1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000,
2 @c 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Extensions to the C Language Family
9 @cindex extensions, C language
10 @cindex C language extensions
13 GNU C provides several language features not found in ISO standard C@.
14 (The @option{-pedantic} option directs GCC to print a warning message if
15 any of these features is used.) To test for the availability of these
16 features in conditional compilation, check for a predefined macro
17 @code{__GNUC__}, which is always defined under GCC@.
19 These extensions are available in C 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 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
62 * Label Attributes:: Specifying attributes of labels and statements.
63 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
64 * Alignment:: Inquiring about the alignment of a type or variable.
65 * Inline:: Defining inline functions (as fast as macros).
66 * Extended Asm:: Assembler instructions with C expressions as operands.
67 (With them you can define ``built-in'' functions.)
68 * Constraints:: Constraints for asm operands
69 * Asm Labels:: Specifying the assembler name to use for a C symbol.
70 * Explicit Reg Vars:: Defining variables residing in specified registers.
71 * Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files.
72 * Incomplete Enums:: @code{enum foo;}, with details to follow.
73 * Function Names:: Printable strings which are the name of the current
75 * Return Address:: Getting the return or frame address of a function.
76 * Vector Extensions:: Using vector instructions through built-in functions.
77 * Offsetof:: Special syntax for implementing @code{offsetof}.
78 * Atomic Builtins:: Built-in functions for atomic memory access.
79 * Object Size Checking:: Built-in functions for limited buffer overflow
81 * Other Builtins:: Other built-in functions.
82 * Target Builtins:: Built-in functions specific to particular targets.
83 * Target Format Checks:: Format checks specific to particular targets.
84 * Pragmas:: Pragmas accepted by GCC.
85 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
86 * Thread-Local:: Per-thread variables.
87 * Binary constants:: Binary constants using the @samp{0b} prefix.
91 @section Statements and Declarations in Expressions
92 @cindex statements inside expressions
93 @cindex declarations inside expressions
94 @cindex expressions containing statements
95 @cindex macros, statements in expressions
97 @c the above section title wrapped and causes an underfull hbox.. i
98 @c changed it from "within" to "in". --mew 4feb93
99 A compound statement enclosed in parentheses may appear as an expression
100 in GNU C@. This allows you to use loops, switches, and local variables
101 within an expression.
103 Recall that a compound statement is a sequence of statements surrounded
104 by braces; in this construct, parentheses go around the braces. For
108 (@{ int y = foo (); int z;
115 is a valid (though slightly more complex than necessary) expression
116 for the absolute value of @code{foo ()}.
118 The last thing in the compound statement should be an expression
119 followed by a semicolon; the value of this subexpression serves as the
120 value of the entire construct. (If you use some other kind of statement
121 last within the braces, the construct has type @code{void}, and thus
122 effectively no value.)
124 This feature is especially useful in making macro definitions ``safe'' (so
125 that they evaluate each operand exactly once). For example, the
126 ``maximum'' function is commonly defined as a macro in standard C as
130 #define max(a,b) ((a) > (b) ? (a) : (b))
134 @cindex side effects, macro argument
135 But this definition computes either @var{a} or @var{b} twice, with bad
136 results if the operand has side effects. In GNU C, if you know the
137 type of the operands (here taken as @code{int}), you can define
138 the macro safely as follows:
141 #define maxint(a,b) \
142 (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
145 Embedded statements are not allowed in constant expressions, such as
146 the value of an enumeration constant, the width of a bit-field, or
147 the initial value of a static variable.
149 If you don't know the type of the operand, you can still do this, but you
150 must use @code{typeof} (@pxref{Typeof}).
152 In G++, the result value of a statement expression undergoes array and
153 function pointer decay, and is returned by value to the enclosing
154 expression. For instance, if @code{A} is a class, then
163 will construct a temporary @code{A} object to hold the result of the
164 statement expression, and that will be used to invoke @code{Foo}.
165 Therefore the @code{this} pointer observed by @code{Foo} will not be the
168 Any temporaries created within a statement within a statement expression
169 will be destroyed at the statement's end. This makes statement
170 expressions inside macros slightly different from function calls. In
171 the latter case temporaries introduced during argument evaluation will
172 be destroyed at the end of the statement that includes the function
173 call. In the statement expression case they will be destroyed during
174 the statement expression. For instance,
177 #define macro(a) (@{__typeof__(a) b = (a); b + 3; @})
178 template<typename T> T function(T a) @{ T b = a; return b + 3; @}
188 will have different places where temporaries are destroyed. For the
189 @code{macro} case, the temporary @code{X} will be destroyed just after
190 the initialization of @code{b}. In the @code{function} case that
191 temporary will be destroyed when the function returns.
193 These considerations mean that it is probably a bad idea to use
194 statement-expressions of this form in header files that are designed to
195 work with C++. (Note that some versions of the GNU C Library contained
196 header files using statement-expression that lead to precisely this
199 Jumping into a statement expression with @code{goto} or using a
200 @code{switch} statement outside the statement expression with a
201 @code{case} or @code{default} label inside the statement expression is
202 not permitted. Jumping into a statement expression with a computed
203 @code{goto} (@pxref{Labels as Values}) yields undefined behavior.
204 Jumping out of a statement expression is permitted, but if the
205 statement expression is part of a larger expression then it is
206 unspecified which other subexpressions of that expression have been
207 evaluated except where the language definition requires certain
208 subexpressions to be evaluated before or after the statement
209 expression. In any case, as with a function call the evaluation of a
210 statement expression is not interleaved with the evaluation of other
211 parts of the containing expression. For example,
214 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
218 will call @code{foo} and @code{bar1} and will not call @code{baz} but
219 may or may not call @code{bar2}. If @code{bar2} is called, it will be
220 called after @code{foo} and before @code{bar1}
223 @section Locally Declared Labels
225 @cindex macros, local labels
227 GCC allows you to declare @dfn{local labels} in any nested block
228 scope. A local label is just like an ordinary label, but you can
229 only reference it (with a @code{goto} statement, or by taking its
230 address) within the block in which it was declared.
232 A local label declaration looks like this:
235 __label__ @var{label};
242 __label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
245 Local label declarations must come at the beginning of the block,
246 before any ordinary declarations or statements.
248 The label declaration defines the label @emph{name}, but does not define
249 the label itself. You must do this in the usual way, with
250 @code{@var{label}:}, within the statements of the statement expression.
252 The local label feature is useful for complex macros. If a macro
253 contains nested loops, a @code{goto} can be useful for breaking out of
254 them. However, an ordinary label whose scope is the whole function
255 cannot be used: if the macro can be expanded several times in one
256 function, the label will be multiply defined in that function. A
257 local label avoids this problem. For example:
260 #define SEARCH(value, array, target) \
263 typeof (target) _SEARCH_target = (target); \
264 typeof (*(array)) *_SEARCH_array = (array); \
267 for (i = 0; i < max; i++) \
268 for (j = 0; j < max; j++) \
269 if (_SEARCH_array[i][j] == _SEARCH_target) \
270 @{ (value) = i; goto found; @} \
276 This could also be written using a statement-expression:
279 #define SEARCH(array, target) \
282 typeof (target) _SEARCH_target = (target); \
283 typeof (*(array)) *_SEARCH_array = (array); \
286 for (i = 0; i < max; i++) \
287 for (j = 0; j < max; j++) \
288 if (_SEARCH_array[i][j] == _SEARCH_target) \
289 @{ value = i; goto found; @} \
296 Local label declarations also make the labels they declare visible to
297 nested functions, if there are any. @xref{Nested Functions}, for details.
299 @node Labels as Values
300 @section Labels as Values
301 @cindex labels as values
302 @cindex computed gotos
303 @cindex goto with computed label
304 @cindex address of a label
306 You can get the address of a label defined in the current function
307 (or a containing function) with the unary operator @samp{&&}. The
308 value has type @code{void *}. This value is a constant and can be used
309 wherever a constant of that type is valid. For example:
317 To use these values, you need to be able to jump to one. This is done
318 with the computed goto statement@footnote{The analogous feature in
319 Fortran is called an assigned goto, but that name seems inappropriate in
320 C, where one can do more than simply store label addresses in label
321 variables.}, @code{goto *@var{exp};}. For example,
328 Any expression of type @code{void *} is allowed.
330 One way of using these constants is in initializing a static array that
331 will serve as a jump table:
334 static void *array[] = @{ &&foo, &&bar, &&hack @};
337 Then you can select a label with indexing, like this:
344 Note that this does not check whether the subscript is in bounds---array
345 indexing in C never does that.
347 Such an array of label values serves a purpose much like that of the
348 @code{switch} statement. The @code{switch} statement is cleaner, so
349 use that rather than an array unless the problem does not fit a
350 @code{switch} statement very well.
352 Another use of label values is in an interpreter for threaded code.
353 The labels within the interpreter function can be stored in the
354 threaded code for super-fast dispatching.
356 You may not use this mechanism to jump to code in a different function.
357 If you do that, totally unpredictable things will happen. The best way to
358 avoid this is to store the label address only in automatic variables and
359 never pass it as an argument.
361 An alternate way to write the above example is
364 static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
366 goto *(&&foo + array[i]);
370 This is more friendly to code living in shared libraries, as it reduces
371 the number of dynamic relocations that are needed, and by consequence,
372 allows the data to be read-only.
374 @node Nested Functions
375 @section Nested Functions
376 @cindex nested functions
377 @cindex downward funargs
380 A @dfn{nested function} is a function defined inside another function.
381 (Nested functions are not supported for GNU C++.) The nested function's
382 name is local to the block where it is defined. For example, here we
383 define a nested function named @code{square}, and call it twice:
387 foo (double a, double b)
389 double square (double z) @{ return z * z; @}
391 return square (a) + square (b);
396 The nested function can access all the variables of the containing
397 function that are visible at the point of its definition. This is
398 called @dfn{lexical scoping}. For example, here we show a nested
399 function which uses an inherited variable named @code{offset}:
403 bar (int *array, int offset, int size)
405 int access (int *array, int index)
406 @{ return array[index + offset]; @}
409 for (i = 0; i < size; i++)
410 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
415 Nested function definitions are permitted within functions in the places
416 where variable definitions are allowed; that is, in any block, mixed
417 with the other declarations and statements in the block.
419 It is possible to call the nested function from outside the scope of its
420 name by storing its address or passing the address to another function:
423 hack (int *array, int size)
425 void store (int index, int value)
426 @{ array[index] = value; @}
428 intermediate (store, size);
432 Here, the function @code{intermediate} receives the address of
433 @code{store} as an argument. If @code{intermediate} calls @code{store},
434 the arguments given to @code{store} are used to store into @code{array}.
435 But this technique works only so long as the containing function
436 (@code{hack}, in this example) does not exit.
438 If you try to call the nested function through its address after the
439 containing function has exited, all hell will break loose. If you try
440 to call it after a containing scope level has exited, and if it refers
441 to some of the variables that are no longer in scope, you may be lucky,
442 but it's not wise to take the risk. If, however, the nested function
443 does not refer to anything that has gone out of scope, you should be
446 GCC implements taking the address of a nested function using a technique
447 called @dfn{trampolines}. A paper describing them is available as
450 @uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
452 A nested function can jump to a label inherited from a containing
453 function, provided the label was explicitly declared in the containing
454 function (@pxref{Local Labels}). Such a jump returns instantly to the
455 containing function, exiting the nested function which did the
456 @code{goto} and any intermediate functions as well. Here is an example:
460 bar (int *array, int offset, int size)
463 int access (int *array, int index)
467 return array[index + offset];
471 for (i = 0; i < size; i++)
472 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
476 /* @r{Control comes here from @code{access}
477 if it detects an error.} */
484 A nested function always has no linkage. Declaring one with
485 @code{extern} or @code{static} is erroneous. If you need to declare the nested function
486 before its definition, use @code{auto} (which is otherwise meaningless
487 for function declarations).
490 bar (int *array, int offset, int size)
493 auto int access (int *, int);
495 int access (int *array, int index)
499 return array[index + offset];
505 @node Constructing Calls
506 @section Constructing Function Calls
507 @cindex constructing calls
508 @cindex forwarding calls
510 Using the built-in functions described below, you can record
511 the arguments a function received, and call another function
512 with the same arguments, without knowing the number or types
515 You can also record the return value of that function call,
516 and later return that value, without knowing what data type
517 the function tried to return (as long as your caller expects
520 However, these built-in functions may interact badly with some
521 sophisticated features or other extensions of the language. It
522 is, therefore, not recommended to use them outside very simple
523 functions acting as mere forwarders for their arguments.
525 @deftypefn {Built-in Function} {void *} __builtin_apply_args ()
526 This built-in function returns a pointer to data
527 describing how to perform a call with the same arguments as were passed
528 to the current function.
530 The function saves the arg pointer register, structure value address,
531 and all registers that might be used to pass arguments to a function
532 into a block of memory allocated on the stack. Then it returns the
533 address of that block.
536 @deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
537 This built-in function invokes @var{function}
538 with a copy of the parameters described by @var{arguments}
541 The value of @var{arguments} should be the value returned by
542 @code{__builtin_apply_args}. The argument @var{size} specifies the size
543 of the stack argument data, in bytes.
545 This function returns a pointer to data describing
546 how to return whatever value was returned by @var{function}. The data
547 is saved in a block of memory allocated on the stack.
549 It is not always simple to compute the proper value for @var{size}. The
550 value is used by @code{__builtin_apply} to compute the amount of data
551 that should be pushed on the stack and copied from the incoming argument
555 @deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
556 This built-in function returns the value described by @var{result} from
557 the containing function. You should specify, for @var{result}, a value
558 returned by @code{__builtin_apply}.
562 @section Referring to a Type with @code{typeof}
565 @cindex macros, types of arguments
567 Another way to refer to the type of an expression is with @code{typeof}.
568 The syntax of using of this keyword looks like @code{sizeof}, but the
569 construct acts semantically like a type name defined with @code{typedef}.
571 There are two ways of writing the argument to @code{typeof}: with an
572 expression or with a type. Here is an example with an expression:
579 This assumes that @code{x} is an array of pointers to functions;
580 the type described is that of the values of the functions.
582 Here is an example with a typename as the argument:
589 Here the type described is that of pointers to @code{int}.
591 If you are writing a header file that must work when included in ISO C
592 programs, write @code{__typeof__} instead of @code{typeof}.
593 @xref{Alternate Keywords}.
595 A @code{typeof}-construct can be used anywhere a typedef name could be
596 used. For example, you can use it in a declaration, in a cast, or inside
597 of @code{sizeof} or @code{typeof}.
599 @code{typeof} is often useful in conjunction with the
600 statements-within-expressions feature. Here is how the two together can
601 be used to define a safe ``maximum'' macro that operates on any
602 arithmetic type and evaluates each of its arguments exactly once:
606 (@{ typeof (a) _a = (a); \
607 typeof (b) _b = (b); \
608 _a > _b ? _a : _b; @})
611 @cindex underscores in variables in macros
612 @cindex @samp{_} in variables in macros
613 @cindex local variables in macros
614 @cindex variables, local, in macros
615 @cindex macros, local variables in
617 The reason for using names that start with underscores for the local
618 variables is to avoid conflicts with variable names that occur within the
619 expressions that are substituted for @code{a} and @code{b}. Eventually we
620 hope to design a new form of declaration syntax that allows you to declare
621 variables whose scopes start only after their initializers; this will be a
622 more reliable way to prevent such conflicts.
625 Some more examples of the use of @code{typeof}:
629 This declares @code{y} with the type of what @code{x} points to.
636 This declares @code{y} as an array of such values.
643 This declares @code{y} as an array of pointers to characters:
646 typeof (typeof (char *)[4]) y;
650 It is equivalent to the following traditional C declaration:
656 To see the meaning of the declaration using @code{typeof}, and why it
657 might be a useful way to write, rewrite it with these macros:
660 #define pointer(T) typeof(T *)
661 #define array(T, N) typeof(T [N])
665 Now the declaration can be rewritten this way:
668 array (pointer (char), 4) y;
672 Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
673 pointers to @code{char}.
676 @emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
677 a more limited extension which permitted one to write
680 typedef @var{T} = @var{expr};
684 with the effect of declaring @var{T} to have the type of the expression
685 @var{expr}. This extension does not work with GCC 3 (versions between
686 3.0 and 3.2 will crash; 3.2.1 and later give an error). Code which
687 relies on it should be rewritten to use @code{typeof}:
690 typedef typeof(@var{expr}) @var{T};
694 This will work with all versions of GCC@.
697 @section Conditionals with Omitted Operands
698 @cindex conditional expressions, extensions
699 @cindex omitted middle-operands
700 @cindex middle-operands, omitted
701 @cindex extensions, @code{?:}
702 @cindex @code{?:} extensions
704 The middle operand in a conditional expression may be omitted. Then
705 if the first operand is nonzero, its value is the value of the conditional
708 Therefore, the expression
715 has the value of @code{x} if that is nonzero; otherwise, the value of
718 This example is perfectly equivalent to
724 @cindex side effect in ?:
725 @cindex ?: side effect
727 In this simple case, the ability to omit the middle operand is not
728 especially useful. When it becomes useful is when the first operand does,
729 or may (if it is a macro argument), contain a side effect. Then repeating
730 the operand in the middle would perform the side effect twice. Omitting
731 the middle operand uses the value already computed without the undesirable
732 effects of recomputing it.
735 @section Double-Word Integers
736 @cindex @code{long long} data types
737 @cindex double-word arithmetic
738 @cindex multiprecision arithmetic
739 @cindex @code{LL} integer suffix
740 @cindex @code{ULL} integer suffix
742 ISO C99 supports data types for integers that are at least 64 bits wide,
743 and as an extension GCC supports them in C89 mode and in C++.
744 Simply write @code{long long int} for a signed integer, or
745 @code{unsigned long long int} for an unsigned integer. To make an
746 integer constant of type @code{long long int}, add the suffix @samp{LL}
747 to the integer. To make an integer constant of type @code{unsigned long
748 long int}, add the suffix @samp{ULL} to the integer.
750 You can use these types in arithmetic like any other integer types.
751 Addition, subtraction, and bitwise boolean operations on these types
752 are open-coded on all types of machines. Multiplication is open-coded
753 if the machine supports fullword-to-doubleword a widening multiply
754 instruction. Division and shifts are open-coded only on machines that
755 provide special support. The operations that are not open-coded use
756 special library routines that come with GCC@.
758 There may be pitfalls when you use @code{long long} types for function
759 arguments, unless you declare function prototypes. If a function
760 expects type @code{int} for its argument, and you pass a value of type
761 @code{long long int}, confusion will result because the caller and the
762 subroutine will disagree about the number of bytes for the argument.
763 Likewise, if the function expects @code{long long int} and you pass
764 @code{int}. The best way to avoid such problems is to use prototypes.
767 @section Complex Numbers
768 @cindex complex numbers
769 @cindex @code{_Complex} keyword
770 @cindex @code{__complex__} keyword
772 ISO C99 supports complex floating data types, and as an extension GCC
773 supports them in C89 mode and in C++, and supports complex integer data
774 types which are not part of ISO C99. You can declare complex types
775 using the keyword @code{_Complex}. As an extension, the older GNU
776 keyword @code{__complex__} is also supported.
778 For example, @samp{_Complex double x;} declares @code{x} as a
779 variable whose real part and imaginary part are both of type
780 @code{double}. @samp{_Complex short int y;} declares @code{y} to
781 have real and imaginary parts of type @code{short int}; this is not
782 likely to be useful, but it shows that the set of complex types is
785 To write a constant with a complex data type, use the suffix @samp{i} or
786 @samp{j} (either one; they are equivalent). For example, @code{2.5fi}
787 has type @code{_Complex float} and @code{3i} has type
788 @code{_Complex int}. Such a constant always has a pure imaginary
789 value, but you can form any complex value you like by adding one to a
790 real constant. This is a GNU extension; if you have an ISO C99
791 conforming C library (such as GNU libc), and want to construct complex
792 constants of floating type, you should include @code{<complex.h>} and
793 use the macros @code{I} or @code{_Complex_I} instead.
795 @cindex @code{__real__} keyword
796 @cindex @code{__imag__} keyword
797 To extract the real part of a complex-valued expression @var{exp}, write
798 @code{__real__ @var{exp}}. Likewise, use @code{__imag__} to
799 extract the imaginary part. This is a GNU extension; for values of
800 floating type, you should use the ISO C99 functions @code{crealf},
801 @code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
802 @code{cimagl}, declared in @code{<complex.h>} and also provided as
803 built-in functions by GCC@.
805 @cindex complex conjugation
806 The operator @samp{~} performs complex conjugation when used on a value
807 with a complex type. This is a GNU extension; for values of
808 floating type, you should use the ISO C99 functions @code{conjf},
809 @code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
810 provided as built-in functions by GCC@.
812 GCC can allocate complex automatic variables in a noncontiguous
813 fashion; it's even possible for the real part to be in a register while
814 the imaginary part is on the stack (or vice-versa). Only the DWARF2
815 debug info format can represent this, so use of DWARF2 is recommended.
816 If you are using the stabs debug info format, GCC describes a noncontiguous
817 complex variable as if it were two separate variables of noncomplex type.
818 If the variable's actual name is @code{foo}, the two fictitious
819 variables are named @code{foo$real} and @code{foo$imag}. You can
820 examine and set these two fictitious variables with your debugger.
823 @section Decimal Floating Types
824 @cindex decimal floating types
825 @cindex @code{_Decimal32} data type
826 @cindex @code{_Decimal64} data type
827 @cindex @code{_Decimal128} data type
828 @cindex @code{df} integer suffix
829 @cindex @code{dd} integer suffix
830 @cindex @code{dl} integer suffix
831 @cindex @code{DF} integer suffix
832 @cindex @code{DD} integer suffix
833 @cindex @code{DL} integer suffix
835 As an extension, the GNU C compiler supports decimal floating types as
836 defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal
837 floating types in GCC will evolve as the draft technical report changes.
838 Calling conventions for any target might also change. Not all targets
839 support decimal floating types.
841 The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
842 @code{_Decimal128}. They use a radix of ten, unlike the floating types
843 @code{float}, @code{double}, and @code{long double} whose radix is not
844 specified by the C standard but is usually two.
846 Support for decimal floating types includes the arithmetic operators
847 add, subtract, multiply, divide; unary arithmetic operators;
848 relational operators; equality operators; and conversions to and from
849 integer and other floating types. Use a suffix @samp{df} or
850 @samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
851 or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
854 GCC support of decimal float as specified by the draft technical report
859 Translation time data type (TTDT) is not supported.
862 Characteristics of decimal floating types are defined in header file
863 @file{decfloat.h} rather than @file{float.h}.
866 When the value of a decimal floating type cannot be represented in the
867 integer type to which it is being converted, the result is undefined
868 rather than the result value specified by the draft technical report.
871 Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
872 are supported by the DWARF2 debug information format.
878 ISO C99 supports floating-point numbers written not only in the usual
879 decimal notation, such as @code{1.55e1}, but also numbers such as
880 @code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC
881 supports this in C89 mode (except in some cases when strictly
882 conforming) and in C++. In that format the
883 @samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
884 mandatory. The exponent is a decimal number that indicates the power of
885 2 by which the significant part will be multiplied. Thus @samp{0x1.f} is
892 @samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
893 is the same as @code{1.55e1}.
895 Unlike for floating-point numbers in the decimal notation the exponent
896 is always required in the hexadecimal notation. Otherwise the compiler
897 would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This
898 could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
899 extension for floating-point constants of type @code{float}.
902 @section Arrays of Length Zero
903 @cindex arrays of length zero
904 @cindex zero-length arrays
905 @cindex length-zero arrays
906 @cindex flexible array members
908 Zero-length arrays are allowed in GNU C@. They are very useful as the
909 last element of a structure which is really a header for a variable-length
918 struct line *thisline = (struct line *)
919 malloc (sizeof (struct line) + this_length);
920 thisline->length = this_length;
923 In ISO C90, you would have to give @code{contents} a length of 1, which
924 means either you waste space or complicate the argument to @code{malloc}.
926 In ISO C99, you would use a @dfn{flexible array member}, which is
927 slightly different in syntax and semantics:
931 Flexible array members are written as @code{contents[]} without
935 Flexible array members have incomplete type, and so the @code{sizeof}
936 operator may not be applied. As a quirk of the original implementation
937 of zero-length arrays, @code{sizeof} evaluates to zero.
940 Flexible array members may only appear as the last member of a
941 @code{struct} that is otherwise non-empty.
944 A structure containing a flexible array member, or a union containing
945 such a structure (possibly recursively), may not be a member of a
946 structure or an element of an array. (However, these uses are
947 permitted by GCC as extensions.)
950 GCC versions before 3.0 allowed zero-length arrays to be statically
951 initialized, as if they were flexible arrays. In addition to those
952 cases that were useful, it also allowed initializations in situations
953 that would corrupt later data. Non-empty initialization of zero-length
954 arrays is now treated like any case where there are more initializer
955 elements than the array holds, in that a suitable warning about "excess
956 elements in array" is given, and the excess elements (all of them, in
957 this case) are ignored.
959 Instead GCC allows static initialization of flexible array members.
960 This is equivalent to defining a new structure containing the original
961 structure followed by an array of sufficient size to contain the data.
962 I.e.@: in the following, @code{f1} is constructed as if it were declared
968 @} f1 = @{ 1, @{ 2, 3, 4 @} @};
971 struct f1 f1; int data[3];
972 @} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
976 The convenience of this extension is that @code{f1} has the desired
977 type, eliminating the need to consistently refer to @code{f2.f1}.
979 This has symmetry with normal static arrays, in that an array of
980 unknown size is also written with @code{[]}.
982 Of course, this extension only makes sense if the extra data comes at
983 the end of a top-level object, as otherwise we would be overwriting
984 data at subsequent offsets. To avoid undue complication and confusion
985 with initialization of deeply nested arrays, we simply disallow any
986 non-empty initialization except when the structure is the top-level
990 struct foo @{ int x; int y[]; @};
991 struct bar @{ struct foo z; @};
993 struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.}
994 struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
995 struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.}
996 struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.}
999 @node Empty Structures
1000 @section Structures With No Members
1001 @cindex empty structures
1002 @cindex zero-size structures
1004 GCC permits a C structure to have no members:
1011 The structure will have size zero. In C++, empty structures are part
1012 of the language. G++ treats empty structures as if they had a single
1013 member of type @code{char}.
1015 @node Variable Length
1016 @section Arrays of Variable Length
1017 @cindex variable-length arrays
1018 @cindex arrays of variable length
1021 Variable-length automatic arrays are allowed in ISO C99, and as an
1022 extension GCC accepts them in C89 mode and in C++. (However, GCC's
1023 implementation of variable-length arrays does not yet conform in detail
1024 to the ISO C99 standard.) These arrays are
1025 declared like any other automatic arrays, but with a length that is not
1026 a constant expression. The storage is allocated at the point of
1027 declaration and deallocated when the brace-level is exited. For
1032 concat_fopen (char *s1, char *s2, char *mode)
1034 char str[strlen (s1) + strlen (s2) + 1];
1037 return fopen (str, mode);
1041 @cindex scope of a variable length array
1042 @cindex variable-length array scope
1043 @cindex deallocating variable length arrays
1044 Jumping or breaking out of the scope of the array name deallocates the
1045 storage. Jumping into the scope is not allowed; you get an error
1048 @cindex @code{alloca} vs variable-length arrays
1049 You can use the function @code{alloca} to get an effect much like
1050 variable-length arrays. The function @code{alloca} is available in
1051 many other C implementations (but not in all). On the other hand,
1052 variable-length arrays are more elegant.
1054 There are other differences between these two methods. Space allocated
1055 with @code{alloca} exists until the containing @emph{function} returns.
1056 The space for a variable-length array is deallocated as soon as the array
1057 name's scope ends. (If you use both variable-length arrays and
1058 @code{alloca} in the same function, deallocation of a variable-length array
1059 will also deallocate anything more recently allocated with @code{alloca}.)
1061 You can also use variable-length arrays as arguments to functions:
1065 tester (int len, char data[len][len])
1071 The length of an array is computed once when the storage is allocated
1072 and is remembered for the scope of the array in case you access it with
1075 If you want to pass the array first and the length afterward, you can
1076 use a forward declaration in the parameter list---another GNU extension.
1080 tester (int len; char data[len][len], int len)
1086 @cindex parameter forward declaration
1087 The @samp{int len} before the semicolon is a @dfn{parameter forward
1088 declaration}, and it serves the purpose of making the name @code{len}
1089 known when the declaration of @code{data} is parsed.
1091 You can write any number of such parameter forward declarations in the
1092 parameter list. They can be separated by commas or semicolons, but the
1093 last one must end with a semicolon, which is followed by the ``real''
1094 parameter declarations. Each forward declaration must match a ``real''
1095 declaration in parameter name and data type. ISO C99 does not support
1096 parameter forward declarations.
1098 @node Variadic Macros
1099 @section Macros with a Variable Number of Arguments.
1100 @cindex variable number of arguments
1101 @cindex macro with variable arguments
1102 @cindex rest argument (in macro)
1103 @cindex variadic macros
1105 In the ISO C standard of 1999, a macro can be declared to accept a
1106 variable number of arguments much as a function can. The syntax for
1107 defining the macro is similar to that of a function. Here is an
1111 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1114 Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of
1115 such a macro, it represents the zero or more tokens until the closing
1116 parenthesis that ends the invocation, including any commas. This set of
1117 tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1118 wherever it appears. See the CPP manual for more information.
1120 GCC has long supported variadic macros, and used a different syntax that
1121 allowed you to give a name to the variable arguments just like any other
1122 argument. Here is an example:
1125 #define debug(format, args...) fprintf (stderr, format, args)
1128 This is in all ways equivalent to the ISO C example above, but arguably
1129 more readable and descriptive.
1131 GNU CPP has two further variadic macro extensions, and permits them to
1132 be used with either of the above forms of macro definition.
1134 In standard C, you are not allowed to leave the variable argument out
1135 entirely; but you are allowed to pass an empty argument. For example,
1136 this invocation is invalid in ISO C, because there is no comma after
1143 GNU CPP permits you to completely omit the variable arguments in this
1144 way. In the above examples, the compiler would complain, though since
1145 the expansion of the macro still has the extra comma after the format
1148 To help solve this problem, CPP behaves specially for variable arguments
1149 used with the token paste operator, @samp{##}. If instead you write
1152 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1155 and if the variable arguments are omitted or empty, the @samp{##}
1156 operator causes the preprocessor to remove the comma before it. If you
1157 do provide some variable arguments in your macro invocation, GNU CPP
1158 does not complain about the paste operation and instead places the
1159 variable arguments after the comma. Just like any other pasted macro
1160 argument, these arguments are not macro expanded.
1162 @node Escaped Newlines
1163 @section Slightly Looser Rules for Escaped Newlines
1164 @cindex escaped newlines
1165 @cindex newlines (escaped)
1167 Recently, the preprocessor has relaxed its treatment of escaped
1168 newlines. Previously, the newline had to immediately follow a
1169 backslash. The current implementation allows whitespace in the form
1170 of spaces, horizontal and vertical tabs, and form feeds between the
1171 backslash and the subsequent newline. The preprocessor issues a
1172 warning, but treats it as a valid escaped newline and combines the two
1173 lines to form a single logical line. This works within comments and
1174 tokens, as well as between tokens. Comments are @emph{not} treated as
1175 whitespace for the purposes of this relaxation, since they have not
1176 yet been replaced with spaces.
1179 @section Non-Lvalue Arrays May Have Subscripts
1180 @cindex subscripting
1181 @cindex arrays, non-lvalue
1183 @cindex subscripting and function values
1184 In ISO C99, arrays that are not lvalues still decay to pointers, and
1185 may be subscripted, although they may not be modified or used after
1186 the next sequence point and the unary @samp{&} operator may not be
1187 applied to them. As an extension, GCC allows such arrays to be
1188 subscripted in C89 mode, though otherwise they do not decay to
1189 pointers outside C99 mode. For example,
1190 this is valid in GNU C though not valid in C89:
1194 struct foo @{int a[4];@};
1200 return f().a[index];
1206 @section Arithmetic on @code{void}- and Function-Pointers
1207 @cindex void pointers, arithmetic
1208 @cindex void, size of pointer to
1209 @cindex function pointers, arithmetic
1210 @cindex function, size of pointer to
1212 In GNU C, addition and subtraction operations are supported on pointers to
1213 @code{void} and on pointers to functions. This is done by treating the
1214 size of a @code{void} or of a function as 1.
1216 A consequence of this is that @code{sizeof} is also allowed on @code{void}
1217 and on function types, and returns 1.
1219 @opindex Wpointer-arith
1220 The option @option{-Wpointer-arith} requests a warning if these extensions
1224 @section Non-Constant Initializers
1225 @cindex initializers, non-constant
1226 @cindex non-constant initializers
1228 As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1229 automatic variable are not required to be constant expressions in GNU C@.
1230 Here is an example of an initializer with run-time varying elements:
1233 foo (float f, float g)
1235 float beat_freqs[2] = @{ f-g, f+g @};
1240 @node Compound Literals
1241 @section Compound Literals
1242 @cindex constructor expressions
1243 @cindex initializations in expressions
1244 @cindex structures, constructor expression
1245 @cindex expressions, constructor
1246 @cindex compound literals
1247 @c The GNU C name for what C99 calls compound literals was "constructor expressions".
1249 ISO C99 supports compound literals. A compound literal looks like
1250 a cast containing an initializer. Its value is an object of the
1251 type specified in the cast, containing the elements specified in
1252 the initializer; it is an lvalue. As an extension, GCC supports
1253 compound literals in C89 mode and in C++.
1255 Usually, the specified type is a structure. Assume that
1256 @code{struct foo} and @code{structure} are declared as shown:
1259 struct foo @{int a; char b[2];@} structure;
1263 Here is an example of constructing a @code{struct foo} with a compound literal:
1266 structure = ((struct foo) @{x + y, 'a', 0@});
1270 This is equivalent to writing the following:
1274 struct foo temp = @{x + y, 'a', 0@};
1279 You can also construct an array. If all the elements of the compound literal
1280 are (made up of) simple constant expressions, suitable for use in
1281 initializers of objects of static storage duration, then the compound
1282 literal can be coerced to a pointer to its first element and used in
1283 such an initializer, as shown here:
1286 char **foo = (char *[]) @{ "x", "y", "z" @};
1289 Compound literals for scalar types and union types are is
1290 also allowed, but then the compound literal is equivalent
1293 As a GNU extension, GCC allows initialization of objects with static storage
1294 duration by compound literals (which is not possible in ISO C99, because
1295 the initializer is not a constant).
1296 It is handled as if the object was initialized only with the bracket
1297 enclosed list if the types of the compound literal and the object match.
1298 The initializer list of the compound literal must be constant.
1299 If the object being initialized has array type of unknown size, the size is
1300 determined by compound literal size.
1303 static struct foo x = (struct foo) @{1, 'a', 'b'@};
1304 static int y[] = (int []) @{1, 2, 3@};
1305 static int z[] = (int [3]) @{1@};
1309 The above lines are equivalent to the following:
1311 static struct foo x = @{1, 'a', 'b'@};
1312 static int y[] = @{1, 2, 3@};
1313 static int z[] = @{1, 0, 0@};
1316 @node Designated Inits
1317 @section Designated Initializers
1318 @cindex initializers with labeled elements
1319 @cindex labeled elements in initializers
1320 @cindex case labels in initializers
1321 @cindex designated initializers
1323 Standard C89 requires the elements of an initializer to appear in a fixed
1324 order, the same as the order of the elements in the array or structure
1327 In ISO C99 you can give the elements in any order, specifying the array
1328 indices or structure field names they apply to, and GNU C allows this as
1329 an extension in C89 mode as well. This extension is not
1330 implemented in GNU C++.
1332 To specify an array index, write
1333 @samp{[@var{index}] =} before the element value. For example,
1336 int a[6] = @{ [4] = 29, [2] = 15 @};
1343 int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1347 The index values must be constant expressions, even if the array being
1348 initialized is automatic.
1350 An alternative syntax for this which has been obsolete since GCC 2.5 but
1351 GCC still accepts is to write @samp{[@var{index}]} before the element
1352 value, with no @samp{=}.
1354 To initialize a range of elements to the same value, write
1355 @samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU
1356 extension. For example,
1359 int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1363 If the value in it has side-effects, the side-effects will happen only once,
1364 not for each initialized field by the range initializer.
1367 Note that the length of the array is the highest value specified
1370 In a structure initializer, specify the name of a field to initialize
1371 with @samp{.@var{fieldname} =} before the element value. For example,
1372 given the following structure,
1375 struct point @{ int x, y; @};
1379 the following initialization
1382 struct point p = @{ .y = yvalue, .x = xvalue @};
1389 struct point p = @{ xvalue, yvalue @};
1392 Another syntax which has the same meaning, obsolete since GCC 2.5, is
1393 @samp{@var{fieldname}:}, as shown here:
1396 struct point p = @{ y: yvalue, x: xvalue @};
1400 The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1401 @dfn{designator}. You can also use a designator (or the obsolete colon
1402 syntax) when initializing a union, to specify which element of the union
1403 should be used. For example,
1406 union foo @{ int i; double d; @};
1408 union foo f = @{ .d = 4 @};
1412 will convert 4 to a @code{double} to store it in the union using
1413 the second element. By contrast, casting 4 to type @code{union foo}
1414 would store it into the union as the integer @code{i}, since it is
1415 an integer. (@xref{Cast to Union}.)
1417 You can combine this technique of naming elements with ordinary C
1418 initialization of successive elements. Each initializer element that
1419 does not have a designator applies to the next consecutive element of the
1420 array or structure. For example,
1423 int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1430 int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1433 Labeling the elements of an array initializer is especially useful
1434 when the indices are characters or belong to an @code{enum} type.
1439 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1440 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1443 @cindex designator lists
1444 You can also write a series of @samp{.@var{fieldname}} and
1445 @samp{[@var{index}]} designators before an @samp{=} to specify a
1446 nested subobject to initialize; the list is taken relative to the
1447 subobject corresponding to the closest surrounding brace pair. For
1448 example, with the @samp{struct point} declaration above:
1451 struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1455 If the same field is initialized multiple times, it will have value from
1456 the last initialization. If any such overridden initialization has
1457 side-effect, it is unspecified whether the side-effect happens or not.
1458 Currently, GCC will discard them and issue a warning.
1461 @section Case Ranges
1463 @cindex ranges in case statements
1465 You can specify a range of consecutive values in a single @code{case} label,
1469 case @var{low} ... @var{high}:
1473 This has the same effect as the proper number of individual @code{case}
1474 labels, one for each integer value from @var{low} to @var{high}, inclusive.
1476 This feature is especially useful for ranges of ASCII character codes:
1482 @strong{Be careful:} Write spaces around the @code{...}, for otherwise
1483 it may be parsed wrong when you use it with integer values. For example,
1498 @section Cast to a Union Type
1499 @cindex cast to a union
1500 @cindex union, casting to a
1502 A cast to union type is similar to other casts, except that the type
1503 specified is a union type. You can specify the type either with
1504 @code{union @var{tag}} or with a typedef name. A cast to union is actually
1505 a constructor though, not a cast, and hence does not yield an lvalue like
1506 normal casts. (@xref{Compound Literals}.)
1508 The types that may be cast to the union type are those of the members
1509 of the union. Thus, given the following union and variables:
1512 union foo @{ int i; double d; @};
1518 both @code{x} and @code{y} can be cast to type @code{union foo}.
1520 Using the cast as the right-hand side of an assignment to a variable of
1521 union type is equivalent to storing in a member of the union:
1526 u = (union foo) x @equiv{} u.i = x
1527 u = (union foo) y @equiv{} u.d = y
1530 You can also use the union cast as a function argument:
1533 void hack (union foo);
1535 hack ((union foo) x);
1538 @node Mixed Declarations
1539 @section Mixed Declarations and Code
1540 @cindex mixed declarations and code
1541 @cindex declarations, mixed with code
1542 @cindex code, mixed with declarations
1544 ISO C99 and ISO C++ allow declarations and code to be freely mixed
1545 within compound statements. As an extension, GCC also allows this in
1546 C89 mode. For example, you could do:
1555 Each identifier is visible from where it is declared until the end of
1556 the enclosing block.
1558 @node Function Attributes
1559 @section Declaring Attributes of Functions
1560 @cindex function attributes
1561 @cindex declaring attributes of functions
1562 @cindex functions that never return
1563 @cindex functions that return more than once
1564 @cindex functions that have no side effects
1565 @cindex functions in arbitrary sections
1566 @cindex functions that behave like malloc
1567 @cindex @code{volatile} applied to function
1568 @cindex @code{const} applied to function
1569 @cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1570 @cindex functions with non-null pointer arguments
1571 @cindex functions that are passed arguments in registers on the 386
1572 @cindex functions that pop the argument stack on the 386
1573 @cindex functions that do not pop the argument stack on the 386
1575 In GNU C, you declare certain things about functions called in your program
1576 which help the compiler optimize function calls and check your code more
1579 The keyword @code{__attribute__} allows you to specify special
1580 attributes when making a declaration. This keyword is followed by an
1581 attribute specification inside double parentheses. The following
1582 attributes are currently defined for functions on all targets:
1584 @code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline},
1585 @code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel},
1586 @code{format}, @code{format_arg}, @code{no_instrument_function},
1587 @code{section}, @code{constructor}, @code{destructor}, @code{used},
1588 @code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1589 @code{alias}, @code{warn_unused_result}, @code{nonnull},
1590 @code{gnu_inline} and @code{externally_visible}. Several other
1591 attributes are defined for functions on particular target systems. Other
1592 attributes, including @code{section} are supported for variables declarations
1593 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
1594 (@pxref{Variable Attributes}), for types (@pxref{Type Attributes}),
1595 and labels (@pxref{Label Attributes}).
1597 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
1598 You may also specify attributes with @samp{__} preceding and following
1599 each keyword. This allows you to use them in header files without
1600 being concerned about a possible macro of the same name. For example,
1601 you may use @code{__noreturn__} instead of @code{noreturn}.
1603 @xref{Attribute Syntax}, for details of the exact syntax for using
1607 @c Keep this table alphabetized by attribute name. Treat _ as space.
1609 @item alias ("@var{target}")
1610 @cindex @code{alias} attribute
1611 The @code{alias} attribute causes the declaration to be emitted as an
1612 alias for another symbol, which must be specified. For instance,
1615 void __f () @{ /* @r{Do something.} */; @}
1616 void f () __attribute__ ((weak, alias ("__f")));
1619 defines @samp{f} to be a weak alias for @samp{__f}. In C++, the
1620 mangled name for the target must be used. It is an error if @samp{__f}
1621 is not defined in the same translation unit.
1623 Not all target machines support this attribute.
1625 @item aligned (@var{alignment})
1626 @cindex @code{aligned} attribute
1627 This attribute specifies a minimum alignment for the function,
1630 You cannot use this attribute to decrease the alignment of a function,
1631 only to increase it. However, when you explicitly specify a function
1632 alignment this will override the effect of the
1633 @option{-falign-functions} (@pxref{Optimize Options}) option for this
1636 Note that the effectiveness of @code{aligned} attributes may be
1637 limited by inherent limitations in your linker. On many systems, the
1638 linker is only able to arrange for functions to be aligned up to a
1639 certain maximum alignment. (For some linkers, the maximum supported
1640 alignment may be very very small.) See your linker documentation for
1641 further information.
1643 The @code{aligned} attribute can also be used for variables and fields
1644 (@pxref{Variable Attributes}.)
1647 @cindex @code{always_inline} function attribute
1648 Generally, functions are not inlined unless optimization is specified.
1649 For functions declared inline, this attribute inlines the function even
1650 if no optimization level was specified.
1653 @cindex @code{gnu_inline} function attribute
1654 This attribute should be used with a function which is also declared
1655 with the @code{inline} keyword. It directs GCC to treat the function
1656 as if it were defined in gnu89 mode even when compiling in C99 or
1659 If the function is declared @code{extern}, then this definition of the
1660 function is used only for inlining. In no case is the function
1661 compiled as a standalone function, not even if you take its address
1662 explicitly. Such an address becomes an external reference, as if you
1663 had only declared the function, and had not defined it. This has
1664 almost the effect of a macro. The way to use this is to put a
1665 function definition in a header file with this attribute, and put
1666 another copy of the function, without @code{extern}, in a library
1667 file. The definition in the header file will cause most calls to the
1668 function to be inlined. If any uses of the function remain, they will
1669 refer to the single copy in the library. Note that the two
1670 definitions of the functions need not be precisely the same, although
1671 if they do not have the same effect your program may behave oddly.
1673 If the function is neither @code{extern} nor @code{static}, then the
1674 function is compiled as a standalone function, as well as being
1675 inlined where possible.
1677 This is how GCC traditionally handled functions declared
1678 @code{inline}. Since ISO C99 specifies a different semantics for
1679 @code{inline}, this function attribute is provided as a transition
1680 measure and as a useful feature in its own right. This attribute is
1681 available in GCC 4.1.3 and later. It is available if either of the
1682 preprocessor macros @code{__GNUC_GNU_INLINE__} or
1683 @code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline
1684 Function is As Fast As a Macro}.
1686 Note that since the first version of GCC to support C99 inline semantics
1687 is 4.3, earlier versions of GCC which accept this attribute effectively
1688 assume that it is always present, whether or not it is given explicitly.
1689 In versions prior to 4.3, the only effect of explicitly including it is
1690 to disable warnings about using inline functions in C99 mode.
1692 @cindex @code{flatten} function attribute
1694 Generally, inlining into a function is limited. For a function marked with
1695 this attribute, every call inside this function will be inlined, if possible.
1696 Whether the function itself is considered for inlining depends on its size and
1697 the current inlining parameters. The @code{flatten} attribute only works
1698 reliably in unit-at-a-time mode.
1701 @cindex functions that do pop the argument stack on the 386
1703 On the Intel 386, the @code{cdecl} attribute causes the compiler to
1704 assume that the calling function will pop off the stack space used to
1705 pass arguments. This is
1706 useful to override the effects of the @option{-mrtd} switch.
1709 @cindex @code{const} function attribute
1710 Many functions do not examine any values except their arguments, and
1711 have no effects except the return value. Basically this is just slightly
1712 more strict class than the @code{pure} attribute below, since function is not
1713 allowed to read global memory.
1715 @cindex pointer arguments
1716 Note that a function that has pointer arguments and examines the data
1717 pointed to must @emph{not} be declared @code{const}. Likewise, a
1718 function that calls a non-@code{const} function usually must not be
1719 @code{const}. It does not make sense for a @code{const} function to
1722 The attribute @code{const} is not implemented in GCC versions earlier
1723 than 2.5. An alternative way to declare that a function has no side
1724 effects, which works in the current version and in some older versions,
1728 typedef int intfn ();
1730 extern const intfn square;
1733 This approach does not work in GNU C++ from 2.6.0 on, since the language
1734 specifies that the @samp{const} must be attached to the return value.
1738 @cindex @code{constructor} function attribute
1739 @cindex @code{destructor} function attribute
1740 The @code{constructor} attribute causes the function to be called
1741 automatically before execution enters @code{main ()}. Similarly, the
1742 @code{destructor} attribute causes the function to be called
1743 automatically after @code{main ()} has completed or @code{exit ()} has
1744 been called. Functions with these attributes are useful for
1745 initializing data that will be used implicitly during the execution of
1748 These attributes are not currently implemented for Objective-C@.
1751 @cindex @code{deprecated} attribute.
1752 The @code{deprecated} attribute results in a warning if the function
1753 is used anywhere in the source file. This is useful when identifying
1754 functions that are expected to be removed in a future version of a
1755 program. The warning also includes the location of the declaration
1756 of the deprecated function, to enable users to easily find further
1757 information about why the function is deprecated, or what they should
1758 do instead. Note that the warnings only occurs for uses:
1761 int old_fn () __attribute__ ((deprecated));
1763 int (*fn_ptr)() = old_fn;
1766 results in a warning on line 3 but not line 2.
1768 The @code{deprecated} attribute can also be used for variables and
1769 types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
1772 @cindex @code{__declspec(dllexport)}
1773 On Microsoft Windows targets and Symbian OS targets the
1774 @code{dllexport} attribute causes the compiler to provide a global
1775 pointer to a pointer in a DLL, so that it can be referenced with the
1776 @code{dllimport} attribute. On Microsoft Windows targets, the pointer
1777 name is formed by combining @code{_imp__} and the function or variable
1780 You can use @code{__declspec(dllexport)} as a synonym for
1781 @code{__attribute__ ((dllexport))} for compatibility with other
1784 On systems that support the @code{visibility} attribute, this
1785 attribute also implies ``default'' visibility, unless a
1786 @code{visibility} attribute is explicitly specified. You should avoid
1787 the use of @code{dllexport} with ``hidden'' or ``internal''
1788 visibility; in the future GCC may issue an error for those cases.
1790 Currently, the @code{dllexport} attribute is ignored for inlined
1791 functions, unless the @option{-fkeep-inline-functions} flag has been
1792 used. The attribute is also ignored for undefined symbols.
1794 When applied to C++ classes, the attribute marks defined non-inlined
1795 member functions and static data members as exports. Static consts
1796 initialized in-class are not marked unless they are also defined
1799 For Microsoft Windows targets there are alternative methods for
1800 including the symbol in the DLL's export table such as using a
1801 @file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
1802 the @option{--export-all} linker flag.
1805 @cindex @code{__declspec(dllimport)}
1806 On Microsoft Windows and Symbian OS targets, the @code{dllimport}
1807 attribute causes the compiler to reference a function or variable via
1808 a global pointer to a pointer that is set up by the DLL exporting the
1809 symbol. The attribute implies @code{extern} storage. On Microsoft
1810 Windows targets, the pointer name is formed by combining @code{_imp__}
1811 and the function or variable name.
1813 You can use @code{__declspec(dllimport)} as a synonym for
1814 @code{__attribute__ ((dllimport))} for compatibility with other
1817 Currently, the attribute is ignored for inlined functions. If the
1818 attribute is applied to a symbol @emph{definition}, an error is reported.
1819 If a symbol previously declared @code{dllimport} is later defined, the
1820 attribute is ignored in subsequent references, and a warning is emitted.
1821 The attribute is also overridden by a subsequent declaration as
1824 When applied to C++ classes, the attribute marks non-inlined
1825 member functions and static data members as imports. However, the
1826 attribute is ignored for virtual methods to allow creation of vtables
1829 On the SH Symbian OS target the @code{dllimport} attribute also has
1830 another affect---it can cause the vtable and run-time type information
1831 for a class to be exported. This happens when the class has a
1832 dllimport'ed constructor or a non-inline, non-pure virtual function
1833 and, for either of those two conditions, the class also has a inline
1834 constructor or destructor and has a key function that is defined in
1835 the current translation unit.
1837 For Microsoft Windows based targets the use of the @code{dllimport}
1838 attribute on functions is not necessary, but provides a small
1839 performance benefit by eliminating a thunk in the DLL@. The use of the
1840 @code{dllimport} attribute on imported variables was required on older
1841 versions of the GNU linker, but can now be avoided by passing the
1842 @option{--enable-auto-import} switch to the GNU linker. As with
1843 functions, using the attribute for a variable eliminates a thunk in
1846 One drawback to using this attribute is that a pointer to a function
1847 or variable marked as @code{dllimport} cannot be used as a constant
1848 address. On Microsoft Windows targets, the attribute can be disabled
1849 for functions by setting the @option{-mnop-fun-dllimport} flag.
1852 @cindex eight bit data on the H8/300, H8/300H, and H8S
1853 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
1854 variable should be placed into the eight bit data section.
1855 The compiler will generate more efficient code for certain operations
1856 on data in the eight bit data area. Note the eight bit data area is limited to
1859 You must use GAS and GLD from GNU binutils version 2.7 or later for
1860 this attribute to work correctly.
1862 @item exception_handler
1863 @cindex exception handler functions on the Blackfin processor
1864 Use this attribute on the Blackfin to indicate that the specified function
1865 is an exception handler. The compiler will generate function entry and
1866 exit sequences suitable for use in an exception handler when this
1867 attribute is present.
1870 @cindex functions which handle memory bank switching
1871 On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
1872 use a calling convention that takes care of switching memory banks when
1873 entering and leaving a function. This calling convention is also the
1874 default when using the @option{-mlong-calls} option.
1876 On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
1877 to call and return from a function.
1879 On 68HC11 the compiler will generate a sequence of instructions
1880 to invoke a board-specific routine to switch the memory bank and call the
1881 real function. The board-specific routine simulates a @code{call}.
1882 At the end of a function, it will jump to a board-specific routine
1883 instead of using @code{rts}. The board-specific return routine simulates
1887 @cindex functions that pop the argument stack on the 386
1888 On the Intel 386, the @code{fastcall} attribute causes the compiler to
1889 pass the first argument (if of integral type) in the register ECX and
1890 the second argument (if of integral type) in the register EDX@. Subsequent
1891 and other typed arguments are passed on the stack. The called function will
1892 pop the arguments off the stack. If the number of arguments is variable all
1893 arguments are pushed on the stack.
1895 @item format (@var{archetype}, @var{string-index}, @var{first-to-check})
1896 @cindex @code{format} function attribute
1898 The @code{format} attribute specifies that a function takes @code{printf},
1899 @code{scanf}, @code{strftime} or @code{strfmon} style arguments which
1900 should be type-checked against a format string. For example, the
1905 my_printf (void *my_object, const char *my_format, ...)
1906 __attribute__ ((format (printf, 2, 3)));
1910 causes the compiler to check the arguments in calls to @code{my_printf}
1911 for consistency with the @code{printf} style format string argument
1914 The parameter @var{archetype} determines how the format string is
1915 interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
1916 or @code{strfmon}. (You can also use @code{__printf__},
1917 @code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The
1918 parameter @var{string-index} specifies which argument is the format
1919 string argument (starting from 1), while @var{first-to-check} is the
1920 number of the first argument to check against the format string. For
1921 functions where the arguments are not available to be checked (such as
1922 @code{vprintf}), specify the third parameter as zero. In this case the
1923 compiler only checks the format string for consistency. For
1924 @code{strftime} formats, the third parameter is required to be zero.
1925 Since non-static C++ methods have an implicit @code{this} argument, the
1926 arguments of such methods should be counted from two, not one, when
1927 giving values for @var{string-index} and @var{first-to-check}.
1929 In the example above, the format string (@code{my_format}) is the second
1930 argument of the function @code{my_print}, and the arguments to check
1931 start with the third argument, so the correct parameters for the format
1932 attribute are 2 and 3.
1934 @opindex ffreestanding
1935 @opindex fno-builtin
1936 The @code{format} attribute allows you to identify your own functions
1937 which take format strings as arguments, so that GCC can check the
1938 calls to these functions for errors. The compiler always (unless
1939 @option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
1940 for the standard library functions @code{printf}, @code{fprintf},
1941 @code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
1942 @code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
1943 warnings are requested (using @option{-Wformat}), so there is no need to
1944 modify the header file @file{stdio.h}. In C99 mode, the functions
1945 @code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
1946 @code{vsscanf} are also checked. Except in strictly conforming C
1947 standard modes, the X/Open function @code{strfmon} is also checked as
1948 are @code{printf_unlocked} and @code{fprintf_unlocked}.
1949 @xref{C Dialect Options,,Options Controlling C Dialect}.
1951 The target may provide additional types of format checks.
1952 @xref{Target Format Checks,,Format Checks Specific to Particular
1955 @item format_arg (@var{string-index})
1956 @cindex @code{format_arg} function attribute
1957 @opindex Wformat-nonliteral
1958 The @code{format_arg} attribute specifies that a function takes a format
1959 string for a @code{printf}, @code{scanf}, @code{strftime} or
1960 @code{strfmon} style function and modifies it (for example, to translate
1961 it into another language), so the result can be passed to a
1962 @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
1963 function (with the remaining arguments to the format function the same
1964 as they would have been for the unmodified string). For example, the
1969 my_dgettext (char *my_domain, const char *my_format)
1970 __attribute__ ((format_arg (2)));
1974 causes the compiler to check the arguments in calls to a @code{printf},
1975 @code{scanf}, @code{strftime} or @code{strfmon} type function, whose
1976 format string argument is a call to the @code{my_dgettext} function, for
1977 consistency with the format string argument @code{my_format}. If the
1978 @code{format_arg} attribute had not been specified, all the compiler
1979 could tell in such calls to format functions would be that the format
1980 string argument is not constant; this would generate a warning when
1981 @option{-Wformat-nonliteral} is used, but the calls could not be checked
1982 without the attribute.
1984 The parameter @var{string-index} specifies which argument is the format
1985 string argument (starting from one). Since non-static C++ methods have
1986 an implicit @code{this} argument, the arguments of such methods should
1987 be counted from two.
1989 The @code{format-arg} attribute allows you to identify your own
1990 functions which modify format strings, so that GCC can check the
1991 calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
1992 type function whose operands are a call to one of your own function.
1993 The compiler always treats @code{gettext}, @code{dgettext}, and
1994 @code{dcgettext} in this manner except when strict ISO C support is
1995 requested by @option{-ansi} or an appropriate @option{-std} option, or
1996 @option{-ffreestanding} or @option{-fno-builtin}
1997 is used. @xref{C Dialect Options,,Options
1998 Controlling C Dialect}.
2000 @item function_vector
2001 @cindex calling functions through the function vector on the H8/300 processors
2002 Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2003 function should be called through the function vector. Calling a
2004 function through the function vector will reduce code size, however;
2005 the function vector has a limited size (maximum 128 entries on the H8/300
2006 and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2008 You must use GAS and GLD from GNU binutils version 2.7 or later for
2009 this attribute to work correctly.
2012 @cindex interrupt handler functions
2013 Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16
2014 ports to indicate that the specified function is an interrupt handler.
2015 The compiler will generate function entry and exit sequences suitable
2016 for use in an interrupt handler when this attribute is present.
2018 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and
2019 SH processors can be specified via the @code{interrupt_handler} attribute.
2021 Note, on the AVR, interrupts will be enabled inside the function.
2023 Note, for the ARM, you can specify the kind of interrupt to be handled by
2024 adding an optional parameter to the interrupt attribute like this:
2027 void f () __attribute__ ((interrupt ("IRQ")));
2030 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2032 @item interrupt_handler
2033 @cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors
2034 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to
2035 indicate that the specified function is an interrupt handler. The compiler
2036 will generate function entry and exit sequences suitable for use in an
2037 interrupt handler when this attribute is present.
2040 @cindex User stack pointer in interrupts on the Blackfin
2041 When used together with @code{interrupt_handler}, @code{exception_handler}
2042 or @code{nmi_handler}, code will be generated to load the stack pointer
2043 from the USP register in the function prologue.
2045 @item long_call/short_call
2046 @cindex indirect calls on ARM
2047 This attribute specifies how a particular function is called on
2048 ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2049 command line switch and @code{#pragma long_calls} settings. The
2050 @code{long_call} attribute indicates that the function might be far
2051 away from the call site and require a different (more expensive)
2052 calling sequence. The @code{short_call} attribute always places
2053 the offset to the function from the call site into the @samp{BL}
2054 instruction directly.
2056 @item longcall/shortcall
2057 @cindex functions called via pointer on the RS/6000 and PowerPC
2058 On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute
2059 indicates that the function might be far away from the call site and
2060 require a different (more expensive) calling sequence. The
2061 @code{shortcall} attribute indicates that the function is always close
2062 enough for the shorter calling sequence to be used. These attributes
2063 override both the @option{-mlongcall} switch and, on the RS/6000 and
2064 PowerPC, the @code{#pragma longcall} setting.
2066 @xref{RS/6000 and PowerPC Options}, for more information on whether long
2067 calls are necessary.
2070 @cindex indirect calls on MIPS
2071 This attribute specifies how a particular function is called on MIPS@.
2072 The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options})
2073 command line switch. This attribute causes the compiler to always call
2074 the function by first loading its address into a register, and then using
2075 the contents of that register.
2078 @cindex @code{malloc} attribute
2079 The @code{malloc} attribute is used to tell the compiler that a function
2080 may be treated as if any non-@code{NULL} pointer it returns cannot
2081 alias any other pointer valid when the function returns.
2082 This will often improve optimization.
2083 Standard functions with this property include @code{malloc} and
2084 @code{calloc}. @code{realloc}-like functions have this property as
2085 long as the old pointer is never referred to (including comparing it
2086 to the new pointer) after the function returns a non-@code{NULL}
2089 @item model (@var{model-name})
2090 @cindex function addressability on the M32R/D
2091 @cindex variable addressability on the IA-64
2093 On the M32R/D, use this attribute to set the addressability of an
2094 object, and of the code generated for a function. The identifier
2095 @var{model-name} is one of @code{small}, @code{medium}, or
2096 @code{large}, representing each of the code models.
2098 Small model objects live in the lower 16MB of memory (so that their
2099 addresses can be loaded with the @code{ld24} instruction), and are
2100 callable with the @code{bl} instruction.
2102 Medium model objects may live anywhere in the 32-bit address space (the
2103 compiler will generate @code{seth/add3} instructions to load their addresses),
2104 and are callable with the @code{bl} instruction.
2106 Large model objects may live anywhere in the 32-bit address space (the
2107 compiler will generate @code{seth/add3} instructions to load their addresses),
2108 and may not be reachable with the @code{bl} instruction (the compiler will
2109 generate the much slower @code{seth/add3/jl} instruction sequence).
2111 On IA-64, use this attribute to set the addressability of an object.
2112 At present, the only supported identifier for @var{model-name} is
2113 @code{small}, indicating addressability via ``small'' (22-bit)
2114 addresses (so that their addresses can be loaded with the @code{addl}
2115 instruction). Caveat: such addressing is by definition not position
2116 independent and hence this attribute must not be used for objects
2117 defined by shared libraries.
2120 @cindex function without a prologue/epilogue code
2121 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2122 specified function does not need prologue/epilogue sequences generated by
2123 the compiler. It is up to the programmer to provide these sequences.
2126 @cindex functions which do not handle memory bank switching on 68HC11/68HC12
2127 On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2128 use the normal calling convention based on @code{jsr} and @code{rts}.
2129 This attribute can be used to cancel the effect of the @option{-mlong-calls}
2133 @cindex Allow nesting in an interrupt handler on the Blackfin processor.
2134 Use this attribute together with @code{interrupt_handler},
2135 @code{exception_handler} or @code{nmi_handler} to indicate that the function
2136 entry code should enable nested interrupts or exceptions.
2139 @cindex NMI handler functions on the Blackfin processor
2140 Use this attribute on the Blackfin to indicate that the specified function
2141 is an NMI handler. The compiler will generate function entry and
2142 exit sequences suitable for use in an NMI handler when this
2143 attribute is present.
2145 @item no_instrument_function
2146 @cindex @code{no_instrument_function} function attribute
2147 @opindex finstrument-functions
2148 If @option{-finstrument-functions} is given, profiling function calls will
2149 be generated at entry and exit of most user-compiled functions.
2150 Functions with this attribute will not be so instrumented.
2153 @cindex @code{noinline} function attribute
2154 This function attribute prevents a function from being considered for
2157 @item nonnull (@var{arg-index}, @dots{})
2158 @cindex @code{nonnull} function attribute
2159 The @code{nonnull} attribute specifies that some function parameters should
2160 be non-null pointers. For instance, the declaration:
2164 my_memcpy (void *dest, const void *src, size_t len)
2165 __attribute__((nonnull (1, 2)));
2169 causes the compiler to check that, in calls to @code{my_memcpy},
2170 arguments @var{dest} and @var{src} are non-null. If the compiler
2171 determines that a null pointer is passed in an argument slot marked
2172 as non-null, and the @option{-Wnonnull} option is enabled, a warning
2173 is issued. The compiler may also choose to make optimizations based
2174 on the knowledge that certain function arguments will not be null.
2176 If no argument index list is given to the @code{nonnull} attribute,
2177 all pointer arguments are marked as non-null. To illustrate, the
2178 following declaration is equivalent to the previous example:
2182 my_memcpy (void *dest, const void *src, size_t len)
2183 __attribute__((nonnull));
2187 @cindex @code{noreturn} function attribute
2188 A few standard library functions, such as @code{abort} and @code{exit},
2189 cannot return. GCC knows this automatically. Some programs define
2190 their own functions that never return. You can declare them
2191 @code{noreturn} to tell the compiler this fact. For example,
2195 void fatal () __attribute__ ((noreturn));
2198 fatal (/* @r{@dots{}} */)
2200 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2206 The @code{noreturn} keyword tells the compiler to assume that
2207 @code{fatal} cannot return. It can then optimize without regard to what
2208 would happen if @code{fatal} ever did return. This makes slightly
2209 better code. More importantly, it helps avoid spurious warnings of
2210 uninitialized variables.
2212 The @code{noreturn} keyword does not affect the exceptional path when that
2213 applies: a @code{noreturn}-marked function may still return to the caller
2214 by throwing an exception or calling @code{longjmp}.
2216 Do not assume that registers saved by the calling function are
2217 restored before calling the @code{noreturn} function.
2219 It does not make sense for a @code{noreturn} function to have a return
2220 type other than @code{void}.
2222 The attribute @code{noreturn} is not implemented in GCC versions
2223 earlier than 2.5. An alternative way to declare that a function does
2224 not return, which works in the current version and in some older
2225 versions, is as follows:
2228 typedef void voidfn ();
2230 volatile voidfn fatal;
2233 This approach does not work in GNU C++.
2236 @cindex @code{nothrow} function attribute
2237 The @code{nothrow} attribute is used to inform the compiler that a
2238 function cannot throw an exception. For example, most functions in
2239 the standard C library can be guaranteed not to throw an exception
2240 with the notable exceptions of @code{qsort} and @code{bsearch} that
2241 take function pointer arguments. The @code{nothrow} attribute is not
2242 implemented in GCC versions earlier than 3.3.
2245 @cindex @code{pure} function attribute
2246 Many functions have no effects except the return value and their
2247 return value depends only on the parameters and/or global variables.
2248 Such a function can be subject
2249 to common subexpression elimination and loop optimization just as an
2250 arithmetic operator would be. These functions should be declared
2251 with the attribute @code{pure}. For example,
2254 int square (int) __attribute__ ((pure));
2258 says that the hypothetical function @code{square} is safe to call
2259 fewer times than the program says.
2261 Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2262 Interesting non-pure functions are functions with infinite loops or those
2263 depending on volatile memory or other system resource, that may change between
2264 two consecutive calls (such as @code{feof} in a multithreading environment).
2266 The attribute @code{pure} is not implemented in GCC versions earlier
2269 @item regparm (@var{number})
2270 @cindex @code{regparm} attribute
2271 @cindex functions that are passed arguments in registers on the 386
2272 On the Intel 386, the @code{regparm} attribute causes the compiler to
2273 pass arguments number one to @var{number} if they are of integral type
2274 in registers EAX, EDX, and ECX instead of on the stack. Functions that
2275 take a variable number of arguments will continue to be passed all of their
2276 arguments on the stack.
2278 Beware that on some ELF systems this attribute is unsuitable for
2279 global functions in shared libraries with lazy binding (which is the
2280 default). Lazy binding will send the first call via resolving code in
2281 the loader, which might assume EAX, EDX and ECX can be clobbered, as
2282 per the standard calling conventions. Solaris 8 is affected by this.
2283 GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2284 safe since the loaders there save all registers. (Lazy binding can be
2285 disabled with the linker or the loader if desired, to avoid the
2289 @cindex @code{sseregparm} attribute
2290 On the Intel 386 with SSE support, the @code{sseregparm} attribute
2291 causes the compiler to pass up to 3 floating point arguments in
2292 SSE registers instead of on the stack. Functions that take a
2293 variable number of arguments will continue to pass all of their
2294 floating point arguments on the stack.
2296 @item force_align_arg_pointer
2297 @cindex @code{force_align_arg_pointer} attribute
2298 On the Intel x86, the @code{force_align_arg_pointer} attribute may be
2299 applied to individual function definitions, generating an alternate
2300 prologue and epilogue that realigns the runtime stack. This supports
2301 mixing legacy codes that run with a 4-byte aligned stack with modern
2302 codes that keep a 16-byte stack for SSE compatibility. The alternate
2303 prologue and epilogue are slower and bigger than the regular ones, and
2304 the alternate prologue requires a scratch register; this lowers the
2305 number of registers available if used in conjunction with the
2306 @code{regparm} attribute. The @code{force_align_arg_pointer}
2307 attribute is incompatible with nested functions; this is considered a
2311 @cindex @code{returns_twice} attribute
2312 The @code{returns_twice} attribute tells the compiler that a function may
2313 return more than one time. The compiler will ensure that all registers
2314 are dead before calling such a function and will emit a warning about
2315 the variables that may be clobbered after the second return from the
2316 function. Examples of such functions are @code{setjmp} and @code{vfork}.
2317 The @code{longjmp}-like counterpart of such function, if any, might need
2318 to be marked with the @code{noreturn} attribute.
2321 @cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S
2322 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that
2323 all registers except the stack pointer should be saved in the prologue
2324 regardless of whether they are used or not.
2326 @item section ("@var{section-name}")
2327 @cindex @code{section} function attribute
2328 Normally, the compiler places the code it generates in the @code{text} section.
2329 Sometimes, however, you need additional sections, or you need certain
2330 particular functions to appear in special sections. The @code{section}
2331 attribute specifies that a function lives in a particular section.
2332 For example, the declaration:
2335 extern void foobar (void) __attribute__ ((section ("bar")));
2339 puts the function @code{foobar} in the @code{bar} section.
2341 Some file formats do not support arbitrary sections so the @code{section}
2342 attribute is not available on all platforms.
2343 If you need to map the entire contents of a module to a particular
2344 section, consider using the facilities of the linker instead.
2347 @cindex @code{sentinel} function attribute
2348 This function attribute ensures that a parameter in a function call is
2349 an explicit @code{NULL}. The attribute is only valid on variadic
2350 functions. By default, the sentinel is located at position zero, the
2351 last parameter of the function call. If an optional integer position
2352 argument P is supplied to the attribute, the sentinel must be located at
2353 position P counting backwards from the end of the argument list.
2356 __attribute__ ((sentinel))
2358 __attribute__ ((sentinel(0)))
2361 The attribute is automatically set with a position of 0 for the built-in
2362 functions @code{execl} and @code{execlp}. The built-in function
2363 @code{execle} has the attribute set with a position of 1.
2365 A valid @code{NULL} in this context is defined as zero with any pointer
2366 type. If your system defines the @code{NULL} macro with an integer type
2367 then you need to add an explicit cast. GCC replaces @code{stddef.h}
2368 with a copy that redefines NULL appropriately.
2370 The warnings for missing or incorrect sentinels are enabled with
2374 See long_call/short_call.
2377 See longcall/shortcall.
2380 @cindex signal handler functions on the AVR processors
2381 Use this attribute on the AVR to indicate that the specified
2382 function is a signal handler. The compiler will generate function
2383 entry and exit sequences suitable for use in a signal handler when this
2384 attribute is present. Interrupts will be disabled inside the function.
2387 Use this attribute on the SH to indicate an @code{interrupt_handler}
2388 function should switch to an alternate stack. It expects a string
2389 argument that names a global variable holding the address of the
2394 void f () __attribute__ ((interrupt_handler,
2395 sp_switch ("alt_stack")));
2399 @cindex functions that pop the argument stack on the 386
2400 On the Intel 386, the @code{stdcall} attribute causes the compiler to
2401 assume that the called function will pop off the stack space used to
2402 pass arguments, unless it takes a variable number of arguments.
2405 @cindex tiny data section on the H8/300H and H8S
2406 Use this attribute on the H8/300H and H8S to indicate that the specified
2407 variable should be placed into the tiny data section.
2408 The compiler will generate more efficient code for loads and stores
2409 on data in the tiny data section. Note the tiny data area is limited to
2410 slightly under 32kbytes of data.
2413 Use this attribute on the SH for an @code{interrupt_handler} to return using
2414 @code{trapa} instead of @code{rte}. This attribute expects an integer
2415 argument specifying the trap number to be used.
2418 @cindex @code{unused} attribute.
2419 This attribute, attached to a function, means that the function is meant
2420 to be possibly unused. GCC will not produce a warning for this
2424 @cindex @code{used} attribute.
2425 This attribute, attached to a function, means that code must be emitted
2426 for the function even if it appears that the function is not referenced.
2427 This is useful, for example, when the function is referenced only in
2430 @item visibility ("@var{visibility_type}")
2431 @cindex @code{visibility} attribute
2432 This attribute affects the linkage of the declaration to which it is attached.
2433 There are four supported @var{visibility_type} values: default,
2434 hidden, protected or internal visibility.
2437 void __attribute__ ((visibility ("protected")))
2438 f () @{ /* @r{Do something.} */; @}
2439 int i __attribute__ ((visibility ("hidden")));
2442 The possible values of @var{visibility_type} correspond to the
2443 visibility settings in the ELF gABI.
2446 @c keep this list of visibilities in alphabetical order.
2449 Default visibility is the normal case for the object file format.
2450 This value is available for the visibility attribute to override other
2451 options that may change the assumed visibility of entities.
2453 On ELF, default visibility means that the declaration is visible to other
2454 modules and, in shared libraries, means that the declared entity may be
2457 On Darwin, default visibility means that the declaration is visible to
2460 Default visibility corresponds to ``external linkage'' in the language.
2463 Hidden visibility indicates that the entity declared will have a new
2464 form of linkage, which we'll call ``hidden linkage''. Two
2465 declarations of an object with hidden linkage refer to the same object
2466 if they are in the same shared object.
2469 Internal visibility is like hidden visibility, but with additional
2470 processor specific semantics. Unless otherwise specified by the
2471 psABI, GCC defines internal visibility to mean that a function is
2472 @emph{never} called from another module. Compare this with hidden
2473 functions which, while they cannot be referenced directly by other
2474 modules, can be referenced indirectly via function pointers. By
2475 indicating that a function cannot be called from outside the module,
2476 GCC may for instance omit the load of a PIC register since it is known
2477 that the calling function loaded the correct value.
2480 Protected visibility is like default visibility except that it
2481 indicates that references within the defining module will bind to the
2482 definition in that module. That is, the declared entity cannot be
2483 overridden by another module.
2487 All visibilities are supported on many, but not all, ELF targets
2488 (supported when the assembler supports the @samp{.visibility}
2489 pseudo-op). Default visibility is supported everywhere. Hidden
2490 visibility is supported on Darwin targets.
2492 The visibility attribute should be applied only to declarations which
2493 would otherwise have external linkage. The attribute should be applied
2494 consistently, so that the same entity should not be declared with
2495 different settings of the attribute.
2497 In C++, the visibility attribute applies to types as well as functions
2498 and objects, because in C++ types have linkage. A class must not have
2499 greater visibility than its non-static data member types and bases,
2500 and class members default to the visibility of their class. Also, a
2501 declaration without explicit visibility is limited to the visibility
2504 In C++, you can mark member functions and static member variables of a
2505 class with the visibility attribute. This is useful if if you know a
2506 particular method or static member variable should only be used from
2507 one shared object; then you can mark it hidden while the rest of the
2508 class has default visibility. Care must be taken to avoid breaking
2509 the One Definition Rule; for example, it is usually not useful to mark
2510 an inline method as hidden without marking the whole class as hidden.
2512 A C++ namespace declaration can also have the visibility attribute.
2513 This attribute applies only to the particular namespace body, not to
2514 other definitions of the same namespace; it is equivalent to using
2515 @samp{#pragma GCC visibility} before and after the namespace
2516 definition (@pxref{Visibility Pragmas}).
2518 In C++, if a template argument has limited visibility, this
2519 restriction is implicitly propagated to the template instantiation.
2520 Otherwise, template instantiations and specializations default to the
2521 visibility of their template.
2523 If both the template and enclosing class have explicit visibility, the
2524 visibility from the template is used.
2526 @item warn_unused_result
2527 @cindex @code{warn_unused_result} attribute
2528 The @code{warn_unused_result} attribute causes a warning to be emitted
2529 if a caller of the function with this attribute does not use its
2530 return value. This is useful for functions where not checking
2531 the result is either a security problem or always a bug, such as
2535 int fn () __attribute__ ((warn_unused_result));
2538 if (fn () < 0) return -1;
2544 results in warning on line 5.
2547 @cindex @code{weak} attribute
2548 The @code{weak} attribute causes the declaration to be emitted as a weak
2549 symbol rather than a global. This is primarily useful in defining
2550 library functions which can be overridden in user code, though it can
2551 also be used with non-function declarations. Weak symbols are supported
2552 for ELF targets, and also for a.out targets when using the GNU assembler
2556 @itemx weakref ("@var{target}")
2557 @cindex @code{weakref} attribute
2558 The @code{weakref} attribute marks a declaration as a weak reference.
2559 Without arguments, it should be accompanied by an @code{alias} attribute
2560 naming the target symbol. Optionally, the @var{target} may be given as
2561 an argument to @code{weakref} itself. In either case, @code{weakref}
2562 implicitly marks the declaration as @code{weak}. Without a
2563 @var{target}, given as an argument to @code{weakref} or to @code{alias},
2564 @code{weakref} is equivalent to @code{weak}.
2567 static int x() __attribute__ ((weakref ("y")));
2568 /* is equivalent to... */
2569 static int x() __attribute__ ((weak, weakref, alias ("y")));
2571 static int x() __attribute__ ((weakref));
2572 static int x() __attribute__ ((alias ("y")));
2575 A weak reference is an alias that does not by itself require a
2576 definition to be given for the target symbol. If the target symbol is
2577 only referenced through weak references, then the becomes a @code{weak}
2578 undefined symbol. If it is directly referenced, however, then such
2579 strong references prevail, and a definition will be required for the
2580 symbol, not necessarily in the same translation unit.
2582 The effect is equivalent to moving all references to the alias to a
2583 separate translation unit, renaming the alias to the aliased symbol,
2584 declaring it as weak, compiling the two separate translation units and
2585 performing a reloadable link on them.
2587 At present, a declaration to which @code{weakref} is attached can
2588 only be @code{static}.
2590 @item externally_visible
2591 @cindex @code{externally_visible} attribute.
2592 This attribute, attached to a global variable or function nullify
2593 effect of @option{-fwhole-program} command line option, so the object
2594 remain visible outside the current compilation unit
2598 You can specify multiple attributes in a declaration by separating them
2599 by commas within the double parentheses or by immediately following an
2600 attribute declaration with another attribute declaration.
2602 @cindex @code{#pragma}, reason for not using
2603 @cindex pragma, reason for not using
2604 Some people object to the @code{__attribute__} feature, suggesting that
2605 ISO C's @code{#pragma} should be used instead. At the time
2606 @code{__attribute__} was designed, there were two reasons for not doing
2611 It is impossible to generate @code{#pragma} commands from a macro.
2614 There is no telling what the same @code{#pragma} might mean in another
2618 These two reasons applied to almost any application that might have been
2619 proposed for @code{#pragma}. It was basically a mistake to use
2620 @code{#pragma} for @emph{anything}.
2622 The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2623 to be generated from macros. In addition, a @code{#pragma GCC}
2624 namespace is now in use for GCC-specific pragmas. However, it has been
2625 found convenient to use @code{__attribute__} to achieve a natural
2626 attachment of attributes to their corresponding declarations, whereas
2627 @code{#pragma GCC} is of use for constructs that do not naturally form
2628 part of the grammar. @xref{Other Directives,,Miscellaneous
2629 Preprocessing Directives, cpp, The GNU C Preprocessor}.
2631 @node Attribute Syntax
2632 @section Attribute Syntax
2633 @cindex attribute syntax
2635 This section describes the syntax with which @code{__attribute__} may be
2636 used, and the constructs to which attribute specifiers bind, for the C
2637 language. Some details may vary for C++ and Objective-C@. Because of
2638 infelicities in the grammar for attributes, some forms described here
2639 may not be successfully parsed in all cases.
2641 There are some problems with the semantics of attributes in C++. For
2642 example, there are no manglings for attributes, although they may affect
2643 code generation, so problems may arise when attributed types are used in
2644 conjunction with templates or overloading. Similarly, @code{typeid}
2645 does not distinguish between types with different attributes. Support
2646 for attributes in C++ may be restricted in future to attributes on
2647 declarations only, but not on nested declarators.
2649 @xref{Function Attributes}, for details of the semantics of attributes
2650 applying to functions. @xref{Variable Attributes}, for details of the
2651 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
2652 semantics of attributes applying to variables. @xref{Type
2653 Attributes}, for details of the semantics of attributes applying to
2654 structure, union and enumerated types. @xref{Label Attributes}, for
2655 details of the semantics of attributes applying to labels and
2658 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
2659 An @dfn{attribute specifier} is of the form
2660 @code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list}
2661 is a possibly empty comma-separated sequence of @dfn{attributes}, where
2662 each attribute is one of the following:
2666 Empty. Empty attributes are ignored.
2669 A word (which may be an identifier such as @code{unused}, or a reserved
2670 word such as @code{const}).
2673 A word, followed by, in parentheses, parameters for the attribute.
2674 These parameters take one of the following forms:
2678 An identifier. For example, @code{mode} attributes use this form.
2681 An identifier followed by a comma and a non-empty comma-separated list
2682 of expressions. For example, @code{format} attributes use this form.
2685 A possibly empty comma-separated list of expressions. For example,
2686 @code{format_arg} attributes use this form with the list being a single
2687 integer constant expression, and @code{alias} attributes use this form
2688 with the list being a single string constant.
2692 An @dfn{attribute specifier list} is a sequence of one or more attribute
2693 specifiers, not separated by any other tokens.
2695 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
2696 In GNU C, an attribute specifier list may appear after the colon
2697 following a label, other than a @code{case} or @code{default} label.
2698 GNU C++ does not permit such placement of attribute lists, as it is
2699 permissible for a declaration, which could begin with an attribute
2700 list, to be labelled in C++. Declarations cannot be labelled in C90
2701 or C99, so the ambiguity does not arise there.
2703 In GNU C an attribute specifier list may also appear after the keyword
2704 @code{while} in a while loop, after @code{do} and after @code{for}.
2706 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
2707 An attribute specifier list may appear as part of a @code{struct},
2708 @code{union} or @code{enum} specifier. It may go either immediately
2709 after the @code{struct}, @code{union} or @code{enum} keyword, or after
2710 the closing brace. The former syntax is preferred.
2711 Where attribute specifiers follow the closing brace, they are considered
2712 to relate to the structure, union or enumerated type defined, not to any
2713 enclosing declaration the type specifier appears in, and the type
2714 defined is not complete until after the attribute specifiers.
2715 @c Otherwise, there would be the following problems: a shift/reduce
2716 @c conflict between attributes binding the struct/union/enum and
2717 @c binding to the list of specifiers/qualifiers; and "aligned"
2718 @c attributes could use sizeof for the structure, but the size could be
2719 @c changed later by "packed" attributes.
2721 Otherwise, an attribute specifier appears as part of a declaration,
2722 counting declarations of unnamed parameters and type names, and relates
2723 to that declaration (which may be nested in another declaration, for
2724 example in the case of a parameter declaration), or to a particular declarator
2725 within a declaration. Where an
2726 attribute specifier is applied to a parameter declared as a function or
2727 an array, it should apply to the function or array rather than the
2728 pointer to which the parameter is implicitly converted, but this is not
2729 yet correctly implemented.
2731 Any list of specifiers and qualifiers at the start of a declaration may
2732 contain attribute specifiers, whether or not such a list may in that
2733 context contain storage class specifiers. (Some attributes, however,
2734 are essentially in the nature of storage class specifiers, and only make
2735 sense where storage class specifiers may be used; for example,
2736 @code{section}.) There is one necessary limitation to this syntax: the
2737 first old-style parameter declaration in a function definition cannot
2738 begin with an attribute specifier, because such an attribute applies to
2739 the function instead by syntax described below (which, however, is not
2740 yet implemented in this case). In some other cases, attribute
2741 specifiers are permitted by this grammar but not yet supported by the
2742 compiler. All attribute specifiers in this place relate to the
2743 declaration as a whole. In the obsolescent usage where a type of
2744 @code{int} is implied by the absence of type specifiers, such a list of
2745 specifiers and qualifiers may be an attribute specifier list with no
2746 other specifiers or qualifiers.
2748 At present, the first parameter in a function prototype must have some
2749 type specifier which is not an attribute specifier; this resolves an
2750 ambiguity in the interpretation of @code{void f(int
2751 (__attribute__((foo)) x))}, but is subject to change. At present, if
2752 the parentheses of a function declarator contain only attributes then
2753 those attributes are ignored, rather than yielding an error or warning
2754 or implying a single parameter of type int, but this is subject to
2757 An attribute specifier list may appear immediately before a declarator
2758 (other than the first) in a comma-separated list of declarators in a
2759 declaration of more than one identifier using a single list of
2760 specifiers and qualifiers. Such attribute specifiers apply
2761 only to the identifier before whose declarator they appear. For
2765 __attribute__((noreturn)) void d0 (void),
2766 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2771 the @code{noreturn} attribute applies to all the functions
2772 declared; the @code{format} attribute only applies to @code{d1}.
2774 An attribute specifier list may appear immediately before the comma,
2775 @code{=} or semicolon terminating the declaration of an identifier other
2776 than a function definition. At present, such attribute specifiers apply
2777 to the declared object or function, but in future they may attach to the
2778 outermost adjacent declarator. In simple cases there is no difference,
2779 but, for example, in
2782 void (****f)(void) __attribute__((noreturn));
2786 at present the @code{noreturn} attribute applies to @code{f}, which
2787 causes a warning since @code{f} is not a function, but in future it may
2788 apply to the function @code{****f}. The precise semantics of what
2789 attributes in such cases will apply to are not yet specified. Where an
2790 assembler name for an object or function is specified (@pxref{Asm
2791 Labels}), at present the attribute must follow the @code{asm}
2792 specification; in future, attributes before the @code{asm} specification
2793 may apply to the adjacent declarator, and those after it to the declared
2796 An attribute specifier list may, in future, be permitted to appear after
2797 the declarator in a function definition (before any old-style parameter
2798 declarations or the function body).
2800 Attribute specifiers may be mixed with type qualifiers appearing inside
2801 the @code{[]} of a parameter array declarator, in the C99 construct by
2802 which such qualifiers are applied to the pointer to which the array is
2803 implicitly converted. Such attribute specifiers apply to the pointer,
2804 not to the array, but at present this is not implemented and they are
2807 An attribute specifier list may appear at the start of a nested
2808 declarator. At present, there are some limitations in this usage: the
2809 attributes correctly apply to the declarator, but for most individual
2810 attributes the semantics this implies are not implemented.
2811 When attribute specifiers follow the @code{*} of a pointer
2812 declarator, they may be mixed with any type qualifiers present.
2813 The following describes the formal semantics of this syntax. It will make the
2814 most sense if you are familiar with the formal specification of
2815 declarators in the ISO C standard.
2817 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2818 D1}, where @code{T} contains declaration specifiers that specify a type
2819 @var{Type} (such as @code{int}) and @code{D1} is a declarator that
2820 contains an identifier @var{ident}. The type specified for @var{ident}
2821 for derived declarators whose type does not include an attribute
2822 specifier is as in the ISO C standard.
2824 If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2825 and the declaration @code{T D} specifies the type
2826 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2827 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2828 @var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2830 If @code{D1} has the form @code{*
2831 @var{type-qualifier-and-attribute-specifier-list} D}, and the
2832 declaration @code{T D} specifies the type
2833 ``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2834 @code{T D1} specifies the type ``@var{derived-declarator-type-list}
2835 @var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2841 void (__attribute__((noreturn)) ****f) (void);
2845 specifies the type ``pointer to pointer to pointer to pointer to
2846 non-returning function returning @code{void}''. As another example,
2849 char *__attribute__((aligned(8))) *f;
2853 specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2854 Note again that this does not work with most attributes; for example,
2855 the usage of @samp{aligned} and @samp{noreturn} attributes given above
2856 is not yet supported.
2858 For compatibility with existing code written for compiler versions that
2859 did not implement attributes on nested declarators, some laxity is
2860 allowed in the placing of attributes. If an attribute that only applies
2861 to types is applied to a declaration, it will be treated as applying to
2862 the type of that declaration. If an attribute that only applies to
2863 declarations is applied to the type of a declaration, it will be treated
2864 as applying to that declaration; and, for compatibility with code
2865 placing the attributes immediately before the identifier declared, such
2866 an attribute applied to a function return type will be treated as
2867 applying to the function type, and such an attribute applied to an array
2868 element type will be treated as applying to the array type. If an
2869 attribute that only applies to function types is applied to a
2870 pointer-to-function type, it will be treated as applying to the pointer
2871 target type; if such an attribute is applied to a function return type
2872 that is not a pointer-to-function type, it will be treated as applying
2873 to the function type.
2875 @node Function Prototypes
2876 @section Prototypes and Old-Style Function Definitions
2877 @cindex function prototype declarations
2878 @cindex old-style function definitions
2879 @cindex promotion of formal parameters
2881 GNU C extends ISO C to allow a function prototype to override a later
2882 old-style non-prototype definition. Consider the following example:
2885 /* @r{Use prototypes unless the compiler is old-fashioned.} */
2892 /* @r{Prototype function declaration.} */
2893 int isroot P((uid_t));
2895 /* @r{Old-style function definition.} */
2897 isroot (x) /* @r{??? lossage here ???} */
2904 Suppose the type @code{uid_t} happens to be @code{short}. ISO C does
2905 not allow this example, because subword arguments in old-style
2906 non-prototype definitions are promoted. Therefore in this example the
2907 function definition's argument is really an @code{int}, which does not
2908 match the prototype argument type of @code{short}.
2910 This restriction of ISO C makes it hard to write code that is portable
2911 to traditional C compilers, because the programmer does not know
2912 whether the @code{uid_t} type is @code{short}, @code{int}, or
2913 @code{long}. Therefore, in cases like these GNU C allows a prototype
2914 to override a later old-style definition. More precisely, in GNU C, a
2915 function prototype argument type overrides the argument type specified
2916 by a later old-style definition if the former type is the same as the
2917 latter type before promotion. Thus in GNU C the above example is
2918 equivalent to the following:
2931 GNU C++ does not support old-style function definitions, so this
2932 extension is irrelevant.
2935 @section C++ Style Comments
2937 @cindex C++ comments
2938 @cindex comments, C++ style
2940 In GNU C, you may use C++ style comments, which start with @samp{//} and
2941 continue until the end of the line. Many other C implementations allow
2942 such comments, and they are included in the 1999 C standard. However,
2943 C++ style comments are not recognized if you specify an @option{-std}
2944 option specifying a version of ISO C before C99, or @option{-ansi}
2945 (equivalent to @option{-std=c89}).
2948 @section Dollar Signs in Identifier Names
2950 @cindex dollar signs in identifier names
2951 @cindex identifier names, dollar signs in
2953 In GNU C, you may normally use dollar signs in identifier names.
2954 This is because many traditional C implementations allow such identifiers.
2955 However, dollar signs in identifiers are not supported on a few target
2956 machines, typically because the target assembler does not allow them.
2958 @node Character Escapes
2959 @section The Character @key{ESC} in Constants
2961 You can use the sequence @samp{\e} in a string or character constant to
2962 stand for the ASCII character @key{ESC}.
2965 @section Inquiring on Alignment of Types or Variables
2967 @cindex type alignment
2968 @cindex variable alignment
2970 The keyword @code{__alignof__} allows you to inquire about how an object
2971 is aligned, or the minimum alignment usually required by a type. Its
2972 syntax is just like @code{sizeof}.
2974 For example, if the target machine requires a @code{double} value to be
2975 aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
2976 This is true on many RISC machines. On more traditional machine
2977 designs, @code{__alignof__ (double)} is 4 or even 2.
2979 Some machines never actually require alignment; they allow reference to any
2980 data type even at an odd address. For these machines, @code{__alignof__}
2981 reports the @emph{recommended} alignment of a type.
2983 If the operand of @code{__alignof__} is an lvalue rather than a type,
2984 its value is the required alignment for its type, taking into account
2985 any minimum alignment specified with GCC's @code{__attribute__}
2986 extension (@pxref{Variable Attributes}). For example, after this
2990 struct foo @{ int x; char y; @} foo1;
2994 the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
2995 alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
2997 It is an error to ask for the alignment of an incomplete type.
2999 @node Variable Attributes
3000 @section Specifying Attributes of Variables
3001 @cindex attribute of variables
3002 @cindex variable attributes
3004 The keyword @code{__attribute__} allows you to specify special
3005 attributes of variables or structure fields. This keyword is followed
3006 by an attribute specification inside double parentheses. Some
3007 attributes are currently defined generically for variables.
3008 Other attributes are defined for variables on particular target
3009 systems. Other attributes are available for functions
3010 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
3011 (@pxref{Function Attributes}), types (@pxref{Type Attributes}) and
3012 labels (@pxref{Label Attributes}). Other front ends might define
3013 more attributes (@pxref{C++ Extensions,,Extensions to the C++ Language}).
3015 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
3016 You may also specify attributes with @samp{__} preceding and following
3017 each keyword. This allows you to use them in header files without
3018 being concerned about a possible macro of the same name. For example,
3019 you may use @code{__aligned__} instead of @code{aligned}.
3021 @xref{Attribute Syntax}, for details of the exact syntax for using
3025 @cindex @code{aligned} attribute
3026 @item aligned (@var{alignment})
3027 This attribute specifies a minimum alignment for the variable or
3028 structure field, measured in bytes. For example, the declaration:
3031 int x __attribute__ ((aligned (16))) = 0;
3035 causes the compiler to allocate the global variable @code{x} on a
3036 16-byte boundary. On a 68040, this could be used in conjunction with
3037 an @code{asm} expression to access the @code{move16} instruction which
3038 requires 16-byte aligned operands.
3040 You can also specify the alignment of structure fields. For example, to
3041 create a double-word aligned @code{int} pair, you could write:
3044 struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3048 This is an alternative to creating a union with a @code{double} member
3049 that forces the union to be double-word aligned.
3051 As in the preceding examples, you can explicitly specify the alignment
3052 (in bytes) that you wish the compiler to use for a given variable or
3053 structure field. Alternatively, you can leave out the alignment factor
3054 and just ask the compiler to align a variable or field to the maximum
3055 useful alignment for the target machine you are compiling for. For
3056 example, you could write:
3059 short array[3] __attribute__ ((aligned));
3062 Whenever you leave out the alignment factor in an @code{aligned} attribute
3063 specification, the compiler automatically sets the alignment for the declared
3064 variable or field to the largest alignment which is ever used for any data
3065 type on the target machine you are compiling for. Doing this can often make
3066 copy operations more efficient, because the compiler can use whatever
3067 instructions copy the biggest chunks of memory when performing copies to
3068 or from the variables or fields that you have aligned this way.
3070 The @code{aligned} attribute can only increase the alignment; but you
3071 can decrease it by specifying @code{packed} as well. See below.
3073 Note that the effectiveness of @code{aligned} attributes may be limited
3074 by inherent limitations in your linker. On many systems, the linker is
3075 only able to arrange for variables to be aligned up to a certain maximum
3076 alignment. (For some linkers, the maximum supported alignment may
3077 be very very small.) If your linker is only able to align variables
3078 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3079 in an @code{__attribute__} will still only provide you with 8 byte
3080 alignment. See your linker documentation for further information.
3082 The @code{aligned} attribute can also be used for functions
3083 (@pxref{Function Attributes}.)
3085 @item cleanup (@var{cleanup_function})
3086 @cindex @code{cleanup} attribute
3087 The @code{cleanup} attribute runs a function when the variable goes
3088 out of scope. This attribute can only be applied to auto function
3089 scope variables; it may not be applied to parameters or variables
3090 with static storage duration. The function must take one parameter,
3091 a pointer to a type compatible with the variable. The return value
3092 of the function (if any) is ignored.
3094 If @option{-fexceptions} is enabled, then @var{cleanup_function}
3095 will be run during the stack unwinding that happens during the
3096 processing of the exception. Note that the @code{cleanup} attribute
3097 does not allow the exception to be caught, only to perform an action.
3098 It is undefined what happens if @var{cleanup_function} does not
3103 @cindex @code{common} attribute
3104 @cindex @code{nocommon} attribute
3107 The @code{common} attribute requests GCC to place a variable in
3108 ``common'' storage. The @code{nocommon} attribute requests the
3109 opposite---to allocate space for it directly.
3111 These attributes override the default chosen by the
3112 @option{-fno-common} and @option{-fcommon} flags respectively.
3115 @cindex @code{deprecated} attribute
3116 The @code{deprecated} attribute results in a warning if the variable
3117 is used anywhere in the source file. This is useful when identifying
3118 variables that are expected to be removed in a future version of a
3119 program. The warning also includes the location of the declaration
3120 of the deprecated variable, to enable users to easily find further
3121 information about why the variable is deprecated, or what they should
3122 do instead. Note that the warning only occurs for uses:
3125 extern int old_var __attribute__ ((deprecated));
3127 int new_fn () @{ return old_var; @}
3130 results in a warning on line 3 but not line 2.
3132 The @code{deprecated} attribute can also be used for functions and
3133 types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3135 @item mode (@var{mode})
3136 @cindex @code{mode} attribute
3137 This attribute specifies the data type for the declaration---whichever
3138 type corresponds to the mode @var{mode}. This in effect lets you
3139 request an integer or floating point type according to its width.
3141 You may also specify a mode of @samp{byte} or @samp{__byte__} to
3142 indicate the mode corresponding to a one-byte integer, @samp{word} or
3143 @samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3144 or @samp{__pointer__} for the mode used to represent pointers.
3147 @cindex @code{packed} attribute
3148 The @code{packed} attribute specifies that a variable or structure field
3149 should have the smallest possible alignment---one byte for a variable,
3150 and one bit for a field, unless you specify a larger value with the
3151 @code{aligned} attribute.
3153 Here is a structure in which the field @code{x} is packed, so that it
3154 immediately follows @code{a}:
3160 int x[2] __attribute__ ((packed));
3164 @item section ("@var{section-name}")
3165 @cindex @code{section} variable attribute
3166 Normally, the compiler places the objects it generates in sections like
3167 @code{data} and @code{bss}. Sometimes, however, you need additional sections,
3168 or you need certain particular variables to appear in special sections,
3169 for example to map to special hardware. The @code{section}
3170 attribute specifies that a variable (or function) lives in a particular
3171 section. For example, this small program uses several specific section names:
3174 struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3175 struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3176 char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3177 int init_data __attribute__ ((section ("INITDATA"))) = 0;
3181 /* @r{Initialize stack pointer} */
3182 init_sp (stack + sizeof (stack));
3184 /* @r{Initialize initialized data} */
3185 memcpy (&init_data, &data, &edata - &data);
3187 /* @r{Turn on the serial ports} */
3194 Use the @code{section} attribute with an @emph{initialized} definition
3195 of a @emph{global} variable, as shown in the example. GCC issues
3196 a warning and otherwise ignores the @code{section} attribute in
3197 uninitialized variable declarations.
3199 You may only use the @code{section} attribute with a fully initialized
3200 global definition because of the way linkers work. The linker requires
3201 each object be defined once, with the exception that uninitialized
3202 variables tentatively go in the @code{common} (or @code{bss}) section
3203 and can be multiply ``defined''. You can force a variable to be
3204 initialized with the @option{-fno-common} flag or the @code{nocommon}
3207 Some file formats do not support arbitrary sections so the @code{section}
3208 attribute is not available on all platforms.
3209 If you need to map the entire contents of a module to a particular
3210 section, consider using the facilities of the linker instead.
3213 @cindex @code{shared} variable attribute
3214 On Microsoft Windows, in addition to putting variable definitions in a named
3215 section, the section can also be shared among all running copies of an
3216 executable or DLL@. For example, this small program defines shared data
3217 by putting it in a named section @code{shared} and marking the section
3221 int foo __attribute__((section ("shared"), shared)) = 0;
3226 /* @r{Read and write foo. All running
3227 copies see the same value.} */
3233 You may only use the @code{shared} attribute along with @code{section}
3234 attribute with a fully initialized global definition because of the way
3235 linkers work. See @code{section} attribute for more information.
3237 The @code{shared} attribute is only available on Microsoft Windows@.
3239 @item tls_model ("@var{tls_model}")
3240 @cindex @code{tls_model} attribute
3241 The @code{tls_model} attribute sets thread-local storage model
3242 (@pxref{Thread-Local}) of a particular @code{__thread} variable,
3243 overriding @option{-ftls-model=} command line switch on a per-variable
3245 The @var{tls_model} argument should be one of @code{global-dynamic},
3246 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3248 Not all targets support this attribute.
3251 This attribute, attached to a variable, means that the variable is meant
3252 to be possibly unused. GCC will not produce a warning for this
3256 This attribute, attached to a variable, means that the variable must be
3257 emitted even if it appears that the variable is not referenced.
3259 @item vector_size (@var{bytes})
3260 This attribute specifies the vector size for the variable, measured in
3261 bytes. For example, the declaration:
3264 int foo __attribute__ ((vector_size (16)));
3268 causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3269 divided into @code{int} sized units. Assuming a 32-bit int (a vector of
3270 4 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3272 This attribute is only applicable to integral and float scalars,
3273 although arrays, pointers, and function return values are allowed in
3274 conjunction with this construct.
3276 Aggregates with this attribute are invalid, even if they are of the same
3277 size as a corresponding scalar. For example, the declaration:
3280 struct S @{ int a; @};
3281 struct S __attribute__ ((vector_size (16))) foo;
3285 is invalid even if the size of the structure is the same as the size of
3289 The @code{selectany} attribute causes an initialized global variable to
3290 have link-once semantics. When multiple definitions of the variable are
3291 encountered by the linker, the first is selected and the remainder are
3292 discarded. Following usage by the Microsoft compiler, the linker is told
3293 @emph{not} to warn about size or content differences of the multiple
3296 Although the primary usage of this attribute is for POD types, the
3297 attribute can also be applied to global C++ objects that are initialized
3298 by a constructor. In this case, the static initialization and destruction
3299 code for the object is emitted in each translation defining the object,
3300 but the calls to the constructor and destructor are protected by a
3301 link-once guard variable.
3303 The @code{selectany} attribute is only available on Microsoft Windows
3304 targets. You can use @code{__declspec (selectany)} as a synonym for
3305 @code{__attribute__ ((selectany))} for compatibility with other
3309 The @code{weak} attribute is described in @xref{Function Attributes}.
3312 The @code{dllimport} attribute is described in @xref{Function Attributes}.
3315 The @code{dllexport} attribute is described in @xref{Function Attributes}.
3319 @subsection M32R/D Variable Attributes
3321 One attribute is currently defined for the M32R/D@.
3324 @item model (@var{model-name})
3325 @cindex variable addressability on the M32R/D
3326 Use this attribute on the M32R/D to set the addressability of an object.
3327 The identifier @var{model-name} is one of @code{small}, @code{medium},
3328 or @code{large}, representing each of the code models.
3330 Small model objects live in the lower 16MB of memory (so that their
3331 addresses can be loaded with the @code{ld24} instruction).
3333 Medium and large model objects may live anywhere in the 32-bit address space
3334 (the compiler will generate @code{seth/add3} instructions to load their
3338 @anchor{i386 Variable Attributes}
3339 @subsection i386 Variable Attributes
3341 Two attributes are currently defined for i386 configurations:
3342 @code{ms_struct} and @code{gcc_struct}
3347 @cindex @code{ms_struct} attribute
3348 @cindex @code{gcc_struct} attribute
3350 If @code{packed} is used on a structure, or if bit-fields are used
3351 it may be that the Microsoft ABI packs them differently
3352 than GCC would normally pack them. Particularly when moving packed
3353 data between functions compiled with GCC and the native Microsoft compiler
3354 (either via function call or as data in a file), it may be necessary to access
3357 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3358 compilers to match the native Microsoft compiler.
3360 The Microsoft structure layout algorithm is fairly simple with the exception
3361 of the bitfield packing:
3363 The padding and alignment of members of structures and whether a bit field
3364 can straddle a storage-unit boundary
3367 @item Structure members are stored sequentially in the order in which they are
3368 declared: the first member has the lowest memory address and the last member
3371 @item Every data object has an alignment-requirement. The alignment-requirement
3372 for all data except structures, unions, and arrays is either the size of the
3373 object or the current packing size (specified with either the aligned attribute
3374 or the pack pragma), whichever is less. For structures, unions, and arrays,
3375 the alignment-requirement is the largest alignment-requirement of its members.
3376 Every object is allocated an offset so that:
3378 offset % alignment-requirement == 0
3380 @item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation
3381 unit if the integral types are the same size and if the next bit field fits
3382 into the current allocation unit without crossing the boundary imposed by the
3383 common alignment requirements of the bit fields.
3386 Handling of zero-length bitfields:
3388 MSVC interprets zero-length bitfields in the following ways:
3391 @item If a zero-length bitfield is inserted between two bitfields that would
3392 normally be coalesced, the bitfields will not be coalesced.
3399 unsigned long bf_1 : 12;
3401 unsigned long bf_2 : 12;
3405 The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the
3406 zero-length bitfield were removed, @code{t1}'s size would be 4 bytes.
3408 @item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the
3409 alignment of the zero-length bitfield is greater than the member that follows it,
3410 @code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield.
3430 For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1.
3431 Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length
3432 bitfield will not affect the alignment of @code{bar} or, as a result, the size
3435 Taking this into account, it is important to note the following:
3438 @item If a zero-length bitfield follows a normal bitfield, the type of the
3439 zero-length bitfield may affect the alignment of the structure as whole. For
3440 example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a
3441 normal bitfield, and is of type short.
3443 @item Even if a zero-length bitfield is not followed by a normal bitfield, it may
3444 still affect the alignment of the structure:
3454 Here, @code{t4} will take up 4 bytes.
3457 @item Zero-length bitfields following non-bitfield members are ignored:
3468 Here, @code{t5} will take up 2 bytes.
3472 @subsection PowerPC Variable Attributes
3474 Three attributes currently are defined for PowerPC configurations:
3475 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3477 For full documentation of the struct attributes please see the
3478 documentation in the @xref{i386 Variable Attributes}, section.
3480 For documentation of @code{altivec} attribute please see the
3481 documentation in the @xref{PowerPC Type Attributes}, section.
3483 @subsection Xstormy16 Variable Attributes
3485 One attribute is currently defined for xstormy16 configurations:
3490 @cindex @code{below100} attribute
3492 If a variable has the @code{below100} attribute (@code{BELOW100} is
3493 allowed also), GCC will place the variable in the first 0x100 bytes of
3494 memory and use special opcodes to access it. Such variables will be
3495 placed in either the @code{.bss_below100} section or the
3496 @code{.data_below100} section.
3500 @node Type Attributes
3501 @section Specifying Attributes of Types
3502 @cindex attribute of types
3503 @cindex type attributes
3505 The keyword @code{__attribute__} allows you to specify special
3506 attributes of @code{struct} and @code{union} types when you define
3507 such types. This keyword is followed by an attribute specification
3508 inside double parentheses. Seven attributes are currently defined for
3509 types: @code{aligned}, @code{packed}, @code{transparent_union},
3510 @code{unused}, @code{deprecated}, @code{visibility}, and
3511 @code{may_alias}. Other attributes are defined for functions
3512 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
3513 (@pxref{Function Attributes}), variables (@pxref{Variable
3514 Attributes}), and labels (@pxref{Label Attributes}).
3516 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
3517 You may also specify any one of these attributes with @samp{__}
3518 preceding and following its keyword. This allows you to use these
3519 attributes in header files without being concerned about a possible
3520 macro of the same name. For example, you may use @code{__aligned__}
3521 instead of @code{aligned}.
3523 You may specify type attributes either in a @code{typedef} declaration
3524 or in an enum, struct or union type declaration or definition.
3526 For an enum, struct or union type, you may specify attributes either
3527 between the enum, struct or union tag and the name of the type, or
3528 just past the closing curly brace of the @emph{definition}. The
3529 former syntax is preferred.
3531 @xref{Attribute Syntax}, for details of the exact syntax for using
3535 @cindex @code{aligned} attribute
3536 @item aligned (@var{alignment})
3537 This attribute specifies a minimum alignment (in bytes) for variables
3538 of the specified type. For example, the declarations:
3541 struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3542 typedef int more_aligned_int __attribute__ ((aligned (8)));
3546 force the compiler to insure (as far as it can) that each variable whose
3547 type is @code{struct S} or @code{more_aligned_int} will be allocated and
3548 aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all
3549 variables of type @code{struct S} aligned to 8-byte boundaries allows
3550 the compiler to use the @code{ldd} and @code{std} (doubleword load and
3551 store) instructions when copying one variable of type @code{struct S} to
3552 another, thus improving run-time efficiency.
3554 Note that the alignment of any given @code{struct} or @code{union} type
3555 is required by the ISO C standard to be at least a perfect multiple of
3556 the lowest common multiple of the alignments of all of the members of
3557 the @code{struct} or @code{union} in question. This means that you @emph{can}
3558 effectively adjust the alignment of a @code{struct} or @code{union}
3559 type by attaching an @code{aligned} attribute to any one of the members
3560 of such a type, but the notation illustrated in the example above is a
3561 more obvious, intuitive, and readable way to request the compiler to
3562 adjust the alignment of an entire @code{struct} or @code{union} type.
3564 As in the preceding example, you can explicitly specify the alignment
3565 (in bytes) that you wish the compiler to use for a given @code{struct}
3566 or @code{union} type. Alternatively, you can leave out the alignment factor
3567 and just ask the compiler to align a type to the maximum
3568 useful alignment for the target machine you are compiling for. For
3569 example, you could write:
3572 struct S @{ short f[3]; @} __attribute__ ((aligned));
3575 Whenever you leave out the alignment factor in an @code{aligned}
3576 attribute specification, the compiler automatically sets the alignment
3577 for the type to the largest alignment which is ever used for any data
3578 type on the target machine you are compiling for. Doing this can often
3579 make copy operations more efficient, because the compiler can use
3580 whatever instructions copy the biggest chunks of memory when performing
3581 copies to or from the variables which have types that you have aligned
3584 In the example above, if the size of each @code{short} is 2 bytes, then
3585 the size of the entire @code{struct S} type is 6 bytes. The smallest
3586 power of two which is greater than or equal to that is 8, so the
3587 compiler sets the alignment for the entire @code{struct S} type to 8
3590 Note that although you can ask the compiler to select a time-efficient
3591 alignment for a given type and then declare only individual stand-alone
3592 objects of that type, the compiler's ability to select a time-efficient
3593 alignment is primarily useful only when you plan to create arrays of
3594 variables having the relevant (efficiently aligned) type. If you
3595 declare or use arrays of variables of an efficiently-aligned type, then
3596 it is likely that your program will also be doing pointer arithmetic (or
3597 subscripting, which amounts to the same thing) on pointers to the
3598 relevant type, and the code that the compiler generates for these
3599 pointer arithmetic operations will often be more efficient for
3600 efficiently-aligned types than for other types.
3602 The @code{aligned} attribute can only increase the alignment; but you
3603 can decrease it by specifying @code{packed} as well. See below.
3605 Note that the effectiveness of @code{aligned} attributes may be limited
3606 by inherent limitations in your linker. On many systems, the linker is
3607 only able to arrange for variables to be aligned up to a certain maximum
3608 alignment. (For some linkers, the maximum supported alignment may
3609 be very very small.) If your linker is only able to align variables
3610 up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3611 in an @code{__attribute__} will still only provide you with 8 byte
3612 alignment. See your linker documentation for further information.
3615 This attribute, attached to @code{struct} or @code{union} type
3616 definition, specifies that each member (other than zero-width bitfields)
3617 of the structure or union is placed to minimize the memory required. When
3618 attached to an @code{enum} definition, it indicates that the smallest
3619 integral type should be used.
3621 @opindex fshort-enums
3622 Specifying this attribute for @code{struct} and @code{union} types is
3623 equivalent to specifying the @code{packed} attribute on each of the
3624 structure or union members. Specifying the @option{-fshort-enums}
3625 flag on the line is equivalent to specifying the @code{packed}
3626 attribute on all @code{enum} definitions.
3628 In the following example @code{struct my_packed_struct}'s members are
3629 packed closely together, but the internal layout of its @code{s} member
3630 is not packed---to do that, @code{struct my_unpacked_struct} would need to
3634 struct my_unpacked_struct
3640 struct __attribute__ ((__packed__)) my_packed_struct
3644 struct my_unpacked_struct s;
3648 You may only specify this attribute on the definition of a @code{enum},
3649 @code{struct} or @code{union}, not on a @code{typedef} which does not
3650 also define the enumerated type, structure or union.
3652 @item transparent_union
3653 This attribute, attached to a @code{union} type definition, indicates
3654 that any function parameter having that union type causes calls to that
3655 function to be treated in a special way.
3657 First, the argument corresponding to a transparent union type can be of
3658 any type in the union; no cast is required. Also, if the union contains
3659 a pointer type, the corresponding argument can be a null pointer
3660 constant or a void pointer expression; and if the union contains a void
3661 pointer type, the corresponding argument can be any pointer expression.
3662 If the union member type is a pointer, qualifiers like @code{const} on
3663 the referenced type must be respected, just as with normal pointer
3666 Second, the argument is passed to the function using the calling
3667 conventions of the first member of the transparent union, not the calling
3668 conventions of the union itself. All members of the union must have the
3669 same machine representation; this is necessary for this argument passing
3672 Transparent unions are designed for library functions that have multiple
3673 interfaces for compatibility reasons. For example, suppose the
3674 @code{wait} function must accept either a value of type @code{int *} to
3675 comply with Posix, or a value of type @code{union wait *} to comply with
3676 the 4.1BSD interface. If @code{wait}'s parameter were @code{void *},
3677 @code{wait} would accept both kinds of arguments, but it would also
3678 accept any other pointer type and this would make argument type checking
3679 less useful. Instead, @code{<sys/wait.h>} might define the interface
3687 @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3689 pid_t wait (wait_status_ptr_t);
3692 This interface allows either @code{int *} or @code{union wait *}
3693 arguments to be passed, using the @code{int *} calling convention.
3694 The program can call @code{wait} with arguments of either type:
3697 int w1 () @{ int w; return wait (&w); @}
3698 int w2 () @{ union wait w; return wait (&w); @}
3701 With this interface, @code{wait}'s implementation might look like this:
3704 pid_t wait (wait_status_ptr_t p)
3706 return waitpid (-1, p.__ip, 0);
3711 When attached to a type (including a @code{union} or a @code{struct}),
3712 this attribute means that variables of that type are meant to appear
3713 possibly unused. GCC will not produce a warning for any variables of
3714 that type, even if the variable appears to do nothing. This is often
3715 the case with lock or thread classes, which are usually defined and then
3716 not referenced, but contain constructors and destructors that have
3717 nontrivial bookkeeping functions.
3720 The @code{deprecated} attribute results in a warning if the type
3721 is used anywhere in the source file. This is useful when identifying
3722 types that are expected to be removed in a future version of a program.
3723 If possible, the warning also includes the location of the declaration
3724 of the deprecated type, to enable users to easily find further
3725 information about why the type is deprecated, or what they should do
3726 instead. Note that the warnings only occur for uses and then only
3727 if the type is being applied to an identifier that itself is not being
3728 declared as deprecated.
3731 typedef int T1 __attribute__ ((deprecated));
3735 typedef T1 T3 __attribute__ ((deprecated));
3736 T3 z __attribute__ ((deprecated));
3739 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
3740 warning is issued for line 4 because T2 is not explicitly
3741 deprecated. Line 5 has no warning because T3 is explicitly
3742 deprecated. Similarly for line 6.
3744 The @code{deprecated} attribute can also be used for functions and
3745 variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3748 Accesses to objects with types with this attribute are not subjected to
3749 type-based alias analysis, but are instead assumed to be able to alias
3750 any other type of objects, just like the @code{char} type. See
3751 @option{-fstrict-aliasing} for more information on aliasing issues.
3756 typedef short __attribute__((__may_alias__)) short_a;
3762 short_a *b = (short_a *) &a;
3766 if (a == 0x12345678)
3773 If you replaced @code{short_a} with @code{short} in the variable
3774 declaration, the above program would abort when compiled with
3775 @option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3776 above in recent GCC versions.
3779 In C++, attribute visibility (@pxref{Function Attributes}) can also be
3780 applied to class, struct, union and enum types. Unlike other type
3781 attributes, the attribute must appear between the initial keyword and
3782 the name of the type; it cannot appear after the body of the type.
3784 Note that the type visibility is applied to vague linkage entities
3785 associated with the class (vtable, typeinfo node, etc.). In
3786 particular, if a class is thrown as an exception in one shared object
3787 and caught in another, the class must have default visibility.
3788 Otherwise the two shared objects will be unable to use the same
3789 typeinfo node and exception handling will break.
3791 @subsection ARM Type Attributes
3793 On those ARM targets that support @code{dllimport} (such as Symbian
3794 OS), you can use the @code{notshared} attribute to indicate that the
3795 virtual table and other similar data for a class should not be
3796 exported from a DLL@. For example:
3799 class __declspec(notshared) C @{
3801 __declspec(dllimport) C();
3805 __declspec(dllexport)
3809 In this code, @code{C::C} is exported from the current DLL, but the
3810 virtual table for @code{C} is not exported. (You can use
3811 @code{__attribute__} instead of @code{__declspec} if you prefer, but
3812 most Symbian OS code uses @code{__declspec}.)
3814 @anchor{i386 Type Attributes}
3815 @subsection i386 Type Attributes
3817 Two attributes are currently defined for i386 configurations:
3818 @code{ms_struct} and @code{gcc_struct}
3822 @cindex @code{ms_struct}
3823 @cindex @code{gcc_struct}
3825 If @code{packed} is used on a structure, or if bit-fields are used
3826 it may be that the Microsoft ABI packs them differently
3827 than GCC would normally pack them. Particularly when moving packed
3828 data between functions compiled with GCC and the native Microsoft compiler
3829 (either via function call or as data in a file), it may be necessary to access
3832 Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3833 compilers to match the native Microsoft compiler.
3836 To specify multiple attributes, separate them by commas within the
3837 double parentheses: for example, @samp{__attribute__ ((aligned (16),
3840 @anchor{PowerPC Type Attributes}
3841 @subsection PowerPC Type Attributes
3843 Three attributes currently are defined for PowerPC configurations:
3844 @code{altivec}, @code{ms_struct} and @code{gcc_struct}.
3846 For full documentation of the struct attributes please see the
3847 documentation in the @xref{i386 Type Attributes}, section.
3849 The @code{altivec} attribute allows one to declare AltiVec vector data
3850 types supported by the AltiVec Programming Interface Manual. The
3851 attribute requires an argument to specify one of three vector types:
3852 @code{vector__}, @code{pixel__} (always followed by unsigned short),
3853 and @code{bool__} (always followed by unsigned).
3856 __attribute__((altivec(vector__)))
3857 __attribute__((altivec(pixel__))) unsigned short
3858 __attribute__((altivec(bool__))) unsigned
3861 These attributes mainly are intended to support the @code{__vector},
3862 @code{__pixel}, and @code{__bool} AltiVec keywords.
3864 @c APPLE LOCAL begin for-fsf-4_4 3274130 5295549
3865 @node Label Attributes
3866 @section Specifying Attributes of Labels and Statements
3867 @cindex attribute of labels
3868 @cindex label attributes
3869 @cindex attribute of statements
3870 @cindex statement attributes
3872 The keyword @code{__attribute__} allows you to specify special
3873 attributes of labels and statements.
3875 Some attributes are currently defined generically for variables.
3876 Other attributes are defined for variables on particular target
3877 systems. Other attributes are available for functions
3878 (@pxref{Function Attributes}), types (@pxref{Type Attributes}) and
3879 variables (@pxref{Variable Attributes}).
3881 You may also specify attributes with @samp{__} preceding and following
3882 each keyword. This allows you to use them in header files without
3883 being concerned about a possible macro of the same name. For example,
3884 you may use @code{__aligned__} instead of @code{aligned}.
3886 @xref{Attribute Syntax}, for details of the exact syntax for using
3890 @cindex @code{aligned} attribute
3891 @item aligned (@var{alignment})
3892 This attribute specifies a minimum alignment for the label,
3893 measured in bytes. For example, the declaration:
3896 some_label: __attribute__((aligned(16)))
3900 requests the compiler to align the label, inserting @code{nop}s as necessary,
3901 to a 16-byte boundary.
3903 The alignment is only a request. The compiler will usually be able to
3904 honour it but sometimes the label will be eliminated by the compiler,
3905 in which case its alignment will be eliminated too.
3907 When applied to loops, the @code{aligned} attribute causes the loop to
3911 When attached to a label this attribute means that the label might not
3912 be used. GCC will not produce a warning for the label, even if the
3913 label doesn't seem to be referenced. This feature is intended for
3914 code generated by programs which contains labels that may be unused
3915 but which is compiled with @option{-Wall}. It would not normally be
3916 appropriate to use in it human-written code, though it could be useful
3917 in cases where the code that jumps to the label is contained within an
3918 @code{#ifdef} conditional.
3920 This attribute can only be applied to labels, not statements, because
3921 there is no warning if a statement is removed.
3924 @c APPLE LOCAL end for-fsf-4_4 3274130 5295549
3926 @section An Inline Function is As Fast As a Macro
3927 @cindex inline functions
3928 @cindex integrating function code
3930 @cindex macros, inline alternative
3932 By declaring a function inline, you can direct GCC to make
3933 calls to that function faster. One way GCC can achieve this is to
3934 integrate that function's code into the code for its callers. This
3935 makes execution faster by eliminating the function-call overhead; in
3936 addition, if any of the actual argument values are constant, their
3937 known values may permit simplifications at compile time so that not
3938 all of the inline function's code needs to be included. The effect on
3939 code size is less predictable; object code may be larger or smaller
3940 with function inlining, depending on the particular case. You can
3941 also direct GCC to try to integrate all ``simple enough'' functions
3942 into their callers with the option @option{-finline-functions}.
3944 GCC implements three different semantics of declaring a function
3945 inline. One is available with @option{-std=gnu89}, another when
3946 @option{-std=c99} or @option{-std=gnu99}, and the third is used when
3949 To declare a function inline, use the @code{inline} keyword in its
3950 declaration, like this:
3960 If you are writing a header file to be included in ISO C89 programs, write
3961 @code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}.
3963 The three types of inlining behave similarly in two important cases:
3964 when the @code{inline} keyword is used on a @code{static} function,
3965 like the example above, and when a function is first declared without
3966 using the @code{inline} keyword and then is defined with
3967 @code{inline}, like this:
3970 extern int inc (int *a);
3978 In both of these common cases, the program behaves the same as if you
3979 had not used the @code{inline} keyword, except for its speed.
3981 @cindex inline functions, omission of
3982 @opindex fkeep-inline-functions
3983 When a function is both inline and @code{static}, if all calls to the
3984 function are integrated into the caller, and the function's address is
3985 never used, then the function's own assembler code is never referenced.
3986 In this case, GCC does not actually output assembler code for the
3987 function, unless you specify the option @option{-fkeep-inline-functions}.
3988 Some calls cannot be integrated for various reasons (in particular,
3989 calls that precede the function's definition cannot be integrated, and
3990 neither can recursive calls within the definition). If there is a
3991 nonintegrated call, then the function is compiled to assembler code as
3992 usual. The function must also be compiled as usual if the program
3993 refers to its address, because that can't be inlined.
3995 @cindex automatic @code{inline} for C++ member fns
3996 @cindex @code{inline} automatic for C++ member fns
3997 @cindex member fns, automatically @code{inline}
3998 @cindex C++ member fns, automatically @code{inline}
3999 @opindex fno-default-inline
4000 As required by ISO C++, GCC considers member functions defined within
4001 the body of a class to be marked inline even if they are
4002 not explicitly declared with the @code{inline} keyword. You can
4003 override this with @option{-fno-default-inline}; @pxref{C++ Dialect
4004 Options,,Options Controlling C++ Dialect}.
4006 GCC does not inline any functions when not optimizing unless you specify
4007 the @samp{always_inline} attribute for the function, like this:
4010 /* @r{Prototype.} */
4011 inline void foo (const char) __attribute__((always_inline));
4014 The remainder of this section is specific to GNU C89 inlining.
4016 @cindex non-static inline function
4017 When an inline function is not @code{static}, then the compiler must assume
4018 that there may be calls from other source files; since a global symbol can
4019 be defined only once in any program, the function must not be defined in
4020 the other source files, so the calls therein cannot be integrated.
4021 Therefore, a non-@code{static} inline function is always compiled on its
4022 own in the usual fashion.
4024 If you specify both @code{inline} and @code{extern} in the function
4025 definition, then the definition is used only for inlining. In no case
4026 is the function compiled on its own, not even if you refer to its
4027 address explicitly. Such an address becomes an external reference, as
4028 if you had only declared the function, and had not defined it.
4030 This combination of @code{inline} and @code{extern} has almost the
4031 effect of a macro. The way to use it is to put a function definition in
4032 a header file with these keywords, and put another copy of the
4033 definition (lacking @code{inline} and @code{extern}) in a library file.
4034 The definition in the header file will cause most calls to the function
4035 to be inlined. If any uses of the function remain, they will refer to
4036 the single copy in the library.
4039 @section Assembler Instructions with C Expression Operands
4040 @cindex extended @code{asm}
4041 @cindex @code{asm} expressions
4042 @cindex assembler instructions
4045 In an assembler instruction using @code{asm}, you can specify the
4046 operands of the instruction using C expressions. This means you need not
4047 guess which registers or memory locations will contain the data you want
4050 You must specify an assembler instruction template much like what
4051 appears in a machine description, plus an operand constraint string for
4054 For example, here is how to use the 68881's @code{fsinx} instruction:
4057 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
4061 Here @code{angle} is the C expression for the input operand while
4062 @code{result} is that of the output operand. Each has @samp{"f"} as its
4063 operand constraint, saying that a floating point register is required.
4064 The @samp{=} in @samp{=f} indicates that the operand is an output; all
4065 output operands' constraints must use @samp{=}. The constraints use the
4066 same language used in the machine description (@pxref{Constraints}).
4068 Each operand is described by an operand-constraint string followed by
4069 the C expression in parentheses. A colon separates the assembler
4070 template from the first output operand and another separates the last
4071 output operand from the first input, if any. Commas separate the
4072 operands within each group. The total number of operands is currently
4073 limited to 30; this limitation may be lifted in some future version of
4076 If there are no output operands but there are input operands, you must
4077 place two consecutive colons surrounding the place where the output
4080 As of GCC version 3.1, it is also possible to specify input and output
4081 operands using symbolic names which can be referenced within the
4082 assembler code. These names are specified inside square brackets
4083 preceding the constraint string, and can be referenced inside the
4084 assembler code using @code{%[@var{name}]} instead of a percentage sign
4085 followed by the operand number. Using named operands the above example
4089 asm ("fsinx %[angle],%[output]"
4090 : [output] "=f" (result)
4091 : [angle] "f" (angle));
4095 Note that the symbolic operand names have no relation whatsoever to
4096 other C identifiers. You may use any name you like, even those of
4097 existing C symbols, but you must ensure that no two operands within the same
4098 assembler construct use the same symbolic name.
4100 Output operand expressions must be lvalues; the compiler can check this.
4101 The input operands need not be lvalues. The compiler cannot check
4102 whether the operands have data types that are reasonable for the
4103 instruction being executed. It does not parse the assembler instruction
4104 template and does not know what it means or even whether it is valid
4105 assembler input. The extended @code{asm} feature is most often used for
4106 machine instructions the compiler itself does not know exist. If
4107 the output expression cannot be directly addressed (for example, it is a
4108 bit-field), your constraint must allow a register. In that case, GCC
4109 will use the register as the output of the @code{asm}, and then store
4110 that register into the output.
4112 The ordinary output operands must be write-only; GCC will assume that
4113 the values in these operands before the instruction are dead and need
4114 not be generated. Extended asm supports input-output or read-write
4115 operands. Use the constraint character @samp{+} to indicate such an
4116 operand and list it with the output operands. You should only use
4117 read-write operands when the constraints for the operand (or the
4118 operand in which only some of the bits are to be changed) allow a
4121 You may, as an alternative, logically split its function into two
4122 separate operands, one input operand and one write-only output
4123 operand. The connection between them is expressed by constraints
4124 which say they need to be in the same location when the instruction
4125 executes. You can use the same C expression for both operands, or
4126 different expressions. For example, here we write the (fictitious)
4127 @samp{combine} instruction with @code{bar} as its read-only source
4128 operand and @code{foo} as its read-write destination:
4131 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
4135 The constraint @samp{"0"} for operand 1 says that it must occupy the
4136 same location as operand 0. A number in constraint is allowed only in
4137 an input operand and it must refer to an output operand.
4139 Only a number in the constraint can guarantee that one operand will be in
4140 the same place as another. The mere fact that @code{foo} is the value
4141 of both operands is not enough to guarantee that they will be in the
4142 same place in the generated assembler code. The following would not
4146 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
4149 Various optimizations or reloading could cause operands 0 and 1 to be in
4150 different registers; GCC knows no reason not to do so. For example, the
4151 compiler might find a copy of the value of @code{foo} in one register and
4152 use it for operand 1, but generate the output operand 0 in a different
4153 register (copying it afterward to @code{foo}'s own address). Of course,
4154 since the register for operand 1 is not even mentioned in the assembler
4155 code, the result will not work, but GCC can't tell that.
4157 As of GCC version 3.1, one may write @code{[@var{name}]} instead of
4158 the operand number for a matching constraint. For example:
4161 asm ("cmoveq %1,%2,%[result]"
4162 : [result] "=r"(result)
4163 : "r" (test), "r"(new), "[result]"(old));
4166 Sometimes you need to make an @code{asm} operand be a specific register,
4167 but there's no matching constraint letter for that register @emph{by
4168 itself}. To force the operand into that register, use a local variable
4169 for the operand and specify the register in the variable declaration.
4170 @xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any
4171 register constraint letter that matches the register:
4174 register int *p1 asm ("r0") = @dots{};
4175 register int *p2 asm ("r1") = @dots{};
4176 register int *result asm ("r0");
4177 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4180 @anchor{Example of asm with clobbered asm reg}
4181 In the above example, beware that a register that is call-clobbered by
4182 the target ABI will be overwritten by any function call in the
4183 assignment, including library calls for arithmetic operators.
4184 Assuming it is a call-clobbered register, this may happen to @code{r0}
4185 above by the assignment to @code{p2}. If you have to use such a
4186 register, use temporary variables for expressions between the register
4191 register int *p1 asm ("r0") = @dots{};
4192 register int *p2 asm ("r1") = t1;
4193 register int *result asm ("r0");
4194 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
4197 Some instructions clobber specific hard registers. To describe this,
4198 write a third colon after the input operands, followed by the names of
4199 the clobbered hard registers (given as strings). Here is a realistic
4200 example for the VAX:
4203 asm volatile ("movc3 %0,%1,%2"
4204 : /* @r{no outputs} */
4205 : "g" (from), "g" (to), "g" (count)
4206 : "r0", "r1", "r2", "r3", "r4", "r5");
4209 You may not write a clobber description in a way that overlaps with an
4210 input or output operand. For example, you may not have an operand
4211 describing a register class with one member if you mention that register
4212 in the clobber list. Variables declared to live in specific registers
4213 (@pxref{Explicit Reg Vars}), and used as asm input or output operands must
4214 have no part mentioned in the clobber description.
4215 There is no way for you to specify that an input
4216 operand is modified without also specifying it as an output
4217 operand. Note that if all the output operands you specify are for this
4218 purpose (and hence unused), you will then also need to specify
4219 @code{volatile} for the @code{asm} construct, as described below, to
4220 prevent GCC from deleting the @code{asm} statement as unused.
4222 If you refer to a particular hardware register from the assembler code,
4223 you will probably have to list the register after the third colon to
4224 tell the compiler the register's value is modified. In some assemblers,
4225 the register names begin with @samp{%}; to produce one @samp{%} in the
4226 assembler code, you must write @samp{%%} in the input.
4228 If your assembler instruction can alter the condition code register, add
4229 @samp{cc} to the list of clobbered registers. GCC on some machines
4230 represents the condition codes as a specific hardware register;
4231 @samp{cc} serves to name this register. On other machines, the
4232 condition code is handled differently, and specifying @samp{cc} has no
4233 effect. But it is valid no matter what the machine.
4235 If your assembler instructions access memory in an unpredictable
4236 fashion, add @samp{memory} to the list of clobbered registers. This
4237 will cause GCC to not keep memory values cached in registers across the
4238 assembler instruction and not optimize stores or loads to that memory.
4239 You will also want to add the @code{volatile} keyword if the memory
4240 affected is not listed in the inputs or outputs of the @code{asm}, as
4241 the @samp{memory} clobber does not count as a side-effect of the
4242 @code{asm}. If you know how large the accessed memory is, you can add
4243 it as input or output but if this is not known, you should add
4244 @samp{memory}. As an example, if you access ten bytes of a string, you
4245 can use a memory input like:
4248 @{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4251 Note that in the following example the memory input is necessary,
4252 otherwise GCC might optimize the store to @code{x} away:
4259 asm ("magic stuff accessing an 'int' pointed to by '%1'"
4260 "=&d" (r) : "a" (y), "m" (*y));
4265 You can put multiple assembler instructions together in a single
4266 @code{asm} template, separated by the characters normally used in assembly
4267 code for the system. A combination that works in most places is a newline
4268 to break the line, plus a tab character to move to the instruction field
4269 (written as @samp{\n\t}). Sometimes semicolons can be used, if the
4270 assembler allows semicolons as a line-breaking character. Note that some
4271 assembler dialects use semicolons to start a comment.
4272 The input operands are guaranteed not to use any of the clobbered
4273 registers, and neither will the output operands' addresses, so you can
4274 read and write the clobbered registers as many times as you like. Here
4275 is an example of multiple instructions in a template; it assumes the
4276 subroutine @code{_foo} accepts arguments in registers 9 and 10:
4279 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4281 : "g" (from), "g" (to)
4285 Unless an output operand has the @samp{&} constraint modifier, GCC
4286 may allocate it in the same register as an unrelated input operand, on
4287 the assumption the inputs are consumed before the outputs are produced.
4288 This assumption may be false if the assembler code actually consists of
4289 more than one instruction. In such a case, use @samp{&} for each output
4290 operand that may not overlap an input. @xref{Modifiers}.
4292 If you want to test the condition code produced by an assembler
4293 instruction, you must include a branch and a label in the @code{asm}
4294 construct, as follows:
4297 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4303 This assumes your assembler supports local labels, as the GNU assembler
4304 and most Unix assemblers do.
4306 Speaking of labels, jumps from one @code{asm} to another are not
4307 supported. The compiler's optimizers do not know about these jumps, and
4308 therefore they cannot take account of them when deciding how to
4311 @cindex macros containing @code{asm}
4312 Usually the most convenient way to use these @code{asm} instructions is to
4313 encapsulate them in macros that look like functions. For example,
4317 (@{ double __value, __arg = (x); \
4318 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
4323 Here the variable @code{__arg} is used to make sure that the instruction
4324 operates on a proper @code{double} value, and to accept only those
4325 arguments @code{x} which can convert automatically to a @code{double}.
4327 Another way to make sure the instruction operates on the correct data
4328 type is to use a cast in the @code{asm}. This is different from using a
4329 variable @code{__arg} in that it converts more different types. For
4330 example, if the desired type were @code{int}, casting the argument to
4331 @code{int} would accept a pointer with no complaint, while assigning the
4332 argument to an @code{int} variable named @code{__arg} would warn about
4333 using a pointer unless the caller explicitly casts it.
4335 If an @code{asm} has output operands, GCC assumes for optimization
4336 purposes the instruction has no side effects except to change the output
4337 operands. This does not mean instructions with a side effect cannot be
4338 used, but you must be careful, because the compiler may eliminate them
4339 if the output operands aren't used, or move them out of loops, or
4340 replace two with one if they constitute a common subexpression. Also,
4341 if your instruction does have a side effect on a variable that otherwise
4342 appears not to change, the old value of the variable may be reused later
4343 if it happens to be found in a register.
4345 You can prevent an @code{asm} instruction from being deleted
4346 by writing the keyword @code{volatile} after
4347 the @code{asm}. For example:
4350 #define get_and_set_priority(new) \
4352 asm volatile ("get_and_set_priority %0, %1" \
4353 : "=g" (__old) : "g" (new)); \
4358 The @code{volatile} keyword indicates that the instruction has
4359 important side-effects. GCC will not delete a volatile @code{asm} if
4360 it is reachable. (The instruction can still be deleted if GCC can
4361 prove that control-flow will never reach the location of the
4362 instruction.) Note that even a volatile @code{asm} instruction
4363 can be moved relative to other code, including across jump
4364 instructions. For example, on many targets there is a system
4365 register which can be set to control the rounding mode of
4366 floating point operations. You might try
4367 setting it with a volatile @code{asm}, like this PowerPC example:
4370 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
4375 This will not work reliably, as the compiler may move the addition back
4376 before the volatile @code{asm}. To make it work you need to add an
4377 artificial dependency to the @code{asm} referencing a variable in the code
4378 you don't want moved, for example:
4381 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
4385 Similarly, you can't expect a
4386 sequence of volatile @code{asm} instructions to remain perfectly
4387 consecutive. If you want consecutive output, use a single @code{asm}.
4388 Also, GCC will perform some optimizations across a volatile @code{asm}
4389 instruction; GCC does not ``forget everything'' when it encounters
4390 a volatile @code{asm} instruction the way some other compilers do.
4392 An @code{asm} instruction without any output operands will be treated
4393 identically to a volatile @code{asm} instruction.
4395 It is a natural idea to look for a way to give access to the condition
4396 code left by the assembler instruction. However, when we attempted to
4397 implement this, we found no way to make it work reliably. The problem
4398 is that output operands might need reloading, which would result in
4399 additional following ``store'' instructions. On most machines, these
4400 instructions would alter the condition code before there was time to
4401 test it. This problem doesn't arise for ordinary ``test'' and
4402 ``compare'' instructions because they don't have any output operands.
4404 For reasons similar to those described above, it is not possible to give
4405 an assembler instruction access to the condition code left by previous
4408 If you are writing a header file that should be includable in ISO C
4409 programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate
4412 @subsection Size of an @code{asm}
4414 Some targets require that GCC track the size of each instruction used in
4415 order to generate correct code. Because the final length of an
4416 @code{asm} is only known by the assembler, GCC must make an estimate as
4417 to how big it will be. The estimate is formed by counting the number of
4418 statements in the pattern of the @code{asm} and multiplying that by the
4419 length of the longest instruction on that processor. Statements in the
4420 @code{asm} are identified by newline characters and whatever statement
4421 separator characters are supported by the assembler; on most processors
4422 this is the `@code{;}' character.
4424 Normally, GCC's estimate is perfectly adequate to ensure that correct
4425 code is generated, but it is possible to confuse the compiler if you use
4426 pseudo instructions or assembler macros that expand into multiple real
4427 instructions or if you use assembler directives that expand to more
4428 space in the object file than would be needed for a single instruction.
4429 If this happens then the assembler will produce a diagnostic saying that
4430 a label is unreachable.
4432 @subsection i386 floating point asm operands
4434 There are several rules on the usage of stack-like regs in
4435 asm_operands insns. These rules apply only to the operands that are
4440 Given a set of input regs that die in an asm_operands, it is
4441 necessary to know which are implicitly popped by the asm, and
4442 which must be explicitly popped by gcc.
4444 An input reg that is implicitly popped by the asm must be
4445 explicitly clobbered, unless it is constrained to match an
4449 For any input reg that is implicitly popped by an asm, it is
4450 necessary to know how to adjust the stack to compensate for the pop.
4451 If any non-popped input is closer to the top of the reg-stack than
4452 the implicitly popped reg, it would not be possible to know what the
4453 stack looked like---it's not clear how the rest of the stack ``slides
4456 All implicitly popped input regs must be closer to the top of
4457 the reg-stack than any input that is not implicitly popped.
4459 It is possible that if an input dies in an insn, reload might
4460 use the input reg for an output reload. Consider this example:
4463 asm ("foo" : "=t" (a) : "f" (b));
4466 This asm says that input B is not popped by the asm, and that
4467 the asm pushes a result onto the reg-stack, i.e., the stack is one
4468 deeper after the asm than it was before. But, it is possible that
4469 reload will think that it can use the same reg for both the input and
4470 the output, if input B dies in this insn.
4472 If any input operand uses the @code{f} constraint, all output reg
4473 constraints must use the @code{&} earlyclobber.
4475 The asm above would be written as
4478 asm ("foo" : "=&t" (a) : "f" (b));
4482 Some operands need to be in particular places on the stack. All
4483 output operands fall in this category---there is no other way to
4484 know which regs the outputs appear in unless the user indicates
4485 this in the constraints.
4487 Output operands must specifically indicate which reg an output
4488 appears in after an asm. @code{=f} is not allowed: the operand
4489 constraints must select a class with a single reg.
4492 Output operands may not be ``inserted'' between existing stack regs.
4493 Since no 387 opcode uses a read/write operand, all output operands
4494 are dead before the asm_operands, and are pushed by the asm_operands.
4495 It makes no sense to push anywhere but the top of the reg-stack.
4497 Output operands must start at the top of the reg-stack: output
4498 operands may not ``skip'' a reg.
4501 Some asm statements may need extra stack space for internal
4502 calculations. This can be guaranteed by clobbering stack registers
4503 unrelated to the inputs and outputs.
4507 Here are a couple of reasonable asms to want to write. This asm
4508 takes one input, which is internally popped, and produces two outputs.
4511 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4514 This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4515 and replaces them with one output. The user must code the @code{st(1)}
4516 clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4519 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4525 @section Controlling Names Used in Assembler Code
4526 @cindex assembler names for identifiers
4527 @cindex names used in assembler code
4528 @cindex identifiers, names in assembler code
4530 You can specify the name to be used in the assembler code for a C
4531 function or variable by writing the @code{asm} (or @code{__asm__})
4532 keyword after the declarator as follows:
4535 int foo asm ("myfoo") = 2;
4539 This specifies that the name to be used for the variable @code{foo} in
4540 the assembler code should be @samp{myfoo} rather than the usual
4543 On systems where an underscore is normally prepended to the name of a C
4544 function or variable, this feature allows you to define names for the
4545 linker that do not start with an underscore.
4547 It does not make sense to use this feature with a non-static local
4548 variable since such variables do not have assembler names. If you are
4549 trying to put the variable in a particular register, see @ref{Explicit
4550 Reg Vars}. GCC presently accepts such code with a warning, but will
4551 probably be changed to issue an error, rather than a warning, in the
4554 You cannot use @code{asm} in this way in a function @emph{definition}; but
4555 you can get the same effect by writing a declaration for the function
4556 before its definition and putting @code{asm} there, like this:
4559 extern func () asm ("FUNC");
4566 It is up to you to make sure that the assembler names you choose do not
4567 conflict with any other assembler symbols. Also, you must not use a
4568 register name; that would produce completely invalid assembler code. GCC
4569 does not as yet have the ability to store static variables in registers.
4570 Perhaps that will be added.
4572 @node Explicit Reg Vars
4573 @section Variables in Specified Registers
4574 @cindex explicit register variables
4575 @cindex variables in specified registers
4576 @cindex specified registers
4577 @cindex registers, global allocation
4579 GNU C allows you to put a few global variables into specified hardware
4580 registers. You can also specify the register in which an ordinary
4581 register variable should be allocated.
4585 Global register variables reserve registers throughout the program.
4586 This may be useful in programs such as programming language
4587 interpreters which have a couple of global variables that are accessed
4591 Local register variables in specific registers do not reserve the
4592 registers, except at the point where they are used as input or output
4593 operands in an @code{asm} statement and the @code{asm} statement itself is
4594 not deleted. The compiler's data flow analysis is capable of determining
4595 where the specified registers contain live values, and where they are
4596 available for other uses. Stores into local register variables may be deleted
4597 when they appear to be dead according to dataflow analysis. References
4598 to local register variables may be deleted or moved or simplified.
4600 These local variables are sometimes convenient for use with the extended
4601 @code{asm} feature (@pxref{Extended Asm}), if you want to write one
4602 output of the assembler instruction directly into a particular register.
4603 (This will work provided the register you specify fits the constraints
4604 specified for that operand in the @code{asm}.)
4612 @node Global Reg Vars
4613 @subsection Defining Global Register Variables
4614 @cindex global register variables
4615 @cindex registers, global variables in
4617 You can define a global register variable in GNU C like this:
4620 register int *foo asm ("a5");
4624 Here @code{a5} is the name of the register which should be used. Choose a
4625 register which is normally saved and restored by function calls on your
4626 machine, so that library routines will not clobber it.
4628 Naturally the register name is cpu-dependent, so you would need to
4629 conditionalize your program according to cpu type. The register
4630 @code{a5} would be a good choice on a 68000 for a variable of pointer
4631 type. On machines with register windows, be sure to choose a ``global''
4632 register that is not affected magically by the function call mechanism.
4634 In addition, operating systems on one type of cpu may differ in how they
4635 name the registers; then you would need additional conditionals. For
4636 example, some 68000 operating systems call this register @code{%a5}.
4638 Eventually there may be a way of asking the compiler to choose a register
4639 automatically, but first we need to figure out how it should choose and
4640 how to enable you to guide the choice. No solution is evident.
4642 Defining a global register variable in a certain register reserves that
4643 register entirely for this use, at least within the current compilation.
4644 The register will not be allocated for any other purpose in the functions
4645 in the current compilation. The register will not be saved and restored by
4646 these functions. Stores into this register are never deleted even if they
4647 would appear to be dead, but references may be deleted or moved or
4650 It is not safe to access the global register variables from signal
4651 handlers, or from more than one thread of control, because the system
4652 library routines may temporarily use the register for other things (unless
4653 you recompile them specially for the task at hand).
4655 @cindex @code{qsort}, and global register variables
4656 It is not safe for one function that uses a global register variable to
4657 call another such function @code{foo} by way of a third function
4658 @code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4659 different source file in which the variable wasn't declared). This is
4660 because @code{lose} might save the register and put some other value there.
4661 For example, you can't expect a global register variable to be available in
4662 the comparison-function that you pass to @code{qsort}, since @code{qsort}
4663 might have put something else in that register. (If you are prepared to
4664 recompile @code{qsort} with the same global register variable, you can
4665 solve this problem.)
4667 If you want to recompile @code{qsort} or other source files which do not
4668 actually use your global register variable, so that they will not use that
4669 register for any other purpose, then it suffices to specify the compiler
4670 option @option{-ffixed-@var{reg}}. You need not actually add a global
4671 register declaration to their source code.
4673 A function which can alter the value of a global register variable cannot
4674 safely be called from a function compiled without this variable, because it
4675 could clobber the value the caller expects to find there on return.
4676 Therefore, the function which is the entry point into the part of the
4677 program that uses the global register variable must explicitly save and
4678 restore the value which belongs to its caller.
4680 @cindex register variable after @code{longjmp}
4681 @cindex global register after @code{longjmp}
4682 @cindex value after @code{longjmp}
4685 On most machines, @code{longjmp} will restore to each global register
4686 variable the value it had at the time of the @code{setjmp}. On some
4687 machines, however, @code{longjmp} will not change the value of global
4688 register variables. To be portable, the function that called @code{setjmp}
4689 should make other arrangements to save the values of the global register
4690 variables, and to restore them in a @code{longjmp}. This way, the same
4691 thing will happen regardless of what @code{longjmp} does.
4693 All global register variable declarations must precede all function
4694 definitions. If such a declaration could appear after function
4695 definitions, the declaration would be too late to prevent the register from
4696 being used for other purposes in the preceding functions.
4698 Global register variables may not have initial values, because an
4699 executable file has no means to supply initial contents for a register.
4701 On the SPARC, there are reports that g3 @dots{} g7 are suitable
4702 registers, but certain library functions, such as @code{getwd}, as well
4703 as the subroutines for division and remainder, modify g3 and g4. g1 and
4704 g2 are local temporaries.
4706 On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4707 Of course, it will not do to use more than a few of those.
4709 @node Local Reg Vars
4710 @subsection Specifying Registers for Local Variables
4711 @cindex local variables, specifying registers
4712 @cindex specifying registers for local variables
4713 @cindex registers for local variables
4715 You can define a local register variable with a specified register
4719 register int *foo asm ("a5");
4723 Here @code{a5} is the name of the register which should be used. Note
4724 that this is the same syntax used for defining global register
4725 variables, but for a local variable it would appear within a function.
4727 Naturally the register name is cpu-dependent, but this is not a
4728 problem, since specific registers are most often useful with explicit
4729 assembler instructions (@pxref{Extended Asm}). Both of these things
4730 generally require that you conditionalize your program according to
4733 In addition, operating systems on one type of cpu may differ in how they
4734 name the registers; then you would need additional conditionals. For
4735 example, some 68000 operating systems call this register @code{%a5}.
4737 Defining such a register variable does not reserve the register; it
4738 remains available for other uses in places where flow control determines
4739 the variable's value is not live.
4741 This option does not guarantee that GCC will generate code that has
4742 this variable in the register you specify at all times. You may not
4743 code an explicit reference to this register in the @emph{assembler
4744 instruction template} part of an @code{asm} statement and assume it will
4745 always refer to this variable. However, using the variable as an
4746 @code{asm} @emph{operand} guarantees that the specified register is used
4749 Stores into local register variables may be deleted when they appear to be dead
4750 according to dataflow analysis. References to local register variables may
4751 be deleted or moved or simplified.
4753 As for global register variables, it's recommended that you choose a
4754 register which is normally saved and restored by function calls on
4755 your machine, so that library routines will not clobber it. A common
4756 pitfall is to initialize multiple call-clobbered registers with
4757 arbitrary expressions, where a function call or library call for an
4758 arithmetic operator will overwrite a register value from a previous
4759 assignment, for example @code{r0} below:
4761 register int *p1 asm ("r0") = @dots{};
4762 register int *p2 asm ("r1") = @dots{};
4764 In those cases, a solution is to use a temporary variable for
4765 each arbitrary expression. @xref{Example of asm with clobbered asm reg}.
4767 @node Alternate Keywords
4768 @section Alternate Keywords
4769 @cindex alternate keywords
4770 @cindex keywords, alternate
4772 @option{-ansi} and the various @option{-std} options disable certain
4773 keywords. This causes trouble when you want to use GNU C extensions, or
4774 a general-purpose header file that should be usable by all programs,
4775 including ISO C programs. The keywords @code{asm}, @code{typeof} and
4776 @code{inline} are not available in programs compiled with
4777 @option{-ansi} or @option{-std} (although @code{inline} can be used in a
4778 program compiled with @option{-std=c99}). The ISO C99 keyword
4779 @code{restrict} is only available when @option{-std=gnu99} (which will
4780 eventually be the default) or @option{-std=c99} (or the equivalent
4781 @option{-std=iso9899:1999}) is used.
4783 The way to solve these problems is to put @samp{__} at the beginning and
4784 end of each problematical keyword. For example, use @code{__asm__}
4785 instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4787 Other C compilers won't accept these alternative keywords; if you want to
4788 compile with another compiler, you can define the alternate keywords as
4789 macros to replace them with the customary keywords. It looks like this:
4797 @findex __extension__
4799 @option{-pedantic} and other options cause warnings for many GNU C extensions.
4801 prevent such warnings within one expression by writing
4802 @code{__extension__} before the expression. @code{__extension__} has no
4803 effect aside from this.
4805 @node Incomplete Enums
4806 @section Incomplete @code{enum} Types
4808 You can define an @code{enum} tag without specifying its possible values.
4809 This results in an incomplete type, much like what you get if you write
4810 @code{struct foo} without describing the elements. A later declaration
4811 which does specify the possible values completes the type.
4813 You can't allocate variables or storage using the type while it is
4814 incomplete. However, you can work with pointers to that type.
4816 This extension may not be very useful, but it makes the handling of
4817 @code{enum} more consistent with the way @code{struct} and @code{union}
4820 This extension is not supported by GNU C++.
4822 @node Function Names
4823 @section Function Names as Strings
4824 @cindex @code{__func__} identifier
4825 @cindex @code{__FUNCTION__} identifier
4826 @cindex @code{__PRETTY_FUNCTION__} identifier
4828 GCC provides three magic variables which hold the name of the current
4829 function, as a string. The first of these is @code{__func__}, which
4830 is part of the C99 standard:
4833 The identifier @code{__func__} is implicitly declared by the translator
4834 as if, immediately following the opening brace of each function
4835 definition, the declaration
4838 static const char __func__[] = "function-name";
4841 appeared, where function-name is the name of the lexically-enclosing
4842 function. This name is the unadorned name of the function.
4845 @code{__FUNCTION__} is another name for @code{__func__}. Older
4846 versions of GCC recognize only this name. However, it is not
4847 standardized. For maximum portability, we recommend you use
4848 @code{__func__}, but provide a fallback definition with the
4852 #if __STDC_VERSION__ < 199901L
4854 # define __func__ __FUNCTION__
4856 # define __func__ "<unknown>"
4861 In C, @code{__PRETTY_FUNCTION__} is yet another name for
4862 @code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains
4863 the type signature of the function as well as its bare name. For
4864 example, this program:
4868 extern int printf (char *, ...);
4875 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4876 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4894 __PRETTY_FUNCTION__ = void a::sub(int)
4897 These identifiers are not preprocessor macros. In GCC 3.3 and
4898 earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4899 were treated as string literals; they could be used to initialize
4900 @code{char} arrays, and they could be concatenated with other string
4901 literals. GCC 3.4 and later treat them as variables, like
4902 @code{__func__}. In C++, @code{__FUNCTION__} and
4903 @code{__PRETTY_FUNCTION__} have always been variables.
4905 @node Return Address
4906 @section Getting the Return or Frame Address of a Function
4908 These functions may be used to get information about the callers of a
4911 @deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4912 This function returns the return address of the current function, or of
4913 one of its callers. The @var{level} argument is number of frames to
4914 scan up the call stack. A value of @code{0} yields the return address
4915 of the current function, a value of @code{1} yields the return address
4916 of the caller of the current function, and so forth. When inlining
4917 the expected behavior is that the function will return the address of
4918 the function that will be returned to. To work around this behavior use
4919 the @code{noinline} function attribute.
4921 The @var{level} argument must be a constant integer.
4923 On some machines it may be impossible to determine the return address of
4924 any function other than the current one; in such cases, or when the top
4925 of the stack has been reached, this function will return @code{0} or a
4926 random value. In addition, @code{__builtin_frame_address} may be used
4927 to determine if the top of the stack has been reached.
4929 This function should only be used with a nonzero argument for debugging
4933 @deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4934 This function is similar to @code{__builtin_return_address}, but it
4935 returns the address of the function frame rather than the return address
4936 of the function. Calling @code{__builtin_frame_address} with a value of
4937 @code{0} yields the frame address of the current function, a value of
4938 @code{1} yields the frame address of the caller of the current function,
4941 The frame is the area on the stack which holds local variables and saved
4942 registers. The frame address is normally the address of the first word
4943 pushed on to the stack by the function. However, the exact definition
4944 depends upon the processor and the calling convention. If the processor
4945 has a dedicated frame pointer register, and the function has a frame,
4946 then @code{__builtin_frame_address} will return the value of the frame
4949 On some machines it may be impossible to determine the frame address of
4950 any function other than the current one; in such cases, or when the top
4951 of the stack has been reached, this function will return @code{0} if
4952 the first frame pointer is properly initialized by the startup code.
4954 This function should only be used with a nonzero argument for debugging
4958 @node Vector Extensions
4959 @section Using vector instructions through built-in functions
4961 On some targets, the instruction set contains SIMD vector instructions that
4962 operate on multiple values contained in one large register at the same time.
4963 For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4966 The first step in using these extensions is to provide the necessary data
4967 types. This should be done using an appropriate @code{typedef}:
4970 typedef int v4si __attribute__ ((vector_size (16)));
4973 The @code{int} type specifies the base type, while the attribute specifies
4974 the vector size for the variable, measured in bytes. For example, the
4975 declaration above causes the compiler to set the mode for the @code{v4si}
4976 type to be 16 bytes wide and divided into @code{int} sized units. For
4977 a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
4978 corresponding mode of @code{foo} will be @acronym{V4SI}.
4980 The @code{vector_size} attribute is only applicable to integral and
4981 float scalars, although arrays, pointers, and function return values
4982 are allowed in conjunction with this construct.
4984 All the basic integer types can be used as base types, both as signed
4985 and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
4986 @code{long long}. In addition, @code{float} and @code{double} can be
4987 used to build floating-point vector types.
4989 Specifying a combination that is not valid for the current architecture
4990 will cause GCC to synthesize the instructions using a narrower mode.
4991 For example, if you specify a variable of type @code{V4SI} and your
4992 architecture does not allow for this specific SIMD type, GCC will
4993 produce code that uses 4 @code{SIs}.
4995 The types defined in this manner can be used with a subset of normal C
4996 operations. Currently, GCC will allow using the following operators
4997 on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4999 The operations behave like C++ @code{valarrays}. Addition is defined as
5000 the addition of the corresponding elements of the operands. For
5001 example, in the code below, each of the 4 elements in @var{a} will be
5002 added to the corresponding 4 elements in @var{b} and the resulting
5003 vector will be stored in @var{c}.
5006 typedef int v4si __attribute__ ((vector_size (16)));
5013 Subtraction, multiplication, division, and the logical operations
5014 operate in a similar manner. Likewise, the result of using the unary
5015 minus or complement operators on a vector type is a vector whose
5016 elements are the negative or complemented values of the corresponding
5017 elements in the operand.
5019 You can declare variables and use them in function calls and returns, as
5020 well as in assignments and some casts. You can specify a vector type as
5021 a return type for a function. Vector types can also be used as function
5022 arguments. It is possible to cast from one vector type to another,
5023 provided they are of the same size (in fact, you can also cast vectors
5024 to and from other datatypes of the same size).
5026 You cannot operate between vectors of different lengths or different
5027 signedness without a cast.
5029 A port that supports hardware vector operations, usually provides a set
5030 of built-in functions that can be used to operate on vectors. For
5031 example, a function to add two vectors and multiply the result by a
5032 third could look like this:
5035 v4si f (v4si a, v4si b, v4si c)
5037 v4si tmp = __builtin_addv4si (a, b);
5038 return __builtin_mulv4si (tmp, c);
5045 @findex __builtin_offsetof
5047 GCC implements for both C and C++ a syntactic extension to implement
5048 the @code{offsetof} macro.
5052 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"
5054 offsetof_member_designator:
5056 | offsetof_member_designator "." @code{identifier}
5057 | offsetof_member_designator "[" @code{expr} "]"
5060 This extension is sufficient such that
5063 #define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member})
5066 is a suitable definition of the @code{offsetof} macro. In C++, @var{type}
5067 may be dependent. In either case, @var{member} may consist of a single
5068 identifier, or a sequence of member accesses and array references.
5070 @node Atomic Builtins
5071 @section Built-in functions for atomic memory access
5073 The following builtins are intended to be compatible with those described
5074 in the @cite{Intel Itanium Processor-specific Application Binary Interface},
5075 section 7.4. As such, they depart from the normal GCC practice of using
5076 the ``__builtin_'' prefix, and further that they are overloaded such that
5077 they work on multiple types.
5079 The definition given in the Intel documentation allows only for the use of
5080 the types @code{int}, @code{long}, @code{long long} as well as their unsigned
5081 counterparts. GCC will allow any integral scalar or pointer type that is
5082 1, 2, 4 or 8 bytes in length.
5084 Not all operations are supported by all target processors. If a particular
5085 operation cannot be implemented on the target processor, a warning will be
5086 generated and a call an external function will be generated. The external
5087 function will carry the same name as the builtin, with an additional suffix
5088 @samp{_@var{n}} where @var{n} is the size of the data type.
5090 @c ??? Should we have a mechanism to suppress this warning? This is almost
5091 @c useful for implementing the operation under the control of an external
5094 In most cases, these builtins are considered a @dfn{full barrier}. That is,
5095 no memory operand will be moved across the operation, either forward or
5096 backward. Further, instructions will be issued as necessary to prevent the
5097 processor from speculating loads across the operation and from queuing stores
5098 after the operation.
5100 All of the routines are are described in the Intel documentation to take
5101 ``an optional list of variables protected by the memory barrier''. It's
5102 not clear what is meant by that; it could mean that @emph{only} the
5103 following variables are protected, or it could mean that these variables
5104 should in addition be protected. At present GCC ignores this list and
5105 protects all variables which are globally accessible. If in the future
5106 we make some use of this list, an empty list will continue to mean all
5107 globally accessible variables.
5110 @item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
5111 @itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
5112 @itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
5113 @itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
5114 @itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
5115 @itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
5116 @findex __sync_fetch_and_add
5117 @findex __sync_fetch_and_sub
5118 @findex __sync_fetch_and_or
5119 @findex __sync_fetch_and_and
5120 @findex __sync_fetch_and_xor
5121 @findex __sync_fetch_and_nand
5122 These builtins perform the operation suggested by the name, and
5123 returns the value that had previously been in memory. That is,
5126 @{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
5127 @{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand
5130 @item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
5131 @itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
5132 @itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
5133 @itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
5134 @itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
5135 @itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
5136 @findex __sync_add_and_fetch
5137 @findex __sync_sub_and_fetch
5138 @findex __sync_or_and_fetch
5139 @findex __sync_and_and_fetch
5140 @findex __sync_xor_and_fetch
5141 @findex __sync_nand_and_fetch
5142 These builtins perform the operation suggested by the name, and
5143 return the new value. That is,
5146 @{ *ptr @var{op}= value; return *ptr; @}
5147 @{ *ptr = ~*ptr & value; return *ptr; @} // nand
5150 @item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5151 @itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...)
5152 @findex __sync_bool_compare_and_swap
5153 @findex __sync_val_compare_and_swap
5154 These builtins perform an atomic compare and swap. That is, if the current
5155 value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
5158 The ``bool'' version returns true if the comparison is successful and
5159 @var{newval} was written. The ``val'' version returns the contents
5160 of @code{*@var{ptr}} before the operation.
5162 @item __sync_synchronize (...)
5163 @findex __sync_synchronize
5164 This builtin issues a full memory barrier.
5166 @item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
5167 @findex __sync_lock_test_and_set
5168 This builtin, as described by Intel, is not a traditional test-and-set
5169 operation, but rather an atomic exchange operation. It writes @var{value}
5170 into @code{*@var{ptr}}, and returns the previous contents of
5173 Many targets have only minimal support for such locks, and do not support
5174 a full exchange operation. In this case, a target may support reduced
5175 functionality here by which the @emph{only} valid value to store is the
5176 immediate constant 1. The exact value actually stored in @code{*@var{ptr}}
5177 is implementation defined.
5179 This builtin is not a full barrier, but rather an @dfn{acquire barrier}.
5180 This means that references after the builtin cannot move to (or be
5181 speculated to) before the builtin, but previous memory stores may not
5182 be globally visible yet, and previous memory loads may not yet be
5185 @item void __sync_lock_release (@var{type} *ptr, ...)
5186 @findex __sync_lock_release
5187 This builtin releases the lock acquired by @code{__sync_lock_test_and_set}.
5188 Normally this means writing the constant 0 to @code{*@var{ptr}}.
5190 This builtin is not a full barrier, but rather a @dfn{release barrier}.
5191 This means that all previous memory stores are globally visible, and all
5192 previous memory loads have been satisfied, but following memory reads
5193 are not prevented from being speculated to before the barrier.
5196 @node Object Size Checking
5197 @section Object Size Checking Builtins
5198 @findex __builtin_object_size
5199 @findex __builtin___memcpy_chk
5200 @findex __builtin___mempcpy_chk
5201 @findex __builtin___memmove_chk
5202 @findex __builtin___memset_chk
5203 @findex __builtin___strcpy_chk
5204 @findex __builtin___stpcpy_chk
5205 @findex __builtin___strncpy_chk
5206 @findex __builtin___strcat_chk
5207 @findex __builtin___strncat_chk
5208 @findex __builtin___sprintf_chk
5209 @findex __builtin___snprintf_chk
5210 @findex __builtin___vsprintf_chk
5211 @findex __builtin___vsnprintf_chk
5212 @findex __builtin___printf_chk
5213 @findex __builtin___vprintf_chk
5214 @findex __builtin___fprintf_chk
5215 @findex __builtin___vfprintf_chk
5217 GCC implements a limited buffer overflow protection mechanism
5218 that can prevent some buffer overflow attacks.
5220 @deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type})
5221 is a built-in construct that returns a constant number of bytes from
5222 @var{ptr} to the end of the object @var{ptr} pointer points to
5223 (if known at compile time). @code{__builtin_object_size} never evaluates
5224 its arguments for side-effects. If there are any side-effects in them, it
5225 returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5226 for @var{type} 2 or 3. If there are multiple objects @var{ptr} can
5227 point to and all of them are known at compile time, the returned number
5228 is the maximum of remaining byte counts in those objects if @var{type} & 2 is
5229 0 and minimum if nonzero. If it is not possible to determine which objects
5230 @var{ptr} points to at compile time, @code{__builtin_object_size} should
5231 return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
5232 for @var{type} 2 or 3.
5234 @var{type} is an integer constant from 0 to 3. If the least significant
5235 bit is clear, objects are whole variables, if it is set, a closest
5236 surrounding subobject is considered the object a pointer points to.
5237 The second bit determines if maximum or minimum of remaining bytes
5241 struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
5242 char *p = &var.buf1[1], *q = &var.b;
5244 /* Here the object p points to is var. */
5245 assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
5246 /* The subobject p points to is var.buf1. */
5247 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
5248 /* The object q points to is var. */
5249 assert (__builtin_object_size (q, 0)
5250 == (char *) (&var + 1) - (char *) &var.b);
5251 /* The subobject q points to is var.b. */
5252 assert (__builtin_object_size (q, 1) == sizeof (var.b));
5256 There are built-in functions added for many common string operation
5257 functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk}
5258 built-in is provided. This built-in has an additional last argument,
5259 which is the number of bytes remaining in object the @var{dest}
5260 argument points to or @code{(size_t) -1} if the size is not known.
5262 The built-in functions are optimized into the normal string functions
5263 like @code{memcpy} if the last argument is @code{(size_t) -1} or if
5264 it is known at compile time that the destination object will not
5265 be overflown. If the compiler can determine at compile time the
5266 object will be always overflown, it issues a warning.
5268 The intended use can be e.g.
5272 #define bos0(dest) __builtin_object_size (dest, 0)
5273 #define memcpy(dest, src, n) \
5274 __builtin___memcpy_chk (dest, src, n, bos0 (dest))
5278 /* It is unknown what object p points to, so this is optimized
5279 into plain memcpy - no checking is possible. */
5280 memcpy (p, "abcde", n);
5281 /* Destination is known and length too. It is known at compile
5282 time there will be no overflow. */
5283 memcpy (&buf[5], "abcde", 5);
5284 /* Destination is known, but the length is not known at compile time.
5285 This will result in __memcpy_chk call that can check for overflow
5287 memcpy (&buf[5], "abcde", n);
5288 /* Destination is known and it is known at compile time there will
5289 be overflow. There will be a warning and __memcpy_chk call that
5290 will abort the program at runtime. */
5291 memcpy (&buf[6], "abcde", 5);
5294 Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
5295 @code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
5296 @code{strcat} and @code{strncat}.
5298 There are also checking built-in functions for formatted output functions.
5300 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
5301 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5302 const char *fmt, ...);
5303 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
5305 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
5306 const char *fmt, va_list ap);
5309 The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
5310 etc. functions and can contain implementation specific flags on what
5311 additional security measures the checking function might take, such as
5312 handling @code{%n} differently.
5314 The @var{os} argument is the object size @var{s} points to, like in the
5315 other built-in functions. There is a small difference in the behavior
5316 though, if @var{os} is @code{(size_t) -1}, the built-in functions are
5317 optimized into the non-checking functions only if @var{flag} is 0, otherwise
5318 the checking function is called with @var{os} argument set to
5321 In addition to this, there are checking built-in functions
5322 @code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
5323 @code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
5324 These have just one additional argument, @var{flag}, right before
5325 format string @var{fmt}. If the compiler is able to optimize them to
5326 @code{fputc} etc. functions, it will, otherwise the checking function
5327 should be called and the @var{flag} argument passed to it.
5329 @node Other Builtins
5330 @section Other built-in functions provided by GCC
5331 @cindex built-in functions
5332 @findex __builtin_isgreater
5333 @findex __builtin_isgreaterequal
5334 @findex __builtin_isless
5335 @findex __builtin_islessequal
5336 @findex __builtin_islessgreater
5337 @findex __builtin_isunordered
5338 @findex __builtin_powi
5339 @findex __builtin_powif
5340 @findex __builtin_powil
5498 @findex fprintf_unlocked
5500 @findex fputs_unlocked
5610 @findex printf_unlocked
5639 @findex significandf
5640 @findex significandl
5711 GCC provides a large number of built-in functions other than the ones
5712 mentioned above. Some of these are for internal use in the processing
5713 of exceptions or variable-length argument lists and will not be
5714 documented here because they may change from time to time; we do not
5715 recommend general use of these functions.
5717 The remaining functions are provided for optimization purposes.
5719 @opindex fno-builtin
5720 GCC includes built-in versions of many of the functions in the standard
5721 C library. The versions prefixed with @code{__builtin_} will always be
5722 treated as having the same meaning as the C library function even if you
5723 specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5724 Many of these functions are only optimized in certain cases; if they are
5725 not optimized in a particular case, a call to the library function will
5730 Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5731 @option{-std=c99}), the functions
5732 @code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5733 @code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5734 @code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5735 @code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5736 @code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5737 @code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0},
5738 @code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn},
5739 @code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10},
5740 @code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl},
5741 @code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl},
5742 @code{significandf}, @code{significandl}, @code{significand},
5743 @code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5744 @code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon},
5745 @code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f},
5746 @code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf},
5747 @code{ynl} and @code{yn}
5748 may be handled as built-in functions.
5749 All these functions have corresponding versions
5750 prefixed with @code{__builtin_}, which may be used even in strict C89
5753 The ISO C99 functions
5754 @code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5755 @code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5756 @code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5757 @code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5758 @code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5759 @code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5760 @code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5761 @code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5762 @code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5763 @code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5764 @code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog},
5765 @code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl},
5766 @code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf},
5767 @code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal},
5768 @code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl},
5769 @code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf},
5770 @code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan},
5771 @code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl},
5772 @code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f},
5773 @code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim},
5774 @code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax},
5775 @code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf},
5776 @code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb},
5777 @code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf},
5778 @code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl},
5779 @code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround},
5780 @code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l},
5781 @code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf},
5782 @code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl},
5783 @code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint},
5784 @code{nextafterf}, @code{nextafterl}, @code{nextafter},
5785 @code{nexttowardf}, @code{nexttowardl}, @code{nexttoward},
5786 @code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof},
5787 @code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint},
5788 @code{roundf}, @code{roundl}, @code{round}, @code{scalblnf},
5789 @code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl},
5790 @code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal},
5791 @code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc},
5792 @code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf}
5793 are handled as built-in functions
5794 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5796 There are also built-in versions of the ISO C99 functions
5797 @code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5798 @code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5799 @code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5800 @code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5801 @code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5802 @code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5803 @code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5804 @code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5805 @code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5806 that are recognized in any mode since ISO C90 reserves these names for
5807 the purpose to which ISO C99 puts them. All these functions have
5808 corresponding versions prefixed with @code{__builtin_}.
5810 The ISO C94 functions
5811 @code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit},
5812 @code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct},
5813 @code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and
5815 are handled as built-in functions
5816 except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5818 The ISO C90 functions
5819 @code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5820 @code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5821 @code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5822 @code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf},
5823 @code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit},
5824 @code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct},
5825 @code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower},
5826 @code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log},
5827 @code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf},
5828 @code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf},
5829 @code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt},
5830 @code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp},
5831 @code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat},
5832 @code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr},
5833 @code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf},
5834 @code{vprintf} and @code{vsprintf}
5835 are all recognized as built-in functions unless
5836 @option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5837 is specified for an individual function). All of these functions have
5838 corresponding versions prefixed with @code{__builtin_}.
5840 GCC provides built-in versions of the ISO C99 floating point comparison
5841 macros that avoid raising exceptions for unordered operands. They have
5842 the same names as the standard macros ( @code{isgreater},
5843 @code{isgreaterequal}, @code{isless}, @code{islessequal},
5844 @code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5845 prefixed. We intend for a library implementor to be able to simply
5846 @code{#define} each standard macro to its built-in equivalent.
5848 @deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5850 You can use the built-in function @code{__builtin_types_compatible_p} to
5851 determine whether two types are the same.
5853 This built-in function returns 1 if the unqualified versions of the
5854 types @var{type1} and @var{type2} (which are types, not expressions) are
5855 compatible, 0 otherwise. The result of this built-in function can be
5856 used in integer constant expressions.
5858 This built-in function ignores top level qualifiers (e.g., @code{const},
5859 @code{volatile}). For example, @code{int} is equivalent to @code{const
5862 The type @code{int[]} and @code{int[5]} are compatible. On the other
5863 hand, @code{int} and @code{char *} are not compatible, even if the size
5864 of their types, on the particular architecture are the same. Also, the
5865 amount of pointer indirection is taken into account when determining
5866 similarity. Consequently, @code{short *} is not similar to
5867 @code{short **}. Furthermore, two types that are typedefed are
5868 considered compatible if their underlying types are compatible.
5870 An @code{enum} type is not considered to be compatible with another
5871 @code{enum} type even if both are compatible with the same integer
5872 type; this is what the C standard specifies.
5873 For example, @code{enum @{foo, bar@}} is not similar to
5874 @code{enum @{hot, dog@}}.
5876 You would typically use this function in code whose execution varies
5877 depending on the arguments' types. For example:
5882 typeof (x) tmp = (x); \
5883 if (__builtin_types_compatible_p (typeof (x), long double)) \
5884 tmp = foo_long_double (tmp); \
5885 else if (__builtin_types_compatible_p (typeof (x), double)) \
5886 tmp = foo_double (tmp); \
5887 else if (__builtin_types_compatible_p (typeof (x), float)) \
5888 tmp = foo_float (tmp); \
5895 @emph{Note:} This construct is only available for C@.
5899 @deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5901 You can use the built-in function @code{__builtin_choose_expr} to
5902 evaluate code depending on the value of a constant expression. This
5903 built-in function returns @var{exp1} if @var{const_exp}, which is a
5904 constant expression that must be able to be determined at compile time,
5905 is nonzero. Otherwise it returns 0.
5907 This built-in function is analogous to the @samp{? :} operator in C,
5908 except that the expression returned has its type unaltered by promotion
5909 rules. Also, the built-in function does not evaluate the expression
5910 that was not chosen. For example, if @var{const_exp} evaluates to true,
5911 @var{exp2} is not evaluated even if it has side-effects.
5913 This built-in function can return an lvalue if the chosen argument is an
5916 If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5917 type. Similarly, if @var{exp2} is returned, its return type is the same
5924 __builtin_choose_expr ( \
5925 __builtin_types_compatible_p (typeof (x), double), \
5927 __builtin_choose_expr ( \
5928 __builtin_types_compatible_p (typeof (x), float), \
5930 /* @r{The void expression results in a compile-time error} \
5931 @r{when assigning the result to something.} */ \
5935 @emph{Note:} This construct is only available for C@. Furthermore, the
5936 unused expression (@var{exp1} or @var{exp2} depending on the value of
5937 @var{const_exp}) may still generate syntax errors. This may change in
5942 @deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5943 You can use the built-in function @code{__builtin_constant_p} to
5944 determine if a value is known to be constant at compile-time and hence
5945 that GCC can perform constant-folding on expressions involving that
5946 value. The argument of the function is the value to test. The function
5947 returns the integer 1 if the argument is known to be a compile-time
5948 constant and 0 if it is not known to be a compile-time constant. A
5949 return of 0 does not indicate that the value is @emph{not} a constant,
5950 but merely that GCC cannot prove it is a constant with the specified
5951 value of the @option{-O} option.
5953 You would typically use this function in an embedded application where
5954 memory was a critical resource. If you have some complex calculation,
5955 you may want it to be folded if it involves constants, but need to call
5956 a function if it does not. For example:
5959 #define Scale_Value(X) \
5960 (__builtin_constant_p (X) \
5961 ? ((X) * SCALE + OFFSET) : Scale (X))
5964 You may use this built-in function in either a macro or an inline
5965 function. However, if you use it in an inlined function and pass an
5966 argument of the function as the argument to the built-in, GCC will
5967 never return 1 when you call the inline function with a string constant
5968 or compound literal (@pxref{Compound Literals}) and will not return 1
5969 when you pass a constant numeric value to the inline function unless you
5970 specify the @option{-O} option.
5972 You may also use @code{__builtin_constant_p} in initializers for static
5973 data. For instance, you can write
5976 static const int table[] = @{
5977 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5983 This is an acceptable initializer even if @var{EXPRESSION} is not a
5984 constant expression. GCC must be more conservative about evaluating the
5985 built-in in this case, because it has no opportunity to perform
5988 Previous versions of GCC did not accept this built-in in data
5989 initializers. The earliest version where it is completely safe is
5993 @deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5994 @opindex fprofile-arcs
5995 You may use @code{__builtin_expect} to provide the compiler with
5996 branch prediction information. In general, you should prefer to
5997 use actual profile feedback for this (@option{-fprofile-arcs}), as
5998 programmers are notoriously bad at predicting how their programs
5999 actually perform. However, there are applications in which this
6000 data is hard to collect.
6002 The return value is the value of @var{exp}, which should be an
6003 integral expression. The value of @var{c} must be a compile-time
6004 constant. The semantics of the built-in are that it is expected
6005 that @var{exp} == @var{c}. For example:
6008 if (__builtin_expect (x, 0))
6013 would indicate that we do not expect to call @code{foo}, since
6014 we expect @code{x} to be zero. Since you are limited to integral
6015 expressions for @var{exp}, you should use constructions such as
6018 if (__builtin_expect (ptr != NULL, 1))
6023 when testing pointer or floating-point values.
6026 @deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
6027 This function is used to minimize cache-miss latency by moving data into
6028 a cache before it is accessed.
6029 You can insert calls to @code{__builtin_prefetch} into code for which
6030 you know addresses of data in memory that is likely to be accessed soon.
6031 If the target supports them, data prefetch instructions will be generated.
6032 If the prefetch is done early enough before the access then the data will
6033 be in the cache by the time it is accessed.
6035 The value of @var{addr} is the address of the memory to prefetch.
6036 There are two optional arguments, @var{rw} and @var{locality}.
6037 The value of @var{rw} is a compile-time constant one or zero; one
6038 means that the prefetch is preparing for a write to the memory address
6039 and zero, the default, means that the prefetch is preparing for a read.
6040 The value @var{locality} must be a compile-time constant integer between
6041 zero and three. A value of zero means that the data has no temporal
6042 locality, so it need not be left in the cache after the access. A value
6043 of three means that the data has a high degree of temporal locality and
6044 should be left in all levels of cache possible. Values of one and two
6045 mean, respectively, a low or moderate degree of temporal locality. The
6049 for (i = 0; i < n; i++)
6052 __builtin_prefetch (&a[i+j], 1, 1);
6053 __builtin_prefetch (&b[i+j], 0, 1);
6058 Data prefetch does not generate faults if @var{addr} is invalid, but
6059 the address expression itself must be valid. For example, a prefetch
6060 of @code{p->next} will not fault if @code{p->next} is not a valid
6061 address, but evaluation will fault if @code{p} is not a valid address.
6063 If the target does not support data prefetch, the address expression
6064 is evaluated if it includes side effects but no other code is generated
6065 and GCC does not issue a warning.
6068 @deftypefn {Built-in Function} double __builtin_huge_val (void)
6069 Returns a positive infinity, if supported by the floating-point format,
6070 else @code{DBL_MAX}. This function is suitable for implementing the
6071 ISO C macro @code{HUGE_VAL}.
6074 @deftypefn {Built-in Function} float __builtin_huge_valf (void)
6075 Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
6078 @deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
6079 Similar to @code{__builtin_huge_val}, except the return
6080 type is @code{long double}.
6083 @deftypefn {Built-in Function} double __builtin_inf (void)
6084 Similar to @code{__builtin_huge_val}, except a warning is generated
6085 if the target floating-point format does not support infinities.
6088 @deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void)
6089 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}.
6092 @deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void)
6093 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}.
6096 @deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void)
6097 Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}.
6100 @deftypefn {Built-in Function} float __builtin_inff (void)
6101 Similar to @code{__builtin_inf}, except the return type is @code{float}.
6102 This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
6105 @deftypefn {Built-in Function} {long double} __builtin_infl (void)
6106 Similar to @code{__builtin_inf}, except the return
6107 type is @code{long double}.
6110 @deftypefn {Built-in Function} double __builtin_nan (const char *str)
6111 This is an implementation of the ISO C99 function @code{nan}.
6113 Since ISO C99 defines this function in terms of @code{strtod}, which we
6114 do not implement, a description of the parsing is in order. The string
6115 is parsed as by @code{strtol}; that is, the base is recognized by
6116 leading @samp{0} or @samp{0x} prefixes. The number parsed is placed
6117 in the significand such that the least significant bit of the number
6118 is at the least significant bit of the significand. The number is
6119 truncated to fit the significand field provided. The significand is
6120 forced to be a quiet NaN@.
6122 This function, if given a string literal all of which would have been
6123 consumed by strtol, is evaluated early enough that it is considered a
6124 compile-time constant.
6127 @deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str)
6128 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}.
6131 @deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str)
6132 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}.
6135 @deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str)
6136 Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}.
6139 @deftypefn {Built-in Function} float __builtin_nanf (const char *str)
6140 Similar to @code{__builtin_nan}, except the return type is @code{float}.
6143 @deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
6144 Similar to @code{__builtin_nan}, except the return type is @code{long double}.
6147 @deftypefn {Built-in Function} double __builtin_nans (const char *str)
6148 Similar to @code{__builtin_nan}, except the significand is forced
6149 to be a signaling NaN@. The @code{nans} function is proposed by
6150 @uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
6153 @deftypefn {Built-in Function} float __builtin_nansf (const char *str)
6154 Similar to @code{__builtin_nans}, except the return type is @code{float}.
6157 @deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
6158 Similar to @code{__builtin_nans}, except the return type is @code{long double}.
6161 @deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
6162 Returns one plus the index of the least significant 1-bit of @var{x}, or
6163 if @var{x} is zero, returns zero.
6166 @deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
6167 Returns the number of leading 0-bits in @var{x}, starting at the most
6168 significant bit position. If @var{x} is 0, the result is undefined.
6171 @deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
6172 Returns the number of trailing 0-bits in @var{x}, starting at the least
6173 significant bit position. If @var{x} is 0, the result is undefined.
6176 @deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
6177 Returns the number of 1-bits in @var{x}.
6180 @deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
6181 Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x}
6185 @deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
6186 Similar to @code{__builtin_ffs}, except the argument type is
6187 @code{unsigned long}.
6190 @deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
6191 Similar to @code{__builtin_clz}, except the argument type is
6192 @code{unsigned long}.
6195 @deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
6196 Similar to @code{__builtin_ctz}, except the argument type is
6197 @code{unsigned long}.
6200 @deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
6201 Similar to @code{__builtin_popcount}, except the argument type is
6202 @code{unsigned long}.
6205 @deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
6206 Similar to @code{__builtin_parity}, except the argument type is
6207 @code{unsigned long}.
6210 @deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
6211 Similar to @code{__builtin_ffs}, except the argument type is
6212 @code{unsigned long long}.
6215 @deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
6216 Similar to @code{__builtin_clz}, except the argument type is
6217 @code{unsigned long long}.
6220 @deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
6221 Similar to @code{__builtin_ctz}, except the argument type is
6222 @code{unsigned long long}.
6225 @deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
6226 Similar to @code{__builtin_popcount}, except the argument type is
6227 @code{unsigned long long}.
6230 @deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
6231 Similar to @code{__builtin_parity}, except the argument type is
6232 @code{unsigned long long}.
6235 @deftypefn {Built-in Function} double __builtin_powi (double, int)
6236 Returns the first argument raised to the power of the second. Unlike the
6237 @code{pow} function no guarantees about precision and rounding are made.
6240 @deftypefn {Built-in Function} float __builtin_powif (float, int)
6241 Similar to @code{__builtin_powi}, except the argument and return types
6245 @deftypefn {Built-in Function} {long double} __builtin_powil (long double, int)
6246 Similar to @code{__builtin_powi}, except the argument and return types
6247 are @code{long double}.
6251 @node Target Builtins
6252 @section Built-in Functions Specific to Particular Target Machines
6254 On some target machines, GCC supports many built-in functions specific
6255 to those machines. Generally these generate calls to specific machine
6256 instructions, but allow the compiler to schedule those calls.
6259 * Alpha Built-in Functions::
6260 * ARM Built-in Functions::
6261 * Blackfin Built-in Functions::
6262 * FR-V Built-in Functions::
6263 * X86 Built-in Functions::
6264 * MIPS DSP Built-in Functions::
6265 * MIPS Paired-Single Support::
6266 * PowerPC AltiVec Built-in Functions::
6267 * SPARC VIS Built-in Functions::
6270 @node Alpha Built-in Functions
6271 @subsection Alpha Built-in Functions
6273 These built-in functions are available for the Alpha family of
6274 processors, depending on the command-line switches used.
6276 The following built-in functions are always available. They
6277 all generate the machine instruction that is part of the name.
6280 long __builtin_alpha_implver (void)
6281 long __builtin_alpha_rpcc (void)
6282 long __builtin_alpha_amask (long)
6283 long __builtin_alpha_cmpbge (long, long)
6284 long __builtin_alpha_extbl (long, long)
6285 long __builtin_alpha_extwl (long, long)
6286 long __builtin_alpha_extll (long, long)
6287 long __builtin_alpha_extql (long, long)
6288 long __builtin_alpha_extwh (long, long)
6289 long __builtin_alpha_extlh (long, long)
6290 long __builtin_alpha_extqh (long, long)
6291 long __builtin_alpha_insbl (long, long)
6292 long __builtin_alpha_inswl (long, long)
6293 long __builtin_alpha_insll (long, long)
6294 long __builtin_alpha_insql (long, long)
6295 long __builtin_alpha_inswh (long, long)
6296 long __builtin_alpha_inslh (long, long)
6297 long __builtin_alpha_insqh (long, long)
6298 long __builtin_alpha_mskbl (long, long)
6299 long __builtin_alpha_mskwl (long, long)
6300 long __builtin_alpha_mskll (long, long)
6301 long __builtin_alpha_mskql (long, long)
6302 long __builtin_alpha_mskwh (long, long)
6303 long __builtin_alpha_msklh (long, long)
6304 long __builtin_alpha_mskqh (long, long)
6305 long __builtin_alpha_umulh (long, long)
6306 long __builtin_alpha_zap (long, long)
6307 long __builtin_alpha_zapnot (long, long)
6310 The following built-in functions are always with @option{-mmax}
6311 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
6312 later. They all generate the machine instruction that is part
6316 long __builtin_alpha_pklb (long)
6317 long __builtin_alpha_pkwb (long)
6318 long __builtin_alpha_unpkbl (long)
6319 long __builtin_alpha_unpkbw (long)
6320 long __builtin_alpha_minub8 (long, long)
6321 long __builtin_alpha_minsb8 (long, long)
6322 long __builtin_alpha_minuw4 (long, long)
6323 long __builtin_alpha_minsw4 (long, long)
6324 long __builtin_alpha_maxub8 (long, long)
6325 long __builtin_alpha_maxsb8 (long, long)
6326 long __builtin_alpha_maxuw4 (long, long)
6327 long __builtin_alpha_maxsw4 (long, long)
6328 long __builtin_alpha_perr (long, long)
6331 The following built-in functions are always with @option{-mcix}
6332 or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
6333 later. They all generate the machine instruction that is part
6337 long __builtin_alpha_cttz (long)
6338 long __builtin_alpha_ctlz (long)
6339 long __builtin_alpha_ctpop (long)
6342 The following builtins are available on systems that use the OSF/1
6343 PALcode. Normally they invoke the @code{rduniq} and @code{wruniq}
6344 PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
6345 @code{rdval} and @code{wrval}.
6348 void *__builtin_thread_pointer (void)
6349 void __builtin_set_thread_pointer (void *)
6352 @node ARM Built-in Functions
6353 @subsection ARM Built-in Functions
6355 These built-in functions are available for the ARM family of
6356 processors, when the @option{-mcpu=iwmmxt} switch is used:
6359 typedef int v2si __attribute__ ((vector_size (8)));
6360 typedef short v4hi __attribute__ ((vector_size (8)));
6361 typedef char v8qi __attribute__ ((vector_size (8)));
6363 int __builtin_arm_getwcx (int)
6364 void __builtin_arm_setwcx (int, int)
6365 int __builtin_arm_textrmsb (v8qi, int)
6366 int __builtin_arm_textrmsh (v4hi, int)
6367 int __builtin_arm_textrmsw (v2si, int)
6368 int __builtin_arm_textrmub (v8qi, int)
6369 int __builtin_arm_textrmuh (v4hi, int)
6370 int __builtin_arm_textrmuw (v2si, int)
6371 v8qi __builtin_arm_tinsrb (v8qi, int)
6372 v4hi __builtin_arm_tinsrh (v4hi, int)
6373 v2si __builtin_arm_tinsrw (v2si, int)
6374 long long __builtin_arm_tmia (long long, int, int)
6375 long long __builtin_arm_tmiabb (long long, int, int)
6376 long long __builtin_arm_tmiabt (long long, int, int)
6377 long long __builtin_arm_tmiaph (long long, int, int)
6378 long long __builtin_arm_tmiatb (long long, int, int)
6379 long long __builtin_arm_tmiatt (long long, int, int)
6380 int __builtin_arm_tmovmskb (v8qi)
6381 int __builtin_arm_tmovmskh (v4hi)
6382 int __builtin_arm_tmovmskw (v2si)
6383 long long __builtin_arm_waccb (v8qi)
6384 long long __builtin_arm_wacch (v4hi)
6385 long long __builtin_arm_waccw (v2si)
6386 v8qi __builtin_arm_waddb (v8qi, v8qi)
6387 v8qi __builtin_arm_waddbss (v8qi, v8qi)
6388 v8qi __builtin_arm_waddbus (v8qi, v8qi)
6389 v4hi __builtin_arm_waddh (v4hi, v4hi)
6390 v4hi __builtin_arm_waddhss (v4hi, v4hi)
6391 v4hi __builtin_arm_waddhus (v4hi, v4hi)
6392 v2si __builtin_arm_waddw (v2si, v2si)
6393 v2si __builtin_arm_waddwss (v2si, v2si)
6394 v2si __builtin_arm_waddwus (v2si, v2si)
6395 v8qi __builtin_arm_walign (v8qi, v8qi, int)
6396 long long __builtin_arm_wand(long long, long long)
6397 long long __builtin_arm_wandn (long long, long long)
6398 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
6399 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
6400 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
6401 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
6402 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
6403 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
6404 v2si __builtin_arm_wcmpeqw (v2si, v2si)
6405 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
6406 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
6407 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
6408 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
6409 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
6410 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
6411 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
6412 long long __builtin_arm_wmacsz (v4hi, v4hi)
6413 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
6414 long long __builtin_arm_wmacuz (v4hi, v4hi)
6415 v4hi __builtin_arm_wmadds (v4hi, v4hi)
6416 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
6417 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
6418 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
6419 v2si __builtin_arm_wmaxsw (v2si, v2si)
6420 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
6421 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
6422 v2si __builtin_arm_wmaxuw (v2si, v2si)
6423 v8qi __builtin_arm_wminsb (v8qi, v8qi)
6424 v4hi __builtin_arm_wminsh (v4hi, v4hi)
6425 v2si __builtin_arm_wminsw (v2si, v2si)
6426 v8qi __builtin_arm_wminub (v8qi, v8qi)
6427 v4hi __builtin_arm_wminuh (v4hi, v4hi)
6428 v2si __builtin_arm_wminuw (v2si, v2si)
6429 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
6430 v4hi __builtin_arm_wmulul (v4hi, v4hi)
6431 v4hi __builtin_arm_wmulum (v4hi, v4hi)
6432 long long __builtin_arm_wor (long long, long long)
6433 v2si __builtin_arm_wpackdss (long long, long long)
6434 v2si __builtin_arm_wpackdus (long long, long long)
6435 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
6436 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
6437 v4hi __builtin_arm_wpackwss (v2si, v2si)
6438 v4hi __builtin_arm_wpackwus (v2si, v2si)
6439 long long __builtin_arm_wrord (long long, long long)
6440 long long __builtin_arm_wrordi (long long, int)
6441 v4hi __builtin_arm_wrorh (v4hi, long long)
6442 v4hi __builtin_arm_wrorhi (v4hi, int)
6443 v2si __builtin_arm_wrorw (v2si, long long)
6444 v2si __builtin_arm_wrorwi (v2si, int)
6445 v2si __builtin_arm_wsadb (v8qi, v8qi)
6446 v2si __builtin_arm_wsadbz (v8qi, v8qi)
6447 v2si __builtin_arm_wsadh (v4hi, v4hi)
6448 v2si __builtin_arm_wsadhz (v4hi, v4hi)
6449 v4hi __builtin_arm_wshufh (v4hi, int)
6450 long long __builtin_arm_wslld (long long, long long)
6451 long long __builtin_arm_wslldi (long long, int)
6452 v4hi __builtin_arm_wsllh (v4hi, long long)
6453 v4hi __builtin_arm_wsllhi (v4hi, int)
6454 v2si __builtin_arm_wsllw (v2si, long long)
6455 v2si __builtin_arm_wsllwi (v2si, int)
6456 long long __builtin_arm_wsrad (long long, long long)
6457 long long __builtin_arm_wsradi (long long, int)
6458 v4hi __builtin_arm_wsrah (v4hi, long long)
6459 v4hi __builtin_arm_wsrahi (v4hi, int)
6460 v2si __builtin_arm_wsraw (v2si, long long)
6461 v2si __builtin_arm_wsrawi (v2si, int)
6462 long long __builtin_arm_wsrld (long long, long long)
6463 long long __builtin_arm_wsrldi (long long, int)
6464 v4hi __builtin_arm_wsrlh (v4hi, long long)
6465 v4hi __builtin_arm_wsrlhi (v4hi, int)
6466 v2si __builtin_arm_wsrlw (v2si, long long)
6467 v2si __builtin_arm_wsrlwi (v2si, int)
6468 v8qi __builtin_arm_wsubb (v8qi, v8qi)
6469 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
6470 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
6471 v4hi __builtin_arm_wsubh (v4hi, v4hi)
6472 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
6473 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
6474 v2si __builtin_arm_wsubw (v2si, v2si)
6475 v2si __builtin_arm_wsubwss (v2si, v2si)
6476 v2si __builtin_arm_wsubwus (v2si, v2si)
6477 v4hi __builtin_arm_wunpckehsb (v8qi)
6478 v2si __builtin_arm_wunpckehsh (v4hi)
6479 long long __builtin_arm_wunpckehsw (v2si)
6480 v4hi __builtin_arm_wunpckehub (v8qi)
6481 v2si __builtin_arm_wunpckehuh (v4hi)
6482 long long __builtin_arm_wunpckehuw (v2si)
6483 v4hi __builtin_arm_wunpckelsb (v8qi)
6484 v2si __builtin_arm_wunpckelsh (v4hi)
6485 long long __builtin_arm_wunpckelsw (v2si)
6486 v4hi __builtin_arm_wunpckelub (v8qi)
6487 v2si __builtin_arm_wunpckeluh (v4hi)
6488 long long __builtin_arm_wunpckeluw (v2si)
6489 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
6490 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
6491 v2si __builtin_arm_wunpckihw (v2si, v2si)
6492 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
6493 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
6494 v2si __builtin_arm_wunpckilw (v2si, v2si)
6495 long long __builtin_arm_wxor (long long, long long)
6496 long long __builtin_arm_wzero ()
6499 @node Blackfin Built-in Functions
6500 @subsection Blackfin Built-in Functions
6502 Currently, there are two Blackfin-specific built-in functions. These are
6503 used for generating @code{CSYNC} and @code{SSYNC} machine insns without
6504 using inline assembly; by using these built-in functions the compiler can
6505 automatically add workarounds for hardware errata involving these
6506 instructions. These functions are named as follows:
6509 void __builtin_bfin_csync (void)
6510 void __builtin_bfin_ssync (void)
6513 @node FR-V Built-in Functions
6514 @subsection FR-V Built-in Functions
6516 GCC provides many FR-V-specific built-in functions. In general,
6517 these functions are intended to be compatible with those described
6518 by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu
6519 Semiconductor}. The two exceptions are @code{__MDUNPACKH} and
6520 @code{__MBTOHE}, the gcc forms of which pass 128-bit values by
6521 pointer rather than by value.
6523 Most of the functions are named after specific FR-V instructions.
6524 Such functions are said to be ``directly mapped'' and are summarized
6525 here in tabular form.
6529 * Directly-mapped Integer Functions::
6530 * Directly-mapped Media Functions::
6531 * Raw read/write Functions::
6532 * Other Built-in Functions::
6535 @node Argument Types
6536 @subsubsection Argument Types
6538 The arguments to the built-in functions can be divided into three groups:
6539 register numbers, compile-time constants and run-time values. In order
6540 to make this classification clear at a glance, the arguments and return
6541 values are given the following pseudo types:
6543 @multitable @columnfractions .20 .30 .15 .35
6544 @item Pseudo type @tab Real C type @tab Constant? @tab Description
6545 @item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword
6546 @item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word
6547 @item @code{sw1} @tab @code{int} @tab No @tab a signed word
6548 @item @code{uw2} @tab @code{unsigned long long} @tab No
6549 @tab an unsigned doubleword
6550 @item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword
6551 @item @code{const} @tab @code{int} @tab Yes @tab an integer constant
6552 @item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number
6553 @item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number
6556 These pseudo types are not defined by GCC, they are simply a notational
6557 convenience used in this manual.
6559 Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2}
6560 and @code{sw2} are evaluated at run time. They correspond to
6561 register operands in the underlying FR-V instructions.
6563 @code{const} arguments represent immediate operands in the underlying
6564 FR-V instructions. They must be compile-time constants.
6566 @code{acc} arguments are evaluated at compile time and specify the number
6567 of an accumulator register. For example, an @code{acc} argument of 2
6568 will select the ACC2 register.
6570 @code{iacc} arguments are similar to @code{acc} arguments but specify the
6571 number of an IACC register. See @pxref{Other Built-in Functions}
6574 @node Directly-mapped Integer Functions
6575 @subsubsection Directly-mapped Integer Functions
6577 The functions listed below map directly to FR-V I-type instructions.
6579 @multitable @columnfractions .45 .32 .23
6580 @item Function prototype @tab Example usage @tab Assembly output
6581 @item @code{sw1 __ADDSS (sw1, sw1)}
6582 @tab @code{@var{c} = __ADDSS (@var{a}, @var{b})}
6583 @tab @code{ADDSS @var{a},@var{b},@var{c}}
6584 @item @code{sw1 __SCAN (sw1, sw1)}
6585 @tab @code{@var{c} = __SCAN (@var{a}, @var{b})}
6586 @tab @code{SCAN @var{a},@var{b},@var{c}}
6587 @item @code{sw1 __SCUTSS (sw1)}
6588 @tab @code{@var{b} = __SCUTSS (@var{a})}
6589 @tab @code{SCUTSS @var{a},@var{b}}
6590 @item @code{sw1 __SLASS (sw1, sw1)}
6591 @tab @code{@var{c} = __SLASS (@var{a}, @var{b})}
6592 @tab @code{SLASS @var{a},@var{b},@var{c}}
6593 @item @code{void __SMASS (sw1, sw1)}
6594 @tab @code{__SMASS (@var{a}, @var{b})}
6595 @tab @code{SMASS @var{a},@var{b}}
6596 @item @code{void __SMSSS (sw1, sw1)}
6597 @tab @code{__SMSSS (@var{a}, @var{b})}
6598 @tab @code{SMSSS @var{a},@var{b}}
6599 @item @code{void __SMU (sw1, sw1)}
6600 @tab @code{__SMU (@var{a}, @var{b})}
6601 @tab @code{SMU @var{a},@var{b}}
6602 @item @code{sw2 __SMUL (sw1, sw1)}
6603 @tab @code{@var{c} = __SMUL (@var{a}, @var{b})}
6604 @tab @code{SMUL @var{a},@var{b},@var{c}}
6605 @item @code{sw1 __SUBSS (sw1, sw1)}
6606 @tab @code{@var{c} = __SUBSS (@var{a}, @var{b})}
6607 @tab @code{SUBSS @var{a},@var{b},@var{c}}
6608 @item @code{uw2 __UMUL (uw1, uw1)}
6609 @tab @code{@var{c} = __UMUL (@var{a}, @var{b})}
6610 @tab @code{UMUL @var{a},@var{b},@var{c}}
6613 @node Directly-mapped Media Functions
6614 @subsubsection Directly-mapped Media Functions
6616 The functions listed below map directly to FR-V M-type instructions.
6618 @multitable @columnfractions .45 .32 .23
6619 @item Function prototype @tab Example usage @tab Assembly output
6620 @item @code{uw1 __MABSHS (sw1)}
6621 @tab @code{@var{b} = __MABSHS (@var{a})}
6622 @tab @code{MABSHS @var{a},@var{b}}
6623 @item @code{void __MADDACCS (acc, acc)}
6624 @tab @code{__MADDACCS (@var{b}, @var{a})}
6625 @tab @code{MADDACCS @var{a},@var{b}}
6626 @item @code{sw1 __MADDHSS (sw1, sw1)}
6627 @tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})}
6628 @tab @code{MADDHSS @var{a},@var{b},@var{c}}
6629 @item @code{uw1 __MADDHUS (uw1, uw1)}
6630 @tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})}
6631 @tab @code{MADDHUS @var{a},@var{b},@var{c}}
6632 @item @code{uw1 __MAND (uw1, uw1)}
6633 @tab @code{@var{c} = __MAND (@var{a}, @var{b})}
6634 @tab @code{MAND @var{a},@var{b},@var{c}}
6635 @item @code{void __MASACCS (acc, acc)}
6636 @tab @code{__MASACCS (@var{b}, @var{a})}
6637 @tab @code{MASACCS @var{a},@var{b}}
6638 @item @code{uw1 __MAVEH (uw1, uw1)}
6639 @tab @code{@var{c} = __MAVEH (@var{a}, @var{b})}
6640 @tab @code{MAVEH @var{a},@var{b},@var{c}}
6641 @item @code{uw2 __MBTOH (uw1)}
6642 @tab @code{@var{b} = __MBTOH (@var{a})}
6643 @tab @code{MBTOH @var{a},@var{b}}
6644 @item @code{void __MBTOHE (uw1 *, uw1)}
6645 @tab @code{__MBTOHE (&@var{b}, @var{a})}
6646 @tab @code{MBTOHE @var{a},@var{b}}
6647 @item @code{void __MCLRACC (acc)}
6648 @tab @code{__MCLRACC (@var{a})}
6649 @tab @code{MCLRACC @var{a}}
6650 @item @code{void __MCLRACCA (void)}
6651 @tab @code{__MCLRACCA ()}
6652 @tab @code{MCLRACCA}
6653 @item @code{uw1 __Mcop1 (uw1, uw1)}
6654 @tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})}
6655 @tab @code{Mcop1 @var{a},@var{b},@var{c}}
6656 @item @code{uw1 __Mcop2 (uw1, uw1)}
6657 @tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})}
6658 @tab @code{Mcop2 @var{a},@var{b},@var{c}}
6659 @item @code{uw1 __MCPLHI (uw2, const)}
6660 @tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})}
6661 @tab @code{MCPLHI @var{a},#@var{b},@var{c}}
6662 @item @code{uw1 __MCPLI (uw2, const)}
6663 @tab @code{@var{c} = __MCPLI (@var{a}, @var{b})}
6664 @tab @code{MCPLI @var{a},#@var{b},@var{c}}
6665 @item @code{void __MCPXIS (acc, sw1, sw1)}
6666 @tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})}
6667 @tab @code{MCPXIS @var{a},@var{b},@var{c}}
6668 @item @code{void __MCPXIU (acc, uw1, uw1)}
6669 @tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})}
6670 @tab @code{MCPXIU @var{a},@var{b},@var{c}}
6671 @item @code{void __MCPXRS (acc, sw1, sw1)}
6672 @tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})}
6673 @tab @code{MCPXRS @var{a},@var{b},@var{c}}
6674 @item @code{void __MCPXRU (acc, uw1, uw1)}
6675 @tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})}
6676 @tab @code{MCPXRU @var{a},@var{b},@var{c}}
6677 @item @code{uw1 __MCUT (acc, uw1)}
6678 @tab @code{@var{c} = __MCUT (@var{a}, @var{b})}
6679 @tab @code{MCUT @var{a},@var{b},@var{c}}
6680 @item @code{uw1 __MCUTSS (acc, sw1)}
6681 @tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})}
6682 @tab @code{MCUTSS @var{a},@var{b},@var{c}}
6683 @item @code{void __MDADDACCS (acc, acc)}
6684 @tab @code{__MDADDACCS (@var{b}, @var{a})}
6685 @tab @code{MDADDACCS @var{a},@var{b}}
6686 @item @code{void __MDASACCS (acc, acc)}
6687 @tab @code{__MDASACCS (@var{b}, @var{a})}
6688 @tab @code{MDASACCS @var{a},@var{b}}
6689 @item @code{uw2 __MDCUTSSI (acc, const)}
6690 @tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})}
6691 @tab @code{MDCUTSSI @var{a},#@var{b},@var{c}}
6692 @item @code{uw2 __MDPACKH (uw2, uw2)}
6693 @tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})}
6694 @tab @code{MDPACKH @var{a},@var{b},@var{c}}
6695 @item @code{uw2 __MDROTLI (uw2, const)}
6696 @tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})}
6697 @tab @code{MDROTLI @var{a},#@var{b},@var{c}}
6698 @item @code{void __MDSUBACCS (acc, acc)}
6699 @tab @code{__MDSUBACCS (@var{b}, @var{a})}
6700 @tab @code{MDSUBACCS @var{a},@var{b}}
6701 @item @code{void __MDUNPACKH (uw1 *, uw2)}
6702 @tab @code{__MDUNPACKH (&@var{b}, @var{a})}
6703 @tab @code{MDUNPACKH @var{a},@var{b}}
6704 @item @code{uw2 __MEXPDHD (uw1, const)}
6705 @tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})}
6706 @tab @code{MEXPDHD @var{a},#@var{b},@var{c}}
6707 @item @code{uw1 __MEXPDHW (uw1, const)}
6708 @tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})}
6709 @tab @code{MEXPDHW @var{a},#@var{b},@var{c}}
6710 @item @code{uw1 __MHDSETH (uw1, const)}
6711 @tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})}
6712 @tab @code{MHDSETH @var{a},#@var{b},@var{c}}
6713 @item @code{sw1 __MHDSETS (const)}
6714 @tab @code{@var{b} = __MHDSETS (@var{a})}
6715 @tab @code{MHDSETS #@var{a},@var{b}}
6716 @item @code{uw1 __MHSETHIH (uw1, const)}
6717 @tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})}
6718 @tab @code{MHSETHIH #@var{a},@var{b}}
6719 @item @code{sw1 __MHSETHIS (sw1, const)}
6720 @tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})}
6721 @tab @code{MHSETHIS #@var{a},@var{b}}
6722 @item @code{uw1 __MHSETLOH (uw1, const)}
6723 @tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})}
6724 @tab @code{MHSETLOH #@var{a},@var{b}}
6725 @item @code{sw1 __MHSETLOS (sw1, const)}
6726 @tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})}
6727 @tab @code{MHSETLOS #@var{a},@var{b}}
6728 @item @code{uw1 __MHTOB (uw2)}
6729 @tab @code{@var{b} = __MHTOB (@var{a})}
6730 @tab @code{MHTOB @var{a},@var{b}}
6731 @item @code{void __MMACHS (acc, sw1, sw1)}
6732 @tab @code{__MMACHS (@var{c}, @var{a}, @var{b})}
6733 @tab @code{MMACHS @var{a},@var{b},@var{c}}
6734 @item @code{void __MMACHU (acc, uw1, uw1)}
6735 @tab @code{__MMACHU (@var{c}, @var{a}, @var{b})}
6736 @tab @code{MMACHU @var{a},@var{b},@var{c}}
6737 @item @code{void __MMRDHS (acc, sw1, sw1)}
6738 @tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})}
6739 @tab @code{MMRDHS @var{a},@var{b},@var{c}}
6740 @item @code{void __MMRDHU (acc, uw1, uw1)}
6741 @tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})}
6742 @tab @code{MMRDHU @var{a},@var{b},@var{c}}
6743 @item @code{void __MMULHS (acc, sw1, sw1)}
6744 @tab @code{__MMULHS (@var{c}, @var{a}, @var{b})}
6745 @tab @code{MMULHS @var{a},@var{b},@var{c}}
6746 @item @code{void __MMULHU (acc, uw1, uw1)}
6747 @tab @code{__MMULHU (@var{c}, @var{a}, @var{b})}
6748 @tab @code{MMULHU @var{a},@var{b},@var{c}}
6749 @item @code{void __MMULXHS (acc, sw1, sw1)}
6750 @tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})}
6751 @tab @code{MMULXHS @var{a},@var{b},@var{c}}
6752 @item @code{void __MMULXHU (acc, uw1, uw1)}
6753 @tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})}
6754 @tab @code{MMULXHU @var{a},@var{b},@var{c}}
6755 @item @code{uw1 __MNOT (uw1)}
6756 @tab @code{@var{b} = __MNOT (@var{a})}
6757 @tab @code{MNOT @var{a},@var{b}}
6758 @item @code{uw1 __MOR (uw1, uw1)}
6759 @tab @code{@var{c} = __MOR (@var{a}, @var{b})}
6760 @tab @code{MOR @var{a},@var{b},@var{c}}
6761 @item @code{uw1 __MPACKH (uh, uh)}
6762 @tab @code{@var{c} = __MPACKH (@var{a}, @var{b})}
6763 @tab @code{MPACKH @var{a},@var{b},@var{c}}
6764 @item @code{sw2 __MQADDHSS (sw2, sw2)}
6765 @tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})}
6766 @tab @code{MQADDHSS @var{a},@var{b},@var{c}}
6767 @item @code{uw2 __MQADDHUS (uw2, uw2)}
6768 @tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})}
6769 @tab @code{MQADDHUS @var{a},@var{b},@var{c}}
6770 @item @code{void __MQCPXIS (acc, sw2, sw2)}
6771 @tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})}
6772 @tab @code{MQCPXIS @var{a},@var{b},@var{c}}
6773 @item @code{void __MQCPXIU (acc, uw2, uw2)}
6774 @tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})}
6775 @tab @code{MQCPXIU @var{a},@var{b},@var{c}}
6776 @item @code{void __MQCPXRS (acc, sw2, sw2)}
6777 @tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})}
6778 @tab @code{MQCPXRS @var{a},@var{b},@var{c}}
6779 @item @code{void __MQCPXRU (acc, uw2, uw2)}
6780 @tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})}
6781 @tab @code{MQCPXRU @var{a},@var{b},@var{c}}
6782 @item @code{sw2 __MQLCLRHS (sw2, sw2)}
6783 @tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})}
6784 @tab @code{MQLCLRHS @var{a},@var{b},@var{c}}
6785 @item @code{sw2 __MQLMTHS (sw2, sw2)}
6786 @tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})}
6787 @tab @code{MQLMTHS @var{a},@var{b},@var{c}}
6788 @item @code{void __MQMACHS (acc, sw2, sw2)}
6789 @tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})}
6790 @tab @code{MQMACHS @var{a},@var{b},@var{c}}
6791 @item @code{void __MQMACHU (acc, uw2, uw2)}
6792 @tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})}
6793 @tab @code{MQMACHU @var{a},@var{b},@var{c}}
6794 @item @code{void __MQMACXHS (acc, sw2, sw2)}
6795 @tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})}
6796 @tab @code{MQMACXHS @var{a},@var{b},@var{c}}
6797 @item @code{void __MQMULHS (acc, sw2, sw2)}
6798 @tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})}
6799 @tab @code{MQMULHS @var{a},@var{b},@var{c}}
6800 @item @code{void __MQMULHU (acc, uw2, uw2)}
6801 @tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})}
6802 @tab @code{MQMULHU @var{a},@var{b},@var{c}}
6803 @item @code{void __MQMULXHS (acc, sw2, sw2)}
6804 @tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})}
6805 @tab @code{MQMULXHS @var{a},@var{b},@var{c}}
6806 @item @code{void __MQMULXHU (acc, uw2, uw2)}
6807 @tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})}
6808 @tab @code{MQMULXHU @var{a},@var{b},@var{c}}
6809 @item @code{sw2 __MQSATHS (sw2, sw2)}
6810 @tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})}
6811 @tab @code{MQSATHS @var{a},@var{b},@var{c}}
6812 @item @code{uw2 __MQSLLHI (uw2, int)}
6813 @tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})}
6814 @tab @code{MQSLLHI @var{a},@var{b},@var{c}}
6815 @item @code{sw2 __MQSRAHI (sw2, int)}
6816 @tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})}
6817 @tab @code{MQSRAHI @var{a},@var{b},@var{c}}
6818 @item @code{sw2 __MQSUBHSS (sw2, sw2)}
6819 @tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})}
6820 @tab @code{MQSUBHSS @var{a},@var{b},@var{c}}
6821 @item @code{uw2 __MQSUBHUS (uw2, uw2)}
6822 @tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})}
6823 @tab @code{MQSUBHUS @var{a},@var{b},@var{c}}
6824 @item @code{void __MQXMACHS (acc, sw2, sw2)}
6825 @tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})}
6826 @tab @code{MQXMACHS @var{a},@var{b},@var{c}}
6827 @item @code{void __MQXMACXHS (acc, sw2, sw2)}
6828 @tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})}
6829 @tab @code{MQXMACXHS @var{a},@var{b},@var{c}}
6830 @item @code{uw1 __MRDACC (acc)}
6831 @tab @code{@var{b} = __MRDACC (@var{a})}
6832 @tab @code{MRDACC @var{a},@var{b}}
6833 @item @code{uw1 __MRDACCG (acc)}
6834 @tab @code{@var{b} = __MRDACCG (@var{a})}
6835 @tab @code{MRDACCG @var{a},@var{b}}
6836 @item @code{uw1 __MROTLI (uw1, const)}
6837 @tab @code{@var{c} = __MROTLI (@var{a}, @var{b})}
6838 @tab @code{MROTLI @var{a},#@var{b},@var{c}}
6839 @item @code{uw1 __MROTRI (uw1, const)}
6840 @tab @code{@var{c} = __MROTRI (@var{a}, @var{b})}
6841 @tab @code{MROTRI @var{a},#@var{b},@var{c}}
6842 @item @code{sw1 __MSATHS (sw1, sw1)}
6843 @tab @code{@var{c} = __MSATHS (@var{a}, @var{b})}
6844 @tab @code{MSATHS @var{a},@var{b},@var{c}}
6845 @item @code{uw1 __MSATHU (uw1, uw1)}
6846 @tab @code{@var{c} = __MSATHU (@var{a}, @var{b})}
6847 @tab @code{MSATHU @var{a},@var{b},@var{c}}
6848 @item @code{uw1 __MSLLHI (uw1, const)}
6849 @tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})}
6850 @tab @code{MSLLHI @var{a},#@var{b},@var{c}}
6851 @item @code{sw1 __MSRAHI (sw1, const)}
6852 @tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})}
6853 @tab @code{MSRAHI @var{a},#@var{b},@var{c}}
6854 @item @code{uw1 __MSRLHI (uw1, const)}
6855 @tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})}
6856 @tab @code{MSRLHI @var{a},#@var{b},@var{c}}
6857 @item @code{void __MSUBACCS (acc, acc)}
6858 @tab @code{__MSUBACCS (@var{b}, @var{a})}
6859 @tab @code{MSUBACCS @var{a},@var{b}}
6860 @item @code{sw1 __MSUBHSS (sw1, sw1)}
6861 @tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})}
6862 @tab @code{MSUBHSS @var{a},@var{b},@var{c}}
6863 @item @code{uw1 __MSUBHUS (uw1, uw1)}
6864 @tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})}
6865 @tab @code{MSUBHUS @var{a},@var{b},@var{c}}
6866 @item @code{void __MTRAP (void)}
6867 @tab @code{__MTRAP ()}
6869 @item @code{uw2 __MUNPACKH (uw1)}
6870 @tab @code{@var{b} = __MUNPACKH (@var{a})}
6871 @tab @code{MUNPACKH @var{a},@var{b}}
6872 @item @code{uw1 __MWCUT (uw2, uw1)}
6873 @tab @code{@var{c} = __MWCUT (@var{a}, @var{b})}
6874 @tab @code{MWCUT @var{a},@var{b},@var{c}}
6875 @item @code{void __MWTACC (acc, uw1)}
6876 @tab @code{__MWTACC (@var{b}, @var{a})}
6877 @tab @code{MWTACC @var{a},@var{b}}
6878 @item @code{void __MWTACCG (acc, uw1)}
6879 @tab @code{__MWTACCG (@var{b}, @var{a})}
6880 @tab @code{MWTACCG @var{a},@var{b}}
6881 @item @code{uw1 __MXOR (uw1, uw1)}
6882 @tab @code{@var{c} = __MXOR (@var{a}, @var{b})}
6883 @tab @code{MXOR @var{a},@var{b},@var{c}}
6886 @node Raw read/write Functions
6887 @subsubsection Raw read/write Functions
6889 This sections describes built-in functions related to read and write
6890 instructions to access memory. These functions generate
6891 @code{membar} instructions to flush the I/O load and stores where
6892 appropriate, as described in Fujitsu's manual described above.
6896 @item unsigned char __builtin_read8 (void *@var{data})
6897 @item unsigned short __builtin_read16 (void *@var{data})
6898 @item unsigned long __builtin_read32 (void *@var{data})
6899 @item unsigned long long __builtin_read64 (void *@var{data})
6901 @item void __builtin_write8 (void *@var{data}, unsigned char @var{datum})
6902 @item void __builtin_write16 (void *@var{data}, unsigned short @var{datum})
6903 @item void __builtin_write32 (void *@var{data}, unsigned long @var{datum})
6904 @item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum})
6907 @node Other Built-in Functions
6908 @subsubsection Other Built-in Functions
6910 This section describes built-in functions that are not named after
6911 a specific FR-V instruction.
6914 @item sw2 __IACCreadll (iacc @var{reg})
6915 Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved
6916 for future expansion and must be 0.
6918 @item sw1 __IACCreadl (iacc @var{reg})
6919 Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1.
6920 Other values of @var{reg} are rejected as invalid.
6922 @item void __IACCsetll (iacc @var{reg}, sw2 @var{x})
6923 Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument
6924 is reserved for future expansion and must be 0.
6926 @item void __IACCsetl (iacc @var{reg}, sw1 @var{x})
6927 Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg}
6928 is 1. Other values of @var{reg} are rejected as invalid.
6930 @item void __data_prefetch0 (const void *@var{x})
6931 Use the @code{dcpl} instruction to load the contents of address @var{x}
6932 into the data cache.
6934 @item void __data_prefetch (const void *@var{x})
6935 Use the @code{nldub} instruction to load the contents of address @var{x}
6936 into the data cache. The instruction will be issued in slot I1@.
6939 @node X86 Built-in Functions
6940 @subsection X86 Built-in Functions
6942 These built-in functions are available for the i386 and x86-64 family
6943 of computers, depending on the command-line switches used.
6945 Note that, if you specify command-line switches such as @option{-msse},
6946 the compiler could use the extended instruction sets even if the built-ins
6947 are not used explicitly in the program. For this reason, applications
6948 which perform runtime CPU detection must compile separate files for each
6949 supported architecture, using the appropriate flags. In particular,
6950 the file containing the CPU detection code should be compiled without
6953 The following machine modes are available for use with MMX built-in functions
6954 (@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
6955 @code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
6956 vector of eight 8-bit integers. Some of the built-in functions operate on
6957 MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
6959 If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
6960 of two 32-bit floating point values.
6962 If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
6963 floating point values. Some instructions use a vector of four 32-bit
6964 integers, these use @code{V4SI}. Finally, some instructions operate on an
6965 entire vector register, interpreting it as a 128-bit integer, these use mode
6968 The following built-in functions are made available by @option{-mmmx}.
6969 All of them generate the machine instruction that is part of the name.
6972 v8qi __builtin_ia32_paddb (v8qi, v8qi)
6973 v4hi __builtin_ia32_paddw (v4hi, v4hi)
6974 v2si __builtin_ia32_paddd (v2si, v2si)
6975 v8qi __builtin_ia32_psubb (v8qi, v8qi)
6976 v4hi __builtin_ia32_psubw (v4hi, v4hi)
6977 v2si __builtin_ia32_psubd (v2si, v2si)
6978 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
6979 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
6980 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
6981 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
6982 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
6983 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
6984 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
6985 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
6986 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
6987 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
6988 di __builtin_ia32_pand (di, di)
6989 di __builtin_ia32_pandn (di,di)
6990 di __builtin_ia32_por (di, di)
6991 di __builtin_ia32_pxor (di, di)
6992 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
6993 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
6994 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
6995 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
6996 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
6997 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
6998 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
6999 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
7000 v2si __builtin_ia32_punpckhdq (v2si, v2si)
7001 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
7002 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
7003 v2si __builtin_ia32_punpckldq (v2si, v2si)
7004 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
7005 v4hi __builtin_ia32_packssdw (v2si, v2si)
7006 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
7009 The following built-in functions are made available either with
7010 @option{-msse}, or with a combination of @option{-m3dnow} and
7011 @option{-march=athlon}. All of them generate the machine
7012 instruction that is part of the name.
7015 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
7016 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
7017 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
7018 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
7019 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
7020 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
7021 v8qi __builtin_ia32_pminub (v8qi, v8qi)
7022 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
7023 int __builtin_ia32_pextrw (v4hi, int)
7024 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
7025 int __builtin_ia32_pmovmskb (v8qi)
7026 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
7027 void __builtin_ia32_movntq (di *, di)
7028 void __builtin_ia32_sfence (void)
7031 The following built-in functions are available when @option{-msse} is used.
7032 All of them generate the machine instruction that is part of the name.
7035 int __builtin_ia32_comieq (v4sf, v4sf)
7036 int __builtin_ia32_comineq (v4sf, v4sf)
7037 int __builtin_ia32_comilt (v4sf, v4sf)
7038 int __builtin_ia32_comile (v4sf, v4sf)
7039 int __builtin_ia32_comigt (v4sf, v4sf)
7040 int __builtin_ia32_comige (v4sf, v4sf)
7041 int __builtin_ia32_ucomieq (v4sf, v4sf)
7042 int __builtin_ia32_ucomineq (v4sf, v4sf)
7043 int __builtin_ia32_ucomilt (v4sf, v4sf)
7044 int __builtin_ia32_ucomile (v4sf, v4sf)
7045 int __builtin_ia32_ucomigt (v4sf, v4sf)
7046 int __builtin_ia32_ucomige (v4sf, v4sf)
7047 v4sf __builtin_ia32_addps (v4sf, v4sf)
7048 v4sf __builtin_ia32_subps (v4sf, v4sf)
7049 v4sf __builtin_ia32_mulps (v4sf, v4sf)
7050 v4sf __builtin_ia32_divps (v4sf, v4sf)
7051 v4sf __builtin_ia32_addss (v4sf, v4sf)
7052 v4sf __builtin_ia32_subss (v4sf, v4sf)
7053 v4sf __builtin_ia32_mulss (v4sf, v4sf)
7054 v4sf __builtin_ia32_divss (v4sf, v4sf)
7055 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
7056 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
7057 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
7058 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
7059 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
7060 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
7061 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
7062 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
7063 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
7064 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
7065 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
7066 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
7067 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
7068 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
7069 v4si __builtin_ia32_cmpless (v4sf, v4sf)
7070 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
7071 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
7072 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
7073 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
7074 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
7075 v4sf __builtin_ia32_maxps (v4sf, v4sf)
7076 v4sf __builtin_ia32_maxss (v4sf, v4sf)
7077 v4sf __builtin_ia32_minps (v4sf, v4sf)
7078 v4sf __builtin_ia32_minss (v4sf, v4sf)
7079 v4sf __builtin_ia32_andps (v4sf, v4sf)
7080 v4sf __builtin_ia32_andnps (v4sf, v4sf)
7081 v4sf __builtin_ia32_orps (v4sf, v4sf)
7082 v4sf __builtin_ia32_xorps (v4sf, v4sf)
7083 v4sf __builtin_ia32_movss (v4sf, v4sf)
7084 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
7085 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
7086 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
7087 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
7088 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
7089 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
7090 v2si __builtin_ia32_cvtps2pi (v4sf)
7091 int __builtin_ia32_cvtss2si (v4sf)
7092 v2si __builtin_ia32_cvttps2pi (v4sf)
7093 int __builtin_ia32_cvttss2si (v4sf)
7094 v4sf __builtin_ia32_rcpps (v4sf)
7095 v4sf __builtin_ia32_rsqrtps (v4sf)
7096 v4sf __builtin_ia32_sqrtps (v4sf)
7097 v4sf __builtin_ia32_rcpss (v4sf)
7098 v4sf __builtin_ia32_rsqrtss (v4sf)
7099 v4sf __builtin_ia32_sqrtss (v4sf)
7100 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
7101 void __builtin_ia32_movntps (float *, v4sf)
7102 int __builtin_ia32_movmskps (v4sf)
7105 The following built-in functions are available when @option{-msse} is used.
7108 @item v4sf __builtin_ia32_loadaps (float *)
7109 Generates the @code{movaps} machine instruction as a load from memory.
7110 @item void __builtin_ia32_storeaps (float *, v4sf)
7111 Generates the @code{movaps} machine instruction as a store to memory.
7112 @item v4sf __builtin_ia32_loadups (float *)
7113 Generates the @code{movups} machine instruction as a load from memory.
7114 @item void __builtin_ia32_storeups (float *, v4sf)
7115 Generates the @code{movups} machine instruction as a store to memory.
7116 @item v4sf __builtin_ia32_loadsss (float *)
7117 Generates the @code{movss} machine instruction as a load from memory.
7118 @item void __builtin_ia32_storess (float *, v4sf)
7119 Generates the @code{movss} machine instruction as a store to memory.
7120 @item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
7121 Generates the @code{movhps} machine instruction as a load from memory.
7122 @item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
7123 Generates the @code{movlps} machine instruction as a load from memory
7124 @item void __builtin_ia32_storehps (v4sf, v2si *)
7125 Generates the @code{movhps} machine instruction as a store to memory.
7126 @item void __builtin_ia32_storelps (v4sf, v2si *)
7127 Generates the @code{movlps} machine instruction as a store to memory.
7130 The following built-in functions are available when @option{-msse2} is used.
7131 All of them generate the machine instruction that is part of the name.
7134 int __builtin_ia32_comisdeq (v2df, v2df)
7135 int __builtin_ia32_comisdlt (v2df, v2df)
7136 int __builtin_ia32_comisdle (v2df, v2df)
7137 int __builtin_ia32_comisdgt (v2df, v2df)
7138 int __builtin_ia32_comisdge (v2df, v2df)
7139 int __builtin_ia32_comisdneq (v2df, v2df)
7140 int __builtin_ia32_ucomisdeq (v2df, v2df)
7141 int __builtin_ia32_ucomisdlt (v2df, v2df)
7142 int __builtin_ia32_ucomisdle (v2df, v2df)
7143 int __builtin_ia32_ucomisdgt (v2df, v2df)
7144 int __builtin_ia32_ucomisdge (v2df, v2df)
7145 int __builtin_ia32_ucomisdneq (v2df, v2df)
7146 v2df __builtin_ia32_cmpeqpd (v2df, v2df)
7147 v2df __builtin_ia32_cmpltpd (v2df, v2df)
7148 v2df __builtin_ia32_cmplepd (v2df, v2df)
7149 v2df __builtin_ia32_cmpgtpd (v2df, v2df)
7150 v2df __builtin_ia32_cmpgepd (v2df, v2df)
7151 v2df __builtin_ia32_cmpunordpd (v2df, v2df)
7152 v2df __builtin_ia32_cmpneqpd (v2df, v2df)
7153 v2df __builtin_ia32_cmpnltpd (v2df, v2df)
7154 v2df __builtin_ia32_cmpnlepd (v2df, v2df)
7155 v2df __builtin_ia32_cmpngtpd (v2df, v2df)
7156 v2df __builtin_ia32_cmpngepd (v2df, v2df)
7157 v2df __builtin_ia32_cmpordpd (v2df, v2df)
7158 v2df __builtin_ia32_cmpeqsd (v2df, v2df)
7159 v2df __builtin_ia32_cmpltsd (v2df, v2df)
7160 v2df __builtin_ia32_cmplesd (v2df, v2df)
7161 v2df __builtin_ia32_cmpunordsd (v2df, v2df)
7162 v2df __builtin_ia32_cmpneqsd (v2df, v2df)
7163 v2df __builtin_ia32_cmpnltsd (v2df, v2df)
7164 v2df __builtin_ia32_cmpnlesd (v2df, v2df)
7165 v2df __builtin_ia32_cmpordsd (v2df, v2df)
7166 v2di __builtin_ia32_paddq (v2di, v2di)
7167 v2di __builtin_ia32_psubq (v2di, v2di)
7168 v2df __builtin_ia32_addpd (v2df, v2df)
7169 v2df __builtin_ia32_subpd (v2df, v2df)
7170 v2df __builtin_ia32_mulpd (v2df, v2df)
7171 v2df __builtin_ia32_divpd (v2df, v2df)
7172 v2df __builtin_ia32_addsd (v2df, v2df)
7173 v2df __builtin_ia32_subsd (v2df, v2df)
7174 v2df __builtin_ia32_mulsd (v2df, v2df)
7175 v2df __builtin_ia32_divsd (v2df, v2df)
7176 v2df __builtin_ia32_minpd (v2df, v2df)
7177 v2df __builtin_ia32_maxpd (v2df, v2df)
7178 v2df __builtin_ia32_minsd (v2df, v2df)
7179 v2df __builtin_ia32_maxsd (v2df, v2df)
7180 v2df __builtin_ia32_andpd (v2df, v2df)
7181 v2df __builtin_ia32_andnpd (v2df, v2df)
7182 v2df __builtin_ia32_orpd (v2df, v2df)
7183 v2df __builtin_ia32_xorpd (v2df, v2df)
7184 v2df __builtin_ia32_movsd (v2df, v2df)
7185 v2df __builtin_ia32_unpckhpd (v2df, v2df)
7186 v2df __builtin_ia32_unpcklpd (v2df, v2df)
7187 v16qi __builtin_ia32_paddb128 (v16qi, v16qi)
7188 v8hi __builtin_ia32_paddw128 (v8hi, v8hi)
7189 v4si __builtin_ia32_paddd128 (v4si, v4si)
7190 v2di __builtin_ia32_paddq128 (v2di, v2di)
7191 v16qi __builtin_ia32_psubb128 (v16qi, v16qi)
7192 v8hi __builtin_ia32_psubw128 (v8hi, v8hi)
7193 v4si __builtin_ia32_psubd128 (v4si, v4si)
7194 v2di __builtin_ia32_psubq128 (v2di, v2di)
7195 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi)
7196 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi)
7197 v2di __builtin_ia32_pand128 (v2di, v2di)
7198 v2di __builtin_ia32_pandn128 (v2di, v2di)
7199 v2di __builtin_ia32_por128 (v2di, v2di)
7200 v2di __builtin_ia32_pxor128 (v2di, v2di)
7201 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi)
7202 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi)
7203 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi)
7204 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi)
7205 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si)
7206 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi)
7207 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi)
7208 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si)
7209 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi)
7210 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi)
7211 v16qi __builtin_ia32_pminub128 (v16qi, v16qi)
7212 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi)
7213 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi)
7214 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi)
7215 v4si __builtin_ia32_punpckhdq128 (v4si, v4si)
7216 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di)
7217 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi)
7218 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi)
7219 v4si __builtin_ia32_punpckldq128 (v4si, v4si)
7220 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di)
7221 v16qi __builtin_ia32_packsswb128 (v16qi, v16qi)
7222 v8hi __builtin_ia32_packssdw128 (v8hi, v8hi)
7223 v16qi __builtin_ia32_packuswb128 (v16qi, v16qi)
7224 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
7225 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
7226 v2df __builtin_ia32_loadupd (double *)
7227 void __builtin_ia32_storeupd (double *, v2df)
7228 v2df __builtin_ia32_loadhpd (v2df, double *)
7229 v2df __builtin_ia32_loadlpd (v2df, double *)
7230 int __builtin_ia32_movmskpd (v2df)
7231 int __builtin_ia32_pmovmskb128 (v16qi)
7232 void __builtin_ia32_movnti (int *, int)
7233 void __builtin_ia32_movntpd (double *, v2df)
7234 void __builtin_ia32_movntdq (v2df *, v2df)
7235 v4si __builtin_ia32_pshufd (v4si, int)
7236 v8hi __builtin_ia32_pshuflw (v8hi, int)
7237 v8hi __builtin_ia32_pshufhw (v8hi, int)
7238 v2di __builtin_ia32_psadbw128 (v16qi, v16qi)
7239 v2df __builtin_ia32_sqrtpd (v2df)
7240 v2df __builtin_ia32_sqrtsd (v2df)
7241 v2df __builtin_ia32_shufpd (v2df, v2df, int)
7242 v2df __builtin_ia32_cvtdq2pd (v4si)
7243 v4sf __builtin_ia32_cvtdq2ps (v4si)
7244 v4si __builtin_ia32_cvtpd2dq (v2df)
7245 v2si __builtin_ia32_cvtpd2pi (v2df)
7246 v4sf __builtin_ia32_cvtpd2ps (v2df)
7247 v4si __builtin_ia32_cvttpd2dq (v2df)
7248 v2si __builtin_ia32_cvttpd2pi (v2df)
7249 v2df __builtin_ia32_cvtpi2pd (v2si)
7250 int __builtin_ia32_cvtsd2si (v2df)
7251 int __builtin_ia32_cvttsd2si (v2df)
7252 long long __builtin_ia32_cvtsd2si64 (v2df)
7253 long long __builtin_ia32_cvttsd2si64 (v2df)
7254 v4si __builtin_ia32_cvtps2dq (v4sf)
7255 v2df __builtin_ia32_cvtps2pd (v4sf)
7256 v4si __builtin_ia32_cvttps2dq (v4sf)
7257 v2df __builtin_ia32_cvtsi2sd (v2df, int)
7258 v2df __builtin_ia32_cvtsi642sd (v2df, long long)
7259 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df)
7260 v2df __builtin_ia32_cvtss2sd (v2df, v4sf)
7261 void __builtin_ia32_clflush (const void *)
7262 void __builtin_ia32_lfence (void)
7263 void __builtin_ia32_mfence (void)
7264 v16qi __builtin_ia32_loaddqu (const char *)
7265 void __builtin_ia32_storedqu (char *, v16qi)
7266 unsigned long long __builtin_ia32_pmuludq (v2si, v2si)
7267 v2di __builtin_ia32_pmuludq128 (v4si, v4si)
7268 v8hi __builtin_ia32_psllw128 (v8hi, v2di)
7269 v4si __builtin_ia32_pslld128 (v4si, v2di)
7270 v2di __builtin_ia32_psllq128 (v4si, v2di)
7271 v8hi __builtin_ia32_psrlw128 (v8hi, v2di)
7272 v4si __builtin_ia32_psrld128 (v4si, v2di)
7273 v2di __builtin_ia32_psrlq128 (v2di, v2di)
7274 v8hi __builtin_ia32_psraw128 (v8hi, v2di)
7275 v4si __builtin_ia32_psrad128 (v4si, v2di)
7276 v2di __builtin_ia32_pslldqi128 (v2di, int)
7277 v8hi __builtin_ia32_psllwi128 (v8hi, int)
7278 v4si __builtin_ia32_pslldi128 (v4si, int)
7279 v2di __builtin_ia32_psllqi128 (v2di, int)
7280 v2di __builtin_ia32_psrldqi128 (v2di, int)
7281 v8hi __builtin_ia32_psrlwi128 (v8hi, int)
7282 v4si __builtin_ia32_psrldi128 (v4si, int)
7283 v2di __builtin_ia32_psrlqi128 (v2di, int)
7284 v8hi __builtin_ia32_psrawi128 (v8hi, int)
7285 v4si __builtin_ia32_psradi128 (v4si, int)
7286 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi)
7289 The following built-in functions are available when @option{-msse3} is used.
7290 All of them generate the machine instruction that is part of the name.
7293 v2df __builtin_ia32_addsubpd (v2df, v2df)
7294 v4sf __builtin_ia32_addsubps (v4sf, v4sf)
7295 v2df __builtin_ia32_haddpd (v2df, v2df)
7296 v4sf __builtin_ia32_haddps (v4sf, v4sf)
7297 v2df __builtin_ia32_hsubpd (v2df, v2df)
7298 v4sf __builtin_ia32_hsubps (v4sf, v4sf)
7299 v16qi __builtin_ia32_lddqu (char const *)
7300 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
7301 v2df __builtin_ia32_movddup (v2df)
7302 v4sf __builtin_ia32_movshdup (v4sf)
7303 v4sf __builtin_ia32_movsldup (v4sf)
7304 void __builtin_ia32_mwait (unsigned int, unsigned int)
7307 The following built-in functions are available when @option{-msse3} is used.
7310 @item v2df __builtin_ia32_loadddup (double const *)
7311 Generates the @code{movddup} machine instruction as a load from memory.
7314 The following built-in functions are available when @option{-mssse3} is used.
7315 All of them generate the machine instruction that is part of the name
7319 v2si __builtin_ia32_phaddd (v2si, v2si)
7320 v4hi __builtin_ia32_phaddw (v4hi, v4hi)
7321 v4hi __builtin_ia32_phaddsw (v4hi, v4hi)
7322 v2si __builtin_ia32_phsubd (v2si, v2si)
7323 v4hi __builtin_ia32_phsubw (v4hi, v4hi)
7324 v4hi __builtin_ia32_phsubsw (v4hi, v4hi)
7325 v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi)
7326 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi)
7327 v8qi __builtin_ia32_pshufb (v8qi, v8qi)
7328 v8qi __builtin_ia32_psignb (v8qi, v8qi)
7329 v2si __builtin_ia32_psignd (v2si, v2si)
7330 v4hi __builtin_ia32_psignw (v4hi, v4hi)
7331 long long __builtin_ia32_palignr (long long, long long, int)
7332 v8qi __builtin_ia32_pabsb (v8qi)
7333 v2si __builtin_ia32_pabsd (v2si)
7334 v4hi __builtin_ia32_pabsw (v4hi)
7337 The following built-in functions are available when @option{-mssse3} is used.
7338 All of them generate the machine instruction that is part of the name
7342 v4si __builtin_ia32_phaddd128 (v4si, v4si)
7343 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi)
7344 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi)
7345 v4si __builtin_ia32_phsubd128 (v4si, v4si)
7346 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi)
7347 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi)
7348 v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi)
7349 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi)
7350 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi)
7351 v16qi __builtin_ia32_psignb128 (v16qi, v16qi)
7352 v4si __builtin_ia32_psignd128 (v4si, v4si)
7353 v8hi __builtin_ia32_psignw128 (v8hi, v8hi)
7354 v2di __builtin_ia32_palignr (v2di, v2di, int)
7355 v16qi __builtin_ia32_pabsb128 (v16qi)
7356 v4si __builtin_ia32_pabsd128 (v4si)
7357 v8hi __builtin_ia32_pabsw128 (v8hi)
7360 The following built-in functions are available when @option{-msse4a} is used.
7363 void _mm_stream_sd (double*,__m128d);
7364 Generates the @code{movntsd} machine instruction.
7365 void _mm_stream_ss (float*,__m128);
7366 Generates the @code{movntss} machine instruction.
7367 __m128i _mm_extract_si64 (__m128i, __m128i);
7368 Generates the @code{extrq} machine instruction with only SSE register operands.
7369 __m128i _mm_extracti_si64 (__m128i, int, int);
7370 Generates the @code{extrq} machine instruction with SSE register and immediate operands.
7371 __m128i _mm_insert_si64 (__m128i, __m128i);
7372 Generates the @code{insertq} machine instruction with only SSE register operands.
7373 __m128i _mm_inserti_si64 (__m128i, __m128i, int, int);
7374 Generates the @code{insertq} machine instruction with SSE register and immediate operands.
7377 The following built-in functions are available when @option{-m3dnow} is used.
7378 All of them generate the machine instruction that is part of the name.
7381 void __builtin_ia32_femms (void)
7382 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
7383 v2si __builtin_ia32_pf2id (v2sf)
7384 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
7385 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
7386 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
7387 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
7388 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
7389 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
7390 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
7391 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
7392 v2sf __builtin_ia32_pfrcp (v2sf)
7393 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
7394 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
7395 v2sf __builtin_ia32_pfrsqrt (v2sf)
7396 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
7397 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
7398 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
7399 v2sf __builtin_ia32_pi2fd (v2si)
7400 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
7403 The following built-in functions are available when both @option{-m3dnow}
7404 and @option{-march=athlon} are used. All of them generate the machine
7405 instruction that is part of the name.
7408 v2si __builtin_ia32_pf2iw (v2sf)
7409 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
7410 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
7411 v2sf __builtin_ia32_pi2fw (v2si)
7412 v2sf __builtin_ia32_pswapdsf (v2sf)
7413 v2si __builtin_ia32_pswapdsi (v2si)
7416 @node MIPS DSP Built-in Functions
7417 @subsection MIPS DSP Built-in Functions
7419 The MIPS DSP Application-Specific Extension (ASE) includes new
7420 instructions that are designed to improve the performance of DSP and
7421 media applications. It provides instructions that operate on packed
7422 8-bit integer data, Q15 fractional data and Q31 fractional data.
7424 GCC supports MIPS DSP operations using both the generic
7425 vector extensions (@pxref{Vector Extensions}) and a collection of
7426 MIPS-specific built-in functions. Both kinds of support are
7427 enabled by the @option{-mdsp} command-line option.
7429 At present, GCC only provides support for operations on 32-bit
7430 vectors. The vector type associated with 8-bit integer data is
7431 usually called @code{v4i8} and the vector type associated with Q15 is
7432 usually called @code{v2q15}. They can be defined in C as follows:
7435 typedef char v4i8 __attribute__ ((vector_size(4)));
7436 typedef short v2q15 __attribute__ ((vector_size(4)));
7439 @code{v4i8} and @code{v2q15} values are initialized in the same way as
7440 aggregates. For example:
7443 v4i8 a = @{1, 2, 3, 4@};
7445 b = (v4i8) @{5, 6, 7, 8@};
7447 v2q15 c = @{0x0fcb, 0x3a75@};
7449 d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@};
7452 @emph{Note:} The CPU's endianness determines the order in which values
7453 are packed. On little-endian targets, the first value is the least
7454 significant and the last value is the most significant. The opposite
7455 order applies to big-endian targets. For example, the code above will
7456 set the lowest byte of @code{a} to @code{1} on little-endian targets
7457 and @code{4} on big-endian targets.
7459 @emph{Note:} Q15 and Q31 values must be initialized with their integer
7460 representation. As shown in this example, the integer representation
7461 of a Q15 value can be obtained by multiplying the fractional value by
7462 @code{0x1.0p15}. The equivalent for Q31 values is to multiply by
7465 The table below lists the @code{v4i8} and @code{v2q15} operations for which
7466 hardware support exists. @code{a} and @code{b} are @code{v4i8} values,
7467 and @code{c} and @code{d} are @code{v2q15} values.
7469 @multitable @columnfractions .50 .50
7470 @item C code @tab MIPS instruction
7471 @item @code{a + b} @tab @code{addu.qb}
7472 @item @code{c + d} @tab @code{addq.ph}
7473 @item @code{a - b} @tab @code{subu.qb}
7474 @item @code{c - d} @tab @code{subq.ph}
7477 It is easier to describe the DSP built-in functions if we first define
7478 the following types:
7483 typedef long long a64;
7486 @code{q31} and @code{i32} are actually the same as @code{int}, but we
7487 use @code{q31} to indicate a Q31 fractional value and @code{i32} to
7488 indicate a 32-bit integer value. Similarly, @code{a64} is the same as
7489 @code{long long}, but we use @code{a64} to indicate values that will
7490 be placed in one of the four DSP accumulators (@code{$ac0},
7491 @code{$ac1}, @code{$ac2} or @code{$ac3}).
7493 Also, some built-in functions prefer or require immediate numbers as
7494 parameters, because the corresponding DSP instructions accept both immediate
7495 numbers and register operands, or accept immediate numbers only. The
7496 immediate parameters are listed as follows.
7504 imm_n32_31: -32 to 31.
7505 imm_n512_511: -512 to 511.
7508 The following built-in functions map directly to a particular MIPS DSP
7509 instruction. Please refer to the architecture specification
7510 for details on what each instruction does.
7513 v2q15 __builtin_mips_addq_ph (v2q15, v2q15)
7514 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15)
7515 q31 __builtin_mips_addq_s_w (q31, q31)
7516 v4i8 __builtin_mips_addu_qb (v4i8, v4i8)
7517 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8)
7518 v2q15 __builtin_mips_subq_ph (v2q15, v2q15)
7519 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15)
7520 q31 __builtin_mips_subq_s_w (q31, q31)
7521 v4i8 __builtin_mips_subu_qb (v4i8, v4i8)
7522 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8)
7523 i32 __builtin_mips_addsc (i32, i32)
7524 i32 __builtin_mips_addwc (i32, i32)
7525 i32 __builtin_mips_modsub (i32, i32)
7526 i32 __builtin_mips_raddu_w_qb (v4i8)
7527 v2q15 __builtin_mips_absq_s_ph (v2q15)
7528 q31 __builtin_mips_absq_s_w (q31)
7529 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15)
7530 v2q15 __builtin_mips_precrq_ph_w (q31, q31)
7531 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31)
7532 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15)
7533 q31 __builtin_mips_preceq_w_phl (v2q15)
7534 q31 __builtin_mips_preceq_w_phr (v2q15)
7535 v2q15 __builtin_mips_precequ_ph_qbl (v4i8)
7536 v2q15 __builtin_mips_precequ_ph_qbr (v4i8)
7537 v2q15 __builtin_mips_precequ_ph_qbla (v4i8)
7538 v2q15 __builtin_mips_precequ_ph_qbra (v4i8)
7539 v2q15 __builtin_mips_preceu_ph_qbl (v4i8)
7540 v2q15 __builtin_mips_preceu_ph_qbr (v4i8)
7541 v2q15 __builtin_mips_preceu_ph_qbla (v4i8)
7542 v2q15 __builtin_mips_preceu_ph_qbra (v4i8)
7543 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7)
7544 v4i8 __builtin_mips_shll_qb (v4i8, i32)
7545 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15)
7546 v2q15 __builtin_mips_shll_ph (v2q15, i32)
7547 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15)
7548 v2q15 __builtin_mips_shll_s_ph (v2q15, i32)
7549 q31 __builtin_mips_shll_s_w (q31, imm0_31)
7550 q31 __builtin_mips_shll_s_w (q31, i32)
7551 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7)
7552 v4i8 __builtin_mips_shrl_qb (v4i8, i32)
7553 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15)
7554 v2q15 __builtin_mips_shra_ph (v2q15, i32)
7555 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15)
7556 v2q15 __builtin_mips_shra_r_ph (v2q15, i32)
7557 q31 __builtin_mips_shra_r_w (q31, imm0_31)
7558 q31 __builtin_mips_shra_r_w (q31, i32)
7559 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15)
7560 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15)
7561 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15)
7562 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15)
7563 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15)
7564 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8)
7565 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8)
7566 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8)
7567 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8)
7568 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15)
7569 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31)
7570 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15)
7571 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31)
7572 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15)
7573 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15)
7574 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15)
7575 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15)
7576 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15)
7577 i32 __builtin_mips_bitrev (i32)
7578 i32 __builtin_mips_insv (i32, i32)
7579 v4i8 __builtin_mips_repl_qb (imm0_255)
7580 v4i8 __builtin_mips_repl_qb (i32)
7581 v2q15 __builtin_mips_repl_ph (imm_n512_511)
7582 v2q15 __builtin_mips_repl_ph (i32)
7583 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8)
7584 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8)
7585 void __builtin_mips_cmpu_le_qb (v4i8, v4i8)
7586 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8)
7587 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8)
7588 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8)
7589 void __builtin_mips_cmp_eq_ph (v2q15, v2q15)
7590 void __builtin_mips_cmp_lt_ph (v2q15, v2q15)
7591 void __builtin_mips_cmp_le_ph (v2q15, v2q15)
7592 v4i8 __builtin_mips_pick_qb (v4i8, v4i8)
7593 v2q15 __builtin_mips_pick_ph (v2q15, v2q15)
7594 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15)
7595 i32 __builtin_mips_extr_w (a64, imm0_31)
7596 i32 __builtin_mips_extr_w (a64, i32)
7597 i32 __builtin_mips_extr_r_w (a64, imm0_31)
7598 i32 __builtin_mips_extr_s_h (a64, i32)
7599 i32 __builtin_mips_extr_rs_w (a64, imm0_31)
7600 i32 __builtin_mips_extr_rs_w (a64, i32)
7601 i32 __builtin_mips_extr_s_h (a64, imm0_31)
7602 i32 __builtin_mips_extr_r_w (a64, i32)
7603 i32 __builtin_mips_extp (a64, imm0_31)
7604 i32 __builtin_mips_extp (a64, i32)
7605 i32 __builtin_mips_extpdp (a64, imm0_31)
7606 i32 __builtin_mips_extpdp (a64, i32)
7607 a64 __builtin_mips_shilo (a64, imm_n32_31)
7608 a64 __builtin_mips_shilo (a64, i32)
7609 a64 __builtin_mips_mthlip (a64, i32)
7610 void __builtin_mips_wrdsp (i32, imm0_63)
7611 i32 __builtin_mips_rddsp (imm0_63)
7612 i32 __builtin_mips_lbux (void *, i32)
7613 i32 __builtin_mips_lhx (void *, i32)
7614 i32 __builtin_mips_lwx (void *, i32)
7615 i32 __builtin_mips_bposge32 (void)
7618 @node MIPS Paired-Single Support
7619 @subsection MIPS Paired-Single Support
7621 The MIPS64 architecture includes a number of instructions that
7622 operate on pairs of single-precision floating-point values.
7623 Each pair is packed into a 64-bit floating-point register,
7624 with one element being designated the ``upper half'' and
7625 the other being designated the ``lower half''.
7627 GCC supports paired-single operations using both the generic
7628 vector extensions (@pxref{Vector Extensions}) and a collection of
7629 MIPS-specific built-in functions. Both kinds of support are
7630 enabled by the @option{-mpaired-single} command-line option.
7632 The vector type associated with paired-single values is usually
7633 called @code{v2sf}. It can be defined in C as follows:
7636 typedef float v2sf __attribute__ ((vector_size (8)));
7639 @code{v2sf} values are initialized in the same way as aggregates.
7643 v2sf a = @{1.5, 9.1@};
7646 b = (v2sf) @{e, f@};
7649 @emph{Note:} The CPU's endianness determines which value is stored in
7650 the upper half of a register and which value is stored in the lower half.
7651 On little-endian targets, the first value is the lower one and the second
7652 value is the upper one. The opposite order applies to big-endian targets.
7653 For example, the code above will set the lower half of @code{a} to
7654 @code{1.5} on little-endian targets and @code{9.1} on big-endian targets.
7657 * Paired-Single Arithmetic::
7658 * Paired-Single Built-in Functions::
7659 * MIPS-3D Built-in Functions::
7662 @node Paired-Single Arithmetic
7663 @subsubsection Paired-Single Arithmetic
7665 The table below lists the @code{v2sf} operations for which hardware
7666 support exists. @code{a}, @code{b} and @code{c} are @code{v2sf}
7667 values and @code{x} is an integral value.
7669 @multitable @columnfractions .50 .50
7670 @item C code @tab MIPS instruction
7671 @item @code{a + b} @tab @code{add.ps}
7672 @item @code{a - b} @tab @code{sub.ps}
7673 @item @code{-a} @tab @code{neg.ps}
7674 @item @code{a * b} @tab @code{mul.ps}
7675 @item @code{a * b + c} @tab @code{madd.ps}
7676 @item @code{a * b - c} @tab @code{msub.ps}
7677 @item @code{-(a * b + c)} @tab @code{nmadd.ps}
7678 @item @code{-(a * b - c)} @tab @code{nmsub.ps}
7679 @item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps}
7682 Note that the multiply-accumulate instructions can be disabled
7683 using the command-line option @code{-mno-fused-madd}.
7685 @node Paired-Single Built-in Functions
7686 @subsubsection Paired-Single Built-in Functions
7688 The following paired-single functions map directly to a particular
7689 MIPS instruction. Please refer to the architecture specification
7690 for details on what each instruction does.
7693 @item v2sf __builtin_mips_pll_ps (v2sf, v2sf)
7694 Pair lower lower (@code{pll.ps}).
7696 @item v2sf __builtin_mips_pul_ps (v2sf, v2sf)
7697 Pair upper lower (@code{pul.ps}).
7699 @item v2sf __builtin_mips_plu_ps (v2sf, v2sf)
7700 Pair lower upper (@code{plu.ps}).
7702 @item v2sf __builtin_mips_puu_ps (v2sf, v2sf)
7703 Pair upper upper (@code{puu.ps}).
7705 @item v2sf __builtin_mips_cvt_ps_s (float, float)
7706 Convert pair to paired single (@code{cvt.ps.s}).
7708 @item float __builtin_mips_cvt_s_pl (v2sf)
7709 Convert pair lower to single (@code{cvt.s.pl}).
7711 @item float __builtin_mips_cvt_s_pu (v2sf)
7712 Convert pair upper to single (@code{cvt.s.pu}).
7714 @item v2sf __builtin_mips_abs_ps (v2sf)
7715 Absolute value (@code{abs.ps}).
7717 @item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)
7718 Align variable (@code{alnv.ps}).
7720 @emph{Note:} The value of the third parameter must be 0 or 4
7721 modulo 8, otherwise the result will be unpredictable. Please read the
7722 instruction description for details.
7725 The following multi-instruction functions are also available.
7726 In each case, @var{cond} can be any of the 16 floating-point conditions:
7727 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7728 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl},
7729 @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7732 @item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7733 @itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7734 Conditional move based on floating point comparison (@code{c.@var{cond}.ps},
7735 @code{movt.ps}/@code{movf.ps}).
7737 The @code{movt} functions return the value @var{x} computed by:
7740 c.@var{cond}.ps @var{cc},@var{a},@var{b}
7741 mov.ps @var{x},@var{c}
7742 movt.ps @var{x},@var{d},@var{cc}
7745 The @code{movf} functions are similar but use @code{movf.ps} instead
7748 @item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7749 @itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7750 Comparison of two paired-single values (@code{c.@var{cond}.ps},
7751 @code{bc1t}/@code{bc1f}).
7753 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7754 and return either the upper or lower half of the result. For example:
7758 if (__builtin_mips_upper_c_eq_ps (a, b))
7759 upper_halves_are_equal ();
7761 upper_halves_are_unequal ();
7763 if (__builtin_mips_lower_c_eq_ps (a, b))
7764 lower_halves_are_equal ();
7766 lower_halves_are_unequal ();
7770 @node MIPS-3D Built-in Functions
7771 @subsubsection MIPS-3D Built-in Functions
7773 The MIPS-3D Application-Specific Extension (ASE) includes additional
7774 paired-single instructions that are designed to improve the performance
7775 of 3D graphics operations. Support for these instructions is controlled
7776 by the @option{-mips3d} command-line option.
7778 The functions listed below map directly to a particular MIPS-3D
7779 instruction. Please refer to the architecture specification for
7780 more details on what each instruction does.
7783 @item v2sf __builtin_mips_addr_ps (v2sf, v2sf)
7784 Reduction add (@code{addr.ps}).
7786 @item v2sf __builtin_mips_mulr_ps (v2sf, v2sf)
7787 Reduction multiply (@code{mulr.ps}).
7789 @item v2sf __builtin_mips_cvt_pw_ps (v2sf)
7790 Convert paired single to paired word (@code{cvt.pw.ps}).
7792 @item v2sf __builtin_mips_cvt_ps_pw (v2sf)
7793 Convert paired word to paired single (@code{cvt.ps.pw}).
7795 @item float __builtin_mips_recip1_s (float)
7796 @itemx double __builtin_mips_recip1_d (double)
7797 @itemx v2sf __builtin_mips_recip1_ps (v2sf)
7798 Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}).
7800 @item float __builtin_mips_recip2_s (float, float)
7801 @itemx double __builtin_mips_recip2_d (double, double)
7802 @itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf)
7803 Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}).
7805 @item float __builtin_mips_rsqrt1_s (float)
7806 @itemx double __builtin_mips_rsqrt1_d (double)
7807 @itemx v2sf __builtin_mips_rsqrt1_ps (v2sf)
7808 Reduced precision reciprocal square root (sequence step 1)
7809 (@code{rsqrt1.@var{fmt}}).
7811 @item float __builtin_mips_rsqrt2_s (float, float)
7812 @itemx double __builtin_mips_rsqrt2_d (double, double)
7813 @itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)
7814 Reduced precision reciprocal square root (sequence step 2)
7815 (@code{rsqrt2.@var{fmt}}).
7818 The following multi-instruction functions are also available.
7819 In each case, @var{cond} can be any of the 16 floating-point conditions:
7820 @code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult},
7821 @code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq},
7822 @code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}.
7825 @item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b})
7826 @itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b})
7827 Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}},
7828 @code{bc1t}/@code{bc1f}).
7830 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s}
7831 or @code{cabs.@var{cond}.d} and return the result as a boolean value.
7836 if (__builtin_mips_cabs_eq_s (a, b))
7842 @item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7843 @itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7844 Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps},
7845 @code{bc1t}/@code{bc1f}).
7847 These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps}
7848 and return either the upper or lower half of the result. For example:
7852 if (__builtin_mips_upper_cabs_eq_ps (a, b))
7853 upper_halves_are_equal ();
7855 upper_halves_are_unequal ();
7857 if (__builtin_mips_lower_cabs_eq_ps (a, b))
7858 lower_halves_are_equal ();
7860 lower_halves_are_unequal ();
7863 @item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7864 @itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7865 Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps},
7866 @code{movt.ps}/@code{movf.ps}).
7868 The @code{movt} functions return the value @var{x} computed by:
7871 cabs.@var{cond}.ps @var{cc},@var{a},@var{b}
7872 mov.ps @var{x},@var{c}
7873 movt.ps @var{x},@var{d},@var{cc}
7876 The @code{movf} functions are similar but use @code{movf.ps} instead
7879 @item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7880 @itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7881 @itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7882 @itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b})
7883 Comparison of two paired-single values
7884 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7885 @code{bc1any2t}/@code{bc1any2f}).
7887 These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps}
7888 or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either
7889 result is true and the @code{all} forms return true if both results are true.
7894 if (__builtin_mips_any_c_eq_ps (a, b))
7899 if (__builtin_mips_all_c_eq_ps (a, b))
7905 @item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7906 @itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7907 @itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7908 @itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d})
7909 Comparison of four paired-single values
7910 (@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps},
7911 @code{bc1any4t}/@code{bc1any4f}).
7913 These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps}
7914 to compare @var{a} with @var{b} and to compare @var{c} with @var{d}.
7915 The @code{any} forms return true if any of the four results are true
7916 and the @code{all} forms return true if all four results are true.
7921 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
7926 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
7933 @node PowerPC AltiVec Built-in Functions
7934 @subsection PowerPC AltiVec Built-in Functions
7936 GCC provides an interface for the PowerPC family of processors to access
7937 the AltiVec operations described in Motorola's AltiVec Programming
7938 Interface Manual. The interface is made available by including
7939 @code{<altivec.h>} and using @option{-maltivec} and
7940 @option{-mabi=altivec}. The interface supports the following vector
7944 vector unsigned char
7948 vector unsigned short
7959 GCC's implementation of the high-level language interface available from
7960 C and C++ code differs from Motorola's documentation in several ways.
7965 A vector constant is a list of constant expressions within curly braces.
7968 A vector initializer requires no cast if the vector constant is of the
7969 same type as the variable it is initializing.
7972 If @code{signed} or @code{unsigned} is omitted, the signedness of the
7973 vector type is the default signedness of the base type. The default
7974 varies depending on the operating system, so a portable program should
7975 always specify the signedness.
7978 Compiling with @option{-maltivec} adds keywords @code{__vector},
7979 @code{__pixel}, and @code{__bool}. Macros @option{vector},
7980 @code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
7984 GCC allows using a @code{typedef} name as the type specifier for a
7988 For C, overloaded functions are implemented with macros so the following
7992 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
7995 Since @code{vec_add} is a macro, the vector constant in the example
7996 is treated as four separate arguments. Wrap the entire argument in
7997 parentheses for this to work.
8000 @emph{Note:} Only the @code{<altivec.h>} interface is supported.
8001 Internally, GCC uses built-in functions to achieve the functionality in
8002 the aforementioned header file, but they are not supported and are
8003 subject to change without notice.
8005 The following interfaces are supported for the generic and specific
8006 AltiVec operations and the AltiVec predicates. In cases where there
8007 is a direct mapping between generic and specific operations, only the
8008 generic names are shown here, although the specific operations can also
8011 Arguments that are documented as @code{const int} require literal
8012 integral values within the range required for that operation.
8015 vector signed char vec_abs (vector signed char);
8016 vector signed short vec_abs (vector signed short);
8017 vector signed int vec_abs (vector signed int);
8018 vector float vec_abs (vector float);
8020 vector signed char vec_abss (vector signed char);
8021 vector signed short vec_abss (vector signed short);
8022 vector signed int vec_abss (vector signed int);
8024 vector signed char vec_add (vector bool char, vector signed char);
8025 vector signed char vec_add (vector signed char, vector bool char);
8026 vector signed char vec_add (vector signed char, vector signed char);
8027 vector unsigned char vec_add (vector bool char, vector unsigned char);
8028 vector unsigned char vec_add (vector unsigned char, vector bool char);
8029 vector unsigned char vec_add (vector unsigned char,
8030 vector unsigned char);
8031 vector signed short vec_add (vector bool short, vector signed short);
8032 vector signed short vec_add (vector signed short, vector bool short);
8033 vector signed short vec_add (vector signed short, vector signed short);
8034 vector unsigned short vec_add (vector bool short,
8035 vector unsigned short);
8036 vector unsigned short vec_add (vector unsigned short,
8038 vector unsigned short vec_add (vector unsigned short,
8039 vector unsigned short);
8040 vector signed int vec_add (vector bool int, vector signed int);
8041 vector signed int vec_add (vector signed int, vector bool int);
8042 vector signed int vec_add (vector signed int, vector signed int);
8043 vector unsigned int vec_add (vector bool int, vector unsigned int);
8044 vector unsigned int vec_add (vector unsigned int, vector bool int);
8045 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
8046 vector float vec_add (vector float, vector float);
8048 vector float vec_vaddfp (vector float, vector float);
8050 vector signed int vec_vadduwm (vector bool int, vector signed int);
8051 vector signed int vec_vadduwm (vector signed int, vector bool int);
8052 vector signed int vec_vadduwm (vector signed int, vector signed int);
8053 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
8054 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
8055 vector unsigned int vec_vadduwm (vector unsigned int,
8056 vector unsigned int);
8058 vector signed short vec_vadduhm (vector bool short,
8059 vector signed short);
8060 vector signed short vec_vadduhm (vector signed short,
8062 vector signed short vec_vadduhm (vector signed short,
8063 vector signed short);
8064 vector unsigned short vec_vadduhm (vector bool short,
8065 vector unsigned short);
8066 vector unsigned short vec_vadduhm (vector unsigned short,
8068 vector unsigned short vec_vadduhm (vector unsigned short,
8069 vector unsigned short);
8071 vector signed char vec_vaddubm (vector bool char, vector signed char);
8072 vector signed char vec_vaddubm (vector signed char, vector bool char);
8073 vector signed char vec_vaddubm (vector signed char, vector signed char);
8074 vector unsigned char vec_vaddubm (vector bool char,
8075 vector unsigned char);
8076 vector unsigned char vec_vaddubm (vector unsigned char,
8078 vector unsigned char vec_vaddubm (vector unsigned char,
8079 vector unsigned char);
8081 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
8083 vector unsigned char vec_adds (vector bool char, vector unsigned char);
8084 vector unsigned char vec_adds (vector unsigned char, vector bool char);
8085 vector unsigned char vec_adds (vector unsigned char,
8086 vector unsigned char);
8087 vector signed char vec_adds (vector bool char, vector signed char);
8088 vector signed char vec_adds (vector signed char, vector bool char);
8089 vector signed char vec_adds (vector signed char, vector signed char);
8090 vector unsigned short vec_adds (vector bool short,
8091 vector unsigned short);
8092 vector unsigned short vec_adds (vector unsigned short,
8094 vector unsigned short vec_adds (vector unsigned short,
8095 vector unsigned short);
8096 vector signed short vec_adds (vector bool short, vector signed short);
8097 vector signed short vec_adds (vector signed short, vector bool short);
8098 vector signed short vec_adds (vector signed short, vector signed short);
8099 vector unsigned int vec_adds (vector bool int, vector unsigned int);
8100 vector unsigned int vec_adds (vector unsigned int, vector bool int);
8101 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
8102 vector signed int vec_adds (vector bool int, vector signed int);
8103 vector signed int vec_adds (vector signed int, vector bool int);
8104 vector signed int vec_adds (vector signed int, vector signed int);
8106 vector signed int vec_vaddsws (vector bool int, vector signed int);
8107 vector signed int vec_vaddsws (vector signed int, vector bool int);
8108 vector signed int vec_vaddsws (vector signed int, vector signed int);
8110 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
8111 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
8112 vector unsigned int vec_vadduws (vector unsigned int,
8113 vector unsigned int);
8115 vector signed short vec_vaddshs (vector bool short,
8116 vector signed short);
8117 vector signed short vec_vaddshs (vector signed short,
8119 vector signed short vec_vaddshs (vector signed short,
8120 vector signed short);
8122 vector unsigned short vec_vadduhs (vector bool short,
8123 vector unsigned short);
8124 vector unsigned short vec_vadduhs (vector unsigned short,
8126 vector unsigned short vec_vadduhs (vector unsigned short,
8127 vector unsigned short);
8129 vector signed char vec_vaddsbs (vector bool char, vector signed char);
8130 vector signed char vec_vaddsbs (vector signed char, vector bool char);
8131 vector signed char vec_vaddsbs (vector signed char, vector signed char);
8133 vector unsigned char vec_vaddubs (vector bool char,
8134 vector unsigned char);
8135 vector unsigned char vec_vaddubs (vector unsigned char,
8137 vector unsigned char vec_vaddubs (vector unsigned char,
8138 vector unsigned char);
8140 vector float vec_and (vector float, vector float);
8141 vector float vec_and (vector float, vector bool int);
8142 vector float vec_and (vector bool int, vector float);
8143 vector bool int vec_and (vector bool int, vector bool int);
8144 vector signed int vec_and (vector bool int, vector signed int);
8145 vector signed int vec_and (vector signed int, vector bool int);
8146 vector signed int vec_and (vector signed int, vector signed int);
8147 vector unsigned int vec_and (vector bool int, vector unsigned int);
8148 vector unsigned int vec_and (vector unsigned int, vector bool int);
8149 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
8150 vector bool short vec_and (vector bool short, vector bool short);
8151 vector signed short vec_and (vector bool short, vector signed short);
8152 vector signed short vec_and (vector signed short, vector bool short);
8153 vector signed short vec_and (vector signed short, vector signed short);
8154 vector unsigned short vec_and (vector bool short,
8155 vector unsigned short);
8156 vector unsigned short vec_and (vector unsigned short,
8158 vector unsigned short vec_and (vector unsigned short,
8159 vector unsigned short);
8160 vector signed char vec_and (vector bool char, vector signed char);
8161 vector bool char vec_and (vector bool char, vector bool char);
8162 vector signed char vec_and (vector signed char, vector bool char);
8163 vector signed char vec_and (vector signed char, vector signed char);
8164 vector unsigned char vec_and (vector bool char, vector unsigned char);
8165 vector unsigned char vec_and (vector unsigned char, vector bool char);
8166 vector unsigned char vec_and (vector unsigned char,
8167 vector unsigned char);
8169 vector float vec_andc (vector float, vector float);
8170 vector float vec_andc (vector float, vector bool int);
8171 vector float vec_andc (vector bool int, vector float);
8172 vector bool int vec_andc (vector bool int, vector bool int);
8173 vector signed int vec_andc (vector bool int, vector signed int);
8174 vector signed int vec_andc (vector signed int, vector bool int);
8175 vector signed int vec_andc (vector signed int, vector signed int);
8176 vector unsigned int vec_andc (vector bool int, vector unsigned int);
8177 vector unsigned int vec_andc (vector unsigned int, vector bool int);
8178 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
8179 vector bool short vec_andc (vector bool short, vector bool short);
8180 vector signed short vec_andc (vector bool short, vector signed short);
8181 vector signed short vec_andc (vector signed short, vector bool short);
8182 vector signed short vec_andc (vector signed short, vector signed short);
8183 vector unsigned short vec_andc (vector bool short,
8184 vector unsigned short);
8185 vector unsigned short vec_andc (vector unsigned short,
8187 vector unsigned short vec_andc (vector unsigned short,
8188 vector unsigned short);
8189 vector signed char vec_andc (vector bool char, vector signed char);
8190 vector bool char vec_andc (vector bool char, vector bool char);
8191 vector signed char vec_andc (vector signed char, vector bool char);
8192 vector signed char vec_andc (vector signed char, vector signed char);
8193 vector unsigned char vec_andc (vector bool char, vector unsigned char);
8194 vector unsigned char vec_andc (vector unsigned char, vector bool char);
8195 vector unsigned char vec_andc (vector unsigned char,
8196 vector unsigned char);
8198 vector unsigned char vec_avg (vector unsigned char,
8199 vector unsigned char);
8200 vector signed char vec_avg (vector signed char, vector signed char);
8201 vector unsigned short vec_avg (vector unsigned short,
8202 vector unsigned short);
8203 vector signed short vec_avg (vector signed short, vector signed short);
8204 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
8205 vector signed int vec_avg (vector signed int, vector signed int);
8207 vector signed int vec_vavgsw (vector signed int, vector signed int);
8209 vector unsigned int vec_vavguw (vector unsigned int,
8210 vector unsigned int);
8212 vector signed short vec_vavgsh (vector signed short,
8213 vector signed short);
8215 vector unsigned short vec_vavguh (vector unsigned short,
8216 vector unsigned short);
8218 vector signed char vec_vavgsb (vector signed char, vector signed char);
8220 vector unsigned char vec_vavgub (vector unsigned char,
8221 vector unsigned char);
8223 vector float vec_ceil (vector float);
8225 vector signed int vec_cmpb (vector float, vector float);
8227 vector bool char vec_cmpeq (vector signed char, vector signed char);
8228 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
8229 vector bool short vec_cmpeq (vector signed short, vector signed short);
8230 vector bool short vec_cmpeq (vector unsigned short,
8231 vector unsigned short);
8232 vector bool int vec_cmpeq (vector signed int, vector signed int);
8233 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
8234 vector bool int vec_cmpeq (vector float, vector float);
8236 vector bool int vec_vcmpeqfp (vector float, vector float);
8238 vector bool int vec_vcmpequw (vector signed int, vector signed int);
8239 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
8241 vector bool short vec_vcmpequh (vector signed short,
8242 vector signed short);
8243 vector bool short vec_vcmpequh (vector unsigned short,
8244 vector unsigned short);
8246 vector bool char vec_vcmpequb (vector signed char, vector signed char);
8247 vector bool char vec_vcmpequb (vector unsigned char,
8248 vector unsigned char);
8250 vector bool int vec_cmpge (vector float, vector float);
8252 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
8253 vector bool char vec_cmpgt (vector signed char, vector signed char);
8254 vector bool short vec_cmpgt (vector unsigned short,
8255 vector unsigned short);
8256 vector bool short vec_cmpgt (vector signed short, vector signed short);
8257 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
8258 vector bool int vec_cmpgt (vector signed int, vector signed int);
8259 vector bool int vec_cmpgt (vector float, vector float);
8261 vector bool int vec_vcmpgtfp (vector float, vector float);
8263 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
8265 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
8267 vector bool short vec_vcmpgtsh (vector signed short,
8268 vector signed short);
8270 vector bool short vec_vcmpgtuh (vector unsigned short,
8271 vector unsigned short);
8273 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
8275 vector bool char vec_vcmpgtub (vector unsigned char,
8276 vector unsigned char);
8278 vector bool int vec_cmple (vector float, vector float);
8280 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
8281 vector bool char vec_cmplt (vector signed char, vector signed char);
8282 vector bool short vec_cmplt (vector unsigned short,
8283 vector unsigned short);
8284 vector bool short vec_cmplt (vector signed short, vector signed short);
8285 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
8286 vector bool int vec_cmplt (vector signed int, vector signed int);
8287 vector bool int vec_cmplt (vector float, vector float);
8289 vector float vec_ctf (vector unsigned int, const int);
8290 vector float vec_ctf (vector signed int, const int);
8292 vector float vec_vcfsx (vector signed int, const int);
8294 vector float vec_vcfux (vector unsigned int, const int);
8296 vector signed int vec_cts (vector float, const int);
8298 vector unsigned int vec_ctu (vector float, const int);
8300 void vec_dss (const int);
8302 void vec_dssall (void);
8304 void vec_dst (const vector unsigned char *, int, const int);
8305 void vec_dst (const vector signed char *, int, const int);
8306 void vec_dst (const vector bool char *, int, const int);
8307 void vec_dst (const vector unsigned short *, int, const int);
8308 void vec_dst (const vector signed short *, int, const int);
8309 void vec_dst (const vector bool short *, int, const int);
8310 void vec_dst (const vector pixel *, int, const int);
8311 void vec_dst (const vector unsigned int *, int, const int);
8312 void vec_dst (const vector signed int *, int, const int);
8313 void vec_dst (const vector bool int *, int, const int);
8314 void vec_dst (const vector float *, int, const int);
8315 void vec_dst (const unsigned char *, int, const int);
8316 void vec_dst (const signed char *, int, const int);
8317 void vec_dst (const unsigned short *, int, const int);
8318 void vec_dst (const short *, int, const int);
8319 void vec_dst (const unsigned int *, int, const int);
8320 void vec_dst (const int *, int, const int);
8321 void vec_dst (const unsigned long *, int, const int);
8322 void vec_dst (const long *, int, const int);
8323 void vec_dst (const float *, int, const int);
8325 void vec_dstst (const vector unsigned char *, int, const int);
8326 void vec_dstst (const vector signed char *, int, const int);
8327 void vec_dstst (const vector bool char *, int, const int);
8328 void vec_dstst (const vector unsigned short *, int, const int);
8329 void vec_dstst (const vector signed short *, int, const int);
8330 void vec_dstst (const vector bool short *, int, const int);
8331 void vec_dstst (const vector pixel *, int, const int);
8332 void vec_dstst (const vector unsigned int *, int, const int);
8333 void vec_dstst (const vector signed int *, int, const int);
8334 void vec_dstst (const vector bool int *, int, const int);
8335 void vec_dstst (const vector float *, int, const int);
8336 void vec_dstst (const unsigned char *, int, const int);
8337 void vec_dstst (const signed char *, int, const int);
8338 void vec_dstst (const unsigned short *, int, const int);
8339 void vec_dstst (const short *, int, const int);
8340 void vec_dstst (const unsigned int *, int, const int);
8341 void vec_dstst (const int *, int, const int);
8342 void vec_dstst (const unsigned long *, int, const int);
8343 void vec_dstst (const long *, int, const int);
8344 void vec_dstst (const float *, int, const int);
8346 void vec_dststt (const vector unsigned char *, int, const int);
8347 void vec_dststt (const vector signed char *, int, const int);
8348 void vec_dststt (const vector bool char *, int, const int);
8349 void vec_dststt (const vector unsigned short *, int, const int);
8350 void vec_dststt (const vector signed short *, int, const int);
8351 void vec_dststt (const vector bool short *, int, const int);
8352 void vec_dststt (const vector pixel *, int, const int);
8353 void vec_dststt (const vector unsigned int *, int, const int);
8354 void vec_dststt (const vector signed int *, int, const int);
8355 void vec_dststt (const vector bool int *, int, const int);
8356 void vec_dststt (const vector float *, int, const int);
8357 void vec_dststt (const unsigned char *, int, const int);
8358 void vec_dststt (const signed char *, int, const int);
8359 void vec_dststt (const unsigned short *, int, const int);
8360 void vec_dststt (const short *, int, const int);
8361 void vec_dststt (const unsigned int *, int, const int);
8362 void vec_dststt (const int *, int, const int);
8363 void vec_dststt (const unsigned long *, int, const int);
8364 void vec_dststt (const long *, int, const int);
8365 void vec_dststt (const float *, int, const int);
8367 void vec_dstt (const vector unsigned char *, int, const int);
8368 void vec_dstt (const vector signed char *, int, const int);
8369 void vec_dstt (const vector bool char *, int, const int);
8370 void vec_dstt (const vector unsigned short *, int, const int);
8371 void vec_dstt (const vector signed short *, int, const int);
8372 void vec_dstt (const vector bool short *, int, const int);
8373 void vec_dstt (const vector pixel *, int, const int);
8374 void vec_dstt (const vector unsigned int *, int, const int);
8375 void vec_dstt (const vector signed int *, int, const int);
8376 void vec_dstt (const vector bool int *, int, const int);
8377 void vec_dstt (const vector float *, int, const int);
8378 void vec_dstt (const unsigned char *, int, const int);
8379 void vec_dstt (const signed char *, int, const int);
8380 void vec_dstt (const unsigned short *, int, const int);
8381 void vec_dstt (const short *, int, const int);
8382 void vec_dstt (const unsigned int *, int, const int);
8383 void vec_dstt (const int *, int, const int);
8384 void vec_dstt (const unsigned long *, int, const int);
8385 void vec_dstt (const long *, int, const int);
8386 void vec_dstt (const float *, int, const int);
8388 vector float vec_expte (vector float);
8390 vector float vec_floor (vector float);
8392 vector float vec_ld (int, const vector float *);
8393 vector float vec_ld (int, const float *);
8394 vector bool int vec_ld (int, const vector bool int *);
8395 vector signed int vec_ld (int, const vector signed int *);
8396 vector signed int vec_ld (int, const int *);
8397 vector signed int vec_ld (int, const long *);
8398 vector unsigned int vec_ld (int, const vector unsigned int *);
8399 vector unsigned int vec_ld (int, const unsigned int *);
8400 vector unsigned int vec_ld (int, const unsigned long *);
8401 vector bool short vec_ld (int, const vector bool short *);
8402 vector pixel vec_ld (int, const vector pixel *);
8403 vector signed short vec_ld (int, const vector signed short *);
8404 vector signed short vec_ld (int, const short *);
8405 vector unsigned short vec_ld (int, const vector unsigned short *);
8406 vector unsigned short vec_ld (int, const unsigned short *);
8407 vector bool char vec_ld (int, const vector bool char *);
8408 vector signed char vec_ld (int, const vector signed char *);
8409 vector signed char vec_ld (int, const signed char *);
8410 vector unsigned char vec_ld (int, const vector unsigned char *);
8411 vector unsigned char vec_ld (int, const unsigned char *);
8413 vector signed char vec_lde (int, const signed char *);
8414 vector unsigned char vec_lde (int, const unsigned char *);
8415 vector signed short vec_lde (int, const short *);
8416 vector unsigned short vec_lde (int, const unsigned short *);
8417 vector float vec_lde (int, const float *);
8418 vector signed int vec_lde (int, const int *);
8419 vector unsigned int vec_lde (int, const unsigned int *);
8420 vector signed int vec_lde (int, const long *);
8421 vector unsigned int vec_lde (int, const unsigned long *);
8423 vector float vec_lvewx (int, float *);
8424 vector signed int vec_lvewx (int, int *);
8425 vector unsigned int vec_lvewx (int, unsigned int *);
8426 vector signed int vec_lvewx (int, long *);
8427 vector unsigned int vec_lvewx (int, unsigned long *);
8429 vector signed short vec_lvehx (int, short *);
8430 vector unsigned short vec_lvehx (int, unsigned short *);
8432 vector signed char vec_lvebx (int, char *);
8433 vector unsigned char vec_lvebx (int, unsigned char *);
8435 vector float vec_ldl (int, const vector float *);
8436 vector float vec_ldl (int, const float *);
8437 vector bool int vec_ldl (int, const vector bool int *);
8438 vector signed int vec_ldl (int, const vector signed int *);
8439 vector signed int vec_ldl (int, const int *);
8440 vector signed int vec_ldl (int, const long *);
8441 vector unsigned int vec_ldl (int, const vector unsigned int *);
8442 vector unsigned int vec_ldl (int, const unsigned int *);
8443 vector unsigned int vec_ldl (int, const unsigned long *);
8444 vector bool short vec_ldl (int, const vector bool short *);
8445 vector pixel vec_ldl (int, const vector pixel *);
8446 vector signed short vec_ldl (int, const vector signed short *);
8447 vector signed short vec_ldl (int, const short *);
8448 vector unsigned short vec_ldl (int, const vector unsigned short *);
8449 vector unsigned short vec_ldl (int, const unsigned short *);
8450 vector bool char vec_ldl (int, const vector bool char *);
8451 vector signed char vec_ldl (int, const vector signed char *);
8452 vector signed char vec_ldl (int, const signed char *);
8453 vector unsigned char vec_ldl (int, const vector unsigned char *);
8454 vector unsigned char vec_ldl (int, const unsigned char *);
8456 vector float vec_loge (vector float);
8458 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
8459 vector unsigned char vec_lvsl (int, const volatile signed char *);
8460 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
8461 vector unsigned char vec_lvsl (int, const volatile short *);
8462 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
8463 vector unsigned char vec_lvsl (int, const volatile int *);
8464 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
8465 vector unsigned char vec_lvsl (int, const volatile long *);
8466 vector unsigned char vec_lvsl (int, const volatile float *);
8468 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
8469 vector unsigned char vec_lvsr (int, const volatile signed char *);
8470 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
8471 vector unsigned char vec_lvsr (int, const volatile short *);
8472 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
8473 vector unsigned char vec_lvsr (int, const volatile int *);
8474 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
8475 vector unsigned char vec_lvsr (int, const volatile long *);
8476 vector unsigned char vec_lvsr (int, const volatile float *);
8478 vector float vec_madd (vector float, vector float, vector float);
8480 vector signed short vec_madds (vector signed short,
8481 vector signed short,
8482 vector signed short);
8484 vector unsigned char vec_max (vector bool char, vector unsigned char);
8485 vector unsigned char vec_max (vector unsigned char, vector bool char);
8486 vector unsigned char vec_max (vector unsigned char,
8487 vector unsigned char);
8488 vector signed char vec_max (vector bool char, vector signed char);
8489 vector signed char vec_max (vector signed char, vector bool char);
8490 vector signed char vec_max (vector signed char, vector signed char);
8491 vector unsigned short vec_max (vector bool short,
8492 vector unsigned short);
8493 vector unsigned short vec_max (vector unsigned short,
8495 vector unsigned short vec_max (vector unsigned short,
8496 vector unsigned short);
8497 vector signed short vec_max (vector bool short, vector signed short);
8498 vector signed short vec_max (vector signed short, vector bool short);
8499 vector signed short vec_max (vector signed short, vector signed short);
8500 vector unsigned int vec_max (vector bool int, vector unsigned int);
8501 vector unsigned int vec_max (vector unsigned int, vector bool int);
8502 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
8503 vector signed int vec_max (vector bool int, vector signed int);
8504 vector signed int vec_max (vector signed int, vector bool int);
8505 vector signed int vec_max (vector signed int, vector signed int);
8506 vector float vec_max (vector float, vector float);
8508 vector float vec_vmaxfp (vector float, vector float);
8510 vector signed int vec_vmaxsw (vector bool int, vector signed int);
8511 vector signed int vec_vmaxsw (vector signed int, vector bool int);
8512 vector signed int vec_vmaxsw (vector signed int, vector signed int);
8514 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
8515 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
8516 vector unsigned int vec_vmaxuw (vector unsigned int,
8517 vector unsigned int);
8519 vector signed short vec_vmaxsh (vector bool short, vector signed short);
8520 vector signed short vec_vmaxsh (vector signed short, vector bool short);
8521 vector signed short vec_vmaxsh (vector signed short,
8522 vector signed short);
8524 vector unsigned short vec_vmaxuh (vector bool short,
8525 vector unsigned short);
8526 vector unsigned short vec_vmaxuh (vector unsigned short,
8528 vector unsigned short vec_vmaxuh (vector unsigned short,
8529 vector unsigned short);
8531 vector signed char vec_vmaxsb (vector bool char, vector signed char);
8532 vector signed char vec_vmaxsb (vector signed char, vector bool char);
8533 vector signed char vec_vmaxsb (vector signed char, vector signed char);
8535 vector unsigned char vec_vmaxub (vector bool char,
8536 vector unsigned char);
8537 vector unsigned char vec_vmaxub (vector unsigned char,
8539 vector unsigned char vec_vmaxub (vector unsigned char,
8540 vector unsigned char);
8542 vector bool char vec_mergeh (vector bool char, vector bool char);
8543 vector signed char vec_mergeh (vector signed char, vector signed char);
8544 vector unsigned char vec_mergeh (vector unsigned char,
8545 vector unsigned char);
8546 vector bool short vec_mergeh (vector bool short, vector bool short);
8547 vector pixel vec_mergeh (vector pixel, vector pixel);
8548 vector signed short vec_mergeh (vector signed short,
8549 vector signed short);
8550 vector unsigned short vec_mergeh (vector unsigned short,
8551 vector unsigned short);
8552 vector float vec_mergeh (vector float, vector float);
8553 vector bool int vec_mergeh (vector bool int, vector bool int);
8554 vector signed int vec_mergeh (vector signed int, vector signed int);
8555 vector unsigned int vec_mergeh (vector unsigned int,
8556 vector unsigned int);
8558 vector float vec_vmrghw (vector float, vector float);
8559 vector bool int vec_vmrghw (vector bool int, vector bool int);
8560 vector signed int vec_vmrghw (vector signed int, vector signed int);
8561 vector unsigned int vec_vmrghw (vector unsigned int,
8562 vector unsigned int);
8564 vector bool short vec_vmrghh (vector bool short, vector bool short);
8565 vector signed short vec_vmrghh (vector signed short,
8566 vector signed short);
8567 vector unsigned short vec_vmrghh (vector unsigned short,
8568 vector unsigned short);
8569 vector pixel vec_vmrghh (vector pixel, vector pixel);
8571 vector bool char vec_vmrghb (vector bool char, vector bool char);
8572 vector signed char vec_vmrghb (vector signed char, vector signed char);
8573 vector unsigned char vec_vmrghb (vector unsigned char,
8574 vector unsigned char);
8576 vector bool char vec_mergel (vector bool char, vector bool char);
8577 vector signed char vec_mergel (vector signed char, vector signed char);
8578 vector unsigned char vec_mergel (vector unsigned char,
8579 vector unsigned char);
8580 vector bool short vec_mergel (vector bool short, vector bool short);
8581 vector pixel vec_mergel (vector pixel, vector pixel);
8582 vector signed short vec_mergel (vector signed short,
8583 vector signed short);
8584 vector unsigned short vec_mergel (vector unsigned short,
8585 vector unsigned short);
8586 vector float vec_mergel (vector float, vector float);
8587 vector bool int vec_mergel (vector bool int, vector bool int);
8588 vector signed int vec_mergel (vector signed int, vector signed int);
8589 vector unsigned int vec_mergel (vector unsigned int,
8590 vector unsigned int);
8592 vector float vec_vmrglw (vector float, vector float);
8593 vector signed int vec_vmrglw (vector signed int, vector signed int);
8594 vector unsigned int vec_vmrglw (vector unsigned int,
8595 vector unsigned int);
8596 vector bool int vec_vmrglw (vector bool int, vector bool int);
8598 vector bool short vec_vmrglh (vector bool short, vector bool short);
8599 vector signed short vec_vmrglh (vector signed short,
8600 vector signed short);
8601 vector unsigned short vec_vmrglh (vector unsigned short,
8602 vector unsigned short);
8603 vector pixel vec_vmrglh (vector pixel, vector pixel);
8605 vector bool char vec_vmrglb (vector bool char, vector bool char);
8606 vector signed char vec_vmrglb (vector signed char, vector signed char);
8607 vector unsigned char vec_vmrglb (vector unsigned char,
8608 vector unsigned char);
8610 vector unsigned short vec_mfvscr (void);
8612 vector unsigned char vec_min (vector bool char, vector unsigned char);
8613 vector unsigned char vec_min (vector unsigned char, vector bool char);
8614 vector unsigned char vec_min (vector unsigned char,
8615 vector unsigned char);
8616 vector signed char vec_min (vector bool char, vector signed char);
8617 vector signed char vec_min (vector signed char, vector bool char);
8618 vector signed char vec_min (vector signed char, vector signed char);
8619 vector unsigned short vec_min (vector bool short,
8620 vector unsigned short);
8621 vector unsigned short vec_min (vector unsigned short,
8623 vector unsigned short vec_min (vector unsigned short,
8624 vector unsigned short);
8625 vector signed short vec_min (vector bool short, vector signed short);
8626 vector signed short vec_min (vector signed short, vector bool short);
8627 vector signed short vec_min (vector signed short, vector signed short);
8628 vector unsigned int vec_min (vector bool int, vector unsigned int);
8629 vector unsigned int vec_min (vector unsigned int, vector bool int);
8630 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
8631 vector signed int vec_min (vector bool int, vector signed int);
8632 vector signed int vec_min (vector signed int, vector bool int);
8633 vector signed int vec_min (vector signed int, vector signed int);
8634 vector float vec_min (vector float, vector float);
8636 vector float vec_vminfp (vector float, vector float);
8638 vector signed int vec_vminsw (vector bool int, vector signed int);
8639 vector signed int vec_vminsw (vector signed int, vector bool int);
8640 vector signed int vec_vminsw (vector signed int, vector signed int);
8642 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
8643 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
8644 vector unsigned int vec_vminuw (vector unsigned int,
8645 vector unsigned int);
8647 vector signed short vec_vminsh (vector bool short, vector signed short);
8648 vector signed short vec_vminsh (vector signed short, vector bool short);
8649 vector signed short vec_vminsh (vector signed short,
8650 vector signed short);
8652 vector unsigned short vec_vminuh (vector bool short,
8653 vector unsigned short);
8654 vector unsigned short vec_vminuh (vector unsigned short,
8656 vector unsigned short vec_vminuh (vector unsigned short,
8657 vector unsigned short);
8659 vector signed char vec_vminsb (vector bool char, vector signed char);
8660 vector signed char vec_vminsb (vector signed char, vector bool char);
8661 vector signed char vec_vminsb (vector signed char, vector signed char);
8663 vector unsigned char vec_vminub (vector bool char,
8664 vector unsigned char);
8665 vector unsigned char vec_vminub (vector unsigned char,
8667 vector unsigned char vec_vminub (vector unsigned char,
8668 vector unsigned char);
8670 vector signed short vec_mladd (vector signed short,
8671 vector signed short,
8672 vector signed short);
8673 vector signed short vec_mladd (vector signed short,
8674 vector unsigned short,
8675 vector unsigned short);
8676 vector signed short vec_mladd (vector unsigned short,
8677 vector signed short,
8678 vector signed short);
8679 vector unsigned short vec_mladd (vector unsigned short,
8680 vector unsigned short,
8681 vector unsigned short);
8683 vector signed short vec_mradds (vector signed short,
8684 vector signed short,
8685 vector signed short);
8687 vector unsigned int vec_msum (vector unsigned char,
8688 vector unsigned char,
8689 vector unsigned int);
8690 vector signed int vec_msum (vector signed char,
8691 vector unsigned char,
8693 vector unsigned int vec_msum (vector unsigned short,
8694 vector unsigned short,
8695 vector unsigned int);
8696 vector signed int vec_msum (vector signed short,
8697 vector signed short,
8700 vector signed int vec_vmsumshm (vector signed short,
8701 vector signed short,
8704 vector unsigned int vec_vmsumuhm (vector unsigned short,
8705 vector unsigned short,
8706 vector unsigned int);
8708 vector signed int vec_vmsummbm (vector signed char,
8709 vector unsigned char,
8712 vector unsigned int vec_vmsumubm (vector unsigned char,
8713 vector unsigned char,
8714 vector unsigned int);
8716 vector unsigned int vec_msums (vector unsigned short,
8717 vector unsigned short,
8718 vector unsigned int);
8719 vector signed int vec_msums (vector signed short,
8720 vector signed short,
8723 vector signed int vec_vmsumshs (vector signed short,
8724 vector signed short,
8727 vector unsigned int vec_vmsumuhs (vector unsigned short,
8728 vector unsigned short,
8729 vector unsigned int);
8731 void vec_mtvscr (vector signed int);
8732 void vec_mtvscr (vector unsigned int);
8733 void vec_mtvscr (vector bool int);
8734 void vec_mtvscr (vector signed short);
8735 void vec_mtvscr (vector unsigned short);
8736 void vec_mtvscr (vector bool short);
8737 void vec_mtvscr (vector pixel);
8738 void vec_mtvscr (vector signed char);
8739 void vec_mtvscr (vector unsigned char);
8740 void vec_mtvscr (vector bool char);
8742 vector unsigned short vec_mule (vector unsigned char,
8743 vector unsigned char);
8744 vector signed short vec_mule (vector signed char,
8745 vector signed char);
8746 vector unsigned int vec_mule (vector unsigned short,
8747 vector unsigned short);
8748 vector signed int vec_mule (vector signed short, vector signed short);
8750 vector signed int vec_vmulesh (vector signed short,
8751 vector signed short);
8753 vector unsigned int vec_vmuleuh (vector unsigned short,
8754 vector unsigned short);
8756 vector signed short vec_vmulesb (vector signed char,
8757 vector signed char);
8759 vector unsigned short vec_vmuleub (vector unsigned char,
8760 vector unsigned char);
8762 vector unsigned short vec_mulo (vector unsigned char,
8763 vector unsigned char);
8764 vector signed short vec_mulo (vector signed char, vector signed char);
8765 vector unsigned int vec_mulo (vector unsigned short,
8766 vector unsigned short);
8767 vector signed int vec_mulo (vector signed short, vector signed short);
8769 vector signed int vec_vmulosh (vector signed short,
8770 vector signed short);
8772 vector unsigned int vec_vmulouh (vector unsigned short,
8773 vector unsigned short);
8775 vector signed short vec_vmulosb (vector signed char,
8776 vector signed char);
8778 vector unsigned short vec_vmuloub (vector unsigned char,
8779 vector unsigned char);
8781 vector float vec_nmsub (vector float, vector float, vector float);
8783 vector float vec_nor (vector float, vector float);
8784 vector signed int vec_nor (vector signed int, vector signed int);
8785 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
8786 vector bool int vec_nor (vector bool int, vector bool int);
8787 vector signed short vec_nor (vector signed short, vector signed short);
8788 vector unsigned short vec_nor (vector unsigned short,
8789 vector unsigned short);
8790 vector bool short vec_nor (vector bool short, vector bool short);
8791 vector signed char vec_nor (vector signed char, vector signed char);
8792 vector unsigned char vec_nor (vector unsigned char,
8793 vector unsigned char);
8794 vector bool char vec_nor (vector bool char, vector bool char);
8796 vector float vec_or (vector float, vector float);
8797 vector float vec_or (vector float, vector bool int);
8798 vector float vec_or (vector bool int, vector float);
8799 vector bool int vec_or (vector bool int, vector bool int);
8800 vector signed int vec_or (vector bool int, vector signed int);
8801 vector signed int vec_or (vector signed int, vector bool int);
8802 vector signed int vec_or (vector signed int, vector signed int);
8803 vector unsigned int vec_or (vector bool int, vector unsigned int);
8804 vector unsigned int vec_or (vector unsigned int, vector bool int);
8805 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
8806 vector bool short vec_or (vector bool short, vector bool short);
8807 vector signed short vec_or (vector bool short, vector signed short);
8808 vector signed short vec_or (vector signed short, vector bool short);
8809 vector signed short vec_or (vector signed short, vector signed short);
8810 vector unsigned short vec_or (vector bool short, vector unsigned short);
8811 vector unsigned short vec_or (vector unsigned short, vector bool short);
8812 vector unsigned short vec_or (vector unsigned short,
8813 vector unsigned short);
8814 vector signed char vec_or (vector bool char, vector signed char);
8815 vector bool char vec_or (vector bool char, vector bool char);
8816 vector signed char vec_or (vector signed char, vector bool char);
8817 vector signed char vec_or (vector signed char, vector signed char);
8818 vector unsigned char vec_or (vector bool char, vector unsigned char);
8819 vector unsigned char vec_or (vector unsigned char, vector bool char);
8820 vector unsigned char vec_or (vector unsigned char,
8821 vector unsigned char);
8823 vector signed char vec_pack (vector signed short, vector signed short);
8824 vector unsigned char vec_pack (vector unsigned short,
8825 vector unsigned short);
8826 vector bool char vec_pack (vector bool short, vector bool short);
8827 vector signed short vec_pack (vector signed int, vector signed int);
8828 vector unsigned short vec_pack (vector unsigned int,
8829 vector unsigned int);
8830 vector bool short vec_pack (vector bool int, vector bool int);
8832 vector bool short vec_vpkuwum (vector bool int, vector bool int);
8833 vector signed short vec_vpkuwum (vector signed int, vector signed int);
8834 vector unsigned short vec_vpkuwum (vector unsigned int,
8835 vector unsigned int);
8837 vector bool char vec_vpkuhum (vector bool short, vector bool short);
8838 vector signed char vec_vpkuhum (vector signed short,
8839 vector signed short);
8840 vector unsigned char vec_vpkuhum (vector unsigned short,
8841 vector unsigned short);
8843 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
8845 vector unsigned char vec_packs (vector unsigned short,
8846 vector unsigned short);
8847 vector signed char vec_packs (vector signed short, vector signed short);
8848 vector unsigned short vec_packs (vector unsigned int,
8849 vector unsigned int);
8850 vector signed short vec_packs (vector signed int, vector signed int);
8852 vector signed short vec_vpkswss (vector signed int, vector signed int);
8854 vector unsigned short vec_vpkuwus (vector unsigned int,
8855 vector unsigned int);
8857 vector signed char vec_vpkshss (vector signed short,
8858 vector signed short);
8860 vector unsigned char vec_vpkuhus (vector unsigned short,
8861 vector unsigned short);
8863 vector unsigned char vec_packsu (vector unsigned short,
8864 vector unsigned short);
8865 vector unsigned char vec_packsu (vector signed short,
8866 vector signed short);
8867 vector unsigned short vec_packsu (vector unsigned int,
8868 vector unsigned int);
8869 vector unsigned short vec_packsu (vector signed int, vector signed int);
8871 vector unsigned short vec_vpkswus (vector signed int,
8874 vector unsigned char vec_vpkshus (vector signed short,
8875 vector signed short);
8877 vector float vec_perm (vector float,
8879 vector unsigned char);
8880 vector signed int vec_perm (vector signed int,
8882 vector unsigned char);
8883 vector unsigned int vec_perm (vector unsigned int,
8884 vector unsigned int,
8885 vector unsigned char);
8886 vector bool int vec_perm (vector bool int,
8888 vector unsigned char);
8889 vector signed short vec_perm (vector signed short,
8890 vector signed short,
8891 vector unsigned char);
8892 vector unsigned short vec_perm (vector unsigned short,
8893 vector unsigned short,
8894 vector unsigned char);
8895 vector bool short vec_perm (vector bool short,
8897 vector unsigned char);
8898 vector pixel vec_perm (vector pixel,
8900 vector unsigned char);
8901 vector signed char vec_perm (vector signed char,
8903 vector unsigned char);
8904 vector unsigned char vec_perm (vector unsigned char,
8905 vector unsigned char,
8906 vector unsigned char);
8907 vector bool char vec_perm (vector bool char,
8909 vector unsigned char);
8911 vector float vec_re (vector float);
8913 vector signed char vec_rl (vector signed char,
8914 vector unsigned char);
8915 vector unsigned char vec_rl (vector unsigned char,
8916 vector unsigned char);
8917 vector signed short vec_rl (vector signed short, vector unsigned short);
8918 vector unsigned short vec_rl (vector unsigned short,
8919 vector unsigned short);
8920 vector signed int vec_rl (vector signed int, vector unsigned int);
8921 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
8923 vector signed int vec_vrlw (vector signed int, vector unsigned int);
8924 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
8926 vector signed short vec_vrlh (vector signed short,
8927 vector unsigned short);
8928 vector unsigned short vec_vrlh (vector unsigned short,
8929 vector unsigned short);
8931 vector signed char vec_vrlb (vector signed char, vector unsigned char);
8932 vector unsigned char vec_vrlb (vector unsigned char,
8933 vector unsigned char);
8935 vector float vec_round (vector float);
8937 vector float vec_rsqrte (vector float);
8939 vector float vec_sel (vector float, vector float, vector bool int);
8940 vector float vec_sel (vector float, vector float, vector unsigned int);
8941 vector signed int vec_sel (vector signed int,
8944 vector signed int vec_sel (vector signed int,
8946 vector unsigned int);
8947 vector unsigned int vec_sel (vector unsigned int,
8948 vector unsigned int,
8950 vector unsigned int vec_sel (vector unsigned int,
8951 vector unsigned int,
8952 vector unsigned int);
8953 vector bool int vec_sel (vector bool int,
8956 vector bool int vec_sel (vector bool int,
8958 vector unsigned int);
8959 vector signed short vec_sel (vector signed short,
8960 vector signed short,
8962 vector signed short vec_sel (vector signed short,
8963 vector signed short,
8964 vector unsigned short);
8965 vector unsigned short vec_sel (vector unsigned short,
8966 vector unsigned short,
8968 vector unsigned short vec_sel (vector unsigned short,
8969 vector unsigned short,
8970 vector unsigned short);
8971 vector bool short vec_sel (vector bool short,
8974 vector bool short vec_sel (vector bool short,
8976 vector unsigned short);
8977 vector signed char vec_sel (vector signed char,
8980 vector signed char vec_sel (vector signed char,
8982 vector unsigned char);
8983 vector unsigned char vec_sel (vector unsigned char,
8984 vector unsigned char,
8986 vector unsigned char vec_sel (vector unsigned char,
8987 vector unsigned char,
8988 vector unsigned char);
8989 vector bool char vec_sel (vector bool char,
8992 vector bool char vec_sel (vector bool char,
8994 vector unsigned char);
8996 vector signed char vec_sl (vector signed char,
8997 vector unsigned char);
8998 vector unsigned char vec_sl (vector unsigned char,
8999 vector unsigned char);
9000 vector signed short vec_sl (vector signed short, vector unsigned short);
9001 vector unsigned short vec_sl (vector unsigned short,
9002 vector unsigned short);
9003 vector signed int vec_sl (vector signed int, vector unsigned int);
9004 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
9006 vector signed int vec_vslw (vector signed int, vector unsigned int);
9007 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
9009 vector signed short vec_vslh (vector signed short,
9010 vector unsigned short);
9011 vector unsigned short vec_vslh (vector unsigned short,
9012 vector unsigned short);
9014 vector signed char vec_vslb (vector signed char, vector unsigned char);
9015 vector unsigned char vec_vslb (vector unsigned char,
9016 vector unsigned char);
9018 vector float vec_sld (vector float, vector float, const int);
9019 vector signed int vec_sld (vector signed int,
9022 vector unsigned int vec_sld (vector unsigned int,
9023 vector unsigned int,
9025 vector bool int vec_sld (vector bool int,
9028 vector signed short vec_sld (vector signed short,
9029 vector signed short,
9031 vector unsigned short vec_sld (vector unsigned short,
9032 vector unsigned short,
9034 vector bool short vec_sld (vector bool short,
9037 vector pixel vec_sld (vector pixel,
9040 vector signed char vec_sld (vector signed char,
9043 vector unsigned char vec_sld (vector unsigned char,
9044 vector unsigned char,
9046 vector bool char vec_sld (vector bool char,
9050 vector signed int vec_sll (vector signed int,
9051 vector unsigned int);
9052 vector signed int vec_sll (vector signed int,
9053 vector unsigned short);
9054 vector signed int vec_sll (vector signed int,
9055 vector unsigned char);
9056 vector unsigned int vec_sll (vector unsigned int,
9057 vector unsigned int);
9058 vector unsigned int vec_sll (vector unsigned int,
9059 vector unsigned short);
9060 vector unsigned int vec_sll (vector unsigned int,
9061 vector unsigned char);
9062 vector bool int vec_sll (vector bool int,
9063 vector unsigned int);
9064 vector bool int vec_sll (vector bool int,
9065 vector unsigned short);
9066 vector bool int vec_sll (vector bool int,
9067 vector unsigned char);
9068 vector signed short vec_sll (vector signed short,
9069 vector unsigned int);
9070 vector signed short vec_sll (vector signed short,
9071 vector unsigned short);
9072 vector signed short vec_sll (vector signed short,
9073 vector unsigned char);
9074 vector unsigned short vec_sll (vector unsigned short,
9075 vector unsigned int);
9076 vector unsigned short vec_sll (vector unsigned short,
9077 vector unsigned short);
9078 vector unsigned short vec_sll (vector unsigned short,
9079 vector unsigned char);
9080 vector bool short vec_sll (vector bool short, vector unsigned int);
9081 vector bool short vec_sll (vector bool short, vector unsigned short);
9082 vector bool short vec_sll (vector bool short, vector unsigned char);
9083 vector pixel vec_sll (vector pixel, vector unsigned int);
9084 vector pixel vec_sll (vector pixel, vector unsigned short);
9085 vector pixel vec_sll (vector pixel, vector unsigned char);
9086 vector signed char vec_sll (vector signed char, vector unsigned int);
9087 vector signed char vec_sll (vector signed char, vector unsigned short);
9088 vector signed char vec_sll (vector signed char, vector unsigned char);
9089 vector unsigned char vec_sll (vector unsigned char,
9090 vector unsigned int);
9091 vector unsigned char vec_sll (vector unsigned char,
9092 vector unsigned short);
9093 vector unsigned char vec_sll (vector unsigned char,
9094 vector unsigned char);
9095 vector bool char vec_sll (vector bool char, vector unsigned int);
9096 vector bool char vec_sll (vector bool char, vector unsigned short);
9097 vector bool char vec_sll (vector bool char, vector unsigned char);
9099 vector float vec_slo (vector float, vector signed char);
9100 vector float vec_slo (vector float, vector unsigned char);
9101 vector signed int vec_slo (vector signed int, vector signed char);
9102 vector signed int vec_slo (vector signed int, vector unsigned char);
9103 vector unsigned int vec_slo (vector unsigned int, vector signed char);
9104 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
9105 vector signed short vec_slo (vector signed short, vector signed char);
9106 vector signed short vec_slo (vector signed short, vector unsigned char);
9107 vector unsigned short vec_slo (vector unsigned short,
9108 vector signed char);
9109 vector unsigned short vec_slo (vector unsigned short,
9110 vector unsigned char);
9111 vector pixel vec_slo (vector pixel, vector signed char);
9112 vector pixel vec_slo (vector pixel, vector unsigned char);
9113 vector signed char vec_slo (vector signed char, vector signed char);
9114 vector signed char vec_slo (vector signed char, vector unsigned char);
9115 vector unsigned char vec_slo (vector unsigned char, vector signed char);
9116 vector unsigned char vec_slo (vector unsigned char,
9117 vector unsigned char);
9119 vector signed char vec_splat (vector signed char, const int);
9120 vector unsigned char vec_splat (vector unsigned char, const int);
9121 vector bool char vec_splat (vector bool char, const int);
9122 vector signed short vec_splat (vector signed short, const int);
9123 vector unsigned short vec_splat (vector unsigned short, const int);
9124 vector bool short vec_splat (vector bool short, const int);
9125 vector pixel vec_splat (vector pixel, const int);
9126 vector float vec_splat (vector float, const int);
9127 vector signed int vec_splat (vector signed int, const int);
9128 vector unsigned int vec_splat (vector unsigned int, const int);
9129 vector bool int vec_splat (vector bool int, const int);
9131 vector float vec_vspltw (vector float, const int);
9132 vector signed int vec_vspltw (vector signed int, const int);
9133 vector unsigned int vec_vspltw (vector unsigned int, const int);
9134 vector bool int vec_vspltw (vector bool int, const int);
9136 vector bool short vec_vsplth (vector bool short, const int);
9137 vector signed short vec_vsplth (vector signed short, const int);
9138 vector unsigned short vec_vsplth (vector unsigned short, const int);
9139 vector pixel vec_vsplth (vector pixel, const int);
9141 vector signed char vec_vspltb (vector signed char, const int);
9142 vector unsigned char vec_vspltb (vector unsigned char, const int);
9143 vector bool char vec_vspltb (vector bool char, const int);
9145 vector signed char vec_splat_s8 (const int);
9147 vector signed short vec_splat_s16 (const int);
9149 vector signed int vec_splat_s32 (const int);
9151 vector unsigned char vec_splat_u8 (const int);
9153 vector unsigned short vec_splat_u16 (const int);
9155 vector unsigned int vec_splat_u32 (const int);
9157 vector signed char vec_sr (vector signed char, vector unsigned char);
9158 vector unsigned char vec_sr (vector unsigned char,
9159 vector unsigned char);
9160 vector signed short vec_sr (vector signed short,
9161 vector unsigned short);
9162 vector unsigned short vec_sr (vector unsigned short,
9163 vector unsigned short);
9164 vector signed int vec_sr (vector signed int, vector unsigned int);
9165 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
9167 vector signed int vec_vsrw (vector signed int, vector unsigned int);
9168 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
9170 vector signed short vec_vsrh (vector signed short,
9171 vector unsigned short);
9172 vector unsigned short vec_vsrh (vector unsigned short,
9173 vector unsigned short);
9175 vector signed char vec_vsrb (vector signed char, vector unsigned char);
9176 vector unsigned char vec_vsrb (vector unsigned char,
9177 vector unsigned char);
9179 vector signed char vec_sra (vector signed char, vector unsigned char);
9180 vector unsigned char vec_sra (vector unsigned char,
9181 vector unsigned char);
9182 vector signed short vec_sra (vector signed short,
9183 vector unsigned short);
9184 vector unsigned short vec_sra (vector unsigned short,
9185 vector unsigned short);
9186 vector signed int vec_sra (vector signed int, vector unsigned int);
9187 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
9189 vector signed int vec_vsraw (vector signed int, vector unsigned int);
9190 vector unsigned int vec_vsraw (vector unsigned int,
9191 vector unsigned int);
9193 vector signed short vec_vsrah (vector signed short,
9194 vector unsigned short);
9195 vector unsigned short vec_vsrah (vector unsigned short,
9196 vector unsigned short);
9198 vector signed char vec_vsrab (vector signed char, vector unsigned char);
9199 vector unsigned char vec_vsrab (vector unsigned char,
9200 vector unsigned char);
9202 vector signed int vec_srl (vector signed int, vector unsigned int);
9203 vector signed int vec_srl (vector signed int, vector unsigned short);
9204 vector signed int vec_srl (vector signed int, vector unsigned char);
9205 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
9206 vector unsigned int vec_srl (vector unsigned int,
9207 vector unsigned short);
9208 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
9209 vector bool int vec_srl (vector bool int, vector unsigned int);
9210 vector bool int vec_srl (vector bool int, vector unsigned short);
9211 vector bool int vec_srl (vector bool int, vector unsigned char);
9212 vector signed short vec_srl (vector signed short, vector unsigned int);
9213 vector signed short vec_srl (vector signed short,
9214 vector unsigned short);
9215 vector signed short vec_srl (vector signed short, vector unsigned char);
9216 vector unsigned short vec_srl (vector unsigned short,
9217 vector unsigned int);
9218 vector unsigned short vec_srl (vector unsigned short,
9219 vector unsigned short);
9220 vector unsigned short vec_srl (vector unsigned short,
9221 vector unsigned char);
9222 vector bool short vec_srl (vector bool short, vector unsigned int);
9223 vector bool short vec_srl (vector bool short, vector unsigned short);
9224 vector bool short vec_srl (vector bool short, vector unsigned char);
9225 vector pixel vec_srl (vector pixel, vector unsigned int);
9226 vector pixel vec_srl (vector pixel, vector unsigned short);
9227 vector pixel vec_srl (vector pixel, vector unsigned char);
9228 vector signed char vec_srl (vector signed char, vector unsigned int);
9229 vector signed char vec_srl (vector signed char, vector unsigned short);
9230 vector signed char vec_srl (vector signed char, vector unsigned char);
9231 vector unsigned char vec_srl (vector unsigned char,
9232 vector unsigned int);
9233 vector unsigned char vec_srl (vector unsigned char,
9234 vector unsigned short);
9235 vector unsigned char vec_srl (vector unsigned char,
9236 vector unsigned char);
9237 vector bool char vec_srl (vector bool char, vector unsigned int);
9238 vector bool char vec_srl (vector bool char, vector unsigned short);
9239 vector bool char vec_srl (vector bool char, vector unsigned char);
9241 vector float vec_sro (vector float, vector signed char);
9242 vector float vec_sro (vector float, vector unsigned char);
9243 vector signed int vec_sro (vector signed int, vector signed char);
9244 vector signed int vec_sro (vector signed int, vector unsigned char);
9245 vector unsigned int vec_sro (vector unsigned int, vector signed char);
9246 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
9247 vector signed short vec_sro (vector signed short, vector signed char);
9248 vector signed short vec_sro (vector signed short, vector unsigned char);
9249 vector unsigned short vec_sro (vector unsigned short,
9250 vector signed char);
9251 vector unsigned short vec_sro (vector unsigned short,
9252 vector unsigned char);
9253 vector pixel vec_sro (vector pixel, vector signed char);
9254 vector pixel vec_sro (vector pixel, vector unsigned char);
9255 vector signed char vec_sro (vector signed char, vector signed char);
9256 vector signed char vec_sro (vector signed char, vector unsigned char);
9257 vector unsigned char vec_sro (vector unsigned char, vector signed char);
9258 vector unsigned char vec_sro (vector unsigned char,
9259 vector unsigned char);
9261 void vec_st (vector float, int, vector float *);
9262 void vec_st (vector float, int, float *);
9263 void vec_st (vector signed int, int, vector signed int *);
9264 void vec_st (vector signed int, int, int *);
9265 void vec_st (vector unsigned int, int, vector unsigned int *);
9266 void vec_st (vector unsigned int, int, unsigned int *);
9267 void vec_st (vector bool int, int, vector bool int *);
9268 void vec_st (vector bool int, int, unsigned int *);
9269 void vec_st (vector bool int, int, int *);
9270 void vec_st (vector signed short, int, vector signed short *);
9271 void vec_st (vector signed short, int, short *);
9272 void vec_st (vector unsigned short, int, vector unsigned short *);
9273 void vec_st (vector unsigned short, int, unsigned short *);
9274 void vec_st (vector bool short, int, vector bool short *);
9275 void vec_st (vector bool short, int, unsigned short *);
9276 void vec_st (vector pixel, int, vector pixel *);
9277 void vec_st (vector pixel, int, unsigned short *);
9278 void vec_st (vector pixel, int, short *);
9279 void vec_st (vector bool short, int, short *);
9280 void vec_st (vector signed char, int, vector signed char *);
9281 void vec_st (vector signed char, int, signed char *);
9282 void vec_st (vector unsigned char, int, vector unsigned char *);
9283 void vec_st (vector unsigned char, int, unsigned char *);
9284 void vec_st (vector bool char, int, vector bool char *);
9285 void vec_st (vector bool char, int, unsigned char *);
9286 void vec_st (vector bool char, int, signed char *);
9288 void vec_ste (vector signed char, int, signed char *);
9289 void vec_ste (vector unsigned char, int, unsigned char *);
9290 void vec_ste (vector bool char, int, signed char *);
9291 void vec_ste (vector bool char, int, unsigned char *);
9292 void vec_ste (vector signed short, int, short *);
9293 void vec_ste (vector unsigned short, int, unsigned short *);
9294 void vec_ste (vector bool short, int, short *);
9295 void vec_ste (vector bool short, int, unsigned short *);
9296 void vec_ste (vector pixel, int, short *);
9297 void vec_ste (vector pixel, int, unsigned short *);
9298 void vec_ste (vector float, int, float *);
9299 void vec_ste (vector signed int, int, int *);
9300 void vec_ste (vector unsigned int, int, unsigned int *);
9301 void vec_ste (vector bool int, int, int *);
9302 void vec_ste (vector bool int, int, unsigned int *);
9304 void vec_stvewx (vector float, int, float *);
9305 void vec_stvewx (vector signed int, int, int *);
9306 void vec_stvewx (vector unsigned int, int, unsigned int *);
9307 void vec_stvewx (vector bool int, int, int *);
9308 void vec_stvewx (vector bool int, int, unsigned int *);
9310 void vec_stvehx (vector signed short, int, short *);
9311 void vec_stvehx (vector unsigned short, int, unsigned short *);
9312 void vec_stvehx (vector bool short, int, short *);
9313 void vec_stvehx (vector bool short, int, unsigned short *);
9314 void vec_stvehx (vector pixel, int, short *);
9315 void vec_stvehx (vector pixel, int, unsigned short *);
9317 void vec_stvebx (vector signed char, int, signed char *);
9318 void vec_stvebx (vector unsigned char, int, unsigned char *);
9319 void vec_stvebx (vector bool char, int, signed char *);
9320 void vec_stvebx (vector bool char, int, unsigned char *);
9322 void vec_stl (vector float, int, vector float *);
9323 void vec_stl (vector float, int, float *);
9324 void vec_stl (vector signed int, int, vector signed int *);
9325 void vec_stl (vector signed int, int, int *);
9326 void vec_stl (vector unsigned int, int, vector unsigned int *);
9327 void vec_stl (vector unsigned int, int, unsigned int *);
9328 void vec_stl (vector bool int, int, vector bool int *);
9329 void vec_stl (vector bool int, int, unsigned int *);
9330 void vec_stl (vector bool int, int, int *);
9331 void vec_stl (vector signed short, int, vector signed short *);
9332 void vec_stl (vector signed short, int, short *);
9333 void vec_stl (vector unsigned short, int, vector unsigned short *);
9334 void vec_stl (vector unsigned short, int, unsigned short *);
9335 void vec_stl (vector bool short, int, vector bool short *);
9336 void vec_stl (vector bool short, int, unsigned short *);
9337 void vec_stl (vector bool short, int, short *);
9338 void vec_stl (vector pixel, int, vector pixel *);
9339 void vec_stl (vector pixel, int, unsigned short *);
9340 void vec_stl (vector pixel, int, short *);
9341 void vec_stl (vector signed char, int, vector signed char *);
9342 void vec_stl (vector signed char, int, signed char *);
9343 void vec_stl (vector unsigned char, int, vector unsigned char *);
9344 void vec_stl (vector unsigned char, int, unsigned char *);
9345 void vec_stl (vector bool char, int, vector bool char *);
9346 void vec_stl (vector bool char, int, unsigned char *);
9347 void vec_stl (vector bool char, int, signed char *);
9349 vector signed char vec_sub (vector bool char, vector signed char);
9350 vector signed char vec_sub (vector signed char, vector bool char);
9351 vector signed char vec_sub (vector signed char, vector signed char);
9352 vector unsigned char vec_sub (vector bool char, vector unsigned char);
9353 vector unsigned char vec_sub (vector unsigned char, vector bool char);
9354 vector unsigned char vec_sub (vector unsigned char,
9355 vector unsigned char);
9356 vector signed short vec_sub (vector bool short, vector signed short);
9357 vector signed short vec_sub (vector signed short, vector bool short);
9358 vector signed short vec_sub (vector signed short, vector signed short);
9359 vector unsigned short vec_sub (vector bool short,
9360 vector unsigned short);
9361 vector unsigned short vec_sub (vector unsigned short,
9363 vector unsigned short vec_sub (vector unsigned short,
9364 vector unsigned short);
9365 vector signed int vec_sub (vector bool int, vector signed int);
9366 vector signed int vec_sub (vector signed int, vector bool int);
9367 vector signed int vec_sub (vector signed int, vector signed int);
9368 vector unsigned int vec_sub (vector bool int, vector unsigned int);
9369 vector unsigned int vec_sub (vector unsigned int, vector bool int);
9370 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
9371 vector float vec_sub (vector float, vector float);
9373 vector float vec_vsubfp (vector float, vector float);
9375 vector signed int vec_vsubuwm (vector bool int, vector signed int);
9376 vector signed int vec_vsubuwm (vector signed int, vector bool int);
9377 vector signed int vec_vsubuwm (vector signed int, vector signed int);
9378 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
9379 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
9380 vector unsigned int vec_vsubuwm (vector unsigned int,
9381 vector unsigned int);
9383 vector signed short vec_vsubuhm (vector bool short,
9384 vector signed short);
9385 vector signed short vec_vsubuhm (vector signed short,
9387 vector signed short vec_vsubuhm (vector signed short,
9388 vector signed short);
9389 vector unsigned short vec_vsubuhm (vector bool short,
9390 vector unsigned short);
9391 vector unsigned short vec_vsubuhm (vector unsigned short,
9393 vector unsigned short vec_vsubuhm (vector unsigned short,
9394 vector unsigned short);
9396 vector signed char vec_vsububm (vector bool char, vector signed char);
9397 vector signed char vec_vsububm (vector signed char, vector bool char);
9398 vector signed char vec_vsububm (vector signed char, vector signed char);
9399 vector unsigned char vec_vsububm (vector bool char,
9400 vector unsigned char);
9401 vector unsigned char vec_vsububm (vector unsigned char,
9403 vector unsigned char vec_vsububm (vector unsigned char,
9404 vector unsigned char);
9406 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
9408 vector unsigned char vec_subs (vector bool char, vector unsigned char);
9409 vector unsigned char vec_subs (vector unsigned char, vector bool char);
9410 vector unsigned char vec_subs (vector unsigned char,
9411 vector unsigned char);
9412 vector signed char vec_subs (vector bool char, vector signed char);
9413 vector signed char vec_subs (vector signed char, vector bool char);
9414 vector signed char vec_subs (vector signed char, vector signed char);
9415 vector unsigned short vec_subs (vector bool short,
9416 vector unsigned short);
9417 vector unsigned short vec_subs (vector unsigned short,
9419 vector unsigned short vec_subs (vector unsigned short,
9420 vector unsigned short);
9421 vector signed short vec_subs (vector bool short, vector signed short);
9422 vector signed short vec_subs (vector signed short, vector bool short);
9423 vector signed short vec_subs (vector signed short, vector signed short);
9424 vector unsigned int vec_subs (vector bool int, vector unsigned int);
9425 vector unsigned int vec_subs (vector unsigned int, vector bool int);
9426 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
9427 vector signed int vec_subs (vector bool int, vector signed int);
9428 vector signed int vec_subs (vector signed int, vector bool int);
9429 vector signed int vec_subs (vector signed int, vector signed int);
9431 vector signed int vec_vsubsws (vector bool int, vector signed int);
9432 vector signed int vec_vsubsws (vector signed int, vector bool int);
9433 vector signed int vec_vsubsws (vector signed int, vector signed int);
9435 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
9436 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
9437 vector unsigned int vec_vsubuws (vector unsigned int,
9438 vector unsigned int);
9440 vector signed short vec_vsubshs (vector bool short,
9441 vector signed short);
9442 vector signed short vec_vsubshs (vector signed short,
9444 vector signed short vec_vsubshs (vector signed short,
9445 vector signed short);
9447 vector unsigned short vec_vsubuhs (vector bool short,
9448 vector unsigned short);
9449 vector unsigned short vec_vsubuhs (vector unsigned short,
9451 vector unsigned short vec_vsubuhs (vector unsigned short,
9452 vector unsigned short);
9454 vector signed char vec_vsubsbs (vector bool char, vector signed char);
9455 vector signed char vec_vsubsbs (vector signed char, vector bool char);
9456 vector signed char vec_vsubsbs (vector signed char, vector signed char);
9458 vector unsigned char vec_vsububs (vector bool char,
9459 vector unsigned char);
9460 vector unsigned char vec_vsububs (vector unsigned char,
9462 vector unsigned char vec_vsububs (vector unsigned char,
9463 vector unsigned char);
9465 vector unsigned int vec_sum4s (vector unsigned char,
9466 vector unsigned int);
9467 vector signed int vec_sum4s (vector signed char, vector signed int);
9468 vector signed int vec_sum4s (vector signed short, vector signed int);
9470 vector signed int vec_vsum4shs (vector signed short, vector signed int);
9472 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
9474 vector unsigned int vec_vsum4ubs (vector unsigned char,
9475 vector unsigned int);
9477 vector signed int vec_sum2s (vector signed int, vector signed int);
9479 vector signed int vec_sums (vector signed int, vector signed int);
9481 vector float vec_trunc (vector float);
9483 vector signed short vec_unpackh (vector signed char);
9484 vector bool short vec_unpackh (vector bool char);
9485 vector signed int vec_unpackh (vector signed short);
9486 vector bool int vec_unpackh (vector bool short);
9487 vector unsigned int vec_unpackh (vector pixel);
9489 vector bool int vec_vupkhsh (vector bool short);
9490 vector signed int vec_vupkhsh (vector signed short);
9492 vector unsigned int vec_vupkhpx (vector pixel);
9494 vector bool short vec_vupkhsb (vector bool char);
9495 vector signed short vec_vupkhsb (vector signed char);
9497 vector signed short vec_unpackl (vector signed char);
9498 vector bool short vec_unpackl (vector bool char);
9499 vector unsigned int vec_unpackl (vector pixel);
9500 vector signed int vec_unpackl (vector signed short);
9501 vector bool int vec_unpackl (vector bool short);
9503 vector unsigned int vec_vupklpx (vector pixel);
9505 vector bool int vec_vupklsh (vector bool short);
9506 vector signed int vec_vupklsh (vector signed short);
9508 vector bool short vec_vupklsb (vector bool char);
9509 vector signed short vec_vupklsb (vector signed char);
9511 vector float vec_xor (vector float, vector float);
9512 vector float vec_xor (vector float, vector bool int);
9513 vector float vec_xor (vector bool int, vector float);
9514 vector bool int vec_xor (vector bool int, vector bool int);
9515 vector signed int vec_xor (vector bool int, vector signed int);
9516 vector signed int vec_xor (vector signed int, vector bool int);
9517 vector signed int vec_xor (vector signed int, vector signed int);
9518 vector unsigned int vec_xor (vector bool int, vector unsigned int);
9519 vector unsigned int vec_xor (vector unsigned int, vector bool int);
9520 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
9521 vector bool short vec_xor (vector bool short, vector bool short);
9522 vector signed short vec_xor (vector bool short, vector signed short);
9523 vector signed short vec_xor (vector signed short, vector bool short);
9524 vector signed short vec_xor (vector signed short, vector signed short);
9525 vector unsigned short vec_xor (vector bool short,
9526 vector unsigned short);
9527 vector unsigned short vec_xor (vector unsigned short,
9529 vector unsigned short vec_xor (vector unsigned short,
9530 vector unsigned short);
9531 vector signed char vec_xor (vector bool char, vector signed char);
9532 vector bool char vec_xor (vector bool char, vector bool char);
9533 vector signed char vec_xor (vector signed char, vector bool char);
9534 vector signed char vec_xor (vector signed char, vector signed char);
9535 vector unsigned char vec_xor (vector bool char, vector unsigned char);
9536 vector unsigned char vec_xor (vector unsigned char, vector bool char);
9537 vector unsigned char vec_xor (vector unsigned char,
9538 vector unsigned char);
9540 int vec_all_eq (vector signed char, vector bool char);
9541 int vec_all_eq (vector signed char, vector signed char);
9542 int vec_all_eq (vector unsigned char, vector bool char);
9543 int vec_all_eq (vector unsigned char, vector unsigned char);
9544 int vec_all_eq (vector bool char, vector bool char);
9545 int vec_all_eq (vector bool char, vector unsigned char);
9546 int vec_all_eq (vector bool char, vector signed char);
9547 int vec_all_eq (vector signed short, vector bool short);
9548 int vec_all_eq (vector signed short, vector signed short);
9549 int vec_all_eq (vector unsigned short, vector bool short);
9550 int vec_all_eq (vector unsigned short, vector unsigned short);
9551 int vec_all_eq (vector bool short, vector bool short);
9552 int vec_all_eq (vector bool short, vector unsigned short);
9553 int vec_all_eq (vector bool short, vector signed short);
9554 int vec_all_eq (vector pixel, vector pixel);
9555 int vec_all_eq (vector signed int, vector bool int);
9556 int vec_all_eq (vector signed int, vector signed int);
9557 int vec_all_eq (vector unsigned int, vector bool int);
9558 int vec_all_eq (vector unsigned int, vector unsigned int);
9559 int vec_all_eq (vector bool int, vector bool int);
9560 int vec_all_eq (vector bool int, vector unsigned int);
9561 int vec_all_eq (vector bool int, vector signed int);
9562 int vec_all_eq (vector float, vector float);
9564 int vec_all_ge (vector bool char, vector unsigned char);
9565 int vec_all_ge (vector unsigned char, vector bool char);
9566 int vec_all_ge (vector unsigned char, vector unsigned char);
9567 int vec_all_ge (vector bool char, vector signed char);
9568 int vec_all_ge (vector signed char, vector bool char);
9569 int vec_all_ge (vector signed char, vector signed char);
9570 int vec_all_ge (vector bool short, vector unsigned short);
9571 int vec_all_ge (vector unsigned short, vector bool short);
9572 int vec_all_ge (vector unsigned short, vector unsigned short);
9573 int vec_all_ge (vector signed short, vector signed short);
9574 int vec_all_ge (vector bool short, vector signed short);
9575 int vec_all_ge (vector signed short, vector bool short);
9576 int vec_all_ge (vector bool int, vector unsigned int);
9577 int vec_all_ge (vector unsigned int, vector bool int);
9578 int vec_all_ge (vector unsigned int, vector unsigned int);
9579 int vec_all_ge (vector bool int, vector signed int);
9580 int vec_all_ge (vector signed int, vector bool int);
9581 int vec_all_ge (vector signed int, vector signed int);
9582 int vec_all_ge (vector float, vector float);
9584 int vec_all_gt (vector bool char, vector unsigned char);
9585 int vec_all_gt (vector unsigned char, vector bool char);
9586 int vec_all_gt (vector unsigned char, vector unsigned char);
9587 int vec_all_gt (vector bool char, vector signed char);
9588 int vec_all_gt (vector signed char, vector bool char);
9589 int vec_all_gt (vector signed char, vector signed char);
9590 int vec_all_gt (vector bool short, vector unsigned short);
9591 int vec_all_gt (vector unsigned short, vector bool short);
9592 int vec_all_gt (vector unsigned short, vector unsigned short);
9593 int vec_all_gt (vector bool short, vector signed short);
9594 int vec_all_gt (vector signed short, vector bool short);
9595 int vec_all_gt (vector signed short, vector signed short);
9596 int vec_all_gt (vector bool int, vector unsigned int);
9597 int vec_all_gt (vector unsigned int, vector bool int);
9598 int vec_all_gt (vector unsigned int, vector unsigned int);
9599 int vec_all_gt (vector bool int, vector signed int);
9600 int vec_all_gt (vector signed int, vector bool int);
9601 int vec_all_gt (vector signed int, vector signed int);
9602 int vec_all_gt (vector float, vector float);
9604 int vec_all_in (vector float, vector float);
9606 int vec_all_le (vector bool char, vector unsigned char);
9607 int vec_all_le (vector unsigned char, vector bool char);
9608 int vec_all_le (vector unsigned char, vector unsigned char);
9609 int vec_all_le (vector bool char, vector signed char);
9610 int vec_all_le (vector signed char, vector bool char);
9611 int vec_all_le (vector signed char, vector signed char);
9612 int vec_all_le (vector bool short, vector unsigned short);
9613 int vec_all_le (vector unsigned short, vector bool short);
9614 int vec_all_le (vector unsigned short, vector unsigned short);
9615 int vec_all_le (vector bool short, vector signed short);
9616 int vec_all_le (vector signed short, vector bool short);
9617 int vec_all_le (vector signed short, vector signed short);
9618 int vec_all_le (vector bool int, vector unsigned int);
9619 int vec_all_le (vector unsigned int, vector bool int);
9620 int vec_all_le (vector unsigned int, vector unsigned int);
9621 int vec_all_le (vector bool int, vector signed int);
9622 int vec_all_le (vector signed int, vector bool int);
9623 int vec_all_le (vector signed int, vector signed int);
9624 int vec_all_le (vector float, vector float);
9626 int vec_all_lt (vector bool char, vector unsigned char);
9627 int vec_all_lt (vector unsigned char, vector bool char);
9628 int vec_all_lt (vector unsigned char, vector unsigned char);
9629 int vec_all_lt (vector bool char, vector signed char);
9630 int vec_all_lt (vector signed char, vector bool char);
9631 int vec_all_lt (vector signed char, vector signed char);
9632 int vec_all_lt (vector bool short, vector unsigned short);
9633 int vec_all_lt (vector unsigned short, vector bool short);
9634 int vec_all_lt (vector unsigned short, vector unsigned short);
9635 int vec_all_lt (vector bool short, vector signed short);
9636 int vec_all_lt (vector signed short, vector bool short);
9637 int vec_all_lt (vector signed short, vector signed short);
9638 int vec_all_lt (vector bool int, vector unsigned int);
9639 int vec_all_lt (vector unsigned int, vector bool int);
9640 int vec_all_lt (vector unsigned int, vector unsigned int);
9641 int vec_all_lt (vector bool int, vector signed int);
9642 int vec_all_lt (vector signed int, vector bool int);
9643 int vec_all_lt (vector signed int, vector signed int);
9644 int vec_all_lt (vector float, vector float);
9646 int vec_all_nan (vector float);
9648 int vec_all_ne (vector signed char, vector bool char);
9649 int vec_all_ne (vector signed char, vector signed char);
9650 int vec_all_ne (vector unsigned char, vector bool char);
9651 int vec_all_ne (vector unsigned char, vector unsigned char);
9652 int vec_all_ne (vector bool char, vector bool char);
9653 int vec_all_ne (vector bool char, vector unsigned char);
9654 int vec_all_ne (vector bool char, vector signed char);
9655 int vec_all_ne (vector signed short, vector bool short);
9656 int vec_all_ne (vector signed short, vector signed short);
9657 int vec_all_ne (vector unsigned short, vector bool short);
9658 int vec_all_ne (vector unsigned short, vector unsigned short);
9659 int vec_all_ne (vector bool short, vector bool short);
9660 int vec_all_ne (vector bool short, vector unsigned short);
9661 int vec_all_ne (vector bool short, vector signed short);
9662 int vec_all_ne (vector pixel, vector pixel);
9663 int vec_all_ne (vector signed int, vector bool int);
9664 int vec_all_ne (vector signed int, vector signed int);
9665 int vec_all_ne (vector unsigned int, vector bool int);
9666 int vec_all_ne (vector unsigned int, vector unsigned int);
9667 int vec_all_ne (vector bool int, vector bool int);
9668 int vec_all_ne (vector bool int, vector unsigned int);
9669 int vec_all_ne (vector bool int, vector signed int);
9670 int vec_all_ne (vector float, vector float);
9672 int vec_all_nge (vector float, vector float);
9674 int vec_all_ngt (vector float, vector float);
9676 int vec_all_nle (vector float, vector float);
9678 int vec_all_nlt (vector float, vector float);
9680 int vec_all_numeric (vector float);
9682 int vec_any_eq (vector signed char, vector bool char);
9683 int vec_any_eq (vector signed char, vector signed char);
9684 int vec_any_eq (vector unsigned char, vector bool char);
9685 int vec_any_eq (vector unsigned char, vector unsigned char);
9686 int vec_any_eq (vector bool char, vector bool char);
9687 int vec_any_eq (vector bool char, vector unsigned char);
9688 int vec_any_eq (vector bool char, vector signed char);
9689 int vec_any_eq (vector signed short, vector bool short);
9690 int vec_any_eq (vector signed short, vector signed short);
9691 int vec_any_eq (vector unsigned short, vector bool short);
9692 int vec_any_eq (vector unsigned short, vector unsigned short);
9693 int vec_any_eq (vector bool short, vector bool short);
9694 int vec_any_eq (vector bool short, vector unsigned short);
9695 int vec_any_eq (vector bool short, vector signed short);
9696 int vec_any_eq (vector pixel, vector pixel);
9697 int vec_any_eq (vector signed int, vector bool int);
9698 int vec_any_eq (vector signed int, vector signed int);
9699 int vec_any_eq (vector unsigned int, vector bool int);
9700 int vec_any_eq (vector unsigned int, vector unsigned int);
9701 int vec_any_eq (vector bool int, vector bool int);
9702 int vec_any_eq (vector bool int, vector unsigned int);
9703 int vec_any_eq (vector bool int, vector signed int);
9704 int vec_any_eq (vector float, vector float);
9706 int vec_any_ge (vector signed char, vector bool char);
9707 int vec_any_ge (vector unsigned char, vector bool char);
9708 int vec_any_ge (vector unsigned char, vector unsigned char);
9709 int vec_any_ge (vector signed char, vector signed char);
9710 int vec_any_ge (vector bool char, vector unsigned char);
9711 int vec_any_ge (vector bool char, vector signed char);
9712 int vec_any_ge (vector unsigned short, vector bool short);
9713 int vec_any_ge (vector unsigned short, vector unsigned short);
9714 int vec_any_ge (vector signed short, vector signed short);
9715 int vec_any_ge (vector signed short, vector bool short);
9716 int vec_any_ge (vector bool short, vector unsigned short);
9717 int vec_any_ge (vector bool short, vector signed short);
9718 int vec_any_ge (vector signed int, vector bool int);
9719 int vec_any_ge (vector unsigned int, vector bool int);
9720 int vec_any_ge (vector unsigned int, vector unsigned int);
9721 int vec_any_ge (vector signed int, vector signed int);
9722 int vec_any_ge (vector bool int, vector unsigned int);
9723 int vec_any_ge (vector bool int, vector signed int);
9724 int vec_any_ge (vector float, vector float);
9726 int vec_any_gt (vector bool char, vector unsigned char);
9727 int vec_any_gt (vector unsigned char, vector bool char);
9728 int vec_any_gt (vector unsigned char, vector unsigned char);
9729 int vec_any_gt (vector bool char, vector signed char);
9730 int vec_any_gt (vector signed char, vector bool char);
9731 int vec_any_gt (vector signed char, vector signed char);
9732 int vec_any_gt (vector bool short, vector unsigned short);
9733 int vec_any_gt (vector unsigned short, vector bool short);
9734 int vec_any_gt (vector unsigned short, vector unsigned short);
9735 int vec_any_gt (vector bool short, vector signed short);
9736 int vec_any_gt (vector signed short, vector bool short);
9737 int vec_any_gt (vector signed short, vector signed short);
9738 int vec_any_gt (vector bool int, vector unsigned int);
9739 int vec_any_gt (vector unsigned int, vector bool int);
9740 int vec_any_gt (vector unsigned int, vector unsigned int);
9741 int vec_any_gt (vector bool int, vector signed int);
9742 int vec_any_gt (vector signed int, vector bool int);
9743 int vec_any_gt (vector signed int, vector signed int);
9744 int vec_any_gt (vector float, vector float);
9746 int vec_any_le (vector bool char, vector unsigned char);
9747 int vec_any_le (vector unsigned char, vector bool char);
9748 int vec_any_le (vector unsigned char, vector unsigned char);
9749 int vec_any_le (vector bool char, vector signed char);
9750 int vec_any_le (vector signed char, vector bool char);
9751 int vec_any_le (vector signed char, vector signed char);
9752 int vec_any_le (vector bool short, vector unsigned short);
9753 int vec_any_le (vector unsigned short, vector bool short);
9754 int vec_any_le (vector unsigned short, vector unsigned short);
9755 int vec_any_le (vector bool short, vector signed short);
9756 int vec_any_le (vector signed short, vector bool short);
9757 int vec_any_le (vector signed short, vector signed short);
9758 int vec_any_le (vector bool int, vector unsigned int);
9759 int vec_any_le (vector unsigned int, vector bool int);
9760 int vec_any_le (vector unsigned int, vector unsigned int);
9761 int vec_any_le (vector bool int, vector signed int);
9762 int vec_any_le (vector signed int, vector bool int);
9763 int vec_any_le (vector signed int, vector signed int);
9764 int vec_any_le (vector float, vector float);
9766 int vec_any_lt (vector bool char, vector unsigned char);
9767 int vec_any_lt (vector unsigned char, vector bool char);
9768 int vec_any_lt (vector unsigned char, vector unsigned char);
9769 int vec_any_lt (vector bool char, vector signed char);
9770 int vec_any_lt (vector signed char, vector bool char);
9771 int vec_any_lt (vector signed char, vector signed char);
9772 int vec_any_lt (vector bool short, vector unsigned short);
9773 int vec_any_lt (vector unsigned short, vector bool short);
9774 int vec_any_lt (vector unsigned short, vector unsigned short);
9775 int vec_any_lt (vector bool short, vector signed short);
9776 int vec_any_lt (vector signed short, vector bool short);
9777 int vec_any_lt (vector signed short, vector signed short);
9778 int vec_any_lt (vector bool int, vector unsigned int);
9779 int vec_any_lt (vector unsigned int, vector bool int);
9780 int vec_any_lt (vector unsigned int, vector unsigned int);
9781 int vec_any_lt (vector bool int, vector signed int);
9782 int vec_any_lt (vector signed int, vector bool int);
9783 int vec_any_lt (vector signed int, vector signed int);
9784 int vec_any_lt (vector float, vector float);
9786 int vec_any_nan (vector float);
9788 int vec_any_ne (vector signed char, vector bool char);
9789 int vec_any_ne (vector signed char, vector signed char);
9790 int vec_any_ne (vector unsigned char, vector bool char);
9791 int vec_any_ne (vector unsigned char, vector unsigned char);
9792 int vec_any_ne (vector bool char, vector bool char);
9793 int vec_any_ne (vector bool char, vector unsigned char);
9794 int vec_any_ne (vector bool char, vector signed char);
9795 int vec_any_ne (vector signed short, vector bool short);
9796 int vec_any_ne (vector signed short, vector signed short);
9797 int vec_any_ne (vector unsigned short, vector bool short);
9798 int vec_any_ne (vector unsigned short, vector unsigned short);
9799 int vec_any_ne (vector bool short, vector bool short);
9800 int vec_any_ne (vector bool short, vector unsigned short);
9801 int vec_any_ne (vector bool short, vector signed short);
9802 int vec_any_ne (vector pixel, vector pixel);
9803 int vec_any_ne (vector signed int, vector bool int);
9804 int vec_any_ne (vector signed int, vector signed int);
9805 int vec_any_ne (vector unsigned int, vector bool int);
9806 int vec_any_ne (vector unsigned int, vector unsigned int);
9807 int vec_any_ne (vector bool int, vector bool int);
9808 int vec_any_ne (vector bool int, vector unsigned int);
9809 int vec_any_ne (vector bool int, vector signed int);
9810 int vec_any_ne (vector float, vector float);
9812 int vec_any_nge (vector float, vector float);
9814 int vec_any_ngt (vector float, vector float);
9816 int vec_any_nle (vector float, vector float);
9818 int vec_any_nlt (vector float, vector float);
9820 int vec_any_numeric (vector float);
9822 int vec_any_out (vector float, vector float);
9825 @node SPARC VIS Built-in Functions
9826 @subsection SPARC VIS Built-in Functions
9828 GCC supports SIMD operations on the SPARC using both the generic vector
9829 extensions (@pxref{Vector Extensions}) as well as built-in functions for
9830 the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis}
9831 switch, the VIS extension is exposed as the following built-in functions:
9834 typedef int v2si __attribute__ ((vector_size (8)));
9835 typedef short v4hi __attribute__ ((vector_size (8)));
9836 typedef short v2hi __attribute__ ((vector_size (4)));
9837 typedef char v8qi __attribute__ ((vector_size (8)));
9838 typedef char v4qi __attribute__ ((vector_size (4)));
9840 void * __builtin_vis_alignaddr (void *, long);
9841 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
9842 v2si __builtin_vis_faligndatav2si (v2si, v2si);
9843 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
9844 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
9846 v4hi __builtin_vis_fexpand (v4qi);
9848 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
9849 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
9850 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
9851 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
9852 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
9853 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
9854 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
9856 v4qi __builtin_vis_fpack16 (v4hi);
9857 v8qi __builtin_vis_fpack32 (v2si, v2si);
9858 v2hi __builtin_vis_fpackfix (v2si);
9859 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
9861 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
9864 @node Target Format Checks
9865 @section Format Checks Specific to Particular Target Machines
9867 For some target machines, GCC supports additional options to the
9869 (@pxref{Function Attributes,,Declaring Attributes of Functions}).
9872 * Solaris Format Checks::
9875 @node Solaris Format Checks
9876 @subsection Solaris Format Checks
9878 Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format
9879 check. @code{cmn_err} accepts a subset of the standard @code{printf}
9880 conversions, and the two-argument @code{%b} conversion for displaying
9881 bit-fields. See the Solaris man page for @code{cmn_err} for more information.
9884 @section Pragmas Accepted by GCC
9888 GCC supports several types of pragmas, primarily in order to compile
9889 code originally written for other compilers. Note that in general
9890 we do not recommend the use of pragmas; @xref{Function Attributes},
9891 for further explanation.
9896 * RS/6000 and PowerPC Pragmas::
9899 * Symbol-Renaming Pragmas::
9900 * Structure-Packing Pragmas::
9902 * Diagnostic Pragmas::
9903 * Visibility Pragmas::
9907 @subsection ARM Pragmas
9909 The ARM target defines pragmas for controlling the default addition of
9910 @code{long_call} and @code{short_call} attributes to functions.
9911 @xref{Function Attributes}, for information about the effects of these
9916 @cindex pragma, long_calls
9917 Set all subsequent functions to have the @code{long_call} attribute.
9920 @cindex pragma, no_long_calls
9921 Set all subsequent functions to have the @code{short_call} attribute.
9923 @item long_calls_off
9924 @cindex pragma, long_calls_off
9925 Do not affect the @code{long_call} or @code{short_call} attributes of
9926 subsequent functions.
9930 @subsection M32C Pragmas
9933 @item memregs @var{number}
9934 @cindex pragma, memregs
9935 Overrides the command line option @code{-memregs=} for the current
9936 file. Use with care! This pragma must be before any function in the
9937 file, and mixing different memregs values in different objects may
9938 make them incompatible. This pragma is useful when a
9939 performance-critical function uses a memreg for temporary values,
9940 as it may allow you to reduce the number of memregs used.
9944 @node RS/6000 and PowerPC Pragmas
9945 @subsection RS/6000 and PowerPC Pragmas
9947 The RS/6000 and PowerPC targets define one pragma for controlling
9948 whether or not the @code{longcall} attribute is added to function
9949 declarations by default. This pragma overrides the @option{-mlongcall}
9950 option, but not the @code{longcall} and @code{shortcall} attributes.
9951 @xref{RS/6000 and PowerPC Options}, for more information about when long
9952 calls are and are not necessary.
9956 @cindex pragma, longcall
9957 Apply the @code{longcall} attribute to all subsequent function
9961 Do not apply the @code{longcall} attribute to subsequent function
9965 @c Describe c4x pragmas here.
9966 @c Describe h8300 pragmas here.
9967 @c Describe sh pragmas here.
9968 @c Describe v850 pragmas here.
9970 @node Darwin Pragmas
9971 @subsection Darwin Pragmas
9973 The following pragmas are available for all architectures running the
9974 Darwin operating system. These are useful for compatibility with other
9978 @item mark @var{tokens}@dots{}
9979 @cindex pragma, mark
9980 This pragma is accepted, but has no effect.
9982 @item options align=@var{alignment}
9983 @cindex pragma, options align
9984 This pragma sets the alignment of fields in structures. The values of
9985 @var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
9986 @code{power}, to emulate PowerPC alignment. Uses of this pragma nest
9987 properly; to restore the previous setting, use @code{reset} for the
9990 @item segment @var{tokens}@dots{}
9991 @cindex pragma, segment
9992 This pragma is accepted, but has no effect.
9994 @item unused (@var{var} [, @var{var}]@dots{})
9995 @cindex pragma, unused
9996 This pragma declares variables to be possibly unused. GCC will not
9997 produce warnings for the listed variables. The effect is similar to
9998 that of the @code{unused} attribute, except that this pragma may appear
9999 anywhere within the variables' scopes.
10002 @node Solaris Pragmas
10003 @subsection Solaris Pragmas
10005 The Solaris target supports @code{#pragma redefine_extname}
10006 (@pxref{Symbol-Renaming Pragmas}). It also supports additional
10007 @code{#pragma} directives for compatibility with the system compiler.
10010 @item align @var{alignment} (@var{variable} [, @var{variable}]...)
10011 @cindex pragma, align
10013 Increase the minimum alignment of each @var{variable} to @var{alignment}.
10014 This is the same as GCC's @code{aligned} attribute @pxref{Variable
10015 Attributes}). Macro expansion occurs on the arguments to this pragma
10016 when compiling C and Objective-C. It does not currently occur when
10017 compiling C++, but this is a bug which may be fixed in a future
10020 @item fini (@var{function} [, @var{function}]...)
10021 @cindex pragma, fini
10023 This pragma causes each listed @var{function} to be called after
10024 main, or during shared module unloading, by adding a call to the
10025 @code{.fini} section.
10027 @item init (@var{function} [, @var{function}]...)
10028 @cindex pragma, init
10030 This pragma causes each listed @var{function} to be called during
10031 initialization (before @code{main}) or during shared module loading, by
10032 adding a call to the @code{.init} section.
10036 @node Symbol-Renaming Pragmas
10037 @subsection Symbol-Renaming Pragmas
10039 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
10040 supports two @code{#pragma} directives which change the name used in
10041 assembly for a given declaration. These pragmas are only available on
10042 platforms whose system headers need them. To get this effect on all
10043 platforms supported by GCC, use the asm labels extension (@pxref{Asm
10047 @item redefine_extname @var{oldname} @var{newname}
10048 @cindex pragma, redefine_extname
10050 This pragma gives the C function @var{oldname} the assembly symbol
10051 @var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME}
10052 will be defined if this pragma is available (currently only on
10055 @item extern_prefix @var{string}
10056 @cindex pragma, extern_prefix
10058 This pragma causes all subsequent external function and variable
10059 declarations to have @var{string} prepended to their assembly symbols.
10060 This effect may be terminated with another @code{extern_prefix} pragma
10061 whose argument is an empty string. The preprocessor macro
10062 @code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is
10063 available (currently only on Tru64 UNIX)@.
10066 These pragmas and the asm labels extension interact in a complicated
10067 manner. Here are some corner cases you may want to be aware of.
10070 @item Both pragmas silently apply only to declarations with external
10071 linkage. Asm labels do not have this restriction.
10073 @item In C++, both pragmas silently apply only to declarations with
10074 ``C'' linkage. Again, asm labels do not have this restriction.
10076 @item If any of the three ways of changing the assembly name of a
10077 declaration is applied to a declaration whose assembly name has
10078 already been determined (either by a previous use of one of these
10079 features, or because the compiler needed the assembly name in order to
10080 generate code), and the new name is different, a warning issues and
10081 the name does not change.
10083 @item The @var{oldname} used by @code{#pragma redefine_extname} is
10084 always the C-language name.
10086 @item If @code{#pragma extern_prefix} is in effect, and a declaration
10087 occurs with an asm label attached, the prefix is silently ignored for
10090 @item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname}
10091 apply to the same declaration, whichever triggered first wins, and a
10092 warning issues if they contradict each other. (We would like to have
10093 @code{#pragma redefine_extname} always win, for consistency with asm
10094 labels, but if @code{#pragma extern_prefix} triggers first we have no
10095 way of knowing that that happened.)
10098 @node Structure-Packing Pragmas
10099 @subsection Structure-Packing Pragmas
10101 For compatibility with Win32, GCC supports a set of @code{#pragma}
10102 directives which change the maximum alignment of members of structures
10103 (other than zero-width bitfields), unions, and classes subsequently
10104 defined. The @var{n} value below always is required to be a small power
10105 of two and specifies the new alignment in bytes.
10108 @item @code{#pragma pack(@var{n})} simply sets the new alignment.
10109 @item @code{#pragma pack()} sets the alignment to the one that was in
10110 effect when compilation started (see also command line option
10111 @option{-fpack-struct[=<n>]} @pxref{Code Gen Options}).
10112 @item @code{#pragma pack(push[,@var{n}])} pushes the current alignment
10113 setting on an internal stack and then optionally sets the new alignment.
10114 @item @code{#pragma pack(pop)} restores the alignment setting to the one
10115 saved at the top of the internal stack (and removes that stack entry).
10116 Note that @code{#pragma pack([@var{n}])} does not influence this internal
10117 stack; thus it is possible to have @code{#pragma pack(push)} followed by
10118 multiple @code{#pragma pack(@var{n})} instances and finalized by a single
10119 @code{#pragma pack(pop)}.
10122 Some targets, e.g. i386 and powerpc, support the @code{ms_struct}
10123 @code{#pragma} which lays out a structure as the documented
10124 @code{__attribute__ ((ms_struct))}.
10126 @item @code{#pragma ms_struct on} turns on the layout for structures
10128 @item @code{#pragma ms_struct off} turns off the layout for structures
10130 @item @code{#pragma ms_struct reset} goes back to the default layout.
10134 @subsection Weak Pragmas
10136 For compatibility with SVR4, GCC supports a set of @code{#pragma}
10137 directives for declaring symbols to be weak, and defining weak
10141 @item #pragma weak @var{symbol}
10142 @cindex pragma, weak
10143 This pragma declares @var{symbol} to be weak, as if the declaration
10144 had the attribute of the same name. The pragma may appear before
10145 or after the declaration of @var{symbol}, but must appear before
10146 either its first use or its definition. It is not an error for
10147 @var{symbol} to never be defined at all.
10149 @item #pragma weak @var{symbol1} = @var{symbol2}
10150 This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}.
10151 It is an error if @var{symbol2} is not defined in the current
10155 @node Diagnostic Pragmas
10156 @subsection Diagnostic Pragmas
10158 GCC allows the user to selectively enable or disable certain types of
10159 diagnostics, and change the kind of the diagnostic. For example, a
10160 project's policy might require that all sources compile with
10161 @option{-Werror} but certain files might have exceptions allowing
10162 specific types of warnings. Or, a project might selectively enable
10163 diagnostics and treat them as errors depending on which preprocessor
10164 macros are defined.
10167 @item #pragma GCC diagnostic @var{kind} @var{option}
10168 @cindex pragma, diagnostic
10170 Modifies the disposition of a diagnostic. Note that not all
10171 diagnostics are modifiable; at the moment only warnings (normally
10172 controlled by @samp{-W...}) can be controlled, and not all of them.
10173 Use @option{-fdiagnostics-show-option} to determine which diagnostics
10174 are controllable and which option controls them.
10176 @var{kind} is @samp{error} to treat this diagnostic as an error,
10177 @samp{warning} to treat it like a warning (even if @option{-Werror} is
10178 in effect), or @samp{ignored} if the diagnostic is to be ignored.
10179 @var{option} is a double quoted string which matches the command line
10183 #pragma GCC diagnostic warning "-Wformat"
10184 #pragma GCC diagnostic error "-Wformat"
10185 #pragma GCC diagnostic ignored "-Wformat"
10188 Note that these pragmas override any command line options. Also,
10189 while it is syntactically valid to put these pragmas anywhere in your
10190 sources, the only supported location for them is before any data or
10191 functions are defined. Doing otherwise may result in unpredictable
10192 results depending on how the optimizer manages your sources. If the
10193 same option is listed multiple times, the last one specified is the
10194 one that is in effect. This pragma is not intended to be a general
10195 purpose replacement for command line options, but for implementing
10196 strict control over project policies.
10200 @node Visibility Pragmas
10201 @subsection Visibility Pragmas
10204 @item #pragma GCC visibility push(@var{visibility})
10205 @itemx #pragma GCC visibility pop
10206 @cindex pragma, visibility
10208 This pragma allows the user to set the visibility for multiple
10209 declarations without having to give each a visibility attribute
10210 @xref{Function Attributes}, for more information about visibility and
10211 the attribute syntax.
10213 In C++, @samp{#pragma GCC visibility} affects only namespace-scope
10214 declarations. Class members and template specializations are not
10215 affected; if you want to override the visibility for a particular
10216 member or instantiation, you must use an attribute.
10220 @node Unnamed Fields
10221 @section Unnamed struct/union fields within structs/unions
10225 For compatibility with other compilers, GCC allows you to define
10226 a structure or union that contains, as fields, structures and unions
10227 without names. For example:
10240 In this example, the user would be able to access members of the unnamed
10241 union with code like @samp{foo.b}. Note that only unnamed structs and
10242 unions are allowed, you may not have, for example, an unnamed
10245 You must never create such structures that cause ambiguous field definitions.
10246 For example, this structure:
10257 It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
10258 Such constructs are not supported and must be avoided. In the future,
10259 such constructs may be detected and treated as compilation errors.
10261 @opindex fms-extensions
10262 Unless @option{-fms-extensions} is used, the unnamed field must be a
10263 structure or union definition without a tag (for example, @samp{struct
10264 @{ int a; @};}). If @option{-fms-extensions} is used, the field may
10265 also be a definition with a tag such as @samp{struct foo @{ int a;
10266 @};}, a reference to a previously defined structure or union such as
10267 @samp{struct foo;}, or a reference to a @code{typedef} name for a
10268 previously defined structure or union type.
10271 @section Thread-Local Storage
10272 @cindex Thread-Local Storage
10273 @cindex @acronym{TLS}
10276 Thread-local storage (@acronym{TLS}) is a mechanism by which variables
10277 are allocated such that there is one instance of the variable per extant
10278 thread. The run-time model GCC uses to implement this originates
10279 in the IA-64 processor-specific ABI, but has since been migrated
10280 to other processors as well. It requires significant support from
10281 the linker (@command{ld}), dynamic linker (@command{ld.so}), and
10282 system libraries (@file{libc.so} and @file{libpthread.so}), so it
10283 is not available everywhere.
10285 At the user level, the extension is visible with a new storage
10286 class keyword: @code{__thread}. For example:
10290 extern __thread struct state s;
10291 static __thread char *p;
10294 The @code{__thread} specifier may be used alone, with the @code{extern}
10295 or @code{static} specifiers, but with no other storage class specifier.
10296 When used with @code{extern} or @code{static}, @code{__thread} must appear
10297 immediately after the other storage class specifier.
10299 The @code{__thread} specifier may be applied to any global, file-scoped
10300 static, function-scoped static, or static data member of a class. It may
10301 not be applied to block-scoped automatic or non-static data member.
10303 When the address-of operator is applied to a thread-local variable, it is
10304 evaluated at run-time and returns the address of the current thread's
10305 instance of that variable. An address so obtained may be used by any
10306 thread. When a thread terminates, any pointers to thread-local variables
10307 in that thread become invalid.
10309 No static initialization may refer to the address of a thread-local variable.
10311 In C++, if an initializer is present for a thread-local variable, it must
10312 be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
10315 See @uref{http://people.redhat.com/drepper/tls.pdf,
10316 ELF Handling For Thread-Local Storage} for a detailed explanation of
10317 the four thread-local storage addressing models, and how the run-time
10318 is expected to function.
10321 * C99 Thread-Local Edits::
10322 * C++98 Thread-Local Edits::
10325 @node C99 Thread-Local Edits
10326 @subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
10328 The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
10329 that document the exact semantics of the language extension.
10333 @cite{5.1.2 Execution environments}
10335 Add new text after paragraph 1
10338 Within either execution environment, a @dfn{thread} is a flow of
10339 control within a program. It is implementation defined whether
10340 or not there may be more than one thread associated with a program.
10341 It is implementation defined how threads beyond the first are
10342 created, the name and type of the function called at thread
10343 startup, and how threads may be terminated. However, objects
10344 with thread storage duration shall be initialized before thread
10349 @cite{6.2.4 Storage durations of objects}
10351 Add new text before paragraph 3
10354 An object whose identifier is declared with the storage-class
10355 specifier @w{@code{__thread}} has @dfn{thread storage duration}.
10356 Its lifetime is the entire execution of the thread, and its
10357 stored value is initialized only once, prior to thread startup.
10361 @cite{6.4.1 Keywords}
10363 Add @code{__thread}.
10366 @cite{6.7.1 Storage-class specifiers}
10368 Add @code{__thread} to the list of storage class specifiers in
10371 Change paragraph 2 to
10374 With the exception of @code{__thread}, at most one storage-class
10375 specifier may be given [@dots{}]. The @code{__thread} specifier may
10376 be used alone, or immediately following @code{extern} or
10380 Add new text after paragraph 6
10383 The declaration of an identifier for a variable that has
10384 block scope that specifies @code{__thread} shall also
10385 specify either @code{extern} or @code{static}.
10387 The @code{__thread} specifier shall be used only with
10392 @node C++98 Thread-Local Edits
10393 @subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
10395 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
10396 that document the exact semantics of the language extension.
10400 @b{[intro.execution]}
10402 New text after paragraph 4
10405 A @dfn{thread} is a flow of control within the abstract machine.
10406 It is implementation defined whether or not there may be more than
10410 New text after paragraph 7
10413 It is unspecified whether additional action must be taken to
10414 ensure when and whether side effects are visible to other threads.
10420 Add @code{__thread}.
10423 @b{[basic.start.main]}
10425 Add after paragraph 5
10428 The thread that begins execution at the @code{main} function is called
10429 the @dfn{main thread}. It is implementation defined how functions
10430 beginning threads other than the main thread are designated or typed.
10431 A function so designated, as well as the @code{main} function, is called
10432 a @dfn{thread startup function}. It is implementation defined what
10433 happens if a thread startup function returns. It is implementation
10434 defined what happens to other threads when any thread calls @code{exit}.
10438 @b{[basic.start.init]}
10440 Add after paragraph 4
10443 The storage for an object of thread storage duration shall be
10444 statically initialized before the first statement of the thread startup
10445 function. An object of thread storage duration shall not require
10446 dynamic initialization.
10450 @b{[basic.start.term]}
10452 Add after paragraph 3
10455 The type of an object with thread storage duration shall not have a
10456 non-trivial destructor, nor shall it be an array type whose elements
10457 (directly or indirectly) have non-trivial destructors.
10463 Add ``thread storage duration'' to the list in paragraph 1.
10468 Thread, static, and automatic storage durations are associated with
10469 objects introduced by declarations [@dots{}].
10472 Add @code{__thread} to the list of specifiers in paragraph 3.
10475 @b{[basic.stc.thread]}
10477 New section before @b{[basic.stc.static]}
10480 The keyword @code{__thread} applied to a non-local object gives the
10481 object thread storage duration.
10483 A local variable or class data member declared both @code{static}
10484 and @code{__thread} gives the variable or member thread storage
10489 @b{[basic.stc.static]}
10494 All objects which have neither thread storage duration, dynamic
10495 storage duration nor are local [@dots{}].
10501 Add @code{__thread} to the list in paragraph 1.
10506 With the exception of @code{__thread}, at most one
10507 @var{storage-class-specifier} shall appear in a given
10508 @var{decl-specifier-seq}. The @code{__thread} specifier may
10509 be used alone, or immediately following the @code{extern} or
10510 @code{static} specifiers. [@dots{}]
10513 Add after paragraph 5
10516 The @code{__thread} specifier can be applied only to the names of objects
10517 and to anonymous unions.
10523 Add after paragraph 6
10526 Non-@code{static} members shall not be @code{__thread}.
10530 @node Binary constants
10531 @section Binary constants using the @samp{0b} prefix
10532 @cindex Binary constants using the @samp{0b} prefix
10534 Integer constants can be written as binary constants, consisting of a
10535 sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or
10536 @samp{0B}. This is particularly useful in environments that operate a
10537 lot on the bit-level (like microcontrollers).
10539 The following statements are identical:
10548 The type of these constants follows the same rules as for octal or
10549 hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL}
10552 @node C++ Extensions
10553 @chapter Extensions to the C++ Language
10554 @cindex extensions, C++ language
10555 @cindex C++ language extensions
10557 The GNU compiler provides these extensions to the C++ language (and you
10558 can also use most of the C language extensions in your C++ programs). If you
10559 want to write code that checks whether these features are available, you can
10560 test for the GNU compiler the same way as for C programs: check for a
10561 predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to
10562 test specifically for GNU C++ (@pxref{Common Predefined Macros,,
10563 Predefined Macros,cpp,The GNU C Preprocessor}).
10566 * Volatiles:: What constitutes an access to a volatile object.
10567 * Restricted Pointers:: C99 restricted pointers and references.
10568 * Vague Linkage:: Where G++ puts inlines, vtables and such.
10569 * C++ Interface:: You can use a single C++ header file for both
10570 declarations and definitions.
10571 * Template Instantiation:: Methods for ensuring that exactly one copy of
10572 each needed template instantiation is emitted.
10573 * Bound member functions:: You can extract a function pointer to the
10574 method denoted by a @samp{->*} or @samp{.*} expression.
10575 * C++ Attributes:: Variable, function, and type attributes for C++ only.
10576 * Namespace Association:: Strong using-directives for namespace association.
10577 * Java Exceptions:: Tweaking exception handling to work with Java.
10578 * Deprecated Features:: Things will disappear from g++.
10579 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
10583 @section When is a Volatile Object Accessed?
10584 @cindex accessing volatiles
10585 @cindex volatile read
10586 @cindex volatile write
10587 @cindex volatile access
10589 Both the C and C++ standard have the concept of volatile objects. These
10590 are normally accessed by pointers and used for accessing hardware. The
10591 standards encourage compilers to refrain from optimizations concerning
10592 accesses to volatile objects. The C standard leaves it implementation
10593 defined as to what constitutes a volatile access. The C++ standard omits
10594 to specify this, except to say that C++ should behave in a similar manner
10595 to C with respect to volatiles, where possible. The minimum either
10596 standard specifies is that at a sequence point all previous accesses to
10597 volatile objects have stabilized and no subsequent accesses have
10598 occurred. Thus an implementation is free to reorder and combine
10599 volatile accesses which occur between sequence points, but cannot do so
10600 for accesses across a sequence point. The use of volatiles does not
10601 allow you to violate the restriction on updating objects multiple times
10602 within a sequence point.
10604 @xref{Qualifiers implementation, , Volatile qualifier and the C compiler}.
10606 The behavior differs slightly between C and C++ in the non-obvious cases:
10609 volatile int *src = @var{somevalue};
10613 With C, such expressions are rvalues, and GCC interprets this either as a
10614 read of the volatile object being pointed to or only as request to evaluate
10615 the side-effects. The C++ standard specifies that such expressions do not
10616 undergo lvalue to rvalue conversion, and that the type of the dereferenced
10617 object may be incomplete. The C++ standard does not specify explicitly
10618 that it is this lvalue to rvalue conversion which may be responsible for
10619 causing an access. However, there is reason to believe that it is,
10620 because otherwise certain simple expressions become undefined. However,
10621 because it would surprise most programmers, G++ treats dereferencing a
10622 pointer to volatile object of complete type when the value is unused as
10623 GCC would do for an equivalent type in C. When the object has incomplete
10624 type, G++ issues a warning; if you wish to force an error, you must
10625 force a conversion to rvalue with, for instance, a static cast.
10627 When using a reference to volatile, G++ does not treat equivalent
10628 expressions as accesses to volatiles, but instead issues a warning that
10629 no volatile is accessed. The rationale for this is that otherwise it
10630 becomes difficult to determine where volatile access occur, and not
10631 possible to ignore the return value from functions returning volatile
10632 references. Again, if you wish to force a read, cast the reference to
10635 @node Restricted Pointers
10636 @section Restricting Pointer Aliasing
10637 @cindex restricted pointers
10638 @cindex restricted references
10639 @cindex restricted this pointer
10641 As with the C front end, G++ understands the C99 feature of restricted pointers,
10642 specified with the @code{__restrict__}, or @code{__restrict} type
10643 qualifier. Because you cannot compile C++ by specifying the @option{-std=c99}
10644 language flag, @code{restrict} is not a keyword in C++.
10646 In addition to allowing restricted pointers, you can specify restricted
10647 references, which indicate that the reference is not aliased in the local
10651 void fn (int *__restrict__ rptr, int &__restrict__ rref)
10658 In the body of @code{fn}, @var{rptr} points to an unaliased integer and
10659 @var{rref} refers to a (different) unaliased integer.
10661 You may also specify whether a member function's @var{this} pointer is
10662 unaliased by using @code{__restrict__} as a member function qualifier.
10665 void T::fn () __restrict__
10672 Within the body of @code{T::fn}, @var{this} will have the effective
10673 definition @code{T *__restrict__ const this}. Notice that the
10674 interpretation of a @code{__restrict__} member function qualifier is
10675 different to that of @code{const} or @code{volatile} qualifier, in that it
10676 is applied to the pointer rather than the object. This is consistent with
10677 other compilers which implement restricted pointers.
10679 As with all outermost parameter qualifiers, @code{__restrict__} is
10680 ignored in function definition matching. This means you only need to
10681 specify @code{__restrict__} in a function definition, rather than
10682 in a function prototype as well.
10684 @node Vague Linkage
10685 @section Vague Linkage
10686 @cindex vague linkage
10688 There are several constructs in C++ which require space in the object
10689 file but are not clearly tied to a single translation unit. We say that
10690 these constructs have ``vague linkage''. Typically such constructs are
10691 emitted wherever they are needed, though sometimes we can be more
10695 @item Inline Functions
10696 Inline functions are typically defined in a header file which can be
10697 included in many different compilations. Hopefully they can usually be
10698 inlined, but sometimes an out-of-line copy is necessary, if the address
10699 of the function is taken or if inlining fails. In general, we emit an
10700 out-of-line copy in all translation units where one is needed. As an
10701 exception, we only emit inline virtual functions with the vtable, since
10702 it will always require a copy.
10704 Local static variables and string constants used in an inline function
10705 are also considered to have vague linkage, since they must be shared
10706 between all inlined and out-of-line instances of the function.
10710 C++ virtual functions are implemented in most compilers using a lookup
10711 table, known as a vtable. The vtable contains pointers to the virtual
10712 functions provided by a class, and each object of the class contains a
10713 pointer to its vtable (or vtables, in some multiple-inheritance
10714 situations). If the class declares any non-inline, non-pure virtual
10715 functions, the first one is chosen as the ``key method'' for the class,
10716 and the vtable is only emitted in the translation unit where the key
10719 @emph{Note:} If the chosen key method is later defined as inline, the
10720 vtable will still be emitted in every translation unit which defines it.
10721 Make sure that any inline virtuals are declared inline in the class
10722 body, even if they are not defined there.
10724 @item type_info objects
10727 C++ requires information about types to be written out in order to
10728 implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
10729 For polymorphic classes (classes with virtual functions), the type_info
10730 object is written out along with the vtable so that @samp{dynamic_cast}
10731 can determine the dynamic type of a class object at runtime. For all
10732 other types, we write out the type_info object when it is used: when
10733 applying @samp{typeid} to an expression, throwing an object, or
10734 referring to a type in a catch clause or exception specification.
10736 @item Template Instantiations
10737 Most everything in this section also applies to template instantiations,
10738 but there are other options as well.
10739 @xref{Template Instantiation,,Where's the Template?}.
10743 When used with GNU ld version 2.8 or later on an ELF system such as
10744 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
10745 these constructs will be discarded at link time. This is known as
10748 On targets that don't support COMDAT, but do support weak symbols, GCC
10749 will use them. This way one copy will override all the others, but
10750 the unused copies will still take up space in the executable.
10752 For targets which do not support either COMDAT or weak symbols,
10753 most entities with vague linkage will be emitted as local symbols to
10754 avoid duplicate definition errors from the linker. This will not happen
10755 for local statics in inlines, however, as having multiple copies will
10756 almost certainly break things.
10758 @xref{C++ Interface,,Declarations and Definitions in One Header}, for
10759 another way to control placement of these constructs.
10761 @node C++ Interface
10762 @section #pragma interface and implementation
10764 @cindex interface and implementation headers, C++
10765 @cindex C++ interface and implementation headers
10766 @cindex pragmas, interface and implementation
10768 @code{#pragma interface} and @code{#pragma implementation} provide the
10769 user with a way of explicitly directing the compiler to emit entities
10770 with vague linkage (and debugging information) in a particular
10773 @emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
10774 most cases, because of COMDAT support and the ``key method'' heuristic
10775 mentioned in @ref{Vague Linkage}. Using them can actually cause your
10776 program to grow due to unnecessary out-of-line copies of inline
10777 functions. Currently (3.4) the only benefit of these
10778 @code{#pragma}s is reduced duplication of debugging information, and
10779 that should be addressed soon on DWARF 2 targets with the use of
10783 @item #pragma interface
10784 @itemx #pragma interface "@var{subdir}/@var{objects}.h"
10785 @kindex #pragma interface
10786 Use this directive in @emph{header files} that define object classes, to save
10787 space in most of the object files that use those classes. Normally,
10788 local copies of certain information (backup copies of inline member
10789 functions, debugging information, and the internal tables that implement
10790 virtual functions) must be kept in each object file that includes class
10791 definitions. You can use this pragma to avoid such duplication. When a
10792 header file containing @samp{#pragma interface} is included in a
10793 compilation, this auxiliary information will not be generated (unless
10794 the main input source file itself uses @samp{#pragma implementation}).
10795 Instead, the object files will contain references to be resolved at link
10798 The second form of this directive is useful for the case where you have
10799 multiple headers with the same name in different directories. If you
10800 use this form, you must specify the same string to @samp{#pragma
10803 @item #pragma implementation
10804 @itemx #pragma implementation "@var{objects}.h"
10805 @kindex #pragma implementation
10806 Use this pragma in a @emph{main input file}, when you want full output from
10807 included header files to be generated (and made globally visible). The
10808 included header file, in turn, should use @samp{#pragma interface}.
10809 Backup copies of inline member functions, debugging information, and the
10810 internal tables used to implement virtual functions are all generated in
10811 implementation files.
10813 @cindex implied @code{#pragma implementation}
10814 @cindex @code{#pragma implementation}, implied
10815 @cindex naming convention, implementation headers
10816 If you use @samp{#pragma implementation} with no argument, it applies to
10817 an include file with the same basename@footnote{A file's @dfn{basename}
10818 was the name stripped of all leading path information and of trailing
10819 suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
10820 file. For example, in @file{allclass.cc}, giving just
10821 @samp{#pragma implementation}
10822 by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
10824 In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
10825 an implementation file whenever you would include it from
10826 @file{allclass.cc} even if you never specified @samp{#pragma
10827 implementation}. This was deemed to be more trouble than it was worth,
10828 however, and disabled.
10830 Use the string argument if you want a single implementation file to
10831 include code from multiple header files. (You must also use
10832 @samp{#include} to include the header file; @samp{#pragma
10833 implementation} only specifies how to use the file---it doesn't actually
10836 There is no way to split up the contents of a single header file into
10837 multiple implementation files.
10840 @cindex inlining and C++ pragmas
10841 @cindex C++ pragmas, effect on inlining
10842 @cindex pragmas in C++, effect on inlining
10843 @samp{#pragma implementation} and @samp{#pragma interface} also have an
10844 effect on function inlining.
10846 If you define a class in a header file marked with @samp{#pragma
10847 interface}, the effect on an inline function defined in that class is
10848 similar to an explicit @code{extern} declaration---the compiler emits
10849 no code at all to define an independent version of the function. Its
10850 definition is used only for inlining with its callers.
10852 @opindex fno-implement-inlines
10853 Conversely, when you include the same header file in a main source file
10854 that declares it as @samp{#pragma implementation}, the compiler emits
10855 code for the function itself; this defines a version of the function
10856 that can be found via pointers (or by callers compiled without
10857 inlining). If all calls to the function can be inlined, you can avoid
10858 emitting the function by compiling with @option{-fno-implement-inlines}.
10859 If any calls were not inlined, you will get linker errors.
10861 @node Template Instantiation
10862 @section Where's the Template?
10863 @cindex template instantiation
10865 C++ templates are the first language feature to require more
10866 intelligence from the environment than one usually finds on a UNIX
10867 system. Somehow the compiler and linker have to make sure that each
10868 template instance occurs exactly once in the executable if it is needed,
10869 and not at all otherwise. There are two basic approaches to this
10870 problem, which are referred to as the Borland model and the Cfront model.
10873 @item Borland model
10874 Borland C++ solved the template instantiation problem by adding the code
10875 equivalent of common blocks to their linker; the compiler emits template
10876 instances in each translation unit that uses them, and the linker
10877 collapses them together. The advantage of this model is that the linker
10878 only has to consider the object files themselves; there is no external
10879 complexity to worry about. This disadvantage is that compilation time
10880 is increased because the template code is being compiled repeatedly.
10881 Code written for this model tends to include definitions of all
10882 templates in the header file, since they must be seen to be
10886 The AT&T C++ translator, Cfront, solved the template instantiation
10887 problem by creating the notion of a template repository, an
10888 automatically maintained place where template instances are stored. A
10889 more modern version of the repository works as follows: As individual
10890 object files are built, the compiler places any template definitions and
10891 instantiations encountered in the repository. At link time, the link
10892 wrapper adds in the objects in the repository and compiles any needed
10893 instances that were not previously emitted. The advantages of this
10894 model are more optimal compilation speed and the ability to use the
10895 system linker; to implement the Borland model a compiler vendor also
10896 needs to replace the linker. The disadvantages are vastly increased
10897 complexity, and thus potential for error; for some code this can be
10898 just as transparent, but in practice it can been very difficult to build
10899 multiple programs in one directory and one program in multiple
10900 directories. Code written for this model tends to separate definitions
10901 of non-inline member templates into a separate file, which should be
10902 compiled separately.
10905 When used with GNU ld version 2.8 or later on an ELF system such as
10906 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
10907 Borland model. On other systems, G++ implements neither automatic
10910 A future version of G++ will support a hybrid model whereby the compiler
10911 will emit any instantiations for which the template definition is
10912 included in the compile, and store template definitions and
10913 instantiation context information into the object file for the rest.
10914 The link wrapper will extract that information as necessary and invoke
10915 the compiler to produce the remaining instantiations. The linker will
10916 then combine duplicate instantiations.
10918 In the mean time, you have the following options for dealing with
10919 template instantiations:
10924 Compile your template-using code with @option{-frepo}. The compiler will
10925 generate files with the extension @samp{.rpo} listing all of the
10926 template instantiations used in the corresponding object files which
10927 could be instantiated there; the link wrapper, @samp{collect2}, will
10928 then update the @samp{.rpo} files to tell the compiler where to place
10929 those instantiations and rebuild any affected object files. The
10930 link-time overhead is negligible after the first pass, as the compiler
10931 will continue to place the instantiations in the same files.
10933 This is your best option for application code written for the Borland
10934 model, as it will just work. Code written for the Cfront model will
10935 need to be modified so that the template definitions are available at
10936 one or more points of instantiation; usually this is as simple as adding
10937 @code{#include <tmethods.cc>} to the end of each template header.
10939 For library code, if you want the library to provide all of the template
10940 instantiations it needs, just try to link all of its object files
10941 together; the link will fail, but cause the instantiations to be
10942 generated as a side effect. Be warned, however, that this may cause
10943 conflicts if multiple libraries try to provide the same instantiations.
10944 For greater control, use explicit instantiation as described in the next
10948 @opindex fno-implicit-templates
10949 Compile your code with @option{-fno-implicit-templates} to disable the
10950 implicit generation of template instances, and explicitly instantiate
10951 all the ones you use. This approach requires more knowledge of exactly
10952 which instances you need than do the others, but it's less
10953 mysterious and allows greater control. You can scatter the explicit
10954 instantiations throughout your program, perhaps putting them in the
10955 translation units where the instances are used or the translation units
10956 that define the templates themselves; you can put all of the explicit
10957 instantiations you need into one big file; or you can create small files
10964 template class Foo<int>;
10965 template ostream& operator <<
10966 (ostream&, const Foo<int>&);
10969 for each of the instances you need, and create a template instantiation
10970 library from those.
10972 If you are using Cfront-model code, you can probably get away with not
10973 using @option{-fno-implicit-templates} when compiling files that don't
10974 @samp{#include} the member template definitions.
10976 If you use one big file to do the instantiations, you may want to
10977 compile it without @option{-fno-implicit-templates} so you get all of the
10978 instances required by your explicit instantiations (but not by any
10979 other files) without having to specify them as well.
10981 G++ has extended the template instantiation syntax given in the ISO
10982 standard to allow forward declaration of explicit instantiations
10983 (with @code{extern}), instantiation of the compiler support data for a
10984 template class (i.e.@: the vtable) without instantiating any of its
10985 members (with @code{inline}), and instantiation of only the static data
10986 members of a template class, without the support data or member
10987 functions (with (@code{static}):
10990 extern template int max (int, int);
10991 inline template class Foo<int>;
10992 static template class Foo<int>;
10996 Do nothing. Pretend G++ does implement automatic instantiation
10997 management. Code written for the Borland model will work fine, but
10998 each translation unit will contain instances of each of the templates it
10999 uses. In a large program, this can lead to an unacceptable amount of code
11003 @node Bound member functions
11004 @section Extracting the function pointer from a bound pointer to member function
11006 @cindex pointer to member function
11007 @cindex bound pointer to member function
11009 In C++, pointer to member functions (PMFs) are implemented using a wide
11010 pointer of sorts to handle all the possible call mechanisms; the PMF
11011 needs to store information about how to adjust the @samp{this} pointer,
11012 and if the function pointed to is virtual, where to find the vtable, and
11013 where in the vtable to look for the member function. If you are using
11014 PMFs in an inner loop, you should really reconsider that decision. If
11015 that is not an option, you can extract the pointer to the function that
11016 would be called for a given object/PMF pair and call it directly inside
11017 the inner loop, to save a bit of time.
11019 Note that you will still be paying the penalty for the call through a
11020 function pointer; on most modern architectures, such a call defeats the
11021 branch prediction features of the CPU@. This is also true of normal
11022 virtual function calls.
11024 The syntax for this extension is
11028 extern int (A::*fp)();
11029 typedef int (*fptr)(A *);
11031 fptr p = (fptr)(a.*fp);
11034 For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
11035 no object is needed to obtain the address of the function. They can be
11036 converted to function pointers directly:
11039 fptr p1 = (fptr)(&A::foo);
11042 @opindex Wno-pmf-conversions
11043 You must specify @option{-Wno-pmf-conversions} to use this extension.
11045 @node C++ Attributes
11046 @section C++-Specific Variable, Function, and Type Attributes
11048 Some attributes only make sense for C++ programs.
11051 @item init_priority (@var{priority})
11052 @cindex init_priority attribute
11055 In Standard C++, objects defined at namespace scope are guaranteed to be
11056 initialized in an order in strict accordance with that of their definitions
11057 @emph{in a given translation unit}. No guarantee is made for initializations
11058 across translation units. However, GNU C++ allows users to control the
11059 order of initialization of objects defined at namespace scope with the
11060 @code{init_priority} attribute by specifying a relative @var{priority},
11061 a constant integral expression currently bounded between 101 and 65535
11062 inclusive. Lower numbers indicate a higher priority.
11064 In the following example, @code{A} would normally be created before
11065 @code{B}, but the @code{init_priority} attribute has reversed that order:
11068 Some_Class A __attribute__ ((init_priority (2000)));
11069 Some_Class B __attribute__ ((init_priority (543)));
11073 Note that the particular values of @var{priority} do not matter; only their
11076 @item java_interface
11077 @cindex java_interface attribute
11079 This type attribute informs C++ that the class is a Java interface. It may
11080 only be applied to classes declared within an @code{extern "Java"} block.
11081 Calls to methods declared in this interface will be dispatched using GCJ's
11082 interface table mechanism, instead of regular virtual table dispatch.
11086 See also @xref{Namespace Association}.
11088 @node Namespace Association
11089 @section Namespace Association
11091 @strong{Caution:} The semantics of this extension are not fully
11092 defined. Users should refrain from using this extension as its
11093 semantics may change subtly over time. It is possible that this
11094 extension will be removed in future versions of G++.
11096 A using-directive with @code{__attribute ((strong))} is stronger
11097 than a normal using-directive in two ways:
11101 Templates from the used namespace can be specialized and explicitly
11102 instantiated as though they were members of the using namespace.
11105 The using namespace is considered an associated namespace of all
11106 templates in the used namespace for purposes of argument-dependent
11110 The used namespace must be nested within the using namespace so that
11111 normal unqualified lookup works properly.
11113 This is useful for composing a namespace transparently from
11114 implementation namespaces. For example:
11119 template <class T> struct A @{ @};
11121 using namespace debug __attribute ((__strong__));
11122 template <> struct A<int> @{ @}; // @r{ok to specialize}
11124 template <class T> void f (A<T>);
11129 f (std::A<float>()); // @r{lookup finds} std::f
11134 @node Java Exceptions
11135 @section Java Exceptions
11137 The Java language uses a slightly different exception handling model
11138 from C++. Normally, GNU C++ will automatically detect when you are
11139 writing C++ code that uses Java exceptions, and handle them
11140 appropriately. However, if C++ code only needs to execute destructors
11141 when Java exceptions are thrown through it, GCC will guess incorrectly.
11142 Sample problematic code is:
11145 struct S @{ ~S(); @};
11146 extern void bar(); // @r{is written in Java, and may throw exceptions}
11155 The usual effect of an incorrect guess is a link failure, complaining of
11156 a missing routine called @samp{__gxx_personality_v0}.
11158 You can inform the compiler that Java exceptions are to be used in a
11159 translation unit, irrespective of what it might think, by writing
11160 @samp{@w{#pragma GCC java_exceptions}} at the head of the file. This
11161 @samp{#pragma} must appear before any functions that throw or catch
11162 exceptions, or run destructors when exceptions are thrown through them.
11164 You cannot mix Java and C++ exceptions in the same translation unit. It
11165 is believed to be safe to throw a C++ exception from one file through
11166 another file compiled for the Java exception model, or vice versa, but
11167 there may be bugs in this area.
11169 @node Deprecated Features
11170 @section Deprecated Features
11172 In the past, the GNU C++ compiler was extended to experiment with new
11173 features, at a time when the C++ language was still evolving. Now that
11174 the C++ standard is complete, some of those features are superseded by
11175 superior alternatives. Using the old features might cause a warning in
11176 some cases that the feature will be dropped in the future. In other
11177 cases, the feature might be gone already.
11179 While the list below is not exhaustive, it documents some of the options
11180 that are now deprecated:
11183 @item -fexternal-templates
11184 @itemx -falt-external-templates
11185 These are two of the many ways for G++ to implement template
11186 instantiation. @xref{Template Instantiation}. The C++ standard clearly
11187 defines how template definitions have to be organized across
11188 implementation units. G++ has an implicit instantiation mechanism that
11189 should work just fine for standard-conforming code.
11191 @item -fstrict-prototype
11192 @itemx -fno-strict-prototype
11193 Previously it was possible to use an empty prototype parameter list to
11194 indicate an unspecified number of parameters (like C), rather than no
11195 parameters, as C++ demands. This feature has been removed, except where
11196 it is required for backwards compatibility @xref{Backwards Compatibility}.
11199 G++ allows a virtual function returning @samp{void *} to be overridden
11200 by one returning a different pointer type. This extension to the
11201 covariant return type rules is now deprecated and will be removed from a
11204 The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and
11205 their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated
11206 and will be removed in a future version. Code using these operators
11207 should be modified to use @code{std::min} and @code{std::max} instead.
11209 The named return value extension has been deprecated, and is now
11212 The use of initializer lists with new expressions has been deprecated,
11213 and is now removed from G++.
11215 Floating and complex non-type template parameters have been deprecated,
11216 and are now removed from G++.
11218 The implicit typename extension has been deprecated and is now
11221 The use of default arguments in function pointers, function typedefs
11222 and other places where they are not permitted by the standard is
11223 deprecated and will be removed from a future version of G++.
11225 G++ allows floating-point literals to appear in integral constant expressions,
11226 e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} }
11227 This extension is deprecated and will be removed from a future version.
11229 G++ allows static data members of const floating-point type to be declared
11230 with an initializer in a class definition. The standard only allows
11231 initializers for static members of const integral types and const
11232 enumeration types so this extension has been deprecated and will be removed
11233 from a future version.
11235 @node Backwards Compatibility
11236 @section Backwards Compatibility
11237 @cindex Backwards Compatibility
11238 @cindex ARM [Annotated C++ Reference Manual]
11240 Now that there is a definitive ISO standard C++, G++ has a specification
11241 to adhere to. The C++ language evolved over time, and features that
11242 used to be acceptable in previous drafts of the standard, such as the ARM
11243 [Annotated C++ Reference Manual], are no longer accepted. In order to allow
11244 compilation of C++ written to such drafts, G++ contains some backwards
11245 compatibilities. @emph{All such backwards compatibility features are
11246 liable to disappear in future versions of G++.} They should be considered
11247 deprecated @xref{Deprecated Features}.
11251 If a variable is declared at for scope, it used to remain in scope until
11252 the end of the scope which contained the for statement (rather than just
11253 within the for scope). G++ retains this, but issues a warning, if such a
11254 variable is accessed outside the for scope.
11256 @item Implicit C language
11257 Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
11258 scope to set the language. On such systems, all header files are
11259 implicitly scoped inside a C language scope. Also, an empty prototype
11260 @code{()} will be treated as an unspecified number of arguments, rather
11261 than no arguments, as C++ demands.