1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C and Objective-C@. Most of them are
20 also available in C++. @xref{C++ Extensions,,Extensions to the
21 C++ Language}, for extensions that apply @emph{only} to C++.
23 Some features that are in ISO C99 but not C89 or C++ are also, as
24 extensions, accepted by GCC in C89 mode and in C++.
27 * Statement Exprs:: Putting statements and declarations inside expressions.
28 * Local Labels:: Labels local to a block.
29 * Labels as Values:: Getting pointers to labels, and computed gotos.
30 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
31 * Constructing Calls:: Dispatching a call to another function.
32 * Typeof:: @code{typeof}: referring to the type of an expression.
33 * Conditionals:: Omitting the middle operand of a @samp{?:} expression.
34 * Long Long:: Double-word integers---@code{long long int}.
35 * Complex:: Data types for complex numbers.
36 * Decimal Float:: Decimal Floating Types.
37 * Hex Floats:: Hexadecimal floating-point constants.
38 * Zero Length:: Zero-length arrays.
39 * Variable Length:: Arrays whose length is computed at run time.
40 * Empty Structures:: Structures with no members.
41 * Variadic Macros:: Macros with a variable number of arguments.
42 * Escaped Newlines:: Slightly looser rules for escaped newlines.
43 * Subscripting:: Any array can be subscripted, even if not an lvalue.
44 * Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers.
45 * Initializers:: Non-constant initializers.
46 * Compound Literals:: Compound literals give structures, unions
48 * Designated Inits:: Labeling elements of initializers.
49 * Cast to Union:: Casting to union type from any member of the union.
50 * Case Ranges:: `case 1 ... 9' and such.
51 * Mixed Declarations:: Mixing declarations and code.
52 * Function Attributes:: Declaring that functions have no side effects,
53 or that they can never return.
54 * Attribute Syntax:: Formal syntax for attributes.
55 * Function Prototypes:: Prototype declarations and old-style definitions.
56 * C++ Comments:: C++ comments are recognized.
57 * Dollar Signs:: Dollar sign is allowed in identifiers.
58 * Character Escapes:: @samp{\e} stands for the character @key{ESC}.
59 * Variable Attributes:: Specifying attributes of variables.
60 * Type Attributes:: Specifying attributes of types.
61 * Alignment:: Inquiring about the alignment of a type or variable.
62 * Inline:: Defining inline functions (as fast as macros).
63 * Extended Asm:: Assembler instructions with C expressions as operands.
64 (With them you can define ``built-in'' functions.)
65 * Constraints:: Constraints for asm operands
66 * Asm Labels:: Specifying the assembler name to use for a C symbol.
67 * Explicit Reg Vars:: Defining variables residing in specified registers.
68 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
69 * Incomplete Enums:: @code{enum foo;}, with details to follow.
70 * Function Names:: Printable strings which are the name of the current
72 * Return Address:: Getting the return or frame address of a function.
73 * Vector Extensions:: Using vector instructions through built-in functions.
74 * Offsetof:: Special syntax for implementing @code{offsetof}.
75 * Atomic Builtins:: Built-in functions for atomic memory access.
76 * Object Size Checking:: Built-in functions for limited buffer overflow
78 * Other Builtins:: Other built-in functions.
79 * Target Builtins:: Built-in functions specific to particular targets.
80 * Target Format Checks:: Format checks specific to particular targets.
81 * Pragmas:: Pragmas accepted by GCC.
82 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
83 * Thread-Local:: Per-thread variables.
87 @section Statements and Declarations in Expressions
88 @cindex statements inside expressions
89 @cindex declarations inside expressions
90 @cindex expressions containing statements
91 @cindex macros, statements in expressions
93 @c the above section title wrapped and causes an underfull hbox.. i
94 @c changed it from "within" to "in". --mew 4feb93
95 A compound statement enclosed in parentheses may appear as an expression
96 in GNU C@. This allows you to use loops, switches, and local variables
99 Recall that a compound statement is a sequence of statements surrounded
100 by braces; in this construct, parentheses go around the braces. For
104 (@{ int y = foo (); int z;
111 is a valid (though slightly more complex than necessary) expression
112 for the absolute value of @code{foo ()}.
114 The last thing in the compound statement should be an expression
115 followed by a semicolon; the value of this subexpression serves as the
116 value of the entire construct. (If you use some other kind of statement
117 last within the braces, the construct has type @code{void}, and thus
118 effectively no value.)
120 This feature is especially useful in making macro definitions ``safe'' (so
121 that they evaluate each operand exactly once). For example, the
122 ``maximum'' function is commonly defined as a macro in standard C as
126 #define max(a,b) ((a) > (b) ? (a) : (b))
130 @cindex side effects, macro argument
131 But this definition computes either @var{a} or @var{b} twice, with bad
132 results if the operand has side effects. In GNU C, if you know the
133 type of the operands (here taken as @code{int}), you can define
134 the macro safely as follows:
137 #define maxint(a,b) \
138 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
141 Embedded statements are not allowed in constant expressions, such as
142 the value of an enumeration constant, the width of a bit-field, or
143 the initial value of a static variable.
145 If you don't know the type of the operand, you can still do this, but you
146 must use @code{typeof} (@pxref{Typeof}).
148 In G++, the result value of a statement expression undergoes array and
149 function pointer decay, and is returned by value to the enclosing
150 expression. For instance, if @code{A} is a class, then
159 will construct a temporary @code{A} object to hold the result of the
160 statement expression, and that will be used to invoke @code{Foo}.
161 Therefore the @code{this} pointer observed by @code{Foo} will not be the
164 Any temporaries created within a statement within a statement expression
165 will be destroyed at the statement's end. This makes statement
166 expressions inside macros slightly different from function calls. In
167 the latter case temporaries introduced during argument evaluation will
168 be destroyed at the end of the statement that includes the function
169 call. In the statement expression case they will be destroyed during
170 the statement expression. For instance,
173 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
174 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
184 will have different places where temporaries are destroyed. For the
185 @code{macro} case, the temporary @code{X} will be destroyed just after
186 the initialization of @code{b}. In the @code{function} case that
187 temporary will be destroyed when the function returns.
189 These considerations mean that it is probably a bad idea to use
190 statement-expressions of this form in header files that are designed to
191 work with C++. (Note that some versions of the GNU C Library contained
192 header files using statement-expression that lead to precisely this
195 Jumping into a statement expression with @code{goto} or using a
196 @code{switch} statement outside the statement expression with a
197 @code{case} or @code{default} label inside the statement expression is
198 not permitted. Jumping into a statement expression with a computed
199 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
200 Jumping out of a statement expression is permitted, but if the
201 statement expression is part of a larger expression then it is
202 unspecified which other subexpressions of that expression have been
203 evaluated except where the language definition requires certain
204 subexpressions to be evaluated before or after the statement
205 expression. In any case, as with a function call the evaluation of a
206 statement expression is not interleaved with the evaluation of other
207 parts of the containing expression. For example,
210 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
214 will call @code{foo} and @code{bar1} and will not call @code{baz} but
215 may or may not call @code{bar2}. If @code{bar2} is called, it will be
216 called after @code{foo} and before @code{bar1}
219 @section Locally Declared Labels
221 @cindex macros, local labels
223 GCC allows you to declare @dfn{local labels} in any nested block
224 scope. A local label is just like an ordinary label, but you can
225 only reference it (with a @code{goto} statement, or by taking its
226 address) within the block in which it was declared.
228 A local label declaration looks like this:
231 __label__ @var{label};
238 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
241 Local label declarations must come at the beginning of the block,
242 before any ordinary declarations or statements.
244 The label declaration defines the label @emph{name}, but does not define
245 the label itself. You must do this in the usual way, with
246 @code{@var{label}:}, within the statements of the statement expression.
248 The local label feature is useful for complex macros. If a macro
249 contains nested loops, a @code{goto} can be useful for breaking out of
250 them. However, an ordinary label whose scope is the whole function
251 cannot be used: if the macro can be expanded several times in one
252 function, the label will be multiply defined in that function. A
253 local label avoids this problem. For example:
256 #define SEARCH(value, array, target) \
259 typeof (target) _SEARCH_target = (target); \
260 typeof (*(array)) *_SEARCH_array = (array); \
263 for (i = 0; i < max; i++) \
264 for (j = 0; j < max; j++) \
265 if (_SEARCH_array[i][j] == _SEARCH_target) \
266 @{ (value) = i; goto found; @} \
272 This could also be written using a statement-expression:
275 #define SEARCH(array, target) \
278 typeof (target) _SEARCH_target = (target); \
279 typeof (*(array)) *_SEARCH_array = (array); \
282 for (i = 0; i < max; i++) \
283 for (j = 0; j < max; j++) \
284 if (_SEARCH_array[i][j] == _SEARCH_target) \
285 @{ value = i; goto found; @} \
292 Local label declarations also make the labels they declare visible to
293 nested functions, if there are any. @xref{Nested Functions}, for details.
295 @node Labels as Values
296 @section Labels as Values
297 @cindex labels as values
298 @cindex computed gotos
299 @cindex goto with computed label
300 @cindex address of a label
302 You can get the address of a label defined in the current function
303 (or a containing function) with the unary operator @samp{&&}. The
304 value has type @code{void *}. This value is a constant and can be used
305 wherever a constant of that type is valid. For example:
313 To use these values, you need to be able to jump to one. This is done
314 with the computed goto statement@footnote{The analogous feature in
315 Fortran is called an assigned goto, but that name seems inappropriate in
316 C, where one can do more than simply store label addresses in label
317 variables.}, @code{goto *@var{exp};}. For example,
324 Any expression of type @code{void *} is allowed.
326 One way of using these constants is in initializing a static array that
327 will serve as a jump table:
330 static void *array[] = @{ &&foo, &&bar, &&hack @};
333 Then you can select a label with indexing, like this:
340 Note that this does not check whether the subscript is in bounds---array
341 indexing in C never does that.
343 Such an array of label values serves a purpose much like that of the
344 @code{switch} statement. The @code{switch} statement is cleaner, so
345 use that rather than an array unless the problem does not fit a
346 @code{switch} statement very well.
348 Another use of label values is in an interpreter for threaded code.
349 The labels within the interpreter function can be stored in the
350 threaded code for super-fast dispatching.
352 You may not use this mechanism to jump to code in a different function.
353 If you do that, totally unpredictable things will happen. The best way to
354 avoid this is to store the label address only in automatic variables and
355 never pass it as an argument.
357 An alternate way to write the above example is
360 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
362 goto *(&&foo + array[i]);
366 This is more friendly to code living in shared libraries, as it reduces
367 the number of dynamic relocations that are needed, and by consequence,
368 allows the data to be read-only.
370 @node Nested Functions
371 @section Nested Functions
372 @cindex nested functions
373 @cindex downward funargs
376 A @dfn{nested function} is a function defined inside another function.
377 (Nested functions are not supported for GNU C++.) The nested function's
378 name is local to the block where it is defined. For example, here we
379 define a nested function named @code{square}, and call it twice:
383 foo (double a, double b)
385 double square (double z) @{ return z * z; @}
387 return square (a) + square (b);
392 The nested function can access all the variables of the containing
393 function that are visible at the point of its definition. This is
394 called @dfn{lexical scoping}. For example, here we show a nested
395 function which uses an inherited variable named @code{offset}:
399 bar (int *array, int offset, int size)
401 int access (int *array, int index)
402 @{ return array[index + offset]; @}
405 for (i = 0; i < size; i++)
406 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
411 Nested function definitions are permitted within functions in the places
412 where variable definitions are allowed; that is, in any block, mixed
413 with the other declarations and statements in the block.
415 It is possible to call the nested function from outside the scope of its
416 name by storing its address or passing the address to another function:
419 hack (int *array, int size)
421 void store (int index, int value)
422 @{ array[index] = value; @}
424 intermediate (store, size);
428 Here, the function @code{intermediate} receives the address of
429 @code{store} as an argument. If @code{intermediate} calls @code{store},
430 the arguments given to @code{store} are used to store into @code{array}.
431 But this technique works only so long as the containing function
432 (@code{hack}, in this example) does not exit.
434 If you try to call the nested function through its address after the
435 containing function has exited, all hell will break loose. If you try
436 to call it after a containing scope level has exited, and if it refers
437 to some of the variables that are no longer in scope, you may be lucky,
438 but it's not wise to take the risk. If, however, the nested function
439 does not refer to anything that has gone out of scope, you should be
442 GCC implements taking the address of a nested function using a technique
443 called @dfn{trampolines}. A paper describing them is available as
446 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
448 A nested function can jump to a label inherited from a containing
449 function, provided the label was explicitly declared in the containing
450 function (@pxref{Local Labels}). Such a jump returns instantly to the
451 containing function, exiting the nested function which did the
452 @code{goto} and any intermediate functions as well. Here is an example:
456 bar (int *array, int offset, int size)
459 int access (int *array, int index)
463 return array[index + offset];
467 for (i = 0; i < size; i++)
468 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
472 /* @r{Control comes here from @code{access}
473 if it detects an error.} */
480 A nested function always has no linkage. Declaring one with
481 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
482 before its definition, use @code{auto} (which is otherwise meaningless
483 for function declarations).
486 bar (int *array, int offset, int size)
489 auto int access (int *, int);
491 int access (int *array, int index)
495 return array[index + offset];
501 @node Constructing Calls
502 @section Constructing Function Calls
503 @cindex constructing calls
504 @cindex forwarding calls
506 Using the built-in functions described below, you can record
507 the arguments a function received, and call another function
508 with the same arguments, without knowing the number or types
511 You can also record the return value of that function call,
512 and later return that value, without knowing what data type
513 the function tried to return (as long as your caller expects
516 However, these built-in functions may interact badly with some
517 sophisticated features or other extensions of the language. It
518 is, therefore, not recommended to use them outside very simple
519 functions acting as mere forwarders for their arguments.
521 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
522 This built-in function returns a pointer to data
523 describing how to perform a call with the same arguments as were passed
524 to the current function.
526 The function saves the arg pointer register, structure value address,
527 and all registers that might be used to pass arguments to a function
528 into a block of memory allocated on the stack. Then it returns the
529 address of that block.
532 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
533 This built-in function invokes @var{function}
534 with a copy of the parameters described by @var{arguments}
537 The value of @var{arguments} should be the value returned by
538 @code{__builtin_apply_args}. The argument @var{size} specifies the size
539 of the stack argument data, in bytes.
541 This function returns a pointer to data describing
542 how to return whatever value was returned by @var{function}. The data
543 is saved in a block of memory allocated on the stack.
545 It is not always simple to compute the proper value for @var{size}. The
546 value is used by @code{__builtin_apply} to compute the amount of data
547 that should be pushed on the stack and copied from the incoming argument
551 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
552 This built-in function returns the value described by @var{result} from
553 the containing function. You should specify, for @var{result}, a value
554 returned by @code{__builtin_apply}.
558 @section Referring to a Type with @code{typeof}
561 @cindex macros, types of arguments
563 Another way to refer to the type of an expression is with @code{typeof}.
564 The syntax of using of this keyword looks like @code{sizeof}, but the
565 construct acts semantically like a type name defined with @code{typedef}.
567 There are two ways of writing the argument to @code{typeof}: with an
568 expression or with a type. Here is an example with an expression:
575 This assumes that @code{x} is an array of pointers to functions;
576 the type described is that of the values of the functions.
578 Here is an example with a typename as the argument:
585 Here the type described is that of pointers to @code{int}.
587 If you are writing a header file that must work when included in ISO C
588 programs, write @code{__typeof__} instead of @code{typeof}.
589 @xref{Alternate Keywords}.
591 A @code{typeof}-construct can be used anywhere a typedef name could be
592 used. For example, you can use it in a declaration, in a cast, or inside
593 of @code{sizeof} or @code{typeof}.
595 @code{typeof} is often useful in conjunction with the
596 statements-within-expressions feature. Here is how the two together can
597 be used to define a safe ``maximum'' macro that operates on any
598 arithmetic type and evaluates each of its arguments exactly once:
602 (@{ typeof (a) _a = (a); \
603 typeof (b) _b = (b); \
604 _a > _b ? _a : _b; @})
607 @cindex underscores in variables in macros
608 @cindex @samp{_} in variables in macros
609 @cindex local variables in macros
610 @cindex variables, local, in macros
611 @cindex macros, local variables in
613 The reason for using names that start with underscores for the local
614 variables is to avoid conflicts with variable names that occur within the
615 expressions that are substituted for @code{a} and @code{b}. Eventually we
616 hope to design a new form of declaration syntax that allows you to declare
617 variables whose scopes start only after their initializers; this will be a
618 more reliable way to prevent such conflicts.
621 Some more examples of the use of @code{typeof}:
625 This declares @code{y} with the type of what @code{x} points to.
632 This declares @code{y} as an array of such values.
639 This declares @code{y} as an array of pointers to characters:
642 typeof (typeof (char *)[4]) y;
646 It is equivalent to the following traditional C declaration:
652 To see the meaning of the declaration using @code{typeof}, and why it
653 might be a useful way to write, rewrite it with these macros:
656 #define pointer(T) typeof(T *)
657 #define array(T, N) typeof(T [N])
661 Now the declaration can be rewritten this way:
664 array (pointer (char), 4) y;
668 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
669 pointers to @code{char}.
672 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
673 a more limited extension which permitted one to write
676 typedef @var{T} = @var{expr};
680 with the effect of declaring @var{T} to have the type of the expression
681 @var{expr}. This extension does not work with GCC 3 (versions between
682 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
683 relies on it should be rewritten to use @code{typeof}:
686 typedef typeof(@var{expr}) @var{T};
690 This will work with all versions of GCC@.
693 @section Conditionals with Omitted Operands
694 @cindex conditional expressions, extensions
695 @cindex omitted middle-operands
696 @cindex middle-operands, omitted
697 @cindex extensions, @code{?:}
698 @cindex @code{?:} extensions
700 The middle operand in a conditional expression may be omitted. Then
701 if the first operand is nonzero, its value is the value of the conditional
704 Therefore, the expression
711 has the value of @code{x} if that is nonzero; otherwise, the value of
714 This example is perfectly equivalent to
720 @cindex side effect in ?:
721 @cindex ?: side effect
723 In this simple case, the ability to omit the middle operand is not
724 especially useful. When it becomes useful is when the first operand does,
725 or may (if it is a macro argument), contain a side effect. Then repeating
726 the operand in the middle would perform the side effect twice. Omitting
727 the middle operand uses the value already computed without the undesirable
728 effects of recomputing it.
731 @section Double-Word Integers
732 @cindex @code{long long} data types
733 @cindex double-word arithmetic
734 @cindex multiprecision arithmetic
735 @cindex @code{LL} integer suffix
736 @cindex @code{ULL} integer suffix
738 ISO C99 supports data types for integers that are at least 64 bits wide,
739 and as an extension GCC supports them in C89 mode and in C++.
740 Simply write @code{long long int} for a signed integer, or
741 @code{unsigned long long int} for an unsigned integer. To make an
742 integer constant of type @code{long long int}, add the suffix @samp{LL}
743 to the integer. To make an integer constant of type @code{unsigned long
744 long int}, add the suffix @samp{ULL} to the integer.
746 You can use these types in arithmetic like any other integer types.
747 Addition, subtraction, and bitwise boolean operations on these types
748 are open-coded on all types of machines. Multiplication is open-coded
749 if the machine supports fullword-to-doubleword a widening multiply
750 instruction. Division and shifts are open-coded only on machines that
751 provide special support. The operations that are not open-coded use
752 special library routines that come with GCC@.
754 There may be pitfalls when you use @code{long long} types for function
755 arguments, unless you declare function prototypes. If a function
756 expects type @code{int} for its argument, and you pass a value of type
757 @code{long long int}, confusion will result because the caller and the
758 subroutine will disagree about the number of bytes for the argument.
759 Likewise, if the function expects @code{long long int} and you pass
760 @code{int}. The best way to avoid such problems is to use prototypes.
763 @section Complex Numbers
764 @cindex complex numbers
765 @cindex @code{_Complex} keyword
766 @cindex @code{__complex__} keyword
768 ISO C99 supports complex floating data types, and as an extension GCC
769 supports them in C89 mode and in C++, and supports complex integer data
770 types which are not part of ISO C99. You can declare complex types
771 using the keyword @code{_Complex}. As an extension, the older GNU
772 keyword @code{__complex__} is also supported.
774 For example, @samp{_Complex double x;} declares @code{x} as a
775 variable whose real part and imaginary part are both of type
776 @code{double}. @samp{_Complex short int y;} declares @code{y} to
777 have real and imaginary parts of type @code{short int}; this is not
778 likely to be useful, but it shows that the set of complex types is
781 To write a constant with a complex data type, use the suffix @samp{i} or
782 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
783 has type @code{_Complex float} and @code{3i} has type
784 @code{_Complex int}. Such a constant always has a pure imaginary
785 value, but you can form any complex value you like by adding one to a
786 real constant. This is a GNU extension; if you have an ISO C99
787 conforming C library (such as GNU libc), and want to construct complex
788 constants of floating type, you should include @code{<complex.h>} and
789 use the macros @code{I} or @code{_Complex_I} instead.
791 @cindex @code{__real__} keyword
792 @cindex @code{__imag__} keyword
793 To extract the real part of a complex-valued expression @var{exp}, write
794 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
795 extract the imaginary part. This is a GNU extension; for values of
796 floating type, you should use the ISO C99 functions @code{crealf},
797 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
798 @code{cimagl}, declared in @code{<complex.h>} and also provided as
799 built-in functions by GCC@.
801 @cindex complex conjugation
802 The operator @samp{~} performs complex conjugation when used on a value
803 with a complex type. This is a GNU extension; for values of
804 floating type, you should use the ISO C99 functions @code{conjf},
805 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
806 provided as built-in functions by GCC@.
808 GCC can allocate complex automatic variables in a noncontiguous
809 fashion; it's even possible for the real part to be in a register while
810 the imaginary part is on the stack (or vice-versa). Only the DWARF2
811 debug info format can represent this, so use of DWARF2 is recommended.
812 If you are using the stabs debug info format, GCC describes a noncontiguous
813 complex variable as if it were two separate variables of noncomplex type.
814 If the variable's actual name is @code{foo}, the two fictitious
815 variables are named @code{foo$real} and @code{foo$imag}. You can
816 examine and set these two fictitious variables with your debugger.
819 @section Decimal Floating Types
820 @cindex decimal floating types
821 @cindex @code{_Decimal32} data type
822 @cindex @code{_Decimal64} data type
823 @cindex @code{_Decimal128} data type
824 @cindex @code{df} integer suffix
825 @cindex @code{dd} integer suffix
826 @cindex @code{dl} integer suffix
827 @cindex @code{DF} integer suffix
828 @cindex @code{DD} integer suffix
829 @cindex @code{DL} integer suffix
831 As an extension, the GNU C compiler supports decimal floating types as
832 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
833 floating types in GCC will evolve as the draft technical report changes.
834 Calling conventions for any target might also change. Not all targets
835 support decimal floating types.
837 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
838 @code{_Decimal128}. They use a radix of ten, unlike the floating types
839 @code{float}, @code{double}, and @code{long double} whose radix is not
840 specified by the C standard but is usually two.
842 Support for decimal floating types includes the arithmetic operators
843 add, subtract, multiply, divide; unary arithmetic operators;
844 relational operators; equality operators; and conversions to and from
845 integer and other floating types. Use a suffix @samp{df} or
846 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
847 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
850 GCC support of decimal float as specified by the draft technical report
855 Translation time data type (TTDT) is not supported.
858 Characteristics of decimal floating types are defined in header file
859 @file{decfloat.h} rather than @file{float.h}.
862 When the value of a decimal floating type cannot be represented in the
863 integer type to which it is being converted, the result is undefined
864 rather than the result value specified by the draft technical report.
867 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
868 are supported by the DWARF2 debug information format.
874 ISO C99 supports floating-point numbers written not only in the usual
875 decimal notation, such as @code{1.55e1}, but also numbers such as
876 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
877 supports this in C89 mode (except in some cases when strictly
878 conforming) and in C++. In that format the
879 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
880 mandatory. The exponent is a decimal number that indicates the power of
881 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
888 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
889 is the same as @code{1.55e1}.
891 Unlike for floating-point numbers in the decimal notation the exponent
892 is always required in the hexadecimal notation. Otherwise the compiler
893 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
894 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
895 extension for floating-point constants of type @code{float}.
898 @section Arrays of Length Zero
899 @cindex arrays of length zero
900 @cindex zero-length arrays
901 @cindex length-zero arrays
902 @cindex flexible array members
904 Zero-length arrays are allowed in GNU C@. They are very useful as the
905 last element of a structure which is really a header for a variable-length
914 struct line *thisline = (struct line *)
915 malloc (sizeof (struct line) + this_length);
916 thisline->length = this_length;
919 In ISO C90, you would have to give @code{contents} a length of 1, which
920 means either you waste space or complicate the argument to @code{malloc}.
922 In ISO C99, you would use a @dfn{flexible array member}, which is
923 slightly different in syntax and semantics:
927 Flexible array members are written as @code{contents[]} without
931 Flexible array members have incomplete type, and so the @code{sizeof}
932 operator may not be applied. As a quirk of the original implementation
933 of zero-length arrays, @code{sizeof} evaluates to zero.
936 Flexible array members may only appear as the last member of a
937 @code{struct} that is otherwise non-empty.
940 A structure containing a flexible array member, or a union containing
941 such a structure (possibly recursively), may not be a member of a
942 structure or an element of an array. (However, these uses are
943 permitted by GCC as extensions.)
946 GCC versions before 3.0 allowed zero-length arrays to be statically
947 initialized, as if they were flexible arrays. In addition to those
948 cases that were useful, it also allowed initializations in situations
949 that would corrupt later data. Non-empty initialization of zero-length
950 arrays is now treated like any case where there are more initializer
951 elements than the array holds, in that a suitable warning about "excess
952 elements in array" is given, and the excess elements (all of them, in
953 this case) are ignored.
955 Instead GCC allows static initialization of flexible array members.
956 This is equivalent to defining a new structure containing the original
957 structure followed by an array of sufficient size to contain the data.
958 I.e.@: in the following, @code{f1} is constructed as if it were declared
964 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
967 struct f1 f1; int data[3];
968 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
972 The convenience of this extension is that @code{f1} has the desired
973 type, eliminating the need to consistently refer to @code{f2.f1}.
975 This has symmetry with normal static arrays, in that an array of
976 unknown size is also written with @code{[]}.
978 Of course, this extension only makes sense if the extra data comes at
979 the end of a top-level object, as otherwise we would be overwriting
980 data at subsequent offsets. To avoid undue complication and confusion
981 with initialization of deeply nested arrays, we simply disallow any
982 non-empty initialization except when the structure is the top-level
986 struct foo @{ int x; int y[]; @};
987 struct bar @{ struct foo z; @};
989 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
990 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
991 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
992 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
995 @node Empty Structures
996 @section Structures With No Members
997 @cindex empty structures
998 @cindex zero-size structures
1000 GCC permits a C structure to have no members:
1007 The structure will have size zero. In C++, empty structures are part
1008 of the language. G++ treats empty structures as if they had a single
1009 member of type @code{char}.
1011 @node Variable Length
1012 @section Arrays of Variable Length
1013 @cindex variable-length arrays
1014 @cindex arrays of variable length
1017 Variable-length automatic arrays are allowed in ISO C99, and as an
1018 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1019 implementation of variable-length arrays does not yet conform in detail
1020 to the ISO C99 standard.) These arrays are
1021 declared like any other automatic arrays, but with a length that is not
1022 a constant expression. The storage is allocated at the point of
1023 declaration and deallocated when the brace-level is exited. For
1028 concat_fopen (char *s1, char *s2, char *mode)
1030 char str[strlen (s1) + strlen (s2) + 1];
1033 return fopen (str, mode);
1037 @cindex scope of a variable length array
1038 @cindex variable-length array scope
1039 @cindex deallocating variable length arrays
1040 Jumping or breaking out of the scope of the array name deallocates the
1041 storage. Jumping into the scope is not allowed; you get an error
1044 @cindex @code{alloca} vs variable-length arrays
1045 You can use the function @code{alloca} to get an effect much like
1046 variable-length arrays. The function @code{alloca} is available in
1047 many other C implementations (but not in all). On the other hand,
1048 variable-length arrays are more elegant.
1050 There are other differences between these two methods. Space allocated
1051 with @code{alloca} exists until the containing @emph{function} returns.
1052 The space for a variable-length array is deallocated as soon as the array
1053 name's scope ends. (If you use both variable-length arrays and
1054 @code{alloca} in the same function, deallocation of a variable-length array
1055 will also deallocate anything more recently allocated with @code{alloca}.)
1057 You can also use variable-length arrays as arguments to functions:
1061 tester (int len, char data[len][len])
1067 The length of an array is computed once when the storage is allocated
1068 and is remembered for the scope of the array in case you access it with
1071 If you want to pass the array first and the length afterward, you can
1072 use a forward declaration in the parameter list---another GNU extension.
1076 tester (int len; char data[len][len], int len)
1082 @cindex parameter forward declaration
1083 The @samp{int len} before the semicolon is a @dfn{parameter forward
1084 declaration}, and it serves the purpose of making the name @code{len}
1085 known when the declaration of @code{data} is parsed.
1087 You can write any number of such parameter forward declarations in the
1088 parameter list. They can be separated by commas or semicolons, but the
1089 last one must end with a semicolon, which is followed by the ``real''
1090 parameter declarations. Each forward declaration must match a ``real''
1091 declaration in parameter name and data type. ISO C99 does not support
1092 parameter forward declarations.
1094 @node Variadic Macros
1095 @section Macros with a Variable Number of Arguments.
1096 @cindex variable number of arguments
1097 @cindex macro with variable arguments
1098 @cindex rest argument (in macro)
1099 @cindex variadic macros
1101 In the ISO C standard of 1999, a macro can be declared to accept a
1102 variable number of arguments much as a function can. The syntax for
1103 defining the macro is similar to that of a function. Here is an
1107 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1110 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1111 such a macro, it represents the zero or more tokens until the closing
1112 parenthesis that ends the invocation, including any commas. This set of
1113 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1114 wherever it appears. See the CPP manual for more information.
1116 GCC has long supported variadic macros, and used a different syntax that
1117 allowed you to give a name to the variable arguments just like any other
1118 argument. Here is an example:
1121 #define debug(format, args...) fprintf (stderr, format, args)
1124 This is in all ways equivalent to the ISO C example above, but arguably
1125 more readable and descriptive.
1127 GNU CPP has two further variadic macro extensions, and permits them to
1128 be used with either of the above forms of macro definition.
1130 In standard C, you are not allowed to leave the variable argument out
1131 entirely; but you are allowed to pass an empty argument. For example,
1132 this invocation is invalid in ISO C, because there is no comma after
1139 GNU CPP permits you to completely omit the variable arguments in this
1140 way. In the above examples, the compiler would complain, though since
1141 the expansion of the macro still has the extra comma after the format
1144 To help solve this problem, CPP behaves specially for variable arguments
1145 used with the token paste operator, @samp{##}. If instead you write
1148 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1151 and if the variable arguments are omitted or empty, the @samp{##}
1152 operator causes the preprocessor to remove the comma before it. If you
1153 do provide some variable arguments in your macro invocation, GNU CPP
1154 does not complain about the paste operation and instead places the
1155 variable arguments after the comma. Just like any other pasted macro
1156 argument, these arguments are not macro expanded.
1158 @node Escaped Newlines
1159 @section Slightly Looser Rules for Escaped Newlines
1160 @cindex escaped newlines
1161 @cindex newlines (escaped)
1163 Recently, the preprocessor has relaxed its treatment of escaped
1164 newlines. Previously, the newline had to immediately follow a
1165 backslash. The current implementation allows whitespace in the form
1166 of spaces, horizontal and vertical tabs, and form feeds between the
1167 backslash and the subsequent newline. The preprocessor issues a
1168 warning, but treats it as a valid escaped newline and combines the two
1169 lines to form a single logical line. This works within comments and
1170 tokens, as well as between tokens. Comments are @emph{not} treated as
1171 whitespace for the purposes of this relaxation, since they have not
1172 yet been replaced with spaces.
1175 @section Non-Lvalue Arrays May Have Subscripts
1176 @cindex subscripting
1177 @cindex arrays, non-lvalue
1179 @cindex subscripting and function values
1180 In ISO C99, arrays that are not lvalues still decay to pointers, and
1181 may be subscripted, although they may not be modified or used after
1182 the next sequence point and the unary @samp{&} operator may not be
1183 applied to them. As an extension, GCC allows such arrays to be
1184 subscripted in C89 mode, though otherwise they do not decay to
1185 pointers outside C99 mode. For example,
1186 this is valid in GNU C though not valid in C89:
1190 struct foo @{int a[4];@};
1196 return f().a[index];
1202 @section Arithmetic on @code{void}- and Function-Pointers
1203 @cindex void pointers, arithmetic
1204 @cindex void, size of pointer to
1205 @cindex function pointers, arithmetic
1206 @cindex function, size of pointer to
1208 In GNU C, addition and subtraction operations are supported on pointers to
1209 @code{void} and on pointers to functions. This is done by treating the
1210 size of a @code{void} or of a function as 1.
1212 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1213 and on function types, and returns 1.
1215 @opindex Wpointer-arith
1216 The option @option{-Wpointer-arith} requests a warning if these extensions
1220 @section Non-Constant Initializers
1221 @cindex initializers, non-constant
1222 @cindex non-constant initializers
1224 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1225 automatic variable are not required to be constant expressions in GNU C@.
1226 Here is an example of an initializer with run-time varying elements:
1229 foo (float f, float g)
1231 float beat_freqs[2] = @{ f-g, f+g @};
1236 @node Compound Literals
1237 @section Compound Literals
1238 @cindex constructor expressions
1239 @cindex initializations in expressions
1240 @cindex structures, constructor expression
1241 @cindex expressions, constructor
1242 @cindex compound literals
1243 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1245 ISO C99 supports compound literals. A compound literal looks like
1246 a cast containing an initializer. Its value is an object of the
1247 type specified in the cast, containing the elements specified in
1248 the initializer; it is an lvalue. As an extension, GCC supports
1249 compound literals in C89 mode and in C++.
1251 Usually, the specified type is a structure. Assume that
1252 @code{struct foo} and @code{structure} are declared as shown:
1255 struct foo @{int a; char b[2];@} structure;
1259 Here is an example of constructing a @code{struct foo} with a compound literal:
1262 structure = ((struct foo) @{x + y, 'a', 0@});
1266 This is equivalent to writing the following:
1270 struct foo temp = @{x + y, 'a', 0@};
1275 You can also construct an array. If all the elements of the compound literal
1276 are (made up of) simple constant expressions, suitable for use in
1277 initializers of objects of static storage duration, then the compound
1278 literal can be coerced to a pointer to its first element and used in
1279 such an initializer, as shown here:
1282 char **foo = (char *[]) @{ "x", "y", "z" @};
1285 Compound literals for scalar types and union types are is
1286 also allowed, but then the compound literal is equivalent
1289 As a GNU extension, GCC allows initialization of objects with static storage
1290 duration by compound literals (which is not possible in ISO C99, because
1291 the initializer is not a constant).
1292 It is handled as if the object was initialized only with the bracket
1293 enclosed list if compound literal's and object types match.
1294 The initializer list of the compound literal must be constant.
1295 If the object being initialized has array type of unknown size, the size is
1296 determined by compound literal size.
1299 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1300 static int y[] = (int []) @{1, 2, 3@};
1301 static int z[] = (int [3]) @{1@};
1305 The above lines are equivalent to the following:
1307 static struct foo x = @{1, 'a', 'b'@};
1308 static int y[] = @{1, 2, 3@};
1309 static int z[] = @{1, 0, 0@};
1312 @node Designated Inits
1313 @section Designated Initializers
1314 @cindex initializers with labeled elements
1315 @cindex labeled elements in initializers
1316 @cindex case labels in initializers
1317 @cindex designated initializers
1319 Standard C89 requires the elements of an initializer to appear in a fixed
1320 order, the same as the order of the elements in the array or structure
1323 In ISO C99 you can give the elements in any order, specifying the array
1324 indices or structure field names they apply to, and GNU C allows this as
1325 an extension in C89 mode as well. This extension is not
1326 implemented in GNU C++.
1328 To specify an array index, write
1329 @samp{[@var{index}] =} before the element value. For example,
1332 int a[6] = @{ [4] = 29, [2] = 15 @};
1339 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1343 The index values must be constant expressions, even if the array being
1344 initialized is automatic.
1346 An alternative syntax for this which has been obsolete since GCC 2.5 but
1347 GCC still accepts is to write @samp{[@var{index}]} before the element
1348 value, with no @samp{=}.
1350 To initialize a range of elements to the same value, write
1351 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1352 extension. For example,
1355 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1359 If the value in it has side-effects, the side-effects will happen only once,
1360 not for each initialized field by the range initializer.
1363 Note that the length of the array is the highest value specified
1366 In a structure initializer, specify the name of a field to initialize
1367 with @samp{.@var{fieldname} =} before the element value. For example,
1368 given the following structure,
1371 struct point @{ int x, y; @};
1375 the following initialization
1378 struct point p = @{ .y = yvalue, .x = xvalue @};
1385 struct point p = @{ xvalue, yvalue @};
1388 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1389 @samp{@var{fieldname}:}, as shown here:
1392 struct point p = @{ y: yvalue, x: xvalue @};
1396 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1397 @dfn{designator}. You can also use a designator (or the obsolete colon
1398 syntax) when initializing a union, to specify which element of the union
1399 should be used. For example,
1402 union foo @{ int i; double d; @};
1404 union foo f = @{ .d = 4 @};
1408 will convert 4 to a @code{double} to store it in the union using
1409 the second element. By contrast, casting 4 to type @code{union foo}
1410 would store it into the union as the integer @code{i}, since it is
1411 an integer. (@xref{Cast to Union}.)
1413 You can combine this technique of naming elements with ordinary C
1414 initialization of successive elements. Each initializer element that
1415 does not have a designator applies to the next consecutive element of the
1416 array or structure. For example,
1419 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1426 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1429 Labeling the elements of an array initializer is especially useful
1430 when the indices are characters or belong to an @code{enum} type.
1435 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1436 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1439 @cindex designator lists
1440 You can also write a series of @samp{.@var{fieldname}} and
1441 @samp{[@var{index}]} designators before an @samp{=} to specify a
1442 nested subobject to initialize; the list is taken relative to the
1443 subobject corresponding to the closest surrounding brace pair. For
1444 example, with the @samp{struct point} declaration above:
1447 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1451 If the same field is initialized multiple times, it will have value from
1452 the last initialization. If any such overridden initialization has
1453 side-effect, it is unspecified whether the side-effect happens or not.
1454 Currently, GCC will discard them and issue a warning.
1457 @section Case Ranges
1459 @cindex ranges in case statements
1461 You can specify a range of consecutive values in a single @code{case} label,
1465 case @var{low} ... @var{high}:
1469 This has the same effect as the proper number of individual @code{case}
1470 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1472 This feature is especially useful for ranges of ASCII character codes:
1478 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1479 it may be parsed wrong when you use it with integer values. For example,
1494 @section Cast to a Union Type
1495 @cindex cast to a union
1496 @cindex union, casting to a
1498 A cast to union type is similar to other casts, except that the type
1499 specified is a union type. You can specify the type either with
1500 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1501 a constructor though, not a cast, and hence does not yield an lvalue like
1502 normal casts. (@xref{Compound Literals}.)
1504 The types that may be cast to the union type are those of the members
1505 of the union. Thus, given the following union and variables:
1508 union foo @{ int i; double d; @};
1514 both @code{x} and @code{y} can be cast to type @code{union foo}.
1516 Using the cast as the right-hand side of an assignment to a variable of
1517 union type is equivalent to storing in a member of the union:
1522 u = (union foo) x @equiv{} u.i = x
1523 u = (union foo) y @equiv{} u.d = y
1526 You can also use the union cast as a function argument:
1529 void hack (union foo);
1531 hack ((union foo) x);
1534 @node Mixed Declarations
1535 @section Mixed Declarations and Code
1536 @cindex mixed declarations and code
1537 @cindex declarations, mixed with code
1538 @cindex code, mixed with declarations
1540 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1541 within compound statements. As an extension, GCC also allows this in
1542 C89 mode. For example, you could do:
1551 Each identifier is visible from where it is declared until the end of
1552 the enclosing block.
1554 @node Function Attributes
1555 @section Declaring Attributes of Functions
1556 @cindex function attributes
1557 @cindex declaring attributes of functions
1558 @cindex functions that never return
1559 @cindex functions that return more than once
1560 @cindex functions that have no side effects
1561 @cindex functions in arbitrary sections
1562 @cindex functions that behave like malloc
1563 @cindex @code{volatile} applied to function
1564 @cindex @code{const} applied to function
1565 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1566 @cindex functions with non-null pointer arguments
1567 @cindex functions that are passed arguments in registers on the 386
1568 @cindex functions that pop the argument stack on the 386
1569 @cindex functions that do not pop the argument stack on the 386
1571 In GNU C, you declare certain things about functions called in your program
1572 which help the compiler optimize function calls and check your code more
1575 The keyword @code{__attribute__} allows you to specify special
1576 attributes when making a declaration. This keyword is followed by an
1577 attribute specification inside double parentheses. The following
1578 attributes are currently defined for functions on all targets:
1579 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1580 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1581 @code{format}, @code{format_arg}, @code{no_instrument_function},
1582 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1583 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1584 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1585 @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 on an inline declaration results in the old GNU C89
1626 inline behavior even in the ISO C99 mode.
1628 @cindex @code{flatten} function attribute
1630 Generally, inlining into a function is limited. For a function marked with
1631 this attribute, every call inside this function will be inlined, if possible.
1632 Whether the function itself is considered for inlining depends on its size and
1633 the current inlining parameters. The @code{flatten} attribute only works
1634 reliably in unit-at-a-time mode.
1637 @cindex functions that do pop the argument stack on the 386
1639 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1640 assume that the calling function will pop off the stack space used to
1641 pass arguments. This is
1642 useful to override the effects of the @option{-mrtd} switch.
1645 @cindex @code{const} function attribute
1646 Many functions do not examine any values except their arguments, and
1647 have no effects except the return value. Basically this is just slightly
1648 more strict class than the @code{pure} attribute below, since function is not
1649 allowed to read global memory.
1651 @cindex pointer arguments
1652 Note that a function that has pointer arguments and examines the data
1653 pointed to must @emph{not} be declared @code{const}. Likewise, a
1654 function that calls a non-@code{const} function usually must not be
1655 @code{const}. It does not make sense for a @code{const} function to
1658 The attribute @code{const} is not implemented in GCC versions earlier
1659 than 2.5. An alternative way to declare that a function has no side
1660 effects, which works in the current version and in some older versions,
1664 typedef int intfn ();
1666 extern const intfn square;
1669 This approach does not work in GNU C++ from 2.6.0 on, since the language
1670 specifies that the @samp{const} must be attached to the return value.
1674 @cindex @code{constructor} function attribute
1675 @cindex @code{destructor} function attribute
1676 The @code{constructor} attribute causes the function to be called
1677 automatically before execution enters @code{main ()}. Similarly, the
1678 @code{destructor} attribute causes the function to be called
1679 automatically after @code{main ()} has completed or @code{exit ()} has
1680 been called. Functions with these attributes are useful for
1681 initializing data that will be used implicitly during the execution of
1684 These attributes are not currently implemented for Objective-C@.
1687 @cindex @code{deprecated} attribute.
1688 The @code{deprecated} attribute results in a warning if the function
1689 is used anywhere in the source file. This is useful when identifying
1690 functions that are expected to be removed in a future version of a
1691 program. The warning also includes the location of the declaration
1692 of the deprecated function, to enable users to easily find further
1693 information about why the function is deprecated, or what they should
1694 do instead. Note that the warnings only occurs for uses:
1697 int old_fn () __attribute__ ((deprecated));
1699 int (*fn_ptr)() = old_fn;
1702 results in a warning on line 3 but not line 2.
1704 The @code{deprecated} attribute can also be used for variables and
1705 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1708 @cindex @code{__declspec(dllexport)}
1709 On Microsoft Windows targets and Symbian OS targets the
1710 @code{dllexport} attribute causes the compiler to provide a global
1711 pointer to a pointer in a DLL, so that it can be referenced with the
1712 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1713 name is formed by combining @code{_imp__} and the function or variable
1716 You can use @code{__declspec(dllexport)} as a synonym for
1717 @code{__attribute__ ((dllexport))} for compatibility with other
1720 On systems that support the @code{visibility} attribute, this
1721 attribute also implies ``default'' visibility, unless a
1722 @code{visibility} attribute is explicitly specified. You should avoid
1723 the use of @code{dllexport} with ``hidden'' or ``internal''
1724 visibility; in the future GCC may issue an error for those cases.
1726 Currently, the @code{dllexport} attribute is ignored for inlined
1727 functions, unless the @option{-fkeep-inline-functions} flag has been
1728 used. The attribute is also ignored for undefined symbols.
1730 When applied to C++ classes, the attribute marks defined non-inlined
1731 member functions and static data members as exports. Static consts
1732 initialized in-class are not marked unless they are also defined
1735 For Microsoft Windows targets there are alternative methods for
1736 including the symbol in the DLL's export table such as using a
1737 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1738 the @option{--export-all} linker flag.
1741 @cindex @code{__declspec(dllimport)}
1742 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1743 attribute causes the compiler to reference a function or variable via
1744 a global pointer to a pointer that is set up by the DLL exporting the
1745 symbol. The attribute implies @code{extern} storage. On Microsoft
1746 Windows targets, the pointer name is formed by combining @code{_imp__}
1747 and the function or variable name.
1749 You can use @code{__declspec(dllimport)} as a synonym for
1750 @code{__attribute__ ((dllimport))} for compatibility with other
1753 Currently, the attribute is ignored for inlined functions. If the
1754 attribute is applied to a symbol @emph{definition}, an error is reported.
1755 If a symbol previously declared @code{dllimport} is later defined, the
1756 attribute is ignored in subsequent references, and a warning is emitted.
1757 The attribute is also overridden by a subsequent declaration as
1760 When applied to C++ classes, the attribute marks non-inlined
1761 member functions and static data members as imports. However, the
1762 attribute is ignored for virtual methods to allow creation of vtables
1765 On the SH Symbian OS target the @code{dllimport} attribute also has
1766 another affect---it can cause the vtable and run-time type information
1767 for a class to be exported. This happens when the class has a
1768 dllimport'ed constructor or a non-inline, non-pure virtual function
1769 and, for either of those two conditions, the class also has a inline
1770 constructor or destructor and has a key function that is defined in
1771 the current translation unit.
1773 For Microsoft Windows based targets the use of the @code{dllimport}
1774 attribute on functions is not necessary, but provides a small
1775 performance benefit by eliminating a thunk in the DLL@. The use of the
1776 @code{dllimport} attribute on imported variables was required on older
1777 versions of the GNU linker, but can now be avoided by passing the
1778 @option{--enable-auto-import} switch to the GNU linker. As with
1779 functions, using the attribute for a variable eliminates a thunk in
1782 One drawback to using this attribute is that a pointer to a function
1783 or variable marked as @code{dllimport} cannot be used as a constant
1784 address. On Microsoft Windows targets, the attribute can be disabled
1785 for functions by setting the @option{-mnop-fun-dllimport} flag.
1788 @cindex eight bit data on the H8/300, H8/300H, and H8S
1789 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1790 variable should be placed into the eight bit data section.
1791 The compiler will generate more efficient code for certain operations
1792 on data in the eight bit data area. Note the eight bit data area is limited to
1795 You must use GAS and GLD from GNU binutils version 2.7 or later for
1796 this attribute to work correctly.
1798 @item exception_handler
1799 @cindex exception handler functions on the Blackfin processor
1800 Use this attribute on the Blackfin to indicate that the specified function
1801 is an exception handler. The compiler will generate function entry and
1802 exit sequences suitable for use in an exception handler when this
1803 attribute is present.
1806 @cindex functions which handle memory bank switching
1807 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1808 use a calling convention that takes care of switching memory banks when
1809 entering and leaving a function. This calling convention is also the
1810 default when using the @option{-mlong-calls} option.
1812 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1813 to call and return from a function.
1815 On 68HC11 the compiler will generate a sequence of instructions
1816 to invoke a board-specific routine to switch the memory bank and call the
1817 real function. The board-specific routine simulates a @code{call}.
1818 At the end of a function, it will jump to a board-specific routine
1819 instead of using @code{rts}. The board-specific return routine simulates
1823 @cindex functions that pop the argument stack on the 386
1824 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1825 pass the first argument (if of integral type) in the register ECX and
1826 the second argument (if of integral type) in the register EDX@. Subsequent
1827 and other typed arguments are passed on the stack. The called function will
1828 pop the arguments off the stack. If the number of arguments is variable all
1829 arguments are pushed on the stack.
1831 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1832 @cindex @code{format} function attribute
1834 The @code{format} attribute specifies that a function takes @code{printf},
1835 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1836 should be type-checked against a format string. For example, the
1841 my_printf (void *my_object, const char *my_format, ...)
1842 __attribute__ ((format (printf, 2, 3)));
1846 causes the compiler to check the arguments in calls to @code{my_printf}
1847 for consistency with the @code{printf} style format string argument
1850 The parameter @var{archetype} determines how the format string is
1851 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1852 or @code{strfmon}. (You can also use @code{__printf__},
1853 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1854 parameter @var{string-index} specifies which argument is the format
1855 string argument (starting from 1), while @var{first-to-check} is the
1856 number of the first argument to check against the format string. For
1857 functions where the arguments are not available to be checked (such as
1858 @code{vprintf}), specify the third parameter as zero. In this case the
1859 compiler only checks the format string for consistency. For
1860 @code{strftime} formats, the third parameter is required to be zero.
1861 Since non-static C++ methods have an implicit @code{this} argument, the
1862 arguments of such methods should be counted from two, not one, when
1863 giving values for @var{string-index} and @var{first-to-check}.
1865 In the example above, the format string (@code{my_format}) is the second
1866 argument of the function @code{my_print}, and the arguments to check
1867 start with the third argument, so the correct parameters for the format
1868 attribute are 2 and 3.
1870 @opindex ffreestanding
1871 @opindex fno-builtin
1872 The @code{format} attribute allows you to identify your own functions
1873 which take format strings as arguments, so that GCC can check the
1874 calls to these functions for errors. The compiler always (unless
1875 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1876 for the standard library functions @code{printf}, @code{fprintf},
1877 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1878 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1879 warnings are requested (using @option{-Wformat}), so there is no need to
1880 modify the header file @file{stdio.h}. In C99 mode, the functions
1881 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1882 @code{vsscanf} are also checked. Except in strictly conforming C
1883 standard modes, the X/Open function @code{strfmon} is also checked as
1884 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1885 @xref{C Dialect Options,,Options Controlling C Dialect}.
1887 The target may provide additional types of format checks.
1888 @xref{Target Format Checks,,Format Checks Specific to Particular
1891 @item format_arg (@var{string-index})
1892 @cindex @code{format_arg} function attribute
1893 @opindex Wformat-nonliteral
1894 The @code{format_arg} attribute specifies that a function takes a format
1895 string for a @code{printf}, @code{scanf}, @code{strftime} or
1896 @code{strfmon} style function and modifies it (for example, to translate
1897 it into another language), so the result can be passed to a
1898 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1899 function (with the remaining arguments to the format function the same
1900 as they would have been for the unmodified string). For example, the
1905 my_dgettext (char *my_domain, const char *my_format)
1906 __attribute__ ((format_arg (2)));
1910 causes the compiler to check the arguments in calls to a @code{printf},
1911 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1912 format string argument is a call to the @code{my_dgettext} function, for
1913 consistency with the format string argument @code{my_format}. If the
1914 @code{format_arg} attribute had not been specified, all the compiler
1915 could tell in such calls to format functions would be that the format
1916 string argument is not constant; this would generate a warning when
1917 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1918 without the attribute.
1920 The parameter @var{string-index} specifies which argument is the format
1921 string argument (starting from one). Since non-static C++ methods have
1922 an implicit @code{this} argument, the arguments of such methods should
1923 be counted from two.
1925 The @code{format-arg} attribute allows you to identify your own
1926 functions which modify format strings, so that GCC can check the
1927 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1928 type function whose operands are a call to one of your own function.
1929 The compiler always treats @code{gettext}, @code{dgettext}, and
1930 @code{dcgettext} in this manner except when strict ISO C support is
1931 requested by @option{-ansi} or an appropriate @option{-std} option, or
1932 @option{-ffreestanding} or @option{-fno-builtin}
1933 is used. @xref{C Dialect Options,,Options
1934 Controlling C Dialect}.
1936 @item function_vector
1937 @cindex calling functions through the function vector on the H8/300 processors
1938 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1939 function should be called through the function vector. Calling a
1940 function through the function vector will reduce code size, however;
1941 the function vector has a limited size (maximum 128 entries on the H8/300
1942 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
1944 You must use GAS and GLD from GNU binutils version 2.7 or later for
1945 this attribute to work correctly.
1948 @cindex interrupt handler functions
1949 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
1950 ports to indicate that the specified function is an interrupt handler.
1951 The compiler will generate function entry and exit sequences suitable
1952 for use in an interrupt handler when this attribute is present.
1954 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
1955 SH processors can be specified via the @code{interrupt_handler} attribute.
1957 Note, on the AVR, interrupts will be enabled inside the function.
1959 Note, for the ARM, you can specify the kind of interrupt to be handled by
1960 adding an optional parameter to the interrupt attribute like this:
1963 void f () __attribute__ ((interrupt ("IRQ")));
1966 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
1968 On ARMv7-M the interrupt type is ignored, and the attibute means the function
1969 may be called with a word aligned stack pointer.
1971 @item interrupt_handler
1972 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
1973 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
1974 indicate that the specified function is an interrupt handler. The compiler
1975 will generate function entry and exit sequences suitable for use in an
1976 interrupt handler when this attribute is present.
1979 @cindex User stack pointer in interrupts on the Blackfin
1980 When used together with @code{interrupt_handler}, @code{exception_handler}
1981 or @code{nmi_handler}, code will be generated to load the stack pointer
1982 from the USP register in the function prologue.
1984 @item long_call/short_call
1985 @cindex indirect calls on ARM
1986 This attribute specifies how a particular function is called on
1987 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
1988 command line switch and @code{#pragma long_calls} settings. The
1989 @code{long_call} attribute indicates that the function might be far
1990 away from the call site and require a different (more expensive)
1991 calling sequence. The @code{short_call} attribute always places
1992 the offset to the function from the call site into the @samp{BL}
1993 instruction directly.
1995 @item longcall/shortcall
1996 @cindex functions called via pointer on the RS/6000 and PowerPC
1997 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
1998 indicates that the function might be far away from the call site and
1999 require a different (more expensive) calling sequence. The
2000 @code{shortcall} attribute indicates that the function is always close
2001 enough for the shorter calling sequence to be used. These attributes
2002 override both the @option{-mlongcall} switch and, on the RS/6000 and
2003 PowerPC, the @code{#pragma longcall} setting.
2005 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2006 calls are necessary.
2009 @cindex indirect calls on MIPS
2010 This attribute specifies how a particular function is called on MIPS@.
2011 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2012 command line switch. This attribute causes the compiler to always call
2013 the function by first loading its address into a register, and then using
2014 the contents of that register.
2017 @cindex @code{malloc} attribute
2018 The @code{malloc} attribute is used to tell the compiler that a function
2019 may be treated as if any non-@code{NULL} pointer it returns cannot
2020 alias any other pointer valid when the function returns.
2021 This will often improve optimization.
2022 Standard functions with this property include @code{malloc} and
2023 @code{calloc}. @code{realloc}-like functions have this property as
2024 long as the old pointer is never referred to (including comparing it
2025 to the new pointer) after the function returns a non-@code{NULL}
2028 @item model (@var{model-name})
2029 @cindex function addressability on the M32R/D
2030 @cindex variable addressability on the IA-64
2032 On the M32R/D, use this attribute to set the addressability of an
2033 object, and of the code generated for a function. The identifier
2034 @var{model-name} is one of @code{small}, @code{medium}, or
2035 @code{large}, representing each of the code models.
2037 Small model objects live in the lower 16MB of memory (so that their
2038 addresses can be loaded with the @code{ld24} instruction), and are
2039 callable with the @code{bl} instruction.
2041 Medium model objects may live anywhere in the 32-bit address space (the
2042 compiler will generate @code{seth/add3} instructions to load their addresses),
2043 and are callable with the @code{bl} instruction.
2045 Large model objects may live anywhere in the 32-bit address space (the
2046 compiler will generate @code{seth/add3} instructions to load their addresses),
2047 and may not be reachable with the @code{bl} instruction (the compiler will
2048 generate the much slower @code{seth/add3/jl} instruction sequence).
2050 On IA-64, use this attribute to set the addressability of an object.
2051 At present, the only supported identifier for @var{model-name} is
2052 @code{small}, indicating addressability via ``small'' (22-bit)
2053 addresses (so that their addresses can be loaded with the @code{addl}
2054 instruction). Caveat: such addressing is by definition not position
2055 independent and hence this attribute must not be used for objects
2056 defined by shared libraries.
2059 @cindex function without a prologue/epilogue code
2060 Use this attribute on the ARM, AVR, C4x, IP2K and SPU ports to indicate that
2061 the specified function does not need prologue/epilogue sequences generated by
2062 the compiler. It is up to the programmer to provide these sequences.
2065 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2066 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2067 use the normal calling convention based on @code{jsr} and @code{rts}.
2068 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2072 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2073 Use this attribute together with @code{interrupt_handler},
2074 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2075 entry code should enable nested interrupts or exceptions.
2078 @cindex NMI handler functions on the Blackfin processor
2079 Use this attribute on the Blackfin to indicate that the specified function
2080 is an NMI handler. The compiler will generate function entry and
2081 exit sequences suitable for use in an NMI handler when this
2082 attribute is present.
2084 @item no_instrument_function
2085 @cindex @code{no_instrument_function} function attribute
2086 @opindex finstrument-functions
2087 If @option{-finstrument-functions} is given, profiling function calls will
2088 be generated at entry and exit of most user-compiled functions.
2089 Functions with this attribute will not be so instrumented.
2092 @cindex @code{noinline} function attribute
2093 This function attribute prevents a function from being considered for
2096 @item nonnull (@var{arg-index}, @dots{})
2097 @cindex @code{nonnull} function attribute
2098 The @code{nonnull} attribute specifies that some function parameters should
2099 be non-null pointers. For instance, the declaration:
2103 my_memcpy (void *dest, const void *src, size_t len)
2104 __attribute__((nonnull (1, 2)));
2108 causes the compiler to check that, in calls to @code{my_memcpy},
2109 arguments @var{dest} and @var{src} are non-null. If the compiler
2110 determines that a null pointer is passed in an argument slot marked
2111 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2112 is issued. The compiler may also choose to make optimizations based
2113 on the knowledge that certain function arguments will not be null.
2115 If no argument index list is given to the @code{nonnull} attribute,
2116 all pointer arguments are marked as non-null. To illustrate, the
2117 following declaration is equivalent to the previous example:
2121 my_memcpy (void *dest, const void *src, size_t len)
2122 __attribute__((nonnull));
2126 @cindex @code{noreturn} function attribute
2127 A few standard library functions, such as @code{abort} and @code{exit},
2128 cannot return. GCC knows this automatically. Some programs define
2129 their own functions that never return. You can declare them
2130 @code{noreturn} to tell the compiler this fact. For example,
2134 void fatal () __attribute__ ((noreturn));
2137 fatal (/* @r{@dots{}} */)
2139 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2145 The @code{noreturn} keyword tells the compiler to assume that
2146 @code{fatal} cannot return. It can then optimize without regard to what
2147 would happen if @code{fatal} ever did return. This makes slightly
2148 better code. More importantly, it helps avoid spurious warnings of
2149 uninitialized variables.
2151 The @code{noreturn} keyword does not affect the exceptional path when that
2152 applies: a @code{noreturn}-marked function may still return to the caller
2153 by throwing an exception or calling @code{longjmp}.
2155 Do not assume that registers saved by the calling function are
2156 restored before calling the @code{noreturn} function.
2158 It does not make sense for a @code{noreturn} function to have a return
2159 type other than @code{void}.
2161 The attribute @code{noreturn} is not implemented in GCC versions
2162 earlier than 2.5. An alternative way to declare that a function does
2163 not return, which works in the current version and in some older
2164 versions, is as follows:
2167 typedef void voidfn ();
2169 volatile voidfn fatal;
2172 This approach does not work in GNU C++.
2175 @cindex @code{nothrow} function attribute
2176 The @code{nothrow} attribute is used to inform the compiler that a
2177 function cannot throw an exception. For example, most functions in
2178 the standard C library can be guaranteed not to throw an exception
2179 with the notable exceptions of @code{qsort} and @code{bsearch} that
2180 take function pointer arguments. The @code{nothrow} attribute is not
2181 implemented in GCC versions earlier than 3.3.
2184 @cindex @code{pure} function attribute
2185 Many functions have no effects except the return value and their
2186 return value depends only on the parameters and/or global variables.
2187 Such a function can be subject
2188 to common subexpression elimination and loop optimization just as an
2189 arithmetic operator would be. These functions should be declared
2190 with the attribute @code{pure}. For example,
2193 int square (int) __attribute__ ((pure));
2197 says that the hypothetical function @code{square} is safe to call
2198 fewer times than the program says.
2200 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2201 Interesting non-pure functions are functions with infinite loops or those
2202 depending on volatile memory or other system resource, that may change between
2203 two consecutive calls (such as @code{feof} in a multithreading environment).
2205 The attribute @code{pure} is not implemented in GCC versions earlier
2208 @item regparm (@var{number})
2209 @cindex @code{regparm} attribute
2210 @cindex functions that are passed arguments in registers on the 386
2211 On the Intel 386, the @code{regparm} attribute causes the compiler to
2212 pass arguments number one to @var{number} if they are of integral type
2213 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2214 take a variable number of arguments will continue to be passed all of their
2215 arguments on the stack.
2217 Beware that on some ELF systems this attribute is unsuitable for
2218 global functions in shared libraries with lazy binding (which is the
2219 default). Lazy binding will send the first call via resolving code in
2220 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2221 per the standard calling conventions. Solaris 8 is affected by this.
2222 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2223 safe since the loaders there save all registers. (Lazy binding can be
2224 disabled with the linker or the loader if desired, to avoid the
2228 @cindex @code{sseregparm} attribute
2229 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2230 causes the compiler to pass up to 3 floating point arguments in
2231 SSE registers instead of on the stack. Functions that take a
2232 variable number of arguments will continue to pass all of their
2233 floating point arguments on the stack.
2235 @item force_align_arg_pointer
2236 @cindex @code{force_align_arg_pointer} attribute
2237 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2238 applied to individual function definitions, generating an alternate
2239 prologue and epilogue that realigns the runtime stack. This supports
2240 mixing legacy codes that run with a 4-byte aligned stack with modern
2241 codes that keep a 16-byte stack for SSE compatibility. The alternate
2242 prologue and epilogue are slower and bigger than the regular ones, and
2243 the alternate prologue requires a scratch register; this lowers the
2244 number of registers available if used in conjunction with the
2245 @code{regparm} attribute. The @code{force_align_arg_pointer}
2246 attribute is incompatible with nested functions; this is considered a
2250 @cindex @code{returns_twice} attribute
2251 The @code{returns_twice} attribute tells the compiler that a function may
2252 return more than one time. The compiler will ensure that all registers
2253 are dead before calling such a function and will emit a warning about
2254 the variables that may be clobbered after the second return from the
2255 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2256 The @code{longjmp}-like counterpart of such function, if any, might need
2257 to be marked with the @code{noreturn} attribute.
2260 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2261 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2262 all registers except the stack pointer should be saved in the prologue
2263 regardless of whether they are used or not.
2265 @item section ("@var{section-name}")
2266 @cindex @code{section} function attribute
2267 Normally, the compiler places the code it generates in the @code{text} section.
2268 Sometimes, however, you need additional sections, or you need certain
2269 particular functions to appear in special sections. The @code{section}
2270 attribute specifies that a function lives in a particular section.
2271 For example, the declaration:
2274 extern void foobar (void) __attribute__ ((section ("bar")));
2278 puts the function @code{foobar} in the @code{bar} section.
2280 Some file formats do not support arbitrary sections so the @code{section}
2281 attribute is not available on all platforms.
2282 If you need to map the entire contents of a module to a particular
2283 section, consider using the facilities of the linker instead.
2286 @cindex @code{sentinel} function attribute
2287 This function attribute ensures that a parameter in a function call is
2288 an explicit @code{NULL}. The attribute is only valid on variadic
2289 functions. By default, the sentinel is located at position zero, the
2290 last parameter of the function call. If an optional integer position
2291 argument P is supplied to the attribute, the sentinel must be located at
2292 position P counting backwards from the end of the argument list.
2295 __attribute__ ((sentinel))
2297 __attribute__ ((sentinel(0)))
2300 The attribute is automatically set with a position of 0 for the built-in
2301 functions @code{execl} and @code{execlp}. The built-in function
2302 @code{execle} has the attribute set with a position of 1.
2304 A valid @code{NULL} in this context is defined as zero with any pointer
2305 type. If your system defines the @code{NULL} macro with an integer type
2306 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2307 with a copy that redefines NULL appropriately.
2309 The warnings for missing or incorrect sentinels are enabled with
2313 See long_call/short_call.
2316 See longcall/shortcall.
2319 @cindex signal handler functions on the AVR processors
2320 Use this attribute on the AVR to indicate that the specified
2321 function is a signal handler. The compiler will generate function
2322 entry and exit sequences suitable for use in a signal handler when this
2323 attribute is present. Interrupts will be disabled inside the function.
2326 Use this attribute on the SH to indicate an @code{interrupt_handler}
2327 function should switch to an alternate stack. It expects a string
2328 argument that names a global variable holding the address of the
2333 void f () __attribute__ ((interrupt_handler,
2334 sp_switch ("alt_stack")));
2338 @cindex functions that pop the argument stack on the 386
2339 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2340 assume that the called function will pop off the stack space used to
2341 pass arguments, unless it takes a variable number of arguments.
2344 @cindex tiny data section on the H8/300H and H8S
2345 Use this attribute on the H8/300H and H8S to indicate that the specified
2346 variable should be placed into the tiny data section.
2347 The compiler will generate more efficient code for loads and stores
2348 on data in the tiny data section. Note the tiny data area is limited to
2349 slightly under 32kbytes of data.
2352 Use this attribute on the SH for an @code{interrupt_handler} to return using
2353 @code{trapa} instead of @code{rte}. This attribute expects an integer
2354 argument specifying the trap number to be used.
2357 @cindex @code{unused} attribute.
2358 This attribute, attached to a function, means that the function is meant
2359 to be possibly unused. GCC will not produce a warning for this
2363 @cindex @code{used} attribute.
2364 This attribute, attached to a function, means that code must be emitted
2365 for the function even if it appears that the function is not referenced.
2366 This is useful, for example, when the function is referenced only in
2369 @item visibility ("@var{visibility_type}")
2370 @cindex @code{visibility} attribute
2371 This attribute affects the linkage of the declaration to which it is attached.
2372 There are four supported @var{visibility_type} values: default,
2373 hidden, protected or internal visibility.
2376 void __attribute__ ((visibility ("protected")))
2377 f () @{ /* @r{Do something.} */; @}
2378 int i __attribute__ ((visibility ("hidden")));
2381 The possible values of @var{visibility_type} correspond to the
2382 visibility settings in the ELF gABI.
2385 @c keep this list of visibilities in alphabetical order.
2388 Default visibility is the normal case for the object file format.
2389 This value is available for the visibility attribute to override other
2390 options that may change the assumed visibility of entities.
2392 On ELF, default visibility means that the declaration is visible to other
2393 modules and, in shared libraries, means that the declared entity may be
2396 On Darwin, default visibility means that the declaration is visible to
2399 Default visibility corresponds to ``external linkage'' in the language.
2402 Hidden visibility indicates that the entity declared will have a new
2403 form of linkage, which we'll call ``hidden linkage''. Two
2404 declarations of an object with hidden linkage refer to the same object
2405 if they are in the same shared object.
2408 Internal visibility is like hidden visibility, but with additional
2409 processor specific semantics. Unless otherwise specified by the
2410 psABI, GCC defines internal visibility to mean that a function is
2411 @emph{never} called from another module. Compare this with hidden
2412 functions which, while they cannot be referenced directly by other
2413 modules, can be referenced indirectly via function pointers. By
2414 indicating that a function cannot be called from outside the module,
2415 GCC may for instance omit the load of a PIC register since it is known
2416 that the calling function loaded the correct value.
2419 Protected visibility is like default visibility except that it
2420 indicates that references within the defining module will bind to the
2421 definition in that module. That is, the declared entity cannot be
2422 overridden by another module.
2426 All visibilities are supported on many, but not all, ELF targets
2427 (supported when the assembler supports the @samp{.visibility}
2428 pseudo-op). Default visibility is supported everywhere. Hidden
2429 visibility is supported on Darwin targets.
2431 The visibility attribute should be applied only to declarations which
2432 would otherwise have external linkage. The attribute should be applied
2433 consistently, so that the same entity should not be declared with
2434 different settings of the attribute.
2436 In C++, the visibility attribute applies to types as well as functions
2437 and objects, because in C++ types have linkage. A class must not have
2438 greater visibility than its non-static data member types and bases,
2439 and class members default to the visibility of their class. Also, a
2440 declaration without explicit visibility is limited to the visibility
2443 In C++, you can mark member functions and static member variables of a
2444 class with the visibility attribute. This is useful if if you know a
2445 particular method or static member variable should only be used from
2446 one shared object; then you can mark it hidden while the rest of the
2447 class has default visibility. Care must be taken to avoid breaking
2448 the One Definition Rule; for example, it is usually not useful to mark
2449 an inline method as hidden without marking the whole class as hidden.
2451 A C++ namespace declaration can also have the visibility attribute.
2452 This attribute applies only to the particular namespace body, not to
2453 other definitions of the same namespace; it is equivalent to using
2454 @samp{#pragma GCC visibility} before and after the namespace
2455 definition (@pxref{Visibility Pragmas}).
2457 In C++, if a template argument has limited visibility, this
2458 restriction is implicitly propagated to the template instantiation.
2459 Otherwise, template instantiations and specializations default to the
2460 visibility of their template.
2462 If both the template and enclosing class have explicit visibility, the
2463 visibility from the template is used.
2465 @item warn_unused_result
2466 @cindex @code{warn_unused_result} attribute
2467 The @code{warn_unused_result} attribute causes a warning to be emitted
2468 if a caller of the function with this attribute does not use its
2469 return value. This is useful for functions where not checking
2470 the result is either a security problem or always a bug, such as
2474 int fn () __attribute__ ((warn_unused_result));
2477 if (fn () < 0) return -1;
2483 results in warning on line 5.
2486 @cindex @code{weak} attribute
2487 The @code{weak} attribute causes the declaration to be emitted as a weak
2488 symbol rather than a global. This is primarily useful in defining
2489 library functions which can be overridden in user code, though it can
2490 also be used with non-function declarations. Weak symbols are supported
2491 for ELF targets, and also for a.out targets when using the GNU assembler
2495 @itemx weakref ("@var{target}")
2496 @cindex @code{weakref} attribute
2497 The @code{weakref} attribute marks a declaration as a weak reference.
2498 Without arguments, it should be accompanied by an @code{alias} attribute
2499 naming the target symbol. Optionally, the @var{target} may be given as
2500 an argument to @code{weakref} itself. In either case, @code{weakref}
2501 implicitly marks the declaration as @code{weak}. Without a
2502 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2503 @code{weakref} is equivalent to @code{weak}.
2506 static int x() __attribute__ ((weakref ("y")));
2507 /* is equivalent to... */
2508 static int x() __attribute__ ((weak, weakref, alias ("y")));
2510 static int x() __attribute__ ((weakref));
2511 static int x() __attribute__ ((alias ("y")));
2514 A weak reference is an alias that does not by itself require a
2515 definition to be given for the target symbol. If the target symbol is
2516 only referenced through weak references, then the becomes a @code{weak}
2517 undefined symbol. If it is directly referenced, however, then such
2518 strong references prevail, and a definition will be required for the
2519 symbol, not necessarily in the same translation unit.
2521 The effect is equivalent to moving all references to the alias to a
2522 separate translation unit, renaming the alias to the aliased symbol,
2523 declaring it as weak, compiling the two separate translation units and
2524 performing a reloadable link on them.
2526 At present, a declaration to which @code{weakref} is attached can
2527 only be @code{static}.
2529 @item externally_visible
2530 @cindex @code{externally_visible} attribute.
2531 This attribute, attached to a global variable or function nullify
2532 effect of @option{-fwhole-program} command line option, so the object
2533 remain visible outside the current compilation unit
2537 You can specify multiple attributes in a declaration by separating them
2538 by commas within the double parentheses or by immediately following an
2539 attribute declaration with another attribute declaration.
2541 @cindex @code{#pragma}, reason for not using
2542 @cindex pragma, reason for not using
2543 Some people object to the @code{__attribute__} feature, suggesting that
2544 ISO C's @code{#pragma} should be used instead. At the time
2545 @code{__attribute__} was designed, there were two reasons for not doing
2550 It is impossible to generate @code{#pragma} commands from a macro.
2553 There is no telling what the same @code{#pragma} might mean in another
2557 These two reasons applied to almost any application that might have been
2558 proposed for @code{#pragma}. It was basically a mistake to use
2559 @code{#pragma} for @emph{anything}.
2561 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2562 to be generated from macros. In addition, a @code{#pragma GCC}
2563 namespace is now in use for GCC-specific pragmas. However, it has been
2564 found convenient to use @code{__attribute__} to achieve a natural
2565 attachment of attributes to their corresponding declarations, whereas
2566 @code{#pragma GCC} is of use for constructs that do not naturally form
2567 part of the grammar. @xref{Other Directives,,Miscellaneous
2568 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2570 @node Attribute Syntax
2571 @section Attribute Syntax
2572 @cindex attribute syntax
2574 This section describes the syntax with which @code{__attribute__} may be
2575 used, and the constructs to which attribute specifiers bind, for the C
2576 language. Some details may vary for C++ and Objective-C@. Because of
2577 infelicities in the grammar for attributes, some forms described here
2578 may not be successfully parsed in all cases.
2580 There are some problems with the semantics of attributes in C++. For
2581 example, there are no manglings for attributes, although they may affect
2582 code generation, so problems may arise when attributed types are used in
2583 conjunction with templates or overloading. Similarly, @code{typeid}
2584 does not distinguish between types with different attributes. Support
2585 for attributes in C++ may be restricted in future to attributes on
2586 declarations only, but not on nested declarators.
2588 @xref{Function Attributes}, for details of the semantics of attributes
2589 applying to functions. @xref{Variable Attributes}, for details of the
2590 semantics of attributes applying to variables. @xref{Type Attributes},
2591 for details of the semantics of attributes applying to structure, union
2592 and enumerated types.
2594 An @dfn{attribute specifier} is of the form
2595 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2596 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2597 each attribute is one of the following:
2601 Empty. Empty attributes are ignored.
2604 A word (which may be an identifier such as @code{unused}, or a reserved
2605 word such as @code{const}).
2608 A word, followed by, in parentheses, parameters for the attribute.
2609 These parameters take one of the following forms:
2613 An identifier. For example, @code{mode} attributes use this form.
2616 An identifier followed by a comma and a non-empty comma-separated list
2617 of expressions. For example, @code{format} attributes use this form.
2620 A possibly empty comma-separated list of expressions. For example,
2621 @code{format_arg} attributes use this form with the list being a single
2622 integer constant expression, and @code{alias} attributes use this form
2623 with the list being a single string constant.
2627 An @dfn{attribute specifier list} is a sequence of one or more attribute
2628 specifiers, not separated by any other tokens.
2630 In GNU C, an attribute specifier list may appear after the colon following a
2631 label, other than a @code{case} or @code{default} label. The only
2632 attribute it makes sense to use after a label is @code{unused}. This
2633 feature is intended for code generated by programs which contains labels
2634 that may be unused but which is compiled with @option{-Wall}. It would
2635 not normally be appropriate to use in it human-written code, though it
2636 could be useful in cases where the code that jumps to the label is
2637 contained within an @code{#ifdef} conditional. GNU C++ does not permit
2638 such placement of attribute lists, as it is permissible for a
2639 declaration, which could begin with an attribute list, to be labelled in
2640 C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2641 does not arise there.
2643 An attribute specifier list may appear as part of a @code{struct},
2644 @code{union} or @code{enum} specifier. It may go either immediately
2645 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2646 the closing brace. The former syntax is preferred.
2647 Where attribute specifiers follow the closing brace, they are considered
2648 to relate to the structure, union or enumerated type defined, not to any
2649 enclosing declaration the type specifier appears in, and the type
2650 defined is not complete until after the attribute specifiers.
2651 @c Otherwise, there would be the following problems: a shift/reduce
2652 @c conflict between attributes binding the struct/union/enum and
2653 @c binding to the list of specifiers/qualifiers; and "aligned"
2654 @c attributes could use sizeof for the structure, but the size could be
2655 @c changed later by "packed" attributes.
2657 Otherwise, an attribute specifier appears as part of a declaration,
2658 counting declarations of unnamed parameters and type names, and relates
2659 to that declaration (which may be nested in another declaration, for
2660 example in the case of a parameter declaration), or to a particular declarator
2661 within a declaration. Where an
2662 attribute specifier is applied to a parameter declared as a function or
2663 an array, it should apply to the function or array rather than the
2664 pointer to which the parameter is implicitly converted, but this is not
2665 yet correctly implemented.
2667 Any list of specifiers and qualifiers at the start of a declaration may
2668 contain attribute specifiers, whether or not such a list may in that
2669 context contain storage class specifiers. (Some attributes, however,
2670 are essentially in the nature of storage class specifiers, and only make
2671 sense where storage class specifiers may be used; for example,
2672 @code{section}.) There is one necessary limitation to this syntax: the
2673 first old-style parameter declaration in a function definition cannot
2674 begin with an attribute specifier, because such an attribute applies to
2675 the function instead by syntax described below (which, however, is not
2676 yet implemented in this case). In some other cases, attribute
2677 specifiers are permitted by this grammar but not yet supported by the
2678 compiler. All attribute specifiers in this place relate to the
2679 declaration as a whole. In the obsolescent usage where a type of
2680 @code{int} is implied by the absence of type specifiers, such a list of
2681 specifiers and qualifiers may be an attribute specifier list with no
2682 other specifiers or qualifiers.
2684 At present, the first parameter in a function prototype must have some
2685 type specifier which is not an attribute specifier; this resolves an
2686 ambiguity in the interpretation of @code{void f(int
2687 (__attribute__((foo)) x))}, but is subject to change. At present, if
2688 the parentheses of a function declarator contain only attributes then
2689 those attributes are ignored, rather than yielding an error or warning
2690 or implying a single parameter of type int, but this is subject to
2693 An attribute specifier list may appear immediately before a declarator
2694 (other than the first) in a comma-separated list of declarators in a
2695 declaration of more than one identifier using a single list of
2696 specifiers and qualifiers. Such attribute specifiers apply
2697 only to the identifier before whose declarator they appear. For
2701 __attribute__((noreturn)) void d0 (void),
2702 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2707 the @code{noreturn} attribute applies to all the functions
2708 declared; the @code{format} attribute only applies to @code{d1}.
2710 An attribute specifier list may appear immediately before the comma,
2711 @code{=} or semicolon terminating the declaration of an identifier other
2712 than a function definition. At present, such attribute specifiers apply
2713 to the declared object or function, but in future they may attach to the
2714 outermost adjacent declarator. In simple cases there is no difference,
2715 but, for example, in
2718 void (****f)(void) __attribute__((noreturn));
2722 at present the @code{noreturn} attribute applies to @code{f}, which
2723 causes a warning since @code{f} is not a function, but in future it may
2724 apply to the function @code{****f}. The precise semantics of what
2725 attributes in such cases will apply to are not yet specified. Where an
2726 assembler name for an object or function is specified (@pxref{Asm
2727 Labels}), at present the attribute must follow the @code{asm}
2728 specification; in future, attributes before the @code{asm} specification
2729 may apply to the adjacent declarator, and those after it to the declared
2732 An attribute specifier list may, in future, be permitted to appear after
2733 the declarator in a function definition (before any old-style parameter
2734 declarations or the function body).
2736 Attribute specifiers may be mixed with type qualifiers appearing inside
2737 the @code{[]} of a parameter array declarator, in the C99 construct by
2738 which such qualifiers are applied to the pointer to which the array is
2739 implicitly converted. Such attribute specifiers apply to the pointer,
2740 not to the array, but at present this is not implemented and they are
2743 An attribute specifier list may appear at the start of a nested
2744 declarator. At present, there are some limitations in this usage: the
2745 attributes correctly apply to the declarator, but for most individual
2746 attributes the semantics this implies are not implemented.
2747 When attribute specifiers follow the @code{*} of a pointer
2748 declarator, they may be mixed with any type qualifiers present.
2749 The following describes the formal semantics of this syntax. It will make the
2750 most sense if you are familiar with the formal specification of
2751 declarators in the ISO C standard.
2753 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2754 D1}, where @code{T} contains declaration specifiers that specify a type
2755 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2756 contains an identifier @var{ident}. The type specified for @var{ident}
2757 for derived declarators whose type does not include an attribute
2758 specifier is as in the ISO C standard.
2760 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2761 and the declaration @code{T D} specifies the type
2762 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2763 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2764 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2766 If @code{D1} has the form @code{*
2767 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2768 declaration @code{T D} specifies the type
2769 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2770 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2771 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2777 void (__attribute__((noreturn)) ****f) (void);
2781 specifies the type ``pointer to pointer to pointer to pointer to
2782 non-returning function returning @code{void}''. As another example,
2785 char *__attribute__((aligned(8))) *f;
2789 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2790 Note again that this does not work with most attributes; for example,
2791 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2792 is not yet supported.
2794 For compatibility with existing code written for compiler versions that
2795 did not implement attributes on nested declarators, some laxity is
2796 allowed in the placing of attributes. If an attribute that only applies
2797 to types is applied to a declaration, it will be treated as applying to
2798 the type of that declaration. If an attribute that only applies to
2799 declarations is applied to the type of a declaration, it will be treated
2800 as applying to that declaration; and, for compatibility with code
2801 placing the attributes immediately before the identifier declared, such
2802 an attribute applied to a function return type will be treated as
2803 applying to the function type, and such an attribute applied to an array
2804 element type will be treated as applying to the array type. If an
2805 attribute that only applies to function types is applied to a
2806 pointer-to-function type, it will be treated as applying to the pointer
2807 target type; if such an attribute is applied to a function return type
2808 that is not a pointer-to-function type, it will be treated as applying
2809 to the function type.
2811 @node Function Prototypes
2812 @section Prototypes and Old-Style Function Definitions
2813 @cindex function prototype declarations
2814 @cindex old-style function definitions
2815 @cindex promotion of formal parameters
2817 GNU C extends ISO C to allow a function prototype to override a later
2818 old-style non-prototype definition. Consider the following example:
2821 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2828 /* @r{Prototype function declaration.} */
2829 int isroot P((uid_t));
2831 /* @r{Old-style function definition.} */
2833 isroot (x) /* @r{??? lossage here ???} */
2840 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2841 not allow this example, because subword arguments in old-style
2842 non-prototype definitions are promoted. Therefore in this example the
2843 function definition's argument is really an @code{int}, which does not
2844 match the prototype argument type of @code{short}.
2846 This restriction of ISO C makes it hard to write code that is portable
2847 to traditional C compilers, because the programmer does not know
2848 whether the @code{uid_t} type is @code{short}, @code{int}, or
2849 @code{long}. Therefore, in cases like these GNU C allows a prototype
2850 to override a later old-style definition. More precisely, in GNU C, a
2851 function prototype argument type overrides the argument type specified
2852 by a later old-style definition if the former type is the same as the
2853 latter type before promotion. Thus in GNU C the above example is
2854 equivalent to the following:
2867 GNU C++ does not support old-style function definitions, so this
2868 extension is irrelevant.
2871 @section C++ Style Comments
2873 @cindex C++ comments
2874 @cindex comments, C++ style
2876 In GNU C, you may use C++ style comments, which start with @samp{//} and
2877 continue until the end of the line. Many other C implementations allow
2878 such comments, and they are included in the 1999 C standard. However,
2879 C++ style comments are not recognized if you specify an @option{-std}
2880 option specifying a version of ISO C before C99, or @option{-ansi}
2881 (equivalent to @option{-std=c89}).
2884 @section Dollar Signs in Identifier Names
2886 @cindex dollar signs in identifier names
2887 @cindex identifier names, dollar signs in
2889 In GNU C, you may normally use dollar signs in identifier names.
2890 This is because many traditional C implementations allow such identifiers.
2891 However, dollar signs in identifiers are not supported on a few target
2892 machines, typically because the target assembler does not allow them.
2894 @node Character Escapes
2895 @section The Character @key{ESC} in Constants
2897 You can use the sequence @samp{\e} in a string or character constant to
2898 stand for the ASCII character @key{ESC}.
2901 @section Inquiring on Alignment of Types or Variables
2903 @cindex type alignment
2904 @cindex variable alignment
2906 The keyword @code{__alignof__} allows you to inquire about how an object
2907 is aligned, or the minimum alignment usually required by a type. Its
2908 syntax is just like @code{sizeof}.
2910 For example, if the target machine requires a @code{double} value to be
2911 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2912 This is true on many RISC machines. On more traditional machine
2913 designs, @code{__alignof__ (double)} is 4 or even 2.
2915 Some machines never actually require alignment; they allow reference to any
2916 data type even at an odd address. For these machines, @code{__alignof__}
2917 reports the @emph{recommended} alignment of a type.
2919 If the operand of @code{__alignof__} is an lvalue rather than a type,
2920 its value is the required alignment for its type, taking into account
2921 any minimum alignment specified with GCC's @code{__attribute__}
2922 extension (@pxref{Variable Attributes}). For example, after this
2926 struct foo @{ int x; char y; @} foo1;
2930 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2931 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2933 It is an error to ask for the alignment of an incomplete type.
2935 @node Variable Attributes
2936 @section Specifying Attributes of Variables
2937 @cindex attribute of variables
2938 @cindex variable attributes
2940 The keyword @code{__attribute__} allows you to specify special
2941 attributes of variables or structure fields. This keyword is followed
2942 by an attribute specification inside double parentheses. Some
2943 attributes are currently defined generically for variables.
2944 Other attributes are defined for variables on particular target
2945 systems. Other attributes are available for functions
2946 (@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
2947 Other front ends might define more attributes
2948 (@pxref{C++ Extensions,,Extensions to the C++ Language}).
2950 You may also specify attributes with @samp{__} preceding and following
2951 each keyword. This allows you to use them in header files without
2952 being concerned about a possible macro of the same name. For example,
2953 you may use @code{__aligned__} instead of @code{aligned}.
2955 @xref{Attribute Syntax}, for details of the exact syntax for using
2959 @cindex @code{aligned} attribute
2960 @item aligned (@var{alignment})
2961 This attribute specifies a minimum alignment for the variable or
2962 structure field, measured in bytes. For example, the declaration:
2965 int x __attribute__ ((aligned (16))) = 0;
2969 causes the compiler to allocate the global variable @code{x} on a
2970 16-byte boundary. On a 68040, this could be used in conjunction with
2971 an @code{asm} expression to access the @code{move16} instruction which
2972 requires 16-byte aligned operands.
2974 You can also specify the alignment of structure fields. For example, to
2975 create a double-word aligned @code{int} pair, you could write:
2978 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
2982 This is an alternative to creating a union with a @code{double} member
2983 that forces the union to be double-word aligned.
2985 As in the preceding examples, you can explicitly specify the alignment
2986 (in bytes) that you wish the compiler to use for a given variable or
2987 structure field. Alternatively, you can leave out the alignment factor
2988 and just ask the compiler to align a variable or field to the maximum
2989 useful alignment for the target machine you are compiling for. For
2990 example, you could write:
2993 short array[3] __attribute__ ((aligned));
2996 Whenever you leave out the alignment factor in an @code{aligned} attribute
2997 specification, the compiler automatically sets the alignment for the declared
2998 variable or field to the largest alignment which is ever used for any data
2999 type on the target machine you are compiling for. Doing this can often make
3000 copy operations more efficient, because the compiler can use whatever
3001 instructions copy the biggest chunks of memory when performing copies to
3002 or from the variables or fields that you have aligned this way.
3004 The @code{aligned} attribute can only increase the alignment; but you
3005 can decrease it by specifying @code{packed} as well. See below.
3007 Note that the effectiveness of @code{aligned} attributes may be limited
3008 by inherent limitations in your linker. On many systems, the linker is
3009 only able to arrange for variables to be aligned up to a certain maximum
3010 alignment. (For some linkers, the maximum supported alignment may
3011 be very very small.) If your linker is only able to align variables
3012 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3013 in an @code{__attribute__} will still only provide you with 8 byte
3014 alignment. See your linker documentation for further information.
3016 @item cleanup (@var{cleanup_function})
3017 @cindex @code{cleanup} attribute
3018 The @code{cleanup} attribute runs a function when the variable goes
3019 out of scope. This attribute can only be applied to auto function
3020 scope variables; it may not be applied to parameters or variables
3021 with static storage duration. The function must take one parameter,
3022 a pointer to a type compatible with the variable. The return value
3023 of the function (if any) is ignored.
3025 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3026 will be run during the stack unwinding that happens during the
3027 processing of the exception. Note that the @code{cleanup} attribute
3028 does not allow the exception to be caught, only to perform an action.
3029 It is undefined what happens if @var{cleanup_function} does not
3034 @cindex @code{common} attribute
3035 @cindex @code{nocommon} attribute
3038 The @code{common} attribute requests GCC to place a variable in
3039 ``common'' storage. The @code{nocommon} attribute requests the
3040 opposite---to allocate space for it directly.
3042 These attributes override the default chosen by the
3043 @option{-fno-common} and @option{-fcommon} flags respectively.
3046 @cindex @code{deprecated} attribute
3047 The @code{deprecated} attribute results in a warning if the variable
3048 is used anywhere in the source file. This is useful when identifying
3049 variables that are expected to be removed in a future version of a
3050 program. The warning also includes the location of the declaration
3051 of the deprecated variable, to enable users to easily find further
3052 information about why the variable is deprecated, or what they should
3053 do instead. Note that the warning only occurs for uses:
3056 extern int old_var __attribute__ ((deprecated));
3058 int new_fn () @{ return old_var; @}
3061 results in a warning on line 3 but not line 2.
3063 The @code{deprecated} attribute can also be used for functions and
3064 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3066 @item mode (@var{mode})
3067 @cindex @code{mode} attribute
3068 This attribute specifies the data type for the declaration---whichever
3069 type corresponds to the mode @var{mode}. This in effect lets you
3070 request an integer or floating point type according to its width.
3072 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3073 indicate the mode corresponding to a one-byte integer, @samp{word} or
3074 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3075 or @samp{__pointer__} for the mode used to represent pointers.
3078 @cindex @code{packed} attribute
3079 The @code{packed} attribute specifies that a variable or structure field
3080 should have the smallest possible alignment---one byte for a variable,
3081 and one bit for a field, unless you specify a larger value with the
3082 @code{aligned} attribute.
3084 Here is a structure in which the field @code{x} is packed, so that it
3085 immediately follows @code{a}:
3091 int x[2] __attribute__ ((packed));
3095 @item section ("@var{section-name}")
3096 @cindex @code{section} variable attribute
3097 Normally, the compiler places the objects it generates in sections like
3098 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3099 or you need certain particular variables to appear in special sections,
3100 for example to map to special hardware. The @code{section}
3101 attribute specifies that a variable (or function) lives in a particular
3102 section. For example, this small program uses several specific section names:
3105 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3106 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3107 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3108 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3112 /* @r{Initialize stack pointer} */
3113 init_sp (stack + sizeof (stack));
3115 /* @r{Initialize initialized data} */
3116 memcpy (&init_data, &data, &edata - &data);
3118 /* @r{Turn on the serial ports} */
3125 Use the @code{section} attribute with an @emph{initialized} definition
3126 of a @emph{global} variable, as shown in the example. GCC issues
3127 a warning and otherwise ignores the @code{section} attribute in
3128 uninitialized variable declarations.
3130 You may only use the @code{section} attribute with a fully initialized
3131 global definition because of the way linkers work. The linker requires
3132 each object be defined once, with the exception that uninitialized
3133 variables tentatively go in the @code{common} (or @code{bss}) section
3134 and can be multiply ``defined''. You can force a variable to be
3135 initialized with the @option{-fno-common} flag or the @code{nocommon}
3138 Some file formats do not support arbitrary sections so the @code{section}
3139 attribute is not available on all platforms.
3140 If you need to map the entire contents of a module to a particular
3141 section, consider using the facilities of the linker instead.
3144 @cindex @code{shared} variable attribute
3145 On Microsoft Windows, in addition to putting variable definitions in a named
3146 section, the section can also be shared among all running copies of an
3147 executable or DLL@. For example, this small program defines shared data
3148 by putting it in a named section @code{shared} and marking the section
3152 int foo __attribute__((section ("shared"), shared)) = 0;
3157 /* @r{Read and write foo. All running
3158 copies see the same value.} */
3164 You may only use the @code{shared} attribute along with @code{section}
3165 attribute with a fully initialized global definition because of the way
3166 linkers work. See @code{section} attribute for more information.
3168 The @code{shared} attribute is only available on Microsoft Windows@.
3170 @item tls_model ("@var{tls_model}")
3171 @cindex @code{tls_model} attribute
3172 The @code{tls_model} attribute sets thread-local storage model
3173 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3174 overriding @option{-ftls-model=} command line switch on a per-variable
3176 The @var{tls_model} argument should be one of @code{global-dynamic},
3177 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3179 Not all targets support this attribute.
3182 This attribute, attached to a variable, means that the variable is meant
3183 to be possibly unused. GCC will not produce a warning for this
3187 This attribute, attached to a variable, means that the variable must be
3188 emitted even if it appears that the variable is not referenced.
3190 @item vector_size (@var{bytes})
3191 This attribute specifies the vector size for the variable, measured in
3192 bytes. For example, the declaration:
3195 int foo __attribute__ ((vector_size (16)));
3199 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3200 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3201 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3203 This attribute is only applicable to integral and float scalars,
3204 although arrays, pointers, and function return values are allowed in
3205 conjunction with this construct.
3207 Aggregates with this attribute are invalid, even if they are of the same
3208 size as a corresponding scalar. For example, the declaration:
3211 struct S @{ int a; @};
3212 struct S __attribute__ ((vector_size (16))) foo;
3216 is invalid even if the size of the structure is the same as the size of
3220 The @code{selectany} attribute causes an initialized global variable to
3221 have link-once semantics. When multiple definitions of the variable are
3222 encountered by the linker, the first is selected and the remainder are
3223 discarded. Following usage by the Microsoft compiler, the linker is told
3224 @emph{not} to warn about size or content differences of the multiple
3227 Although the primary usage of this attribute is for POD types, the
3228 attribute can also be applied to global C++ objects that are initialized
3229 by a constructor. In this case, the static initialization and destruction
3230 code for the object is emitted in each translation defining the object,
3231 but the calls to the constructor and destructor are protected by a
3232 link-once guard variable.
3234 The @code{selectany} attribute is only available on Microsoft Windows
3235 targets. You can use @code{__declspec (selectany)} as a synonym for
3236 @code{__attribute__ ((selectany))} for compatibility with other
3240 The @code{weak} attribute is described in @xref{Function Attributes}.
3243 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3246 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3250 @subsection M32R/D Variable Attributes
3252 One attribute is currently defined for the M32R/D@.
3255 @item model (@var{model-name})
3256 @cindex variable addressability on the M32R/D
3257 Use this attribute on the M32R/D to set the addressability of an object.
3258 The identifier @var{model-name} is one of @code{small}, @code{medium},
3259 or @code{large}, representing each of the code models.
3261 Small model objects live in the lower 16MB of memory (so that their
3262 addresses can be loaded with the @code{ld24} instruction).
3264 Medium and large model objects may live anywhere in the 32-bit address space
3265 (the compiler will generate @code{seth/add3} instructions to load their
3269 @anchor{i386 Variable Attributes}
3270 @subsection i386 Variable Attributes
3272 Two attributes are currently defined for i386 configurations:
3273 @code{ms_struct} and @code{gcc_struct}
3278 @cindex @code{ms_struct} attribute
3279 @cindex @code{gcc_struct} attribute
3281 If @code{packed} is used on a structure, or if bit-fields are used
3282 it may be that the Microsoft ABI packs them differently
3283 than GCC would normally pack them. Particularly when moving packed
3284 data between functions compiled with GCC and the native Microsoft compiler
3285 (either via function call or as data in a file), it may be necessary to access
3288 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3289 compilers to match the native Microsoft compiler.
3291 The Microsoft structure layout algorithm is fairly simple with the exception
3292 of the bitfield packing:
3294 The padding and alignment of members of structures and whether a bit field
3295 can straddle a storage-unit boundary
3298 @item Structure members are stored sequentially in the order in which they are
3299 declared: the first member has the lowest memory address and the last member
3302 @item Every data object has an alignment-requirement. The alignment-requirement
3303 for all data except structures, unions, and arrays is either the size of the
3304 object or the current packing size (specified with either the aligned attribute
3305 or the pack pragma), whichever is less. For structures, unions, and arrays,
3306 the alignment-requirement is the largest alignment-requirement of its members.
3307 Every object is allocated an offset so that:
3309 offset % alignment-requirement == 0
3311 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3312 unit if the integral types are the same size and if the next bit field fits
3313 into the current allocation unit without crossing the boundary imposed by the
3314 common alignment requirements of the bit fields.
3317 Handling of zero-length bitfields:
3319 MSVC interprets zero-length bitfields in the following ways:
3322 @item If a zero-length bitfield is inserted between two bitfields that would
3323 normally be coalesced, the bitfields will not be coalesced.
3330 unsigned long bf_1 : 12;
3332 unsigned long bf_2 : 12;
3336 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3337 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3339 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3340 alignment of the zero-length bitfield is greater than the member that follows it,
3341 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3361 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3362 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3363 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3366 Taking this into account, it is important to note the following:
3369 @item If a zero-length bitfield follows a normal bitfield, the type of the
3370 zero-length bitfield may affect the alignment of the structure as whole. For
3371 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3372 normal bitfield, and is of type short.
3374 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3375 still affect the alignment of the structure:
3385 Here, @code{t4} will take up 4 bytes.
3388 @item Zero-length bitfields following non-bitfield members are ignored:
3399 Here, @code{t5} will take up 2 bytes.
3403 @subsection PowerPC Variable Attributes
3405 Three attributes currently are defined for PowerPC configurations:
3406 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3408 For full documentation of the struct attributes please see the
3409 documentation in the @xref{i386 Variable Attributes}, section.
3411 For documentation of @code{altivec} attribute please see the
3412 documentation in the @xref{PowerPC Type Attributes}, section.
3414 @subsection SPU Variable Attributes
3416 The SPU supports the @code{spu_vector} attribute for variables. For
3417 documentation of this attribute please see the documentation in the
3418 @xref{SPU Type Attributes}, section.
3420 @subsection Xstormy16 Variable Attributes
3422 One attribute is currently defined for xstormy16 configurations:
3427 @cindex @code{below100} attribute
3429 If a variable has the @code{below100} attribute (@code{BELOW100} is
3430 allowed also), GCC will place the variable in the first 0x100 bytes of
3431 memory and use special opcodes to access it. Such variables will be
3432 placed in either the @code{.bss_below100} section or the
3433 @code{.data_below100} section.
3437 @node Type Attributes
3438 @section Specifying Attributes of Types
3439 @cindex attribute of types
3440 @cindex type attributes
3442 The keyword @code{__attribute__} allows you to specify special
3443 attributes of @code{struct} and @code{union} types when you define
3444 such types. This keyword is followed by an attribute specification
3445 inside double parentheses. Seven attributes are currently defined for
3446 types: @code{aligned}, @code{packed}, @code{transparent_union},
3447 @code{unused}, @code{deprecated}, @code{visibility}, and
3448 @code{may_alias}. Other attributes are defined for functions
3449 (@pxref{Function Attributes}) and for variables (@pxref{Variable
3452 You may also specify any one of these attributes with @samp{__}
3453 preceding and following its keyword. This allows you to use these
3454 attributes in header files without being concerned about a possible
3455 macro of the same name. For example, you may use @code{__aligned__}
3456 instead of @code{aligned}.
3458 You may specify type attributes either in a @code{typedef} declaration
3459 or in an enum, struct or union type declaration or definition.
3461 For an enum, struct or union type, you may specify attributes either
3462 between the enum, struct or union tag and the name of the type, or
3463 just past the closing curly brace of the @emph{definition}. The
3464 former syntax is preferred.
3466 @xref{Attribute Syntax}, for details of the exact syntax for using
3470 @cindex @code{aligned} attribute
3471 @item aligned (@var{alignment})
3472 This attribute specifies a minimum alignment (in bytes) for variables
3473 of the specified type. For example, the declarations:
3476 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3477 typedef int more_aligned_int __attribute__ ((aligned (8)));
3481 force the compiler to insure (as far as it can) that each variable whose
3482 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3483 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3484 variables of type @code{struct S} aligned to 8-byte boundaries allows
3485 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3486 store) instructions when copying one variable of type @code{struct S} to
3487 another, thus improving run-time efficiency.
3489 Note that the alignment of any given @code{struct} or @code{union} type
3490 is required by the ISO C standard to be at least a perfect multiple of
3491 the lowest common multiple of the alignments of all of the members of
3492 the @code{struct} or @code{union} in question. This means that you @emph{can}
3493 effectively adjust the alignment of a @code{struct} or @code{union}
3494 type by attaching an @code{aligned} attribute to any one of the members
3495 of such a type, but the notation illustrated in the example above is a
3496 more obvious, intuitive, and readable way to request the compiler to
3497 adjust the alignment of an entire @code{struct} or @code{union} type.
3499 As in the preceding example, you can explicitly specify the alignment
3500 (in bytes) that you wish the compiler to use for a given @code{struct}
3501 or @code{union} type. Alternatively, you can leave out the alignment factor
3502 and just ask the compiler to align a type to the maximum
3503 useful alignment for the target machine you are compiling for. For
3504 example, you could write:
3507 struct S @{ short f[3]; @} __attribute__ ((aligned));
3510 Whenever you leave out the alignment factor in an @code{aligned}
3511 attribute specification, the compiler automatically sets the alignment
3512 for the type to the largest alignment which is ever used for any data
3513 type on the target machine you are compiling for. Doing this can often
3514 make copy operations more efficient, because the compiler can use
3515 whatever instructions copy the biggest chunks of memory when performing
3516 copies to or from the variables which have types that you have aligned
3519 In the example above, if the size of each @code{short} is 2 bytes, then
3520 the size of the entire @code{struct S} type is 6 bytes. The smallest
3521 power of two which is greater than or equal to that is 8, so the
3522 compiler sets the alignment for the entire @code{struct S} type to 8
3525 Note that although you can ask the compiler to select a time-efficient
3526 alignment for a given type and then declare only individual stand-alone
3527 objects of that type, the compiler's ability to select a time-efficient
3528 alignment is primarily useful only when you plan to create arrays of
3529 variables having the relevant (efficiently aligned) type. If you
3530 declare or use arrays of variables of an efficiently-aligned type, then
3531 it is likely that your program will also be doing pointer arithmetic (or
3532 subscripting, which amounts to the same thing) on pointers to the
3533 relevant type, and the code that the compiler generates for these
3534 pointer arithmetic operations will often be more efficient for
3535 efficiently-aligned types than for other types.
3537 The @code{aligned} attribute can only increase the alignment; but you
3538 can decrease it by specifying @code{packed} as well. See below.
3540 Note that the effectiveness of @code{aligned} attributes may be limited
3541 by inherent limitations in your linker. On many systems, the linker is
3542 only able to arrange for variables to be aligned up to a certain maximum
3543 alignment. (For some linkers, the maximum supported alignment may
3544 be very very small.) If your linker is only able to align variables
3545 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3546 in an @code{__attribute__} will still only provide you with 8 byte
3547 alignment. See your linker documentation for further information.
3550 This attribute, attached to @code{struct} or @code{union} type
3551 definition, specifies that each member (other than zero-width bitfields)
3552 of the structure or union is placed to minimize the memory required. When
3553 attached to an @code{enum} definition, it indicates that the smallest
3554 integral type should be used.
3556 @opindex fshort-enums
3557 Specifying this attribute for @code{struct} and @code{union} types is
3558 equivalent to specifying the @code{packed} attribute on each of the
3559 structure or union members. Specifying the @option{-fshort-enums}
3560 flag on the line is equivalent to specifying the @code{packed}
3561 attribute on all @code{enum} definitions.
3563 In the following example @code{struct my_packed_struct}'s members are
3564 packed closely together, but the internal layout of its @code{s} member
3565 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3569 struct my_unpacked_struct
3575 struct __attribute__ ((__packed__)) my_packed_struct
3579 struct my_unpacked_struct s;
3583 You may only specify this attribute on the definition of a @code{enum},
3584 @code{struct} or @code{union}, not on a @code{typedef} which does not
3585 also define the enumerated type, structure or union.
3587 @item transparent_union
3588 This attribute, attached to a @code{union} type definition, indicates
3589 that any function parameter having that union type causes calls to that
3590 function to be treated in a special way.
3592 First, the argument corresponding to a transparent union type can be of
3593 any type in the union; no cast is required. Also, if the union contains
3594 a pointer type, the corresponding argument can be a null pointer
3595 constant or a void pointer expression; and if the union contains a void
3596 pointer type, the corresponding argument can be any pointer expression.
3597 If the union member type is a pointer, qualifiers like @code{const} on
3598 the referenced type must be respected, just as with normal pointer
3601 Second, the argument is passed to the function using the calling
3602 conventions of the first member of the transparent union, not the calling
3603 conventions of the union itself. All members of the union must have the
3604 same machine representation; this is necessary for this argument passing
3607 Transparent unions are designed for library functions that have multiple
3608 interfaces for compatibility reasons. For example, suppose the
3609 @code{wait} function must accept either a value of type @code{int *} to
3610 comply with Posix, or a value of type @code{union wait *} to comply with
3611 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3612 @code{wait} would accept both kinds of arguments, but it would also
3613 accept any other pointer type and this would make argument type checking
3614 less useful. Instead, @code{<sys/wait.h>} might define the interface
3622 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3624 pid_t wait (wait_status_ptr_t);
3627 This interface allows either @code{int *} or @code{union wait *}
3628 arguments to be passed, using the @code{int *} calling convention.
3629 The program can call @code{wait} with arguments of either type:
3632 int w1 () @{ int w; return wait (&w); @}
3633 int w2 () @{ union wait w; return wait (&w); @}
3636 With this interface, @code{wait}'s implementation might look like this:
3639 pid_t wait (wait_status_ptr_t p)
3641 return waitpid (-1, p.__ip, 0);
3646 When attached to a type (including a @code{union} or a @code{struct}),
3647 this attribute means that variables of that type are meant to appear
3648 possibly unused. GCC will not produce a warning for any variables of
3649 that type, even if the variable appears to do nothing. This is often
3650 the case with lock or thread classes, which are usually defined and then
3651 not referenced, but contain constructors and destructors that have
3652 nontrivial bookkeeping functions.
3655 The @code{deprecated} attribute results in a warning if the type
3656 is used anywhere in the source file. This is useful when identifying
3657 types that are expected to be removed in a future version of a program.
3658 If possible, the warning also includes the location of the declaration
3659 of the deprecated type, to enable users to easily find further
3660 information about why the type is deprecated, or what they should do
3661 instead. Note that the warnings only occur for uses and then only
3662 if the type is being applied to an identifier that itself is not being
3663 declared as deprecated.
3666 typedef int T1 __attribute__ ((deprecated));
3670 typedef T1 T3 __attribute__ ((deprecated));
3671 T3 z __attribute__ ((deprecated));
3674 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3675 warning is issued for line 4 because T2 is not explicitly
3676 deprecated. Line 5 has no warning because T3 is explicitly
3677 deprecated. Similarly for line 6.
3679 The @code{deprecated} attribute can also be used for functions and
3680 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3683 Accesses to objects with types with this attribute are not subjected to
3684 type-based alias analysis, but are instead assumed to be able to alias
3685 any other type of objects, just like the @code{char} type. See
3686 @option{-fstrict-aliasing} for more information on aliasing issues.
3691 typedef short __attribute__((__may_alias__)) short_a;
3697 short_a *b = (short_a *) &a;
3701 if (a == 0x12345678)
3708 If you replaced @code{short_a} with @code{short} in the variable
3709 declaration, the above program would abort when compiled with
3710 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3711 above in recent GCC versions.
3714 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3715 applied to class, struct, union and enum types. Unlike other type
3716 attributes, the attribute must appear between the initial keyword and
3717 the name of the type; it cannot appear after the body of the type.
3719 Note that the type visibility is applied to vague linkage entities
3720 associated with the class (vtable, typeinfo node, etc.). In
3721 particular, if a class is thrown as an exception in one shared object
3722 and caught in another, the class must have default visibility.
3723 Otherwise the two shared objects will be unable to use the same
3724 typeinfo node and exception handling will break.
3726 @subsection ARM Type Attributes
3728 On those ARM targets that support @code{dllimport} (such as Symbian
3729 OS), you can use the @code{notshared} attribute to indicate that the
3730 virtual table and other similar data for a class should not be
3731 exported from a DLL@. For example:
3734 class __declspec(notshared) C @{
3736 __declspec(dllimport) C();
3740 __declspec(dllexport)
3744 In this code, @code{C::C} is exported from the current DLL, but the
3745 virtual table for @code{C} is not exported. (You can use
3746 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3747 most Symbian OS code uses @code{__declspec}.)
3749 @anchor{i386 Type Attributes}
3750 @subsection i386 Type Attributes
3752 Two attributes are currently defined for i386 configurations:
3753 @code{ms_struct} and @code{gcc_struct}
3757 @cindex @code{ms_struct}
3758 @cindex @code{gcc_struct}
3760 If @code{packed} is used on a structure, or if bit-fields are used
3761 it may be that the Microsoft ABI packs them differently
3762 than GCC would normally pack them. Particularly when moving packed
3763 data between functions compiled with GCC and the native Microsoft compiler
3764 (either via function call or as data in a file), it may be necessary to access
3767 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3768 compilers to match the native Microsoft compiler.
3771 To specify multiple attributes, separate them by commas within the
3772 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3775 @anchor{PowerPC Type Attributes}
3776 @subsection PowerPC Type Attributes
3778 Three attributes currently are defined for PowerPC configurations:
3779 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3781 For full documentation of the struct attributes please see the
3782 documentation in the @xref{i386 Type Attributes}, section.
3784 The @code{altivec} attribute allows one to declare AltiVec vector data
3785 types supported by the AltiVec Programming Interface Manual. The
3786 attribute requires an argument to specify one of three vector types:
3787 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3788 and @code{bool__} (always followed by unsigned).
3791 __attribute__((altivec(vector__)))
3792 __attribute__((altivec(pixel__))) unsigned short
3793 __attribute__((altivec(bool__))) unsigned
3796 These attributes mainly are intended to support the @code{__vector},
3797 @code{__pixel}, and @code{__bool} AltiVec keywords.
3799 @anchor{SPU Type Attributes}
3800 @subsection SPU Type Attributes
3802 The SPU supports the @code{spu_vector} attribute for types. This attribute
3803 allows one to declare vector data types supported by the Sony/Toshiba/IBM SPU
3804 Language Extensions Specification. It is intended to support the
3805 @code{__vector} keyword.
3809 @section An Inline Function is As Fast As a Macro
3810 @cindex inline functions
3811 @cindex integrating function code
3813 @cindex macros, inline alternative
3815 By declaring a function inline, you can direct GCC to make
3816 calls to that function faster. One way GCC can achieve this is to
3817 integrate that function's code into the code for its callers. This
3818 makes execution faster by eliminating the function-call overhead; in
3819 addition, if any of the actual argument values are constant, their
3820 known values may permit simplifications at compile time so that not
3821 all of the inline function's code needs to be included. The effect on
3822 code size is less predictable; object code may be larger or smaller
3823 with function inlining, depending on the particular case. You can
3824 also direct GCC to try to integrate all ``simple enough'' functions
3825 into their callers with the option @option{-finline-functions}.
3827 GCC implements three different semantics of declaring a function
3828 inline. One is available with @option{-std=gnu89} or when @code{gnu_inline}
3829 attribute is present on all inline declarations, another when
3830 @option{-std=c99} or @option{-std=gnu99}, and the third is used when
3833 To declare a function inline, use the @code{inline} keyword in its
3834 declaration, like this:
3844 If you are writing a header file to be included in ISO C89 programs, write
3845 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3847 The three types of inlining behave similarly in two important cases:
3848 when the @code{inline} keyword is used on a @code{static} function,
3849 like the example above, and when a function is first declared without
3850 using the @code{inline} keyword and then is defined with
3851 @code{inline}, like this:
3854 extern int inc (int *a);
3862 In both of these common cases, the program behaves the same as if you
3863 had not used the @code{inline} keyword, except for its speed.
3865 @cindex inline functions, omission of
3866 @opindex fkeep-inline-functions
3867 When a function is both inline and @code{static}, if all calls to the
3868 function are integrated into the caller, and the function's address is
3869 never used, then the function's own assembler code is never referenced.
3870 In this case, GCC does not actually output assembler code for the
3871 function, unless you specify the option @option{-fkeep-inline-functions}.
3872 Some calls cannot be integrated for various reasons (in particular,
3873 calls that precede the function's definition cannot be integrated, and
3874 neither can recursive calls within the definition). If there is a
3875 nonintegrated call, then the function is compiled to assembler code as
3876 usual. The function must also be compiled as usual if the program
3877 refers to its address, because that can't be inlined.
3880 Note that certain usages in a function definition can make it unsuitable
3881 for inline substitution. Among these usages are: use of varargs, use of
3882 alloca, use of variable sized data types (@pxref{Variable Length}),
3883 use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3884 and nested functions (@pxref{Nested Functions}). Using @option{-Winline}
3885 will warn when a function marked @code{inline} could not be substituted,
3886 and will give the reason for the failure.
3888 @cindex automatic @code{inline} for C++ member fns
3889 @cindex @code{inline} automatic for C++ member fns
3890 @cindex member fns, automatically @code{inline}
3891 @cindex C++ member fns, automatically @code{inline}
3892 @opindex fno-default-inline
3893 As required by ISO C++, GCC considers member functions defined within
3894 the body of a class to be marked inline even if they are
3895 not explicitly declared with the @code{inline} keyword. You can
3896 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
3897 Options,,Options Controlling C++ Dialect}.
3899 GCC does not inline any functions when not optimizing unless you specify
3900 the @samp{always_inline} attribute for the function, like this:
3903 /* @r{Prototype.} */
3904 inline void foo (const char) __attribute__((always_inline));
3907 The remainder of this section is specific to GNU C89 inlining.
3909 @cindex non-static inline function
3910 When an inline function is not @code{static}, then the compiler must assume
3911 that there may be calls from other source files; since a global symbol can
3912 be defined only once in any program, the function must not be defined in
3913 the other source files, so the calls therein cannot be integrated.
3914 Therefore, a non-@code{static} inline function is always compiled on its
3915 own in the usual fashion.
3917 If you specify both @code{inline} and @code{extern} in the function
3918 definition, then the definition is used only for inlining. In no case
3919 is the function compiled on its own, not even if you refer to its
3920 address explicitly. Such an address becomes an external reference, as
3921 if you had only declared the function, and had not defined it.
3923 This combination of @code{inline} and @code{extern} has almost the
3924 effect of a macro. The way to use it is to put a function definition in
3925 a header file with these keywords, and put another copy of the
3926 definition (lacking @code{inline} and @code{extern}) in a library file.
3927 The definition in the header file will cause most calls to the function
3928 to be inlined. If any uses of the function remain, they will refer to
3929 the single copy in the library.
3932 @section Assembler Instructions with C Expression Operands
3933 @cindex extended @code{asm}
3934 @cindex @code{asm} expressions
3935 @cindex assembler instructions
3938 In an assembler instruction using @code{asm}, you can specify the
3939 operands of the instruction using C expressions. This means you need not
3940 guess which registers or memory locations will contain the data you want
3943 You must specify an assembler instruction template much like what
3944 appears in a machine description, plus an operand constraint string for
3947 For example, here is how to use the 68881's @code{fsinx} instruction:
3950 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3954 Here @code{angle} is the C expression for the input operand while
3955 @code{result} is that of the output operand. Each has @samp{"f"} as its
3956 operand constraint, saying that a floating point register is required.
3957 The @samp{=} in @samp{=f} indicates that the operand is an output; all
3958 output operands' constraints must use @samp{=}. The constraints use the
3959 same language used in the machine description (@pxref{Constraints}).
3961 Each operand is described by an operand-constraint string followed by
3962 the C expression in parentheses. A colon separates the assembler
3963 template from the first output operand and another separates the last
3964 output operand from the first input, if any. Commas separate the
3965 operands within each group. The total number of operands is currently
3966 limited to 30; this limitation may be lifted in some future version of
3969 If there are no output operands but there are input operands, you must
3970 place two consecutive colons surrounding the place where the output
3973 As of GCC version 3.1, it is also possible to specify input and output
3974 operands using symbolic names which can be referenced within the
3975 assembler code. These names are specified inside square brackets
3976 preceding the constraint string, and can be referenced inside the
3977 assembler code using @code{%[@var{name}]} instead of a percentage sign
3978 followed by the operand number. Using named operands the above example
3982 asm ("fsinx %[angle],%[output]"
3983 : [output] "=f" (result)
3984 : [angle] "f" (angle));
3988 Note that the symbolic operand names have no relation whatsoever to
3989 other C identifiers. You may use any name you like, even those of
3990 existing C symbols, but you must ensure that no two operands within the same
3991 assembler construct use the same symbolic name.
3993 Output operand expressions must be lvalues; the compiler can check this.
3994 The input operands need not be lvalues. The compiler cannot check
3995 whether the operands have data types that are reasonable for the
3996 instruction being executed. It does not parse the assembler instruction
3997 template and does not know what it means or even whether it is valid
3998 assembler input. The extended @code{asm} feature is most often used for
3999 machine instructions the compiler itself does not know exist. If
4000 the output expression cannot be directly addressed (for example, it is a
4001 bit-field), your constraint must allow a register. In that case, GCC
4002 will use the register as the output of the @code{asm}, and then store
4003 that register into the output.
4005 The ordinary output operands must be write-only; GCC will assume that
4006 the values in these operands before the instruction are dead and need
4007 not be generated. Extended asm supports input-output or read-write
4008 operands. Use the constraint character @samp{+} to indicate such an
4009 operand and list it with the output operands. You should only use
4010 read-write operands when the constraints for the operand (or the
4011 operand in which only some of the bits are to be changed) allow a
4014 You may, as an alternative, logically split its function into two
4015 separate operands, one input operand and one write-only output
4016 operand. The connection between them is expressed by constraints
4017 which say they need to be in the same location when the instruction
4018 executes. You can use the same C expression for both operands, or
4019 different expressions. For example, here we write the (fictitious)
4020 @samp{combine} instruction with @code{bar} as its read-only source
4021 operand and @code{foo} as its read-write destination:
4024 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4028 The constraint @samp{"0"} for operand 1 says that it must occupy the
4029 same location as operand 0. A number in constraint is allowed only in
4030 an input operand and it must refer to an output operand.
4032 Only a number in the constraint can guarantee that one operand will be in
4033 the same place as another. The mere fact that @code{foo} is the value
4034 of both operands is not enough to guarantee that they will be in the
4035 same place in the generated assembler code. The following would not
4039 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4042 Various optimizations or reloading could cause operands 0 and 1 to be in
4043 different registers; GCC knows no reason not to do so. For example, the
4044 compiler might find a copy of the value of @code{foo} in one register and
4045 use it for operand 1, but generate the output operand 0 in a different
4046 register (copying it afterward to @code{foo}'s own address). Of course,
4047 since the register for operand 1 is not even mentioned in the assembler
4048 code, the result will not work, but GCC can't tell that.
4050 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4051 the operand number for a matching constraint. For example:
4054 asm ("cmoveq %1,%2,%[result]"
4055 : [result] "=r"(result)
4056 : "r" (test), "r"(new), "[result]"(old));
4059 Sometimes you need to make an @code{asm} operand be a specific register,
4060 but there's no matching constraint letter for that register @emph{by
4061 itself}. To force the operand into that register, use a local variable
4062 for the operand and specify the register in the variable declaration.
4063 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4064 register constraint letter that matches the register:
4067 register int *p1 asm ("r0") = @dots{};
4068 register int *p2 asm ("r1") = @dots{};
4069 register int *result asm ("r0");
4070 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4073 @anchor{Example of asm with clobbered asm reg}
4074 In the above example, beware that a register that is call-clobbered by
4075 the target ABI will be overwritten by any function call in the
4076 assignment, including library calls for arithmetic operators.
4077 Assuming it is a call-clobbered register, this may happen to @code{r0}
4078 above by the assignment to @code{p2}. If you have to use such a
4079 register, use temporary variables for expressions between the register
4084 register int *p1 asm ("r0") = @dots{};
4085 register int *p2 asm ("r1") = t1;
4086 register int *result asm ("r0");
4087 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4090 Some instructions clobber specific hard registers. To describe this,
4091 write a third colon after the input operands, followed by the names of
4092 the clobbered hard registers (given as strings). Here is a realistic
4093 example for the VAX:
4096 asm volatile ("movc3 %0,%1,%2"
4097 : /* @r{no outputs} */
4098 : "g" (from), "g" (to), "g" (count)
4099 : "r0", "r1", "r2", "r3", "r4", "r5");
4102 You may not write a clobber description in a way that overlaps with an
4103 input or output operand. For example, you may not have an operand
4104 describing a register class with one member if you mention that register
4105 in the clobber list. Variables declared to live in specific registers
4106 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4107 have no part mentioned in the clobber description.
4108 There is no way for you to specify that an input
4109 operand is modified without also specifying it as an output
4110 operand. Note that if all the output operands you specify are for this
4111 purpose (and hence unused), you will then also need to specify
4112 @code{volatile} for the @code{asm} construct, as described below, to
4113 prevent GCC from deleting the @code{asm} statement as unused.
4115 If you refer to a particular hardware register from the assembler code,
4116 you will probably have to list the register after the third colon to
4117 tell the compiler the register's value is modified. In some assemblers,
4118 the register names begin with @samp{%}; to produce one @samp{%} in the
4119 assembler code, you must write @samp{%%} in the input.
4121 If your assembler instruction can alter the condition code register, add
4122 @samp{cc} to the list of clobbered registers. GCC on some machines
4123 represents the condition codes as a specific hardware register;
4124 @samp{cc} serves to name this register. On other machines, the
4125 condition code is handled differently, and specifying @samp{cc} has no
4126 effect. But it is valid no matter what the machine.
4128 If your assembler instructions access memory in an unpredictable
4129 fashion, add @samp{memory} to the list of clobbered registers. This
4130 will cause GCC to not keep memory values cached in registers across the
4131 assembler instruction and not optimize stores or loads to that memory.
4132 You will also want to add the @code{volatile} keyword if the memory
4133 affected is not listed in the inputs or outputs of the @code{asm}, as
4134 the @samp{memory} clobber does not count as a side-effect of the
4135 @code{asm}. If you know how large the accessed memory is, you can add
4136 it as input or output but if this is not known, you should add
4137 @samp{memory}. As an example, if you access ten bytes of a string, you
4138 can use a memory input like:
4141 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4144 Note that in the following example the memory input is necessary,
4145 otherwise GCC might optimize the store to @code{x} away:
4152 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4153 "=&d" (r) : "a" (y), "m" (*y));
4158 You can put multiple assembler instructions together in a single
4159 @code{asm} template, separated by the characters normally used in assembly
4160 code for the system. A combination that works in most places is a newline
4161 to break the line, plus a tab character to move to the instruction field
4162 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4163 assembler allows semicolons as a line-breaking character. Note that some
4164 assembler dialects use semicolons to start a comment.
4165 The input operands are guaranteed not to use any of the clobbered
4166 registers, and neither will the output operands' addresses, so you can
4167 read and write the clobbered registers as many times as you like. Here
4168 is an example of multiple instructions in a template; it assumes the
4169 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4172 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4174 : "g" (from), "g" (to)
4178 Unless an output operand has the @samp{&} constraint modifier, GCC
4179 may allocate it in the same register as an unrelated input operand, on
4180 the assumption the inputs are consumed before the outputs are produced.
4181 This assumption may be false if the assembler code actually consists of
4182 more than one instruction. In such a case, use @samp{&} for each output
4183 operand that may not overlap an input. @xref{Modifiers}.
4185 If you want to test the condition code produced by an assembler
4186 instruction, you must include a branch and a label in the @code{asm}
4187 construct, as follows:
4190 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4196 This assumes your assembler supports local labels, as the GNU assembler
4197 and most Unix assemblers do.
4199 Speaking of labels, jumps from one @code{asm} to another are not
4200 supported. The compiler's optimizers do not know about these jumps, and
4201 therefore they cannot take account of them when deciding how to
4204 @cindex macros containing @code{asm}
4205 Usually the most convenient way to use these @code{asm} instructions is to
4206 encapsulate them in macros that look like functions. For example,
4210 (@{ double __value, __arg = (x); \
4211 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4216 Here the variable @code{__arg} is used to make sure that the instruction
4217 operates on a proper @code{double} value, and to accept only those
4218 arguments @code{x} which can convert automatically to a @code{double}.
4220 Another way to make sure the instruction operates on the correct data
4221 type is to use a cast in the @code{asm}. This is different from using a
4222 variable @code{__arg} in that it converts more different types. For
4223 example, if the desired type were @code{int}, casting the argument to
4224 @code{int} would accept a pointer with no complaint, while assigning the
4225 argument to an @code{int} variable named @code{__arg} would warn about
4226 using a pointer unless the caller explicitly casts it.
4228 If an @code{asm} has output operands, GCC assumes for optimization
4229 purposes the instruction has no side effects except to change the output
4230 operands. This does not mean instructions with a side effect cannot be
4231 used, but you must be careful, because the compiler may eliminate them
4232 if the output operands aren't used, or move them out of loops, or
4233 replace two with one if they constitute a common subexpression. Also,
4234 if your instruction does have a side effect on a variable that otherwise
4235 appears not to change, the old value of the variable may be reused later
4236 if it happens to be found in a register.
4238 You can prevent an @code{asm} instruction from being deleted
4239 by writing the keyword @code{volatile} after
4240 the @code{asm}. For example:
4243 #define get_and_set_priority(new) \
4245 asm volatile ("get_and_set_priority %0, %1" \
4246 : "=g" (__old) : "g" (new)); \
4251 The @code{volatile} keyword indicates that the instruction has
4252 important side-effects. GCC will not delete a volatile @code{asm} if
4253 it is reachable. (The instruction can still be deleted if GCC can
4254 prove that control-flow will never reach the location of the
4255 instruction.) Note that even a volatile @code{asm} instruction
4256 can be moved relative to other code, including across jump
4257 instructions. For example, on many targets there is a system
4258 register which can be set to control the rounding mode of
4259 floating point operations. You might try
4260 setting it with a volatile @code{asm}, like this PowerPC example:
4263 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4268 This will not work reliably, as the compiler may move the addition back
4269 before the volatile @code{asm}. To make it work you need to add an
4270 artificial dependency to the @code{asm} referencing a variable in the code
4271 you don't want moved, for example:
4274 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4278 Similarly, you can't expect a
4279 sequence of volatile @code{asm} instructions to remain perfectly
4280 consecutive. If you want consecutive output, use a single @code{asm}.
4281 Also, GCC will perform some optimizations across a volatile @code{asm}
4282 instruction; GCC does not ``forget everything'' when it encounters
4283 a volatile @code{asm} instruction the way some other compilers do.
4285 An @code{asm} instruction without any output operands will be treated
4286 identically to a volatile @code{asm} instruction.
4288 It is a natural idea to look for a way to give access to the condition
4289 code left by the assembler instruction. However, when we attempted to
4290 implement this, we found no way to make it work reliably. The problem
4291 is that output operands might need reloading, which would result in
4292 additional following ``store'' instructions. On most machines, these
4293 instructions would alter the condition code before there was time to
4294 test it. This problem doesn't arise for ordinary ``test'' and
4295 ``compare'' instructions because they don't have any output operands.
4297 For reasons similar to those described above, it is not possible to give
4298 an assembler instruction access to the condition code left by previous
4301 If you are writing a header file that should be includable in ISO C
4302 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4305 @subsection Size of an @code{asm}
4307 Some targets require that GCC track the size of each instruction used in
4308 order to generate correct code. Because the final length of an
4309 @code{asm} is only known by the assembler, GCC must make an estimate as
4310 to how big it will be. The estimate is formed by counting the number of
4311 statements in the pattern of the @code{asm} and multiplying that by the
4312 length of the longest instruction on that processor. Statements in the
4313 @code{asm} are identified by newline characters and whatever statement
4314 separator characters are supported by the assembler; on most processors
4315 this is the `@code{;}' character.
4317 Normally, GCC's estimate is perfectly adequate to ensure that correct
4318 code is generated, but it is possible to confuse the compiler if you use
4319 pseudo instructions or assembler macros that expand into multiple real
4320 instructions or if you use assembler directives that expand to more
4321 space in the object file than would be needed for a single instruction.
4322 If this happens then the assembler will produce a diagnostic saying that
4323 a label is unreachable.
4325 @subsection i386 floating point asm operands
4327 There are several rules on the usage of stack-like regs in
4328 asm_operands insns. These rules apply only to the operands that are
4333 Given a set of input regs that die in an asm_operands, it is
4334 necessary to know which are implicitly popped by the asm, and
4335 which must be explicitly popped by gcc.
4337 An input reg that is implicitly popped by the asm must be
4338 explicitly clobbered, unless it is constrained to match an
4342 For any input reg that is implicitly popped by an asm, it is
4343 necessary to know how to adjust the stack to compensate for the pop.
4344 If any non-popped input is closer to the top of the reg-stack than
4345 the implicitly popped reg, it would not be possible to know what the
4346 stack looked like---it's not clear how the rest of the stack ``slides
4349 All implicitly popped input regs must be closer to the top of
4350 the reg-stack than any input that is not implicitly popped.
4352 It is possible that if an input dies in an insn, reload might
4353 use the input reg for an output reload. Consider this example:
4356 asm ("foo" : "=t" (a) : "f" (b));
4359 This asm says that input B is not popped by the asm, and that
4360 the asm pushes a result onto the reg-stack, i.e., the stack is one
4361 deeper after the asm than it was before. But, it is possible that
4362 reload will think that it can use the same reg for both the input and
4363 the output, if input B dies in this insn.
4365 If any input operand uses the @code{f} constraint, all output reg
4366 constraints must use the @code{&} earlyclobber.
4368 The asm above would be written as
4371 asm ("foo" : "=&t" (a) : "f" (b));
4375 Some operands need to be in particular places on the stack. All
4376 output operands fall in this category---there is no other way to
4377 know which regs the outputs appear in unless the user indicates
4378 this in the constraints.
4380 Output operands must specifically indicate which reg an output
4381 appears in after an asm. @code{=f} is not allowed: the operand
4382 constraints must select a class with a single reg.
4385 Output operands may not be ``inserted'' between existing stack regs.
4386 Since no 387 opcode uses a read/write operand, all output operands
4387 are dead before the asm_operands, and are pushed by the asm_operands.
4388 It makes no sense to push anywhere but the top of the reg-stack.
4390 Output operands must start at the top of the reg-stack: output
4391 operands may not ``skip'' a reg.
4394 Some asm statements may need extra stack space for internal
4395 calculations. This can be guaranteed by clobbering stack registers
4396 unrelated to the inputs and outputs.
4400 Here are a couple of reasonable asms to want to write. This asm
4401 takes one input, which is internally popped, and produces two outputs.
4404 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4407 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4408 and replaces them with one output. The user must code the @code{st(1)}
4409 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4412 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4418 @section Controlling Names Used in Assembler Code
4419 @cindex assembler names for identifiers
4420 @cindex names used in assembler code
4421 @cindex identifiers, names in assembler code
4423 You can specify the name to be used in the assembler code for a C
4424 function or variable by writing the @code{asm} (or @code{__asm__})
4425 keyword after the declarator as follows:
4428 int foo asm ("myfoo") = 2;
4432 This specifies that the name to be used for the variable @code{foo} in
4433 the assembler code should be @samp{myfoo} rather than the usual
4436 On systems where an underscore is normally prepended to the name of a C
4437 function or variable, this feature allows you to define names for the
4438 linker that do not start with an underscore.
4440 It does not make sense to use this feature with a non-static local
4441 variable since such variables do not have assembler names. If you are
4442 trying to put the variable in a particular register, see @ref{Explicit
4443 Reg Vars}. GCC presently accepts such code with a warning, but will
4444 probably be changed to issue an error, rather than a warning, in the
4447 You cannot use @code{asm} in this way in a function @emph{definition}; but
4448 you can get the same effect by writing a declaration for the function
4449 before its definition and putting @code{asm} there, like this:
4452 extern func () asm ("FUNC");
4459 It is up to you to make sure that the assembler names you choose do not
4460 conflict with any other assembler symbols. Also, you must not use a
4461 register name; that would produce completely invalid assembler code. GCC
4462 does not as yet have the ability to store static variables in registers.
4463 Perhaps that will be added.
4465 @node Explicit Reg Vars
4466 @section Variables in Specified Registers
4467 @cindex explicit register variables
4468 @cindex variables in specified registers
4469 @cindex specified registers
4470 @cindex registers, global allocation
4472 GNU C allows you to put a few global variables into specified hardware
4473 registers. You can also specify the register in which an ordinary
4474 register variable should be allocated.
4478 Global register variables reserve registers throughout the program.
4479 This may be useful in programs such as programming language
4480 interpreters which have a couple of global variables that are accessed
4484 Local register variables in specific registers do not reserve the
4485 registers, except at the point where they are used as input or output
4486 operands in an @code{asm} statement and the @code{asm} statement itself is
4487 not deleted. The compiler's data flow analysis is capable of determining
4488 where the specified registers contain live values, and where they are
4489 available for other uses. Stores into local register variables may be deleted
4490 when they appear to be dead according to dataflow analysis. References
4491 to local register variables may be deleted or moved or simplified.
4493 These local variables are sometimes convenient for use with the extended
4494 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4495 output of the assembler instruction directly into a particular register.
4496 (This will work provided the register you specify fits the constraints
4497 specified for that operand in the @code{asm}.)
4505 @node Global Reg Vars
4506 @subsection Defining Global Register Variables
4507 @cindex global register variables
4508 @cindex registers, global variables in
4510 You can define a global register variable in GNU C like this:
4513 register int *foo asm ("a5");
4517 Here @code{a5} is the name of the register which should be used. Choose a
4518 register which is normally saved and restored by function calls on your
4519 machine, so that library routines will not clobber it.
4521 Naturally the register name is cpu-dependent, so you would need to
4522 conditionalize your program according to cpu type. The register
4523 @code{a5} would be a good choice on a 68000 for a variable of pointer
4524 type. On machines with register windows, be sure to choose a ``global''
4525 register that is not affected magically by the function call mechanism.
4527 In addition, operating systems on one type of cpu may differ in how they
4528 name the registers; then you would need additional conditionals. For
4529 example, some 68000 operating systems call this register @code{%a5}.
4531 Eventually there may be a way of asking the compiler to choose a register
4532 automatically, but first we need to figure out how it should choose and
4533 how to enable you to guide the choice. No solution is evident.
4535 Defining a global register variable in a certain register reserves that
4536 register entirely for this use, at least within the current compilation.
4537 The register will not be allocated for any other purpose in the functions
4538 in the current compilation. The register will not be saved and restored by
4539 these functions. Stores into this register are never deleted even if they
4540 would appear to be dead, but references may be deleted or moved or
4543 It is not safe to access the global register variables from signal
4544 handlers, or from more than one thread of control, because the system
4545 library routines may temporarily use the register for other things (unless
4546 you recompile them specially for the task at hand).
4548 @cindex @code{qsort}, and global register variables
4549 It is not safe for one function that uses a global register variable to
4550 call another such function @code{foo} by way of a third function
4551 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4552 different source file in which the variable wasn't declared). This is
4553 because @code{lose} might save the register and put some other value there.
4554 For example, you can't expect a global register variable to be available in
4555 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4556 might have put something else in that register. (If you are prepared to
4557 recompile @code{qsort} with the same global register variable, you can
4558 solve this problem.)
4560 If you want to recompile @code{qsort} or other source files which do not
4561 actually use your global register variable, so that they will not use that
4562 register for any other purpose, then it suffices to specify the compiler
4563 option @option{-ffixed-@var{reg}}. You need not actually add a global
4564 register declaration to their source code.
4566 A function which can alter the value of a global register variable cannot
4567 safely be called from a function compiled without this variable, because it
4568 could clobber the value the caller expects to find there on return.
4569 Therefore, the function which is the entry point into the part of the
4570 program that uses the global register variable must explicitly save and
4571 restore the value which belongs to its caller.
4573 @cindex register variable after @code{longjmp}
4574 @cindex global register after @code{longjmp}
4575 @cindex value after @code{longjmp}
4578 On most machines, @code{longjmp} will restore to each global register
4579 variable the value it had at the time of the @code{setjmp}. On some
4580 machines, however, @code{longjmp} will not change the value of global
4581 register variables. To be portable, the function that called @code{setjmp}
4582 should make other arrangements to save the values of the global register
4583 variables, and to restore them in a @code{longjmp}. This way, the same
4584 thing will happen regardless of what @code{longjmp} does.
4586 All global register variable declarations must precede all function
4587 definitions. If such a declaration could appear after function
4588 definitions, the declaration would be too late to prevent the register from
4589 being used for other purposes in the preceding functions.
4591 Global register variables may not have initial values, because an
4592 executable file has no means to supply initial contents for a register.
4594 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4595 registers, but certain library functions, such as @code{getwd}, as well
4596 as the subroutines for division and remainder, modify g3 and g4. g1 and
4597 g2 are local temporaries.
4599 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4600 Of course, it will not do to use more than a few of those.
4602 @node Local Reg Vars
4603 @subsection Specifying Registers for Local Variables
4604 @cindex local variables, specifying registers
4605 @cindex specifying registers for local variables
4606 @cindex registers for local variables
4608 You can define a local register variable with a specified register
4612 register int *foo asm ("a5");
4616 Here @code{a5} is the name of the register which should be used. Note
4617 that this is the same syntax used for defining global register
4618 variables, but for a local variable it would appear within a function.
4620 Naturally the register name is cpu-dependent, but this is not a
4621 problem, since specific registers are most often useful with explicit
4622 assembler instructions (@pxref{Extended Asm}). Both of these things
4623 generally require that you conditionalize your program according to
4626 In addition, operating systems on one type of cpu may differ in how they
4627 name the registers; then you would need additional conditionals. For
4628 example, some 68000 operating systems call this register @code{%a5}.
4630 Defining such a register variable does not reserve the register; it
4631 remains available for other uses in places where flow control determines
4632 the variable's value is not live.
4634 This option does not guarantee that GCC will generate code that has
4635 this variable in the register you specify at all times. You may not
4636 code an explicit reference to this register in the @emph{assembler
4637 instruction template} part of an @code{asm} statement and assume it will
4638 always refer to this variable. However, using the variable as an
4639 @code{asm} @emph{operand} guarantees that the specified register is used
4642 Stores into local register variables may be deleted when they appear to be dead
4643 according to dataflow analysis. References to local register variables may
4644 be deleted or moved or simplified.
4646 As for global register variables, it's recommended that you choose a
4647 register which is normally saved and restored by function calls on
4648 your machine, so that library routines will not clobber it. A common
4649 pitfall is to initialize multiple call-clobbered registers with
4650 arbitrary expressions, where a function call or library call for an
4651 arithmetic operator will overwrite a register value from a previous
4652 assignment, for example @code{r0} below:
4654 register int *p1 asm ("r0") = @dots{};
4655 register int *p2 asm ("r1") = @dots{};
4657 In those cases, a solution is to use a temporary variable for
4658 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4660 @node Alternate Keywords
4661 @section Alternate Keywords
4662 @cindex alternate keywords
4663 @cindex keywords, alternate
4665 @option{-ansi} and the various @option{-std} options disable certain
4666 keywords. This causes trouble when you want to use GNU C extensions, or
4667 a general-purpose header file that should be usable by all programs,
4668 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4669 @code{inline} are not available in programs compiled with
4670 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4671 program compiled with @option{-std=c99}). The ISO C99 keyword
4672 @code{restrict} is only available when @option{-std=gnu99} (which will
4673 eventually be the default) or @option{-std=c99} (or the equivalent
4674 @option{-std=iso9899:1999}) is used.
4676 The way to solve these problems is to put @samp{__} at the beginning and
4677 end of each problematical keyword. For example, use @code{__asm__}
4678 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4680 Other C compilers won't accept these alternative keywords; if you want to
4681 compile with another compiler, you can define the alternate keywords as
4682 macros to replace them with the customary keywords. It looks like this:
4690 @findex __extension__
4692 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4694 prevent such warnings within one expression by writing
4695 @code{__extension__} before the expression. @code{__extension__} has no
4696 effect aside from this.
4698 @node Incomplete Enums
4699 @section Incomplete @code{enum} Types
4701 You can define an @code{enum} tag without specifying its possible values.
4702 This results in an incomplete type, much like what you get if you write
4703 @code{struct foo} without describing the elements. A later declaration
4704 which does specify the possible values completes the type.
4706 You can't allocate variables or storage using the type while it is
4707 incomplete. However, you can work with pointers to that type.
4709 This extension may not be very useful, but it makes the handling of
4710 @code{enum} more consistent with the way @code{struct} and @code{union}
4713 This extension is not supported by GNU C++.
4715 @node Function Names
4716 @section Function Names as Strings
4717 @cindex @code{__func__} identifier
4718 @cindex @code{__FUNCTION__} identifier
4719 @cindex @code{__PRETTY_FUNCTION__} identifier
4721 GCC provides three magic variables which hold the name of the current
4722 function, as a string. The first of these is @code{__func__}, which
4723 is part of the C99 standard:
4726 The identifier @code{__func__} is implicitly declared by the translator
4727 as if, immediately following the opening brace of each function
4728 definition, the declaration
4731 static const char __func__[] = "function-name";
4734 appeared, where function-name is the name of the lexically-enclosing
4735 function. This name is the unadorned name of the function.
4738 @code{__FUNCTION__} is another name for @code{__func__}. Older
4739 versions of GCC recognize only this name. However, it is not
4740 standardized. For maximum portability, we recommend you use
4741 @code{__func__}, but provide a fallback definition with the
4745 #if __STDC_VERSION__ < 199901L
4747 # define __func__ __FUNCTION__
4749 # define __func__ "<unknown>"
4754 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4755 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4756 the type signature of the function as well as its bare name. For
4757 example, this program:
4761 extern int printf (char *, ...);
4768 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4769 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4787 __PRETTY_FUNCTION__ = void a::sub(int)
4790 These identifiers are not preprocessor macros. In GCC 3.3 and
4791 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4792 were treated as string literals; they could be used to initialize
4793 @code{char} arrays, and they could be concatenated with other string
4794 literals. GCC 3.4 and later treat them as variables, like
4795 @code{__func__}. In C++, @code{__FUNCTION__} and
4796 @code{__PRETTY_FUNCTION__} have always been variables.
4798 @node Return Address
4799 @section Getting the Return or Frame Address of a Function
4801 These functions may be used to get information about the callers of a
4804 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4805 This function returns the return address of the current function, or of
4806 one of its callers. The @var{level} argument is number of frames to
4807 scan up the call stack. A value of @code{0} yields the return address
4808 of the current function, a value of @code{1} yields the return address
4809 of the caller of the current function, and so forth. When inlining
4810 the expected behavior is that the function will return the address of
4811 the function that will be returned to. To work around this behavior use
4812 the @code{noinline} function attribute.
4814 The @var{level} argument must be a constant integer.
4816 On some machines it may be impossible to determine the return address of
4817 any function other than the current one; in such cases, or when the top
4818 of the stack has been reached, this function will return @code{0} or a
4819 random value. In addition, @code{__builtin_frame_address} may be used
4820 to determine if the top of the stack has been reached.
4822 This function should only be used with a nonzero argument for debugging
4826 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4827 This function is similar to @code{__builtin_return_address}, but it
4828 returns the address of the function frame rather than the return address
4829 of the function. Calling @code{__builtin_frame_address} with a value of
4830 @code{0} yields the frame address of the current function, a value of
4831 @code{1} yields the frame address of the caller of the current function,
4834 The frame is the area on the stack which holds local variables and saved
4835 registers. The frame address is normally the address of the first word
4836 pushed on to the stack by the function. However, the exact definition
4837 depends upon the processor and the calling convention. If the processor
4838 has a dedicated frame pointer register, and the function has a frame,
4839 then @code{__builtin_frame_address} will return the value of the frame
4842 On some machines it may be impossible to determine the frame address of
4843 any function other than the current one; in such cases, or when the top
4844 of the stack has been reached, this function will return @code{0} if
4845 the first frame pointer is properly initialized by the startup code.
4847 This function should only be used with a nonzero argument for debugging
4851 @node Vector Extensions
4852 @section Using vector instructions through built-in functions
4854 On some targets, the instruction set contains SIMD vector instructions that
4855 operate on multiple values contained in one large register at the same time.
4856 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4859 The first step in using these extensions is to provide the necessary data
4860 types. This should be done using an appropriate @code{typedef}:
4863 typedef int v4si __attribute__ ((vector_size (16)));
4866 The @code{int} type specifies the base type, while the attribute specifies
4867 the vector size for the variable, measured in bytes. For example, the
4868 declaration above causes the compiler to set the mode for the @code{v4si}
4869 type to be 16 bytes wide and divided into @code{int} sized units. For
4870 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4871 corresponding mode of @code{foo} will be @acronym{V4SI}.
4873 The @code{vector_size} attribute is only applicable to integral and
4874 float scalars, although arrays, pointers, and function return values
4875 are allowed in conjunction with this construct.
4877 All the basic integer types can be used as base types, both as signed
4878 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4879 @code{long long}. In addition, @code{float} and @code{double} can be
4880 used to build floating-point vector types.
4882 Specifying a combination that is not valid for the current architecture
4883 will cause GCC to synthesize the instructions using a narrower mode.
4884 For example, if you specify a variable of type @code{V4SI} and your
4885 architecture does not allow for this specific SIMD type, GCC will
4886 produce code that uses 4 @code{SIs}.
4888 The types defined in this manner can be used with a subset of normal C
4889 operations. Currently, GCC will allow using the following operators
4890 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4892 The operations behave like C++ @code{valarrays}. Addition is defined as
4893 the addition of the corresponding elements of the operands. For
4894 example, in the code below, each of the 4 elements in @var{a} will be
4895 added to the corresponding 4 elements in @var{b} and the resulting
4896 vector will be stored in @var{c}.
4899 typedef int v4si __attribute__ ((vector_size (16)));
4906 Subtraction, multiplication, division, and the logical operations
4907 operate in a similar manner. Likewise, the result of using the unary
4908 minus or complement operators on a vector type is a vector whose
4909 elements are the negative or complemented values of the corresponding
4910 elements in the operand.
4912 You can declare variables and use them in function calls and returns, as
4913 well as in assignments and some casts. You can specify a vector type as
4914 a return type for a function. Vector types can also be used as function
4915 arguments. It is possible to cast from one vector type to another,
4916 provided they are of the same size (in fact, you can also cast vectors
4917 to and from other datatypes of the same size).
4919 You cannot operate between vectors of different lengths or different
4920 signedness without a cast.
4922 A port that supports hardware vector operations, usually provides a set
4923 of built-in functions that can be used to operate on vectors. For
4924 example, a function to add two vectors and multiply the result by a
4925 third could look like this:
4928 v4si f (v4si a, v4si b, v4si c)
4930 v4si tmp = __builtin_addv4si (a, b);
4931 return __builtin_mulv4si (tmp, c);
4938 @findex __builtin_offsetof
4940 GCC implements for both C and C++ a syntactic extension to implement
4941 the @code{offsetof} macro.
4945 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
4947 offsetof_member_designator:
4949 | offsetof_member_designator "." @code{identifier}
4950 | offsetof_member_designator "[" @code{expr} "]"
4953 This extension is sufficient such that
4956 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
4959 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
4960 may be dependent. In either case, @var{member} may consist of a single
4961 identifier, or a sequence of member accesses and array references.
4963 @node Atomic Builtins
4964 @section Built-in functions for atomic memory access
4966 The following builtins are intended to be compatible with those described
4967 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
4968 section 7.4. As such, they depart from the normal GCC practice of using
4969 the ``__builtin_'' prefix, and further that they are overloaded such that
4970 they work on multiple types.
4972 The definition given in the Intel documentation allows only for the use of
4973 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
4974 counterparts. GCC will allow any integral scalar or pointer type that is
4975 1, 2, 4 or 8 bytes in length.
4977 Not all operations are supported by all target processors. If a particular
4978 operation cannot be implemented on the target processor, a warning will be
4979 generated and a call an external function will be generated. The external
4980 function will carry the same name as the builtin, with an additional suffix
4981 @samp{_@var{n}} where @var{n} is the size of the data type.
4983 @c ??? Should we have a mechanism to suppress this warning? This is almost
4984 @c useful for implementing the operation under the control of an external
4987 In most cases, these builtins are considered a @dfn{full barrier}. That is,
4988 no memory operand will be moved across the operation, either forward or
4989 backward. Further, instructions will be issued as necessary to prevent the
4990 processor from speculating loads across the operation and from queuing stores
4991 after the operation.
4993 All of the routines are are described in the Intel documentation to take
4994 ``an optional list of variables protected by the memory barrier''. It's
4995 not clear what is meant by that; it could mean that @emph{only} the
4996 following variables are protected, or it could mean that these variables
4997 should in addition be protected. At present GCC ignores this list and
4998 protects all variables which are globally accessible. If in the future
4999 we make some use of this list, an empty list will continue to mean all
5000 globally accessible variables.
5003 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5004 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5005 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5006 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5007 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5008 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5009 @findex __sync_fetch_and_add
5010 @findex __sync_fetch_and_sub
5011 @findex __sync_fetch_and_or
5012 @findex __sync_fetch_and_and
5013 @findex __sync_fetch_and_xor
5014 @findex __sync_fetch_and_nand
5015 These builtins perform the operation suggested by the name, and
5016 returns the value that had previously been in memory. That is,
5019 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5020 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5023 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5024 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5025 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5026 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5027 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5028 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5029 @findex __sync_add_and_fetch
5030 @findex __sync_sub_and_fetch
5031 @findex __sync_or_and_fetch
5032 @findex __sync_and_and_fetch
5033 @findex __sync_xor_and_fetch
5034 @findex __sync_nand_and_fetch
5035 These builtins perform the operation suggested by the name, and
5036 return the new value. That is,
5039 @{ *ptr @var{op}= value; return *ptr; @}
5040 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5043 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5044 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5045 @findex __sync_bool_compare_and_swap
5046 @findex __sync_val_compare_and_swap
5047 These builtins perform an atomic compare and swap. That is, if the current
5048 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5051 The ``bool'' version returns true if the comparison is successful and
5052 @var{newval} was written. The ``val'' version returns the contents
5053 of @code{*@var{ptr}} before the operation.
5055 @item __sync_synchronize (...)
5056 @findex __sync_synchronize
5057 This builtin issues a full memory barrier.
5059 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5060 @findex __sync_lock_test_and_set
5061 This builtin, as described by Intel, is not a traditional test-and-set
5062 operation, but rather an atomic exchange operation. It writes @var{value}
5063 into @code{*@var{ptr}}, and returns the previous contents of
5066 Many targets have only minimal support for such locks, and do not support
5067 a full exchange operation. In this case, a target may support reduced
5068 functionality here by which the @emph{only} valid value to store is the
5069 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5070 is implementation defined.
5072 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5073 This means that references after the builtin cannot move to (or be
5074 speculated to) before the builtin, but previous memory stores may not
5075 be globally visible yet, and previous memory loads may not yet be
5078 @item void __sync_lock_release (@var{type} *ptr, ...)
5079 @findex __sync_lock_release
5080 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5081 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5083 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5084 This means that all previous memory stores are globally visible, and all
5085 previous memory loads have been satisfied, but following memory reads
5086 are not prevented from being speculated to before the barrier.
5089 @node Object Size Checking
5090 @section Object Size Checking Builtins
5091 @findex __builtin_object_size
5092 @findex __builtin___memcpy_chk
5093 @findex __builtin___mempcpy_chk
5094 @findex __builtin___memmove_chk
5095 @findex __builtin___memset_chk
5096 @findex __builtin___strcpy_chk
5097 @findex __builtin___stpcpy_chk
5098 @findex __builtin___strncpy_chk
5099 @findex __builtin___strcat_chk
5100 @findex __builtin___strncat_chk
5101 @findex __builtin___sprintf_chk
5102 @findex __builtin___snprintf_chk
5103 @findex __builtin___vsprintf_chk
5104 @findex __builtin___vsnprintf_chk
5105 @findex __builtin___printf_chk
5106 @findex __builtin___vprintf_chk
5107 @findex __builtin___fprintf_chk
5108 @findex __builtin___vfprintf_chk
5110 GCC implements a limited buffer overflow protection mechanism
5111 that can prevent some buffer overflow attacks.
5113 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5114 is a built-in construct that returns a constant number of bytes from
5115 @var{ptr} to the end of the object @var{ptr} pointer points to
5116 (if known at compile time). @code{__builtin_object_size} never evaluates
5117 its arguments for side-effects. If there are any side-effects in them, it
5118 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5119 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5120 point to and all of them are known at compile time, the returned number
5121 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5122 0 and minimum if nonzero. If it is not possible to determine which objects
5123 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5124 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5125 for @var{type} 2 or 3.
5127 @var{type} is an integer constant from 0 to 3. If the least significant
5128 bit is clear, objects are whole variables, if it is set, a closest
5129 surrounding subobject is considered the object a pointer points to.
5130 The second bit determines if maximum or minimum of remaining bytes
5134 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5135 char *p = &var.buf1[1], *q = &var.b;
5137 /* Here the object p points to is var. */
5138 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5139 /* The subobject p points to is var.buf1. */
5140 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5141 /* The object q points to is var. */
5142 assert (__builtin_object_size (q, 0)
5143 == (char *) (&var + 1) - (char *) &var.b);
5144 /* The subobject q points to is var.b. */
5145 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5149 There are built-in functions added for many common string operation
5150 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5151 built-in is provided. This built-in has an additional last argument,
5152 which is the number of bytes remaining in object the @var{dest}
5153 argument points to or @code{(size_t) -1} if the size is not known.
5155 The built-in functions are optimized into the normal string functions
5156 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5157 it is known at compile time that the destination object will not
5158 be overflown. If the compiler can determine at compile time the
5159 object will be always overflown, it issues a warning.
5161 The intended use can be e.g.
5165 #define bos0(dest) __builtin_object_size (dest, 0)
5166 #define memcpy(dest, src, n) \
5167 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5171 /* It is unknown what object p points to, so this is optimized
5172 into plain memcpy - no checking is possible. */
5173 memcpy (p, "abcde", n);
5174 /* Destination is known and length too. It is known at compile
5175 time there will be no overflow. */
5176 memcpy (&buf[5], "abcde", 5);
5177 /* Destination is known, but the length is not known at compile time.
5178 This will result in __memcpy_chk call that can check for overflow
5180 memcpy (&buf[5], "abcde", n);
5181 /* Destination is known and it is known at compile time there will
5182 be overflow. There will be a warning and __memcpy_chk call that
5183 will abort the program at runtime. */
5184 memcpy (&buf[6], "abcde", 5);
5187 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5188 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5189 @code{strcat} and @code{strncat}.
5191 There are also checking built-in functions for formatted output functions.
5193 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5194 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5195 const char *fmt, ...);
5196 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5198 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5199 const char *fmt, va_list ap);
5202 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5203 etc. functions and can contain implementation specific flags on what
5204 additional security measures the checking function might take, such as
5205 handling @code{%n} differently.
5207 The @var{os} argument is the object size @var{s} points to, like in the
5208 other built-in functions. There is a small difference in the behavior
5209 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5210 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5211 the checking function is called with @var{os} argument set to
5214 In addition to this, there are checking built-in functions
5215 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5216 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5217 These have just one additional argument, @var{flag}, right before
5218 format string @var{fmt}. If the compiler is able to optimize them to
5219 @code{fputc} etc. functions, it will, otherwise the checking function
5220 should be called and the @var{flag} argument passed to it.
5222 @node Other Builtins
5223 @section Other built-in functions provided by GCC
5224 @cindex built-in functions
5225 @findex __builtin_isgreater
5226 @findex __builtin_isgreaterequal
5227 @findex __builtin_isless
5228 @findex __builtin_islessequal
5229 @findex __builtin_islessgreater
5230 @findex __builtin_isunordered
5231 @findex __builtin_powi
5232 @findex __builtin_powif
5233 @findex __builtin_powil
5391 @findex fprintf_unlocked
5393 @findex fputs_unlocked
5503 @findex printf_unlocked
5532 @findex significandf
5533 @findex significandl
5604 GCC provides a large number of built-in functions other than the ones
5605 mentioned above. Some of these are for internal use in the processing
5606 of exceptions or variable-length argument lists and will not be
5607 documented here because they may change from time to time; we do not
5608 recommend general use of these functions.
5610 The remaining functions are provided for optimization purposes.
5612 @opindex fno-builtin
5613 GCC includes built-in versions of many of the functions in the standard
5614 C library. The versions prefixed with @code{__builtin_} will always be
5615 treated as having the same meaning as the C library function even if you
5616 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5617 Many of these functions are only optimized in certain cases; if they are
5618 not optimized in a particular case, a call to the library function will
5623 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5624 @option{-std=c99}), the functions
5625 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5626 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5627 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5628 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5629 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5630 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5631 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5632 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5633 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5634 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5635 @code{significandf}, @code{significandl}, @code{significand},
5636 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5637 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5638 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5639 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5640 @code{ynl} and @code{yn}
5641 may be handled as built-in functions.
5642 All these functions have corresponding versions
5643 prefixed with @code{__builtin_}, which may be used even in strict C89
5646 The ISO C99 functions
5647 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5648 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5649 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5650 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5651 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5652 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5653 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5654 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5655 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5656 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5657 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5658 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5659 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5660 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5661 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5662 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5663 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5664 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5665 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5666 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5667 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5668 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5669 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5670 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5671 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5672 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5673 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5674 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5675 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5676 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5677 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5678 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5679 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5680 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5681 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5682 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5683 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5684 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5685 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5686 are handled as built-in functions
5687 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5689 There are also built-in versions of the ISO C99 functions
5690 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5691 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5692 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5693 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5694 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5695 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5696 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5697 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5698 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5699 that are recognized in any mode since ISO C90 reserves these names for
5700 the purpose to which ISO C99 puts them. All these functions have
5701 corresponding versions prefixed with @code{__builtin_}.
5703 The ISO C94 functions
5704 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5705 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5706 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5708 are handled as built-in functions
5709 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5711 The ISO C90 functions
5712 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5713 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5714 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5715 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5716 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5717 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5718 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5719 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5720 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5721 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5722 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5723 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5724 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5725 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5726 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5727 @code{vprintf} and @code{vsprintf}
5728 are all recognized as built-in functions unless
5729 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5730 is specified for an individual function). All of these functions have
5731 corresponding versions prefixed with @code{__builtin_}.
5733 GCC provides built-in versions of the ISO C99 floating point comparison
5734 macros that avoid raising exceptions for unordered operands. They have
5735 the same names as the standard macros ( @code{isgreater},
5736 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5737 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5738 prefixed. We intend for a library implementor to be able to simply
5739 @code{#define} each standard macro to its built-in equivalent.
5741 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5743 You can use the built-in function @code{__builtin_types_compatible_p} to
5744 determine whether two types are the same.
5746 This built-in function returns 1 if the unqualified versions of the
5747 types @var{type1} and @var{type2} (which are types, not expressions) are
5748 compatible, 0 otherwise. The result of this built-in function can be
5749 used in integer constant expressions.
5751 This built-in function ignores top level qualifiers (e.g., @code{const},
5752 @code{volatile}). For example, @code{int} is equivalent to @code{const
5755 The type @code{int[]} and @code{int[5]} are compatible. On the other
5756 hand, @code{int} and @code{char *} are not compatible, even if the size
5757 of their types, on the particular architecture are the same. Also, the
5758 amount of pointer indirection is taken into account when determining
5759 similarity. Consequently, @code{short *} is not similar to
5760 @code{short **}. Furthermore, two types that are typedefed are
5761 considered compatible if their underlying types are compatible.
5763 An @code{enum} type is not considered to be compatible with another
5764 @code{enum} type even if both are compatible with the same integer
5765 type; this is what the C standard specifies.
5766 For example, @code{enum @{foo, bar@}} is not similar to
5767 @code{enum @{hot, dog@}}.
5769 You would typically use this function in code whose execution varies
5770 depending on the arguments' types. For example:
5775 typeof (x) tmp = (x); \
5776 if (__builtin_types_compatible_p (typeof (x), long double)) \
5777 tmp = foo_long_double (tmp); \
5778 else if (__builtin_types_compatible_p (typeof (x), double)) \
5779 tmp = foo_double (tmp); \
5780 else if (__builtin_types_compatible_p (typeof (x), float)) \
5781 tmp = foo_float (tmp); \
5788 @emph{Note:} This construct is only available for C@.
5792 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5794 You can use the built-in function @code{__builtin_choose_expr} to
5795 evaluate code depending on the value of a constant expression. This
5796 built-in function returns @var{exp1} if @var{const_exp}, which is a
5797 constant expression that must be able to be determined at compile time,
5798 is nonzero. Otherwise it returns 0.
5800 This built-in function is analogous to the @samp{? :} operator in C,
5801 except that the expression returned has its type unaltered by promotion
5802 rules. Also, the built-in function does not evaluate the expression
5803 that was not chosen. For example, if @var{const_exp} evaluates to true,
5804 @var{exp2} is not evaluated even if it has side-effects.
5806 This built-in function can return an lvalue if the chosen argument is an
5809 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5810 type. Similarly, if @var{exp2} is returned, its return type is the same
5817 __builtin_choose_expr ( \
5818 __builtin_types_compatible_p (typeof (x), double), \
5820 __builtin_choose_expr ( \
5821 __builtin_types_compatible_p (typeof (x), float), \
5823 /* @r{The void expression results in a compile-time error} \
5824 @r{when assigning the result to something.} */ \
5828 @emph{Note:} This construct is only available for C@. Furthermore, the
5829 unused expression (@var{exp1} or @var{exp2} depending on the value of
5830 @var{const_exp}) may still generate syntax errors. This may change in
5835 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5836 You can use the built-in function @code{__builtin_constant_p} to
5837 determine if a value is known to be constant at compile-time and hence
5838 that GCC can perform constant-folding on expressions involving that
5839 value. The argument of the function is the value to test. The function
5840 returns the integer 1 if the argument is known to be a compile-time
5841 constant and 0 if it is not known to be a compile-time constant. A
5842 return of 0 does not indicate that the value is @emph{not} a constant,
5843 but merely that GCC cannot prove it is a constant with the specified
5844 value of the @option{-O} option.
5846 You would typically use this function in an embedded application where
5847 memory was a critical resource. If you have some complex calculation,
5848 you may want it to be folded if it involves constants, but need to call
5849 a function if it does not. For example:
5852 #define Scale_Value(X) \
5853 (__builtin_constant_p (X) \
5854 ? ((X) * SCALE + OFFSET) : Scale (X))
5857 You may use this built-in function in either a macro or an inline
5858 function. However, if you use it in an inlined function and pass an
5859 argument of the function as the argument to the built-in, GCC will
5860 never return 1 when you call the inline function with a string constant
5861 or compound literal (@pxref{Compound Literals}) and will not return 1
5862 when you pass a constant numeric value to the inline function unless you
5863 specify the @option{-O} option.
5865 You may also use @code{__builtin_constant_p} in initializers for static
5866 data. For instance, you can write
5869 static const int table[] = @{
5870 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5876 This is an acceptable initializer even if @var{EXPRESSION} is not a
5877 constant expression. GCC must be more conservative about evaluating the
5878 built-in in this case, because it has no opportunity to perform
5881 Previous versions of GCC did not accept this built-in in data
5882 initializers. The earliest version where it is completely safe is
5886 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5887 @opindex fprofile-arcs
5888 You may use @code{__builtin_expect} to provide the compiler with
5889 branch prediction information. In general, you should prefer to
5890 use actual profile feedback for this (@option{-fprofile-arcs}), as
5891 programmers are notoriously bad at predicting how their programs
5892 actually perform. However, there are applications in which this
5893 data is hard to collect.
5895 The return value is the value of @var{exp}, which should be an integral
5896 expression. The semantics of the built-in are that it is expected that
5897 @var{exp} == @var{c}. For example:
5900 if (__builtin_expect (x, 0))
5905 would indicate that we do not expect to call @code{foo}, since
5906 we expect @code{x} to be zero. Since you are limited to integral
5907 expressions for @var{exp}, you should use constructions such as
5910 if (__builtin_expect (ptr != NULL, 1))
5915 when testing pointer or floating-point values.
5918 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5919 This function is used to minimize cache-miss latency by moving data into
5920 a cache before it is accessed.
5921 You can insert calls to @code{__builtin_prefetch} into code for which
5922 you know addresses of data in memory that is likely to be accessed soon.
5923 If the target supports them, data prefetch instructions will be generated.
5924 If the prefetch is done early enough before the access then the data will
5925 be in the cache by the time it is accessed.
5927 The value of @var{addr} is the address of the memory to prefetch.
5928 There are two optional arguments, @var{rw} and @var{locality}.
5929 The value of @var{rw} is a compile-time constant one or zero; one
5930 means that the prefetch is preparing for a write to the memory address
5931 and zero, the default, means that the prefetch is preparing for a read.
5932 The value @var{locality} must be a compile-time constant integer between
5933 zero and three. A value of zero means that the data has no temporal
5934 locality, so it need not be left in the cache after the access. A value
5935 of three means that the data has a high degree of temporal locality and
5936 should be left in all levels of cache possible. Values of one and two
5937 mean, respectively, a low or moderate degree of temporal locality. The
5941 for (i = 0; i < n; i++)
5944 __builtin_prefetch (&a[i+j], 1, 1);
5945 __builtin_prefetch (&b[i+j], 0, 1);
5950 Data prefetch does not generate faults if @var{addr} is invalid, but
5951 the address expression itself must be valid. For example, a prefetch
5952 of @code{p->next} will not fault if @code{p->next} is not a valid
5953 address, but evaluation will fault if @code{p} is not a valid address.
5955 If the target does not support data prefetch, the address expression
5956 is evaluated if it includes side effects but no other code is generated
5957 and GCC does not issue a warning.
5960 @deftypefn {Built-in Function} double __builtin_huge_val (void)
5961 Returns a positive infinity, if supported by the floating-point format,
5962 else @code{DBL_MAX}. This function is suitable for implementing the
5963 ISO C macro @code{HUGE_VAL}.
5966 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
5967 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5970 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5971 Similar to @code{__builtin_huge_val}, except the return
5972 type is @code{long double}.
5975 @deftypefn {Built-in Function} double __builtin_inf (void)
5976 Similar to @code{__builtin_huge_val}, except a warning is generated
5977 if the target floating-point format does not support infinities.
5980 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
5981 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
5984 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
5985 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
5988 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
5989 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
5992 @deftypefn {Built-in Function} float __builtin_inff (void)
5993 Similar to @code{__builtin_inf}, except the return type is @code{float}.
5994 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5997 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
5998 Similar to @code{__builtin_inf}, except the return
5999 type is @code{long double}.
6002 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6003 This is an implementation of the ISO C99 function @code{nan}.
6005 Since ISO C99 defines this function in terms of @code{strtod}, which we
6006 do not implement, a description of the parsing is in order. The string
6007 is parsed as by @code{strtol}; that is, the base is recognized by
6008 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6009 in the significand such that the least significant bit of the number
6010 is at the least significant bit of the significand. The number is
6011 truncated to fit the significand field provided. The significand is
6012 forced to be a quiet NaN@.
6014 This function, if given a string literal all of which would have been
6015 consumed by strtol, is evaluated early enough that it is considered a
6016 compile-time constant.
6019 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6020 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6023 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6024 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6027 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6028 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6031 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6032 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6035 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6036 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6039 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6040 Similar to @code{__builtin_nan}, except the significand is forced
6041 to be a signaling NaN@. The @code{nans} function is proposed by
6042 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6045 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6046 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6049 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6050 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6053 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6054 Returns one plus the index of the least significant 1-bit of @var{x}, or
6055 if @var{x} is zero, returns zero.
6058 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6059 Returns the number of leading 0-bits in @var{x}, starting at the most
6060 significant bit position. If @var{x} is 0, the result is undefined.
6063 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6064 Returns the number of trailing 0-bits in @var{x}, starting at the least
6065 significant bit position. If @var{x} is 0, the result is undefined.
6068 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6069 Returns the number of 1-bits in @var{x}.
6072 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6073 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6077 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6078 Similar to @code{__builtin_ffs}, except the argument type is
6079 @code{unsigned long}.
6082 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6083 Similar to @code{__builtin_clz}, except the argument type is
6084 @code{unsigned long}.
6087 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6088 Similar to @code{__builtin_ctz}, except the argument type is
6089 @code{unsigned long}.
6092 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6093 Similar to @code{__builtin_popcount}, except the argument type is
6094 @code{unsigned long}.
6097 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6098 Similar to @code{__builtin_parity}, except the argument type is
6099 @code{unsigned long}.
6102 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6103 Similar to @code{__builtin_ffs}, except the argument type is
6104 @code{unsigned long long}.
6107 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6108 Similar to @code{__builtin_clz}, except the argument type is
6109 @code{unsigned long long}.
6112 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6113 Similar to @code{__builtin_ctz}, except the argument type is
6114 @code{unsigned long long}.
6117 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6118 Similar to @code{__builtin_popcount}, except the argument type is
6119 @code{unsigned long long}.
6122 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6123 Similar to @code{__builtin_parity}, except the argument type is
6124 @code{unsigned long long}.
6127 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6128 Returns the first argument raised to the power of the second. Unlike the
6129 @code{pow} function no guarantees about precision and rounding are made.
6132 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6133 Similar to @code{__builtin_powi}, except the argument and return types
6137 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6138 Similar to @code{__builtin_powi}, except the argument and return types
6139 are @code{long double}.
6142 @deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x)
6143 Returns @var{x} with the order of the bytes reversed; for example,
6144 @code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means
6148 @deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x)
6149 Similar to @code{__builtin_bswap32}, except the argument and return types
6153 @node Target Builtins
6154 @section Built-in Functions Specific to Particular Target Machines
6156 On some target machines, GCC supports many built-in functions specific
6157 to those machines. Generally these generate calls to specific machine
6158 instructions, but allow the compiler to schedule those calls.
6161 * Alpha Built-in Functions::
6162 * ARM Built-in Functions::
6163 * Blackfin Built-in Functions::
6164 * FR-V Built-in Functions::
6165 * X86 Built-in Functions::
6166 * MIPS DSP Built-in Functions::
6167 * MIPS Paired-Single Support::
6168 * PowerPC AltiVec Built-in Functions::
6169 * SPARC VIS Built-in Functions::
6170 * SPU Built-in Functions::
6173 @node Alpha Built-in Functions
6174 @subsection Alpha Built-in Functions
6176 These built-in functions are available for the Alpha family of
6177 processors, depending on the command-line switches used.
6179 The following built-in functions are always available. They
6180 all generate the machine instruction that is part of the name.
6183 long __builtin_alpha_implver (void)
6184 long __builtin_alpha_rpcc (void)
6185 long __builtin_alpha_amask (long)
6186 long __builtin_alpha_cmpbge (long, long)
6187 long __builtin_alpha_extbl (long, long)
6188 long __builtin_alpha_extwl (long, long)
6189 long __builtin_alpha_extll (long, long)
6190 long __builtin_alpha_extql (long, long)
6191 long __builtin_alpha_extwh (long, long)
6192 long __builtin_alpha_extlh (long, long)
6193 long __builtin_alpha_extqh (long, long)
6194 long __builtin_alpha_insbl (long, long)
6195 long __builtin_alpha_inswl (long, long)
6196 long __builtin_alpha_insll (long, long)
6197 long __builtin_alpha_insql (long, long)
6198 long __builtin_alpha_inswh (long, long)
6199 long __builtin_alpha_inslh (long, long)
6200 long __builtin_alpha_insqh (long, long)
6201 long __builtin_alpha_mskbl (long, long)
6202 long __builtin_alpha_mskwl (long, long)
6203 long __builtin_alpha_mskll (long, long)
6204 long __builtin_alpha_mskql (long, long)
6205 long __builtin_alpha_mskwh (long, long)
6206 long __builtin_alpha_msklh (long, long)
6207 long __builtin_alpha_mskqh (long, long)
6208 long __builtin_alpha_umulh (long, long)
6209 long __builtin_alpha_zap (long, long)
6210 long __builtin_alpha_zapnot (long, long)
6213 The following built-in functions are always with @option{-mmax}
6214 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6215 later. They all generate the machine instruction that is part
6219 long __builtin_alpha_pklb (long)
6220 long __builtin_alpha_pkwb (long)
6221 long __builtin_alpha_unpkbl (long)
6222 long __builtin_alpha_unpkbw (long)
6223 long __builtin_alpha_minub8 (long, long)
6224 long __builtin_alpha_minsb8 (long, long)
6225 long __builtin_alpha_minuw4 (long, long)
6226 long __builtin_alpha_minsw4 (long, long)
6227 long __builtin_alpha_maxub8 (long, long)
6228 long __builtin_alpha_maxsb8 (long, long)
6229 long __builtin_alpha_maxuw4 (long, long)
6230 long __builtin_alpha_maxsw4 (long, long)
6231 long __builtin_alpha_perr (long, long)
6234 The following built-in functions are always with @option{-mcix}
6235 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6236 later. They all generate the machine instruction that is part
6240 long __builtin_alpha_cttz (long)
6241 long __builtin_alpha_ctlz (long)
6242 long __builtin_alpha_ctpop (long)
6245 The following builtins are available on systems that use the OSF/1
6246 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6247 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6248 @code{rdval} and @code{wrval}.
6251 void *__builtin_thread_pointer (void)
6252 void __builtin_set_thread_pointer (void *)
6255 @node ARM Built-in Functions
6256 @subsection ARM Built-in Functions
6258 These built-in functions are available for the ARM family of
6259 processors, when the @option{-mcpu=iwmmxt} switch is used:
6262 typedef int v2si __attribute__ ((vector_size (8)));
6263 typedef short v4hi __attribute__ ((vector_size (8)));
6264 typedef char v8qi __attribute__ ((vector_size (8)));
6266 int __builtin_arm_getwcx (int)
6267 void __builtin_arm_setwcx (int, int)
6268 int __builtin_arm_textrmsb (v8qi, int)
6269 int __builtin_arm_textrmsh (v4hi, int)
6270 int __builtin_arm_textrmsw (v2si, int)
6271 int __builtin_arm_textrmub (v8qi, int)
6272 int __builtin_arm_textrmuh (v4hi, int)
6273 int __builtin_arm_textrmuw (v2si, int)
6274 v8qi __builtin_arm_tinsrb (v8qi, int)
6275 v4hi __builtin_arm_tinsrh (v4hi, int)
6276 v2si __builtin_arm_tinsrw (v2si, int)
6277 long long __builtin_arm_tmia (long long, int, int)
6278 long long __builtin_arm_tmiabb (long long, int, int)
6279 long long __builtin_arm_tmiabt (long long, int, int)
6280 long long __builtin_arm_tmiaph (long long, int, int)
6281 long long __builtin_arm_tmiatb (long long, int, int)
6282 long long __builtin_arm_tmiatt (long long, int, int)
6283 int __builtin_arm_tmovmskb (v8qi)
6284 int __builtin_arm_tmovmskh (v4hi)
6285 int __builtin_arm_tmovmskw (v2si)
6286 long long __builtin_arm_waccb (v8qi)
6287 long long __builtin_arm_wacch (v4hi)
6288 long long __builtin_arm_waccw (v2si)
6289 v8qi __builtin_arm_waddb (v8qi, v8qi)
6290 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6291 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6292 v4hi __builtin_arm_waddh (v4hi, v4hi)
6293 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6294 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6295 v2si __builtin_arm_waddw (v2si, v2si)
6296 v2si __builtin_arm_waddwss (v2si, v2si)
6297 v2si __builtin_arm_waddwus (v2si, v2si)
6298 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6299 long long __builtin_arm_wand(long long, long long)
6300 long long __builtin_arm_wandn (long long, long long)
6301 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6302 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6303 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6304 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6305 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6306 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6307 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6308 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6309 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6310 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6311 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6312 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6313 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6314 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6315 long long __builtin_arm_wmacsz (v4hi, v4hi)
6316 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6317 long long __builtin_arm_wmacuz (v4hi, v4hi)
6318 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6319 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6320 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6321 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6322 v2si __builtin_arm_wmaxsw (v2si, v2si)
6323 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6324 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6325 v2si __builtin_arm_wmaxuw (v2si, v2si)
6326 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6327 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6328 v2si __builtin_arm_wminsw (v2si, v2si)
6329 v8qi __builtin_arm_wminub (v8qi, v8qi)
6330 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6331 v2si __builtin_arm_wminuw (v2si, v2si)
6332 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6333 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6334 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6335 long long __builtin_arm_wor (long long, long long)
6336 v2si __builtin_arm_wpackdss (long long, long long)
6337 v2si __builtin_arm_wpackdus (long long, long long)
6338 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6339 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6340 v4hi __builtin_arm_wpackwss (v2si, v2si)
6341 v4hi __builtin_arm_wpackwus (v2si, v2si)
6342 long long __builtin_arm_wrord (long long, long long)
6343 long long __builtin_arm_wrordi (long long, int)
6344 v4hi __builtin_arm_wrorh (v4hi, long long)
6345 v4hi __builtin_arm_wrorhi (v4hi, int)
6346 v2si __builtin_arm_wrorw (v2si, long long)
6347 v2si __builtin_arm_wrorwi (v2si, int)
6348 v2si __builtin_arm_wsadb (v8qi, v8qi)
6349 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6350 v2si __builtin_arm_wsadh (v4hi, v4hi)
6351 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6352 v4hi __builtin_arm_wshufh (v4hi, int)
6353 long long __builtin_arm_wslld (long long, long long)
6354 long long __builtin_arm_wslldi (long long, int)
6355 v4hi __builtin_arm_wsllh (v4hi, long long)
6356 v4hi __builtin_arm_wsllhi (v4hi, int)
6357 v2si __builtin_arm_wsllw (v2si, long long)
6358 v2si __builtin_arm_wsllwi (v2si, int)
6359 long long __builtin_arm_wsrad (long long, long long)
6360 long long __builtin_arm_wsradi (long long, int)
6361 v4hi __builtin_arm_wsrah (v4hi, long long)
6362 v4hi __builtin_arm_wsrahi (v4hi, int)
6363 v2si __builtin_arm_wsraw (v2si, long long)
6364 v2si __builtin_arm_wsrawi (v2si, int)
6365 long long __builtin_arm_wsrld (long long, long long)
6366 long long __builtin_arm_wsrldi (long long, int)
6367 v4hi __builtin_arm_wsrlh (v4hi, long long)
6368 v4hi __builtin_arm_wsrlhi (v4hi, int)
6369 v2si __builtin_arm_wsrlw (v2si, long long)
6370 v2si __builtin_arm_wsrlwi (v2si, int)
6371 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6372 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6373 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6374 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6375 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6376 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6377 v2si __builtin_arm_wsubw (v2si, v2si)
6378 v2si __builtin_arm_wsubwss (v2si, v2si)
6379 v2si __builtin_arm_wsubwus (v2si, v2si)
6380 v4hi __builtin_arm_wunpckehsb (v8qi)
6381 v2si __builtin_arm_wunpckehsh (v4hi)
6382 long long __builtin_arm_wunpckehsw (v2si)
6383 v4hi __builtin_arm_wunpckehub (v8qi)
6384 v2si __builtin_arm_wunpckehuh (v4hi)
6385 long long __builtin_arm_wunpckehuw (v2si)
6386 v4hi __builtin_arm_wunpckelsb (v8qi)
6387 v2si __builtin_arm_wunpckelsh (v4hi)
6388 long long __builtin_arm_wunpckelsw (v2si)
6389 v4hi __builtin_arm_wunpckelub (v8qi)
6390 v2si __builtin_arm_wunpckeluh (v4hi)
6391 long long __builtin_arm_wunpckeluw (v2si)
6392 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6393 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6394 v2si __builtin_arm_wunpckihw (v2si, v2si)
6395 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6396 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6397 v2si __builtin_arm_wunpckilw (v2si, v2si)
6398 long long __builtin_arm_wxor (long long, long long)
6399 long long __builtin_arm_wzero ()
6402 @node Blackfin Built-in Functions
6403 @subsection Blackfin Built-in Functions
6405 Currently, there are two Blackfin-specific built-in functions. These are
6406 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6407 using inline assembly; by using these built-in functions the compiler can
6408 automatically add workarounds for hardware errata involving these
6409 instructions. These functions are named as follows:
6412 void __builtin_bfin_csync (void)
6413 void __builtin_bfin_ssync (void)
6416 @node FR-V Built-in Functions
6417 @subsection FR-V Built-in Functions
6419 GCC provides many FR-V-specific built-in functions. In general,
6420 these functions are intended to be compatible with those described
6421 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6422 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6423 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6424 pointer rather than by value.
6426 Most of the functions are named after specific FR-V instructions.
6427 Such functions are said to be ``directly mapped'' and are summarized
6428 here in tabular form.
6432 * Directly-mapped Integer Functions::
6433 * Directly-mapped Media Functions::
6434 * Raw read/write Functions::
6435 * Other Built-in Functions::
6438 @node Argument Types
6439 @subsubsection Argument Types
6441 The arguments to the built-in functions can be divided into three groups:
6442 register numbers, compile-time constants and run-time values. In order
6443 to make this classification clear at a glance, the arguments and return
6444 values are given the following pseudo types:
6446 @multitable @columnfractions .20 .30 .15 .35
6447 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6448 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6449 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6450 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6451 @item @code{uw2} @tab @code{unsigned long long} @tab No
6452 @tab an unsigned doubleword
6453 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6454 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6455 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6456 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6459 These pseudo types are not defined by GCC, they are simply a notational
6460 convenience used in this manual.
6462 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6463 and @code{sw2} are evaluated at run time. They correspond to
6464 register operands in the underlying FR-V instructions.
6466 @code{const} arguments represent immediate operands in the underlying
6467 FR-V instructions. They must be compile-time constants.
6469 @code{acc} arguments are evaluated at compile time and specify the number
6470 of an accumulator register. For example, an @code{acc} argument of 2
6471 will select the ACC2 register.
6473 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6474 number of an IACC register. See @pxref{Other Built-in Functions}
6477 @node Directly-mapped Integer Functions
6478 @subsubsection Directly-mapped Integer Functions
6480 The functions listed below map directly to FR-V I-type instructions.
6482 @multitable @columnfractions .45 .32 .23
6483 @item Function prototype @tab Example usage @tab Assembly output
6484 @item @code{sw1 __ADDSS (sw1, sw1)}
6485 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6486 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6487 @item @code{sw1 __SCAN (sw1, sw1)}
6488 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6489 @tab @code{SCAN @var{a},@var{b},@var{c}}
6490 @item @code{sw1 __SCUTSS (sw1)}
6491 @tab @code{@var{b} = __SCUTSS (@var{a})}
6492 @tab @code{SCUTSS @var{a},@var{b}}
6493 @item @code{sw1 __SLASS (sw1, sw1)}
6494 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6495 @tab @code{SLASS @var{a},@var{b},@var{c}}
6496 @item @code{void __SMASS (sw1, sw1)}
6497 @tab @code{__SMASS (@var{a}, @var{b})}
6498 @tab @code{SMASS @var{a},@var{b}}
6499 @item @code{void __SMSSS (sw1, sw1)}
6500 @tab @code{__SMSSS (@var{a}, @var{b})}
6501 @tab @code{SMSSS @var{a},@var{b}}
6502 @item @code{void __SMU (sw1, sw1)}
6503 @tab @code{__SMU (@var{a}, @var{b})}
6504 @tab @code{SMU @var{a},@var{b}}
6505 @item @code{sw2 __SMUL (sw1, sw1)}
6506 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6507 @tab @code{SMUL @var{a},@var{b},@var{c}}
6508 @item @code{sw1 __SUBSS (sw1, sw1)}
6509 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6510 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6511 @item @code{uw2 __UMUL (uw1, uw1)}
6512 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6513 @tab @code{UMUL @var{a},@var{b},@var{c}}
6516 @node Directly-mapped Media Functions
6517 @subsubsection Directly-mapped Media Functions
6519 The functions listed below map directly to FR-V M-type instructions.
6521 @multitable @columnfractions .45 .32 .23
6522 @item Function prototype @tab Example usage @tab Assembly output
6523 @item @code{uw1 __MABSHS (sw1)}
6524 @tab @code{@var{b} = __MABSHS (@var{a})}
6525 @tab @code{MABSHS @var{a},@var{b}}
6526 @item @code{void __MADDACCS (acc, acc)}
6527 @tab @code{__MADDACCS (@var{b}, @var{a})}
6528 @tab @code{MADDACCS @var{a},@var{b}}
6529 @item @code{sw1 __MADDHSS (sw1, sw1)}
6530 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6531 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6532 @item @code{uw1 __MADDHUS (uw1, uw1)}
6533 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6534 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6535 @item @code{uw1 __MAND (uw1, uw1)}
6536 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6537 @tab @code{MAND @var{a},@var{b},@var{c}}
6538 @item @code{void __MASACCS (acc, acc)}
6539 @tab @code{__MASACCS (@var{b}, @var{a})}
6540 @tab @code{MASACCS @var{a},@var{b}}
6541 @item @code{uw1 __MAVEH (uw1, uw1)}
6542 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6543 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6544 @item @code{uw2 __MBTOH (uw1)}
6545 @tab @code{@var{b} = __MBTOH (@var{a})}
6546 @tab @code{MBTOH @var{a},@var{b}}
6547 @item @code{void __MBTOHE (uw1 *, uw1)}
6548 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6549 @tab @code{MBTOHE @var{a},@var{b}}
6550 @item @code{void __MCLRACC (acc)}
6551 @tab @code{__MCLRACC (@var{a})}
6552 @tab @code{MCLRACC @var{a}}
6553 @item @code{void __MCLRACCA (void)}
6554 @tab @code{__MCLRACCA ()}
6555 @tab @code{MCLRACCA}
6556 @item @code{uw1 __Mcop1 (uw1, uw1)}
6557 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6558 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6559 @item @code{uw1 __Mcop2 (uw1, uw1)}
6560 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6561 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6562 @item @code{uw1 __MCPLHI (uw2, const)}
6563 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6564 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6565 @item @code{uw1 __MCPLI (uw2, const)}
6566 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6567 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6568 @item @code{void __MCPXIS (acc, sw1, sw1)}
6569 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6570 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6571 @item @code{void __MCPXIU (acc, uw1, uw1)}
6572 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6573 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6574 @item @code{void __MCPXRS (acc, sw1, sw1)}
6575 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6576 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6577 @item @code{void __MCPXRU (acc, uw1, uw1)}
6578 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6579 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6580 @item @code{uw1 __MCUT (acc, uw1)}
6581 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6582 @tab @code{MCUT @var{a},@var{b},@var{c}}
6583 @item @code{uw1 __MCUTSS (acc, sw1)}
6584 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6585 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6586 @item @code{void __MDADDACCS (acc, acc)}
6587 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6588 @tab @code{MDADDACCS @var{a},@var{b}}
6589 @item @code{void __MDASACCS (acc, acc)}
6590 @tab @code{__MDASACCS (@var{b}, @var{a})}
6591 @tab @code{MDASACCS @var{a},@var{b}}
6592 @item @code{uw2 __MDCUTSSI (acc, const)}
6593 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6594 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6595 @item @code{uw2 __MDPACKH (uw2, uw2)}
6596 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6597 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6598 @item @code{uw2 __MDROTLI (uw2, const)}
6599 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6600 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6601 @item @code{void __MDSUBACCS (acc, acc)}
6602 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6603 @tab @code{MDSUBACCS @var{a},@var{b}}
6604 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6605 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6606 @tab @code{MDUNPACKH @var{a},@var{b}}
6607 @item @code{uw2 __MEXPDHD (uw1, const)}
6608 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6609 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6610 @item @code{uw1 __MEXPDHW (uw1, const)}
6611 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6612 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6613 @item @code{uw1 __MHDSETH (uw1, const)}
6614 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6615 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6616 @item @code{sw1 __MHDSETS (const)}
6617 @tab @code{@var{b} = __MHDSETS (@var{a})}
6618 @tab @code{MHDSETS #@var{a},@var{b}}
6619 @item @code{uw1 __MHSETHIH (uw1, const)}
6620 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6621 @tab @code{MHSETHIH #@var{a},@var{b}}
6622 @item @code{sw1 __MHSETHIS (sw1, const)}
6623 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6624 @tab @code{MHSETHIS #@var{a},@var{b}}
6625 @item @code{uw1 __MHSETLOH (uw1, const)}
6626 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6627 @tab @code{MHSETLOH #@var{a},@var{b}}
6628 @item @code{sw1 __MHSETLOS (sw1, const)}
6629 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6630 @tab @code{MHSETLOS #@var{a},@var{b}}
6631 @item @code{uw1 __MHTOB (uw2)}
6632 @tab @code{@var{b} = __MHTOB (@var{a})}
6633 @tab @code{MHTOB @var{a},@var{b}}
6634 @item @code{void __MMACHS (acc, sw1, sw1)}
6635 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6636 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6637 @item @code{void __MMACHU (acc, uw1, uw1)}
6638 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6639 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6640 @item @code{void __MMRDHS (acc, sw1, sw1)}
6641 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6642 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6643 @item @code{void __MMRDHU (acc, uw1, uw1)}
6644 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6645 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6646 @item @code{void __MMULHS (acc, sw1, sw1)}
6647 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6648 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6649 @item @code{void __MMULHU (acc, uw1, uw1)}
6650 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6651 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6652 @item @code{void __MMULXHS (acc, sw1, sw1)}
6653 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6654 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6655 @item @code{void __MMULXHU (acc, uw1, uw1)}
6656 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6657 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6658 @item @code{uw1 __MNOT (uw1)}
6659 @tab @code{@var{b} = __MNOT (@var{a})}
6660 @tab @code{MNOT @var{a},@var{b}}
6661 @item @code{uw1 __MOR (uw1, uw1)}
6662 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6663 @tab @code{MOR @var{a},@var{b},@var{c}}
6664 @item @code{uw1 __MPACKH (uh, uh)}
6665 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6666 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6667 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6668 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6669 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6670 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6671 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6672 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6673 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6674 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6675 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6676 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6677 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6678 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6679 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6680 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6681 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6682 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6683 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6684 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6685 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6686 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6687 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6688 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6689 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6690 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6691 @item @code{void __MQMACHS (acc, sw2, sw2)}
6692 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6693 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6694 @item @code{void __MQMACHU (acc, uw2, uw2)}
6695 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6696 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6697 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6698 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6699 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6700 @item @code{void __MQMULHS (acc, sw2, sw2)}
6701 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6702 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6703 @item @code{void __MQMULHU (acc, uw2, uw2)}
6704 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6705 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6706 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6707 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6708 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6709 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6710 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6711 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6712 @item @code{sw2 __MQSATHS (sw2, sw2)}
6713 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6714 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6715 @item @code{uw2 __MQSLLHI (uw2, int)}
6716 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6717 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6718 @item @code{sw2 __MQSRAHI (sw2, int)}
6719 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6720 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6721 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6722 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6723 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6724 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6725 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6726 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6727 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6728 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6729 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6730 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6731 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6732 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6733 @item @code{uw1 __MRDACC (acc)}
6734 @tab @code{@var{b} = __MRDACC (@var{a})}
6735 @tab @code{MRDACC @var{a},@var{b}}
6736 @item @code{uw1 __MRDACCG (acc)}
6737 @tab @code{@var{b} = __MRDACCG (@var{a})}
6738 @tab @code{MRDACCG @var{a},@var{b}}
6739 @item @code{uw1 __MROTLI (uw1, const)}
6740 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6741 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6742 @item @code{uw1 __MROTRI (uw1, const)}
6743 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6744 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6745 @item @code{sw1 __MSATHS (sw1, sw1)}
6746 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6747 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6748 @item @code{uw1 __MSATHU (uw1, uw1)}
6749 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6750 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6751 @item @code{uw1 __MSLLHI (uw1, const)}
6752 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6753 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6754 @item @code{sw1 __MSRAHI (sw1, const)}
6755 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6756 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6757 @item @code{uw1 __MSRLHI (uw1, const)}
6758 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6759 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6760 @item @code{void __MSUBACCS (acc, acc)}
6761 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6762 @tab @code{MSUBACCS @var{a},@var{b}}
6763 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6764 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6765 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6766 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6767 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6768 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6769 @item @code{void __MTRAP (void)}
6770 @tab @code{__MTRAP ()}
6772 @item @code{uw2 __MUNPACKH (uw1)}
6773 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6774 @tab @code{MUNPACKH @var{a},@var{b}}
6775 @item @code{uw1 __MWCUT (uw2, uw1)}
6776 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6777 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6778 @item @code{void __MWTACC (acc, uw1)}
6779 @tab @code{__MWTACC (@var{b}, @var{a})}
6780 @tab @code{MWTACC @var{a},@var{b}}
6781 @item @code{void __MWTACCG (acc, uw1)}
6782 @tab @code{__MWTACCG (@var{b}, @var{a})}
6783 @tab @code{MWTACCG @var{a},@var{b}}
6784 @item @code{uw1 __MXOR (uw1, uw1)}
6785 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6786 @tab @code{MXOR @var{a},@var{b},@var{c}}
6789 @node Raw read/write Functions
6790 @subsubsection Raw read/write Functions
6792 This sections describes built-in functions related to read and write
6793 instructions to access memory. These functions generate
6794 @code{membar} instructions to flush the I/O load and stores where
6795 appropriate, as described in Fujitsu's manual described above.
6799 @item unsigned char __builtin_read8 (void *@var{data})
6800 @item unsigned short __builtin_read16 (void *@var{data})
6801 @item unsigned long __builtin_read32 (void *@var{data})
6802 @item unsigned long long __builtin_read64 (void *@var{data})
6804 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6805 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6806 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6807 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6810 @node Other Built-in Functions
6811 @subsubsection Other Built-in Functions
6813 This section describes built-in functions that are not named after
6814 a specific FR-V instruction.
6817 @item sw2 __IACCreadll (iacc @var{reg})
6818 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6819 for future expansion and must be 0.
6821 @item sw1 __IACCreadl (iacc @var{reg})
6822 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6823 Other values of @var{reg} are rejected as invalid.
6825 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6826 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6827 is reserved for future expansion and must be 0.
6829 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6830 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6831 is 1. Other values of @var{reg} are rejected as invalid.
6833 @item void __data_prefetch0 (const void *@var{x})
6834 Use the @code{dcpl} instruction to load the contents of address @var{x}
6835 into the data cache.
6837 @item void __data_prefetch (const void *@var{x})
6838 Use the @code{nldub} instruction to load the contents of address @var{x}
6839 into the data cache. The instruction will be issued in slot I1@.
6842 @node X86 Built-in Functions
6843 @subsection X86 Built-in Functions
6845 These built-in functions are available for the i386 and x86-64 family
6846 of computers, depending on the command-line switches used.
6848 Note that, if you specify command-line switches such as @option{-msse},
6849 the compiler could use the extended instruction sets even if the built-ins
6850 are not used explicitly in the program. For this reason, applications
6851 which perform runtime CPU detection must compile separate files for each
6852 supported architecture, using the appropriate flags. In particular,
6853 the file containing the CPU detection code should be compiled without
6856 The following machine modes are available for use with MMX built-in functions
6857 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6858 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6859 vector of eight 8-bit integers. Some of the built-in functions operate on
6860 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6862 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6863 of two 32-bit floating point values.
6865 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6866 floating point values. Some instructions use a vector of four 32-bit
6867 integers, these use @code{V4SI}. Finally, some instructions operate on an
6868 entire vector register, interpreting it as a 128-bit integer, these use mode
6871 The following built-in functions are made available by @option{-mmmx}.
6872 All of them generate the machine instruction that is part of the name.
6875 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6876 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6877 v2si __builtin_ia32_paddd (v2si, v2si)
6878 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6879 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6880 v2si __builtin_ia32_psubd (v2si, v2si)
6881 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6882 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6883 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6884 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6885 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6886 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6887 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6888 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6889 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6890 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6891 di __builtin_ia32_pand (di, di)
6892 di __builtin_ia32_pandn (di,di)
6893 di __builtin_ia32_por (di, di)
6894 di __builtin_ia32_pxor (di, di)
6895 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6896 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6897 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6898 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6899 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6900 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6901 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6902 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
6903 v2si __builtin_ia32_punpckhdq (v2si, v2si)
6904 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
6905 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
6906 v2si __builtin_ia32_punpckldq (v2si, v2si)
6907 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
6908 v4hi __builtin_ia32_packssdw (v2si, v2si)
6909 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
6912 The following built-in functions are made available either with
6913 @option{-msse}, or with a combination of @option{-m3dnow} and
6914 @option{-march=athlon}. All of them generate the machine
6915 instruction that is part of the name.
6918 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
6919 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
6920 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
6921 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
6922 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
6923 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
6924 v8qi __builtin_ia32_pminub (v8qi, v8qi)
6925 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
6926 int __builtin_ia32_pextrw (v4hi, int)
6927 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
6928 int __builtin_ia32_pmovmskb (v8qi)
6929 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
6930 void __builtin_ia32_movntq (di *, di)
6931 void __builtin_ia32_sfence (void)
6934 The following built-in functions are available when @option{-msse} is used.
6935 All of them generate the machine instruction that is part of the name.
6938 int __builtin_ia32_comieq (v4sf, v4sf)
6939 int __builtin_ia32_comineq (v4sf, v4sf)
6940 int __builtin_ia32_comilt (v4sf, v4sf)
6941 int __builtin_ia32_comile (v4sf, v4sf)
6942 int __builtin_ia32_comigt (v4sf, v4sf)
6943 int __builtin_ia32_comige (v4sf, v4sf)
6944 int __builtin_ia32_ucomieq (v4sf, v4sf)
6945 int __builtin_ia32_ucomineq (v4sf, v4sf)
6946 int __builtin_ia32_ucomilt (v4sf, v4sf)
6947 int __builtin_ia32_ucomile (v4sf, v4sf)
6948 int __builtin_ia32_ucomigt (v4sf, v4sf)
6949 int __builtin_ia32_ucomige (v4sf, v4sf)
6950 v4sf __builtin_ia32_addps (v4sf, v4sf)
6951 v4sf __builtin_ia32_subps (v4sf, v4sf)
6952 v4sf __builtin_ia32_mulps (v4sf, v4sf)
6953 v4sf __builtin_ia32_divps (v4sf, v4sf)
6954 v4sf __builtin_ia32_addss (v4sf, v4sf)
6955 v4sf __builtin_ia32_subss (v4sf, v4sf)
6956 v4sf __builtin_ia32_mulss (v4sf, v4sf)
6957 v4sf __builtin_ia32_divss (v4sf, v4sf)
6958 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
6959 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
6960 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
6961 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
6962 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
6963 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
6964 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
6965 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
6966 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
6967 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
6968 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
6969 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
6970 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
6971 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
6972 v4si __builtin_ia32_cmpless (v4sf, v4sf)
6973 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
6974 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
6975 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6976 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6977 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6978 v4sf __builtin_ia32_maxps (v4sf, v4sf)
6979 v4sf __builtin_ia32_maxss (v4sf, v4sf)
6980 v4sf __builtin_ia32_minps (v4sf, v4sf)
6981 v4sf __builtin_ia32_minss (v4sf, v4sf)
6982 v4sf __builtin_ia32_andps (v4sf, v4sf)
6983 v4sf __builtin_ia32_andnps (v4sf, v4sf)
6984 v4sf __builtin_ia32_orps (v4sf, v4sf)
6985 v4sf __builtin_ia32_xorps (v4sf, v4sf)
6986 v4sf __builtin_ia32_movss (v4sf, v4sf)
6987 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6988 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6989 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6990 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6991 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6992 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6993 v2si __builtin_ia32_cvtps2pi (v4sf)
6994 int __builtin_ia32_cvtss2si (v4sf)
6995 v2si __builtin_ia32_cvttps2pi (v4sf)
6996 int __builtin_ia32_cvttss2si (v4sf)
6997 v4sf __builtin_ia32_rcpps (v4sf)
6998 v4sf __builtin_ia32_rsqrtps (v4sf)
6999 v4sf __builtin_ia32_sqrtps (v4sf)
7000 v4sf __builtin_ia32_rcpss (v4sf)
7001 v4sf __builtin_ia32_rsqrtss (v4sf)
7002 v4sf __builtin_ia32_sqrtss (v4sf)
7003 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7004 void __builtin_ia32_movntps (float *, v4sf)
7005 int __builtin_ia32_movmskps (v4sf)
7008 The following built-in functions are available when @option{-msse} is used.
7011 @item v4sf __builtin_ia32_loadaps (float *)
7012 Generates the @code{movaps} machine instruction as a load from memory.
7013 @item void __builtin_ia32_storeaps (float *, v4sf)
7014 Generates the @code{movaps} machine instruction as a store to memory.
7015 @item v4sf __builtin_ia32_loadups (float *)
7016 Generates the @code{movups} machine instruction as a load from memory.
7017 @item void __builtin_ia32_storeups (float *, v4sf)
7018 Generates the @code{movups} machine instruction as a store to memory.
7019 @item v4sf __builtin_ia32_loadsss (float *)
7020 Generates the @code{movss} machine instruction as a load from memory.
7021 @item void __builtin_ia32_storess (float *, v4sf)
7022 Generates the @code{movss} machine instruction as a store to memory.
7023 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7024 Generates the @code{movhps} machine instruction as a load from memory.
7025 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7026 Generates the @code{movlps} machine instruction as a load from memory
7027 @item void __builtin_ia32_storehps (v4sf, v2si *)
7028 Generates the @code{movhps} machine instruction as a store to memory.
7029 @item void __builtin_ia32_storelps (v4sf, v2si *)
7030 Generates the @code{movlps} machine instruction as a store to memory.
7033 The following built-in functions are available when @option{-msse2} is used.
7034 All of them generate the machine instruction that is part of the name.
7037 int __builtin_ia32_comisdeq (v2df, v2df)
7038 int __builtin_ia32_comisdlt (v2df, v2df)
7039 int __builtin_ia32_comisdle (v2df, v2df)
7040 int __builtin_ia32_comisdgt (v2df, v2df)
7041 int __builtin_ia32_comisdge (v2df, v2df)
7042 int __builtin_ia32_comisdneq (v2df, v2df)
7043 int __builtin_ia32_ucomisdeq (v2df, v2df)
7044 int __builtin_ia32_ucomisdlt (v2df, v2df)
7045 int __builtin_ia32_ucomisdle (v2df, v2df)
7046 int __builtin_ia32_ucomisdgt (v2df, v2df)
7047 int __builtin_ia32_ucomisdge (v2df, v2df)
7048 int __builtin_ia32_ucomisdneq (v2df, v2df)
7049 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7050 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7051 v2df __builtin_ia32_cmplepd (v2df, v2df)
7052 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7053 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7054 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7055 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7056 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7057 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7058 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7059 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7060 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7061 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7062 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7063 v2df __builtin_ia32_cmplesd (v2df, v2df)
7064 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7065 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7066 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7067 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7068 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7069 v2di __builtin_ia32_paddq (v2di, v2di)
7070 v2di __builtin_ia32_psubq (v2di, v2di)
7071 v2df __builtin_ia32_addpd (v2df, v2df)
7072 v2df __builtin_ia32_subpd (v2df, v2df)
7073 v2df __builtin_ia32_mulpd (v2df, v2df)
7074 v2df __builtin_ia32_divpd (v2df, v2df)
7075 v2df __builtin_ia32_addsd (v2df, v2df)
7076 v2df __builtin_ia32_subsd (v2df, v2df)
7077 v2df __builtin_ia32_mulsd (v2df, v2df)
7078 v2df __builtin_ia32_divsd (v2df, v2df)
7079 v2df __builtin_ia32_minpd (v2df, v2df)
7080 v2df __builtin_ia32_maxpd (v2df, v2df)
7081 v2df __builtin_ia32_minsd (v2df, v2df)
7082 v2df __builtin_ia32_maxsd (v2df, v2df)
7083 v2df __builtin_ia32_andpd (v2df, v2df)
7084 v2df __builtin_ia32_andnpd (v2df, v2df)
7085 v2df __builtin_ia32_orpd (v2df, v2df)
7086 v2df __builtin_ia32_xorpd (v2df, v2df)
7087 v2df __builtin_ia32_movsd (v2df, v2df)
7088 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7089 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7090 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7091 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7092 v4si __builtin_ia32_paddd128 (v4si, v4si)
7093 v2di __builtin_ia32_paddq128 (v2di, v2di)
7094 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7095 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7096 v4si __builtin_ia32_psubd128 (v4si, v4si)
7097 v2di __builtin_ia32_psubq128 (v2di, v2di)
7098 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7099 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7100 v2di __builtin_ia32_pand128 (v2di, v2di)
7101 v2di __builtin_ia32_pandn128 (v2di, v2di)
7102 v2di __builtin_ia32_por128 (v2di, v2di)
7103 v2di __builtin_ia32_pxor128 (v2di, v2di)
7104 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7105 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7106 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7107 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7108 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7109 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7110 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7111 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7112 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7113 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7114 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7115 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7116 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7117 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7118 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7119 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7120 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7121 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7122 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7123 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7124 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7125 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7126 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7127 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7128 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7129 v2df __builtin_ia32_loadupd (double *)
7130 void __builtin_ia32_storeupd (double *, v2df)
7131 v2df __builtin_ia32_loadhpd (v2df, double *)
7132 v2df __builtin_ia32_loadlpd (v2df, double *)
7133 int __builtin_ia32_movmskpd (v2df)
7134 int __builtin_ia32_pmovmskb128 (v16qi)
7135 void __builtin_ia32_movnti (int *, int)
7136 void __builtin_ia32_movntpd (double *, v2df)
7137 void __builtin_ia32_movntdq (v2df *, v2df)
7138 v4si __builtin_ia32_pshufd (v4si, int)
7139 v8hi __builtin_ia32_pshuflw (v8hi, int)
7140 v8hi __builtin_ia32_pshufhw (v8hi, int)
7141 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7142 v2df __builtin_ia32_sqrtpd (v2df)
7143 v2df __builtin_ia32_sqrtsd (v2df)
7144 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7145 v2df __builtin_ia32_cvtdq2pd (v4si)
7146 v4sf __builtin_ia32_cvtdq2ps (v4si)
7147 v4si __builtin_ia32_cvtpd2dq (v2df)
7148 v2si __builtin_ia32_cvtpd2pi (v2df)
7149 v4sf __builtin_ia32_cvtpd2ps (v2df)
7150 v4si __builtin_ia32_cvttpd2dq (v2df)
7151 v2si __builtin_ia32_cvttpd2pi (v2df)
7152 v2df __builtin_ia32_cvtpi2pd (v2si)
7153 int __builtin_ia32_cvtsd2si (v2df)
7154 int __builtin_ia32_cvttsd2si (v2df)
7155 long long __builtin_ia32_cvtsd2si64 (v2df)
7156 long long __builtin_ia32_cvttsd2si64 (v2df)
7157 v4si __builtin_ia32_cvtps2dq (v4sf)
7158 v2df __builtin_ia32_cvtps2pd (v4sf)
7159 v4si __builtin_ia32_cvttps2dq (v4sf)
7160 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7161 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7162 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7163 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7164 void __builtin_ia32_clflush (const void *)
7165 void __builtin_ia32_lfence (void)
7166 void __builtin_ia32_mfence (void)
7167 v16qi __builtin_ia32_loaddqu (const char *)
7168 void __builtin_ia32_storedqu (char *, v16qi)
7169 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7170 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7171 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7172 v4si __builtin_ia32_pslld128 (v4si, v2di)
7173 v2di __builtin_ia32_psllq128 (v4si, v2di)
7174 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7175 v4si __builtin_ia32_psrld128 (v4si, v2di)
7176 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7177 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7178 v4si __builtin_ia32_psrad128 (v4si, v2di)
7179 v2di __builtin_ia32_pslldqi128 (v2di, int)
7180 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7181 v4si __builtin_ia32_pslldi128 (v4si, int)
7182 v2di __builtin_ia32_psllqi128 (v2di, int)
7183 v2di __builtin_ia32_psrldqi128 (v2di, int)
7184 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7185 v4si __builtin_ia32_psrldi128 (v4si, int)
7186 v2di __builtin_ia32_psrlqi128 (v2di, int)
7187 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7188 v4si __builtin_ia32_psradi128 (v4si, int)
7189 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7192 The following built-in functions are available when @option{-msse3} is used.
7193 All of them generate the machine instruction that is part of the name.
7196 v2df __builtin_ia32_addsubpd (v2df, v2df)
7197 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7198 v2df __builtin_ia32_haddpd (v2df, v2df)
7199 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7200 v2df __builtin_ia32_hsubpd (v2df, v2df)
7201 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7202 v16qi __builtin_ia32_lddqu (char const *)
7203 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7204 v2df __builtin_ia32_movddup (v2df)
7205 v4sf __builtin_ia32_movshdup (v4sf)
7206 v4sf __builtin_ia32_movsldup (v4sf)
7207 void __builtin_ia32_mwait (unsigned int, unsigned int)
7210 The following built-in functions are available when @option{-msse3} is used.
7213 @item v2df __builtin_ia32_loadddup (double const *)
7214 Generates the @code{movddup} machine instruction as a load from memory.
7217 The following built-in functions are available when @option{-mssse3} is used.
7218 All of them generate the machine instruction that is part of the name
7222 v2si __builtin_ia32_phaddd (v2si, v2si)
7223 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7224 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7225 v2si __builtin_ia32_phsubd (v2si, v2si)
7226 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7227 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7228 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7229 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7230 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7231 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7232 v2si __builtin_ia32_psignd (v2si, v2si)
7233 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7234 long long __builtin_ia32_palignr (long long, long long, int)
7235 v8qi __builtin_ia32_pabsb (v8qi)
7236 v2si __builtin_ia32_pabsd (v2si)
7237 v4hi __builtin_ia32_pabsw (v4hi)
7240 The following built-in functions are available when @option{-mssse3} is used.
7241 All of them generate the machine instruction that is part of the name
7245 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7246 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7247 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7248 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7249 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7250 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7251 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7252 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7253 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7254 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7255 v4si __builtin_ia32_psignd128 (v4si, v4si)
7256 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7257 v2di __builtin_ia32_palignr (v2di, v2di, int)
7258 v16qi __builtin_ia32_pabsb128 (v16qi)
7259 v4si __builtin_ia32_pabsd128 (v4si)
7260 v8hi __builtin_ia32_pabsw128 (v8hi)
7263 The following built-in functions are available when @option{-m3dnow} is used.
7264 All of them generate the machine instruction that is part of the name.
7267 void __builtin_ia32_femms (void)
7268 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7269 v2si __builtin_ia32_pf2id (v2sf)
7270 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7271 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7272 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7273 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7274 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7275 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7276 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7277 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7278 v2sf __builtin_ia32_pfrcp (v2sf)
7279 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7280 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7281 v2sf __builtin_ia32_pfrsqrt (v2sf)
7282 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7283 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7284 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7285 v2sf __builtin_ia32_pi2fd (v2si)
7286 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7289 The following built-in functions are available when both @option{-m3dnow}
7290 and @option{-march=athlon} are used. All of them generate the machine
7291 instruction that is part of the name.
7294 v2si __builtin_ia32_pf2iw (v2sf)
7295 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7296 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7297 v2sf __builtin_ia32_pi2fw (v2si)
7298 v2sf __builtin_ia32_pswapdsf (v2sf)
7299 v2si __builtin_ia32_pswapdsi (v2si)
7302 @node MIPS DSP Built-in Functions
7303 @subsection MIPS DSP Built-in Functions
7305 The MIPS DSP Application-Specific Extension (ASE) includes new
7306 instructions that are designed to improve the performance of DSP and
7307 media applications. It provides instructions that operate on packed
7308 8-bit integer data, Q15 fractional data and Q31 fractional data.
7310 GCC supports MIPS DSP operations using both the generic
7311 vector extensions (@pxref{Vector Extensions}) and a collection of
7312 MIPS-specific built-in functions. Both kinds of support are
7313 enabled by the @option{-mdsp} command-line option.
7315 At present, GCC only provides support for operations on 32-bit
7316 vectors. The vector type associated with 8-bit integer data is
7317 usually called @code{v4i8} and the vector type associated with Q15 is
7318 usually called @code{v2q15}. They can be defined in C as follows:
7321 typedef char v4i8 __attribute__ ((vector_size(4)));
7322 typedef short v2q15 __attribute__ ((vector_size(4)));
7325 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7326 aggregates. For example:
7329 v4i8 a = @{1, 2, 3, 4@};
7331 b = (v4i8) @{5, 6, 7, 8@};
7333 v2q15 c = @{0x0fcb, 0x3a75@};
7335 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7338 @emph{Note:} The CPU's endianness determines the order in which values
7339 are packed. On little-endian targets, the first value is the least
7340 significant and the last value is the most significant. The opposite
7341 order applies to big-endian targets. For example, the code above will
7342 set the lowest byte of @code{a} to @code{1} on little-endian targets
7343 and @code{4} on big-endian targets.
7345 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7346 representation. As shown in this example, the integer representation
7347 of a Q15 value can be obtained by multiplying the fractional value by
7348 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7351 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7352 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7353 and @code{c} and @code{d} are @code{v2q15} values.
7355 @multitable @columnfractions .50 .50
7356 @item C code @tab MIPS instruction
7357 @item @code{a + b} @tab @code{addu.qb}
7358 @item @code{c + d} @tab @code{addq.ph}
7359 @item @code{a - b} @tab @code{subu.qb}
7360 @item @code{c - d} @tab @code{subq.ph}
7363 It is easier to describe the DSP built-in functions if we first define
7364 the following types:
7369 typedef long long a64;
7372 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7373 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7374 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7375 @code{long long}, but we use @code{a64} to indicate values that will
7376 be placed in one of the four DSP accumulators (@code{$ac0},
7377 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7379 Also, some built-in functions prefer or require immediate numbers as
7380 parameters, because the corresponding DSP instructions accept both immediate
7381 numbers and register operands, or accept immediate numbers only. The
7382 immediate parameters are listed as follows.
7390 imm_n32_31: -32 to 31.
7391 imm_n512_511: -512 to 511.
7394 The following built-in functions map directly to a particular MIPS DSP
7395 instruction. Please refer to the architecture specification
7396 for details on what each instruction does.
7399 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7400 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7401 q31 __builtin_mips_addq_s_w (q31, q31)
7402 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7403 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7404 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7405 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7406 q31 __builtin_mips_subq_s_w (q31, q31)
7407 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7408 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7409 i32 __builtin_mips_addsc (i32, i32)
7410 i32 __builtin_mips_addwc (i32, i32)
7411 i32 __builtin_mips_modsub (i32, i32)
7412 i32 __builtin_mips_raddu_w_qb (v4i8)
7413 v2q15 __builtin_mips_absq_s_ph (v2q15)
7414 q31 __builtin_mips_absq_s_w (q31)
7415 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7416 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7417 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7418 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7419 q31 __builtin_mips_preceq_w_phl (v2q15)
7420 q31 __builtin_mips_preceq_w_phr (v2q15)
7421 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7422 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7423 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7424 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7425 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7426 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7427 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7428 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7429 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7430 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7431 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7432 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7433 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7434 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7435 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7436 q31 __builtin_mips_shll_s_w (q31, i32)
7437 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7438 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7439 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7440 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7441 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7442 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7443 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7444 q31 __builtin_mips_shra_r_w (q31, i32)
7445 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7446 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7447 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7448 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7449 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7450 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7451 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7452 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7453 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7454 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7455 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7456 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7457 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7458 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7459 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7460 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7461 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7462 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7463 i32 __builtin_mips_bitrev (i32)
7464 i32 __builtin_mips_insv (i32, i32)
7465 v4i8 __builtin_mips_repl_qb (imm0_255)
7466 v4i8 __builtin_mips_repl_qb (i32)
7467 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7468 v2q15 __builtin_mips_repl_ph (i32)
7469 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7470 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7471 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7472 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7473 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7474 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7475 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7476 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7477 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7478 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7479 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7480 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7481 i32 __builtin_mips_extr_w (a64, imm0_31)
7482 i32 __builtin_mips_extr_w (a64, i32)
7483 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7484 i32 __builtin_mips_extr_s_h (a64, i32)
7485 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7486 i32 __builtin_mips_extr_rs_w (a64, i32)
7487 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7488 i32 __builtin_mips_extr_r_w (a64, i32)
7489 i32 __builtin_mips_extp (a64, imm0_31)
7490 i32 __builtin_mips_extp (a64, i32)
7491 i32 __builtin_mips_extpdp (a64, imm0_31)
7492 i32 __builtin_mips_extpdp (a64, i32)
7493 a64 __builtin_mips_shilo (a64, imm_n32_31)
7494 a64 __builtin_mips_shilo (a64, i32)
7495 a64 __builtin_mips_mthlip (a64, i32)
7496 void __builtin_mips_wrdsp (i32, imm0_63)
7497 i32 __builtin_mips_rddsp (imm0_63)
7498 i32 __builtin_mips_lbux (void *, i32)
7499 i32 __builtin_mips_lhx (void *, i32)
7500 i32 __builtin_mips_lwx (void *, i32)
7501 i32 __builtin_mips_bposge32 (void)
7504 @node MIPS Paired-Single Support
7505 @subsection MIPS Paired-Single Support
7507 The MIPS64 architecture includes a number of instructions that
7508 operate on pairs of single-precision floating-point values.
7509 Each pair is packed into a 64-bit floating-point register,
7510 with one element being designated the ``upper half'' and
7511 the other being designated the ``lower half''.
7513 GCC supports paired-single operations using both the generic
7514 vector extensions (@pxref{Vector Extensions}) and a collection of
7515 MIPS-specific built-in functions. Both kinds of support are
7516 enabled by the @option{-mpaired-single} command-line option.
7518 The vector type associated with paired-single values is usually
7519 called @code{v2sf}. It can be defined in C as follows:
7522 typedef float v2sf __attribute__ ((vector_size (8)));
7525 @code{v2sf} values are initialized in the same way as aggregates.
7529 v2sf a = @{1.5, 9.1@};
7532 b = (v2sf) @{e, f@};
7535 @emph{Note:} The CPU's endianness determines which value is stored in
7536 the upper half of a register and which value is stored in the lower half.
7537 On little-endian targets, the first value is the lower one and the second
7538 value is the upper one. The opposite order applies to big-endian targets.
7539 For example, the code above will set the lower half of @code{a} to
7540 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7543 * Paired-Single Arithmetic::
7544 * Paired-Single Built-in Functions::
7545 * MIPS-3D Built-in Functions::
7548 @node Paired-Single Arithmetic
7549 @subsubsection Paired-Single Arithmetic
7551 The table below lists the @code{v2sf} operations for which hardware
7552 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7553 values and @code{x} is an integral value.
7555 @multitable @columnfractions .50 .50
7556 @item C code @tab MIPS instruction
7557 @item @code{a + b} @tab @code{add.ps}
7558 @item @code{a - b} @tab @code{sub.ps}
7559 @item @code{-a} @tab @code{neg.ps}
7560 @item @code{a * b} @tab @code{mul.ps}
7561 @item @code{a * b + c} @tab @code{madd.ps}
7562 @item @code{a * b - c} @tab @code{msub.ps}
7563 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7564 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7565 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7568 Note that the multiply-accumulate instructions can be disabled
7569 using the command-line option @code{-mno-fused-madd}.
7571 @node Paired-Single Built-in Functions
7572 @subsubsection Paired-Single Built-in Functions
7574 The following paired-single functions map directly to a particular
7575 MIPS instruction. Please refer to the architecture specification
7576 for details on what each instruction does.
7579 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7580 Pair lower lower (@code{pll.ps}).
7582 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7583 Pair upper lower (@code{pul.ps}).
7585 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7586 Pair lower upper (@code{plu.ps}).
7588 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7589 Pair upper upper (@code{puu.ps}).
7591 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7592 Convert pair to paired single (@code{cvt.ps.s}).
7594 @item float __builtin_mips_cvt_s_pl (v2sf)
7595 Convert pair lower to single (@code{cvt.s.pl}).
7597 @item float __builtin_mips_cvt_s_pu (v2sf)
7598 Convert pair upper to single (@code{cvt.s.pu}).
7600 @item v2sf __builtin_mips_abs_ps (v2sf)
7601 Absolute value (@code{abs.ps}).
7603 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7604 Align variable (@code{alnv.ps}).
7606 @emph{Note:} The value of the third parameter must be 0 or 4
7607 modulo 8, otherwise the result will be unpredictable. Please read the
7608 instruction description for details.
7611 The following multi-instruction functions are also available.
7612 In each case, @var{cond} can be any of the 16 floating-point conditions:
7613 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7614 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7615 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7618 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7619 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7620 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7621 @code{movt.ps}/@code{movf.ps}).
7623 The @code{movt} functions return the value @var{x} computed by:
7626 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7627 mov.ps @var{x},@var{c}
7628 movt.ps @var{x},@var{d},@var{cc}
7631 The @code{movf} functions are similar but use @code{movf.ps} instead
7634 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7635 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7636 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7637 @code{bc1t}/@code{bc1f}).
7639 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7640 and return either the upper or lower half of the result. For example:
7644 if (__builtin_mips_upper_c_eq_ps (a, b))
7645 upper_halves_are_equal ();
7647 upper_halves_are_unequal ();
7649 if (__builtin_mips_lower_c_eq_ps (a, b))
7650 lower_halves_are_equal ();
7652 lower_halves_are_unequal ();
7656 @node MIPS-3D Built-in Functions
7657 @subsubsection MIPS-3D Built-in Functions
7659 The MIPS-3D Application-Specific Extension (ASE) includes additional
7660 paired-single instructions that are designed to improve the performance
7661 of 3D graphics operations. Support for these instructions is controlled
7662 by the @option{-mips3d} command-line option.
7664 The functions listed below map directly to a particular MIPS-3D
7665 instruction. Please refer to the architecture specification for
7666 more details on what each instruction does.
7669 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7670 Reduction add (@code{addr.ps}).
7672 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7673 Reduction multiply (@code{mulr.ps}).
7675 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7676 Convert paired single to paired word (@code{cvt.pw.ps}).
7678 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7679 Convert paired word to paired single (@code{cvt.ps.pw}).
7681 @item float __builtin_mips_recip1_s (float)
7682 @itemx double __builtin_mips_recip1_d (double)
7683 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7684 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7686 @item float __builtin_mips_recip2_s (float, float)
7687 @itemx double __builtin_mips_recip2_d (double, double)
7688 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7689 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7691 @item float __builtin_mips_rsqrt1_s (float)
7692 @itemx double __builtin_mips_rsqrt1_d (double)
7693 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7694 Reduced precision reciprocal square root (sequence step 1)
7695 (@code{rsqrt1.@var{fmt}}).
7697 @item float __builtin_mips_rsqrt2_s (float, float)
7698 @itemx double __builtin_mips_rsqrt2_d (double, double)
7699 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7700 Reduced precision reciprocal square root (sequence step 2)
7701 (@code{rsqrt2.@var{fmt}}).
7704 The following multi-instruction functions are also available.
7705 In each case, @var{cond} can be any of the 16 floating-point conditions:
7706 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7707 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7708 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7711 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7712 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7713 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7714 @code{bc1t}/@code{bc1f}).
7716 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7717 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7722 if (__builtin_mips_cabs_eq_s (a, b))
7728 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7729 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7730 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7731 @code{bc1t}/@code{bc1f}).
7733 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7734 and return either the upper or lower half of the result. For example:
7738 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7739 upper_halves_are_equal ();
7741 upper_halves_are_unequal ();
7743 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7744 lower_halves_are_equal ();
7746 lower_halves_are_unequal ();
7749 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7750 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7751 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7752 @code{movt.ps}/@code{movf.ps}).
7754 The @code{movt} functions return the value @var{x} computed by:
7757 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7758 mov.ps @var{x},@var{c}
7759 movt.ps @var{x},@var{d},@var{cc}
7762 The @code{movf} functions are similar but use @code{movf.ps} instead
7765 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7766 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7767 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7768 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7769 Comparison of two paired-single values
7770 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7771 @code{bc1any2t}/@code{bc1any2f}).
7773 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7774 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7775 result is true and the @code{all} forms return true if both results are true.
7780 if (__builtin_mips_any_c_eq_ps (a, b))
7785 if (__builtin_mips_all_c_eq_ps (a, b))
7791 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7792 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7793 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7794 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7795 Comparison of four paired-single values
7796 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7797 @code{bc1any4t}/@code{bc1any4f}).
7799 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7800 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7801 The @code{any} forms return true if any of the four results are true
7802 and the @code{all} forms return true if all four results are true.
7807 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7812 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7819 @node PowerPC AltiVec Built-in Functions
7820 @subsection PowerPC AltiVec Built-in Functions
7822 GCC provides an interface for the PowerPC family of processors to access
7823 the AltiVec operations described in Motorola's AltiVec Programming
7824 Interface Manual. The interface is made available by including
7825 @code{<altivec.h>} and using @option{-maltivec} and
7826 @option{-mabi=altivec}. The interface supports the following vector
7830 vector unsigned char
7834 vector unsigned short
7845 GCC's implementation of the high-level language interface available from
7846 C and C++ code differs from Motorola's documentation in several ways.
7851 A vector constant is a list of constant expressions within curly braces.
7854 A vector initializer requires no cast if the vector constant is of the
7855 same type as the variable it is initializing.
7858 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7859 vector type is the default signedness of the base type. The default
7860 varies depending on the operating system, so a portable program should
7861 always specify the signedness.
7864 Compiling with @option{-maltivec} adds keywords @code{__vector},
7865 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7866 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7870 GCC allows using a @code{typedef} name as the type specifier for a
7874 For C, overloaded functions are implemented with macros so the following
7878 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7881 Since @code{vec_add} is a macro, the vector constant in the example
7882 is treated as four separate arguments. Wrap the entire argument in
7883 parentheses for this to work.
7886 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
7887 Internally, GCC uses built-in functions to achieve the functionality in
7888 the aforementioned header file, but they are not supported and are
7889 subject to change without notice.
7891 The following interfaces are supported for the generic and specific
7892 AltiVec operations and the AltiVec predicates. In cases where there
7893 is a direct mapping between generic and specific operations, only the
7894 generic names are shown here, although the specific operations can also
7897 Arguments that are documented as @code{const int} require literal
7898 integral values within the range required for that operation.
7901 vector signed char vec_abs (vector signed char);
7902 vector signed short vec_abs (vector signed short);
7903 vector signed int vec_abs (vector signed int);
7904 vector float vec_abs (vector float);
7906 vector signed char vec_abss (vector signed char);
7907 vector signed short vec_abss (vector signed short);
7908 vector signed int vec_abss (vector signed int);
7910 vector signed char vec_add (vector bool char, vector signed char);
7911 vector signed char vec_add (vector signed char, vector bool char);
7912 vector signed char vec_add (vector signed char, vector signed char);
7913 vector unsigned char vec_add (vector bool char, vector unsigned char);
7914 vector unsigned char vec_add (vector unsigned char, vector bool char);
7915 vector unsigned char vec_add (vector unsigned char,
7916 vector unsigned char);
7917 vector signed short vec_add (vector bool short, vector signed short);
7918 vector signed short vec_add (vector signed short, vector bool short);
7919 vector signed short vec_add (vector signed short, vector signed short);
7920 vector unsigned short vec_add (vector bool short,
7921 vector unsigned short);
7922 vector unsigned short vec_add (vector unsigned short,
7924 vector unsigned short vec_add (vector unsigned short,
7925 vector unsigned short);
7926 vector signed int vec_add (vector bool int, vector signed int);
7927 vector signed int vec_add (vector signed int, vector bool int);
7928 vector signed int vec_add (vector signed int, vector signed int);
7929 vector unsigned int vec_add (vector bool int, vector unsigned int);
7930 vector unsigned int vec_add (vector unsigned int, vector bool int);
7931 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
7932 vector float vec_add (vector float, vector float);
7934 vector float vec_vaddfp (vector float, vector float);
7936 vector signed int vec_vadduwm (vector bool int, vector signed int);
7937 vector signed int vec_vadduwm (vector signed int, vector bool int);
7938 vector signed int vec_vadduwm (vector signed int, vector signed int);
7939 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
7940 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
7941 vector unsigned int vec_vadduwm (vector unsigned int,
7942 vector unsigned int);
7944 vector signed short vec_vadduhm (vector bool short,
7945 vector signed short);
7946 vector signed short vec_vadduhm (vector signed short,
7948 vector signed short vec_vadduhm (vector signed short,
7949 vector signed short);
7950 vector unsigned short vec_vadduhm (vector bool short,
7951 vector unsigned short);
7952 vector unsigned short vec_vadduhm (vector unsigned short,
7954 vector unsigned short vec_vadduhm (vector unsigned short,
7955 vector unsigned short);
7957 vector signed char vec_vaddubm (vector bool char, vector signed char);
7958 vector signed char vec_vaddubm (vector signed char, vector bool char);
7959 vector signed char vec_vaddubm (vector signed char, vector signed char);
7960 vector unsigned char vec_vaddubm (vector bool char,
7961 vector unsigned char);
7962 vector unsigned char vec_vaddubm (vector unsigned char,
7964 vector unsigned char vec_vaddubm (vector unsigned char,
7965 vector unsigned char);
7967 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
7969 vector unsigned char vec_adds (vector bool char, vector unsigned char);
7970 vector unsigned char vec_adds (vector unsigned char, vector bool char);
7971 vector unsigned char vec_adds (vector unsigned char,
7972 vector unsigned char);
7973 vector signed char vec_adds (vector bool char, vector signed char);
7974 vector signed char vec_adds (vector signed char, vector bool char);
7975 vector signed char vec_adds (vector signed char, vector signed char);
7976 vector unsigned short vec_adds (vector bool short,
7977 vector unsigned short);
7978 vector unsigned short vec_adds (vector unsigned short,
7980 vector unsigned short vec_adds (vector unsigned short,
7981 vector unsigned short);
7982 vector signed short vec_adds (vector bool short, vector signed short);
7983 vector signed short vec_adds (vector signed short, vector bool short);
7984 vector signed short vec_adds (vector signed short, vector signed short);
7985 vector unsigned int vec_adds (vector bool int, vector unsigned int);
7986 vector unsigned int vec_adds (vector unsigned int, vector bool int);
7987 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
7988 vector signed int vec_adds (vector bool int, vector signed int);
7989 vector signed int vec_adds (vector signed int, vector bool int);
7990 vector signed int vec_adds (vector signed int, vector signed int);
7992 vector signed int vec_vaddsws (vector bool int, vector signed int);
7993 vector signed int vec_vaddsws (vector signed int, vector bool int);
7994 vector signed int vec_vaddsws (vector signed int, vector signed int);
7996 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
7997 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
7998 vector unsigned int vec_vadduws (vector unsigned int,
7999 vector unsigned int);
8001 vector signed short vec_vaddshs (vector bool short,
8002 vector signed short);
8003 vector signed short vec_vaddshs (vector signed short,
8005 vector signed short vec_vaddshs (vector signed short,
8006 vector signed short);
8008 vector unsigned short vec_vadduhs (vector bool short,
8009 vector unsigned short);
8010 vector unsigned short vec_vadduhs (vector unsigned short,
8012 vector unsigned short vec_vadduhs (vector unsigned short,
8013 vector unsigned short);
8015 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8016 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8017 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8019 vector unsigned char vec_vaddubs (vector bool char,
8020 vector unsigned char);
8021 vector unsigned char vec_vaddubs (vector unsigned char,
8023 vector unsigned char vec_vaddubs (vector unsigned char,
8024 vector unsigned char);
8026 vector float vec_and (vector float, vector float);
8027 vector float vec_and (vector float, vector bool int);
8028 vector float vec_and (vector bool int, vector float);
8029 vector bool int vec_and (vector bool int, vector bool int);
8030 vector signed int vec_and (vector bool int, vector signed int);
8031 vector signed int vec_and (vector signed int, vector bool int);
8032 vector signed int vec_and (vector signed int, vector signed int);
8033 vector unsigned int vec_and (vector bool int, vector unsigned int);
8034 vector unsigned int vec_and (vector unsigned int, vector bool int);
8035 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8036 vector bool short vec_and (vector bool short, vector bool short);
8037 vector signed short vec_and (vector bool short, vector signed short);
8038 vector signed short vec_and (vector signed short, vector bool short);
8039 vector signed short vec_and (vector signed short, vector signed short);
8040 vector unsigned short vec_and (vector bool short,
8041 vector unsigned short);
8042 vector unsigned short vec_and (vector unsigned short,
8044 vector unsigned short vec_and (vector unsigned short,
8045 vector unsigned short);
8046 vector signed char vec_and (vector bool char, vector signed char);
8047 vector bool char vec_and (vector bool char, vector bool char);
8048 vector signed char vec_and (vector signed char, vector bool char);
8049 vector signed char vec_and (vector signed char, vector signed char);
8050 vector unsigned char vec_and (vector bool char, vector unsigned char);
8051 vector unsigned char vec_and (vector unsigned char, vector bool char);
8052 vector unsigned char vec_and (vector unsigned char,
8053 vector unsigned char);
8055 vector float vec_andc (vector float, vector float);
8056 vector float vec_andc (vector float, vector bool int);
8057 vector float vec_andc (vector bool int, vector float);
8058 vector bool int vec_andc (vector bool int, vector bool int);
8059 vector signed int vec_andc (vector bool int, vector signed int);
8060 vector signed int vec_andc (vector signed int, vector bool int);
8061 vector signed int vec_andc (vector signed int, vector signed int);
8062 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8063 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8064 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8065 vector bool short vec_andc (vector bool short, vector bool short);
8066 vector signed short vec_andc (vector bool short, vector signed short);
8067 vector signed short vec_andc (vector signed short, vector bool short);
8068 vector signed short vec_andc (vector signed short, vector signed short);
8069 vector unsigned short vec_andc (vector bool short,
8070 vector unsigned short);
8071 vector unsigned short vec_andc (vector unsigned short,
8073 vector unsigned short vec_andc (vector unsigned short,
8074 vector unsigned short);
8075 vector signed char vec_andc (vector bool char, vector signed char);
8076 vector bool char vec_andc (vector bool char, vector bool char);
8077 vector signed char vec_andc (vector signed char, vector bool char);
8078 vector signed char vec_andc (vector signed char, vector signed char);
8079 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8080 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8081 vector unsigned char vec_andc (vector unsigned char,
8082 vector unsigned char);
8084 vector unsigned char vec_avg (vector unsigned char,
8085 vector unsigned char);
8086 vector signed char vec_avg (vector signed char, vector signed char);
8087 vector unsigned short vec_avg (vector unsigned short,
8088 vector unsigned short);
8089 vector signed short vec_avg (vector signed short, vector signed short);
8090 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8091 vector signed int vec_avg (vector signed int, vector signed int);
8093 vector signed int vec_vavgsw (vector signed int, vector signed int);
8095 vector unsigned int vec_vavguw (vector unsigned int,
8096 vector unsigned int);
8098 vector signed short vec_vavgsh (vector signed short,
8099 vector signed short);
8101 vector unsigned short vec_vavguh (vector unsigned short,
8102 vector unsigned short);
8104 vector signed char vec_vavgsb (vector signed char, vector signed char);
8106 vector unsigned char vec_vavgub (vector unsigned char,
8107 vector unsigned char);
8109 vector float vec_ceil (vector float);
8111 vector signed int vec_cmpb (vector float, vector float);
8113 vector bool char vec_cmpeq (vector signed char, vector signed char);
8114 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8115 vector bool short vec_cmpeq (vector signed short, vector signed short);
8116 vector bool short vec_cmpeq (vector unsigned short,
8117 vector unsigned short);
8118 vector bool int vec_cmpeq (vector signed int, vector signed int);
8119 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8120 vector bool int vec_cmpeq (vector float, vector float);
8122 vector bool int vec_vcmpeqfp (vector float, vector float);
8124 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8125 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8127 vector bool short vec_vcmpequh (vector signed short,
8128 vector signed short);
8129 vector bool short vec_vcmpequh (vector unsigned short,
8130 vector unsigned short);
8132 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8133 vector bool char vec_vcmpequb (vector unsigned char,
8134 vector unsigned char);
8136 vector bool int vec_cmpge (vector float, vector float);
8138 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8139 vector bool char vec_cmpgt (vector signed char, vector signed char);
8140 vector bool short vec_cmpgt (vector unsigned short,
8141 vector unsigned short);
8142 vector bool short vec_cmpgt (vector signed short, vector signed short);
8143 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8144 vector bool int vec_cmpgt (vector signed int, vector signed int);
8145 vector bool int vec_cmpgt (vector float, vector float);
8147 vector bool int vec_vcmpgtfp (vector float, vector float);
8149 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8151 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8153 vector bool short vec_vcmpgtsh (vector signed short,
8154 vector signed short);
8156 vector bool short vec_vcmpgtuh (vector unsigned short,
8157 vector unsigned short);
8159 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8161 vector bool char vec_vcmpgtub (vector unsigned char,
8162 vector unsigned char);
8164 vector bool int vec_cmple (vector float, vector float);
8166 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8167 vector bool char vec_cmplt (vector signed char, vector signed char);
8168 vector bool short vec_cmplt (vector unsigned short,
8169 vector unsigned short);
8170 vector bool short vec_cmplt (vector signed short, vector signed short);
8171 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8172 vector bool int vec_cmplt (vector signed int, vector signed int);
8173 vector bool int vec_cmplt (vector float, vector float);
8175 vector float vec_ctf (vector unsigned int, const int);
8176 vector float vec_ctf (vector signed int, const int);
8178 vector float vec_vcfsx (vector signed int, const int);
8180 vector float vec_vcfux (vector unsigned int, const int);
8182 vector signed int vec_cts (vector float, const int);
8184 vector unsigned int vec_ctu (vector float, const int);
8186 void vec_dss (const int);
8188 void vec_dssall (void);
8190 void vec_dst (const vector unsigned char *, int, const int);
8191 void vec_dst (const vector signed char *, int, const int);
8192 void vec_dst (const vector bool char *, int, const int);
8193 void vec_dst (const vector unsigned short *, int, const int);
8194 void vec_dst (const vector signed short *, int, const int);
8195 void vec_dst (const vector bool short *, int, const int);
8196 void vec_dst (const vector pixel *, int, const int);
8197 void vec_dst (const vector unsigned int *, int, const int);
8198 void vec_dst (const vector signed int *, int, const int);
8199 void vec_dst (const vector bool int *, int, const int);
8200 void vec_dst (const vector float *, int, const int);
8201 void vec_dst (const unsigned char *, int, const int);
8202 void vec_dst (const signed char *, int, const int);
8203 void vec_dst (const unsigned short *, int, const int);
8204 void vec_dst (const short *, int, const int);
8205 void vec_dst (const unsigned int *, int, const int);
8206 void vec_dst (const int *, int, const int);
8207 void vec_dst (const unsigned long *, int, const int);
8208 void vec_dst (const long *, int, const int);
8209 void vec_dst (const float *, int, const int);
8211 void vec_dstst (const vector unsigned char *, int, const int);
8212 void vec_dstst (const vector signed char *, int, const int);
8213 void vec_dstst (const vector bool char *, int, const int);
8214 void vec_dstst (const vector unsigned short *, int, const int);
8215 void vec_dstst (const vector signed short *, int, const int);
8216 void vec_dstst (const vector bool short *, int, const int);
8217 void vec_dstst (const vector pixel *, int, const int);
8218 void vec_dstst (const vector unsigned int *, int, const int);
8219 void vec_dstst (const vector signed int *, int, const int);
8220 void vec_dstst (const vector bool int *, int, const int);
8221 void vec_dstst (const vector float *, int, const int);
8222 void vec_dstst (const unsigned char *, int, const int);
8223 void vec_dstst (const signed char *, int, const int);
8224 void vec_dstst (const unsigned short *, int, const int);
8225 void vec_dstst (const short *, int, const int);
8226 void vec_dstst (const unsigned int *, int, const int);
8227 void vec_dstst (const int *, int, const int);
8228 void vec_dstst (const unsigned long *, int, const int);
8229 void vec_dstst (const long *, int, const int);
8230 void vec_dstst (const float *, int, const int);
8232 void vec_dststt (const vector unsigned char *, int, const int);
8233 void vec_dststt (const vector signed char *, int, const int);
8234 void vec_dststt (const vector bool char *, int, const int);
8235 void vec_dststt (const vector unsigned short *, int, const int);
8236 void vec_dststt (const vector signed short *, int, const int);
8237 void vec_dststt (const vector bool short *, int, const int);
8238 void vec_dststt (const vector pixel *, int, const int);
8239 void vec_dststt (const vector unsigned int *, int, const int);
8240 void vec_dststt (const vector signed int *, int, const int);
8241 void vec_dststt (const vector bool int *, int, const int);
8242 void vec_dststt (const vector float *, int, const int);
8243 void vec_dststt (const unsigned char *, int, const int);
8244 void vec_dststt (const signed char *, int, const int);
8245 void vec_dststt (const unsigned short *, int, const int);
8246 void vec_dststt (const short *, int, const int);
8247 void vec_dststt (const unsigned int *, int, const int);
8248 void vec_dststt (const int *, int, const int);
8249 void vec_dststt (const unsigned long *, int, const int);
8250 void vec_dststt (const long *, int, const int);
8251 void vec_dststt (const float *, int, const int);
8253 void vec_dstt (const vector unsigned char *, int, const int);
8254 void vec_dstt (const vector signed char *, int, const int);
8255 void vec_dstt (const vector bool char *, int, const int);
8256 void vec_dstt (const vector unsigned short *, int, const int);
8257 void vec_dstt (const vector signed short *, int, const int);
8258 void vec_dstt (const vector bool short *, int, const int);
8259 void vec_dstt (const vector pixel *, int, const int);
8260 void vec_dstt (const vector unsigned int *, int, const int);
8261 void vec_dstt (const vector signed int *, int, const int);
8262 void vec_dstt (const vector bool int *, int, const int);
8263 void vec_dstt (const vector float *, int, const int);
8264 void vec_dstt (const unsigned char *, int, const int);
8265 void vec_dstt (const signed char *, int, const int);
8266 void vec_dstt (const unsigned short *, int, const int);
8267 void vec_dstt (const short *, int, const int);
8268 void vec_dstt (const unsigned int *, int, const int);
8269 void vec_dstt (const int *, int, const int);
8270 void vec_dstt (const unsigned long *, int, const int);
8271 void vec_dstt (const long *, int, const int);
8272 void vec_dstt (const float *, int, const int);
8274 vector float vec_expte (vector float);
8276 vector float vec_floor (vector float);
8278 vector float vec_ld (int, const vector float *);
8279 vector float vec_ld (int, const float *);
8280 vector bool int vec_ld (int, const vector bool int *);
8281 vector signed int vec_ld (int, const vector signed int *);
8282 vector signed int vec_ld (int, const int *);
8283 vector signed int vec_ld (int, const long *);
8284 vector unsigned int vec_ld (int, const vector unsigned int *);
8285 vector unsigned int vec_ld (int, const unsigned int *);
8286 vector unsigned int vec_ld (int, const unsigned long *);
8287 vector bool short vec_ld (int, const vector bool short *);
8288 vector pixel vec_ld (int, const vector pixel *);
8289 vector signed short vec_ld (int, const vector signed short *);
8290 vector signed short vec_ld (int, const short *);
8291 vector unsigned short vec_ld (int, const vector unsigned short *);
8292 vector unsigned short vec_ld (int, const unsigned short *);
8293 vector bool char vec_ld (int, const vector bool char *);
8294 vector signed char vec_ld (int, const vector signed char *);
8295 vector signed char vec_ld (int, const signed char *);
8296 vector unsigned char vec_ld (int, const vector unsigned char *);
8297 vector unsigned char vec_ld (int, const unsigned char *);
8299 vector signed char vec_lde (int, const signed char *);
8300 vector unsigned char vec_lde (int, const unsigned char *);
8301 vector signed short vec_lde (int, const short *);
8302 vector unsigned short vec_lde (int, const unsigned short *);
8303 vector float vec_lde (int, const float *);
8304 vector signed int vec_lde (int, const int *);
8305 vector unsigned int vec_lde (int, const unsigned int *);
8306 vector signed int vec_lde (int, const long *);
8307 vector unsigned int vec_lde (int, const unsigned long *);
8309 vector float vec_lvewx (int, float *);
8310 vector signed int vec_lvewx (int, int *);
8311 vector unsigned int vec_lvewx (int, unsigned int *);
8312 vector signed int vec_lvewx (int, long *);
8313 vector unsigned int vec_lvewx (int, unsigned long *);
8315 vector signed short vec_lvehx (int, short *);
8316 vector unsigned short vec_lvehx (int, unsigned short *);
8318 vector signed char vec_lvebx (int, char *);
8319 vector unsigned char vec_lvebx (int, unsigned char *);
8321 vector float vec_ldl (int, const vector float *);
8322 vector float vec_ldl (int, const float *);
8323 vector bool int vec_ldl (int, const vector bool int *);
8324 vector signed int vec_ldl (int, const vector signed int *);
8325 vector signed int vec_ldl (int, const int *);
8326 vector signed int vec_ldl (int, const long *);
8327 vector unsigned int vec_ldl (int, const vector unsigned int *);
8328 vector unsigned int vec_ldl (int, const unsigned int *);
8329 vector unsigned int vec_ldl (int, const unsigned long *);
8330 vector bool short vec_ldl (int, const vector bool short *);
8331 vector pixel vec_ldl (int, const vector pixel *);
8332 vector signed short vec_ldl (int, const vector signed short *);
8333 vector signed short vec_ldl (int, const short *);
8334 vector unsigned short vec_ldl (int, const vector unsigned short *);
8335 vector unsigned short vec_ldl (int, const unsigned short *);
8336 vector bool char vec_ldl (int, const vector bool char *);
8337 vector signed char vec_ldl (int, const vector signed char *);
8338 vector signed char vec_ldl (int, const signed char *);
8339 vector unsigned char vec_ldl (int, const vector unsigned char *);
8340 vector unsigned char vec_ldl (int, const unsigned char *);
8342 vector float vec_loge (vector float);
8344 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8345 vector unsigned char vec_lvsl (int, const volatile signed char *);
8346 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8347 vector unsigned char vec_lvsl (int, const volatile short *);
8348 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8349 vector unsigned char vec_lvsl (int, const volatile int *);
8350 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8351 vector unsigned char vec_lvsl (int, const volatile long *);
8352 vector unsigned char vec_lvsl (int, const volatile float *);
8354 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8355 vector unsigned char vec_lvsr (int, const volatile signed char *);
8356 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8357 vector unsigned char vec_lvsr (int, const volatile short *);
8358 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8359 vector unsigned char vec_lvsr (int, const volatile int *);
8360 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8361 vector unsigned char vec_lvsr (int, const volatile long *);
8362 vector unsigned char vec_lvsr (int, const volatile float *);
8364 vector float vec_madd (vector float, vector float, vector float);
8366 vector signed short vec_madds (vector signed short,
8367 vector signed short,
8368 vector signed short);
8370 vector unsigned char vec_max (vector bool char, vector unsigned char);
8371 vector unsigned char vec_max (vector unsigned char, vector bool char);
8372 vector unsigned char vec_max (vector unsigned char,
8373 vector unsigned char);
8374 vector signed char vec_max (vector bool char, vector signed char);
8375 vector signed char vec_max (vector signed char, vector bool char);
8376 vector signed char vec_max (vector signed char, vector signed char);
8377 vector unsigned short vec_max (vector bool short,
8378 vector unsigned short);
8379 vector unsigned short vec_max (vector unsigned short,
8381 vector unsigned short vec_max (vector unsigned short,
8382 vector unsigned short);
8383 vector signed short vec_max (vector bool short, vector signed short);
8384 vector signed short vec_max (vector signed short, vector bool short);
8385 vector signed short vec_max (vector signed short, vector signed short);
8386 vector unsigned int vec_max (vector bool int, vector unsigned int);
8387 vector unsigned int vec_max (vector unsigned int, vector bool int);
8388 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8389 vector signed int vec_max (vector bool int, vector signed int);
8390 vector signed int vec_max (vector signed int, vector bool int);
8391 vector signed int vec_max (vector signed int, vector signed int);
8392 vector float vec_max (vector float, vector float);
8394 vector float vec_vmaxfp (vector float, vector float);
8396 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8397 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8398 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8400 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8401 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8402 vector unsigned int vec_vmaxuw (vector unsigned int,
8403 vector unsigned int);
8405 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8406 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8407 vector signed short vec_vmaxsh (vector signed short,
8408 vector signed short);
8410 vector unsigned short vec_vmaxuh (vector bool short,
8411 vector unsigned short);
8412 vector unsigned short vec_vmaxuh (vector unsigned short,
8414 vector unsigned short vec_vmaxuh (vector unsigned short,
8415 vector unsigned short);
8417 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8418 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8419 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8421 vector unsigned char vec_vmaxub (vector bool char,
8422 vector unsigned char);
8423 vector unsigned char vec_vmaxub (vector unsigned char,
8425 vector unsigned char vec_vmaxub (vector unsigned char,
8426 vector unsigned char);
8428 vector bool char vec_mergeh (vector bool char, vector bool char);
8429 vector signed char vec_mergeh (vector signed char, vector signed char);
8430 vector unsigned char vec_mergeh (vector unsigned char,
8431 vector unsigned char);
8432 vector bool short vec_mergeh (vector bool short, vector bool short);
8433 vector pixel vec_mergeh (vector pixel, vector pixel);
8434 vector signed short vec_mergeh (vector signed short,
8435 vector signed short);
8436 vector unsigned short vec_mergeh (vector unsigned short,
8437 vector unsigned short);
8438 vector float vec_mergeh (vector float, vector float);
8439 vector bool int vec_mergeh (vector bool int, vector bool int);
8440 vector signed int vec_mergeh (vector signed int, vector signed int);
8441 vector unsigned int vec_mergeh (vector unsigned int,
8442 vector unsigned int);
8444 vector float vec_vmrghw (vector float, vector float);
8445 vector bool int vec_vmrghw (vector bool int, vector bool int);
8446 vector signed int vec_vmrghw (vector signed int, vector signed int);
8447 vector unsigned int vec_vmrghw (vector unsigned int,
8448 vector unsigned int);
8450 vector bool short vec_vmrghh (vector bool short, vector bool short);
8451 vector signed short vec_vmrghh (vector signed short,
8452 vector signed short);
8453 vector unsigned short vec_vmrghh (vector unsigned short,
8454 vector unsigned short);
8455 vector pixel vec_vmrghh (vector pixel, vector pixel);
8457 vector bool char vec_vmrghb (vector bool char, vector bool char);
8458 vector signed char vec_vmrghb (vector signed char, vector signed char);
8459 vector unsigned char vec_vmrghb (vector unsigned char,
8460 vector unsigned char);
8462 vector bool char vec_mergel (vector bool char, vector bool char);
8463 vector signed char vec_mergel (vector signed char, vector signed char);
8464 vector unsigned char vec_mergel (vector unsigned char,
8465 vector unsigned char);
8466 vector bool short vec_mergel (vector bool short, vector bool short);
8467 vector pixel vec_mergel (vector pixel, vector pixel);
8468 vector signed short vec_mergel (vector signed short,
8469 vector signed short);
8470 vector unsigned short vec_mergel (vector unsigned short,
8471 vector unsigned short);
8472 vector float vec_mergel (vector float, vector float);
8473 vector bool int vec_mergel (vector bool int, vector bool int);
8474 vector signed int vec_mergel (vector signed int, vector signed int);
8475 vector unsigned int vec_mergel (vector unsigned int,
8476 vector unsigned int);
8478 vector float vec_vmrglw (vector float, vector float);
8479 vector signed int vec_vmrglw (vector signed int, vector signed int);
8480 vector unsigned int vec_vmrglw (vector unsigned int,
8481 vector unsigned int);
8482 vector bool int vec_vmrglw (vector bool int, vector bool int);
8484 vector bool short vec_vmrglh (vector bool short, vector bool short);
8485 vector signed short vec_vmrglh (vector signed short,
8486 vector signed short);
8487 vector unsigned short vec_vmrglh (vector unsigned short,
8488 vector unsigned short);
8489 vector pixel vec_vmrglh (vector pixel, vector pixel);
8491 vector bool char vec_vmrglb (vector bool char, vector bool char);
8492 vector signed char vec_vmrglb (vector signed char, vector signed char);
8493 vector unsigned char vec_vmrglb (vector unsigned char,
8494 vector unsigned char);
8496 vector unsigned short vec_mfvscr (void);
8498 vector unsigned char vec_min (vector bool char, vector unsigned char);
8499 vector unsigned char vec_min (vector unsigned char, vector bool char);
8500 vector unsigned char vec_min (vector unsigned char,
8501 vector unsigned char);
8502 vector signed char vec_min (vector bool char, vector signed char);
8503 vector signed char vec_min (vector signed char, vector bool char);
8504 vector signed char vec_min (vector signed char, vector signed char);
8505 vector unsigned short vec_min (vector bool short,
8506 vector unsigned short);
8507 vector unsigned short vec_min (vector unsigned short,
8509 vector unsigned short vec_min (vector unsigned short,
8510 vector unsigned short);
8511 vector signed short vec_min (vector bool short, vector signed short);
8512 vector signed short vec_min (vector signed short, vector bool short);
8513 vector signed short vec_min (vector signed short, vector signed short);
8514 vector unsigned int vec_min (vector bool int, vector unsigned int);
8515 vector unsigned int vec_min (vector unsigned int, vector bool int);
8516 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8517 vector signed int vec_min (vector bool int, vector signed int);
8518 vector signed int vec_min (vector signed int, vector bool int);
8519 vector signed int vec_min (vector signed int, vector signed int);
8520 vector float vec_min (vector float, vector float);
8522 vector float vec_vminfp (vector float, vector float);
8524 vector signed int vec_vminsw (vector bool int, vector signed int);
8525 vector signed int vec_vminsw (vector signed int, vector bool int);
8526 vector signed int vec_vminsw (vector signed int, vector signed int);
8528 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8529 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8530 vector unsigned int vec_vminuw (vector unsigned int,
8531 vector unsigned int);
8533 vector signed short vec_vminsh (vector bool short, vector signed short);
8534 vector signed short vec_vminsh (vector signed short, vector bool short);
8535 vector signed short vec_vminsh (vector signed short,
8536 vector signed short);
8538 vector unsigned short vec_vminuh (vector bool short,
8539 vector unsigned short);
8540 vector unsigned short vec_vminuh (vector unsigned short,
8542 vector unsigned short vec_vminuh (vector unsigned short,
8543 vector unsigned short);
8545 vector signed char vec_vminsb (vector bool char, vector signed char);
8546 vector signed char vec_vminsb (vector signed char, vector bool char);
8547 vector signed char vec_vminsb (vector signed char, vector signed char);
8549 vector unsigned char vec_vminub (vector bool char,
8550 vector unsigned char);
8551 vector unsigned char vec_vminub (vector unsigned char,
8553 vector unsigned char vec_vminub (vector unsigned char,
8554 vector unsigned char);
8556 vector signed short vec_mladd (vector signed short,
8557 vector signed short,
8558 vector signed short);
8559 vector signed short vec_mladd (vector signed short,
8560 vector unsigned short,
8561 vector unsigned short);
8562 vector signed short vec_mladd (vector unsigned short,
8563 vector signed short,
8564 vector signed short);
8565 vector unsigned short vec_mladd (vector unsigned short,
8566 vector unsigned short,
8567 vector unsigned short);
8569 vector signed short vec_mradds (vector signed short,
8570 vector signed short,
8571 vector signed short);
8573 vector unsigned int vec_msum (vector unsigned char,
8574 vector unsigned char,
8575 vector unsigned int);
8576 vector signed int vec_msum (vector signed char,
8577 vector unsigned char,
8579 vector unsigned int vec_msum (vector unsigned short,
8580 vector unsigned short,
8581 vector unsigned int);
8582 vector signed int vec_msum (vector signed short,
8583 vector signed short,
8586 vector signed int vec_vmsumshm (vector signed short,
8587 vector signed short,
8590 vector unsigned int vec_vmsumuhm (vector unsigned short,
8591 vector unsigned short,
8592 vector unsigned int);
8594 vector signed int vec_vmsummbm (vector signed char,
8595 vector unsigned char,
8598 vector unsigned int vec_vmsumubm (vector unsigned char,
8599 vector unsigned char,
8600 vector unsigned int);
8602 vector unsigned int vec_msums (vector unsigned short,
8603 vector unsigned short,
8604 vector unsigned int);
8605 vector signed int vec_msums (vector signed short,
8606 vector signed short,
8609 vector signed int vec_vmsumshs (vector signed short,
8610 vector signed short,
8613 vector unsigned int vec_vmsumuhs (vector unsigned short,
8614 vector unsigned short,
8615 vector unsigned int);
8617 void vec_mtvscr (vector signed int);
8618 void vec_mtvscr (vector unsigned int);
8619 void vec_mtvscr (vector bool int);
8620 void vec_mtvscr (vector signed short);
8621 void vec_mtvscr (vector unsigned short);
8622 void vec_mtvscr (vector bool short);
8623 void vec_mtvscr (vector pixel);
8624 void vec_mtvscr (vector signed char);
8625 void vec_mtvscr (vector unsigned char);
8626 void vec_mtvscr (vector bool char);
8628 vector unsigned short vec_mule (vector unsigned char,
8629 vector unsigned char);
8630 vector signed short vec_mule (vector signed char,
8631 vector signed char);
8632 vector unsigned int vec_mule (vector unsigned short,
8633 vector unsigned short);
8634 vector signed int vec_mule (vector signed short, vector signed short);
8636 vector signed int vec_vmulesh (vector signed short,
8637 vector signed short);
8639 vector unsigned int vec_vmuleuh (vector unsigned short,
8640 vector unsigned short);
8642 vector signed short vec_vmulesb (vector signed char,
8643 vector signed char);
8645 vector unsigned short vec_vmuleub (vector unsigned char,
8646 vector unsigned char);
8648 vector unsigned short vec_mulo (vector unsigned char,
8649 vector unsigned char);
8650 vector signed short vec_mulo (vector signed char, vector signed char);
8651 vector unsigned int vec_mulo (vector unsigned short,
8652 vector unsigned short);
8653 vector signed int vec_mulo (vector signed short, vector signed short);
8655 vector signed int vec_vmulosh (vector signed short,
8656 vector signed short);
8658 vector unsigned int vec_vmulouh (vector unsigned short,
8659 vector unsigned short);
8661 vector signed short vec_vmulosb (vector signed char,
8662 vector signed char);
8664 vector unsigned short vec_vmuloub (vector unsigned char,
8665 vector unsigned char);
8667 vector float vec_nmsub (vector float, vector float, vector float);
8669 vector float vec_nor (vector float, vector float);
8670 vector signed int vec_nor (vector signed int, vector signed int);
8671 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8672 vector bool int vec_nor (vector bool int, vector bool int);
8673 vector signed short vec_nor (vector signed short, vector signed short);
8674 vector unsigned short vec_nor (vector unsigned short,
8675 vector unsigned short);
8676 vector bool short vec_nor (vector bool short, vector bool short);
8677 vector signed char vec_nor (vector signed char, vector signed char);
8678 vector unsigned char vec_nor (vector unsigned char,
8679 vector unsigned char);
8680 vector bool char vec_nor (vector bool char, vector bool char);
8682 vector float vec_or (vector float, vector float);
8683 vector float vec_or (vector float, vector bool int);
8684 vector float vec_or (vector bool int, vector float);
8685 vector bool int vec_or (vector bool int, vector bool int);
8686 vector signed int vec_or (vector bool int, vector signed int);
8687 vector signed int vec_or (vector signed int, vector bool int);
8688 vector signed int vec_or (vector signed int, vector signed int);
8689 vector unsigned int vec_or (vector bool int, vector unsigned int);
8690 vector unsigned int vec_or (vector unsigned int, vector bool int);
8691 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8692 vector bool short vec_or (vector bool short, vector bool short);
8693 vector signed short vec_or (vector bool short, vector signed short);
8694 vector signed short vec_or (vector signed short, vector bool short);
8695 vector signed short vec_or (vector signed short, vector signed short);
8696 vector unsigned short vec_or (vector bool short, vector unsigned short);
8697 vector unsigned short vec_or (vector unsigned short, vector bool short);
8698 vector unsigned short vec_or (vector unsigned short,
8699 vector unsigned short);
8700 vector signed char vec_or (vector bool char, vector signed char);
8701 vector bool char vec_or (vector bool char, vector bool char);
8702 vector signed char vec_or (vector signed char, vector bool char);
8703 vector signed char vec_or (vector signed char, vector signed char);
8704 vector unsigned char vec_or (vector bool char, vector unsigned char);
8705 vector unsigned char vec_or (vector unsigned char, vector bool char);
8706 vector unsigned char vec_or (vector unsigned char,
8707 vector unsigned char);
8709 vector signed char vec_pack (vector signed short, vector signed short);
8710 vector unsigned char vec_pack (vector unsigned short,
8711 vector unsigned short);
8712 vector bool char vec_pack (vector bool short, vector bool short);
8713 vector signed short vec_pack (vector signed int, vector signed int);
8714 vector unsigned short vec_pack (vector unsigned int,
8715 vector unsigned int);
8716 vector bool short vec_pack (vector bool int, vector bool int);
8718 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8719 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8720 vector unsigned short vec_vpkuwum (vector unsigned int,
8721 vector unsigned int);
8723 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8724 vector signed char vec_vpkuhum (vector signed short,
8725 vector signed short);
8726 vector unsigned char vec_vpkuhum (vector unsigned short,
8727 vector unsigned short);
8729 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8731 vector unsigned char vec_packs (vector unsigned short,
8732 vector unsigned short);
8733 vector signed char vec_packs (vector signed short, vector signed short);
8734 vector unsigned short vec_packs (vector unsigned int,
8735 vector unsigned int);
8736 vector signed short vec_packs (vector signed int, vector signed int);
8738 vector signed short vec_vpkswss (vector signed int, vector signed int);
8740 vector unsigned short vec_vpkuwus (vector unsigned int,
8741 vector unsigned int);
8743 vector signed char vec_vpkshss (vector signed short,
8744 vector signed short);
8746 vector unsigned char vec_vpkuhus (vector unsigned short,
8747 vector unsigned short);
8749 vector unsigned char vec_packsu (vector unsigned short,
8750 vector unsigned short);
8751 vector unsigned char vec_packsu (vector signed short,
8752 vector signed short);
8753 vector unsigned short vec_packsu (vector unsigned int,
8754 vector unsigned int);
8755 vector unsigned short vec_packsu (vector signed int, vector signed int);
8757 vector unsigned short vec_vpkswus (vector signed int,
8760 vector unsigned char vec_vpkshus (vector signed short,
8761 vector signed short);
8763 vector float vec_perm (vector float,
8765 vector unsigned char);
8766 vector signed int vec_perm (vector signed int,
8768 vector unsigned char);
8769 vector unsigned int vec_perm (vector unsigned int,
8770 vector unsigned int,
8771 vector unsigned char);
8772 vector bool int vec_perm (vector bool int,
8774 vector unsigned char);
8775 vector signed short vec_perm (vector signed short,
8776 vector signed short,
8777 vector unsigned char);
8778 vector unsigned short vec_perm (vector unsigned short,
8779 vector unsigned short,
8780 vector unsigned char);
8781 vector bool short vec_perm (vector bool short,
8783 vector unsigned char);
8784 vector pixel vec_perm (vector pixel,
8786 vector unsigned char);
8787 vector signed char vec_perm (vector signed char,
8789 vector unsigned char);
8790 vector unsigned char vec_perm (vector unsigned char,
8791 vector unsigned char,
8792 vector unsigned char);
8793 vector bool char vec_perm (vector bool char,
8795 vector unsigned char);
8797 vector float vec_re (vector float);
8799 vector signed char vec_rl (vector signed char,
8800 vector unsigned char);
8801 vector unsigned char vec_rl (vector unsigned char,
8802 vector unsigned char);
8803 vector signed short vec_rl (vector signed short, vector unsigned short);
8804 vector unsigned short vec_rl (vector unsigned short,
8805 vector unsigned short);
8806 vector signed int vec_rl (vector signed int, vector unsigned int);
8807 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8809 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8810 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8812 vector signed short vec_vrlh (vector signed short,
8813 vector unsigned short);
8814 vector unsigned short vec_vrlh (vector unsigned short,
8815 vector unsigned short);
8817 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8818 vector unsigned char vec_vrlb (vector unsigned char,
8819 vector unsigned char);
8821 vector float vec_round (vector float);
8823 vector float vec_rsqrte (vector float);
8825 vector float vec_sel (vector float, vector float, vector bool int);
8826 vector float vec_sel (vector float, vector float, vector unsigned int);
8827 vector signed int vec_sel (vector signed int,
8830 vector signed int vec_sel (vector signed int,
8832 vector unsigned int);
8833 vector unsigned int vec_sel (vector unsigned int,
8834 vector unsigned int,
8836 vector unsigned int vec_sel (vector unsigned int,
8837 vector unsigned int,
8838 vector unsigned int);
8839 vector bool int vec_sel (vector bool int,
8842 vector bool int vec_sel (vector bool int,
8844 vector unsigned int);
8845 vector signed short vec_sel (vector signed short,
8846 vector signed short,
8848 vector signed short vec_sel (vector signed short,
8849 vector signed short,
8850 vector unsigned short);
8851 vector unsigned short vec_sel (vector unsigned short,
8852 vector unsigned short,
8854 vector unsigned short vec_sel (vector unsigned short,
8855 vector unsigned short,
8856 vector unsigned short);
8857 vector bool short vec_sel (vector bool short,
8860 vector bool short vec_sel (vector bool short,
8862 vector unsigned short);
8863 vector signed char vec_sel (vector signed char,
8866 vector signed char vec_sel (vector signed char,
8868 vector unsigned char);
8869 vector unsigned char vec_sel (vector unsigned char,
8870 vector unsigned char,
8872 vector unsigned char vec_sel (vector unsigned char,
8873 vector unsigned char,
8874 vector unsigned char);
8875 vector bool char vec_sel (vector bool char,
8878 vector bool char vec_sel (vector bool char,
8880 vector unsigned char);
8882 vector signed char vec_sl (vector signed char,
8883 vector unsigned char);
8884 vector unsigned char vec_sl (vector unsigned char,
8885 vector unsigned char);
8886 vector signed short vec_sl (vector signed short, vector unsigned short);
8887 vector unsigned short vec_sl (vector unsigned short,
8888 vector unsigned short);
8889 vector signed int vec_sl (vector signed int, vector unsigned int);
8890 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
8892 vector signed int vec_vslw (vector signed int, vector unsigned int);
8893 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
8895 vector signed short vec_vslh (vector signed short,
8896 vector unsigned short);
8897 vector unsigned short vec_vslh (vector unsigned short,
8898 vector unsigned short);
8900 vector signed char vec_vslb (vector signed char, vector unsigned char);
8901 vector unsigned char vec_vslb (vector unsigned char,
8902 vector unsigned char);
8904 vector float vec_sld (vector float, vector float, const int);
8905 vector signed int vec_sld (vector signed int,
8908 vector unsigned int vec_sld (vector unsigned int,
8909 vector unsigned int,
8911 vector bool int vec_sld (vector bool int,
8914 vector signed short vec_sld (vector signed short,
8915 vector signed short,
8917 vector unsigned short vec_sld (vector unsigned short,
8918 vector unsigned short,
8920 vector bool short vec_sld (vector bool short,
8923 vector pixel vec_sld (vector pixel,
8926 vector signed char vec_sld (vector signed char,
8929 vector unsigned char vec_sld (vector unsigned char,
8930 vector unsigned char,
8932 vector bool char vec_sld (vector bool char,
8936 vector signed int vec_sll (vector signed int,
8937 vector unsigned int);
8938 vector signed int vec_sll (vector signed int,
8939 vector unsigned short);
8940 vector signed int vec_sll (vector signed int,
8941 vector unsigned char);
8942 vector unsigned int vec_sll (vector unsigned int,
8943 vector unsigned int);
8944 vector unsigned int vec_sll (vector unsigned int,
8945 vector unsigned short);
8946 vector unsigned int vec_sll (vector unsigned int,
8947 vector unsigned char);
8948 vector bool int vec_sll (vector bool int,
8949 vector unsigned int);
8950 vector bool int vec_sll (vector bool int,
8951 vector unsigned short);
8952 vector bool int vec_sll (vector bool int,
8953 vector unsigned char);
8954 vector signed short vec_sll (vector signed short,
8955 vector unsigned int);
8956 vector signed short vec_sll (vector signed short,
8957 vector unsigned short);
8958 vector signed short vec_sll (vector signed short,
8959 vector unsigned char);
8960 vector unsigned short vec_sll (vector unsigned short,
8961 vector unsigned int);
8962 vector unsigned short vec_sll (vector unsigned short,
8963 vector unsigned short);
8964 vector unsigned short vec_sll (vector unsigned short,
8965 vector unsigned char);
8966 vector bool short vec_sll (vector bool short, vector unsigned int);
8967 vector bool short vec_sll (vector bool short, vector unsigned short);
8968 vector bool short vec_sll (vector bool short, vector unsigned char);
8969 vector pixel vec_sll (vector pixel, vector unsigned int);
8970 vector pixel vec_sll (vector pixel, vector unsigned short);
8971 vector pixel vec_sll (vector pixel, vector unsigned char);
8972 vector signed char vec_sll (vector signed char, vector unsigned int);
8973 vector signed char vec_sll (vector signed char, vector unsigned short);
8974 vector signed char vec_sll (vector signed char, vector unsigned char);
8975 vector unsigned char vec_sll (vector unsigned char,
8976 vector unsigned int);
8977 vector unsigned char vec_sll (vector unsigned char,
8978 vector unsigned short);
8979 vector unsigned char vec_sll (vector unsigned char,
8980 vector unsigned char);
8981 vector bool char vec_sll (vector bool char, vector unsigned int);
8982 vector bool char vec_sll (vector bool char, vector unsigned short);
8983 vector bool char vec_sll (vector bool char, vector unsigned char);
8985 vector float vec_slo (vector float, vector signed char);
8986 vector float vec_slo (vector float, vector unsigned char);
8987 vector signed int vec_slo (vector signed int, vector signed char);
8988 vector signed int vec_slo (vector signed int, vector unsigned char);
8989 vector unsigned int vec_slo (vector unsigned int, vector signed char);
8990 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
8991 vector signed short vec_slo (vector signed short, vector signed char);
8992 vector signed short vec_slo (vector signed short, vector unsigned char);
8993 vector unsigned short vec_slo (vector unsigned short,
8994 vector signed char);
8995 vector unsigned short vec_slo (vector unsigned short,
8996 vector unsigned char);
8997 vector pixel vec_slo (vector pixel, vector signed char);
8998 vector pixel vec_slo (vector pixel, vector unsigned char);
8999 vector signed char vec_slo (vector signed char, vector signed char);
9000 vector signed char vec_slo (vector signed char, vector unsigned char);
9001 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9002 vector unsigned char vec_slo (vector unsigned char,
9003 vector unsigned char);
9005 vector signed char vec_splat (vector signed char, const int);
9006 vector unsigned char vec_splat (vector unsigned char, const int);
9007 vector bool char vec_splat (vector bool char, const int);
9008 vector signed short vec_splat (vector signed short, const int);
9009 vector unsigned short vec_splat (vector unsigned short, const int);
9010 vector bool short vec_splat (vector bool short, const int);
9011 vector pixel vec_splat (vector pixel, const int);
9012 vector float vec_splat (vector float, const int);
9013 vector signed int vec_splat (vector signed int, const int);
9014 vector unsigned int vec_splat (vector unsigned int, const int);
9015 vector bool int vec_splat (vector bool int, const int);
9017 vector float vec_vspltw (vector float, const int);
9018 vector signed int vec_vspltw (vector signed int, const int);
9019 vector unsigned int vec_vspltw (vector unsigned int, const int);
9020 vector bool int vec_vspltw (vector bool int, const int);
9022 vector bool short vec_vsplth (vector bool short, const int);
9023 vector signed short vec_vsplth (vector signed short, const int);
9024 vector unsigned short vec_vsplth (vector unsigned short, const int);
9025 vector pixel vec_vsplth (vector pixel, const int);
9027 vector signed char vec_vspltb (vector signed char, const int);
9028 vector unsigned char vec_vspltb (vector unsigned char, const int);
9029 vector bool char vec_vspltb (vector bool char, const int);
9031 vector signed char vec_splat_s8 (const int);
9033 vector signed short vec_splat_s16 (const int);
9035 vector signed int vec_splat_s32 (const int);
9037 vector unsigned char vec_splat_u8 (const int);
9039 vector unsigned short vec_splat_u16 (const int);
9041 vector unsigned int vec_splat_u32 (const int);
9043 vector signed char vec_sr (vector signed char, vector unsigned char);
9044 vector unsigned char vec_sr (vector unsigned char,
9045 vector unsigned char);
9046 vector signed short vec_sr (vector signed short,
9047 vector unsigned short);
9048 vector unsigned short vec_sr (vector unsigned short,
9049 vector unsigned short);
9050 vector signed int vec_sr (vector signed int, vector unsigned int);
9051 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9053 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9054 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9056 vector signed short vec_vsrh (vector signed short,
9057 vector unsigned short);
9058 vector unsigned short vec_vsrh (vector unsigned short,
9059 vector unsigned short);
9061 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9062 vector unsigned char vec_vsrb (vector unsigned char,
9063 vector unsigned char);
9065 vector signed char vec_sra (vector signed char, vector unsigned char);
9066 vector unsigned char vec_sra (vector unsigned char,
9067 vector unsigned char);
9068 vector signed short vec_sra (vector signed short,
9069 vector unsigned short);
9070 vector unsigned short vec_sra (vector unsigned short,
9071 vector unsigned short);
9072 vector signed int vec_sra (vector signed int, vector unsigned int);
9073 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9075 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9076 vector unsigned int vec_vsraw (vector unsigned int,
9077 vector unsigned int);
9079 vector signed short vec_vsrah (vector signed short,
9080 vector unsigned short);
9081 vector unsigned short vec_vsrah (vector unsigned short,
9082 vector unsigned short);
9084 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9085 vector unsigned char vec_vsrab (vector unsigned char,
9086 vector unsigned char);
9088 vector signed int vec_srl (vector signed int, vector unsigned int);
9089 vector signed int vec_srl (vector signed int, vector unsigned short);
9090 vector signed int vec_srl (vector signed int, vector unsigned char);
9091 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9092 vector unsigned int vec_srl (vector unsigned int,
9093 vector unsigned short);
9094 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9095 vector bool int vec_srl (vector bool int, vector unsigned int);
9096 vector bool int vec_srl (vector bool int, vector unsigned short);
9097 vector bool int vec_srl (vector bool int, vector unsigned char);
9098 vector signed short vec_srl (vector signed short, vector unsigned int);
9099 vector signed short vec_srl (vector signed short,
9100 vector unsigned short);
9101 vector signed short vec_srl (vector signed short, vector unsigned char);
9102 vector unsigned short vec_srl (vector unsigned short,
9103 vector unsigned int);
9104 vector unsigned short vec_srl (vector unsigned short,
9105 vector unsigned short);
9106 vector unsigned short vec_srl (vector unsigned short,
9107 vector unsigned char);
9108 vector bool short vec_srl (vector bool short, vector unsigned int);
9109 vector bool short vec_srl (vector bool short, vector unsigned short);
9110 vector bool short vec_srl (vector bool short, vector unsigned char);
9111 vector pixel vec_srl (vector pixel, vector unsigned int);
9112 vector pixel vec_srl (vector pixel, vector unsigned short);
9113 vector pixel vec_srl (vector pixel, vector unsigned char);
9114 vector signed char vec_srl (vector signed char, vector unsigned int);
9115 vector signed char vec_srl (vector signed char, vector unsigned short);
9116 vector signed char vec_srl (vector signed char, vector unsigned char);
9117 vector unsigned char vec_srl (vector unsigned char,
9118 vector unsigned int);
9119 vector unsigned char vec_srl (vector unsigned char,
9120 vector unsigned short);
9121 vector unsigned char vec_srl (vector unsigned char,
9122 vector unsigned char);
9123 vector bool char vec_srl (vector bool char, vector unsigned int);
9124 vector bool char vec_srl (vector bool char, vector unsigned short);
9125 vector bool char vec_srl (vector bool char, vector unsigned char);
9127 vector float vec_sro (vector float, vector signed char);
9128 vector float vec_sro (vector float, vector unsigned char);
9129 vector signed int vec_sro (vector signed int, vector signed char);
9130 vector signed int vec_sro (vector signed int, vector unsigned char);
9131 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9132 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9133 vector signed short vec_sro (vector signed short, vector signed char);
9134 vector signed short vec_sro (vector signed short, vector unsigned char);
9135 vector unsigned short vec_sro (vector unsigned short,
9136 vector signed char);
9137 vector unsigned short vec_sro (vector unsigned short,
9138 vector unsigned char);
9139 vector pixel vec_sro (vector pixel, vector signed char);
9140 vector pixel vec_sro (vector pixel, vector unsigned char);
9141 vector signed char vec_sro (vector signed char, vector signed char);
9142 vector signed char vec_sro (vector signed char, vector unsigned char);
9143 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9144 vector unsigned char vec_sro (vector unsigned char,
9145 vector unsigned char);
9147 void vec_st (vector float, int, vector float *);
9148 void vec_st (vector float, int, float *);
9149 void vec_st (vector signed int, int, vector signed int *);
9150 void vec_st (vector signed int, int, int *);
9151 void vec_st (vector unsigned int, int, vector unsigned int *);
9152 void vec_st (vector unsigned int, int, unsigned int *);
9153 void vec_st (vector bool int, int, vector bool int *);
9154 void vec_st (vector bool int, int, unsigned int *);
9155 void vec_st (vector bool int, int, int *);
9156 void vec_st (vector signed short, int, vector signed short *);
9157 void vec_st (vector signed short, int, short *);
9158 void vec_st (vector unsigned short, int, vector unsigned short *);
9159 void vec_st (vector unsigned short, int, unsigned short *);
9160 void vec_st (vector bool short, int, vector bool short *);
9161 void vec_st (vector bool short, int, unsigned short *);
9162 void vec_st (vector pixel, int, vector pixel *);
9163 void vec_st (vector pixel, int, unsigned short *);
9164 void vec_st (vector pixel, int, short *);
9165 void vec_st (vector bool short, int, short *);
9166 void vec_st (vector signed char, int, vector signed char *);
9167 void vec_st (vector signed char, int, signed char *);
9168 void vec_st (vector unsigned char, int, vector unsigned char *);
9169 void vec_st (vector unsigned char, int, unsigned char *);
9170 void vec_st (vector bool char, int, vector bool char *);
9171 void vec_st (vector bool char, int, unsigned char *);
9172 void vec_st (vector bool char, int, signed char *);
9174 void vec_ste (vector signed char, int, signed char *);
9175 void vec_ste (vector unsigned char, int, unsigned char *);
9176 void vec_ste (vector bool char, int, signed char *);
9177 void vec_ste (vector bool char, int, unsigned char *);
9178 void vec_ste (vector signed short, int, short *);
9179 void vec_ste (vector unsigned short, int, unsigned short *);
9180 void vec_ste (vector bool short, int, short *);
9181 void vec_ste (vector bool short, int, unsigned short *);
9182 void vec_ste (vector pixel, int, short *);
9183 void vec_ste (vector pixel, int, unsigned short *);
9184 void vec_ste (vector float, int, float *);
9185 void vec_ste (vector signed int, int, int *);
9186 void vec_ste (vector unsigned int, int, unsigned int *);
9187 void vec_ste (vector bool int, int, int *);
9188 void vec_ste (vector bool int, int, unsigned int *);
9190 void vec_stvewx (vector float, int, float *);
9191 void vec_stvewx (vector signed int, int, int *);
9192 void vec_stvewx (vector unsigned int, int, unsigned int *);
9193 void vec_stvewx (vector bool int, int, int *);
9194 void vec_stvewx (vector bool int, int, unsigned int *);
9196 void vec_stvehx (vector signed short, int, short *);
9197 void vec_stvehx (vector unsigned short, int, unsigned short *);
9198 void vec_stvehx (vector bool short, int, short *);
9199 void vec_stvehx (vector bool short, int, unsigned short *);
9200 void vec_stvehx (vector pixel, int, short *);
9201 void vec_stvehx (vector pixel, int, unsigned short *);
9203 void vec_stvebx (vector signed char, int, signed char *);
9204 void vec_stvebx (vector unsigned char, int, unsigned char *);
9205 void vec_stvebx (vector bool char, int, signed char *);
9206 void vec_stvebx (vector bool char, int, unsigned char *);
9208 void vec_stl (vector float, int, vector float *);
9209 void vec_stl (vector float, int, float *);
9210 void vec_stl (vector signed int, int, vector signed int *);
9211 void vec_stl (vector signed int, int, int *);
9212 void vec_stl (vector unsigned int, int, vector unsigned int *);
9213 void vec_stl (vector unsigned int, int, unsigned int *);
9214 void vec_stl (vector bool int, int, vector bool int *);
9215 void vec_stl (vector bool int, int, unsigned int *);
9216 void vec_stl (vector bool int, int, int *);
9217 void vec_stl (vector signed short, int, vector signed short *);
9218 void vec_stl (vector signed short, int, short *);
9219 void vec_stl (vector unsigned short, int, vector unsigned short *);
9220 void vec_stl (vector unsigned short, int, unsigned short *);
9221 void vec_stl (vector bool short, int, vector bool short *);
9222 void vec_stl (vector bool short, int, unsigned short *);
9223 void vec_stl (vector bool short, int, short *);
9224 void vec_stl (vector pixel, int, vector pixel *);
9225 void vec_stl (vector pixel, int, unsigned short *);
9226 void vec_stl (vector pixel, int, short *);
9227 void vec_stl (vector signed char, int, vector signed char *);
9228 void vec_stl (vector signed char, int, signed char *);
9229 void vec_stl (vector unsigned char, int, vector unsigned char *);
9230 void vec_stl (vector unsigned char, int, unsigned char *);
9231 void vec_stl (vector bool char, int, vector bool char *);
9232 void vec_stl (vector bool char, int, unsigned char *);
9233 void vec_stl (vector bool char, int, signed char *);
9235 vector signed char vec_sub (vector bool char, vector signed char);
9236 vector signed char vec_sub (vector signed char, vector bool char);
9237 vector signed char vec_sub (vector signed char, vector signed char);
9238 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9239 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9240 vector unsigned char vec_sub (vector unsigned char,
9241 vector unsigned char);
9242 vector signed short vec_sub (vector bool short, vector signed short);
9243 vector signed short vec_sub (vector signed short, vector bool short);
9244 vector signed short vec_sub (vector signed short, vector signed short);
9245 vector unsigned short vec_sub (vector bool short,
9246 vector unsigned short);
9247 vector unsigned short vec_sub (vector unsigned short,
9249 vector unsigned short vec_sub (vector unsigned short,
9250 vector unsigned short);
9251 vector signed int vec_sub (vector bool int, vector signed int);
9252 vector signed int vec_sub (vector signed int, vector bool int);
9253 vector signed int vec_sub (vector signed int, vector signed int);
9254 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9255 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9256 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9257 vector float vec_sub (vector float, vector float);
9259 vector float vec_vsubfp (vector float, vector float);
9261 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9262 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9263 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9264 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9265 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9266 vector unsigned int vec_vsubuwm (vector unsigned int,
9267 vector unsigned int);
9269 vector signed short vec_vsubuhm (vector bool short,
9270 vector signed short);
9271 vector signed short vec_vsubuhm (vector signed short,
9273 vector signed short vec_vsubuhm (vector signed short,
9274 vector signed short);
9275 vector unsigned short vec_vsubuhm (vector bool short,
9276 vector unsigned short);
9277 vector unsigned short vec_vsubuhm (vector unsigned short,
9279 vector unsigned short vec_vsubuhm (vector unsigned short,
9280 vector unsigned short);
9282 vector signed char vec_vsububm (vector bool char, vector signed char);
9283 vector signed char vec_vsububm (vector signed char, vector bool char);
9284 vector signed char vec_vsububm (vector signed char, vector signed char);
9285 vector unsigned char vec_vsububm (vector bool char,
9286 vector unsigned char);
9287 vector unsigned char vec_vsububm (vector unsigned char,
9289 vector unsigned char vec_vsububm (vector unsigned char,
9290 vector unsigned char);
9292 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9294 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9295 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9296 vector unsigned char vec_subs (vector unsigned char,
9297 vector unsigned char);
9298 vector signed char vec_subs (vector bool char, vector signed char);
9299 vector signed char vec_subs (vector signed char, vector bool char);
9300 vector signed char vec_subs (vector signed char, vector signed char);
9301 vector unsigned short vec_subs (vector bool short,
9302 vector unsigned short);
9303 vector unsigned short vec_subs (vector unsigned short,
9305 vector unsigned short vec_subs (vector unsigned short,
9306 vector unsigned short);
9307 vector signed short vec_subs (vector bool short, vector signed short);
9308 vector signed short vec_subs (vector signed short, vector bool short);
9309 vector signed short vec_subs (vector signed short, vector signed short);
9310 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9311 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9312 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9313 vector signed int vec_subs (vector bool int, vector signed int);
9314 vector signed int vec_subs (vector signed int, vector bool int);
9315 vector signed int vec_subs (vector signed int, vector signed int);
9317 vector signed int vec_vsubsws (vector bool int, vector signed int);
9318 vector signed int vec_vsubsws (vector signed int, vector bool int);
9319 vector signed int vec_vsubsws (vector signed int, vector signed int);
9321 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9322 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9323 vector unsigned int vec_vsubuws (vector unsigned int,
9324 vector unsigned int);
9326 vector signed short vec_vsubshs (vector bool short,
9327 vector signed short);
9328 vector signed short vec_vsubshs (vector signed short,
9330 vector signed short vec_vsubshs (vector signed short,
9331 vector signed short);
9333 vector unsigned short vec_vsubuhs (vector bool short,
9334 vector unsigned short);
9335 vector unsigned short vec_vsubuhs (vector unsigned short,
9337 vector unsigned short vec_vsubuhs (vector unsigned short,
9338 vector unsigned short);
9340 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9341 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9342 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9344 vector unsigned char vec_vsububs (vector bool char,
9345 vector unsigned char);
9346 vector unsigned char vec_vsububs (vector unsigned char,
9348 vector unsigned char vec_vsububs (vector unsigned char,
9349 vector unsigned char);
9351 vector unsigned int vec_sum4s (vector unsigned char,
9352 vector unsigned int);
9353 vector signed int vec_sum4s (vector signed char, vector signed int);
9354 vector signed int vec_sum4s (vector signed short, vector signed int);
9356 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9358 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9360 vector unsigned int vec_vsum4ubs (vector unsigned char,
9361 vector unsigned int);
9363 vector signed int vec_sum2s (vector signed int, vector signed int);
9365 vector signed int vec_sums (vector signed int, vector signed int);
9367 vector float vec_trunc (vector float);
9369 vector signed short vec_unpackh (vector signed char);
9370 vector bool short vec_unpackh (vector bool char);
9371 vector signed int vec_unpackh (vector signed short);
9372 vector bool int vec_unpackh (vector bool short);
9373 vector unsigned int vec_unpackh (vector pixel);
9375 vector bool int vec_vupkhsh (vector bool short);
9376 vector signed int vec_vupkhsh (vector signed short);
9378 vector unsigned int vec_vupkhpx (vector pixel);
9380 vector bool short vec_vupkhsb (vector bool char);
9381 vector signed short vec_vupkhsb (vector signed char);
9383 vector signed short vec_unpackl (vector signed char);
9384 vector bool short vec_unpackl (vector bool char);
9385 vector unsigned int vec_unpackl (vector pixel);
9386 vector signed int vec_unpackl (vector signed short);
9387 vector bool int vec_unpackl (vector bool short);
9389 vector unsigned int vec_vupklpx (vector pixel);
9391 vector bool int vec_vupklsh (vector bool short);
9392 vector signed int vec_vupklsh (vector signed short);
9394 vector bool short vec_vupklsb (vector bool char);
9395 vector signed short vec_vupklsb (vector signed char);
9397 vector float vec_xor (vector float, vector float);
9398 vector float vec_xor (vector float, vector bool int);
9399 vector float vec_xor (vector bool int, vector float);
9400 vector bool int vec_xor (vector bool int, vector bool int);
9401 vector signed int vec_xor (vector bool int, vector signed int);
9402 vector signed int vec_xor (vector signed int, vector bool int);
9403 vector signed int vec_xor (vector signed int, vector signed int);
9404 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9405 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9406 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9407 vector bool short vec_xor (vector bool short, vector bool short);
9408 vector signed short vec_xor (vector bool short, vector signed short);
9409 vector signed short vec_xor (vector signed short, vector bool short);
9410 vector signed short vec_xor (vector signed short, vector signed short);
9411 vector unsigned short vec_xor (vector bool short,
9412 vector unsigned short);
9413 vector unsigned short vec_xor (vector unsigned short,
9415 vector unsigned short vec_xor (vector unsigned short,
9416 vector unsigned short);
9417 vector signed char vec_xor (vector bool char, vector signed char);
9418 vector bool char vec_xor (vector bool char, vector bool char);
9419 vector signed char vec_xor (vector signed char, vector bool char);
9420 vector signed char vec_xor (vector signed char, vector signed char);
9421 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9422 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9423 vector unsigned char vec_xor (vector unsigned char,
9424 vector unsigned char);
9426 int vec_all_eq (vector signed char, vector bool char);
9427 int vec_all_eq (vector signed char, vector signed char);
9428 int vec_all_eq (vector unsigned char, vector bool char);
9429 int vec_all_eq (vector unsigned char, vector unsigned char);
9430 int vec_all_eq (vector bool char, vector bool char);
9431 int vec_all_eq (vector bool char, vector unsigned char);
9432 int vec_all_eq (vector bool char, vector signed char);
9433 int vec_all_eq (vector signed short, vector bool short);
9434 int vec_all_eq (vector signed short, vector signed short);
9435 int vec_all_eq (vector unsigned short, vector bool short);
9436 int vec_all_eq (vector unsigned short, vector unsigned short);
9437 int vec_all_eq (vector bool short, vector bool short);
9438 int vec_all_eq (vector bool short, vector unsigned short);
9439 int vec_all_eq (vector bool short, vector signed short);
9440 int vec_all_eq (vector pixel, vector pixel);
9441 int vec_all_eq (vector signed int, vector bool int);
9442 int vec_all_eq (vector signed int, vector signed int);
9443 int vec_all_eq (vector unsigned int, vector bool int);
9444 int vec_all_eq (vector unsigned int, vector unsigned int);
9445 int vec_all_eq (vector bool int, vector bool int);
9446 int vec_all_eq (vector bool int, vector unsigned int);
9447 int vec_all_eq (vector bool int, vector signed int);
9448 int vec_all_eq (vector float, vector float);
9450 int vec_all_ge (vector bool char, vector unsigned char);
9451 int vec_all_ge (vector unsigned char, vector bool char);
9452 int vec_all_ge (vector unsigned char, vector unsigned char);
9453 int vec_all_ge (vector bool char, vector signed char);
9454 int vec_all_ge (vector signed char, vector bool char);
9455 int vec_all_ge (vector signed char, vector signed char);
9456 int vec_all_ge (vector bool short, vector unsigned short);
9457 int vec_all_ge (vector unsigned short, vector bool short);
9458 int vec_all_ge (vector unsigned short, vector unsigned short);
9459 int vec_all_ge (vector signed short, vector signed short);
9460 int vec_all_ge (vector bool short, vector signed short);
9461 int vec_all_ge (vector signed short, vector bool short);
9462 int vec_all_ge (vector bool int, vector unsigned int);
9463 int vec_all_ge (vector unsigned int, vector bool int);
9464 int vec_all_ge (vector unsigned int, vector unsigned int);
9465 int vec_all_ge (vector bool int, vector signed int);
9466 int vec_all_ge (vector signed int, vector bool int);
9467 int vec_all_ge (vector signed int, vector signed int);
9468 int vec_all_ge (vector float, vector float);
9470 int vec_all_gt (vector bool char, vector unsigned char);
9471 int vec_all_gt (vector unsigned char, vector bool char);
9472 int vec_all_gt (vector unsigned char, vector unsigned char);
9473 int vec_all_gt (vector bool char, vector signed char);
9474 int vec_all_gt (vector signed char, vector bool char);
9475 int vec_all_gt (vector signed char, vector signed char);
9476 int vec_all_gt (vector bool short, vector unsigned short);
9477 int vec_all_gt (vector unsigned short, vector bool short);
9478 int vec_all_gt (vector unsigned short, vector unsigned short);
9479 int vec_all_gt (vector bool short, vector signed short);
9480 int vec_all_gt (vector signed short, vector bool short);
9481 int vec_all_gt (vector signed short, vector signed short);
9482 int vec_all_gt (vector bool int, vector unsigned int);
9483 int vec_all_gt (vector unsigned int, vector bool int);
9484 int vec_all_gt (vector unsigned int, vector unsigned int);
9485 int vec_all_gt (vector bool int, vector signed int);
9486 int vec_all_gt (vector signed int, vector bool int);
9487 int vec_all_gt (vector signed int, vector signed int);
9488 int vec_all_gt (vector float, vector float);
9490 int vec_all_in (vector float, vector float);
9492 int vec_all_le (vector bool char, vector unsigned char);
9493 int vec_all_le (vector unsigned char, vector bool char);
9494 int vec_all_le (vector unsigned char, vector unsigned char);
9495 int vec_all_le (vector bool char, vector signed char);
9496 int vec_all_le (vector signed char, vector bool char);
9497 int vec_all_le (vector signed char, vector signed char);
9498 int vec_all_le (vector bool short, vector unsigned short);
9499 int vec_all_le (vector unsigned short, vector bool short);
9500 int vec_all_le (vector unsigned short, vector unsigned short);
9501 int vec_all_le (vector bool short, vector signed short);
9502 int vec_all_le (vector signed short, vector bool short);
9503 int vec_all_le (vector signed short, vector signed short);
9504 int vec_all_le (vector bool int, vector unsigned int);
9505 int vec_all_le (vector unsigned int, vector bool int);
9506 int vec_all_le (vector unsigned int, vector unsigned int);
9507 int vec_all_le (vector bool int, vector signed int);
9508 int vec_all_le (vector signed int, vector bool int);
9509 int vec_all_le (vector signed int, vector signed int);
9510 int vec_all_le (vector float, vector float);
9512 int vec_all_lt (vector bool char, vector unsigned char);
9513 int vec_all_lt (vector unsigned char, vector bool char);
9514 int vec_all_lt (vector unsigned char, vector unsigned char);
9515 int vec_all_lt (vector bool char, vector signed char);
9516 int vec_all_lt (vector signed char, vector bool char);
9517 int vec_all_lt (vector signed char, vector signed char);
9518 int vec_all_lt (vector bool short, vector unsigned short);
9519 int vec_all_lt (vector unsigned short, vector bool short);
9520 int vec_all_lt (vector unsigned short, vector unsigned short);
9521 int vec_all_lt (vector bool short, vector signed short);
9522 int vec_all_lt (vector signed short, vector bool short);
9523 int vec_all_lt (vector signed short, vector signed short);
9524 int vec_all_lt (vector bool int, vector unsigned int);
9525 int vec_all_lt (vector unsigned int, vector bool int);
9526 int vec_all_lt (vector unsigned int, vector unsigned int);
9527 int vec_all_lt (vector bool int, vector signed int);
9528 int vec_all_lt (vector signed int, vector bool int);
9529 int vec_all_lt (vector signed int, vector signed int);
9530 int vec_all_lt (vector float, vector float);
9532 int vec_all_nan (vector float);
9534 int vec_all_ne (vector signed char, vector bool char);
9535 int vec_all_ne (vector signed char, vector signed char);
9536 int vec_all_ne (vector unsigned char, vector bool char);
9537 int vec_all_ne (vector unsigned char, vector unsigned char);
9538 int vec_all_ne (vector bool char, vector bool char);
9539 int vec_all_ne (vector bool char, vector unsigned char);
9540 int vec_all_ne (vector bool char, vector signed char);
9541 int vec_all_ne (vector signed short, vector bool short);
9542 int vec_all_ne (vector signed short, vector signed short);
9543 int vec_all_ne (vector unsigned short, vector bool short);
9544 int vec_all_ne (vector unsigned short, vector unsigned short);
9545 int vec_all_ne (vector bool short, vector bool short);
9546 int vec_all_ne (vector bool short, vector unsigned short);
9547 int vec_all_ne (vector bool short, vector signed short);
9548 int vec_all_ne (vector pixel, vector pixel);
9549 int vec_all_ne (vector signed int, vector bool int);
9550 int vec_all_ne (vector signed int, vector signed int);
9551 int vec_all_ne (vector unsigned int, vector bool int);
9552 int vec_all_ne (vector unsigned int, vector unsigned int);
9553 int vec_all_ne (vector bool int, vector bool int);
9554 int vec_all_ne (vector bool int, vector unsigned int);
9555 int vec_all_ne (vector bool int, vector signed int);
9556 int vec_all_ne (vector float, vector float);
9558 int vec_all_nge (vector float, vector float);
9560 int vec_all_ngt (vector float, vector float);
9562 int vec_all_nle (vector float, vector float);
9564 int vec_all_nlt (vector float, vector float);
9566 int vec_all_numeric (vector float);
9568 int vec_any_eq (vector signed char, vector bool char);
9569 int vec_any_eq (vector signed char, vector signed char);
9570 int vec_any_eq (vector unsigned char, vector bool char);
9571 int vec_any_eq (vector unsigned char, vector unsigned char);
9572 int vec_any_eq (vector bool char, vector bool char);
9573 int vec_any_eq (vector bool char, vector unsigned char);
9574 int vec_any_eq (vector bool char, vector signed char);
9575 int vec_any_eq (vector signed short, vector bool short);
9576 int vec_any_eq (vector signed short, vector signed short);
9577 int vec_any_eq (vector unsigned short, vector bool short);
9578 int vec_any_eq (vector unsigned short, vector unsigned short);
9579 int vec_any_eq (vector bool short, vector bool short);
9580 int vec_any_eq (vector bool short, vector unsigned short);
9581 int vec_any_eq (vector bool short, vector signed short);
9582 int vec_any_eq (vector pixel, vector pixel);
9583 int vec_any_eq (vector signed int, vector bool int);
9584 int vec_any_eq (vector signed int, vector signed int);
9585 int vec_any_eq (vector unsigned int, vector bool int);
9586 int vec_any_eq (vector unsigned int, vector unsigned int);
9587 int vec_any_eq (vector bool int, vector bool int);
9588 int vec_any_eq (vector bool int, vector unsigned int);
9589 int vec_any_eq (vector bool int, vector signed int);
9590 int vec_any_eq (vector float, vector float);
9592 int vec_any_ge (vector signed char, vector bool char);
9593 int vec_any_ge (vector unsigned char, vector bool char);
9594 int vec_any_ge (vector unsigned char, vector unsigned char);
9595 int vec_any_ge (vector signed char, vector signed char);
9596 int vec_any_ge (vector bool char, vector unsigned char);
9597 int vec_any_ge (vector bool char, vector signed char);
9598 int vec_any_ge (vector unsigned short, vector bool short);
9599 int vec_any_ge (vector unsigned short, vector unsigned short);
9600 int vec_any_ge (vector signed short, vector signed short);
9601 int vec_any_ge (vector signed short, vector bool short);
9602 int vec_any_ge (vector bool short, vector unsigned short);
9603 int vec_any_ge (vector bool short, vector signed short);
9604 int vec_any_ge (vector signed int, vector bool int);
9605 int vec_any_ge (vector unsigned int, vector bool int);
9606 int vec_any_ge (vector unsigned int, vector unsigned int);
9607 int vec_any_ge (vector signed int, vector signed int);
9608 int vec_any_ge (vector bool int, vector unsigned int);
9609 int vec_any_ge (vector bool int, vector signed int);
9610 int vec_any_ge (vector float, vector float);
9612 int vec_any_gt (vector bool char, vector unsigned char);
9613 int vec_any_gt (vector unsigned char, vector bool char);
9614 int vec_any_gt (vector unsigned char, vector unsigned char);
9615 int vec_any_gt (vector bool char, vector signed char);
9616 int vec_any_gt (vector signed char, vector bool char);
9617 int vec_any_gt (vector signed char, vector signed char);
9618 int vec_any_gt (vector bool short, vector unsigned short);
9619 int vec_any_gt (vector unsigned short, vector bool short);
9620 int vec_any_gt (vector unsigned short, vector unsigned short);
9621 int vec_any_gt (vector bool short, vector signed short);
9622 int vec_any_gt (vector signed short, vector bool short);
9623 int vec_any_gt (vector signed short, vector signed short);
9624 int vec_any_gt (vector bool int, vector unsigned int);
9625 int vec_any_gt (vector unsigned int, vector bool int);
9626 int vec_any_gt (vector unsigned int, vector unsigned int);
9627 int vec_any_gt (vector bool int, vector signed int);
9628 int vec_any_gt (vector signed int, vector bool int);
9629 int vec_any_gt (vector signed int, vector signed int);
9630 int vec_any_gt (vector float, vector float);
9632 int vec_any_le (vector bool char, vector unsigned char);
9633 int vec_any_le (vector unsigned char, vector bool char);
9634 int vec_any_le (vector unsigned char, vector unsigned char);
9635 int vec_any_le (vector bool char, vector signed char);
9636 int vec_any_le (vector signed char, vector bool char);
9637 int vec_any_le (vector signed char, vector signed char);
9638 int vec_any_le (vector bool short, vector unsigned short);
9639 int vec_any_le (vector unsigned short, vector bool short);
9640 int vec_any_le (vector unsigned short, vector unsigned short);
9641 int vec_any_le (vector bool short, vector signed short);
9642 int vec_any_le (vector signed short, vector bool short);
9643 int vec_any_le (vector signed short, vector signed short);
9644 int vec_any_le (vector bool int, vector unsigned int);
9645 int vec_any_le (vector unsigned int, vector bool int);
9646 int vec_any_le (vector unsigned int, vector unsigned int);
9647 int vec_any_le (vector bool int, vector signed int);
9648 int vec_any_le (vector signed int, vector bool int);
9649 int vec_any_le (vector signed int, vector signed int);
9650 int vec_any_le (vector float, vector float);
9652 int vec_any_lt (vector bool char, vector unsigned char);
9653 int vec_any_lt (vector unsigned char, vector bool char);
9654 int vec_any_lt (vector unsigned char, vector unsigned char);
9655 int vec_any_lt (vector bool char, vector signed char);
9656 int vec_any_lt (vector signed char, vector bool char);
9657 int vec_any_lt (vector signed char, vector signed char);
9658 int vec_any_lt (vector bool short, vector unsigned short);
9659 int vec_any_lt (vector unsigned short, vector bool short);
9660 int vec_any_lt (vector unsigned short, vector unsigned short);
9661 int vec_any_lt (vector bool short, vector signed short);
9662 int vec_any_lt (vector signed short, vector bool short);
9663 int vec_any_lt (vector signed short, vector signed short);
9664 int vec_any_lt (vector bool int, vector unsigned int);
9665 int vec_any_lt (vector unsigned int, vector bool int);
9666 int vec_any_lt (vector unsigned int, vector unsigned int);
9667 int vec_any_lt (vector bool int, vector signed int);
9668 int vec_any_lt (vector signed int, vector bool int);
9669 int vec_any_lt (vector signed int, vector signed int);
9670 int vec_any_lt (vector float, vector float);
9672 int vec_any_nan (vector float);
9674 int vec_any_ne (vector signed char, vector bool char);
9675 int vec_any_ne (vector signed char, vector signed char);
9676 int vec_any_ne (vector unsigned char, vector bool char);
9677 int vec_any_ne (vector unsigned char, vector unsigned char);
9678 int vec_any_ne (vector bool char, vector bool char);
9679 int vec_any_ne (vector bool char, vector unsigned char);
9680 int vec_any_ne (vector bool char, vector signed char);
9681 int vec_any_ne (vector signed short, vector bool short);
9682 int vec_any_ne (vector signed short, vector signed short);
9683 int vec_any_ne (vector unsigned short, vector bool short);
9684 int vec_any_ne (vector unsigned short, vector unsigned short);
9685 int vec_any_ne (vector bool short, vector bool short);
9686 int vec_any_ne (vector bool short, vector unsigned short);
9687 int vec_any_ne (vector bool short, vector signed short);
9688 int vec_any_ne (vector pixel, vector pixel);
9689 int vec_any_ne (vector signed int, vector bool int);
9690 int vec_any_ne (vector signed int, vector signed int);
9691 int vec_any_ne (vector unsigned int, vector bool int);
9692 int vec_any_ne (vector unsigned int, vector unsigned int);
9693 int vec_any_ne (vector bool int, vector bool int);
9694 int vec_any_ne (vector bool int, vector unsigned int);
9695 int vec_any_ne (vector bool int, vector signed int);
9696 int vec_any_ne (vector float, vector float);
9698 int vec_any_nge (vector float, vector float);
9700 int vec_any_ngt (vector float, vector float);
9702 int vec_any_nle (vector float, vector float);
9704 int vec_any_nlt (vector float, vector float);
9706 int vec_any_numeric (vector float);
9708 int vec_any_out (vector float, vector float);
9711 @node SPARC VIS Built-in Functions
9712 @subsection SPARC VIS Built-in Functions
9714 GCC supports SIMD operations on the SPARC using both the generic vector
9715 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9716 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9717 switch, the VIS extension is exposed as the following built-in functions:
9720 typedef int v2si __attribute__ ((vector_size (8)));
9721 typedef short v4hi __attribute__ ((vector_size (8)));
9722 typedef short v2hi __attribute__ ((vector_size (4)));
9723 typedef char v8qi __attribute__ ((vector_size (8)));
9724 typedef char v4qi __attribute__ ((vector_size (4)));
9726 void * __builtin_vis_alignaddr (void *, long);
9727 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9728 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9729 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9730 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9732 v4hi __builtin_vis_fexpand (v4qi);
9734 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9735 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9736 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9737 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9738 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9739 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9740 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9742 v4qi __builtin_vis_fpack16 (v4hi);
9743 v8qi __builtin_vis_fpack32 (v2si, v2si);
9744 v2hi __builtin_vis_fpackfix (v2si);
9745 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9747 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9750 @node SPU Built-in Functions
9751 @subsection SPU Built-in Functions
9753 GCC provides extensions for the SPU processor as described in the
9754 Sony/Toshiba/IBM SPU Language Extensions Specification, which can be
9755 found at @uref{http://cell.scei.co.jp/} or
9756 @uref{http://www.ibm.com/developerworks/power/cell/}. GCC's
9757 implementation differs in several ways.
9762 The optional extension of specifying vector constants in parentheses is
9766 A vector initializer requires no cast if the vector constant is of the
9767 same type as the variable it is initializing.
9770 If @code{signed} or @code{unsigned} is omitted, the signedness of the
9771 vector type is the default signedness of the base type. The default
9772 varies depending on the operating system, so a portable program should
9773 always specify the signedness.
9776 By default, the keyword @code{__vector} is added. The macro
9777 @code{vector} is defined in @code{<spu_intrinsics.h>} and can be
9781 GCC allows using a @code{typedef} name as the type specifier for a
9785 For C, overloaded functions are implemented with macros so the following
9789 spu_add ((vector signed int)@{1, 2, 3, 4@}, foo);
9792 Since @code{spu_add} is a macro, the vector constant in the example
9793 is treated as four separate arguments. Wrap the entire argument in
9794 parentheses for this to work.
9797 The extended version of @code{__builtin_expect} is not supported.
9801 @emph{Note:} Only the interface described in the aforementioned
9802 specification is supported. Internally, GCC uses built-in functions to
9803 implement the required functionality, but these are not supported and
9804 are subject to change without notice.
9806 @node Target Format Checks
9807 @section Format Checks Specific to Particular Target Machines
9809 For some target machines, GCC supports additional options to the
9811 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9814 * Solaris Format Checks::
9817 @node Solaris Format Checks
9818 @subsection Solaris Format Checks
9820 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9821 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9822 conversions, and the two-argument @code{%b} conversion for displaying
9823 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9826 @section Pragmas Accepted by GCC
9830 GCC supports several types of pragmas, primarily in order to compile
9831 code originally written for other compilers. Note that in general
9832 we do not recommend the use of pragmas; @xref{Function Attributes},
9833 for further explanation.
9838 * RS/6000 and PowerPC Pragmas::
9841 * Symbol-Renaming Pragmas::
9842 * Structure-Packing Pragmas::
9844 * Diagnostic Pragmas::
9845 * Visibility Pragmas::
9849 @subsection ARM Pragmas
9851 The ARM target defines pragmas for controlling the default addition of
9852 @code{long_call} and @code{short_call} attributes to functions.
9853 @xref{Function Attributes}, for information about the effects of these
9858 @cindex pragma, long_calls
9859 Set all subsequent functions to have the @code{long_call} attribute.
9862 @cindex pragma, no_long_calls
9863 Set all subsequent functions to have the @code{short_call} attribute.
9865 @item long_calls_off
9866 @cindex pragma, long_calls_off
9867 Do not affect the @code{long_call} or @code{short_call} attributes of
9868 subsequent functions.
9872 @subsection M32C Pragmas
9875 @item memregs @var{number}
9876 @cindex pragma, memregs
9877 Overrides the command line option @code{-memregs=} for the current
9878 file. Use with care! This pragma must be before any function in the
9879 file, and mixing different memregs values in different objects may
9880 make them incompatible. This pragma is useful when a
9881 performance-critical function uses a memreg for temporary values,
9882 as it may allow you to reduce the number of memregs used.
9886 @node RS/6000 and PowerPC Pragmas
9887 @subsection RS/6000 and PowerPC Pragmas
9889 The RS/6000 and PowerPC targets define one pragma for controlling
9890 whether or not the @code{longcall} attribute is added to function
9891 declarations by default. This pragma overrides the @option{-mlongcall}
9892 option, but not the @code{longcall} and @code{shortcall} attributes.
9893 @xref{RS/6000 and PowerPC Options}, for more information about when long
9894 calls are and are not necessary.
9898 @cindex pragma, longcall
9899 Apply the @code{longcall} attribute to all subsequent function
9903 Do not apply the @code{longcall} attribute to subsequent function
9907 @c Describe c4x pragmas here.
9908 @c Describe h8300 pragmas here.
9909 @c Describe sh pragmas here.
9910 @c Describe v850 pragmas here.
9912 @node Darwin Pragmas
9913 @subsection Darwin Pragmas
9915 The following pragmas are available for all architectures running the
9916 Darwin operating system. These are useful for compatibility with other
9920 @item mark @var{tokens}@dots{}
9921 @cindex pragma, mark
9922 This pragma is accepted, but has no effect.
9924 @item options align=@var{alignment}
9925 @cindex pragma, options align
9926 This pragma sets the alignment of fields in structures. The values of
9927 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9928 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9929 properly; to restore the previous setting, use @code{reset} for the
9932 @item segment @var{tokens}@dots{}
9933 @cindex pragma, segment
9934 This pragma is accepted, but has no effect.
9936 @item unused (@var{var} [, @var{var}]@dots{})
9937 @cindex pragma, unused
9938 This pragma declares variables to be possibly unused. GCC will not
9939 produce warnings for the listed variables. The effect is similar to
9940 that of the @code{unused} attribute, except that this pragma may appear
9941 anywhere within the variables' scopes.
9944 @node Solaris Pragmas
9945 @subsection Solaris Pragmas
9947 The Solaris target supports @code{#pragma redefine_extname}
9948 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
9949 @code{#pragma} directives for compatibility with the system compiler.
9952 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
9953 @cindex pragma, align
9955 Increase the minimum alignment of each @var{variable} to @var{alignment}.
9956 This is the same as GCC's @code{aligned} attribute @pxref{Variable
9957 Attributes}). Macro expansion occurs on the arguments to this pragma
9958 when compiling C and Objective-C. It does not currently occur when
9959 compiling C++, but this is a bug which may be fixed in a future
9962 @item fini (@var{function} [, @var{function}]...)
9963 @cindex pragma, fini
9965 This pragma causes each listed @var{function} to be called after
9966 main, or during shared module unloading, by adding a call to the
9967 @code{.fini} section.
9969 @item init (@var{function} [, @var{function}]...)
9970 @cindex pragma, init
9972 This pragma causes each listed @var{function} to be called during
9973 initialization (before @code{main}) or during shared module loading, by
9974 adding a call to the @code{.init} section.
9978 @node Symbol-Renaming Pragmas
9979 @subsection Symbol-Renaming Pragmas
9981 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
9982 supports two @code{#pragma} directives which change the name used in
9983 assembly for a given declaration. These pragmas are only available on
9984 platforms whose system headers need them. To get this effect on all
9985 platforms supported by GCC, use the asm labels extension (@pxref{Asm
9989 @item redefine_extname @var{oldname} @var{newname}
9990 @cindex pragma, redefine_extname
9992 This pragma gives the C function @var{oldname} the assembly symbol
9993 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
9994 will be defined if this pragma is available (currently only on
9997 @item extern_prefix @var{string}
9998 @cindex pragma, extern_prefix
10000 This pragma causes all subsequent external function and variable
10001 declarations to have @var{string} prepended to their assembly symbols.
10002 This effect may be terminated with another @code{extern_prefix} pragma
10003 whose argument is an empty string. The preprocessor macro
10004 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10005 available (currently only on Tru64 UNIX)@.
10008 These pragmas and the asm labels extension interact in a complicated
10009 manner. Here are some corner cases you may want to be aware of.
10012 @item Both pragmas silently apply only to declarations with external
10013 linkage. Asm labels do not have this restriction.
10015 @item In C++, both pragmas silently apply only to declarations with
10016 ``C'' linkage. Again, asm labels do not have this restriction.
10018 @item If any of the three ways of changing the assembly name of a
10019 declaration is applied to a declaration whose assembly name has
10020 already been determined (either by a previous use of one of these
10021 features, or because the compiler needed the assembly name in order to
10022 generate code), and the new name is different, a warning issues and
10023 the name does not change.
10025 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10026 always the C-language name.
10028 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10029 occurs with an asm label attached, the prefix is silently ignored for
10032 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10033 apply to the same declaration, whichever triggered first wins, and a
10034 warning issues if they contradict each other. (We would like to have
10035 @code{#pragma redefine_extname} always win, for consistency with asm
10036 labels, but if @code{#pragma extern_prefix} triggers first we have no
10037 way of knowing that that happened.)
10040 @node Structure-Packing Pragmas
10041 @subsection Structure-Packing Pragmas
10043 For compatibility with Win32, GCC supports a set of @code{#pragma}
10044 directives which change the maximum alignment of members of structures
10045 (other than zero-width bitfields), unions, and classes subsequently
10046 defined. The @var{n} value below always is required to be a small power
10047 of two and specifies the new alignment in bytes.
10050 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10051 @item @code{#pragma pack()} sets the alignment to the one that was in
10052 effect when compilation started (see also command line option
10053 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10054 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10055 setting on an internal stack and then optionally sets the new alignment.
10056 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10057 saved at the top of the internal stack (and removes that stack entry).
10058 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10059 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10060 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10061 @code{#pragma pack(pop)}.
10064 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10065 @code{#pragma} which lays out a structure as the documented
10066 @code{__attribute__ ((ms_struct))}.
10068 @item @code{#pragma ms_struct on} turns on the layout for structures
10070 @item @code{#pragma ms_struct off} turns off the layout for structures
10072 @item @code{#pragma ms_struct reset} goes back to the default layout.
10076 @subsection Weak Pragmas
10078 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10079 directives for declaring symbols to be weak, and defining weak
10083 @item #pragma weak @var{symbol}
10084 @cindex pragma, weak
10085 This pragma declares @var{symbol} to be weak, as if the declaration
10086 had the attribute of the same name. The pragma may appear before
10087 or after the declaration of @var{symbol}, but must appear before
10088 either its first use or its definition. It is not an error for
10089 @var{symbol} to never be defined at all.
10091 @item #pragma weak @var{symbol1} = @var{symbol2}
10092 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10093 It is an error if @var{symbol2} is not defined in the current
10097 @node Diagnostic Pragmas
10098 @subsection Diagnostic Pragmas
10100 GCC allows the user to selectively enable or disable certain types of
10101 diagnostics, and change the kind of the diagnostic. For example, a
10102 project's policy might require that all sources compile with
10103 @option{-Werror} but certain files might have exceptions allowing
10104 specific types of warnings. Or, a project might selectively enable
10105 diagnostics and treat them as errors depending on which preprocessor
10106 macros are defined.
10109 @item #pragma GCC diagnostic @var{kind} @var{option}
10110 @cindex pragma, diagnostic
10112 Modifies the disposition of a diagnostic. Note that not all
10113 diagnostics are modifyiable; at the moment only warnings (normally
10114 controlled by @samp{-W...}) can be controlled, and not all of them.
10115 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10116 are controllable and which option controls them.
10118 @var{kind} is @samp{error} to treat this diagnostic as an error,
10119 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10120 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10121 @var{option} is a double quoted string which matches the command line
10125 #pragma GCC diagnostic warning "-Wformat"
10126 #pragma GCC diagnostic error "-Walways-true"
10127 #pragma GCC diagnostic ignored "-Walways-true"
10130 Note that these pragmas override any command line options. Also,
10131 while it is syntactically valid to put these pragmas anywhere in your
10132 sources, the only supported location for them is before any data or
10133 functions are defined. Doing otherwise may result in unpredictable
10134 results depending on how the optimizer manages your sources. If the
10135 same option is listed multiple times, the last one specified is the
10136 one that is in effect. This pragma is not intended to be a general
10137 purpose replacement for command line options, but for implementing
10138 strict control over project policies.
10142 @node Visibility Pragmas
10143 @subsection Visibility Pragmas
10146 @item #pragma GCC visibility push(@var{visibility})
10147 @itemx #pragma GCC visibility pop
10148 @cindex pragma, visibility
10150 This pragma allows the user to set the visibility for multiple
10151 declarations without having to give each a visibility attribute
10152 @xref{Function Attributes}, for more information about visibility and
10153 the attribute syntax.
10155 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10156 declarations. Class members and template specializations are not
10157 affected; if you want to override the visibility for a particular
10158 member or instantiation, you must use an attribute.
10162 @node Unnamed Fields
10163 @section Unnamed struct/union fields within structs/unions
10167 For compatibility with other compilers, GCC allows you to define
10168 a structure or union that contains, as fields, structures and unions
10169 without names. For example:
10182 In this example, the user would be able to access members of the unnamed
10183 union with code like @samp{foo.b}. Note that only unnamed structs and
10184 unions are allowed, you may not have, for example, an unnamed
10187 You must never create such structures that cause ambiguous field definitions.
10188 For example, this structure:
10199 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10200 Such constructs are not supported and must be avoided. In the future,
10201 such constructs may be detected and treated as compilation errors.
10203 @opindex fms-extensions
10204 Unless @option{-fms-extensions} is used, the unnamed field must be a
10205 structure or union definition without a tag (for example, @samp{struct
10206 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10207 also be a definition with a tag such as @samp{struct foo @{ int a;
10208 @};}, a reference to a previously defined structure or union such as
10209 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10210 previously defined structure or union type.
10213 @section Thread-Local Storage
10214 @cindex Thread-Local Storage
10215 @cindex @acronym{TLS}
10218 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10219 are allocated such that there is one instance of the variable per extant
10220 thread. The run-time model GCC uses to implement this originates
10221 in the IA-64 processor-specific ABI, but has since been migrated
10222 to other processors as well. It requires significant support from
10223 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10224 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10225 is not available everywhere.
10227 At the user level, the extension is visible with a new storage
10228 class keyword: @code{__thread}. For example:
10232 extern __thread struct state s;
10233 static __thread char *p;
10236 The @code{__thread} specifier may be used alone, with the @code{extern}
10237 or @code{static} specifiers, but with no other storage class specifier.
10238 When used with @code{extern} or @code{static}, @code{__thread} must appear
10239 immediately after the other storage class specifier.
10241 The @code{__thread} specifier may be applied to any global, file-scoped
10242 static, function-scoped static, or static data member of a class. It may
10243 not be applied to block-scoped automatic or non-static data member.
10245 When the address-of operator is applied to a thread-local variable, it is
10246 evaluated at run-time and returns the address of the current thread's
10247 instance of that variable. An address so obtained may be used by any
10248 thread. When a thread terminates, any pointers to thread-local variables
10249 in that thread become invalid.
10251 No static initialization may refer to the address of a thread-local variable.
10253 In C++, if an initializer is present for a thread-local variable, it must
10254 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10257 See @uref{http://people.redhat.com/drepper/tls.pdf,
10258 ELF Handling For Thread-Local Storage} for a detailed explanation of
10259 the four thread-local storage addressing models, and how the run-time
10260 is expected to function.
10263 * C99 Thread-Local Edits::
10264 * C++98 Thread-Local Edits::
10267 @node C99 Thread-Local Edits
10268 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10270 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10271 that document the exact semantics of the language extension.
10275 @cite{5.1.2 Execution environments}
10277 Add new text after paragraph 1
10280 Within either execution environment, a @dfn{thread} is a flow of
10281 control within a program. It is implementation defined whether
10282 or not there may be more than one thread associated with a program.
10283 It is implementation defined how threads beyond the first are
10284 created, the name and type of the function called at thread
10285 startup, and how threads may be terminated. However, objects
10286 with thread storage duration shall be initialized before thread
10291 @cite{6.2.4 Storage durations of objects}
10293 Add new text before paragraph 3
10296 An object whose identifier is declared with the storage-class
10297 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10298 Its lifetime is the entire execution of the thread, and its
10299 stored value is initialized only once, prior to thread startup.
10303 @cite{6.4.1 Keywords}
10305 Add @code{__thread}.
10308 @cite{6.7.1 Storage-class specifiers}
10310 Add @code{__thread} to the list of storage class specifiers in
10313 Change paragraph 2 to
10316 With the exception of @code{__thread}, at most one storage-class
10317 specifier may be given [@dots{}]. The @code{__thread} specifier may
10318 be used alone, or immediately following @code{extern} or
10322 Add new text after paragraph 6
10325 The declaration of an identifier for a variable that has
10326 block scope that specifies @code{__thread} shall also
10327 specify either @code{extern} or @code{static}.
10329 The @code{__thread} specifier shall be used only with
10334 @node C++98 Thread-Local Edits
10335 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10337 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10338 that document the exact semantics of the language extension.
10342 @b{[intro.execution]}
10344 New text after paragraph 4
10347 A @dfn{thread} is a flow of control within the abstract machine.
10348 It is implementation defined whether or not there may be more than
10352 New text after paragraph 7
10355 It is unspecified whether additional action must be taken to
10356 ensure when and whether side effects are visible to other threads.
10362 Add @code{__thread}.
10365 @b{[basic.start.main]}
10367 Add after paragraph 5
10370 The thread that begins execution at the @code{main} function is called
10371 the @dfn{main thread}. It is implementation defined how functions
10372 beginning threads other than the main thread are designated or typed.
10373 A function so designated, as well as the @code{main} function, is called
10374 a @dfn{thread startup function}. It is implementation defined what
10375 happens if a thread startup function returns. It is implementation
10376 defined what happens to other threads when any thread calls @code{exit}.
10380 @b{[basic.start.init]}
10382 Add after paragraph 4
10385 The storage for an object of thread storage duration shall be
10386 statically initialized before the first statement of the thread startup
10387 function. An object of thread storage duration shall not require
10388 dynamic initialization.
10392 @b{[basic.start.term]}
10394 Add after paragraph 3
10397 The type of an object with thread storage duration shall not have a
10398 non-trivial destructor, nor shall it be an array type whose elements
10399 (directly or indirectly) have non-trivial destructors.
10405 Add ``thread storage duration'' to the list in paragraph 1.
10410 Thread, static, and automatic storage durations are associated with
10411 objects introduced by declarations [@dots{}].
10414 Add @code{__thread} to the list of specifiers in paragraph 3.
10417 @b{[basic.stc.thread]}
10419 New section before @b{[basic.stc.static]}
10422 The keyword @code{__thread} applied to a non-local object gives the
10423 object thread storage duration.
10425 A local variable or class data member declared both @code{static}
10426 and @code{__thread} gives the variable or member thread storage
10431 @b{[basic.stc.static]}
10436 All objects which have neither thread storage duration, dynamic
10437 storage duration nor are local [@dots{}].
10443 Add @code{__thread} to the list in paragraph 1.
10448 With the exception of @code{__thread}, at most one
10449 @var{storage-class-specifier} shall appear in a given
10450 @var{decl-specifier-seq}. The @code{__thread} specifier may
10451 be used alone, or immediately following the @code{extern} or
10452 @code{static} specifiers. [@dots{}]
10455 Add after paragraph 5
10458 The @code{__thread} specifier can be applied only to the names of objects
10459 and to anonymous unions.
10465 Add after paragraph 6
10468 Non-@code{static} members shall not be @code{__thread}.
10472 @node C++ Extensions
10473 @chapter Extensions to the C++ Language
10474 @cindex extensions, C++ language
10475 @cindex C++ language extensions
10477 The GNU compiler provides these extensions to the C++ language (and you
10478 can also use most of the C language extensions in your C++ programs). If you
10479 want to write code that checks whether these features are available, you can
10480 test for the GNU compiler the same way as for C programs: check for a
10481 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10482 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10483 Predefined Macros,cpp,The GNU C Preprocessor}).
10486 * Volatiles:: What constitutes an access to a volatile object.
10487 * Restricted Pointers:: C99 restricted pointers and references.
10488 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10489 * C++ Interface:: You can use a single C++ header file for both
10490 declarations and definitions.
10491 * Template Instantiation:: Methods for ensuring that exactly one copy of
10492 each needed template instantiation is emitted.
10493 * Bound member functions:: You can extract a function pointer to the
10494 method denoted by a @samp{->*} or @samp{.*} expression.
10495 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10496 * Namespace Association:: Strong using-directives for namespace association.
10497 * Java Exceptions:: Tweaking exception handling to work with Java.
10498 * Deprecated Features:: Things will disappear from g++.
10499 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10503 @section When is a Volatile Object Accessed?
10504 @cindex accessing volatiles
10505 @cindex volatile read
10506 @cindex volatile write
10507 @cindex volatile access
10509 Both the C and C++ standard have the concept of volatile objects. These
10510 are normally accessed by pointers and used for accessing hardware. The
10511 standards encourage compilers to refrain from optimizations concerning
10512 accesses to volatile objects. The C standard leaves it implementation
10513 defined as to what constitutes a volatile access. The C++ standard omits
10514 to specify this, except to say that C++ should behave in a similar manner
10515 to C with respect to volatiles, where possible. The minimum either
10516 standard specifies is that at a sequence point all previous accesses to
10517 volatile objects have stabilized and no subsequent accesses have
10518 occurred. Thus an implementation is free to reorder and combine
10519 volatile accesses which occur between sequence points, but cannot do so
10520 for accesses across a sequence point. The use of volatiles does not
10521 allow you to violate the restriction on updating objects multiple times
10522 within a sequence point.
10524 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10526 The behavior differs slightly between C and C++ in the non-obvious cases:
10529 volatile int *src = @var{somevalue};
10533 With C, such expressions are rvalues, and GCC interprets this either as a
10534 read of the volatile object being pointed to or only as request to evaluate
10535 the side-effects. The C++ standard specifies that such expressions do not
10536 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10537 object may be incomplete. The C++ standard does not specify explicitly
10538 that it is this lvalue to rvalue conversion which may be responsible for
10539 causing an access. However, there is reason to believe that it is,
10540 because otherwise certain simple expressions become undefined. However,
10541 because it would surprise most programmers, G++ treats dereferencing a
10542 pointer to volatile object of complete type when the value is unused as
10543 GCC would do for an equivalent type in C. When the object has incomplete
10544 type, G++ issues a warning; if you wish to force an error, you must
10545 force a conversion to rvalue with, for instance, a static cast.
10547 When using a reference to volatile, G++ does not treat equivalent
10548 expressions as accesses to volatiles, but instead issues a warning that
10549 no volatile is accessed. The rationale for this is that otherwise it
10550 becomes difficult to determine where volatile access occur, and not
10551 possible to ignore the return value from functions returning volatile
10552 references. Again, if you wish to force a read, cast the reference to
10555 @node Restricted Pointers
10556 @section Restricting Pointer Aliasing
10557 @cindex restricted pointers
10558 @cindex restricted references
10559 @cindex restricted this pointer
10561 As with the C front end, G++ understands the C99 feature of restricted pointers,
10562 specified with the @code{__restrict__}, or @code{__restrict} type
10563 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10564 language flag, @code{restrict} is not a keyword in C++.
10566 In addition to allowing restricted pointers, you can specify restricted
10567 references, which indicate that the reference is not aliased in the local
10571 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10578 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10579 @var{rref} refers to a (different) unaliased integer.
10581 You may also specify whether a member function's @var{this} pointer is
10582 unaliased by using @code{__restrict__} as a member function qualifier.
10585 void T::fn () __restrict__
10592 Within the body of @code{T::fn}, @var{this} will have the effective
10593 definition @code{T *__restrict__ const this}. Notice that the
10594 interpretation of a @code{__restrict__} member function qualifier is
10595 different to that of @code{const} or @code{volatile} qualifier, in that it
10596 is applied to the pointer rather than the object. This is consistent with
10597 other compilers which implement restricted pointers.
10599 As with all outermost parameter qualifiers, @code{__restrict__} is
10600 ignored in function definition matching. This means you only need to
10601 specify @code{__restrict__} in a function definition, rather than
10602 in a function prototype as well.
10604 @node Vague Linkage
10605 @section Vague Linkage
10606 @cindex vague linkage
10608 There are several constructs in C++ which require space in the object
10609 file but are not clearly tied to a single translation unit. We say that
10610 these constructs have ``vague linkage''. Typically such constructs are
10611 emitted wherever they are needed, though sometimes we can be more
10615 @item Inline Functions
10616 Inline functions are typically defined in a header file which can be
10617 included in many different compilations. Hopefully they can usually be
10618 inlined, but sometimes an out-of-line copy is necessary, if the address
10619 of the function is taken or if inlining fails. In general, we emit an
10620 out-of-line copy in all translation units where one is needed. As an
10621 exception, we only emit inline virtual functions with the vtable, since
10622 it will always require a copy.
10624 Local static variables and string constants used in an inline function
10625 are also considered to have vague linkage, since they must be shared
10626 between all inlined and out-of-line instances of the function.
10630 C++ virtual functions are implemented in most compilers using a lookup
10631 table, known as a vtable. The vtable contains pointers to the virtual
10632 functions provided by a class, and each object of the class contains a
10633 pointer to its vtable (or vtables, in some multiple-inheritance
10634 situations). If the class declares any non-inline, non-pure virtual
10635 functions, the first one is chosen as the ``key method'' for the class,
10636 and the vtable is only emitted in the translation unit where the key
10639 @emph{Note:} If the chosen key method is later defined as inline, the
10640 vtable will still be emitted in every translation unit which defines it.
10641 Make sure that any inline virtuals are declared inline in the class
10642 body, even if they are not defined there.
10644 @item type_info objects
10647 C++ requires information about types to be written out in order to
10648 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10649 For polymorphic classes (classes with virtual functions), the type_info
10650 object is written out along with the vtable so that @samp{dynamic_cast}
10651 can determine the dynamic type of a class object at runtime. For all
10652 other types, we write out the type_info object when it is used: when
10653 applying @samp{typeid} to an expression, throwing an object, or
10654 referring to a type in a catch clause or exception specification.
10656 @item Template Instantiations
10657 Most everything in this section also applies to template instantiations,
10658 but there are other options as well.
10659 @xref{Template Instantiation,,Where's the Template?}.
10663 When used with GNU ld version 2.8 or later on an ELF system such as
10664 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10665 these constructs will be discarded at link time. This is known as
10668 On targets that don't support COMDAT, but do support weak symbols, GCC
10669 will use them. This way one copy will override all the others, but
10670 the unused copies will still take up space in the executable.
10672 For targets which do not support either COMDAT or weak symbols,
10673 most entities with vague linkage will be emitted as local symbols to
10674 avoid duplicate definition errors from the linker. This will not happen
10675 for local statics in inlines, however, as having multiple copies will
10676 almost certainly break things.
10678 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10679 another way to control placement of these constructs.
10681 @node C++ Interface
10682 @section #pragma interface and implementation
10684 @cindex interface and implementation headers, C++
10685 @cindex C++ interface and implementation headers
10686 @cindex pragmas, interface and implementation
10688 @code{#pragma interface} and @code{#pragma implementation} provide the
10689 user with a way of explicitly directing the compiler to emit entities
10690 with vague linkage (and debugging information) in a particular
10693 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10694 most cases, because of COMDAT support and the ``key method'' heuristic
10695 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10696 program to grow due to unnecessary out-of-line copies of inline
10697 functions. Currently (3.4) the only benefit of these
10698 @code{#pragma}s is reduced duplication of debugging information, and
10699 that should be addressed soon on DWARF 2 targets with the use of
10703 @item #pragma interface
10704 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10705 @kindex #pragma interface
10706 Use this directive in @emph{header files} that define object classes, to save
10707 space in most of the object files that use those classes. Normally,
10708 local copies of certain information (backup copies of inline member
10709 functions, debugging information, and the internal tables that implement
10710 virtual functions) must be kept in each object file that includes class
10711 definitions. You can use this pragma to avoid such duplication. When a
10712 header file containing @samp{#pragma interface} is included in a
10713 compilation, this auxiliary information will not be generated (unless
10714 the main input source file itself uses @samp{#pragma implementation}).
10715 Instead, the object files will contain references to be resolved at link
10718 The second form of this directive is useful for the case where you have
10719 multiple headers with the same name in different directories. If you
10720 use this form, you must specify the same string to @samp{#pragma
10723 @item #pragma implementation
10724 @itemx #pragma implementation "@var{objects}.h"
10725 @kindex #pragma implementation
10726 Use this pragma in a @emph{main input file}, when you want full output from
10727 included header files to be generated (and made globally visible). The
10728 included header file, in turn, should use @samp{#pragma interface}.
10729 Backup copies of inline member functions, debugging information, and the
10730 internal tables used to implement virtual functions are all generated in
10731 implementation files.
10733 @cindex implied @code{#pragma implementation}
10734 @cindex @code{#pragma implementation}, implied
10735 @cindex naming convention, implementation headers
10736 If you use @samp{#pragma implementation} with no argument, it applies to
10737 an include file with the same basename@footnote{A file's @dfn{basename}
10738 was the name stripped of all leading path information and of trailing
10739 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10740 file. For example, in @file{allclass.cc}, giving just
10741 @samp{#pragma implementation}
10742 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10744 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10745 an implementation file whenever you would include it from
10746 @file{allclass.cc} even if you never specified @samp{#pragma
10747 implementation}. This was deemed to be more trouble than it was worth,
10748 however, and disabled.
10750 Use the string argument if you want a single implementation file to
10751 include code from multiple header files. (You must also use
10752 @samp{#include} to include the header file; @samp{#pragma
10753 implementation} only specifies how to use the file---it doesn't actually
10756 There is no way to split up the contents of a single header file into
10757 multiple implementation files.
10760 @cindex inlining and C++ pragmas
10761 @cindex C++ pragmas, effect on inlining
10762 @cindex pragmas in C++, effect on inlining
10763 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10764 effect on function inlining.
10766 If you define a class in a header file marked with @samp{#pragma
10767 interface}, the effect on an inline function defined in that class is
10768 similar to an explicit @code{extern} declaration---the compiler emits
10769 no code at all to define an independent version of the function. Its
10770 definition is used only for inlining with its callers.
10772 @opindex fno-implement-inlines
10773 Conversely, when you include the same header file in a main source file
10774 that declares it as @samp{#pragma implementation}, the compiler emits
10775 code for the function itself; this defines a version of the function
10776 that can be found via pointers (or by callers compiled without
10777 inlining). If all calls to the function can be inlined, you can avoid
10778 emitting the function by compiling with @option{-fno-implement-inlines}.
10779 If any calls were not inlined, you will get linker errors.
10781 @node Template Instantiation
10782 @section Where's the Template?
10783 @cindex template instantiation
10785 C++ templates are the first language feature to require more
10786 intelligence from the environment than one usually finds on a UNIX
10787 system. Somehow the compiler and linker have to make sure that each
10788 template instance occurs exactly once in the executable if it is needed,
10789 and not at all otherwise. There are two basic approaches to this
10790 problem, which are referred to as the Borland model and the Cfront model.
10793 @item Borland model
10794 Borland C++ solved the template instantiation problem by adding the code
10795 equivalent of common blocks to their linker; the compiler emits template
10796 instances in each translation unit that uses them, and the linker
10797 collapses them together. The advantage of this model is that the linker
10798 only has to consider the object files themselves; there is no external
10799 complexity to worry about. This disadvantage is that compilation time
10800 is increased because the template code is being compiled repeatedly.
10801 Code written for this model tends to include definitions of all
10802 templates in the header file, since they must be seen to be
10806 The AT&T C++ translator, Cfront, solved the template instantiation
10807 problem by creating the notion of a template repository, an
10808 automatically maintained place where template instances are stored. A
10809 more modern version of the repository works as follows: As individual
10810 object files are built, the compiler places any template definitions and
10811 instantiations encountered in the repository. At link time, the link
10812 wrapper adds in the objects in the repository and compiles any needed
10813 instances that were not previously emitted. The advantages of this
10814 model are more optimal compilation speed and the ability to use the
10815 system linker; to implement the Borland model a compiler vendor also
10816 needs to replace the linker. The disadvantages are vastly increased
10817 complexity, and thus potential for error; for some code this can be
10818 just as transparent, but in practice it can been very difficult to build
10819 multiple programs in one directory and one program in multiple
10820 directories. Code written for this model tends to separate definitions
10821 of non-inline member templates into a separate file, which should be
10822 compiled separately.
10825 When used with GNU ld version 2.8 or later on an ELF system such as
10826 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10827 Borland model. On other systems, G++ implements neither automatic
10830 A future version of G++ will support a hybrid model whereby the compiler
10831 will emit any instantiations for which the template definition is
10832 included in the compile, and store template definitions and
10833 instantiation context information into the object file for the rest.
10834 The link wrapper will extract that information as necessary and invoke
10835 the compiler to produce the remaining instantiations. The linker will
10836 then combine duplicate instantiations.
10838 In the mean time, you have the following options for dealing with
10839 template instantiations:
10844 Compile your template-using code with @option{-frepo}. The compiler will
10845 generate files with the extension @samp{.rpo} listing all of the
10846 template instantiations used in the corresponding object files which
10847 could be instantiated there; the link wrapper, @samp{collect2}, will
10848 then update the @samp{.rpo} files to tell the compiler where to place
10849 those instantiations and rebuild any affected object files. The
10850 link-time overhead is negligible after the first pass, as the compiler
10851 will continue to place the instantiations in the same files.
10853 This is your best option for application code written for the Borland
10854 model, as it will just work. Code written for the Cfront model will
10855 need to be modified so that the template definitions are available at
10856 one or more points of instantiation; usually this is as simple as adding
10857 @code{#include <tmethods.cc>} to the end of each template header.
10859 For library code, if you want the library to provide all of the template
10860 instantiations it needs, just try to link all of its object files
10861 together; the link will fail, but cause the instantiations to be
10862 generated as a side effect. Be warned, however, that this may cause
10863 conflicts if multiple libraries try to provide the same instantiations.
10864 For greater control, use explicit instantiation as described in the next
10868 @opindex fno-implicit-templates
10869 Compile your code with @option{-fno-implicit-templates} to disable the
10870 implicit generation of template instances, and explicitly instantiate
10871 all the ones you use. This approach requires more knowledge of exactly
10872 which instances you need than do the others, but it's less
10873 mysterious and allows greater control. You can scatter the explicit
10874 instantiations throughout your program, perhaps putting them in the
10875 translation units where the instances are used or the translation units
10876 that define the templates themselves; you can put all of the explicit
10877 instantiations you need into one big file; or you can create small files
10884 template class Foo<int>;
10885 template ostream& operator <<
10886 (ostream&, const Foo<int>&);
10889 for each of the instances you need, and create a template instantiation
10890 library from those.
10892 If you are using Cfront-model code, you can probably get away with not
10893 using @option{-fno-implicit-templates} when compiling files that don't
10894 @samp{#include} the member template definitions.
10896 If you use one big file to do the instantiations, you may want to
10897 compile it without @option{-fno-implicit-templates} so you get all of the
10898 instances required by your explicit instantiations (but not by any
10899 other files) without having to specify them as well.
10901 G++ has extended the template instantiation syntax given in the ISO
10902 standard to allow forward declaration of explicit instantiations
10903 (with @code{extern}), instantiation of the compiler support data for a
10904 template class (i.e.@: the vtable) without instantiating any of its
10905 members (with @code{inline}), and instantiation of only the static data
10906 members of a template class, without the support data or member
10907 functions (with (@code{static}):
10910 extern template int max (int, int);
10911 inline template class Foo<int>;
10912 static template class Foo<int>;
10916 Do nothing. Pretend G++ does implement automatic instantiation
10917 management. Code written for the Borland model will work fine, but
10918 each translation unit will contain instances of each of the templates it
10919 uses. In a large program, this can lead to an unacceptable amount of code
10923 @node Bound member functions
10924 @section Extracting the function pointer from a bound pointer to member function
10926 @cindex pointer to member function
10927 @cindex bound pointer to member function
10929 In C++, pointer to member functions (PMFs) are implemented using a wide
10930 pointer of sorts to handle all the possible call mechanisms; the PMF
10931 needs to store information about how to adjust the @samp{this} pointer,
10932 and if the function pointed to is virtual, where to find the vtable, and
10933 where in the vtable to look for the member function. If you are using
10934 PMFs in an inner loop, you should really reconsider that decision. If
10935 that is not an option, you can extract the pointer to the function that
10936 would be called for a given object/PMF pair and call it directly inside
10937 the inner loop, to save a bit of time.
10939 Note that you will still be paying the penalty for the call through a
10940 function pointer; on most modern architectures, such a call defeats the
10941 branch prediction features of the CPU@. This is also true of normal
10942 virtual function calls.
10944 The syntax for this extension is
10948 extern int (A::*fp)();
10949 typedef int (*fptr)(A *);
10951 fptr p = (fptr)(a.*fp);
10954 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
10955 no object is needed to obtain the address of the function. They can be
10956 converted to function pointers directly:
10959 fptr p1 = (fptr)(&A::foo);
10962 @opindex Wno-pmf-conversions
10963 You must specify @option{-Wno-pmf-conversions} to use this extension.
10965 @node C++ Attributes
10966 @section C++-Specific Variable, Function, and Type Attributes
10968 Some attributes only make sense for C++ programs.
10971 @item init_priority (@var{priority})
10972 @cindex init_priority attribute
10975 In Standard C++, objects defined at namespace scope are guaranteed to be
10976 initialized in an order in strict accordance with that of their definitions
10977 @emph{in a given translation unit}. No guarantee is made for initializations
10978 across translation units. However, GNU C++ allows users to control the
10979 order of initialization of objects defined at namespace scope with the
10980 @code{init_priority} attribute by specifying a relative @var{priority},
10981 a constant integral expression currently bounded between 101 and 65535
10982 inclusive. Lower numbers indicate a higher priority.
10984 In the following example, @code{A} would normally be created before
10985 @code{B}, but the @code{init_priority} attribute has reversed that order:
10988 Some_Class A __attribute__ ((init_priority (2000)));
10989 Some_Class B __attribute__ ((init_priority (543)));
10993 Note that the particular values of @var{priority} do not matter; only their
10996 @item java_interface
10997 @cindex java_interface attribute
10999 This type attribute informs C++ that the class is a Java interface. It may
11000 only be applied to classes declared within an @code{extern "Java"} block.
11001 Calls to methods declared in this interface will be dispatched using GCJ's
11002 interface table mechanism, instead of regular virtual table dispatch.
11006 See also @xref{Namespace Association}.
11008 @node Namespace Association
11009 @section Namespace Association
11011 @strong{Caution:} The semantics of this extension are not fully
11012 defined. Users should refrain from using this extension as its
11013 semantics may change subtly over time. It is possible that this
11014 extension will be removed in future versions of G++.
11016 A using-directive with @code{__attribute ((strong))} is stronger
11017 than a normal using-directive in two ways:
11021 Templates from the used namespace can be specialized and explicitly
11022 instantiated as though they were members of the using namespace.
11025 The using namespace is considered an associated namespace of all
11026 templates in the used namespace for purposes of argument-dependent
11030 The used namespace must be nested within the using namespace so that
11031 normal unqualified lookup works properly.
11033 This is useful for composing a namespace transparently from
11034 implementation namespaces. For example:
11039 template <class T> struct A @{ @};
11041 using namespace debug __attribute ((__strong__));
11042 template <> struct A<int> @{ @}; // @r{ok to specialize}
11044 template <class T> void f (A<T>);
11049 f (std::A<float>()); // @r{lookup finds} std::f
11054 @node Java Exceptions
11055 @section Java Exceptions
11057 The Java language uses a slightly different exception handling model
11058 from C++. Normally, GNU C++ will automatically detect when you are
11059 writing C++ code that uses Java exceptions, and handle them
11060 appropriately. However, if C++ code only needs to execute destructors
11061 when Java exceptions are thrown through it, GCC will guess incorrectly.
11062 Sample problematic code is:
11065 struct S @{ ~S(); @};
11066 extern void bar(); // @r{is written in Java, and may throw exceptions}
11075 The usual effect of an incorrect guess is a link failure, complaining of
11076 a missing routine called @samp{__gxx_personality_v0}.
11078 You can inform the compiler that Java exceptions are to be used in a
11079 translation unit, irrespective of what it might think, by writing
11080 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11081 @samp{#pragma} must appear before any functions that throw or catch
11082 exceptions, or run destructors when exceptions are thrown through them.
11084 You cannot mix Java and C++ exceptions in the same translation unit. It
11085 is believed to be safe to throw a C++ exception from one file through
11086 another file compiled for the Java exception model, or vice versa, but
11087 there may be bugs in this area.
11089 @node Deprecated Features
11090 @section Deprecated Features
11092 In the past, the GNU C++ compiler was extended to experiment with new
11093 features, at a time when the C++ language was still evolving. Now that
11094 the C++ standard is complete, some of those features are superseded by
11095 superior alternatives. Using the old features might cause a warning in
11096 some cases that the feature will be dropped in the future. In other
11097 cases, the feature might be gone already.
11099 While the list below is not exhaustive, it documents some of the options
11100 that are now deprecated:
11103 @item -fexternal-templates
11104 @itemx -falt-external-templates
11105 These are two of the many ways for G++ to implement template
11106 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11107 defines how template definitions have to be organized across
11108 implementation units. G++ has an implicit instantiation mechanism that
11109 should work just fine for standard-conforming code.
11111 @item -fstrict-prototype
11112 @itemx -fno-strict-prototype
11113 Previously it was possible to use an empty prototype parameter list to
11114 indicate an unspecified number of parameters (like C), rather than no
11115 parameters, as C++ demands. This feature has been removed, except where
11116 it is required for backwards compatibility @xref{Backwards Compatibility}.
11119 G++ allows a virtual function returning @samp{void *} to be overridden
11120 by one returning a different pointer type. This extension to the
11121 covariant return type rules is now deprecated and will be removed from a
11124 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11125 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11126 and will be removed in a future version. Code using these operators
11127 should be modified to use @code{std::min} and @code{std::max} instead.
11129 The named return value extension has been deprecated, and is now
11132 The use of initializer lists with new expressions has been deprecated,
11133 and is now removed from G++.
11135 Floating and complex non-type template parameters have been deprecated,
11136 and are now removed from G++.
11138 The implicit typename extension has been deprecated and is now
11141 The use of default arguments in function pointers, function typedefs and
11142 and other places where they are not permitted by the standard is
11143 deprecated and will be removed from a future version of G++.
11145 G++ allows floating-point literals to appear in integral constant expressions,
11146 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11147 This extension is deprecated and will be removed from a future version.
11149 G++ allows static data members of const floating-point type to be declared
11150 with an initializer in a class definition. The standard only allows
11151 initializers for static members of const integral types and const
11152 enumeration types so this extension has been deprecated and will be removed
11153 from a future version.
11155 @node Backwards Compatibility
11156 @section Backwards Compatibility
11157 @cindex Backwards Compatibility
11158 @cindex ARM [Annotated C++ Reference Manual]
11160 Now that there is a definitive ISO standard C++, G++ has a specification
11161 to adhere to. The C++ language evolved over time, and features that
11162 used to be acceptable in previous drafts of the standard, such as the ARM
11163 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11164 compilation of C++ written to such drafts, G++ contains some backwards
11165 compatibilities. @emph{All such backwards compatibility features are
11166 liable to disappear in future versions of G++.} They should be considered
11167 deprecated @xref{Deprecated Features}.
11171 If a variable is declared at for scope, it used to remain in scope until
11172 the end of the scope which contained the for statement (rather than just
11173 within the for scope). G++ retains this, but issues a warning, if such a
11174 variable is accessed outside the for scope.
11176 @item Implicit C language
11177 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11178 scope to set the language. On such systems, all header files are
11179 implicitly scoped inside a C language scope. Also, an empty prototype
11180 @code{()} will be treated as an unspecified number of arguments, rather
11181 than no arguments, as C++ demands.