsigned char);
8793 vector bool short vec_perm (vector bool short,
8795 vector unsigned char);
8796 vector pixel vec_perm (vector pixel,
8798 vector unsigned char);
8799 vector signed char vec_perm (vector signed char,
8801 vector unsigned char);
8802 vector unsigned char vec_perm (vector unsigned char,
8803 vector unsigned char,
8804 vector unsigned char);
8805 vector bool char vec_perm (vector bool char,
8807 vector unsigned char);
8809 vector float vec_re (vector float);
8811 vector signed char vec_rl (vector signed char,
8812 vector unsigned char);
8813 vector unsigned char vec_rl (vector unsigned char,
8814 vector unsigned char);
8815 vector signed short vec_rl (vector signed short, vector unsigned short);
8816 vector unsigned short vec_rl (vector unsigned short,
8817 vector unsigned short);
8818 vector signed int vec_rl (vector signed int, vector unsigned int);
8819 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8821 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8822 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8824 vector signed short vec_vrlh (vector signed short,
8825 vector unsigned short);
8826 vector unsigned short vec_vrlh (vector unsigned short,
8827 vector unsigned short);
8829 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8830 vector unsigned char vec_vrlb (vector unsigned char,
8831 vector unsigned char);
8833 vector float vec_round (vector float);
8835 vector float vec_rsqrte (vector float);
8837 vector float vec_sel (vector float, vector float, vector bool int);
8838 vector float vec_sel (vector float, vector float, vector unsigned int);
8839 vector signed int vec_sel (vector signed int,
8842 vector signed int vec_sel (vector signed int,
8844 vector unsigned int);
8845 vector unsigned int vec_sel (vector unsigned int,
8846 vector unsigned int,
8848 vector unsigned int vec_sel (vector unsigned int,
8849 vector unsigned int,
8850 vector unsigned int);
8851 vector bool int vec_sel (vector bool int,
8854 vector bool int vec_sel (vector bool int,
8856 vector unsigned int);
8857 vector signed short vec_sel (vector signed short,
8858 vector signed short,
8860 vector signed short vec_sel (vector signed short,
8861 vector signed short,
8862 vector unsigned short);
8863 vector unsigned short vec_sel (vector unsigned short,
8864 vector unsigned short,
8866 vector unsigned short vec_sel (vector unsigned short,
8867 vector unsigned short,
8868 vector unsigned short);
8869 vector bool short vec_sel (vector bool short,
8872 vector bool short vec_sel (vector bool short,
8874 vector unsigned short);
8875 vector signed char vec_sel (vector signed char,
8878 vector signed char vec_sel (vector signed char,
8880 vector unsigned char);
8881 vector unsigned char vec_sel (vector unsigned char,
8882 vector unsigned char,
8884 vector unsigned char vec_sel (vector unsigned char,
8885 vector unsigned char,
8886 vector unsigned char);
8887 vector bool char vec_sel (vector bool char,
8890 vector bool char vec_sel (vector bool char,
8892 vector unsigned char);
8894 vector signed char vec_sl (vector signed char,
8895 vector unsigned char);
8896 vector unsigned char vec_sl (vector unsigned char,
8897 vector unsigned char);
8898 vector signed short vec_sl (vector signed short, vector unsigned short);
8899 vector unsigned short vec_sl (vector unsigned short,
8900 vector unsigned short);
8901 vector signed int vec_sl (vector signed int, vector unsigned int);
8902 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
8904 vector signed int vec_vslw (vector signed int, vector unsigned int);
8905 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
8907 vector signed short vec_vslh (vector signed short,
8908 vector unsigned short);
8909 vector unsigned short vec_vslh (vector unsigned short,
8910 vector unsigned short);
8912 vector signed char vec_vslb (vector signed char, vector unsigned char);
8913 vector unsigned char vec_vslb (vector unsigned char,
8914 vector unsigned char);
8916 vector float vec_sld (vector float, vector float, const int);
8917 vector signed int vec_sld (vector signed int,
8920 vector unsigned int vec_sld (vector unsigned int,
8921 vector unsigned int,
8923 vector bool int vec_sld (vector bool int,
8926 vector signed short vec_sld (vector signed short,
8927 vector signed short,
8929 vector unsigned short vec_sld (vector unsigned short,
8930 vector unsigned short,
8932 vector bool short vec_sld (vector bool short,
8935 vector pixel vec_sld (vector pixel,
8938 vector signed char vec_sld (vector signed char,
8941 vector unsigned char vec_sld (vector unsigned char,
8942 vector unsigned char,
8944 vector bool char vec_sld (vector bool char,
8948 vector signed int vec_sll (vector signed int,
8949 vector unsigned int);
8950 vector signed int vec_sll (vector signed int,
8951 vector unsigned short);
8952 vector signed int vec_sll (vector signed int,
8953 vector unsigned char);
8954 vector unsigned int vec_sll (vector unsigned int,
8955 vector unsigned int);
8956 vector unsigned int vec_sll (vector unsigned int,
8957 vector unsigned short);
8958 vector unsigned int vec_sll (vector unsigned int,
8959 vector unsigned char);
8960 vector bool int vec_sll (vector bool int,
8961 vector unsigned int);
8962 vector bool int vec_sll (vector bool int,
8963 vector unsigned short);
8964 vector bool int vec_sll (vector bool int,
8965 vector unsigned char);
8966 vector signed short vec_sll (vector signed short,
8967 vector unsigned int);
8968 vector signed short vec_sll (vector signed short,
8969 vector unsigned short);
8970 vector signed short vec_sll (vector signed short,
8971 vector unsigned char);
8972 vector unsigned short vec_sll (vector unsigned short,
8973 vector unsigned int);
8974 vector unsigned short vec_sll (vector unsigned short,
8975 vector unsigned short);
8976 vector unsigned short vec_sll (vector unsigned short,
8977 vector unsigned char);
8978 vector bool short vec_sll (vector bool short, vector unsigned int);
8979 vector bool short vec_sll (vector bool short, vector unsigned short);
8980 vector bool short vec_sll (vector bool short, vector unsigned char);
8981 vector pixel vec_sll (vector pixel, vector unsigned int);
8982 vector pixel vec_sll (vector pixel, vector unsigned short);
8983 vector pixel vec_sll (vector pixel, vector unsigned char);
8984 vector signed char vec_sll (vector signed char, vector unsigned int);
8985 vector signed char vec_sll (vector signed char, vector unsigned short);
8986 vector signed char vec_sll (vector signed char, vector unsigned char);
8987 vector unsigned char vec_sll (vector unsigned char,
8988 vector unsigned int);
8989 vector unsigned char vec_sll (vector unsigned char,
8990 vector unsigned short);
8991 vector unsigned char vec_sll (vector unsigned char,
8992 vector unsigned char);
8993 vector bool char vec_sll (vector bool char, vector unsigned int);
8994 vector bool char vec_sll (vector bool char, vector unsigned short);
8995 vector bool char vec_sll (vector bool char, vector unsigned char);
8997 vector float vec_slo (vector float, vector signed char);
8998 vector float vec_slo (vector float, vector unsigned char);
8999 vector signed int vec_slo (vector signed int, vector signed char);
9000 vector signed int vec_slo (vector signed int, vector unsigned char);
9001 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9002 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9003 vector signed short vec_slo (vector signed short, vector signed char);
9004 vector signed short vec_slo (vector signed short, vector unsigned char);
9005 vector unsigned short vec_slo (vector unsigned short,
9006 vector signed char);
9007 vector unsigned short vec_slo (vector unsigned short,
9008 vector unsigned char);
9009 vector pixel vec_slo (vector pixel, vector signed char);
9010 vector pixel vec_slo (vector pixel, vector unsigned char);
9011 vector signed char vec_slo (vector signed char, vector signed char);
9012 vector signed char vec_slo (vector signed char, vector unsigned char);
9013 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9014 vector unsigned char vec_slo (vector unsigned char,
9015 vector unsigned char);
9017 vector signed char vec_splat (vector signed char, const int);
9018 vector unsigned char vec_splat (vector unsigned char, const int);
9019 vector bool char vec_splat (vector bool char, const int);
9020 vector signed short vec_splat (vector signed short, const int);
9021 vector unsigned short vec_splat (vector unsigned short, const int);
9022 vector bool short vec_splat (vector bool short, const int);
9023 vector pixel vec_splat (vector pixel, const int);
9024 vector float vec_splat (vector float, const int);
9025 vector signed int vec_splat (vector signed int, const int);
9026 vector unsigned int vec_splat (vector unsigned int, const int);
9027 vector bool int vec_splat (vector bool int, const int);
9029 vector float vec_vspltw (vector float, const int);
9030 vector signed int vec_vspltw (vector signed int, const int);
9031 vector unsigned int vec_vspltw (vector unsigned int, const int);
9032 vector bool int vec_vspltw (vector bool int, const int);
9034 vector bool short vec_vsplth (vector bool short, const int);
9035 vector signed short vec_vsplth (vector signed short, const int);
9036 vector unsigned short vec_vsplth (vector unsigned short, const int);
9037 vector pixel vec_vsplth (vector pixel, const int);
9039 vector signed char vec_vspltb (vector signed char, const int);
9040 vector unsigned char vec_vspltb (vector unsigned char, const int);
9041 vector bool char vec_vspltb (vector bool char, const int);
9043 vector signed char vec_splat_s8 (const int);
9045 vector signed short vec_splat_s16 (const int);
9047 vector signed int vec_splat_s32 (const int);
9049 vector unsigned char vec_splat_u8 (const int);
9051 vector unsigned short vec_splat_u16 (const int);
9053 vector unsigned int vec_splat_u32 (const int);
9055 vector signed char vec_sr (vector signed char, vector unsigned char);
9056 vector unsigned char vec_sr (vector unsigned char,
9057 vector unsigned char);
9058 vector signed short vec_sr (vector signed short,
9059 vector unsigned short);
9060 vector unsigned short vec_sr (vector unsigned short,
9061 vector unsigned short);
9062 vector signed int vec_sr (vector signed int, vector unsigned int);
9063 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9065 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9066 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9068 vector signed short vec_vsrh (vector signed short,
9069 vector unsigned short);
9070 vector unsigned short vec_vsrh (vector unsigned short,
9071 vector unsigned short);
9073 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9074 vector unsigned char vec_vsrb (vector unsigned char,
9075 vector unsigned char);
9077 vector signed char vec_sra (vector signed char, vector unsigned char);
9078 vector unsigned char vec_sra (vector unsigned char,
9079 vector unsigned char);
9080 vector signed short vec_sra (vector signed short,
9081 vector unsigned short);
9082 vector unsigned short vec_sra (vector unsigned short,
9083 vector unsigned short);
9084 vector signed int vec_sra (vector signed int, vector unsigned int);
9085 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9087 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9088 vector unsigned int vec_vsraw (vector unsigned int,
9089 vector unsigned int);
9091 vector signed short vec_vsrah (vector signed short,
9092 vector unsigned short);
9093 vector unsigned short vec_vsrah (vector unsigned short,
9094 vector unsigned short);
9096 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9097 vector unsigned char vec_vsrab (vector unsigned char,
9098 vector unsigned char);
9100 vector signed int vec_srl (vector signed int, vector unsigned int);
9101 vector signed int vec_srl (vector signed int, vector unsigned short);
9102 vector signed int vec_srl (vector signed int, vector unsigned char);
9103 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9104 vector unsigned int vec_srl (vector unsigned int,
9105 vector unsigned short);
9106 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9107 vector bool int vec_srl (vector bool int, vector unsigned int);
9108 vector bool int vec_srl (vector bool int, vector unsigned short);
9109 vector bool int vec_srl (vector bool int, vector unsigned char);
9110 vector signed short vec_srl (vector signed short, vector unsigned int);
9111 vector signed short vec_srl (vector signed short,
9112 vector unsigned short);
9113 vector signed short vec_srl (vector signed short, vector unsigned char);
9114 vector unsigned short vec_srl (vector unsigned short,
9115 vector unsigned int);
9116 vector unsigned short vec_srl (vector unsigned short,
9117 vector unsigned short);
9118 vector unsigned short vec_srl (vector unsigned short,
9119 vector unsigned char);
9120 vector bool short vec_srl (vector bool short, vector unsigned int);
9121 vector bool short vec_srl (vector bool short, vector unsigned short);
9122 vector bool short vec_srl (vector bool short, vector unsigned char);
9123 vector pixel vec_srl (vector pixel, vector unsigned int);
9124 vector pixel vec_srl (vector pixel, vector unsigned short);
9125 vector pixel vec_srl (vector pixel, vector unsigned char);
9126 vector signed char vec_srl (vector signed char, vector unsigned int);
9127 vector signed char vec_srl (vector signed char, vector unsigned short);
9128 vector signed char vec_srl (vector signed char, vector unsigned char);
9129 vector unsigned char vec_srl (vector unsigned char,
9130 vector unsigned int);
9131 vector unsigned char vec_srl (vector unsigned char,
9132 vector unsigned short);
9133 vector unsigned char vec_srl (vector unsigned char,
9134 vector unsigned char);
9135 vector bool char vec_srl (vector bool char, vector unsigned int);
9136 vector bool char vec_srl (vector bool char, vector unsigned short);
9137 vector bool char vec_srl (vector bool char, vector unsigned char);
9139 vector float vec_sro (vector float, vector signed char);
9140 vector float vec_sro (vector float, vector unsigned char);
9141 vector signed int vec_sro (vector signed int, vector signed char);
9142 vector signed int vec_sro (vector signed int, vector unsigned char);
9143 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9144 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9145 vector signed short vec_sro (vector signed short, vector signed char);
9146 vector signed short vec_sro (vector signed short, vector unsigned char);
9147 vector unsigned short vec_sro (vector unsigned short,
9148 vector signed char);
9149 vector unsigned short vec_sro (vector unsigned short,
9150 vector unsigned char);
9151 vector pixel vec_sro (vector pixel, vector signed char);
9152 vector pixel vec_sro (vector pixel, vector unsigned char);
9153 vector signed char vec_sro (vector signed char, vector signed char);
9154 vector signed char vec_sro (vector signed char, vector unsigned char);
9155 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9156 vector unsigned char vec_sro (vector unsigned char,
9157 vector unsigned char);
9159 void vec_st (vector float, int, vector float *);
9160 void vec_st (vector float, int, float *);
9161 void vec_st (vector signed int, int, vector signed int *);
9162 void vec_st (vector signed int, int, int *);
9163 void vec_st (vector unsigned int, int, vector unsigned int *);
9164 void vec_st (vector unsigned int, int, unsigned int *);
9165 void vec_st (vector bool int, int, vector bool int *);
9166 void vec_st (vector bool int, int, unsigned int *);
9167 void vec_st (vector bool int, int, int *);
9168 void vec_st (vector signed short, int, vector signed short *);
9169 void vec_st (vector signed short, int, short *);
9170 void vec_st (vector unsigned short, int, vector unsigned short *);
9171 void vec_st (vector unsigned short, int, unsigned short *);
9172 void vec_st (vector bool short, int, vector bool short *);
9173 void vec_st (vector bool short, int, unsigned short *);
9174 void vec_st (vector pixel, int, vector pixel *);
9175 void vec_st (vector pixel, int, unsigned short *);
9176 void vec_st (vector pixel, int, short *);
9177 void vec_st (vector bool short, int, short *);
9178 void vec_st (vector signed char, int, vector signed char *);
9179 void vec_st (vector signed char, int, signed char *);
9180 void vec_st (vector unsigned char, int, vector unsigned char *);
9181 void vec_st (vector unsigned char, int, unsigned char *);
9182 void vec_st (vector bool char, int, vector bool char *);
9183 void vec_st (vector bool char, int, unsigned char *);
9184 void vec_st (vector bool char, int, signed char *);
9186 void vec_ste (vector signed char, int, signed char *);
9187 void vec_ste (vector unsigned char, int, unsigned char *);
9188 void vec_ste (vector bool char, int, signed char *);
9189 void vec_ste (vector bool char, int, unsigned char *);
9190 void vec_ste (vector signed short, int, short *);
9191 void vec_ste (vector unsigned short, int, unsigned short *);
9192 void vec_ste (vector bool short, int, short *);
9193 void vec_ste (vector bool short, int, unsigned short *);
9194 void vec_ste (vector pixel, int, short *);
9195 void vec_ste (vector pixel, int, unsigned short *);
9196 void vec_ste (vector float, int, float *);
9197 void vec_ste (vector signed int, int, int *);
9198 void vec_ste (vector unsigned int, int, unsigned int *);
9199 void vec_ste (vector bool int, int, int *);
9200 void vec_ste (vector bool int, int, unsigned int *);
9202 void vec_stvewx (vector float, int, float *);
9203 void vec_stvewx (vector signed int, int, int *);
9204 void vec_stvewx (vector unsigned int, int, unsigned int *);
9205 void vec_stvewx (vector bool int, int, int *);
9206 void vec_stvewx (vector bool int, int, unsigned int *);
9208 void vec_stvehx (vector signed short, int, short *);
9209 void vec_stvehx (vector unsigned short, int, unsigned short *);
9210 void vec_stvehx (vector bool short, int, short *);
9211 void vec_stvehx (vector bool short, int, unsigned short *);
9212 void vec_stvehx (vector pixel, int, short *);
9213 void vec_stvehx (vector pixel, int, unsigned short *);
9215 void vec_stvebx (vector signed char, int, signed char *);
9216 void vec_stvebx (vector unsigned char, int, unsigned char *);
9217 void vec_stvebx (vector bool char, int, signed char *);
9218 void vec_stvebx (vector bool char, int, unsigned char *);
9220 void vec_stl (vector float, int, vector float *);
9221 void vec_stl (vector float, int, float *);
9222 void vec_stl (vector signed int, int, vector signed int *);
9223 void vec_stl (vector signed int, int, int *);
9224 void vec_stl (vector unsigned int, int, vector unsigned int *);
9225 void vec_stl (vector unsigned int, int, unsigned int *);
9226 void vec_stl (vector bool int, int, vector bool int *);
9227 void vec_stl (vector bool int, int, unsigned int *);
9228 void vec_stl (vector bool int, int, int *);
9229 void vec_stl (vector signed short, int, vector signed short *);
9230 void vec_stl (vector signed short, int, short *);
9231 void vec_stl (vector unsigned short, int, vector unsigned short *);
9232 void vec_stl (vector unsigned short, int, unsigned short *);
9233 void vec_stl (vector bool short, int, vector bool short *);
9234 void vec_stl (vector bool short, int, unsigned short *);
9235 void vec_stl (vector bool short, int, short *);
9236 void vec_stl (vector pixel, int, vector pixel *);
9237 void vec_stl (vector pixel, int, unsigned short *);
9238 void vec_stl (vector pixel, int, short *);
9239 void vec_stl (vector signed char, int, vector signed char *);
9240 void vec_stl (vector signed char, int, signed char *);
9241 void vec_stl (vector unsigned char, int, vector unsigned char *);
9242 void vec_stl (vector unsigned char, int, unsigned char *);
9243 void vec_stl (vector bool char, int, vector bool char *);
9244 void vec_stl (vector bool char, int, unsigned char *);
9245 void vec_stl (vector bool char, int, signed char *);
9247 vector signed char vec_sub (vector bool char, vector signed char);
9248 vector signed char vec_sub (vector signed char, vector bool char);
9249 vector signed char vec_sub (vector signed char, vector signed char);
9250 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9251 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9252 vector unsigned char vec_sub (vector unsigned char,
9253 vector unsigned char);
9254 vector signed short vec_sub (vector bool short, vector signed short);
9255 vector signed short vec_sub (vector signed short, vector bool short);
9256 vector signed short vec_sub (vector signed short, vector signed short);
9257 vector unsigned short vec_sub (vector bool short,
9258 vector unsigned short);
9259 vector unsigned short vec_sub (vector unsigned short,
9261 vector unsigned short vec_sub (vector unsigned short,
9262 vector unsigned short);
9263 vector signed int vec_sub (vector bool int, vector signed int);
9264 vector signed int vec_sub (vector signed int, vector bool int);
9265 vector signed int vec_sub (vector signed int, vector signed int);
9266 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9267 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9268 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9269 vector float vec_sub (vector float, vector float);
9271 vector float vec_vsubfp (vector float, vector float);
9273 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9274 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9275 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9276 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9277 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9278 vector unsigned int vec_vsubuwm (vector unsigned int,
9279 vector unsigned int);
9281 vector signed short vec_vsubuhm (vector bool short,
9282 vector signed short);
9283 vector signed short vec_vsubuhm (vector signed short,
9285 vector signed short vec_vsubuhm (vector signed short,
9286 vector signed short);
9287 vector unsigned short vec_vsubuhm (vector bool short,
9288 vector unsigned short);
9289 vector unsigned short vec_vsubuhm (vector unsigned short,
9291 vector unsigned short vec_vsubuhm (vector unsigned short,
9292 vector unsigned short);
9294 vector signed char vec_vsububm (vector bool char, vector signed char);
9295 vector signed char vec_vsububm (vector signed char, vector bool char);
9296 vector signed char vec_vsububm (vector signed char, vector signed char);
9297 vector unsigned char vec_vsububm (vector bool char,
9298 vector unsigned char);
9299 vector unsigned char vec_vsububm (vector unsigned char,
9301 vector unsigned char vec_vsububm (vector unsigned char,
9302 vector unsigned char);
9304 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9306 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9307 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9308 vector unsigned char vec_subs (vector unsigned char,
9309 vector unsigned char);
9310 vector signed char vec_subs (vector bool char, vector signed char);
9311 vector signed char vec_subs (vector signed char, vector bool char);
9312 vector signed char vec_subs (vector signed char, vector signed char);
9313 vector unsigned short vec_subs (vector bool short,
9314 vector unsigned short);
9315 vector unsigned short vec_subs (vector unsigned short,
9317 vector unsigned short vec_subs (vector unsigned short,
9318 vector unsigned short);
9319 vector signed short vec_subs (vector bool short, vector signed short);
9320 vector signed short vec_subs (vector signed short, vector bool short);
9321 vector signed short vec_subs (vector signed short, vector signed short);
9322 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9323 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9324 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9325 vector signed int vec_subs (vector bool int, vector signed int);
9326 vector signed int vec_subs (vector signed int, vector bool int);
9327 vector signed int vec_subs (vector signed int, vector signed int);
9329 vector signed int vec_vsubsws (vector bool int, vector signed int);
9330 vector signed int vec_vsubsws (vector signed int, vector bool int);
9331 vector signed int vec_vsubsws (vector signed int, vector signed int);
9333 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9334 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9335 vector unsigned int vec_vsubuws (vector unsigned int,
9336 vector unsigned int);
9338 vector signed short vec_vsubshs (vector bool short,
9339 vector signed short);
9340 vector signed short vec_vsubshs (vector signed short,
9342 vector signed short vec_vsubshs (vector signed short,
9343 vector signed short);
9345 vector unsigned short vec_vsubuhs (vector bool short,
9346 vector unsigned short);
9347 vector unsigned short vec_vsubuhs (vector unsigned short,
9349 vector unsigned short vec_vsubuhs (vector unsigned short,
9350 vector unsigned short);
9352 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9353 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9354 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9356 vector unsigned char vec_vsububs (vector bool char,
9357 vector unsigned char);
9358 vector unsigned char vec_vsububs (vector unsigned char,
9360 vector unsigned char vec_vsububs (vector unsigned char,
9361 vector unsigned char);
9363 vector unsigned int vec_sum4s (vector unsigned char,
9364 vector unsigned int);
9365 vector signed int vec_sum4s (vector signed char, vector signed int);
9366 vector signed int vec_sum4s (vector signed short, vector signed int);
9368 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9370 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9372 vector unsigned int vec_vsum4ubs (vector unsigned char,
9373 vector unsigned int);
9375 vector signed int vec_sum2s (vector signed int, vector signed int);
9377 vector signed int vec_sums (vector signed int, vector signed int);
9379 vector float vec_trunc (vector float);
9381 vector signed short vec_unpackh (vector signed char);
9382 vector bool short vec_unpackh (vector bool char);
9383 vector signed int vec_unpackh (vector signed short);
9384 vector bool int vec_unpackh (vector bool short);
9385 vector unsigned int vec_unpackh (vector pixel);
9387 vector bool int vec_vupkhsh (vector bool short);
9388 vector signed int vec_vupkhsh (vector signed short);
9390 vector unsigned int vec_vupkhpx (vector pixel);
9392 vector bool short vec_vupkhsb (vector bool char);
9393 vector signed short vec_vupkhsb (vector signed char);
9395 vector signed short vec_unpackl (vector signed char);
9396 vector bool short vec_unpackl (vector bool char);
9397 vector unsigned int vec_unpackl (vector pixel);
9398 vector signed int vec_unpackl (vector signed short);
9399 vector bool int vec_unpackl (vector bool short);
9401 vector unsigned int vec_vupklpx (vector pixel);
9403 vector bool int vec_vupklsh (vector bool short);
9404 vector signed int vec_vupklsh (vector signed short);
9406 vector bool short vec_vupklsb (vector bool char);
9407 vector signed short vec_vupklsb (vector signed char);
9409 vector float vec_xor (vector float, vector float);
9410 vector float vec_xor (vector float, vector bool int);
9411 vector float vec_xor (vector bool int, vector float);
9412 vector bool int vec_xor (vector bool int, vector bool int);
9413 vector signed int vec_xor (vector bool int, vector signed int);
9414 vector signed int vec_xor (vector signed int, vector bool int);
9415 vector signed int vec_xor (vector signed int, vector signed int);
9416 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9417 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9418 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9419 vector bool short vec_xor (vector bool short, vector bool short);
9420 vector signed short vec_xor (vector bool short, vector signed short);
9421 vector signed short vec_xor (vector signed short, vector bool short);
9422 vector signed short vec_xor (vector signed short, vector signed short);
9423 vector unsigned short vec_xor (vector bool short,
9424 vector unsigned short);
9425 vector unsigned short vec_xor (vector unsigned short,
9427 vector unsigned short vec_xor (vector unsigned short,
9428 vector unsigned short);
9429 vector signed char vec_xor (vector bool char, vector signed char);
9430 vector bool char vec_xor (vector bool char, vector bool char);
9431 vector signed char vec_xor (vector signed char, vector bool char);
9432 vector signed char vec_xor (vector signed char, vector signed char);
9433 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9434 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9435 vector unsigned char vec_xor (vector unsigned char,
9436 vector unsigned char);
9438 int vec_all_eq (vector signed char, vector bool char);
9439 int vec_all_eq (vector signed char, vector signed char);
9440 int vec_all_eq (vector unsigned char, vector bool char);
9441 int vec_all_eq (vector unsigned char, vector unsigned char);
9442 int vec_all_eq (vector bool char, vector bool char);
9443 int vec_all_eq (vector bool char, vector unsigned char);
9444 int vec_all_eq (vector bool char, vector signed char);
9445 int vec_all_eq (vector signed short, vector bool short);
9446 int vec_all_eq (vector signed short, vector signed short);
9447 int vec_all_eq (vector unsigned short, vector bool short);
9448 int vec_all_eq (vector unsigned short, vector unsigned short);
9449 int vec_all_eq (vector bool short, vector bool short);
9450 int vec_all_eq (vector bool short, vector unsigned short);
9451 int vec_all_eq (vector bool short, vector signed short);
9452 int vec_all_eq (vector pixel, vector pixel);
9453 int vec_all_eq (vector signed int, vector bool int);
9454 int vec_all_eq (vector signed int, vector signed int);
9455 int vec_all_eq (vector unsigned int, vector bool int);
9456 int vec_all_eq (vector unsigned int, vector unsigned int);
9457 int vec_all_eq (vector bool int, vector bool int);
9458 int vec_all_eq (vector bool int, vector unsigned int);
9459 int vec_all_eq (vector bool int, vector signed int);
9460 int vec_all_eq (vector float, vector float);
9462 int vec_all_ge (vector bool char, vector unsigned char);
9463 int vec_all_ge (vector unsigned char, vector bool char);
9464 int vec_all_ge (vector unsigned char, vector unsigned char);
9465 int vec_all_ge (vector bool char, vector signed char);
9466 int vec_all_ge (vector signed char, vector bool char);
9467 int vec_all_ge (vector signed char, vector signed char);
9468 int vec_all_ge (vector bool short, vector unsigned short);
9469 int vec_all_ge (vector unsigned short, vector bool short);
9470 int vec_all_ge (vector unsigned short, vector unsigned short);
9471 int vec_all_ge (vector signed short, vector signed short);
9472 int vec_all_ge (vector bool short, vector signed short);
9473 int vec_all_ge (vector signed short, vector bool short);
9474 int vec_all_ge (vector bool int, vector unsigned int);
9475 int vec_all_ge (vector unsigned int, vector bool int);
9476 int vec_all_ge (vector unsigned int, vector unsigned int);
9477 int vec_all_ge (vector bool int, vector signed int);
9478 int vec_all_ge (vector signed int, vector bool int);
9479 int vec_all_ge (vector signed int, vector signed int);
9480 int vec_all_ge (vector float, vector float);
9482 int vec_all_gt (vector bool char, vector unsigned char);
9483 int vec_all_gt (vector unsigned char, vector bool char);
9484 int vec_all_gt (vector unsigned char, vector unsigned char);
9485 int vec_all_gt (vector bool char, vector signed char);
9486 int vec_all_gt (vector signed char, vector bool char);
9487 int vec_all_gt (vector signed char, vector signed char);
9488 int vec_all_gt (vector bool short, vector unsigned short);
9489 int vec_all_gt (vector unsigned short, vector bool short);
9490 int vec_all_gt (vector unsigned short, vector unsigned short);
9491 int vec_all_gt (vector bool short, vector signed short);
9492 int vec_all_gt (vector signed short, vector bool short);
9493 int vec_all_gt (vector signed short, vector signed short);
9494 int vec_all_gt (vector bool int, vector unsigned int);
9495 int vec_all_gt (vector unsigned int, vector bool int);
9496 int vec_all_gt (vector unsigned int, vector unsigned int);
9497 int vec_all_gt (vector bool int, vector signed int);
9498 int vec_all_gt (vector signed int, vector bool int);
9499 int vec_all_gt (vector signed int, vector signed int);
9500 int vec_all_gt (vector float, vector float);
9502 int vec_all_in (vector float, vector float);
9504 int vec_all_le (vector bool char, vector unsigned char);
9505 int vec_all_le (vector unsigned char, vector bool char);
9506 int vec_all_le (vector unsigned char, vector unsigned char);
9507 int vec_all_le (vector bool char, vector signed char);
9508 int vec_all_le (vector signed char, vector bool char);
9509 int vec_all_le (vector signed char, vector signed char);
9510 int vec_all_le (vector bool short, vector unsigned short);
9511 int vec_all_le (vector unsigned short, vector bool short);
9512 int vec_all_le (vector unsigned short, vector unsigned short);
9513 int vec_all_le (vector bool short, vector signed short);
9514 int vec_all_le (vector signed short, vector bool short);
9515 int vec_all_le (vector signed short, vector signed short);
9516 int vec_all_le (vector bool int, vector unsigned int);
9517 int vec_all_le (vector unsigned int, vector bool int);
9518 int vec_all_le (vector unsigned int, vector unsigned int);
9519 int vec_all_le (vector bool int, vector signed int);
9520 int vec_all_le (vector signed int, vector bool int);
9521 int vec_all_le (vector signed int, vector signed int);
9522 int vec_all_le (vector float, vector float);
9524 int vec_all_lt (vector bool char, vector unsigned char);
9525 int vec_all_lt (vector unsigned char, vector bool char);
9526 int vec_all_lt (vector unsigned char, vector unsigned char);
9527 int vec_all_lt (vector bool char, vector signed char);
9528 int vec_all_lt (vector signed char, vector bool char);
9529 int vec_all_lt (vector signed char, vector signed char);
9530 int vec_all_lt (vector bool short, vector unsigned short);
9531 int vec_all_lt (vector unsigned short, vector bool short);
9532 int vec_all_lt (vector unsigned short, vector unsigned short);
9533 int vec_all_lt (vector bool short, vector signed short);
9534 int vec_all_lt (vector signed short, vector bool short);
9535 int vec_all_lt (vector signed short, vector signed short);
9536 int vec_all_lt (vector bool int, vector unsigned int);
9537 int vec_all_lt (vector unsigned int, vector bool int);
9538 int vec_all_lt (vector unsigned int, vector unsigned int);
9539 int vec_all_lt (vector bool int, vector signed int);
9540 int vec_all_lt (vector signed int, vector bool int);
9541 int vec_all_lt (vector signed int, vector signed int);
9542 int vec_all_lt (vector float, vector float);
9544 int vec_all_nan (vector float);
9546 int vec_all_ne (vector signed char, vector bool char);
9547 int vec_all_ne (vector signed char, vector signed char);
9548 int vec_all_ne (vector unsigned char, vector bool char);
9549 int vec_all_ne (vector unsigned char, vector unsigned char);
9550 int vec_all_ne (vector bool char, vector bool char);
9551 int vec_all_ne (vector bool char, vector unsigned char);
9552 int vec_all_ne (vector bool char, vector signed char);
9553 int vec_all_ne (vector signed short, vector bool short);
9554 int vec_all_ne (vector signed short, vector signed short);
9555 int vec_all_ne (vector unsigned short, vector bool short);
9556 int vec_all_ne (vector unsigned short, vector unsigned short);
9557 int vec_all_ne (vector bool short, vector bool short);
9558 int vec_all_ne (vector bool short, vector unsigned short);
9559 int vec_all_ne (vector bool short, vector signed short);
9560 int vec_all_ne (vector pixel, vector pixel);
9561 int vec_all_ne (vector signed int, vector bool int);
9562 int vec_all_ne (vector signed int, vector signed int);
9563 int vec_all_ne (vector unsigned int, vector bool int);
9564 int vec_all_ne (vector unsigned int, vector unsigned int);
9565 int vec_all_ne (vector bool int, vector bool int);
9566 int vec_all_ne (vector bool int, vector unsigned int);
9567 int vec_all_ne (vector bool int, vector signed int);
9568 int vec_all_ne (vector float, vector float);
9570 int vec_all_nge (vector float, vector float);
9572 int vec_all_ngt (vector float, vector float);
9574 int vec_all_nle (vector float, vector float);
9576 int vec_all_nlt (vector float, vector float);
9578 int vec_all_numeric (vector float);
9580 int vec_any_eq (vector signed char, vector bool char);
9581 int vec_any_eq (vector signed char, vector signed char);
9582 int vec_any_eq (vector unsigned char, vector bool char);
9583 int vec_any_eq (vector unsigned char, vector unsigned char);
9584 int vec_any_eq (vector bool char, vector bool char);
9585 int vec_any_eq (vector bool char, vector unsigned char);
9586 int vec_any_eq (vector bool char, vector signed char);
9587 int vec_any_eq (vector signed short, vector bool short);
9588 int vec_any_eq (vector signed short, vector signed short);
9589 int vec_any_eq (vector unsigned short, vector bool short);
9590 int vec_any_eq (vector unsigned short, vector unsigned short);
9591 int vec_any_eq (vector bool short, vector bool short);
9592 int vec_any_eq (vector bool short, vector unsigned short);
9593 int vec_any_eq (vector bool short, vector signed short);
9594 int vec_any_eq (vector pixel, vector pixel);
9595 int vec_any_eq (vector signed int, vector bool int);
9596 int vec_any_eq (vector signed int, vector signed int);
9597 int vec_any_eq (vector unsigned int, vector bool int);
9598 int vec_any_eq (vector unsigned int, vector unsigned int);
9599 int vec_any_eq (vector bool int, vector bool int);
9600 int vec_any_eq (vector bool int, vector unsigned int);
9601 int vec_any_eq (vector bool int, vector signed int);
9602 int vec_any_eq (vector float, vector float);
9604 int vec_any_ge (vector signed char, vector bool char);
9605 int vec_any_ge (vector unsigned char, vector bool char);
9606 int vec_any_ge (vector unsigned char, vector unsigned char);
9607 int vec_any_ge (vector signed char, vector signed char);
9608 int vec_any_ge (vector bool char, vector unsigned char);
9609 int vec_any_ge (vector bool char, vector signed char);
9610 int vec_any_ge (vector unsigned short, vector bool short);
9611 int vec_any_ge (vector unsigned short, vector unsigned short);
9612 int vec_any_ge (vector signed short, vector signed short);
9613 int vec_any_ge (vector signed short, vector bool short);
9614 int vec_any_ge (vector bool short, vector unsigned short);
9615 int vec_any_ge (vector bool short, vector signed short);
9616 int vec_any_ge (vector signed int, vector bool int);
9617 int vec_any_ge (vector unsigned int, vector bool int);
9618 int vec_any_ge (vector unsigned int, vector unsigned int);
9619 int vec_any_ge (vector signed int, vector signed int);
9620 int vec_any_ge (vector bool int, vector unsigned int);
9621 int vec_any_ge (vector bool int, vector signed int);
9622 int vec_any_ge (vector float, vector float);
9624 int vec_any_gt (vector bool char, vector unsigned char);
9625 int vec_any_gt (vector unsigned char, vector bool char);
9626 int vec_any_gt (vector unsigned char, vector unsigned char);
9627 int vec_any_gt (vector bool char, vector signed char);
9628 int vec_any_gt (vector signed char, vector bool char);
9629 int vec_any_gt (vector signed char, vector signed char);
9630 int vec_any_gt (vector bool short, vector unsigned short);
9631 int vec_any_gt (vector unsigned short, vector bool short);
9632 int vec_any_gt (vector unsigned short, vector unsigned short);
9633 int vec_any_gt (vector bool short, vector signed short);
9634 int vec_any_gt (vector signed short, vector bool short);
9635 int vec_any_gt (vector signed short, vector signed short);
9636 int vec_any_gt (vector bool int, vector unsigned int);
9637 int vec_any_gt (vector unsigned int, vector bool int);
9638 int vec_any_gt (vector unsigned int, vector unsigned int);
9639 int vec_any_gt (vector bool int, vector signed int);
9640 int vec_any_gt (vector signed int, vector bool int);
9641 int vec_any_gt (vector signed int, vector signed int);
9642 int vec_any_gt (vector float, vector float);
9644 int vec_any_le (vector bool char, vector unsigned char);
9645 int vec_any_le (vector unsigned char, vector bool char);
9646 int vec_any_le (vector unsigned char, vector unsigned char);
9647 int vec_any_le (vector bool char, vector signed char);
9648 int vec_any_le (vector signed char, vector bool char);
9649 int vec_any_le (vector signed char, vector signed char);
9650 int vec_any_le (vector bool short, vector unsigned short);
9651 int vec_any_le (vector unsigned short, vector bool short);
9652 int vec_any_le (vector unsigned short, vector unsigned short);
9653 int vec_any_le (vector bool short, vector signed short);
9654 int vec_any_le (vector signed short, vector bool short);
9655 int vec_any_le (vector signed short, vector signed short);
9656 int vec_any_le (vector bool int, vector unsigned int);
9657 int vec_any_le (vector unsigned int, vector bool int);
9658 int vec_any_le (vector unsigned int, vector unsigned int);
9659 int vec_any_le (vector bool int, vector signed int);
9660 int vec_any_le (vector signed int, vector bool int);
9661 int vec_any_le (vector signed int, vector signed int);
9662 int vec_any_le (vector float, vector float);
9664 int vec_any_lt (vector bool char, vector unsigned char);
9665 int vec_any_lt (vector unsigned char, vector bool char);
9666 int vec_any_lt (vector unsigned char, vector unsigned char);
9667 int vec_any_lt (vector bool char, vector signed char);
9668 int vec_any_lt (vector signed char, vector bool char);
9669 int vec_any_lt (vector signed char, vector signed char);
9670 int vec_any_lt (vector bool short, vector unsigned short);
9671 int vec_any_lt (vector unsigned short, vector bool short);
9672 int vec_any_lt (vector unsigned short, vector unsigned short);
9673 int vec_any_lt (vector bool short, vector signed short);
9674 int vec_any_lt (vector signed short, vector bool short);
9675 int vec_any_lt (vector signed short, vector signed short);
9676 int vec_any_lt (vector bool int, vector unsigned int);
9677 int vec_any_lt (vector unsigned int, vector bool int);
9678 int vec_any_lt (vector unsigned int, vector unsigned int);
9679 int vec_any_lt (vector bool int, vector signed int);
9680 int vec_any_lt (vector signed int, vector bool int);
9681 int vec_any_lt (vector signed int, vector signed int);
9682 int vec_any_lt (vector float, vector float);
9684 int vec_any_nan (vector float);
9686 int vec_any_ne (vector signed char, vector bool char);
9687 int vec_any_ne (vector signed char, vector signed char);
9688 int vec_any_ne (vector unsigned char, vector bool char);
9689 int vec_any_ne (vector unsigned char, vector unsigned char);
9690 int vec_any_ne (vector bool char, vector bool char);
9691 int vec_any_ne (vector bool char, vector unsigned char);
9692 int vec_any_ne (vector bool char, vector signed char);
9693 int vec_any_ne (vector signed short, vector bool short);
9694 int vec_any_ne (vector signed short, vector signed short);
9695 int vec_any_ne (vector unsigned short, vector bool short);
9696 int vec_any_ne (vector unsigned short, vector unsigned short);
9697 int vec_any_ne (vector bool short, vector bool short);
9698 int vec_any_ne (vector bool short, vector unsigned short);
9699 int vec_any_ne (vector bool short, vector signed short);
9700 int vec_any_ne (vector pixel, vector pixel);
9701 int vec_any_ne (vector signed int, vector bool int);
9702 int vec_any_ne (vector signed int, vector signed int);
9703 int vec_any_ne (vector unsigned int, vector bool int);
9704 int vec_any_ne (vector unsigned int, vector unsigned int);
9705 int vec_any_ne (vector bool int, vector bool int);
9706 int vec_any_ne (vector bool int, vector unsigned int);
9707 int vec_any_ne (vector bool int, vector signed int);
9708 int vec_any_ne (vector float, vector float);
9710 int vec_any_nge (vector float, vector float);
9712 int vec_any_ngt (vector float, vector float);
9714 int vec_any_nle (vector float, vector float);
9716 int vec_any_nlt (vector float, vector float);
9718 int vec_any_numeric (vector float);
9720 int vec_any_out (vector float, vector float);
9723 @node SPARC VIS Built-in Functions
9724 @subsection SPARC VIS Built-in Functions
9726 GCC supports SIMD operations on the SPARC using both the generic vector
9727 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9728 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9729 switch, the VIS extension is exposed as the following built-in functions:
9732 typedef int v2si __attribute__ ((vector_size (8)));
9733 typedef short v4hi __attribute__ ((vector_size (8)));
9734 typedef short v2hi __attribute__ ((vector_size (4)));
9735 typedef char v8qi __attribute__ ((vector_size (8)));
9736 typedef char v4qi __attribute__ ((vector_size (4)));
9738 void * __builtin_vis_alignaddr (void *, long);
9739 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9740 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9741 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9742 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9744 v4hi __builtin_vis_fexpand (v4qi);
9746 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9747 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9748 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9749 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9750 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9751 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9752 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9754 v4qi __builtin_vis_fpack16 (v4hi);
9755 v8qi __builtin_vis_fpack32 (v2si, v2si);
9756 v2hi __builtin_vis_fpackfix (v2si);
9757 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9759 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9762 @node Target Format Checks
9763 @section Format Checks Specific to Particular Target Machines
9765 For some target machines, GCC supports additional options to the
9767 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9770 * Solaris Format Checks::
9773 @node Solaris Format Checks
9774 @subsection Solaris Format Checks
9776 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9777 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9778 conversions, and the two-argument @code{%b} conversion for displaying
9779 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9782 @section Pragmas Accepted by GCC
9786 GCC supports several types of pragmas, primarily in order to compile
9787 code originally written for other compilers. Note that in general
9788 we do not recommend the use of pragmas; @xref{Function Attributes},
9789 for further explanation.
9794 * RS/6000 and PowerPC Pragmas::
9797 * Symbol-Renaming Pragmas::
9798 * Structure-Packing Pragmas::
9800 * Diagnostic Pragmas::
9801 * Visibility Pragmas::
9805 @subsection ARM Pragmas
9807 The ARM target defines pragmas for controlling the default addition of
9808 @code{long_call} and @code{short_call} attributes to functions.
9809 @xref{Function Attributes}, for information about the effects of these
9814 @cindex pragma, long_calls
9815 Set all subsequent functions to have the @code{long_call} attribute.
9818 @cindex pragma, no_long_calls
9819 Set all subsequent functions to have the @code{short_call} attribute.
9821 @item long_calls_off
9822 @cindex pragma, long_calls_off
9823 Do not affect the @code{long_call} or @code{short_call} attributes of
9824 subsequent functions.
9828 @subsection M32C Pragmas
9831 @item memregs @var{number}
9832 @cindex pragma, memregs
9833 Overrides the command line option @code{-memregs=} for the current
9834 file. Use with care! This pragma must be before any function in the
9835 file, and mixing different memregs values in different objects may
9836 make them incompatible. This pragma is useful when a
9837 performance-critical function uses a memreg for temporary values,
9838 as it may allow you to reduce the number of memregs used.
9842 @node RS/6000 and PowerPC Pragmas
9843 @subsection RS/6000 and PowerPC Pragmas
9845 The RS/6000 and PowerPC targets define one pragma for controlling
9846 whether or not the @code{longcall} attribute is added to function
9847 declarations by default. This pragma overrides the @option{-mlongcall}
9848 option, but not the @code{longcall} and @code{shortcall} attributes.
9849 @xref{RS/6000 and PowerPC Options}, for more information about when long
9850 calls are and are not necessary.
9854 @cindex pragma, longcall
9855 Apply the @code{longcall} attribute to all subsequent function
9859 Do not apply the @code{longcall} attribute to subsequent function
9863 @c Describe c4x pragmas here.
9864 @c Describe h8300 pragmas here.
9865 @c Describe sh pragmas here.
9866 @c Describe v850 pragmas here.
9868 @node Darwin Pragmas
9869 @subsection Darwin Pragmas
9871 The following pragmas are available for all architectures running the
9872 Darwin operating system. These are useful for compatibility with other
9876 @item mark @var{tokens}@dots{}
9877 @cindex pragma, mark
9878 This pragma is accepted, but has no effect.
9880 @item options align=@var{alignment}
9881 @cindex pragma, options align
9882 This pragma sets the alignment of fields in structures. The values of
9883 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9884 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9885 properly; to restore the previous setting, use @code{reset} for the
9888 @item segment @var{tokens}@dots{}
9889 @cindex pragma, segment
9890 This pragma is accepted, but has no effect.
9892 @item unused (@var{var} [, @var{var}]@dots{})
9893 @cindex pragma, unused
9894 This pragma declares variables to be possibly unused. GCC will not
9895 produce warnings for the listed variables. The effect is similar to
9896 that of the @code{unused} attribute, except that this pragma may appear
9897 anywhere within the variables' scopes.
9900 @node Solaris Pragmas
9901 @subsection Solaris Pragmas
9903 The Solaris target supports @code{#pragma redefine_extname}
9904 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
9905 @code{#pragma} directives for compatibility with the system compiler.
9908 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
9909 @cindex pragma, align
9911 Increase the minimum alignment of each @var{variable} to @var{alignment}.
9912 This is the same as GCC's @code{aligned} attribute @pxref{Variable
9913 Attributes}). Macro expansion occurs on the arguments to this pragma
9914 when compiling C. It does not currently occur when compiling C++, but
9915 this is a bug which may be fixed in a future release.
9917 @item fini (@var{function} [, @var{function}]...)
9918 @cindex pragma, fini
9920 This pragma causes each listed @var{function} to be called after
9921 main, or during shared module unloading, by adding a call to the
9922 @code{.fini} section.
9924 @item init (@var{function} [, @var{function}]...)
9925 @cindex pragma, init
9927 This pragma causes each listed @var{function} to be called during
9928 initialization (before @code{main}) or during shared module loading, by
9929 adding a call to the @code{.init} section.
9933 @node Symbol-Renaming Pragmas
9934 @subsection Symbol-Renaming Pragmas
9936 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
9937 supports two @code{#pragma} directives which change the name used in
9938 assembly for a given declaration. These pragmas are only available on
9939 platforms whose system headers need them. To get this effect on all
9940 platforms supported by GCC, use the asm labels extension (@pxref{Asm
9944 @item redefine_extname @var{oldname} @var{newname}
9945 @cindex pragma, redefine_extname
9947 This pragma gives the C function @var{oldname} the assembly symbol
9948 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
9949 will be defined if this pragma is available (currently only on
9952 @item extern_prefix @var{string}
9953 @cindex pragma, extern_prefix
9955 This pragma causes all subsequent external function and variable
9956 declarations to have @var{string} prepended to their assembly symbols.
9957 This effect may be terminated with another @code{extern_prefix} pragma
9958 whose argument is an empty string. The preprocessor macro
9959 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
9960 available (currently only on Tru64 UNIX)@.
9963 These pragmas and the asm labels extension interact in a complicated
9964 manner. Here are some corner cases you may want to be aware of.
9967 @item Both pragmas silently apply only to declarations with external
9968 linkage. Asm labels do not have this restriction.
9970 @item In C++, both pragmas silently apply only to declarations with
9971 ``C'' linkage. Again, asm labels do not have this restriction.
9973 @item If any of the three ways of changing the assembly name of a
9974 declaration is applied to a declaration whose assembly name has
9975 already been determined (either by a previous use of one of these
9976 features, or because the compiler needed the assembly name in order to
9977 generate code), and the new name is different, a warning issues and
9978 the name does not change.
9980 @item The @var{oldname} used by @code{#pragma redefine_extname} is
9981 always the C-language name.
9983 @item If @code{#pragma extern_prefix} is in effect, and a declaration
9984 occurs with an asm label attached, the prefix is silently ignored for
9987 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
9988 apply to the same declaration, whichever triggered first wins, and a
9989 warning issues if they contradict each other. (We would like to have
9990 @code{#pragma redefine_extname} always win, for consistency with asm
9991 labels, but if @code{#pragma extern_prefix} triggers first we have no
9992 way of knowing that that happened.)
9995 @node Structure-Packing Pragmas
9996 @subsection Structure-Packing Pragmas
9998 For compatibility with Win32, GCC supports a set of @code{#pragma}
9999 directives which change the maximum alignment of members of structures
10000 (other than zero-width bitfields), unions, and classes subsequently
10001 defined. The @var{n} value below always is required to be a small power
10002 of two and specifies the new alignment in bytes.
10005 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10006 @item @code{#pragma pack()} sets the alignment to the one that was in
10007 effect when compilation started (see also command line option
10008 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10009 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10010 setting on an internal stack and then optionally sets the new alignment.
10011 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10012 saved at the top of the internal stack (and removes that stack entry).
10013 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10014 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10015 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10016 @code{#pragma pack(pop)}.
10019 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10020 @code{#pragma} which lays out a structure as the documented
10021 @code{__attribute__ ((ms_struct))}.
10023 @item @code{#pragma ms_struct on} turns on the layout for structures
10025 @item @code{#pragma ms_struct off} turns off the layout for structures
10027 @item @code{#pragma ms_struct reset} goes back to the default layout.
10031 @subsection Weak Pragmas
10033 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10034 directives for declaring symbols to be weak, and defining weak
10038 @item #pragma weak @var{symbol}
10039 @cindex pragma, weak
10040 This pragma declares @var{symbol} to be weak, as if the declaration
10041 had the attribute of the same name. The pragma may appear before
10042 or after the declaration of @var{symbol}, but must appear before
10043 either its first use or its definition. It is not an error for
10044 @var{symbol} to never be defined at all.
10046 @item #pragma weak @var{symbol1} = @var{symbol2}
10047 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10048 It is an error if @var{symbol2} is not defined in the current
10052 @node Diagnostic Pragmas
10053 @subsection Diagnostic Pragmas
10055 GCC allows the user to selectively enable or disable certain types of
10056 diagnostics, and change the kind of the diagnostic. For example, a
10057 project's policy might require that all sources compile with
10058 @option{-Werror} but certain files might have exceptions allowing
10059 specific types of warnings. Or, a project might selectively enable
10060 diagnostics and treat them as errors depending on which preprocessor
10061 macros are defined.
10064 @item #pragma GCC diagnostic @var{kind} @var{option}
10065 @cindex pragma, diagnostic
10067 Modifies the disposition of a diagnostic. Note that not all
10068 diagnostics are modifiable; at the moment only warnings (normally
10069 controlled by @samp{-W...}) can be controlled, and not all of them.
10070 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10071 are controllable and which option controls them.
10073 @var{kind} is @samp{error} to treat this diagnostic as an error,
10074 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10075 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10076 @var{option} is a double quoted string which matches the command line
10080 #pragma GCC diagnostic warning "-Wformat"
10081 #pragma GCC diagnostic error "-Wformat"
10082 #pragma GCC diagnostic ignored "-Wformat"
10085 Note that these pragmas override any command line options. Also,
10086 while it is syntactically valid to put these pragmas anywhere in your
10087 sources, the only supported location for them is before any data or
10088 functions are defined. Doing otherwise may result in unpredictable
10089 results depending on how the optimizer manages your sources. If the
10090 same option is listed multiple times, the last one specified is the
10091 one that is in effect. This pragma is not intended to be a general
10092 purpose replacement for command line options, but for implementing
10093 strict control over project policies.
10097 @node Visibility Pragmas
10098 @subsection Visibility Pragmas
10101 @item #pragma GCC visibility push(@var{visibility})
10102 @itemx #pragma GCC visibility pop
10103 @cindex pragma, visibility
10105 This pragma allows the user to set the visibility for multiple
10106 declarations without having to give each a visibility attribute
10107 @xref{Function Attributes}, for more information about visibility and
10108 the attribute syntax.
10110 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10111 declarations. Class members and template specializations are not
10112 affected; if you want to override the visibility for a particular
10113 member or instantiation, you must use an attribute.
10117 @node Unnamed Fields
10118 @section Unnamed struct/union fields within structs/unions
10122 For compatibility with other compilers, GCC allows you to define
10123 a structure or union that contains, as fields, structures and unions
10124 without names. For example:
10137 In this example, the user would be able to access members of the unnamed
10138 union with code like @samp{foo.b}. Note that only unnamed structs and
10139 unions are allowed, you may not have, for example, an unnamed
10142 You must never create such structures that cause ambiguous field definitions.
10143 For example, this structure:
10154 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10155 Such constructs are not supported and must be avoided. In the future,
10156 such constructs may be detected and treated as compilation errors.
10158 @opindex fms-extensions
10159 Unless @option{-fms-extensions} is used, the unnamed field must be a
10160 structure or union definition without a tag (for example, @samp{struct
10161 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10162 also be a definition with a tag such as @samp{struct foo @{ int a;
10163 @};}, a reference to a previously defined structure or union such as
10164 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10165 previously defined structure or union type.
10168 @section Thread-Local Storage
10169 @cindex Thread-Local Storage
10170 @cindex @acronym{TLS}
10173 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10174 are allocated such that there is one instance of the variable per extant
10175 thread. The run-time model GCC uses to implement this originates
10176 in the IA-64 processor-specific ABI, but has since been migrated
10177 to other processors as well. It requires significant support from
10178 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10179 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10180 is not available everywhere.
10182 At the user level, the extension is visible with a new storage
10183 class keyword: @code{__thread}. For example:
10187 extern __thread struct state s;
10188 static __thread char *p;
10191 The @code{__thread} specifier may be used alone, with the @code{extern}
10192 or @code{static} specifiers, but with no other storage class specifier.
10193 When used with @code{extern} or @code{static}, @code{__thread} must appear
10194 immediately after the other storage class specifier.
10196 The @code{__thread} specifier may be applied to any global, file-scoped
10197 static, function-scoped static, or static data member of a class. It may
10198 not be applied to block-scoped automatic or non-static data member.
10200 When the address-of operator is applied to a thread-local variable, it is
10201 evaluated at run-time and returns the address of the current thread's
10202 instance of that variable. An address so obtained may be used by any
10203 thread. When a thread terminates, any pointers to thread-local variables
10204 in that thread become invalid.
10206 No static initialization may refer to the address of a thread-local variable.
10208 In C++, if an initializer is present for a thread-local variable, it must
10209 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10212 See @uref{http://people.redhat.com/drepper/tls.pdf,
10213 ELF Handling For Thread-Local Storage} for a detailed explanation of
10214 the four thread-local storage addressing models, and how the run-time
10215 is expected to function.
10218 * C99 Thread-Local Edits::
10219 * C++98 Thread-Local Edits::
10222 @node C99 Thread-Local Edits
10223 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10225 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10226 that document the exact semantics of the language extension.
10230 @cite{5.1.2 Execution environments}
10232 Add new text after paragraph 1
10235 Within either execution environment, a @dfn{thread} is a flow of
10236 control within a program. It is implementation defined whether
10237 or not there may be more than one thread associated with a program.
10238 It is implementation defined how threads beyond the first are
10239 created, the name and type of the function called at thread
10240 startup, and how threads may be terminated. However, objects
10241 with thread storage duration shall be initialized before thread
10246 @cite{6.2.4 Storage durations of objects}
10248 Add new text before paragraph 3
10251 An object whose identifier is declared with the storage-class
10252 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10253 Its lifetime is the entire execution of the thread, and its
10254 stored value is initialized only once, prior to thread startup.
10258 @cite{6.4.1 Keywords}
10260 Add @code{__thread}.
10263 @cite{6.7.1 Storage-class specifiers}
10265 Add @code{__thread} to the list of storage class specifiers in
10268 Change paragraph 2 to
10271 With the exception of @code{__thread}, at most one storage-class
10272 specifier may be given [@dots{}]. The @code{__thread} specifier may
10273 be used alone, or immediately following @code{extern} or
10277 Add new text after paragraph 6
10280 The declaration of an identifier for a variable that has
10281 block scope that specifies @code{__thread} shall also
10282 specify either @code{extern} or @code{static}.
10284 The @code{__thread} specifier shall be used only with
10289 @node C++98 Thread-Local Edits
10290 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10292 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10293 that document the exact semantics of the language extension.
10297 @b{[intro.execution]}
10299 New text after paragraph 4
10302 A @dfn{thread} is a flow of control within the abstract machine.
10303 It is implementation defined whether or not there may be more than
10307 New text after paragraph 7
10310 It is unspecified whether additional action must be taken to
10311 ensure when and whether side effects are visible to other threads.
10317 Add @code{__thread}.
10320 @b{[basic.start.main]}
10322 Add after paragraph 5
10325 The thread that begins execution at the @code{main} function is called
10326 the @dfn{main thread}. It is implementation defined how functions
10327 beginning threads other than the main thread are designated or typed.
10328 A function so designated, as well as the @code{main} function, is called
10329 a @dfn{thread startup function}. It is implementation defined what
10330 happens if a thread startup function returns. It is implementation
10331 defined what happens to other threads when any thread calls @code{exit}.
10335 @b{[basic.start.init]}
10337 Add after paragraph 4
10340 The storage for an object of thread storage duration shall be
10341 statically initialized before the first statement of the thread startup
10342 function. An object of thread storage duration shall not require
10343 dynamic initialization.
10347 @b{[basic.start.term]}
10349 Add after paragraph 3
10352 The type of an object with thread storage duration shall not have a
10353 non-trivial destructor, nor shall it be an array type whose elements
10354 (directly or indirectly) have non-trivial destructors.
10360 Add ``thread storage duration'' to the list in paragraph 1.
10365 Thread, static, and automatic storage durations are associated with
10366 objects introduced by declarations [@dots{}].
10369 Add @code{__thread} to the list of specifiers in paragraph 3.
10372 @b{[basic.stc.thread]}
10374 New section before @b{[basic.stc.static]}
10377 The keyword @code{__thread} applied to a non-local object gives the
10378 object thread storage duration.
10380 A local variable or class data member declared both @code{static}
10381 and @code{__thread} gives the variable or member thread storage
10386 @b{[basic.stc.static]}
10391 All objects which have neither thread storage duration, dynamic
10392 storage duration nor are local [@dots{}].
10398 Add @code{__thread} to the list in paragraph 1.
10403 With the exception of @code{__thread}, at most one
10404 @var{storage-class-specifier} shall appear in a given
10405 @var{decl-specifier-seq}. The @code{__thread} specifier may
10406 be used alone, or immediately following the @code{extern} or
10407 @code{static} specifiers. [@dots{}]
10410 Add after paragraph 5
10413 The @code{__thread} specifier can be applied only to the names of objects
10414 and to anonymous unions.
10420 Add after paragraph 6
10423 Non-@code{static} members shall not be @code{__thread}.
10427 @node C++ Extensions
10428 @chapter Extensions to the C++ Language
10429 @cindex extensions, C++ language
10430 @cindex C++ language extensions
10432 The GNU compiler provides these extensions to the C++ language (and you
10433 can also use most of the C language extensions in your C++ programs). If you
10434 want to write code that checks whether these features are available, you can
10435 test for the GNU compiler the same way as for C programs: check for a
10436 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10437 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10438 Predefined Macros,cpp,The GNU C Preprocessor}).
10441 * Volatiles:: What constitutes an access to a volatile object.
10442 * Restricted Pointers:: C99 restricted pointers and references.
10443 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10444 * C++ Interface:: You can use a single C++ header file for both
10445 declarations and definitions.
10446 * Template Instantiation:: Methods for ensuring that exactly one copy of
10447 each needed template instantiation is emitted.
10448 * Bound member functions:: You can extract a function pointer to the
10449 method denoted by a @samp{->*} or @samp{.*} expression.
10450 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10451 * Namespace Association:: Strong using-directives for namespace association.
10452 * Java Exceptions:: Tweaking exception handling to work with Java.
10453 * Deprecated Features:: Things will disappear from g++.
10454 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10458 @section When is a Volatile Object Accessed?
10459 @cindex accessing volatiles
10460 @cindex volatile read
10461 @cindex volatile write
10462 @cindex volatile access
10464 Both the C and C++ standard have the concept of volatile objects. These
10465 are normally accessed by pointers and used for accessing hardware. The
10466 standards encourage compilers to refrain from optimizations concerning
10467 accesses to volatile objects. The C standard leaves it implementation
10468 defined as to what constitutes a volatile access. The C++ standard omits
10469 to specify this, except to say that C++ should behave in a similar manner
10470 to C with respect to volatiles, where possible. The minimum either
10471 standard specifies is that at a sequence point all previous accesses to
10472 volatile objects have stabilized and no subsequent accesses have
10473 occurred. Thus an implementation is free to reorder and combine
10474 volatile accesses which occur between sequence points, but cannot do so
10475 for accesses across a sequence point. The use of volatiles does not
10476 allow you to violate the restriction on updating objects multiple times
10477 within a sequence point.
10479 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10481 The behavior differs slightly between C and C++ in the non-obvious cases:
10484 volatile int *src = @var{somevalue};
10488 With C, such expressions are rvalues, and GCC interprets this either as a
10489 read of the volatile object being pointed to or only as request to evaluate
10490 the side-effects. The C++ standard specifies that such expressions do not
10491 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10492 object may be incomplete. The C++ standard does not specify explicitly
10493 that it is this lvalue to rvalue conversion which may be responsible for
10494 causing an access. However, there is reason to believe that it is,
10495 because otherwise certain simple expressions become undefined. However,
10496 because it would surprise most programmers, G++ treats dereferencing a
10497 pointer to volatile object of complete type when the value is unused as
10498 GCC would do for an equivalent type in C. When the object has incomplete
10499 type, G++ issues a warning; if you wish to force an error, you must
10500 force a conversion to rvalue with, for instance, a static cast.
10502 When using a reference to volatile, G++ does not treat equivalent
10503 expressions as accesses to volatiles, but instead issues a warning that
10504 no volatile is accessed. The rationale for this is that otherwise it
10505 becomes difficult to determine where volatile access occur, and not
10506 possible to ignore the return value from functions returning volatile
10507 references. Again, if you wish to force a read, cast the reference to
10510 @node Restricted Pointers
10511 @section Restricting Pointer Aliasing
10512 @cindex restricted pointers
10513 @cindex restricted references
10514 @cindex restricted this pointer
10516 As with the C front end, G++ understands the C99 feature of restricted pointers,
10517 specified with the @code{__restrict__}, or @code{__restrict} type
10518 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10519 language flag, @code{restrict} is not a keyword in C++.
10521 In addition to allowing restricted pointers, you can specify restricted
10522 references, which indicate that the reference is not aliased in the local
10526 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10533 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10534 @var{rref} refers to a (different) unaliased integer.
10536 You may also specify whether a member function's @var{this} pointer is
10537 unaliased by using @code{__restrict__} as a member function qualifier.
10540 void T::fn () __restrict__
10547 Within the body of @code{T::fn}, @var{this} will have the effective
10548 definition @code{T *__restrict__ const this}. Notice that the
10549 interpretation of a @code{__restrict__} member function qualifier is
10550 different to that of @code{const} or @code{volatile} qualifier, in that it
10551 is applied to the pointer rather than the object. This is consistent with
10552 other compilers which implement restricted pointers.
10554 As with all outermost parameter qualifiers, @code{__restrict__} is
10555 ignored in function definition matching. This means you only need to
10556 specify @code{__restrict__} in a function definition, rather than
10557 in a function prototype as well.
10559 @node Vague Linkage
10560 @section Vague Linkage
10561 @cindex vague linkage
10563 There are several constructs in C++ which require space in the object
10564 file but are not clearly tied to a single translation unit. We say that
10565 these constructs have ``vague linkage''. Typically such constructs are
10566 emitted wherever they are needed, though sometimes we can be more
10570 @item Inline Functions
10571 Inline functions are typically defined in a header file which can be
10572 included in many different compilations. Hopefully they can usually be
10573 inlined, but sometimes an out-of-line copy is necessary, if the address
10574 of the function is taken or if inlining fails. In general, we emit an
10575 out-of-line copy in all translation units where one is needed. As an
10576 exception, we only emit inline virtual functions with the vtable, since
10577 it will always require a copy.
10579 Local static variables and string constants used in an inline function
10580 are also considered to have vague linkage, since they must be shared
10581 between all inlined and out-of-line instances of the function.
10585 C++ virtual functions are implemented in most compilers using a lookup
10586 table, known as a vtable. The vtable contains pointers to the virtual
10587 functions provided by a class, and each object of the class contains a
10588 pointer to its vtable (or vtables, in some multiple-inheritance
10589 situations). If the class declares any non-inline, non-pure virtual
10590 functions, the first one is chosen as the ``key method'' for the class,
10591 and the vtable is only emitted in the translation unit where the key
10594 @emph{Note:} If the chosen key method is later defined as inline, the
10595 vtable will still be emitted in every translation unit which defines it.
10596 Make sure that any inline virtuals are declared inline in the class
10597 body, even if they are not defined there.
10599 @item type_info objects
10602 C++ requires information about types to be written out in order to
10603 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10604 For polymorphic classes (classes with virtual functions), the type_info
10605 object is written out along with the vtable so that @samp{dynamic_cast}
10606 can determine the dynamic type of a class object at runtime. For all
10607 other types, we write out the type_info object when it is used: when
10608 applying @samp{typeid} to an expression, throwing an object, or
10609 referring to a type in a catch clause or exception specification.
10611 @item Template Instantiations
10612 Most everything in this section also applies to template instantiations,
10613 but there are other options as well.
10614 @xref{Template Instantiation,,Where's the Template?}.
10618 When used with GNU ld version 2.8 or later on an ELF system such as
10619 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10620 these constructs will be discarded at link time. This is known as
10623 On targets that don't support COMDAT, but do support weak symbols, GCC
10624 will use them. This way one copy will override all the others, but
10625 the unused copies will still take up space in the executable.
10627 For targets which do not support either COMDAT or weak symbols,
10628 most entities with vague linkage will be emitted as local symbols to
10629 avoid duplicate definition errors from the linker. This will not happen
10630 for local statics in inlines, however, as having multiple copies will
10631 almost certainly break things.
10633 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10634 another way to control placement of these constructs.
10636 @node C++ Interface
10637 @section #pragma interface and implementation
10639 @cindex interface and implementation headers, C++
10640 @cindex C++ interface and implementation headers
10641 @cindex pragmas, interface and implementation
10643 @code{#pragma interface} and @code{#pragma implementation} provide the
10644 user with a way of explicitly directing the compiler to emit entities
10645 with vague linkage (and debugging information) in a particular
10648 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10649 most cases, because of COMDAT support and the ``key method'' heuristic
10650 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10651 program to grow due to unnecessary out-of-line copies of inline
10652 functions. Currently (3.4) the only benefit of these
10653 @code{#pragma}s is reduced duplication of debugging information, and
10654 that should be addressed soon on DWARF 2 targets with the use of
10658 @item #pragma interface
10659 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10660 @kindex #pragma interface
10661 Use this directive in @emph{header files} that define object classes, to save
10662 space in most of the object files that use those classes. Normally,
10663 local copies of certain information (backup copies of inline member
10664 functions, debugging information, and the internal tables that implement
10665 virtual functions) must be kept in each object file that includes class
10666 definitions. You can use this pragma to avoid such duplication. When a
10667 header file containing @samp{#pragma interface} is included in a
10668 compilation, this auxiliary information will not be generated (unless
10669 the main input source file itself uses @samp{#pragma implementation}).
10670 Instead, the object files will contain references to be resolved at link
10673 The second form of this directive is useful for the case where you have
10674 multiple headers with the same name in different directories. If you
10675 use this form, you must specify the same string to @samp{#pragma
10678 @item #pragma implementation
10679 @itemx #pragma implementation "@var{objects}.h"
10680 @kindex #pragma implementation
10681 Use this pragma in a @emph{main input file}, when you want full output from
10682 included header files to be generated (and made globally visible). The
10683 included header file, in turn, should use @samp{#pragma interface}.
10684 Backup copies of inline member functions, debugging information, and the
10685 internal tables used to implement virtual functions are all generated in
10686 implementation files.
10688 @cindex implied @code{#pragma implementation}
10689 @cindex @code{#pragma implementation}, implied
10690 @cindex naming convention, implementation headers
10691 If you use @samp{#pragma implementation} with no argument, it applies to
10692 an include file with the same basename@footnote{A file's @dfn{basename}
10693 was the name stripped of all leading path information and of trailing
10694 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10695 file. For example, in @file{allclass.cc}, giving just
10696 @samp{#pragma implementation}
10697 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10699 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10700 an implementation file whenever you would include it from
10701 @file{allclass.cc} even if you never specified @samp{#pragma
10702 implementation}. This was deemed to be more trouble than it was worth,
10703 however, and disabled.
10705 Use the string argument if you want a single implementation file to
10706 include code from multiple header files. (You must also use
10707 @samp{#include} to include the header file; @samp{#pragma
10708 implementation} only specifies how to use the file---it doesn't actually
10711 There is no way to split up the contents of a single header file into
10712 multiple implementation files.
10715 @cindex inlining and C++ pragmas
10716 @cindex C++ pragmas, effect on inlining
10717 @cindex pragmas in C++, effect on inlining
10718 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10719 effect on function inlining.
10721 If you define a class in a header file marked with @samp{#pragma
10722 interface}, the effect on an inline function defined in that class is
10723 similar to an explicit @code{extern} declaration---the compiler emits
10724 no code at all to define an independent version of the function. Its
10725 definition is used only for inlining with its callers.
10727 @opindex fno-implement-inlines
10728 Conversely, when you include the same header file in a main source file
10729 that declares it as @samp{#pragma implementation}, the compiler emits
10730 code for the function itself; this defines a version of the function
10731 that can be found via pointers (or by callers compiled without
10732 inlining). If all calls to the function can be inlined, you can avoid
10733 emitting the function by compiling with @option{-fno-implement-inlines}.
10734 If any calls were not inlined, you will get linker errors.
10736 @node Template Instantiation
10737 @section Where's the Template?
10738 @cindex template instantiation
10740 C++ templates are the first language feature to require more
10741 intelligence from the environment than one usually finds on a UNIX
10742 system. Somehow the compiler and linker have to make sure that each
10743 template instance occurs exactly once in the executable if it is needed,
10744 and not at all otherwise. There are two basic approaches to this
10745 problem, which are referred to as the Borland model and the Cfront model.
10748 @item Borland model
10749 Borland C++ solved the template instantiation problem by adding the code
10750 equivalent of common blocks to their linker; the compiler emits template
10751 instances in each translation unit that uses them, and the linker
10752 collapses them together. The advantage of this model is that the linker
10753 only has to consider the object files themselves; there is no external
10754 complexity to worry about. This disadvantage is that compilation time
10755 is increased because the template code is being compiled repeatedly.
10756 Code written for this model tends to include definitions of all
10757 templates in the header file, since they must be seen to be
10761 The AT&T C++ translator, Cfront, solved the template instantiation
10762 problem by creating the notion of a template repository, an
10763 automatically maintained place where template instances are stored. A
10764 more modern version of the repository works as follows: As individual
10765 object files are built, the compiler places any template definitions and
10766 instantiations encountered in the repository. At link time, the link
10767 wrapper adds in the objects in the repository and compiles any needed
10768 instances that were not previously emitted. The advantages of this
10769 model are more optimal compilation speed and the ability to use the
10770 system linker; to implement the Borland model a compiler vendor also
10771 needs to replace the linker. The disadvantages are vastly increased
10772 complexity, and thus potential for error; for some code this can be
10773 just as transparent, but in practice it can been very difficult to build
10774 multiple programs in one directory and one program in multiple
10775 directories. Code written for this model tends to separate definitions
10776 of non-inline member templates into a separate file, which should be
10777 compiled separately.
10780 When used with GNU ld version 2.8 or later on an ELF system such as
10781 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10782 Borland model. On other systems, G++ implements neither automatic
10785 A future version of G++ will support a hybrid model whereby the compiler
10786 will emit any instantiations for which the template definition is
10787 included in the compile, and store template definitions and
10788 instantiation context information into the object file for the rest.
10789 The link wrapper will extract that information as necessary and invoke
10790 the compiler to produce the remaining instantiations. The linker will
10791 then combine duplicate instantiations.
10793 In the mean time, you have the following options for dealing with
10794 template instantiations:
10799 Compile your template-using code with @option{-frepo}. The compiler will
10800 generate files with the extension @samp{.rpo} listing all of the
10801 template instantiations used in the corresponding object files which
10802 could be instantiated there; the link wrapper, @samp{collect2}, will
10803 then update the @samp{.rpo} files to tell the compiler where to place
10804 those instantiations and rebuild any affected object files. The
10805 link-time overhead is negligible after the first pass, as the compiler
10806 will continue to place the instantiations in the same files.
10808 This is your best option for application code written for the Borland
10809 model, as it will just work. Code written for the Cfront model will
10810 need to be modified so that the template definitions are available at
10811 one or more points of instantiation; usually this is as simple as adding
10812 @code{#include <tmethods.cc>} to the end of each template header.
10814 For library code, if you want the library to provide all of the template
10815 instantiations it needs, just try to link all of its object files
10816 together; the link will fail, but cause the instantiations to be
10817 generated as a side effect. Be warned, however, that this may cause
10818 conflicts if multiple libraries try to provide the same instantiations.
10819 For greater control, use explicit instantiation as described in the next
10823 @opindex fno-implicit-templates
10824 Compile your code with @option{-fno-implicit-templates} to disable the
10825 implicit generation of template instances, and explicitly instantiate
10826 all the ones you use. This approach requires more knowledge of exactly
10827 which instances you need than do the others, but it's less
10828 mysterious and allows greater control. You can scatter the explicit
10829 instantiations throughout your program, perhaps putting them in the
10830 translation units where the instances are used or the translation units
10831 that define the templates themselves; you can put all of the explicit
10832 instantiations you need into one big file; or you can create small files
10839 template class Foo<int>;
10840 template ostream& operator <<
10841 (ostream&, const Foo<int>&);
10844 for each of the instances you need, and create a template instantiation
10845 library from those.
10847 If you are using Cfront-model code, you can probably get away with not
10848 using @option{-fno-implicit-templates} when compiling files that don't
10849 @samp{#include} the member template definitions.
10851 If you use one big file to do the instantiations, you may want to
10852 compile it without @option{-fno-implicit-templates} so you get all of the
10853 instances required by your explicit instantiations (but not by any
10854 other files) without having to specify them as well.
10856 G++ has extended the template instantiation syntax given in the ISO
10857 standard to allow forward declaration of explicit instantiations
10858 (with @code{extern}), instantiation of the compiler support data for a
10859 template class (i.e.@: the vtable) without instantiating any of its
10860 members (with @code{inline}), and instantiation of only the static data
10861 members of a template class, without the support data or member
10862 functions (with (@code{static}):
10865 extern template int max (int, int);
10866 inline template class Foo<int>;
10867 static template class Foo<int>;
10871 Do nothing. Pretend G++ does implement automatic instantiation
10872 management. Code written for the Borland model will work fine, but
10873 each translation unit will contain instances of each of the templates it
10874 uses. In a large program, this can lead to an unacceptable amount of code
10878 @node Bound member functions
10879 @section Extracting the function pointer from a bound pointer to member function
10881 @cindex pointer to member function
10882 @cindex bound pointer to member function
10884 In C++, pointer to member functions (PMFs) are implemented using a wide
10885 pointer of sorts to handle all the possible call mechanisms; the PMF
10886 needs to store information about how to adjust the @samp{this} pointer,
10887 and if the function pointed to is virtual, where to find the vtable, and
10888 where in the vtable to look for the member function. If you are using
10889 PMFs in an inner loop, you should really reconsider that decision. If
10890 that is not an option, you can extract the pointer to the function that
10891 would be called for a given object/PMF pair and call it directly inside
10892 the inner loop, to save a bit of time.
10894 Note that you will still be paying the penalty for the call through a
10895 function pointer; on most modern architectures, such a call defeats the
10896 branch prediction features of the CPU@. This is also true of normal
10897 virtual function calls.
10899 The syntax for this extension is
10903 extern int (A::*fp)();
10904 typedef int (*fptr)(A *);
10906 fptr p = (fptr)(a.*fp);
10909 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
10910 no object is needed to obtain the address of the function. They can be
10911 converted to function pointers directly:
10914 fptr p1 = (fptr)(&A::foo);
10917 @opindex Wno-pmf-conversions
10918 You must specify @option{-Wno-pmf-conversions} to use this extension.
10920 @node C++ Attributes
10921 @section C++-Specific Variable, Function, and Type Attributes
10923 Some attributes only make sense for C++ programs.
10926 @item init_priority (@var{priority})
10927 @cindex init_priority attribute
10930 In Standard C++, objects defined at namespace scope are guaranteed to be
10931 initialized in an order in strict accordance with that of their definitions
10932 @emph{in a given translation unit}. No guarantee is made for initializations
10933 across translation units. However, GNU C++ allows users to control the
10934 order of initialization of objects defined at namespace scope with the
10935 @code{init_priority} attribute by specifying a relative @var{priority},
10936 a constant integral expression currently bounded between 101 and 65535
10937 inclusive. Lower numbers indicate a higher priority.
10939 In the following example, @code{A} would normally be created before
10940 @code{B}, but the @code{init_priority} attribute has reversed that order:
10943 Some_Class A __attribute__ ((init_priority (2000)));
10944 Some_Class B __attribute__ ((init_priority (543)));
10948 Note that the particular values of @var{priority} do not matter; only their
10951 @item java_interface
10952 @cindex java_interface attribute
10954 This type attribute informs C++ that the class is a Java interface. It may
10955 only be applied to classes declared within an @code{extern "Java"} block.
10956 Calls to methods declared in this interface will be dispatched using GCJ's
10957 interface table mechanism, instead of regular virtual table dispatch.
10961 See also @xref{Namespace Association}.
10963 @node Namespace Association
10964 @section Namespace Association
10966 @strong{Caution:} The semantics of this extension are not fully
10967 defined. Users should refrain from using this extension as its
10968 semantics may change subtly over time. It is possible that this
10969 extension will be removed in future versions of G++.
10971 A using-directive with @code{__attribute ((strong))} is stronger
10972 than a normal using-directive in two ways:
10976 Templates from the used namespace can be specialized and explicitly
10977 instantiated as though they were members of the using namespace.
10980 The using namespace is considered an associated namespace of all
10981 templates in the used namespace for purposes of argument-dependent
10985 The used namespace must be nested within the using namespace so that
10986 normal unqualified lookup works properly.
10988 This is useful for composing a namespace transparently from
10989 implementation namespaces. For example:
10994 template <class T> struct A @{ @};
10996 using namespace debug __attribute ((__strong__));
10997 template <> struct A<int> @{ @}; // @r{ok to specialize}
10999 template <class T> void f (A<T>);
11004 f (std::A<float>()); // @r{lookup finds} std::f
11009 @node Java Exceptions
11010 @section Java Exceptions
11012 The Java language uses a slightly different exception handling model
11013 from C++. Normally, GNU C++ will automatically detect when you are
11014 writing C++ code that uses Java exceptions, and handle them
11015 appropriately. However, if C++ code only needs to execute destructors
11016 when Java exceptions are thrown through it, GCC will guess incorrectly.
11017 Sample problematic code is:
11020 struct S @{ ~S(); @};
11021 extern void bar(); // @r{is written in Java, and may throw exceptions}
11030 The usual effect of an incorrect guess is a link failure, complaining of
11031 a missing routine called @samp{__gxx_personality_v0}.
11033 You can inform the compiler that Java exceptions are to be used in a
11034 translation unit, irrespective of what it might think, by writing
11035 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11036 @samp{#pragma} must appear before any functions that throw or catch
11037 exceptions, or run destructors when exceptions are thrown through them.
11039 You cannot mix Java and C++ exceptions in the same translation unit. It
11040 is believed to be safe to throw a C++ exception from one file through
11041 another file compiled for the Java exception model, or vice versa, but
11042 there may be bugs in this area.
11044 @node Deprecated Features
11045 @section Deprecated Features
11047 In the past, the GNU C++ compiler was extended to experiment with new
11048 features, at a time when the C++ language was still evolving. Now that
11049 the C++ standard is complete, some of those features are superseded by
11050 superior alternatives. Using the old features might cause a warning in
11051 some cases that the feature will be dropped in the future. In other
11052 cases, the feature might be gone already.
11054 While the list below is not exhaustive, it documents some of the options
11055 that are now deprecated:
11058 @item -fexternal-templates
11059 @itemx -falt-external-templates
11060 These are two of the many ways for G++ to implement template
11061 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11062 defines how template definitions have to be organized across
11063 implementation units. G++ has an implicit instantiation mechanism that
11064 should work just fine for standard-conforming code.
11066 @item -fstrict-prototype
11067 @itemx -fno-strict-prototype
11068 Previously it was possible to use an empty prototype parameter list to
11069 indicate an unspecified number of parameters (like C), rather than no
11070 parameters, as C++ demands. This feature has been removed, except where
11071 it is required for backwards compatibility @xref{Backwards Compatibility}.
11074 G++ allows a virtual function returning @samp{void *} to be overridden
11075 by one returning a different pointer type. This extension to the
11076 covariant return type rules is now deprecated and will be removed from a
11079 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11080 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11081 and will be removed in a future version. Code using these operators
11082 should be modified to use @code{std::min} and @code{std::max} instead.
11084 The named return value extension has been deprecated, and is now
11087 The use of initializer lists with new expressions has been deprecated,
11088 and is now removed from G++.
11090 Floating and complex non-type template parameters have been deprecated,
11091 and are now removed from G++.
11093 The implicit typename extension has been deprecated and is now
11096 The use of default arguments in function pointers, function typedefs
11097 and other places where they are not permitted by the standard is
11098 deprecated and will be removed from a future version of G++.
11100 G++ allows floating-point literals to appear in integral constant expressions,
11101 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11102 This extension is deprecated and will be removed from a future version.
11104 G++ allows static data members of const floating-point type to be declared
11105 with an initializer in a class definition. The standard only allows
11106 initializers for static members of const integral types and const
11107 enumeration types so this extension has been deprecated and will be removed
11108 from a future version.
11110 @node Backwards Compatibility
11111 @section Backwards Compatibility
11112 @cindex Backwards Compatibility
11113 @cindex ARM [Annotated C++ Reference Manual]
11115 Now that there is a definitive ISO standard C++, G++ has a specification
11116 to adhere to. The C++ language evolved over time, and features that
11117 used to be acceptable in previous drafts of the standard, such as the ARM
11118 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11119 compilation of C++ written to such drafts, G++ contains some backwards
11120 compatibilities. @emph{All such backwards compatibility features are
11121 liable to disappear in future versions of G++.} They should be considered
11122 deprecated @xref{Deprecated Features}.
11126 If a variable is declared at for scope, it used to remain in scope until
11127 the end of the scope which contained the for statement (rather than just
11128 within the for scope). G++ retains this, but issues a warning, if such a
11129 variable is accessed outside the for scope.
11131 @item Implicit C language
11132 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11133 scope to set the language. On such systems, all header files are
11134 implicitly scoped inside a C language scope. Also, an empty prototype
11135 @code{()} will be treated as an unspecified number of arguments, rather
11136 than no arguments, as C++ demands.