view gcc/doc/extend.texi @ 158:494b0b89df80 default tip

...
author Shinji KONO <kono@ie.u-ryukyu.ac.jp>
date Mon, 25 May 2020 18:13:55 +0900
parents 1830386684a0
children
line wrap: on
line source

c Copyright (C) 1988-2020 Free Software Foundation, Inc.

@c This is part of the GCC manual.
@c For copying conditions, see the file gcc.texi.

@node C Extensions
@chapter Extensions to the C Language Family
@cindex extensions, C language
@cindex C language extensions

@opindex pedantic
GNU C provides several language features not found in ISO standard C@.
(The @option{-pedantic} option directs GCC to print a warning message if
any of these features is used.)  To test for the availability of these
features in conditional compilation, check for a predefined macro
@code{__GNUC__}, which is always defined under GCC@.

These extensions are available in C and Objective-C@.  Most of them are
also available in C++.  @xref{C++ Extensions,,Extensions to the
C++ Language}, for extensions that apply @emph{only} to C++.

Some features that are in ISO C99 but not C90 or C++ are also, as
extensions, accepted by GCC in C90 mode and in C++.

@menu
* Statement Exprs::     Putting statements and declarations inside expressions.
* Local Labels::        Labels local to a block.
* Labels as Values::    Getting pointers to labels, and computed gotos.
* Nested Functions::    Nested function in GNU C.
* Nonlocal Gotos::      Nonlocal gotos.
* Constructing Calls::  Dispatching a call to another function.
* Typeof::              @code{typeof}: referring to the type of an expression.
* Conditionals::        Omitting the middle operand of a @samp{?:} expression.
* __int128::		128-bit integers---@code{__int128}.
* Long Long::           Double-word integers---@code{long long int}.
* Complex::             Data types for complex numbers.
* Floating Types::      Additional Floating Types.
* Half-Precision::      Half-Precision Floating Point.
* Decimal Float::       Decimal Floating Types.
* Hex Floats::          Hexadecimal floating-point constants.
* Fixed-Point::         Fixed-Point Types.
* Named Address Spaces::Named address spaces.
* Zero Length::         Zero-length arrays.
* Empty Structures::    Structures with no members.
* Variable Length::     Arrays whose length is computed at run time.
* Variadic Macros::     Macros with a variable number of arguments.
* Escaped Newlines::    Slightly looser rules for escaped newlines.
* Subscripting::        Any array can be subscripted, even if not an lvalue.
* Pointer Arith::       Arithmetic on @code{void}-pointers and function pointers.
* Variadic Pointer Args::  Pointer arguments to variadic functions.
* Pointers to Arrays::  Pointers to arrays with qualifiers work as expected.
* Initializers::        Non-constant initializers.
* Compound Literals::   Compound literals give structures, unions
                        or arrays as values.
* Designated Inits::    Labeling elements of initializers.
* Case Ranges::         `case 1 ... 9' and such.
* Cast to Union::       Casting to union type from any member of the union.
* Mixed Declarations::  Mixing declarations and code.
* Function Attributes:: Declaring that functions have no side effects,
                        or that they can never return.
* Variable Attributes:: Specifying attributes of variables.
* Type Attributes::     Specifying attributes of types.
* Label Attributes::    Specifying attributes on labels.
* Enumerator Attributes:: Specifying attributes on enumerators.
* Statement Attributes:: Specifying attributes on statements.
* Attribute Syntax::    Formal syntax for attributes.
* Function Prototypes:: Prototype declarations and old-style definitions.
* C++ Comments::        C++ comments are recognized.
* Dollar Signs::        Dollar sign is allowed in identifiers.
* Character Escapes::   @samp{\e} stands for the character @key{ESC}.
* Alignment::           Determining the alignment of a function, type or variable.
* Inline::              Defining inline functions (as fast as macros).
* Volatiles::           What constitutes an access to a volatile object.
* Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler.
* Alternate Keywords::  @code{__const__}, @code{__asm__}, etc., for header files.
* Incomplete Enums::    @code{enum foo;}, with details to follow.
* Function Names::      Printable strings which are the name of the current
                        function.
* Return Address::      Getting the return or frame address of a function.
* Vector Extensions::   Using vector instructions through built-in functions.
* Offsetof::            Special syntax for implementing @code{offsetof}.
* __sync Builtins::     Legacy built-in functions for atomic memory access.
* __atomic Builtins::   Atomic built-in functions with memory model.
* Integer Overflow Builtins:: Built-in functions to perform arithmetics and
                        arithmetic overflow checking.
* x86 specific memory model extensions for transactional memory:: x86 memory models.
* Object Size Checking:: Built-in functions for limited buffer overflow
                        checking.
* Other Builtins::      Other built-in functions.
* Target Builtins::     Built-in functions specific to particular targets.
* Target Format Checks:: Format checks specific to particular targets.
* Pragmas::             Pragmas accepted by GCC.
* Unnamed Fields::      Unnamed struct/union fields within structs/unions.
* Thread-Local::        Per-thread variables.
* Binary constants::    Binary constants using the @samp{0b} prefix.
@end menu

@node Statement Exprs
@section Statements and Declarations in Expressions
@cindex statements inside expressions
@cindex declarations inside expressions
@cindex expressions containing statements
@cindex macros, statements in expressions

@c the above section title wrapped and causes an underfull hbox.. i
@c changed it from "within" to "in". --mew 4feb93
A compound statement enclosed in parentheses may appear as an expression
in GNU C@.  This allows you to use loops, switches, and local variables
within an expression.

Recall that a compound statement is a sequence of statements surrounded
by braces; in this construct, parentheses go around the braces.  For
example:

@smallexample
(@{ int y = foo (); int z;
   if (y > 0) z = y;
   else z = - y;
   z; @})
@end smallexample

@noindent
is a valid (though slightly more complex than necessary) expression
for the absolute value of @code{foo ()}.

The last thing in the compound statement should be an expression
followed by a semicolon; the value of this subexpression serves as the
value of the entire construct.  (If you use some other kind of statement
last within the braces, the construct has type @code{void}, and thus
effectively no value.)

This feature is especially useful in making macro definitions ``safe'' (so
that they evaluate each operand exactly once).  For example, the
``maximum'' function is commonly defined as a macro in standard C as
follows:

@smallexample
#define max(a,b) ((a) > (b) ? (a) : (b))
@end smallexample

@noindent
@cindex side effects, macro argument
But this definition computes either @var{a} or @var{b} twice, with bad
results if the operand has side effects.  In GNU C, if you know the
type of the operands (here taken as @code{int}), you can avoid this
problem by defining the macro as follows:

@smallexample
#define maxint(a,b) \
  (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
@end smallexample

Note that introducing variable declarations (as we do in @code{maxint}) can
cause variable shadowing, so while this example using the @code{max} macro
produces correct results:
@smallexample
int _a = 1, _b = 2, c;
c = max (_a, _b);
@end smallexample
@noindent
this example using maxint will not:
@smallexample
int _a = 1, _b = 2, c;
c = maxint (_a, _b);
@end smallexample

This problem may for instance occur when we use this pattern recursively, like
so:

@smallexample
#define maxint3(a, b, c) \
  (@{int _a = (a), _b = (b), _c = (c); maxint (maxint (_a, _b), _c); @})
@end smallexample

Embedded statements are not allowed in constant expressions, such as
the value of an enumeration constant, the width of a bit-field, or
the initial value of a static variable.

If you don't know the type of the operand, you can still do this, but you
must use @code{typeof} or @code{__auto_type} (@pxref{Typeof}).

In G++, the result value of a statement expression undergoes array and
function pointer decay, and is returned by value to the enclosing
expression.  For instance, if @code{A} is a class, then

@smallexample
        A a;

        (@{a;@}).Foo ()
@end smallexample

@noindent
constructs a temporary @code{A} object to hold the result of the
statement expression, and that is used to invoke @code{Foo}.
Therefore the @code{this} pointer observed by @code{Foo} is not the
address of @code{a}.

In a statement expression, any temporaries created within a statement
are destroyed at that statement's end.  This makes statement
expressions inside macros slightly different from function calls.  In
the latter case temporaries introduced during argument evaluation are
destroyed at the end of the statement that includes the function
call.  In the statement expression case they are destroyed during
the statement expression.  For instance,

@smallexample
#define macro(a)  (@{__typeof__(a) b = (a); b + 3; @})
template<typename T> T function(T a) @{ T b = a; return b + 3; @}

void foo ()
@{
  macro (X ());
  function (X ());
@}
@end smallexample

@noindent
has different places where temporaries are destroyed.  For the
@code{macro} case, the temporary @code{X} is destroyed just after
the initialization of @code{b}.  In the @code{function} case that
temporary is destroyed when the function returns.

These considerations mean that it is probably a bad idea to use
statement expressions of this form in header files that are designed to
work with C++.  (Note that some versions of the GNU C Library contained
header files using statement expressions that lead to precisely this
bug.)

Jumping into a statement expression with @code{goto} or using a
@code{switch} statement outside the statement expression with a
@code{case} or @code{default} label inside the statement expression is
not permitted.  Jumping into a statement expression with a computed
@code{goto} (@pxref{Labels as Values}) has undefined behavior.
Jumping out of a statement expression is permitted, but if the
statement expression is part of a larger expression then it is
unspecified which other subexpressions of that expression have been
evaluated except where the language definition requires certain
subexpressions to be evaluated before or after the statement
expression.  A @code{break} or @code{continue} statement inside of
a statement expression used in @code{while}, @code{do} or @code{for}
loop or @code{switch} statement condition
or @code{for} statement init or increment expressions jumps to an
outer loop or @code{switch} statement if any (otherwise it is an error),
rather than to the loop or @code{switch} statement in whose condition
or init or increment expression it appears.
In any case, as with a function call, the evaluation of a
statement expression is not interleaved with the evaluation of other
parts of the containing expression.  For example,

@smallexample
  foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz();
@end smallexample

@noindent
calls @code{foo} and @code{bar1} and does not call @code{baz} but
may or may not call @code{bar2}.  If @code{bar2} is called, it is
called after @code{foo} and before @code{bar1}.

@node Local Labels
@section Locally Declared Labels
@cindex local labels
@cindex macros, local labels

GCC allows you to declare @dfn{local labels} in any nested block
scope.  A local label is just like an ordinary label, but you can
only reference it (with a @code{goto} statement, or by taking its
address) within the block in which it is declared.

A local label declaration looks like this:

@smallexample
__label__ @var{label};
@end smallexample

@noindent
or

@smallexample
__label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
@end smallexample

Local label declarations must come at the beginning of the block,
before any ordinary declarations or statements.

The label declaration defines the label @emph{name}, but does not define
the label itself.  You must do this in the usual way, with
@code{@var{label}:}, within the statements of the statement expression.

The local label feature is useful for complex macros.  If a macro
contains nested loops, a @code{goto} can be useful for breaking out of
them.  However, an ordinary label whose scope is the whole function
cannot be used: if the macro can be expanded several times in one
function, the label is multiply defined in that function.  A
local label avoids this problem.  For example:

@smallexample
#define SEARCH(value, array, target)              \
do @{                                              \
  __label__ found;                                \
  typeof (target) _SEARCH_target = (target);      \
  typeof (*(array)) *_SEARCH_array = (array);     \
  int i, j;                                       \
  int value;                                      \
  for (i = 0; i < max; i++)                       \
    for (j = 0; j < max; j++)                     \
      if (_SEARCH_array[i][j] == _SEARCH_target)  \
        @{ (value) = i; goto found; @}              \
  (value) = -1;                                   \
 found:;                                          \
@} while (0)
@end smallexample

This could also be written using a statement expression:

@smallexample
#define SEARCH(array, target)                     \
(@{                                                \
  __label__ found;                                \
  typeof (target) _SEARCH_target = (target);      \
  typeof (*(array)) *_SEARCH_array = (array);     \
  int i, j;                                       \
  int value;                                      \
  for (i = 0; i < max; i++)                       \
    for (j = 0; j < max; j++)                     \
      if (_SEARCH_array[i][j] == _SEARCH_target)  \
        @{ value = i; goto found; @}                \
  value = -1;                                     \
 found:                                           \
  value;                                          \
@})
@end smallexample

Local label declarations also make the labels they declare visible to
nested functions, if there are any.  @xref{Nested Functions}, for details.

@node Labels as Values
@section Labels as Values
@cindex labels as values
@cindex computed gotos
@cindex goto with computed label
@cindex address of a label

You can get the address of a label defined in the current function
(or a containing function) with the unary operator @samp{&&}.  The
value has type @code{void *}.  This value is a constant and can be used
wherever a constant of that type is valid.  For example:

@smallexample
void *ptr;
/* @r{@dots{}} */
ptr = &&foo;
@end smallexample

To use these values, you need to be able to jump to one.  This is done
with the computed goto statement@footnote{The analogous feature in
Fortran is called an assigned goto, but that name seems inappropriate in
C, where one can do more than simply store label addresses in label
variables.}, @code{goto *@var{exp};}.  For example,

@smallexample
goto *ptr;
@end smallexample

@noindent
Any expression of type @code{void *} is allowed.

One way of using these constants is in initializing a static array that
serves as a jump table:

@smallexample
static void *array[] = @{ &&foo, &&bar, &&hack @};
@end smallexample

@noindent
Then you can select a label with indexing, like this:

@smallexample
goto *array[i];
@end smallexample

@noindent
Note that this does not check whether the subscript is in bounds---array
indexing in C never does that.

Such an array of label values serves a purpose much like that of the
@code{switch} statement.  The @code{switch} statement is cleaner, so
use that rather than an array unless the problem does not fit a
@code{switch} statement very well.

Another use of label values is in an interpreter for threaded code.
The labels within the interpreter function can be stored in the
threaded code for super-fast dispatching.

You may not use this mechanism to jump to code in a different function.
If you do that, totally unpredictable things happen.  The best way to
avoid this is to store the label address only in automatic variables and
never pass it as an argument.

An alternate way to write the above example is

@smallexample
static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
                             &&hack - &&foo @};
goto *(&&foo + array[i]);
@end smallexample

@noindent
This is more friendly to code living in shared libraries, as it reduces
the number of dynamic relocations that are needed, and by consequence,
allows the data to be read-only.
This alternative with label differences is not supported for the AVR target,
please use the first approach for AVR programs.

The @code{&&foo} expressions for the same label might have different
values if the containing function is inlined or cloned.  If a program
relies on them being always the same,
@code{__attribute__((__noinline__,__noclone__))} should be used to
prevent inlining and cloning.  If @code{&&foo} is used in a static
variable initializer, inlining and cloning is forbidden.

@node Nested Functions
@section Nested Functions
@cindex nested functions
@cindex downward funargs
@cindex thunks

A @dfn{nested function} is a function defined inside another function.
Nested functions are supported as an extension in GNU C, but are not
supported by GNU C++.

The nested function's name is local to the block where it is defined.
For example, here we define a nested function named @code{square}, and
call it twice:

@smallexample
@group
foo (double a, double b)
@{
  double square (double z) @{ return z * z; @}

  return square (a) + square (b);
@}
@end group
@end smallexample

The nested function can access all the variables of the containing
function that are visible at the point of its definition.  This is
called @dfn{lexical scoping}.  For example, here we show a nested
function which uses an inherited variable named @code{offset}:

@smallexample
@group
bar (int *array, int offset, int size)
@{
  int access (int *array, int index)
    @{ return array[index + offset]; @}
  int i;
  /* @r{@dots{}} */
  for (i = 0; i < size; i++)
    /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
@}
@end group
@end smallexample

Nested function definitions are permitted within functions in the places
where variable definitions are allowed; that is, in any block, mixed
with the other declarations and statements in the block.

It is possible to call the nested function from outside the scope of its
name by storing its address or passing the address to another function:

@smallexample
hack (int *array, int size)
@{
  void store (int index, int value)
    @{ array[index] = value; @}

  intermediate (store, size);
@}
@end smallexample

Here, the function @code{intermediate} receives the address of
@code{store} as an argument.  If @code{intermediate} calls @code{store},
the arguments given to @code{store} are used to store into @code{array}.
But this technique works only so long as the containing function
(@code{hack}, in this example) does not exit.

If you try to call the nested function through its address after the
containing function exits, all hell breaks loose.  If you try
to call it after a containing scope level exits, and if it refers
to some of the variables that are no longer in scope, you may be lucky,
but it's not wise to take the risk.  If, however, the nested function
does not refer to anything that has gone out of scope, you should be
safe.

GCC implements taking the address of a nested function using a technique
called @dfn{trampolines}.  This technique was described in
@cite{Lexical Closures for C++} (Thomas M. Breuel, USENIX
C++ Conference Proceedings, October 17-21, 1988).

A nested function can jump to a label inherited from a containing
function, provided the label is explicitly declared in the containing
function (@pxref{Local Labels}).  Such a jump returns instantly to the
containing function, exiting the nested function that did the
@code{goto} and any intermediate functions as well.  Here is an example:

@smallexample
@group
bar (int *array, int offset, int size)
@{
  __label__ failure;
  int access (int *array, int index)
    @{
      if (index > size)
        goto failure;
      return array[index + offset];
    @}
  int i;
  /* @r{@dots{}} */
  for (i = 0; i < size; i++)
    /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
  /* @r{@dots{}} */
  return 0;

 /* @r{Control comes here from @code{access}
    if it detects an error.}  */
 failure:
  return -1;
@}
@end group
@end smallexample

A nested function always has no linkage.  Declaring one with
@code{extern} or @code{static} is erroneous.  If you need to declare the nested function
before its definition, use @code{auto} (which is otherwise meaningless
for function declarations).

@smallexample
bar (int *array, int offset, int size)
@{
  __label__ failure;
  auto int access (int *, int);
  /* @r{@dots{}} */
  int access (int *array, int index)
    @{
      if (index > size)
        goto failure;
      return array[index + offset];
    @}
  /* @r{@dots{}} */
@}
@end smallexample

@node Nonlocal Gotos
@section Nonlocal Gotos
@cindex nonlocal gotos

GCC provides the built-in functions @code{__builtin_setjmp} and
@code{__builtin_longjmp} which are similar to, but not interchangeable
with, the C library functions @code{setjmp} and @code{longjmp}.  
The built-in versions are used internally by GCC's libraries
to implement exception handling on some targets.  You should use the 
standard C library functions declared in @code{<setjmp.h>} in user code
instead of the builtins.

The built-in versions of these functions use GCC's normal
mechanisms to save and restore registers using the stack on function
entry and exit.  The jump buffer argument @var{buf} holds only the
information needed to restore the stack frame, rather than the entire 
set of saved register values.  

An important caveat is that GCC arranges to save and restore only
those registers known to the specific architecture variant being
compiled for.  This can make @code{__builtin_setjmp} and
@code{__builtin_longjmp} more efficient than their library
counterparts in some cases, but it can also cause incorrect and
mysterious behavior when mixing with code that uses the full register
set.

You should declare the jump buffer argument @var{buf} to the
built-in functions as:

@smallexample
#include <stdint.h>
intptr_t @var{buf}[5];
@end smallexample

@deftypefn {Built-in Function} {int} __builtin_setjmp (intptr_t *@var{buf})
This function saves the current stack context in @var{buf}.  
@code{__builtin_setjmp} returns 0 when returning directly,
and 1 when returning from @code{__builtin_longjmp} using the same
@var{buf}.
@end deftypefn

@deftypefn {Built-in Function} {void} __builtin_longjmp (intptr_t *@var{buf}, int @var{val})
This function restores the stack context in @var{buf}, 
saved by a previous call to @code{__builtin_setjmp}.  After
@code{__builtin_longjmp} is finished, the program resumes execution as
if the matching @code{__builtin_setjmp} returns the value @var{val},
which must be 1.

Because @code{__builtin_longjmp} depends on the function return
mechanism to restore the stack context, it cannot be called
from the same function calling @code{__builtin_setjmp} to
initialize @var{buf}.  It can only be called from a function called
(directly or indirectly) from the function calling @code{__builtin_setjmp}.
@end deftypefn

@node Constructing Calls
@section Constructing Function Calls
@cindex constructing calls
@cindex forwarding calls

Using the built-in functions described below, you can record
the arguments a function received, and call another function
with the same arguments, without knowing the number or types
of the arguments.

You can also record the return value of that function call,
and later return that value, without knowing what data type
the function tried to return (as long as your caller expects
that data type).

However, these built-in functions may interact badly with some
sophisticated features or other extensions of the language.  It
is, therefore, not recommended to use them outside very simple
functions acting as mere forwarders for their arguments.

@deftypefn {Built-in Function} {void *} __builtin_apply_args ()
This built-in function returns a pointer to data
describing how to perform a call with the same arguments as are passed
to the current function.

The function saves the arg pointer register, structure value address,
and all registers that might be used to pass arguments to a function
into a block of memory allocated on the stack.  Then it returns the
address of that block.
@end deftypefn

@deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
This built-in function invokes @var{function}
with a copy of the parameters described by @var{arguments}
and @var{size}.

The value of @var{arguments} should be the value returned by
@code{__builtin_apply_args}.  The argument @var{size} specifies the size
of the stack argument data, in bytes.

This function returns a pointer to data describing
how to return whatever value is returned by @var{function}.  The data
is saved in a block of memory allocated on the stack.

It is not always simple to compute the proper value for @var{size}.  The
value is used by @code{__builtin_apply} to compute the amount of data
that should be pushed on the stack and copied from the incoming argument
area.
@end deftypefn

@deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
This built-in function returns the value described by @var{result} from
the containing function.  You should specify, for @var{result}, a value
returned by @code{__builtin_apply}.
@end deftypefn

@deftypefn {Built-in Function} {} __builtin_va_arg_pack ()
This built-in function represents all anonymous arguments of an inline
function.  It can be used only in inline functions that are always
inlined, never compiled as a separate function, such as those using
@code{__attribute__ ((__always_inline__))} or
@code{__attribute__ ((__gnu_inline__))} extern inline functions.
It must be only passed as last argument to some other function
with variable arguments.  This is useful for writing small wrapper
inlines for variable argument functions, when using preprocessor
macros is undesirable.  For example:
@smallexample
extern int myprintf (FILE *f, const char *format, ...);
extern inline __attribute__ ((__gnu_inline__)) int
myprintf (FILE *f, const char *format, ...)
@{
  int r = fprintf (f, "myprintf: ");
  if (r < 0)
    return r;
  int s = fprintf (f, format, __builtin_va_arg_pack ());
  if (s < 0)
    return s;
  return r + s;
@}
@end smallexample
@end deftypefn

@deftypefn {Built-in Function} {size_t} __builtin_va_arg_pack_len ()
This built-in function returns the number of anonymous arguments of
an inline function.  It can be used only in inline functions that
are always inlined, never compiled as a separate function, such
as those using @code{__attribute__ ((__always_inline__))} or
@code{__attribute__ ((__gnu_inline__))} extern inline functions.
For example following does link- or run-time checking of open
arguments for optimized code:
@smallexample
#ifdef __OPTIMIZE__
extern inline __attribute__((__gnu_inline__)) int
myopen (const char *path, int oflag, ...)
@{
  if (__builtin_va_arg_pack_len () > 1)
    warn_open_too_many_arguments ();

  if (__builtin_constant_p (oflag))
    @{
      if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1)
        @{
          warn_open_missing_mode ();
          return __open_2 (path, oflag);
        @}
      return open (path, oflag, __builtin_va_arg_pack ());
    @}

  if (__builtin_va_arg_pack_len () < 1)
    return __open_2 (path, oflag);

  return open (path, oflag, __builtin_va_arg_pack ());
@}
#endif
@end smallexample
@end deftypefn

@node Typeof
@section Referring to a Type with @code{typeof}
@findex typeof
@findex sizeof
@cindex macros, types of arguments

Another way to refer to the type of an expression is with @code{typeof}.
The syntax of using of this keyword looks like @code{sizeof}, but the
construct acts semantically like a type name defined with @code{typedef}.

There are two ways of writing the argument to @code{typeof}: with an
expression or with a type.  Here is an example with an expression:

@smallexample
typeof (x[0](1))
@end smallexample

@noindent
This assumes that @code{x} is an array of pointers to functions;
the type described is that of the values of the functions.

Here is an example with a typename as the argument:

@smallexample
typeof (int *)
@end smallexample

@noindent
Here the type described is that of pointers to @code{int}.

If you are writing a header file that must work when included in ISO C
programs, write @code{__typeof__} instead of @code{typeof}.
@xref{Alternate Keywords}.

A @code{typeof} construct can be used anywhere a typedef name can be
used.  For example, you can use it in a declaration, in a cast, or inside
of @code{sizeof} or @code{typeof}.

The operand of @code{typeof} is evaluated for its side effects if and
only if it is an expression of variably modified type or the name of
such a type.

@code{typeof} is often useful in conjunction with
statement expressions (@pxref{Statement Exprs}).
Here is how the two together can
be used to define a safe ``maximum'' macro which operates on any
arithmetic type and evaluates each of its arguments exactly once:

@smallexample
#define max(a,b) \
  (@{ typeof (a) _a = (a); \
      typeof (b) _b = (b); \
    _a > _b ? _a : _b; @})
@end smallexample

@cindex underscores in variables in macros
@cindex @samp{_} in variables in macros
@cindex local variables in macros
@cindex variables, local, in macros
@cindex macros, local variables in

The reason for using names that start with underscores for the local
variables is to avoid conflicts with variable names that occur within the
expressions that are substituted for @code{a} and @code{b}.  Eventually we
hope to design a new form of declaration syntax that allows you to declare
variables whose scopes start only after their initializers; this will be a
more reliable way to prevent such conflicts.

@noindent
Some more examples of the use of @code{typeof}:

@itemize @bullet
@item
This declares @code{y} with the type of what @code{x} points to.

@smallexample
typeof (*x) y;
@end smallexample

@item
This declares @code{y} as an array of such values.

@smallexample
typeof (*x) y[4];
@end smallexample

@item
This declares @code{y} as an array of pointers to characters:

@smallexample
typeof (typeof (char *)[4]) y;
@end smallexample

@noindent
It is equivalent to the following traditional C declaration:

@smallexample
char *y[4];
@end smallexample

To see the meaning of the declaration using @code{typeof}, and why it
might be a useful way to write, rewrite it with these macros:

@smallexample
#define pointer(T)  typeof(T *)
#define array(T, N) typeof(T [N])
@end smallexample

@noindent
Now the declaration can be rewritten this way:

@smallexample
array (pointer (char), 4) y;
@end smallexample

@noindent
Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
pointers to @code{char}.
@end itemize

In GNU C, but not GNU C++, you may also declare the type of a variable
as @code{__auto_type}.  In that case, the declaration must declare
only one variable, whose declarator must just be an identifier, the
declaration must be initialized, and the type of the variable is
determined by the initializer; the name of the variable is not in
scope until after the initializer.  (In C++, you should use C++11
@code{auto} for this purpose.)  Using @code{__auto_type}, the
``maximum'' macro above could be written as:

@smallexample
#define max(a,b) \
  (@{ __auto_type _a = (a); \
      __auto_type _b = (b); \
    _a > _b ? _a : _b; @})
@end smallexample

Using @code{__auto_type} instead of @code{typeof} has two advantages:

@itemize @bullet
@item Each argument to the macro appears only once in the expansion of
the macro.  This prevents the size of the macro expansion growing
exponentially when calls to such macros are nested inside arguments of
such macros.

@item If the argument to the macro has variably modified type, it is
evaluated only once when using @code{__auto_type}, but twice if
@code{typeof} is used.
@end itemize

@node Conditionals
@section Conditionals with Omitted Operands
@cindex conditional expressions, extensions
@cindex omitted middle-operands
@cindex middle-operands, omitted
@cindex extensions, @code{?:}
@cindex @code{?:} extensions

The middle operand in a conditional expression may be omitted.  Then
if the first operand is nonzero, its value is the value of the conditional
expression.

Therefore, the expression

@smallexample
x ? : y
@end smallexample

@noindent
has the value of @code{x} if that is nonzero; otherwise, the value of
@code{y}.

This example is perfectly equivalent to

@smallexample
x ? x : y
@end smallexample

@cindex side effect in @code{?:}
@cindex @code{?:} side effect
@noindent
In this simple case, the ability to omit the middle operand is not
especially useful.  When it becomes useful is when the first operand does,
or may (if it is a macro argument), contain a side effect.  Then repeating
the operand in the middle would perform the side effect twice.  Omitting
the middle operand uses the value already computed without the undesirable
effects of recomputing it.

@node __int128
@section 128-bit Integers
@cindex @code{__int128} data types

As an extension the integer scalar type @code{__int128} is supported for
targets which have an integer mode wide enough to hold 128 bits.
Simply write @code{__int128} for a signed 128-bit integer, or
@code{unsigned __int128} for an unsigned 128-bit integer.  There is no
support in GCC for expressing an integer constant of type @code{__int128}
for targets with @code{long long} integer less than 128 bits wide.

@node Long Long
@section Double-Word Integers
@cindex @code{long long} data types
@cindex double-word arithmetic
@cindex multiprecision arithmetic
@cindex @code{LL} integer suffix
@cindex @code{ULL} integer suffix

ISO C99 and ISO C++11 support data types for integers that are at least
64 bits wide, and as an extension GCC supports them in C90 and C++98 modes.
Simply write @code{long long int} for a signed integer, or
@code{unsigned long long int} for an unsigned integer.  To make an
integer constant of type @code{long long int}, add the suffix @samp{LL}
to the integer.  To make an integer constant of type @code{unsigned long
long int}, add the suffix @samp{ULL} to the integer.

You can use these types in arithmetic like any other integer types.
Addition, subtraction, and bitwise boolean operations on these types
are open-coded on all types of machines.  Multiplication is open-coded
if the machine supports a fullword-to-doubleword widening multiply
instruction.  Division and shifts are open-coded only on machines that
provide special support.  The operations that are not open-coded use
special library routines that come with GCC@.

There may be pitfalls when you use @code{long long} types for function
arguments without function prototypes.  If a function
expects type @code{int} for its argument, and you pass a value of type
@code{long long int}, confusion results because the caller and the
subroutine disagree about the number of bytes for the argument.
Likewise, if the function expects @code{long long int} and you pass
@code{int}.  The best way to avoid such problems is to use prototypes.

@node Complex
@section Complex Numbers
@cindex complex numbers
@cindex @code{_Complex} keyword
@cindex @code{__complex__} keyword

ISO C99 supports complex floating data types, and as an extension GCC
supports them in C90 mode and in C++.  GCC also supports complex integer data
types which are not part of ISO C99.  You can declare complex types
using the keyword @code{_Complex}.  As an extension, the older GNU
keyword @code{__complex__} is also supported.

For example, @samp{_Complex double x;} declares @code{x} as a
variable whose real part and imaginary part are both of type
@code{double}.  @samp{_Complex short int y;} declares @code{y} to
have real and imaginary parts of type @code{short int}; this is not
likely to be useful, but it shows that the set of complex types is
complete.

To write a constant with a complex data type, use the suffix @samp{i} or
@samp{j} (either one; they are equivalent).  For example, @code{2.5fi}
has type @code{_Complex float} and @code{3i} has type
@code{_Complex int}.  Such a constant always has a pure imaginary
value, but you can form any complex value you like by adding one to a
real constant.  This is a GNU extension; if you have an ISO C99
conforming C library (such as the GNU C Library), and want to construct complex
constants of floating type, you should include @code{<complex.h>} and
use the macros @code{I} or @code{_Complex_I} instead.

The ISO C++14 library also defines the @samp{i} suffix, so C++14 code
that includes the @samp{<complex>} header cannot use @samp{i} for the
GNU extension.  The @samp{j} suffix still has the GNU meaning.

@cindex @code{__real__} keyword
@cindex @code{__imag__} keyword
To extract the real part of a complex-valued expression @var{exp}, write
@code{__real__ @var{exp}}.  Likewise, use @code{__imag__} to
extract the imaginary part.  This is a GNU extension; for values of
floating type, you should use the ISO C99 functions @code{crealf},
@code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
@code{cimagl}, declared in @code{<complex.h>} and also provided as
built-in functions by GCC@.

@cindex complex conjugation
The operator @samp{~} performs complex conjugation when used on a value
with a complex type.  This is a GNU extension; for values of
floating type, you should use the ISO C99 functions @code{conjf},
@code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
provided as built-in functions by GCC@.

GCC can allocate complex automatic variables in a noncontiguous
fashion; it's even possible for the real part to be in a register while
the imaginary part is on the stack (or vice versa).  Only the DWARF
debug info format can represent this, so use of DWARF is recommended.
If you are using the stabs debug info format, GCC describes a noncontiguous
complex variable as if it were two separate variables of noncomplex type.
If the variable's actual name is @code{foo}, the two fictitious
variables are named @code{foo$real} and @code{foo$imag}.  You can
examine and set these two fictitious variables with your debugger.

@node Floating Types
@section Additional Floating Types
@cindex additional floating types
@cindex @code{_Float@var{n}} data types
@cindex @code{_Float@var{n}x} data types
@cindex @code{__float80} data type
@cindex @code{__float128} data type
@cindex @code{__ibm128} data type
@cindex @code{w} floating point suffix
@cindex @code{q} floating point suffix
@cindex @code{W} floating point suffix
@cindex @code{Q} floating point suffix

ISO/IEC TS 18661-3:2015 defines C support for additional floating
types @code{_Float@var{n}} and @code{_Float@var{n}x}, and GCC supports
these type names; the set of types supported depends on the target
architecture.  These types are not supported when compiling C++.
Constants with these types use suffixes @code{f@var{n}} or
@code{F@var{n}} and @code{f@var{n}x} or @code{F@var{n}x}.  These type
names can be used together with @code{_Complex} to declare complex
types.

As an extension, GNU C and GNU C++ support additional floating
types, which are not supported by all targets.
@itemize @bullet
@item @code{__float128} is available on i386, x86_64, IA-64, and
hppa HP-UX, as well as on PowerPC GNU/Linux targets that enable
the vector scalar (VSX) instruction set.  @code{__float128} supports
the 128-bit floating type.  On i386, x86_64, PowerPC, and IA-64
other than HP-UX, @code{__float128} is an alias for @code{_Float128}.
On hppa and IA-64 HP-UX, @code{__float128} is an alias for @code{long
double}.

@item @code{__float80} is available on the i386, x86_64, and IA-64
targets, and supports the 80-bit (@code{XFmode}) floating type.  It is
an alias for the type name @code{_Float64x} on these targets.

@item @code{__ibm128} is available on PowerPC targets, and provides
access to the IBM extended double format which is the current format
used for @code{long double}.  When @code{long double} transitions to
@code{__float128} on PowerPC in the future, @code{__ibm128} will remain
for use in conversions between the two types.
@end itemize

Support for these additional types includes the arithmetic operators:
add, subtract, multiply, divide; unary arithmetic operators;
relational operators; equality operators; and conversions to and from
integer and other floating types.  Use a suffix @samp{w} or @samp{W}
in a literal constant of type @code{__float80} or type
@code{__ibm128}.  Use a suffix @samp{q} or @samp{Q} for @code{_float128}.

In order to use @code{_Float128}, @code{__float128}, and @code{__ibm128}
on PowerPC Linux systems, you must use the @option{-mfloat128} option. It is
expected in future versions of GCC that @code{_Float128} and @code{__float128}
will be enabled automatically.

The @code{_Float128} type is supported on all systems where
@code{__float128} is supported or where @code{long double} has the
IEEE binary128 format.  The @code{_Float64x} type is supported on all
systems where @code{__float128} is supported.  The @code{_Float32}
type is supported on all systems supporting IEEE binary32; the
@code{_Float64} and @code{_Float32x} types are supported on all systems
supporting IEEE binary64.  The @code{_Float16} type is supported on AArch64
systems by default, and on ARM systems when the IEEE format for 16-bit
floating-point types is selected with @option{-mfp16-format=ieee}.
GCC does not currently support @code{_Float128x} on any systems.

On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex
types using the corresponding internal complex type, @code{XCmode} for
@code{__float80} type and @code{TCmode} for @code{__float128} type:

@smallexample
typedef _Complex float __attribute__((mode(TC))) _Complex128;
typedef _Complex float __attribute__((mode(XC))) _Complex80;
@end smallexample

On the PowerPC Linux VSX targets, you can declare complex types using
the corresponding internal complex type, @code{KCmode} for
@code{__float128} type and @code{ICmode} for @code{__ibm128} type:

@smallexample
typedef _Complex float __attribute__((mode(KC))) _Complex_float128;
typedef _Complex float __attribute__((mode(IC))) _Complex_ibm128;
@end smallexample

@node Half-Precision
@section Half-Precision Floating Point
@cindex half-precision floating point
@cindex @code{__fp16} data type

On ARM and AArch64 targets, GCC supports half-precision (16-bit) floating
point via the @code{__fp16} type defined in the ARM C Language Extensions.
On ARM systems, you must enable this type explicitly with the
@option{-mfp16-format} command-line option in order to use it.

ARM targets support two incompatible representations for half-precision
floating-point values.  You must choose one of the representations and
use it consistently in your program.

Specifying @option{-mfp16-format=ieee} selects the IEEE 754-2008 format.
This format can represent normalized values in the range of @math{2^{-14}} to 65504.
There are 11 bits of significand precision, approximately 3
decimal digits.

Specifying @option{-mfp16-format=alternative} selects the ARM
alternative format.  This representation is similar to the IEEE
format, but does not support infinities or NaNs.  Instead, the range
of exponents is extended, so that this format can represent normalized
values in the range of @math{2^{-14}} to 131008.

The GCC port for AArch64 only supports the IEEE 754-2008 format, and does
not require use of the @option{-mfp16-format} command-line option.

The @code{__fp16} type may only be used as an argument to intrinsics defined
in @code{<arm_fp16.h>}, or as a storage format.  For purposes of
arithmetic and other operations, @code{__fp16} values in C or C++
expressions are automatically promoted to @code{float}.

The ARM target provides hardware support for conversions between
@code{__fp16} and @code{float} values
as an extension to VFP and NEON (Advanced SIMD), and from ARMv8-A provides
hardware support for conversions between @code{__fp16} and @code{double}
values.  GCC generates code using these hardware instructions if you
compile with options to select an FPU that provides them;
for example, @option{-mfpu=neon-fp16 -mfloat-abi=softfp},
in addition to the @option{-mfp16-format} option to select
a half-precision format.

Language-level support for the @code{__fp16} data type is
independent of whether GCC generates code using hardware floating-point
instructions.  In cases where hardware support is not specified, GCC
implements conversions between @code{__fp16} and other types as library
calls.

It is recommended that portable code use the @code{_Float16} type defined
by ISO/IEC TS 18661-3:2015.  @xref{Floating Types}.

@node Decimal Float
@section Decimal Floating Types
@cindex decimal floating types
@cindex @code{_Decimal32} data type
@cindex @code{_Decimal64} data type
@cindex @code{_Decimal128} data type
@cindex @code{df} integer suffix
@cindex @code{dd} integer suffix
@cindex @code{dl} integer suffix
@cindex @code{DF} integer suffix
@cindex @code{DD} integer suffix
@cindex @code{DL} integer suffix

As an extension, GNU C supports decimal floating types as
defined in the N1312 draft of ISO/IEC WDTR24732.  Support for decimal
floating types in GCC will evolve as the draft technical report changes.
Calling conventions for any target might also change.  Not all targets
support decimal floating types.

The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and
@code{_Decimal128}.  They use a radix of ten, unlike the floating types
@code{float}, @code{double}, and @code{long double} whose radix is not
specified by the C standard but is usually two.

Support for decimal floating types includes the arithmetic operators
add, subtract, multiply, divide; unary arithmetic operators;
relational operators; equality operators; and conversions to and from
integer and other floating types.  Use a suffix @samp{df} or
@samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd}
or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for
@code{_Decimal128}.

GCC support of decimal float as specified by the draft technical report
is incomplete:

@itemize @bullet
@item
When the value of a decimal floating type cannot be represented in the
integer type to which it is being converted, the result is undefined
rather than the result value specified by the draft technical report.

@item
GCC does not provide the C library functionality associated with
@file{math.h}, @file{fenv.h}, @file{stdio.h}, @file{stdlib.h}, and
@file{wchar.h}, which must come from a separate C library implementation.
Because of this the GNU C compiler does not define macro
@code{__STDC_DEC_FP__} to indicate that the implementation conforms to
the technical report.
@end itemize

Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128}
are supported by the DWARF debug information format.

@node Hex Floats
@section Hex Floats
@cindex hex floats

ISO C99 and ISO C++17 support floating-point numbers written not only in
the usual decimal notation, such as @code{1.55e1}, but also numbers such as
@code{0x1.fp3} written in hexadecimal format.  As a GNU extension, GCC
supports this in C90 mode (except in some cases when strictly
conforming) and in C++98, C++11 and C++14 modes.  In that format the
@samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
mandatory.  The exponent is a decimal number that indicates the power of
2 by which the significant part is multiplied.  Thus @samp{0x1.f} is
@tex
$1 {15\over16}$,
@end tex
@ifnottex
1 15/16,
@end ifnottex
@samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
is the same as @code{1.55e1}.

Unlike for floating-point numbers in the decimal notation the exponent
is always required in the hexadecimal notation.  Otherwise the compiler
would not be able to resolve the ambiguity of, e.g., @code{0x1.f}.  This
could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
extension for floating-point constants of type @code{float}.

@node Fixed-Point
@section Fixed-Point Types
@cindex fixed-point types
@cindex @code{_Fract} data type
@cindex @code{_Accum} data type
@cindex @code{_Sat} data type
@cindex @code{hr} fixed-suffix
@cindex @code{r} fixed-suffix
@cindex @code{lr} fixed-suffix
@cindex @code{llr} fixed-suffix
@cindex @code{uhr} fixed-suffix
@cindex @code{ur} fixed-suffix
@cindex @code{ulr} fixed-suffix
@cindex @code{ullr} fixed-suffix
@cindex @code{hk} fixed-suffix
@cindex @code{k} fixed-suffix
@cindex @code{lk} fixed-suffix
@cindex @code{llk} fixed-suffix
@cindex @code{uhk} fixed-suffix
@cindex @code{uk} fixed-suffix
@cindex @code{ulk} fixed-suffix
@cindex @code{ullk} fixed-suffix
@cindex @code{HR} fixed-suffix
@cindex @code{R} fixed-suffix
@cindex @code{LR} fixed-suffix
@cindex @code{LLR} fixed-suffix
@cindex @code{UHR} fixed-suffix
@cindex @code{UR} fixed-suffix
@cindex @code{ULR} fixed-suffix
@cindex @code{ULLR} fixed-suffix
@cindex @code{HK} fixed-suffix
@cindex @code{K} fixed-suffix
@cindex @code{LK} fixed-suffix
@cindex @code{LLK} fixed-suffix
@cindex @code{UHK} fixed-suffix
@cindex @code{UK} fixed-suffix
@cindex @code{ULK} fixed-suffix
@cindex @code{ULLK} fixed-suffix

As an extension, GNU C supports fixed-point types as
defined in the N1169 draft of ISO/IEC DTR 18037.  Support for fixed-point
types in GCC will evolve as the draft technical report changes.
Calling conventions for any target might also change.  Not all targets
support fixed-point types.

The fixed-point types are
@code{short _Fract},
@code{_Fract},
@code{long _Fract},
@code{long long _Fract},
@code{unsigned short _Fract},
@code{unsigned _Fract},
@code{unsigned long _Fract},
@code{unsigned long long _Fract},
@code{_Sat short _Fract},
@code{_Sat _Fract},
@code{_Sat long _Fract},
@code{_Sat long long _Fract},
@code{_Sat unsigned short _Fract},
@code{_Sat unsigned _Fract},
@code{_Sat unsigned long _Fract},
@code{_Sat unsigned long long _Fract},
@code{short _Accum},
@code{_Accum},
@code{long _Accum},
@code{long long _Accum},
@code{unsigned short _Accum},
@code{unsigned _Accum},
@code{unsigned long _Accum},
@code{unsigned long long _Accum},
@code{_Sat short _Accum},
@code{_Sat _Accum},
@code{_Sat long _Accum},
@code{_Sat long long _Accum},
@code{_Sat unsigned short _Accum},
@code{_Sat unsigned _Accum},
@code{_Sat unsigned long _Accum},
@code{_Sat unsigned long long _Accum}.

Fixed-point data values contain fractional and optional integral parts.
The format of fixed-point data varies and depends on the target machine.

Support for fixed-point types includes:
@itemize @bullet
@item
prefix and postfix increment and decrement operators (@code{++}, @code{--})
@item
unary arithmetic operators (@code{+}, @code{-}, @code{!})
@item
binary arithmetic operators (@code{+}, @code{-}, @code{*}, @code{/})
@item
binary shift operators (@code{<<}, @code{>>})
@item
relational operators (@code{<}, @code{<=}, @code{>=}, @code{>})
@item
equality operators (@code{==}, @code{!=})
@item
assignment operators (@code{+=}, @code{-=}, @code{*=}, @code{/=},
@code{<<=}, @code{>>=})
@item
conversions to and from integer, floating-point, or fixed-point types
@end itemize

Use a suffix in a fixed-point literal constant:
@itemize
@item @samp{hr} or @samp{HR} for @code{short _Fract} and
@code{_Sat short _Fract}
@item @samp{r} or @samp{R} for @code{_Fract} and @code{_Sat _Fract}
@item @samp{lr} or @samp{LR} for @code{long _Fract} and
@code{_Sat long _Fract}
@item @samp{llr} or @samp{LLR} for @code{long long _Fract} and
@code{_Sat long long _Fract}
@item @samp{uhr} or @samp{UHR} for @code{unsigned short _Fract} and
@code{_Sat unsigned short _Fract}
@item @samp{ur} or @samp{UR} for @code{unsigned _Fract} and
@code{_Sat unsigned _Fract}
@item @samp{ulr} or @samp{ULR} for @code{unsigned long _Fract} and
@code{_Sat unsigned long _Fract}
@item @samp{ullr} or @samp{ULLR} for @code{unsigned long long _Fract}
and @code{_Sat unsigned long long _Fract}
@item @samp{hk} or @samp{HK} for @code{short _Accum} and
@code{_Sat short _Accum}
@item @samp{k} or @samp{K} for @code{_Accum} and @code{_Sat _Accum}
@item @samp{lk} or @samp{LK} for @code{long _Accum} and
@code{_Sat long _Accum}
@item @samp{llk} or @samp{LLK} for @code{long long _Accum} and
@code{_Sat long long _Accum}
@item @samp{uhk} or @samp{UHK} for @code{unsigned short _Accum} and
@code{_Sat unsigned short _Accum}
@item @samp{uk} or @samp{UK} for @code{unsigned _Accum} and
@code{_Sat unsigned _Accum}
@item @samp{ulk} or @samp{ULK} for @code{unsigned long _Accum} and
@code{_Sat unsigned long _Accum}
@item @samp{ullk} or @samp{ULLK} for @code{unsigned long long _Accum}
and @code{_Sat unsigned long long _Accum}
@end itemize

GCC support of fixed-point types as specified by the draft technical report
is incomplete:

@itemize @bullet
@item
Pragmas to control overflow and rounding behaviors are not implemented.
@end itemize

Fixed-point types are supported by the DWARF debug information format.

@node Named Address Spaces
@section Named Address Spaces
@cindex Named Address Spaces

As an extension, GNU C supports named address spaces as
defined in the N1275 draft of ISO/IEC DTR 18037.  Support for named
address spaces in GCC will evolve as the draft technical report
changes.  Calling conventions for any target might also change.  At
present, only the AVR, M32C, RL78, and x86 targets support
address spaces other than the generic address space.

Address space identifiers may be used exactly like any other C type
qualifier (e.g., @code{const} or @code{volatile}).  See the N1275
document for more details.

@anchor{AVR Named Address Spaces}
@subsection AVR Named Address Spaces

On the AVR target, there are several address spaces that can be used
in order to put read-only data into the flash memory and access that
data by means of the special instructions @code{LPM} or @code{ELPM}
needed to read from flash.

Devices belonging to @code{avrtiny} and @code{avrxmega3} can access
flash memory by means of @code{LD*} instructions because the flash
memory is mapped into the RAM address space.  There is @emph{no need}
for language extensions like @code{__flash} or attribute
@ref{AVR Variable Attributes,,@code{progmem}}.
The default linker description files for these devices cater for that
feature and @code{.rodata} stays in flash: The compiler just generates
@code{LD*} instructions, and the linker script adds core specific
offsets to all @code{.rodata} symbols: @code{0x4000} in the case of
@code{avrtiny} and @code{0x8000} in the case of @code{avrxmega3}.
See @ref{AVR Options} for a list of respective devices.

For devices not in @code{avrtiny} or @code{avrxmega3},
any data including read-only data is located in RAM (the generic
address space) because flash memory is not visible in the RAM address
space.  In order to locate read-only data in flash memory @emph{and}
to generate the right instructions to access this data without
using (inline) assembler code, special address spaces are needed.

@table @code
@item __flash
@cindex @code{__flash} AVR Named Address Spaces
The @code{__flash} qualifier locates data in the
@code{.progmem.data} section. Data is read using the @code{LPM}
instruction. Pointers to this address space are 16 bits wide.

@item __flash1
@itemx __flash2
@itemx __flash3
@itemx __flash4
@itemx __flash5
@cindex @code{__flash1} AVR Named Address Spaces
@cindex @code{__flash2} AVR Named Address Spaces
@cindex @code{__flash3} AVR Named Address Spaces
@cindex @code{__flash4} AVR Named Address Spaces
@cindex @code{__flash5} AVR Named Address Spaces
These are 16-bit address spaces locating data in section
@code{.progmem@var{N}.data} where @var{N} refers to
address space @code{__flash@var{N}}.
The compiler sets the @code{RAMPZ} segment register appropriately 
before reading data by means of the @code{ELPM} instruction.

@item __memx
@cindex @code{__memx} AVR Named Address Spaces
This is a 24-bit address space that linearizes flash and RAM:
If the high bit of the address is set, data is read from
RAM using the lower two bytes as RAM address.
If the high bit of the address is clear, data is read from flash
with @code{RAMPZ} set according to the high byte of the address.
@xref{AVR Built-in Functions,,@code{__builtin_avr_flash_segment}}.

Objects in this address space are located in @code{.progmemx.data}.
@end table

@b{Example}

@smallexample
char my_read (const __flash char ** p)
@{
    /* p is a pointer to RAM that points to a pointer to flash.
       The first indirection of p reads that flash pointer
       from RAM and the second indirection reads a char from this
       flash address.  */

    return **p;
@}

/* Locate array[] in flash memory */
const __flash int array[] = @{ 3, 5, 7, 11, 13, 17, 19 @};

int i = 1;

int main (void)
@{
   /* Return 17 by reading from flash memory */
   return array[array[i]];
@}
@end smallexample

@noindent
For each named address space supported by avr-gcc there is an equally
named but uppercase built-in macro defined. 
The purpose is to facilitate testing if respective address space
support is available or not:

@smallexample
#ifdef __FLASH
const __flash int var = 1;

int read_var (void)
@{
    return var;
@}
#else
#include <avr/pgmspace.h> /* From AVR-LibC */

const int var PROGMEM = 1;

int read_var (void)
@{
    return (int) pgm_read_word (&var);
@}
#endif /* __FLASH */
@end smallexample

@noindent
Notice that attribute @ref{AVR Variable Attributes,,@code{progmem}}
locates data in flash but
accesses to these data read from generic address space, i.e.@:
from RAM,
so that you need special accessors like @code{pgm_read_byte}
from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}}
together with attribute @code{progmem}.

@noindent
@b{Limitations and caveats}

@itemize
@item
Reading across the 64@tie{}KiB section boundary of
the @code{__flash} or @code{__flash@var{N}} address spaces
shows undefined behavior. The only address space that
supports reading across the 64@tie{}KiB flash segment boundaries is
@code{__memx}.

@item
If you use one of the @code{__flash@var{N}} address spaces
you must arrange your linker script to locate the
@code{.progmem@var{N}.data} sections according to your needs.

@item
Any data or pointers to the non-generic address spaces must
be qualified as @code{const}, i.e.@: as read-only data.
This still applies if the data in one of these address
spaces like software version number or calibration lookup table are intended to
be changed after load time by, say, a boot loader. In this case
the right qualification is @code{const} @code{volatile} so that the compiler
must not optimize away known values or insert them
as immediates into operands of instructions.

@item
The following code initializes a variable @code{pfoo}
located in static storage with a 24-bit address:
@smallexample
extern const __memx char foo;
const __memx void *pfoo = &foo;
@end smallexample

@item
On the reduced Tiny devices like ATtiny40, no address spaces are supported.
Just use vanilla C / C++ code without overhead as outlined above.
Attribute @code{progmem} is supported but works differently,
see @ref{AVR Variable Attributes}.

@end itemize

@subsection M32C Named Address Spaces
@cindex @code{__far} M32C Named Address Spaces

On the M32C target, with the R8C and M16C CPU variants, variables
qualified with @code{__far} are accessed using 32-bit addresses in
order to access memory beyond the first 64@tie{}Ki bytes.  If
@code{__far} is used with the M32CM or M32C CPU variants, it has no
effect.

@subsection RL78 Named Address Spaces
@cindex @code{__far} RL78 Named Address Spaces

On the RL78 target, variables qualified with @code{__far} are accessed
with 32-bit pointers (20-bit addresses) rather than the default 16-bit
addresses.  Non-far variables are assumed to appear in the topmost
64@tie{}KiB of the address space.

@subsection x86 Named Address Spaces
@cindex x86 named address spaces

On the x86 target, variables may be declared as being relative
to the @code{%fs} or @code{%gs} segments.

@table @code
@item __seg_fs
@itemx __seg_gs
@cindex @code{__seg_fs} x86 named address space
@cindex @code{__seg_gs} x86 named address space
The object is accessed with the respective segment override prefix.

The respective segment base must be set via some method specific to
the operating system.  Rather than require an expensive system call
to retrieve the segment base, these address spaces are not considered
to be subspaces of the generic (flat) address space.  This means that
explicit casts are required to convert pointers between these address
spaces and the generic address space.  In practice the application
should cast to @code{uintptr_t} and apply the segment base offset
that it installed previously.

The preprocessor symbols @code{__SEG_FS} and @code{__SEG_GS} are
defined when these address spaces are supported.
@end table

@node Zero Length
@section Arrays of Length Zero
@cindex arrays of length zero
@cindex zero-length arrays
@cindex length-zero arrays
@cindex flexible array members

Declaring zero-length arrays is allowed in GNU C as an extension.
A zero-length array can be useful as the last element of a structure
that is really a header for a variable-length object:

@smallexample
struct line @{
  int length;
  char contents[0];
@};

struct line *thisline = (struct line *)
  malloc (sizeof (struct line) + this_length);
thisline->length = this_length;
@end smallexample

Although the size of a zero-length array is zero, an array member of
this kind may increase the size of the enclosing type as a result of tail
padding.  The offset of a zero-length array member from the beginning
of the enclosing structure is the same as the offset of an array with
one or more elements of the same type.  The alignment of a zero-length
array is the same as the alignment of its elements.

Declaring zero-length arrays in other contexts, including as interior
members of structure objects or as non-member objects, is discouraged.
Accessing elements of zero-length arrays declared in such contexts is
undefined and may be diagnosed.

In the absence of the zero-length array extension, in ISO C90
the @code{contents} array in the example above would typically be declared
to have a single element.  Unlike a zero-length array which only contributes
to the size of the enclosing structure for the purposes of alignment,
a one-element array always occupies at least as much space as a single
object of the type.  Although using one-element arrays this way is
discouraged, GCC handles accesses to trailing one-element array members
analogously to zero-length arrays.

The preferred mechanism to declare variable-length types like
@code{struct line} above is the ISO C99 @dfn{flexible array member},
with slightly different syntax and semantics:

@itemize @bullet
@item
Flexible array members are written as @code{contents[]} without
the @code{0}.

@item
Flexible array members have incomplete type, and so the @code{sizeof}
operator may not be applied.  As a quirk of the original implementation
of zero-length arrays, @code{sizeof} evaluates to zero.

@item
Flexible array members may only appear as the last member of a
@code{struct} that is otherwise non-empty.

@item
A structure containing a flexible array member, or a union containing
such a structure (possibly recursively), may not be a member of a
structure or an element of an array.  (However, these uses are
permitted by GCC as extensions.)
@end itemize

Non-empty initialization of zero-length
arrays is treated like any case where there are more initializer
elements than the array holds, in that a suitable warning about ``excess
elements in array'' is given, and the excess elements (all of them, in
this case) are ignored.

GCC allows static initialization of flexible array members.
This is equivalent to defining a new structure containing the original
structure followed by an array of sufficient size to contain the data.
E.g.@: in the following, @code{f1} is constructed as if it were declared
like @code{f2}.

@smallexample
struct f1 @{
  int x; int y[];
@} f1 = @{ 1, @{ 2, 3, 4 @} @};

struct f2 @{
  struct f1 f1; int data[3];
@} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
@end smallexample

@noindent
The convenience of this extension is that @code{f1} has the desired
type, eliminating the need to consistently refer to @code{f2.f1}.

This has symmetry with normal static arrays, in that an array of
unknown size is also written with @code{[]}.

Of course, this extension only makes sense if the extra data comes at
the end of a top-level object, as otherwise we would be overwriting
data at subsequent offsets.  To avoid undue complication and confusion
with initialization of deeply nested arrays, we simply disallow any
non-empty initialization except when the structure is the top-level
object.  For example:

@smallexample
struct foo @{ int x; int y[]; @};
struct bar @{ struct foo z; @};

struct foo a = @{ 1, @{ 2, 3, 4 @} @};        // @r{Valid.}
struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @};    // @r{Invalid.}
struct bar c = @{ @{ 1, @{ @} @} @};            // @r{Valid.}
struct foo d[1] = @{ @{ 1, @{ 2, 3, 4 @} @} @};  // @r{Invalid.}
@end smallexample

@node Empty Structures
@section Structures with No Members
@cindex empty structures
@cindex zero-size structures

GCC permits a C structure to have no members:

@smallexample
struct empty @{
@};
@end smallexample

The structure has size zero.  In C++, empty structures are part
of the language.  G++ treats empty structures as if they had a single
member of type @code{char}.

@node Variable Length
@section Arrays of Variable Length
@cindex variable-length arrays
@cindex arrays of variable length
@cindex VLAs

Variable-length automatic arrays are allowed in ISO C99, and as an
extension GCC accepts them in C90 mode and in C++.  These arrays are
declared like any other automatic arrays, but with a length that is not
a constant expression.  The storage is allocated at the point of
declaration and deallocated when the block scope containing the declaration
exits.  For
example:

@smallexample
FILE *
concat_fopen (char *s1, char *s2, char *mode)
@{
  char str[strlen (s1) + strlen (s2) + 1];
  strcpy (str, s1);
  strcat (str, s2);
  return fopen (str, mode);
@}
@end smallexample

@cindex scope of a variable length array
@cindex variable-length array scope
@cindex deallocating variable length arrays
Jumping or breaking out of the scope of the array name deallocates the
storage.  Jumping into the scope is not allowed; you get an error
message for it.

@cindex variable-length array in a structure
As an extension, GCC accepts variable-length arrays as a member of
a structure or a union.  For example:

@smallexample
void
foo (int n)
@{
  struct S @{ int x[n]; @};
@}
@end smallexample

@cindex @code{alloca} vs variable-length arrays
You can use the function @code{alloca} to get an effect much like
variable-length arrays.  The function @code{alloca} is available in
many other C implementations (but not in all).  On the other hand,
variable-length arrays are more elegant.

There are other differences between these two methods.  Space allocated
with @code{alloca} exists until the containing @emph{function} returns.
The space for a variable-length array is deallocated as soon as the array
name's scope ends, unless you also use @code{alloca} in this scope.

You can also use variable-length arrays as arguments to functions:

@smallexample
struct entry
tester (int len, char data[len][len])
@{
  /* @r{@dots{}} */
@}
@end smallexample

The length of an array is computed once when the storage is allocated
and is remembered for the scope of the array in case you access it with
@code{sizeof}.

If you want to pass the array first and the length afterward, you can
use a forward declaration in the parameter list---another GNU extension.

@smallexample
struct entry
tester (int len; char data[len][len], int len)
@{
  /* @r{@dots{}} */
@}
@end smallexample

@cindex parameter forward declaration
The @samp{int len} before the semicolon is a @dfn{parameter forward
declaration}, and it serves the purpose of making the name @code{len}
known when the declaration of @code{data} is parsed.

You can write any number of such parameter forward declarations in the
parameter list.  They can be separated by commas or semicolons, but the
last one must end with a semicolon, which is followed by the ``real''
parameter declarations.  Each forward declaration must match a ``real''
declaration in parameter name and data type.  ISO C99 does not support
parameter forward declarations.

@node Variadic Macros
@section Macros with a Variable Number of Arguments.
@cindex variable number of arguments
@cindex macro with variable arguments
@cindex rest argument (in macro)
@cindex variadic macros

In the ISO C standard of 1999, a macro can be declared to accept a
variable number of arguments much as a function can.  The syntax for
defining the macro is similar to that of a function.  Here is an
example:

@smallexample
#define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
@end smallexample

@noindent
Here @samp{@dots{}} is a @dfn{variable argument}.  In the invocation of
such a macro, it represents the zero or more tokens until the closing
parenthesis that ends the invocation, including any commas.  This set of
tokens replaces the identifier @code{__VA_ARGS__} in the macro body
wherever it appears.  See the CPP manual for more information.

GCC has long supported variadic macros, and used a different syntax that
allowed you to give a name to the variable arguments just like any other
argument.  Here is an example:

@smallexample
#define debug(format, args...) fprintf (stderr, format, args)
@end smallexample

@noindent
This is in all ways equivalent to the ISO C example above, but arguably
more readable and descriptive.

GNU CPP has two further variadic macro extensions, and permits them to
be used with either of the above forms of macro definition.

In standard C, you are not allowed to leave the variable argument out
entirely; but you are allowed to pass an empty argument.  For example,
this invocation is invalid in ISO C, because there is no comma after
the string:

@smallexample
debug ("A message")
@end smallexample

GNU CPP permits you to completely omit the variable arguments in this
way.  In the above examples, the compiler would complain, though since
the expansion of the macro still has the extra comma after the format
string.

To help solve this problem, CPP behaves specially for variable arguments
used with the token paste operator, @samp{##}.  If instead you write

@smallexample
#define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
@end smallexample

@noindent
and if the variable arguments are omitted or empty, the @samp{##}
operator causes the preprocessor to remove the comma before it.  If you
do provide some variable arguments in your macro invocation, GNU CPP
does not complain about the paste operation and instead places the
variable arguments after the comma.  Just like any other pasted macro
argument, these arguments are not macro expanded.

@node Escaped Newlines
@section Slightly Looser Rules for Escaped Newlines
@cindex escaped newlines
@cindex newlines (escaped)

The preprocessor treatment of escaped newlines is more relaxed 
than that specified by the C90 standard, which requires the newline
to immediately follow a backslash.  
GCC's implementation allows whitespace in the form
of spaces, horizontal and vertical tabs, and form feeds between the
backslash and the subsequent newline.  The preprocessor issues a
warning, but treats it as a valid escaped newline and combines the two
lines to form a single logical line.  This works within comments and
tokens, as well as between tokens.  Comments are @emph{not} treated as
whitespace for the purposes of this relaxation, since they have not
yet been replaced with spaces.

@node Subscripting
@section Non-Lvalue Arrays May Have Subscripts
@cindex subscripting
@cindex arrays, non-lvalue

@cindex subscripting and function values
In ISO C99, arrays that are not lvalues still decay to pointers, and
may be subscripted, although they may not be modified or used after
the next sequence point and the unary @samp{&} operator may not be
applied to them.  As an extension, GNU C allows such arrays to be
subscripted in C90 mode, though otherwise they do not decay to
pointers outside C99 mode.  For example,
this is valid in GNU C though not valid in C90:

@smallexample
@group
struct foo @{int a[4];@};

struct foo f();

bar (int index)
@{
  return f().a[index];
@}
@end group
@end smallexample

@node Pointer Arith
@section Arithmetic on @code{void}- and Function-Pointers
@cindex void pointers, arithmetic
@cindex void, size of pointer to
@cindex function pointers, arithmetic
@cindex function, size of pointer to

In GNU C, addition and subtraction operations are supported on pointers to
@code{void} and on pointers to functions.  This is done by treating the
size of a @code{void} or of a function as 1.

A consequence of this is that @code{sizeof} is also allowed on @code{void}
and on function types, and returns 1.

@opindex Wpointer-arith
The option @option{-Wpointer-arith} requests a warning if these extensions
are used.

@node Variadic Pointer Args
@section Pointer Arguments in Variadic Functions
@cindex pointer arguments in variadic functions
@cindex variadic functions, pointer arguments

Standard C requires that pointer types used with @code{va_arg} in
functions with variable argument lists either must be compatible with
that of the actual argument, or that one type must be a pointer to
@code{void} and the other a pointer to a character type.  GNU C
implements the POSIX XSI extension that additionally permits the use
of @code{va_arg} with a pointer type to receive arguments of any other
pointer type.

In particular, in GNU C @samp{va_arg (ap, void *)} can safely be used
to consume an argument of any pointer type.

@node Pointers to Arrays
@section Pointers to Arrays with Qualifiers Work as Expected
@cindex pointers to arrays
@cindex const qualifier

In GNU C, pointers to arrays with qualifiers work similar to pointers
to other qualified types. For example, a value of type @code{int (*)[5]}
can be used to initialize a variable of type @code{const int (*)[5]}.
These types are incompatible in ISO C because the @code{const} qualifier
is formally attached to the element type of the array and not the
array itself.

@smallexample
extern void
transpose (int N, int M, double out[M][N], const double in[N][M]);
double x[3][2];
double y[2][3];
@r{@dots{}}
transpose(3, 2, y, x);
@end smallexample

@node Initializers
@section Non-Constant Initializers
@cindex initializers, non-constant
@cindex non-constant initializers

As in standard C++ and ISO C99, the elements of an aggregate initializer for an
automatic variable are not required to be constant expressions in GNU C@.
Here is an example of an initializer with run-time varying elements:

@smallexample
foo (float f, float g)
@{
  float beat_freqs[2] = @{ f-g, f+g @};
  /* @r{@dots{}} */
@}
@end smallexample

@node Compound Literals
@section Compound Literals
@cindex constructor expressions
@cindex initializations in expressions
@cindex structures, constructor expression
@cindex expressions, constructor
@cindex compound literals
@c The GNU C name for what C99 calls compound literals was "constructor expressions".

A compound literal looks like a cast of a brace-enclosed aggregate
initializer list.  Its value is an object of the type specified in
the cast, containing the elements specified in the initializer.
Unlike the result of a cast, a compound literal is an lvalue.  ISO
C99 and later support compound literals.  As an extension, GCC
supports compound literals also in C90 mode and in C++, although
as explained below, the C++ semantics are somewhat different.

Usually, the specified type of a compound literal is a structure.  Assume
that @code{struct foo} and @code{structure} are declared as shown:

@smallexample
struct foo @{int a; char b[2];@} structure;
@end smallexample

@noindent
Here is an example of constructing a @code{struct foo} with a compound literal:

@smallexample
structure = ((struct foo) @{x + y, 'a', 0@});
@end smallexample

@noindent
This is equivalent to writing the following:

@smallexample
@{
  struct foo temp = @{x + y, 'a', 0@};
  structure = temp;
@}
@end smallexample

You can also construct an array, though this is dangerous in C++, as
explained below.  If all the elements of the compound literal are
(made up of) simple constant expressions suitable for use in
initializers of objects of static storage duration, then the compound
literal can be coerced to a pointer to its first element and used in
such an initializer, as shown here:

@smallexample
char **foo = (char *[]) @{ "x", "y", "z" @};
@end smallexample

Compound literals for scalar types and union types are also allowed.  In
the following example the variable @code{i} is initialized to the value
@code{2}, the result of incrementing the unnamed object created by
the compound literal.

@smallexample
int i = ++(int) @{ 1 @};
@end smallexample

As a GNU extension, GCC allows initialization of objects with static storage
duration by compound literals (which is not possible in ISO C99 because
the initializer is not a constant).
It is handled as if the object were initialized only with the brace-enclosed
list if the types of the compound literal and the object match.
The elements of the compound literal must be constant.
If the object being initialized has array type of unknown size, the size is
determined by the size of the compound literal.

@smallexample
static struct foo x = (struct foo) @{1, 'a', 'b'@};
static int y[] = (int []) @{1, 2, 3@};
static int z[] = (int [3]) @{1@};
@end smallexample

@noindent
The above lines are equivalent to the following:
@smallexample
static struct foo x = @{1, 'a', 'b'@};
static int y[] = @{1, 2, 3@};
static int z[] = @{1, 0, 0@};
@end smallexample

In C, a compound literal designates an unnamed object with static or
automatic storage duration.  In C++, a compound literal designates a
temporary object that only lives until the end of its full-expression.
As a result, well-defined C code that takes the address of a subobject
of a compound literal can be undefined in C++, so G++ rejects
the conversion of a temporary array to a pointer.  For instance, if
the array compound literal example above appeared inside a function,
any subsequent use of @code{foo} in C++ would have undefined behavior
because the lifetime of the array ends after the declaration of @code{foo}.

As an optimization, G++ sometimes gives array compound literals longer
lifetimes: when the array either appears outside a function or has
a @code{const}-qualified type.  If @code{foo} and its initializer had
elements of type @code{char *const} rather than @code{char *}, or if
@code{foo} were a global variable, the array would have static storage
duration.  But it is probably safest just to avoid the use of array
compound literals in C++ code.

@node Designated Inits
@section Designated Initializers
@cindex initializers with labeled elements
@cindex labeled elements in initializers
@cindex case labels in initializers
@cindex designated initializers

Standard C90 requires the elements of an initializer to appear in a fixed
order, the same as the order of the elements in the array or structure
being initialized.

In ISO C99 you can give the elements in any order, specifying the array
indices or structure field names they apply to, and GNU C allows this as
an extension in C90 mode as well.  This extension is not
implemented in GNU C++.

To specify an array index, write
@samp{[@var{index}] =} before the element value.  For example,

@smallexample
int a[6] = @{ [4] = 29, [2] = 15 @};
@end smallexample

@noindent
is equivalent to

@smallexample
int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
@end smallexample

@noindent
The index values must be constant expressions, even if the array being
initialized is automatic.

An alternative syntax for this that has been obsolete since GCC 2.5 but
GCC still accepts is to write @samp{[@var{index}]} before the element
value, with no @samp{=}.

To initialize a range of elements to the same value, write
@samp{[@var{first} ... @var{last}] = @var{value}}.  This is a GNU
extension.  For example,

@smallexample
int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
@end smallexample

@noindent
If the value in it has side effects, the side effects happen only once,
not for each initialized field by the range initializer.

@noindent
Note that the length of the array is the highest value specified
plus one.

In a structure initializer, specify the name of a field to initialize
with @samp{.@var{fieldname} =} before the element value.  For example,
given the following structure,

@smallexample
struct point @{ int x, y; @};
@end smallexample

@noindent
the following initialization

@smallexample
struct point p = @{ .y = yvalue, .x = xvalue @};
@end smallexample

@noindent
is equivalent to

@smallexample
struct point p = @{ xvalue, yvalue @};
@end smallexample

Another syntax that has the same meaning, obsolete since GCC 2.5, is
@samp{@var{fieldname}:}, as shown here:

@smallexample
struct point p = @{ y: yvalue, x: xvalue @};
@end smallexample

Omitted fields are implicitly initialized the same as for objects
that have static storage duration.

@cindex designators
The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
@dfn{designator}.  You can also use a designator (or the obsolete colon
syntax) when initializing a union, to specify which element of the union
should be used.  For example,

@smallexample
union foo @{ int i; double d; @};

union foo f = @{ .d = 4 @};
@end smallexample

@noindent
converts 4 to a @code{double} to store it in the union using
the second element.  By contrast, casting 4 to type @code{union foo}
stores it into the union as the integer @code{i}, since it is
an integer.  @xref{Cast to Union}.

You can combine this technique of naming elements with ordinary C
initialization of successive elements.  Each initializer element that
does not have a designator applies to the next consecutive element of the
array or structure.  For example,

@smallexample
int a[6] = @{ [1] = v1, v2, [4] = v4 @};
@end smallexample

@noindent
is equivalent to

@smallexample
int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
@end smallexample

Labeling the elements of an array initializer is especially useful
when the indices are characters or belong to an @code{enum} type.
For example:

@smallexample
int whitespace[256]
  = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
      ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
@end smallexample

@cindex designator lists
You can also write a series of @samp{.@var{fieldname}} and
@samp{[@var{index}]} designators before an @samp{=} to specify a
nested subobject to initialize; the list is taken relative to the
subobject corresponding to the closest surrounding brace pair.  For
example, with the @samp{struct point} declaration above:

@smallexample
struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
@end smallexample

If the same field is initialized multiple times, or overlapping
fields of a union are initialized, the value from the last
initialization is used.  When a field of a union is itself a structure, 
the entire structure from the last field initialized is used.  If any previous
initializer has side effect, it is unspecified whether the side effect
happens or not.  Currently, GCC discards the side-effecting
initializer expressions and issues a warning.

@node Case Ranges
@section Case Ranges
@cindex case ranges
@cindex ranges in case statements

You can specify a range of consecutive values in a single @code{case} label,
like this:

@smallexample
case @var{low} ... @var{high}:
@end smallexample

@noindent
This has the same effect as the proper number of individual @code{case}
labels, one for each integer value from @var{low} to @var{high}, inclusive.

This feature is especially useful for ranges of ASCII character codes:

@smallexample
case 'A' ... 'Z':
@end smallexample

@strong{Be careful:} Write spaces around the @code{...}, for otherwise
it may be parsed wrong when you use it with integer values.  For example,
write this:

@smallexample
case 1 ... 5:
@end smallexample

@noindent
rather than this:

@smallexample
case 1...5:
@end smallexample

@node Cast to Union
@section Cast to a Union Type
@cindex cast to a union
@cindex union, casting to a

A cast to a union type is a C extension not available in C++.  It looks
just like ordinary casts with the constraint that the type specified is
a union type.  You can specify the type either with the @code{union}
keyword or with a @code{typedef} name that refers to a union.  The result
of a cast to a union is a temporary rvalue of the union type with a member
whose type matches that of the operand initialized to the value of
the operand.  The effect of a cast to a union is similar to a compound
literal except that it yields an rvalue like standard casts do.
@xref{Compound Literals}.

Expressions that may be cast to the union type are those whose type matches
at least one of the members of the union.  Thus, given the following union
and variables:

@smallexample
union foo @{ int i; double d; @};
int x;
double y;
union foo z;
@end smallexample

@noindent
both @code{x} and @code{y} can be cast to type @code{union foo} and
the following assignments
@smallexample
  z = (union foo) x;
  z = (union foo) y;
@end smallexample
are shorthand equivalents of these
@smallexample
  z = (union foo) @{ .i = x @};
  z = (union foo) @{ .d = y @};
@end smallexample

However, @code{(union foo) FLT_MAX;} is not a valid cast because the union
has no member of type @code{float}.

Using the cast as the right-hand side of an assignment to a variable of
union type is equivalent to storing in a member of the union with
the same type

@smallexample
union foo u;
/* @r{@dots{}} */
u = (union foo) x  @equiv{}  u.i = x
u = (union foo) y  @equiv{}  u.d = y
@end smallexample

You can also use the union cast as a function argument:

@smallexample
void hack (union foo);
/* @r{@dots{}} */
hack ((union foo) x);
@end smallexample

@node Mixed Declarations
@section Mixed Declarations and Code
@cindex mixed declarations and code
@cindex declarations, mixed with code
@cindex code, mixed with declarations

ISO C99 and ISO C++ allow declarations and code to be freely mixed
within compound statements.  As an extension, GNU C also allows this in
C90 mode.  For example, you could do:

@smallexample
int i;
/* @r{@dots{}} */
i++;
int j = i + 2;
@end smallexample

Each identifier is visible from where it is declared until the end of
the enclosing block.

@node Function Attributes
@section Declaring Attributes of Functions
@cindex function attributes
@cindex declaring attributes of functions
@cindex @code{volatile} applied to function
@cindex @code{const} applied to function

In GNU C and C++, you can use function attributes to specify certain
function properties that may help the compiler optimize calls or
check code more carefully for correctness.  For example, you
can use attributes to specify that a function never returns
(@code{noreturn}), returns a value depending only on the values of
its arguments (@code{const}), or has @code{printf}-style arguments
(@code{format}).

You can also use attributes to control memory placement, code
generation options or call/return conventions within the function
being annotated.  Many of these attributes are target-specific.  For
example, many targets support attributes for defining interrupt
handler functions, which typically must follow special register usage
and return conventions.  Such attributes are described in the subsection
for each target.  However, a considerable number of attributes are
supported by most, if not all targets.  Those are described in
the @ref{Common Function Attributes} section.

Function attributes are introduced by the @code{__attribute__} keyword
in the declaration of a function, followed by an attribute specification
enclosed in double parentheses.  You can specify multiple attributes in
a declaration by separating them by commas within the double parentheses
or by immediately following one attribute specification with another.
@xref{Attribute Syntax}, for the exact rules on attribute syntax and
placement.  Compatible attribute specifications on distinct declarations
of the same function are merged.  An attribute specification that is not
compatible with attributes already applied to a declaration of the same
function is ignored with a warning.

Some function attributes take one or more arguments that refer to
the function's parameters by their positions within the function parameter
list.  Such attribute arguments are referred to as @dfn{positional arguments}.
Unless specified otherwise, positional arguments that specify properties
of parameters with pointer types can also specify the same properties of
the implicit C++ @code{this} argument in non-static member functions, and
of parameters of reference to a pointer type.  For ordinary functions,
position one refers to the first parameter on the list.  In C++ non-static
member functions, position one refers to the implicit @code{this} pointer.
The same restrictions and effects apply to function attributes used with
ordinary functions or C++ member functions.

GCC also supports attributes on
variable declarations (@pxref{Variable Attributes}),
labels (@pxref{Label Attributes}),
enumerators (@pxref{Enumerator Attributes}),
statements (@pxref{Statement Attributes}),
and types (@pxref{Type Attributes}).

There is some overlap between the purposes of attributes and pragmas
(@pxref{Pragmas,,Pragmas Accepted by GCC}).  It has been
found convenient to use @code{__attribute__} to achieve a natural
attachment of attributes to their corresponding declarations, whereas
@code{#pragma} is of use for compatibility with other compilers
or constructs that do not naturally form part of the grammar.

In addition to the attributes documented here,
GCC plugins may provide their own attributes.

@menu
* Common Function Attributes::
* AArch64 Function Attributes::
* AMD GCN Function Attributes::
* ARC Function Attributes::
* ARM Function Attributes::
* AVR Function Attributes::
* Blackfin Function Attributes::
* CR16 Function Attributes::
* C-SKY Function Attributes::
* Epiphany Function Attributes::
* H8/300 Function Attributes::
* IA-64 Function Attributes::
* M32C Function Attributes::
* M32R/D Function Attributes::
* m68k Function Attributes::
* MCORE Function Attributes::
* MeP Function Attributes::
* MicroBlaze Function Attributes::
* Microsoft Windows Function Attributes::
* MIPS Function Attributes::
* MSP430 Function Attributes::
* NDS32 Function Attributes::
* Nios II Function Attributes::
* Nvidia PTX Function Attributes::
* PowerPC Function Attributes::
* RISC-V Function Attributes::
* RL78 Function Attributes::
* RX Function Attributes::
* S/390 Function Attributes::
* SH Function Attributes::
* Symbian OS Function Attributes::
* V850 Function Attributes::
* Visium Function Attributes::
* x86 Function Attributes::
* Xstormy16 Function Attributes::
@end menu

@node Common Function Attributes
@subsection Common Function Attributes

The following attributes are supported on most targets.

@table @code
@c Keep this table alphabetized by attribute name.  Treat _ as space.

@item access
@itemx access (@var{access-mode}, @var{ref-index})
@itemx access (@var{access-mode}, @var{ref-index}, @var{size-index})

The @code{access} attribute enables the detection of invalid or unsafe
accesses by functions to which they apply or their callers, as well as
write-only accesses to objects that are never read from.  Such accesses
may be diagnosed by warnings such as @option{-Wstringop-overflow},
@option{-Wuninitialized}, @option{-Wunused}, and others.

The @code{access} attribute specifies that a function to whose by-reference
arguments the attribute applies accesses the referenced object according to
@var{access-mode}.  The @var{access-mode} argument is required and must be
one of three names: @code{read_only}, @code{read_write}, or @code{write_only}.
The remaining two are positional arguments.

The required @var{ref-index} positional argument  denotes a function
argument of pointer (or in C++, reference) type that is subject to
the access.  The same pointer argument can be referenced by at most one
distinct @code{access} attribute.

The optional @var{size-index} positional argument denotes a function
argument of integer type that specifies the maximum size of the access.
The size is the number of elements of the type referenced by @var{ref-index},
or the number of bytes when the pointer type is @code{void*}.  When no
@var{size-index} argument is specified, the pointer argument must be either
null or point to a space that is suitably aligned and large for at least one
object of the referenced type (this implies that a past-the-end pointer is
not a valid argument).  The actual size of the access may be less but it
must not be more.

The @code{read_only} access mode specifies that the pointer to which it
applies is used to read the referenced object but not write to it.  Unless
the argument specifying the size of the access denoted by @var{size-index}
is zero, the referenced object must be initialized.  The mode implies
a stronger guarantee than the @code{const} qualifier which, when cast away
from a pointer, does not prevent the pointed-to object from being modified.
Examples of the use of the @code{read_only} access mode is the argument to
the @code{puts} function, or the second and third arguments to
the @code{memcpy} function.

@smallexample
__attribute__ ((access (read_only))) int puts (const char*);
__attribute__ ((access (read_only, 1, 2))) void* memcpy (void*, const void*, size_t);
@end smallexample

The @code{read_write} access mode applies to arguments of pointer types
without the @code{const} qualifier.  It specifies that the pointer to which
it applies is used to both read and write the referenced object.  Unless
the argument specifying the size of the access denoted by @var{size-index}
is zero, the object referenced by the pointer must be initialized.  An example
of the use of the @code{read_write} access mode is the first argument to
the @code{strcat} function.

@smallexample
__attribute__ ((access (read_write, 1), access (read_only, 2))) char* strcat (char*, const char*);
@end smallexample

The @code{write_only} access mode applies to arguments of pointer types
without the @code{const} qualifier.  It specifies that the pointer to which
it applies is used to write to the referenced object but not read from it.
The object refrenced by the pointer need not be initialized.  An example
of the use of the @code{write_only} access mode is the first argument to
the @code{strcpy} function, or the first two arguments to the @code{fgets}
function.

@smallexample
__attribute__ ((access (write_only, 1), access (read_only, 2))) char* strcpy (char*, const char*);
__attribute__ ((access (write_only, 1, 2), access (read_write, 3))) int fgets (char*, int, FILE*);
@end smallexample

@item alias ("@var{target}")
@cindex @code{alias} function attribute
The @code{alias} attribute causes the declaration to be emitted as an
alias for another symbol, which must be specified.  For instance,

@smallexample
void __f () @{ /* @r{Do something.} */; @}
void f () __attribute__ ((weak, alias ("__f")));
@end smallexample

@noindent
defines @samp{f} to be a weak alias for @samp{__f}.  In C++, the
mangled name for the target must be used.  It is an error if @samp{__f}
is not defined in the same translation unit.

This attribute requires assembler and object file support,
and may not be available on all targets.

@item aligned
@itemx aligned (@var{alignment})
@cindex @code{aligned} function attribute
The @code{aligned} attribute specifies a minimum alignment for
the first instruction of the function, measured in bytes.  When specified,
@var{alignment} must be an integer constant power of 2.  Specifying no
@var{alignment} argument implies the ideal alignment for the target.
The @code{__alignof__} operator can be used to determine what that is
(@pxref{Alignment}).  The attribute has no effect when a definition for
the function is not provided in the same translation unit.

The attribute cannot be used to decrease the alignment of a function
previously declared with a more restrictive alignment; only to increase
it.  Attempts to do otherwise are diagnosed.  Some targets specify
a minimum default alignment for functions that is greater than 1.  On
such targets, specifying a less restrictive alignment is silently ignored.
Using the attribute overrides the effect of the @option{-falign-functions}
(@pxref{Optimize Options}) option for this function.

Note that the effectiveness of @code{aligned} attributes may be
limited by inherent limitations in the system linker 
and/or object file format.  On some systems, the
linker is only able to arrange for functions to be aligned up to a
certain maximum alignment.  (For some linkers, the maximum supported
alignment may be very very small.)  See your linker documentation for
further information.

The @code{aligned} attribute can also be used for variables and fields
(@pxref{Variable Attributes}.)

@item alloc_align (@var{position})
@cindex @code{alloc_align} function attribute
The @code{alloc_align} attribute may be applied to a function that
returns a pointer and takes at least one argument of an integer or
enumerated type.
It indicates that the returned pointer is aligned on a boundary given
by the function argument at @var{position}.  Meaningful alignments are
powers of 2 greater than one.  GCC uses this information to improve
pointer alignment analysis.

The function parameter denoting the allocated alignment is specified by
one constant integer argument whose number is the argument of the attribute.
Argument numbering starts at one.

For instance,

@smallexample
void* my_memalign (size_t, size_t) __attribute__ ((alloc_align (1)));
@end smallexample

@noindent
declares that @code{my_memalign} returns memory with minimum alignment
given by parameter 1.

@item alloc_size (@var{position})
@itemx alloc_size (@var{position-1}, @var{position-2})
@cindex @code{alloc_size} function attribute
The @code{alloc_size} attribute may be applied to a function that
returns a pointer and takes at least one argument of an integer or
enumerated type.
It indicates that the returned pointer points to memory whose size is
given by the function argument at @var{position-1}, or by the product
of the arguments at @var{position-1} and @var{position-2}.  Meaningful
sizes are positive values less than @code{PTRDIFF_MAX}.  GCC uses this
information to improve the results of @code{__builtin_object_size}.

The function parameter(s) denoting the allocated size are specified by
one or two integer arguments supplied to the attribute.  The allocated size
is either the value of the single function argument specified or the product
of the two function arguments specified.  Argument numbering starts at
one for ordinary functions, and at two for C++ non-static member functions.

For instance,

@smallexample
void* my_calloc (size_t, size_t) __attribute__ ((alloc_size (1, 2)));
void* my_realloc (void*, size_t) __attribute__ ((alloc_size (2)));
@end smallexample

@noindent
declares that @code{my_calloc} returns memory of the size given by
the product of parameter 1 and 2 and that @code{my_realloc} returns memory
of the size given by parameter 2.

@item always_inline
@cindex @code{always_inline} function attribute
Generally, functions are not inlined unless optimization is specified.
For functions declared inline, this attribute inlines the function
independent of any restrictions that otherwise apply to inlining.
Failure to inline such a function is diagnosed as an error.
Note that if such a function is called indirectly the compiler may
or may not inline it depending on optimization level and a failure
to inline an indirect call may or may not be diagnosed.

@item artificial
@cindex @code{artificial} function attribute
This attribute is useful for small inline wrappers that if possible
should appear during debugging as a unit.  Depending on the debug
info format it either means marking the function as artificial
or using the caller location for all instructions within the inlined
body.

@item assume_aligned (@var{alignment})
@itemx assume_aligned (@var{alignment}, @var{offset})
@cindex @code{assume_aligned} function attribute
The @code{assume_aligned} attribute may be applied to a function that
returns a pointer.  It indicates that the returned pointer is aligned
on a boundary given by @var{alignment}.  If the attribute has two
arguments, the second argument is misalignment @var{offset}.  Meaningful
values of @var{alignment} are powers of 2 greater than one.  Meaningful
values of @var{offset} are greater than zero and less than @var{alignment}.

For instance

@smallexample
void* my_alloc1 (size_t) __attribute__((assume_aligned (16)));
void* my_alloc2 (size_t) __attribute__((assume_aligned (32, 8)));
@end smallexample

@noindent
declares that @code{my_alloc1} returns 16-byte aligned pointers and
that @code{my_alloc2} returns a pointer whose value modulo 32 is equal
to 8.

@item cold
@cindex @code{cold} function attribute
The @code{cold} attribute on functions is used to inform the compiler that
the function is unlikely to be executed.  The function is optimized for
size rather than speed and on many targets it is placed into a special
subsection of the text section so all cold functions appear close together,
improving code locality of non-cold parts of program.  The paths leading
to calls of cold functions within code are marked as unlikely by the branch
prediction mechanism.  It is thus useful to mark functions used to handle
unlikely conditions, such as @code{perror}, as cold to improve optimization
of hot functions that do call marked functions in rare occasions.

When profile feedback is available, via @option{-fprofile-use}, cold functions
are automatically detected and this attribute is ignored.

@item const
@cindex @code{const} function attribute
@cindex functions that have no side effects
Calls to functions whose return value is not affected by changes to
the observable state of the program and that have no observable effects
on such state other than to return a value may lend themselves to
optimizations such as common subexpression elimination.  Declaring such
functions with the @code{const} attribute allows GCC to avoid emitting
some calls in repeated invocations of the function with the same argument
values.

For example,

@smallexample
int square (int) __attribute__ ((const));
@end smallexample

@noindent
tells GCC that subsequent calls to function @code{square} with the same
argument value can be replaced by the result of the first call regardless
of the statements in between.

The @code{const} attribute prohibits a function from reading objects
that affect its return value between successive invocations.  However,
functions declared with the attribute can safely read objects that do
not change their return value, such as non-volatile constants.

The @code{const} attribute imposes greater restrictions on a function's
definition than the similar @code{pure} attribute.  Declaring the same
function with both the @code{const} and the @code{pure} attribute is
diagnosed.  Because a const function cannot have any observable side
effects it does not make sense for it to return @code{void}.  Declaring
such a function is diagnosed.

@cindex pointer arguments
Note that a function that has pointer arguments and examines the data
pointed to must @emph{not} be declared @code{const} if the pointed-to
data might change between successive invocations of the function.  In
general, since a function cannot distinguish data that might change
from data that cannot, const functions should never take pointer or,
in C++, reference arguments. Likewise, a function that calls a non-const
function usually must not be const itself.

@item constructor
@itemx destructor
@itemx constructor (@var{priority})
@itemx destructor (@var{priority})
@cindex @code{constructor} function attribute
@cindex @code{destructor} function attribute
The @code{constructor} attribute causes the function to be called
automatically before execution enters @code{main ()}.  Similarly, the
@code{destructor} attribute causes the function to be called
automatically after @code{main ()} completes or @code{exit ()} is
called.  Functions with these attributes are useful for
initializing data that is used implicitly during the execution of
the program.

On some targets the attributes also accept an integer argument to
specify a priority to control the order in which constructor and
destructor functions are run.  A constructor
with a smaller priority number runs before a constructor with a larger
priority number; the opposite relationship holds for destructors.  So,
if you have a constructor that allocates a resource and a destructor
that deallocates the same resource, both functions typically have the
same priority.  The priorities for constructor and destructor
functions are the same as those specified for namespace-scope C++
objects (@pxref{C++ Attributes}).  However, at present, the order in which
constructors for C++ objects with static storage duration and functions
decorated with attribute @code{constructor} are invoked is unspecified.
In mixed declarations, attribute @code{init_priority} can be used to
impose a specific ordering.

Using the argument forms of the @code{constructor} and @code{destructor}
attributes on targets where the feature is not supported is rejected with
an error.

@item copy
@itemx copy (@var{function})
@cindex @code{copy} function attribute
The @code{copy} attribute applies the set of attributes with which
@var{function} has been declared to the declaration of the function
to which the attribute is applied.  The attribute is designed for
libraries that define aliases or function resolvers that are expected
to specify the same set of attributes as their targets.  The @code{copy}
attribute can be used with functions, variables, or types.  However,
the kind of symbol to which the attribute is applied (either function
or variable) must match the kind of symbol to which the argument refers.
The @code{copy} attribute copies only syntactic and semantic attributes
but not attributes that affect a symbol's linkage or visibility such as
@code{alias}, @code{visibility}, or @code{weak}.  The @code{deprecated}
and @code{target_clones} attribute are also not copied.
@xref{Common Type Attributes}.
@xref{Common Variable Attributes}.

For example, the @var{StrongAlias} macro below makes use of the @code{alias}
and @code{copy} attributes to define an alias named @var{alloc} for function
@var{allocate} declared with attributes @var{alloc_size}, @var{malloc}, and
@var{nothrow}.  Thanks to the @code{__typeof__} operator the alias has
the same type as the target function.  As a result of the @code{copy}
attribute the alias also shares the same attributes as the target.

@smallexample
#define StrongAlias(TagetFunc, AliasDecl)   \
  extern __typeof__ (TargetFunc) AliasDecl  \
    __attribute__ ((alias (#TargetFunc), copy (TargetFunc)));

extern __attribute__ ((alloc_size (1), malloc, nothrow))
  void* allocate (size_t);
StrongAlias (allocate, alloc);
@end smallexample

@item deprecated
@itemx deprecated (@var{msg})
@cindex @code{deprecated} function attribute
The @code{deprecated} attribute results in a warning if the function
is used anywhere in the source file.  This is useful when identifying
functions that are expected to be removed in a future version of a
program.  The warning also includes the location of the declaration
of the deprecated function, to enable users to easily find further
information about why the function is deprecated, or what they should
do instead.  Note that the warnings only occurs for uses:

@smallexample
int old_fn () __attribute__ ((deprecated));
int old_fn ();
int (*fn_ptr)() = old_fn;
@end smallexample

@noindent
results in a warning on line 3 but not line 2.  The optional @var{msg}
argument, which must be a string, is printed in the warning if
present.

The @code{deprecated} attribute can also be used for variables and
types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)

The message attached to the attribute is affected by the setting of
the @option{-fmessage-length} option.

@item error ("@var{message}")
@itemx warning ("@var{message}")
@cindex @code{error} function attribute
@cindex @code{warning} function attribute
If the @code{error} or @code{warning} attribute 
is used on a function declaration and a call to such a function
is not eliminated through dead code elimination or other optimizations, 
an error or warning (respectively) that includes @var{message} is diagnosed.  
This is useful
for compile-time checking, especially together with @code{__builtin_constant_p}
and inline functions where checking the inline function arguments is not
possible through @code{extern char [(condition) ? 1 : -1];} tricks.

While it is possible to leave the function undefined and thus invoke
a link failure (to define the function with
a message in @code{.gnu.warning*} section),
when using these attributes the problem is diagnosed
earlier and with exact location of the call even in presence of inline
functions or when not emitting debugging information.

@item externally_visible
@cindex @code{externally_visible} function attribute
This attribute, attached to a global variable or function, nullifies
the effect of the @option{-fwhole-program} command-line option, so the
object remains visible outside the current compilation unit.

If @option{-fwhole-program} is used together with @option{-flto} and 
@command{gold} is used as the linker plugin, 
@code{externally_visible} attributes are automatically added to functions 
(not variable yet due to a current @command{gold} issue) 
that are accessed outside of LTO objects according to resolution file
produced by @command{gold}.
For other linkers that cannot generate resolution file,
explicit @code{externally_visible} attributes are still necessary.

@item flatten
@cindex @code{flatten} function attribute
Generally, inlining into a function is limited.  For a function marked with
this attribute, every call inside this function is inlined, if possible.
Functions declared with attribute @code{noinline} and similar are not
inlined.  Whether the function itself is considered for inlining depends
on its size and the current inlining parameters.

@item format (@var{archetype}, @var{string-index}, @var{first-to-check})
@cindex @code{format} function attribute
@cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
@opindex Wformat
The @code{format} attribute specifies that a function takes @code{printf},
@code{scanf}, @code{strftime} or @code{strfmon} style arguments that
should be type-checked against a format string.  For example, the
declaration:

@smallexample
extern int
my_printf (void *my_object, const char *my_format, ...)
      __attribute__ ((format (printf, 2, 3)));
@end smallexample

@noindent
causes the compiler to check the arguments in calls to @code{my_printf}
for consistency with the @code{printf} style format string argument
@code{my_format}.

The parameter @var{archetype} determines how the format string is
interpreted, and should be @code{printf}, @code{scanf}, @code{strftime},
@code{gnu_printf}, @code{gnu_scanf}, @code{gnu_strftime} or
@code{strfmon}.  (You can also use @code{__printf__},
@code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.)  On
MinGW targets, @code{ms_printf}, @code{ms_scanf}, and
@code{ms_strftime} are also present.
@var{archetype} values such as @code{printf} refer to the formats accepted
by the system's C runtime library,
while values prefixed with @samp{gnu_} always refer
to the formats accepted by the GNU C Library.  On Microsoft Windows
targets, values prefixed with @samp{ms_} refer to the formats accepted by the
@file{msvcrt.dll} library.
The parameter @var{string-index}
specifies which argument is the format string argument (starting
from 1), while @var{first-to-check} is the number of the first
argument to check against the format string.  For functions
where the arguments are not available to be checked (such as
@code{vprintf}), specify the third parameter as zero.  In this case the
compiler only checks the format string for consistency.  For
@code{strftime} formats, the third parameter is required to be zero.
Since non-static C++ methods have an implicit @code{this} argument, the
arguments of such methods should be counted from two, not one, when
giving values for @var{string-index} and @var{first-to-check}.

In the example above, the format string (@code{my_format}) is the second
argument of the function @code{my_print}, and the arguments to check
start with the third argument, so the correct parameters for the format
attribute are 2 and 3.

@opindex ffreestanding
@opindex fno-builtin
The @code{format} attribute allows you to identify your own functions
that take format strings as arguments, so that GCC can check the
calls to these functions for errors.  The compiler always (unless
@option{-ffreestanding} or @option{-fno-builtin} is used) checks formats
for the standard library functions @code{printf}, @code{fprintf},
@code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
@code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
warnings are requested (using @option{-Wformat}), so there is no need to
modify the header file @file{stdio.h}.  In C99 mode, the functions
@code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
@code{vsscanf} are also checked.  Except in strictly conforming C
standard modes, the X/Open function @code{strfmon} is also checked as
are @code{printf_unlocked} and @code{fprintf_unlocked}.
@xref{C Dialect Options,,Options Controlling C Dialect}.

For Objective-C dialects, @code{NSString} (or @code{__NSString__}) is
recognized in the same context.  Declarations including these format attributes
are parsed for correct syntax, however the result of checking of such format
strings is not yet defined, and is not carried out by this version of the
compiler.

The target may also provide additional types of format checks.
@xref{Target Format Checks,,Format Checks Specific to Particular
Target Machines}.

@item format_arg (@var{string-index})
@cindex @code{format_arg} function attribute
@opindex Wformat-nonliteral
The @code{format_arg} attribute specifies that a function takes one or
more format strings for a @code{printf}, @code{scanf}, @code{strftime} or
@code{strfmon} style function and modifies it (for example, to translate
it into another language), so the result can be passed to a
@code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
function (with the remaining arguments to the format function the same
as they would have been for the unmodified string).  Multiple
@code{format_arg} attributes may be applied to the same function, each
designating a distinct parameter as a format string.  For example, the
declaration:

@smallexample
extern char *
my_dgettext (char *my_domain, const char *my_format)
      __attribute__ ((format_arg (2)));
@end smallexample

@noindent
causes the compiler to check the arguments in calls to a @code{printf},
@code{scanf}, @code{strftime} or @code{strfmon} type function, whose
format string argument is a call to the @code{my_dgettext} function, for
consistency with the format string argument @code{my_format}.  If the
@code{format_arg} attribute had not been specified, all the compiler
could tell in such calls to format functions would be that the format
string argument is not constant; this would generate a warning when
@option{-Wformat-nonliteral} is used, but the calls could not be checked
without the attribute.

In calls to a function declared with more than one @code{format_arg}
attribute, each with a distinct argument value, the corresponding
actual function arguments are checked against all format strings
designated by the attributes.  This capability is designed to support
the GNU @code{ngettext} family of functions.

The parameter @var{string-index} specifies which argument is the format
string argument (starting from one).  Since non-static C++ methods have
an implicit @code{this} argument, the arguments of such methods should
be counted from two.

The @code{format_arg} attribute allows you to identify your own
functions that modify format strings, so that GCC can check the
calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
type function whose operands are a call to one of your own function.
The compiler always treats @code{gettext}, @code{dgettext}, and
@code{dcgettext} in this manner except when strict ISO C support is
requested by @option{-ansi} or an appropriate @option{-std} option, or
@option{-ffreestanding} or @option{-fno-builtin}
is used.  @xref{C Dialect Options,,Options
Controlling C Dialect}.

For Objective-C dialects, the @code{format-arg} attribute may refer to an
@code{NSString} reference for compatibility with the @code{format} attribute
above.

The target may also allow additional types in @code{format-arg} attributes.
@xref{Target Format Checks,,Format Checks Specific to Particular
Target Machines}.

@item gnu_inline
@cindex @code{gnu_inline} function attribute
This attribute should be used with a function that is also declared
with the @code{inline} keyword.  It directs GCC to treat the function
as if it were defined in gnu90 mode even when compiling in C99 or
gnu99 mode.

If the function is declared @code{extern}, then this definition of the
function is used only for inlining.  In no case is the function
compiled as a standalone function, not even if you take its address
explicitly.  Such an address becomes an external reference, as if you
had only declared the function, and had not defined it.  This has
almost the effect of a macro.  The way to use this is to put a
function definition in a header file with this attribute, and put
another copy of the function, without @code{extern}, in a library
file.  The definition in the header file causes most calls to the
function to be inlined.  If any uses of the function remain, they
refer to the single copy in the library.  Note that the two
definitions of the functions need not be precisely the same, although
if they do not have the same effect your program may behave oddly.

In C, if the function is neither @code{extern} nor @code{static}, then
the function is compiled as a standalone function, as well as being
inlined where possible.

This is how GCC traditionally handled functions declared
@code{inline}.  Since ISO C99 specifies a different semantics for
@code{inline}, this function attribute is provided as a transition
measure and as a useful feature in its own right.  This attribute is
available in GCC 4.1.3 and later.  It is available if either of the
preprocessor macros @code{__GNUC_GNU_INLINE__} or
@code{__GNUC_STDC_INLINE__} are defined.  @xref{Inline,,An Inline
Function is As Fast As a Macro}.

In C++, this attribute does not depend on @code{extern} in any way,
but it still requires the @code{inline} keyword to enable its special
behavior.

@item hot
@cindex @code{hot} function attribute
The @code{hot} attribute on a function is used to inform the compiler that
the function is a hot spot of the compiled program.  The function is
optimized more aggressively and on many targets it is placed into a special
subsection of the text section so all hot functions appear close together,
improving locality.

When profile feedback is available, via @option{-fprofile-use}, hot functions
are automatically detected and this attribute is ignored.

@item ifunc ("@var{resolver}")
@cindex @code{ifunc} function attribute
@cindex indirect functions
@cindex functions that are dynamically resolved
The @code{ifunc} attribute is used to mark a function as an indirect
function using the STT_GNU_IFUNC symbol type extension to the ELF
standard.  This allows the resolution of the symbol value to be
determined dynamically at load time, and an optimized version of the
routine to be selected for the particular processor or other system
characteristics determined then.  To use this attribute, first define
the implementation functions available, and a resolver function that
returns a pointer to the selected implementation function.  The
implementation functions' declarations must match the API of the
function being implemented.  The resolver should be declared to
be a function taking no arguments and returning a pointer to
a function of the same type as the implementation.  For example:

@smallexample
void *my_memcpy (void *dst, const void *src, size_t len)
@{
  @dots{}
  return dst;
@}

static void * (*resolve_memcpy (void))(void *, const void *, size_t)
@{
  return my_memcpy; // we will just always select this routine
@}
@end smallexample

@noindent
The exported header file declaring the function the user calls would
contain:

@smallexample
extern void *memcpy (void *, const void *, size_t);
@end smallexample

@noindent
allowing the user to call @code{memcpy} as a regular function, unaware of
the actual implementation.  Finally, the indirect function needs to be
defined in the same translation unit as the resolver function:

@smallexample
void *memcpy (void *, const void *, size_t)
     __attribute__ ((ifunc ("resolve_memcpy")));
@end smallexample

In C++, the @code{ifunc} attribute takes a string that is the mangled name
of the resolver function.  A C++ resolver for a non-static member function
of class @code{C} should be declared to return a pointer to a non-member
function taking pointer to @code{C} as the first argument, followed by
the same arguments as of the implementation function.  G++ checks
the signatures of the two functions and issues
a @option{-Wattribute-alias} warning for mismatches.  To suppress a warning
for the necessary cast from a pointer to the implementation member function
to the type of the corresponding non-member function use
the @option{-Wno-pmf-conversions} option.  For example:

@smallexample
class S
@{
private:
  int debug_impl (int);
  int optimized_impl (int);

  typedef int Func (S*, int);

  static Func* resolver ();
public:

  int interface (int);
@};

int S::debug_impl (int) @{ /* @r{@dots{}} */ @}
int S::optimized_impl (int) @{ /* @r{@dots{}} */ @}

S::Func* S::resolver ()
@{
  int (S::*pimpl) (int)
    = getenv ("DEBUG") ? &S::debug_impl : &S::optimized_impl;

  // Cast triggers -Wno-pmf-conversions.
  return reinterpret_cast<Func*>(pimpl);
@}

int S::interface (int) __attribute__ ((ifunc ("_ZN1S8resolverEv")));
@end smallexample

Indirect functions cannot be weak.  Binutils version 2.20.1 or higher
and GNU C Library version 2.11.1 are required to use this feature.

@item interrupt
@itemx interrupt_handler
Many GCC back ends support attributes to indicate that a function is
an interrupt handler, which tells the compiler to generate function
entry and exit sequences that differ from those from regular
functions.  The exact syntax and behavior are target-specific;
refer to the following subsections for details.

@item leaf
@cindex @code{leaf} function attribute
Calls to external functions with this attribute must return to the
current compilation unit only by return or by exception handling.  In
particular, a leaf function is not allowed to invoke callback functions
passed to it from the current compilation unit, directly call functions
exported by the unit, or @code{longjmp} into the unit.  Leaf functions
might still call functions from other compilation units and thus they
are not necessarily leaf in the sense that they contain no function
calls at all.

The attribute is intended for library functions to improve dataflow
analysis.  The compiler takes the hint that any data not escaping the
current compilation unit cannot be used or modified by the leaf
function.  For example, the @code{sin} function is a leaf function, but
@code{qsort} is not.

Note that leaf functions might indirectly run a signal handler defined
in the current compilation unit that uses static variables.  Similarly,
when lazy symbol resolution is in effect, leaf functions might invoke
indirect functions whose resolver function or implementation function is
defined in the current compilation unit and uses static variables.  There
is no standard-compliant way to write such a signal handler, resolver
function, or implementation function, and the best that you can do is to
remove the @code{leaf} attribute or mark all such static variables
@code{volatile}.  Lastly, for ELF-based systems that support symbol
interposition, care should be taken that functions defined in the
current compilation unit do not unexpectedly interpose other symbols
based on the defined standards mode and defined feature test macros;
otherwise an inadvertent callback would be added.

The attribute has no effect on functions defined within the current
compilation unit.  This is to allow easy merging of multiple compilation
units into one, for example, by using the link-time optimization.  For
this reason the attribute is not allowed on types to annotate indirect
calls.

@item malloc
@cindex @code{malloc} function attribute
@cindex functions that behave like malloc
This tells the compiler that a function is @code{malloc}-like, i.e.,
that the pointer @var{P} returned by the function cannot alias any
other pointer valid when the function returns, and moreover no
pointers to valid objects occur in any storage addressed by @var{P}.

Using this attribute can improve optimization.  Compiler predicts
that a function with the attribute returns non-null in most cases.
Functions like
@code{malloc} and @code{calloc} have this property because they return
a pointer to uninitialized or zeroed-out storage.  However, functions
like @code{realloc} do not have this property, as they can return a
pointer to storage containing pointers.

@item no_icf
@cindex @code{no_icf} function attribute
This function attribute prevents a functions from being merged with another
semantically equivalent function.

@item no_instrument_function
@cindex @code{no_instrument_function} function attribute
@opindex finstrument-functions
@opindex p
@opindex pg
If any of @option{-finstrument-functions}, @option{-p}, or @option{-pg} are 
given, profiling function calls are
generated at entry and exit of most user-compiled functions.
Functions with this attribute are not so instrumented.

@item no_profile_instrument_function
@cindex @code{no_profile_instrument_function} function attribute
The @code{no_profile_instrument_function} attribute on functions is used
to inform the compiler that it should not process any profile feedback based
optimization code instrumentation.

@item no_reorder
@cindex @code{no_reorder} function attribute
Do not reorder functions or variables marked @code{no_reorder}
against each other or top level assembler statements the executable.
The actual order in the program will depend on the linker command
line. Static variables marked like this are also not removed.
This has a similar effect
as the @option{-fno-toplevel-reorder} option, but only applies to the
marked symbols.

@item no_sanitize ("@var{sanitize_option}")
@cindex @code{no_sanitize} function attribute
The @code{no_sanitize} attribute on functions is used
to inform the compiler that it should not do sanitization of any option
mentioned in @var{sanitize_option}.  A list of values acceptable by
the @option{-fsanitize} option can be provided.

@smallexample
void __attribute__ ((no_sanitize ("alignment", "object-size")))
f () @{ /* @r{Do something.} */; @}
void __attribute__ ((no_sanitize ("alignment,object-size")))
g () @{ /* @r{Do something.} */; @}
@end smallexample

@item no_sanitize_address
@itemx no_address_safety_analysis
@cindex @code{no_sanitize_address} function attribute
The @code{no_sanitize_address} attribute on functions is used
to inform the compiler that it should not instrument memory accesses
in the function when compiling with the @option{-fsanitize=address} option.
The @code{no_address_safety_analysis} is a deprecated alias of the
@code{no_sanitize_address} attribute, new code should use
@code{no_sanitize_address}.

@item no_sanitize_thread
@cindex @code{no_sanitize_thread} function attribute
The @code{no_sanitize_thread} attribute on functions is used
to inform the compiler that it should not instrument memory accesses
in the function when compiling with the @option{-fsanitize=thread} option.

@item no_sanitize_undefined
@cindex @code{no_sanitize_undefined} function attribute
The @code{no_sanitize_undefined} attribute on functions is used
to inform the compiler that it should not check for undefined behavior
in the function when compiling with the @option{-fsanitize=undefined} option.

@item no_split_stack
@cindex @code{no_split_stack} function attribute
@opindex fsplit-stack
If @option{-fsplit-stack} is given, functions have a small
prologue which decides whether to split the stack.  Functions with the
@code{no_split_stack} attribute do not have that prologue, and thus
may run with only a small amount of stack space available.

@item no_stack_limit
@cindex @code{no_stack_limit} function attribute
This attribute locally overrides the @option{-fstack-limit-register}
and @option{-fstack-limit-symbol} command-line options; it has the effect
of disabling stack limit checking in the function it applies to.

@item noclone
@cindex @code{noclone} function attribute
This function attribute prevents a function from being considered for
cloning---a mechanism that produces specialized copies of functions
and which is (currently) performed by interprocedural constant
propagation.

@item noinline
@cindex @code{noinline} function attribute
This function attribute prevents a function from being considered for
inlining.
@c Don't enumerate the optimizations by name here; we try to be
@c future-compatible with this mechanism.
If the function does not have side effects, there are optimizations
other than inlining that cause function calls to be optimized away,
although the function call is live.  To keep such calls from being
optimized away, put
@smallexample
asm ("");
@end smallexample

@noindent
(@pxref{Extended Asm}) in the called function, to serve as a special
side effect.

@item noipa
@cindex @code{noipa} function attribute
Disable interprocedural optimizations between the function with this
attribute and its callers, as if the body of the function is not available
when optimizing callers and the callers are unavailable when optimizing
the body.  This attribute implies @code{noinline}, @code{noclone} and
@code{no_icf} attributes.    However, this attribute is not equivalent
to a combination of other attributes, because its purpose is to suppress
existing and future optimizations employing interprocedural analysis,
including those that do not have an attribute suitable for disabling
them individually.  This attribute is supported mainly for the purpose
of testing the compiler.

@item nonnull
@itemx nonnull (@var{arg-index}, @dots{})
@cindex @code{nonnull} function attribute
@cindex functions with non-null pointer arguments
The @code{nonnull} attribute may be applied to a function that takes at
least one argument of a pointer type.  It indicates that the referenced
arguments must be non-null pointers.  For instance, the declaration:

@smallexample
extern void *
my_memcpy (void *dest, const void *src, size_t len)
        __attribute__((nonnull (1, 2)));
@end smallexample

@noindent
causes the compiler to check that, in calls to @code{my_memcpy},
arguments @var{dest} and @var{src} are non-null.  If the compiler
determines that a null pointer is passed in an argument slot marked
as non-null, and the @option{-Wnonnull} option is enabled, a warning
is issued.  @xref{Warning Options}.  Unless disabled by
the @option{-fno-delete-null-pointer-checks} option the compiler may
also perform optimizations based on the knowledge that certain function
arguments cannot be null. In addition,
the @option{-fisolate-erroneous-paths-attribute} option can be specified
to have GCC transform calls with null arguments to non-null functions
into traps. @xref{Optimize Options}.

If no @var{arg-index} is given to the @code{nonnull} attribute,
all pointer arguments are marked as non-null.  To illustrate, the
following declaration is equivalent to the previous example:

@smallexample
extern void *
my_memcpy (void *dest, const void *src, size_t len)
        __attribute__((nonnull));
@end smallexample

@item noplt
@cindex @code{noplt} function attribute
The @code{noplt} attribute is the counterpart to option @option{-fno-plt}.
Calls to functions marked with this attribute in position-independent code
do not use the PLT.

@smallexample
@group
/* Externally defined function foo.  */
int foo () __attribute__ ((noplt));

int
main (/* @r{@dots{}} */)
@{
  /* @r{@dots{}} */
  foo ();
  /* @r{@dots{}} */
@}
@end group
@end smallexample

The @code{noplt} attribute on function @code{foo}
tells the compiler to assume that
the function @code{foo} is externally defined and that the call to
@code{foo} must avoid the PLT
in position-independent code.

In position-dependent code, a few targets also convert calls to
functions that are marked to not use the PLT to use the GOT instead.

@item noreturn
@cindex @code{noreturn} function attribute
@cindex functions that never return
A few standard library functions, such as @code{abort} and @code{exit},
cannot return.  GCC knows this automatically.  Some programs define
their own functions that never return.  You can declare them
@code{noreturn} to tell the compiler this fact.  For example,

@smallexample
@group
void fatal () __attribute__ ((noreturn));

void
fatal (/* @r{@dots{}} */)
@{
  /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
  exit (1);
@}
@end group
@end smallexample

The @code{noreturn} keyword tells the compiler to assume that
@code{fatal} cannot return.  It can then optimize without regard to what
would happen if @code{fatal} ever did return.  This makes slightly
better code.  More importantly, it helps avoid spurious warnings of
uninitialized variables.

The @code{noreturn} keyword does not affect the exceptional path when that
applies: a @code{noreturn}-marked function may still return to the caller
by throwing an exception or calling @code{longjmp}.

In order to preserve backtraces, GCC will never turn calls to
@code{noreturn} functions into tail calls.

Do not assume that registers saved by the calling function are
restored before calling the @code{noreturn} function.

It does not make sense for a @code{noreturn} function to have a return
type other than @code{void}.

@item nothrow
@cindex @code{nothrow} function attribute
The @code{nothrow} attribute is used to inform the compiler that a
function cannot throw an exception.  For example, most functions in
the standard C library can be guaranteed not to throw an exception
with the notable exceptions of @code{qsort} and @code{bsearch} that
take function pointer arguments.

@item optimize (@var{level}, @dots{})
@item optimize (@var{string}, @dots{})
@cindex @code{optimize} function attribute
The @code{optimize} attribute is used to specify that a function is to
be compiled with different optimization options than specified on the
command line.  Valid arguments are constant non-negative integers and
strings.  Each numeric argument specifies an optimization @var{level}.
Each @var{string} argument consists of one or more comma-separated
substrings.  Each substring that begins with the letter @code{O} refers
to an optimization option such as @option{-O0} or @option{-Os}.  Other
substrings are taken as suffixes to the @code{-f} prefix jointly
forming the name of an optimization option.  @xref{Optimize Options}.

@samp{#pragma GCC optimize} can be used to set optimization options
for more than one function.  @xref{Function Specific Option Pragmas},
for details about the pragma.

Providing multiple strings as arguments separated by commas to specify
multiple options is equivalent to separating the option suffixes with
a comma (@samp{,}) within a single string.  Spaces are not permitted
within the strings.

Not every optimization option that starts with the @var{-f} prefix
specified by the attribute necessarily has an effect on the function.
The @code{optimize} attribute should be used for debugging purposes only.
It is not suitable in production code.

@item patchable_function_entry
@cindex @code{patchable_function_entry} function attribute
@cindex extra NOP instructions at the function entry point
In case the target's text segment can be made writable at run time by
any means, padding the function entry with a number of NOPs can be
used to provide a universal tool for instrumentation.

The @code{patchable_function_entry} function attribute can be used to
change the number of NOPs to any desired value.  The two-value syntax
is the same as for the command-line switch
@option{-fpatchable-function-entry=N,M}, generating @var{N} NOPs, with
the function entry point before the @var{M}th NOP instruction.
@var{M} defaults to 0 if omitted e.g.@: function entry point is before
the first NOP.

If patchable function entries are enabled globally using the command-line
option @option{-fpatchable-function-entry=N,M}, then you must disable
instrumentation on all functions that are part of the instrumentation
framework with the attribute @code{patchable_function_entry (0)}
to prevent recursion.

@item pure
@cindex @code{pure} function attribute
@cindex functions that have no side effects

Calls to functions that have no observable effects on the state of
the program other than to return a value may lend themselves to optimizations
such as common subexpression elimination.  Declaring such functions with
the @code{pure} attribute allows GCC to avoid emitting some calls in repeated
invocations of the function with the same argument values.

The @code{pure} attribute prohibits a function from modifying the state
of the program that is observable by means other than inspecting
the function's return value.  However, functions declared with the @code{pure}
attribute can safely read any non-volatile objects, and modify the value of
objects in a way that does not affect their return value or the observable
state of the program.

For example,

@smallexample
int hash (char *) __attribute__ ((pure));
@end smallexample

@noindent
tells GCC that subsequent calls to the function @code{hash} with the same
string can be replaced by the result of the first call provided the state
of the program observable by @code{hash}, including the contents of the array
itself, does not change in between.  Even though @code{hash} takes a non-const
pointer argument it must not modify the array it points to, or any other object
whose value the rest of the program may depend on.  However, the caller may
safely change the contents of the array between successive calls to
the function (doing so disables the optimization).  The restriction also
applies to member objects referenced by the @code{this} pointer in C++
non-static member functions.

Some common examples of pure functions are @code{strlen} or @code{memcmp}.
Interesting non-pure functions are functions with infinite loops or those
depending on volatile memory or other system resource, that may change between
consecutive calls (such as the standard C @code{feof} function in
a multithreading environment).

The @code{pure} attribute imposes similar but looser restrictions on
a function's definition than the @code{const} attribute: @code{pure}
allows the function to read any non-volatile memory, even if it changes
in between successive invocations of the function.  Declaring the same
function with both the @code{pure} and the @code{const} attribute is
diagnosed.  Because a pure function cannot have any observable side
effects it does not make sense for such a function to return @code{void}.
Declaring such a function is diagnosed.

@item returns_nonnull
@cindex @code{returns_nonnull} function attribute
The @code{returns_nonnull} attribute specifies that the function
return value should be a non-null pointer.  For instance, the declaration:

@smallexample
extern void *
mymalloc (size_t len) __attribute__((returns_nonnull));
@end smallexample

@noindent
lets the compiler optimize callers based on the knowledge
that the return value will never be null.

@item returns_twice
@cindex @code{returns_twice} function attribute
@cindex functions that return more than once
The @code{returns_twice} attribute tells the compiler that a function may
return more than one time.  The compiler ensures that all registers
are dead before calling such a function and emits a warning about
the variables that may be clobbered after the second return from the
function.  Examples of such functions are @code{setjmp} and @code{vfork}.
The @code{longjmp}-like counterpart of such function, if any, might need
to be marked with the @code{noreturn} attribute.

@item section ("@var{section-name}")
@cindex @code{section} function attribute
@cindex functions in arbitrary sections
Normally, the compiler places the code it generates in the @code{text} section.
Sometimes, however, you need additional sections, or you need certain
particular functions to appear in special sections.  The @code{section}
attribute specifies that a function lives in a particular section.
For example, the declaration:

@smallexample
extern void foobar (void) __attribute__ ((section ("bar")));
@end smallexample

@noindent
puts the function @code{foobar} in the @code{bar} section.

Some file formats do not support arbitrary sections so the @code{section}
attribute is not available on all platforms.
If you need to map the entire contents of a module to a particular
section, consider using the facilities of the linker instead.

@item sentinel
@itemx sentinel (@var{position})
@cindex @code{sentinel} function attribute
This function attribute indicates that an argument in a call to the function
is expected to be an explicit @code{NULL}.  The attribute is only valid on
variadic functions.  By default, the sentinel is expected to be the last
argument of the function call.  If the optional @var{position} argument
is specified to the attribute, the sentinel must be located at
@var{position} counting backwards from the end of the argument list.

@smallexample
__attribute__ ((sentinel))
is equivalent to
__attribute__ ((sentinel(0)))
@end smallexample

The attribute is automatically set with a position of 0 for the built-in
functions @code{execl} and @code{execlp}.  The built-in function
@code{execle} has the attribute set with a position of 1.

A valid @code{NULL} in this context is defined as zero with any object
pointer type.  If your system defines the @code{NULL} macro with
an integer type then you need to add an explicit cast.  During
installation GCC replaces the system @code{<stddef.h>} header with
a copy that redefines NULL appropriately.

The warnings for missing or incorrect sentinels are enabled with
@option{-Wformat}.

@item simd
@itemx simd("@var{mask}")
@cindex @code{simd} function attribute
This attribute enables creation of one or more function versions that
can process multiple arguments using SIMD instructions from a
single invocation.  Specifying this attribute allows compiler to
assume that such versions are available at link time (provided
in the same or another translation unit).  Generated versions are
target-dependent and described in the corresponding Vector ABI document.  For
x86_64 target this document can be found
@w{@uref{https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt,here}}.

The optional argument @var{mask} may have the value
@code{notinbranch} or @code{inbranch},
and instructs the compiler to generate non-masked or masked
clones correspondingly. By default, all clones are generated.

If the attribute is specified and @code{#pragma omp declare simd} is
present on a declaration and the @option{-fopenmp} or @option{-fopenmp-simd}
switch is specified, then the attribute is ignored.

@item stack_protect
@cindex @code{stack_protect} function attribute
This attribute adds stack protection code to the function if 
flags @option{-fstack-protector}, @option{-fstack-protector-strong}
or @option{-fstack-protector-explicit} are set.

@item target (@var{string}, @dots{})
@cindex @code{target} function attribute
Multiple target back ends implement the @code{target} attribute
to specify that a function is to
be compiled with different target options than specified on the
command line.  One or more strings can be provided as arguments.
Each string consists of one or more comma-separated suffixes to
the @code{-m} prefix jointly forming the name of a machine-dependent
option.  @xref{Submodel Options,,Machine-Dependent Options}.

The @code{target} attribute can be used for instance to have a function
compiled with a different ISA (instruction set architecture) than the
default.  @samp{#pragma GCC target} can be used to specify target-specific
options for more than one function.  @xref{Function Specific Option Pragmas},
for details about the pragma.

For instance, on an x86, you could declare one function with the
@code{target("sse4.1,arch=core2")} attribute and another with
@code{target("sse4a,arch=amdfam10")}.  This is equivalent to
compiling the first function with @option{-msse4.1} and
@option{-march=core2} options, and the second function with
@option{-msse4a} and @option{-march=amdfam10} options.  It is up to you
to make sure that a function is only invoked on a machine that
supports the particular ISA it is compiled for (for example by using
@code{cpuid} on x86 to determine what feature bits and architecture
family are used).

@smallexample
int core2_func (void) __attribute__ ((__target__ ("arch=core2")));
int sse3_func (void) __attribute__ ((__target__ ("sse3")));
@end smallexample

Providing multiple strings as arguments separated by commas to specify
multiple options is equivalent to separating the option suffixes with
a comma (@samp{,}) within a single string.  Spaces are not permitted
within the strings.

The options supported are specific to each target; refer to @ref{x86
Function Attributes}, @ref{PowerPC Function Attributes},
@ref{ARM Function Attributes}, @ref{AArch64 Function Attributes},
@ref{Nios II Function Attributes}, and @ref{S/390 Function Attributes}
for details.

@item symver ("@var{name2}@@@var{nodename}")
On ELF targets this attribute creates a symbol version.  The @var{name2} part
of the parameter is the actual name of the symbol by which it will be
externally referenced.  The @code{nodename} portion should be the name of a
node specified in the version script supplied to the linker when building a
shared library.  Versioned symbol must be defined and must be exported with
default visibility.

@smallexample
__attribute__ ((__symver__ ("foo@@VERS_1"))) int
foo_v1 (void)
@{
@}
@end smallexample

Will produce a @code{.symver foo_v1, foo@@VERS_1} directive in the assembler
output. 

It's an error to define multiple version of a given symbol.  In such case
an alias can be used.

@smallexample
__attribute__ ((__symver__ ("foo@@VERS_2")))
__attribute__ ((alias ("foo_v1")))
int symver_foo_v1 (void);
@end smallexample

This example creates an alias of @code{foo_v1} with symbol name
@code{symver_foo_v1} which will be version @code{VERS_2} of @code{foo}.

Finally if the parameter is @code{"@var{name2}@@@@@var{nodename}"} then in
addition to creating a symbol version (as if
@code{"@var{name2}@@@var{nodename}"} was used) the version will be also used
to resolve @var{name2} by the linker.

@item target_clones (@var{options})
@cindex @code{target_clones} function attribute
The @code{target_clones} attribute is used to specify that a function
be cloned into multiple versions compiled with different target options
than specified on the command line.  The supported options and restrictions
are the same as for @code{target} attribute.

For instance, on an x86, you could compile a function with
@code{target_clones("sse4.1,avx")}.  GCC creates two function clones,
one compiled with @option{-msse4.1} and another with @option{-mavx}.

On a PowerPC, you can compile a function with
@code{target_clones("cpu=power9,default")}.  GCC will create two
function clones, one compiled with @option{-mcpu=power9} and another
with the default options.  GCC must be configured to use GLIBC 2.23 or
newer in order to use the @code{target_clones} attribute.

It also creates a resolver function (see
the @code{ifunc} attribute above) that dynamically selects a clone
suitable for current architecture.  The resolver is created only if there
is a usage of a function with @code{target_clones} attribute.

Note that any subsequent call of a function without @code{target_clone}
from a @code{target_clone} caller will not lead to copying
(target clone) of the called function.
If you want to enforce such behaviour,
we recommend declaring the calling function with the @code{flatten} attribute?

@item unused
@cindex @code{unused} function attribute
This attribute, attached to a function, means that the function is meant
to be possibly unused.  GCC does not produce a warning for this
function.

@item used
@cindex @code{used} function attribute
This attribute, attached to a function, means that code must be emitted
for the function even if it appears that the function is not referenced.
This is useful, for example, when the function is referenced only in
inline assembly.

When applied to a member function of a C++ class template, the
attribute also means that the function is instantiated if the
class itself is instantiated.

@item visibility ("@var{visibility_type}")
@cindex @code{visibility} function attribute
This attribute affects the linkage of the declaration to which it is attached.
It can be applied to variables (@pxref{Common Variable Attributes}) and types
(@pxref{Common Type Attributes}) as well as functions.

There are four supported @var{visibility_type} values: default,
hidden, protected or internal visibility.

@smallexample
void __attribute__ ((visibility ("protected")))
f () @{ /* @r{Do something.} */; @}
int i __attribute__ ((visibility ("hidden")));
@end smallexample

The possible values of @var{visibility_type} correspond to the
visibility settings in the ELF gABI.

@table @code
@c keep this list of visibilities in alphabetical order.

@item default
Default visibility is the normal case for the object file format.
This value is available for the visibility attribute to override other
options that may change the assumed visibility of entities.

On ELF, default visibility means that the declaration is visible to other
modules and, in shared libraries, means that the declared entity may be
overridden.

On Darwin, default visibility means that the declaration is visible to
other modules.

Default visibility corresponds to ``external linkage'' in the language.

@item hidden
Hidden visibility indicates that the entity declared has a new
form of linkage, which we call ``hidden linkage''.  Two
declarations of an object with hidden linkage refer to the same object
if they are in the same shared object.

@item internal
Internal visibility is like hidden visibility, but with additional
processor specific semantics.  Unless otherwise specified by the
psABI, GCC defines internal visibility to mean that a function is
@emph{never} called from another module.  Compare this with hidden
functions which, while they cannot be referenced directly by other
modules, can be referenced indirectly via function pointers.  By
indicating that a function cannot be called from outside the module,
GCC may for instance omit the load of a PIC register since it is known
that the calling function loaded the correct value.

@item protected
Protected visibility is like default visibility except that it
indicates that references within the defining module bind to the
definition in that module.  That is, the declared entity cannot be
overridden by another module.

@end table

All visibilities are supported on many, but not all, ELF targets
(supported when the assembler supports the @samp{.visibility}
pseudo-op).  Default visibility is supported everywhere.  Hidden
visibility is supported on Darwin targets.

The visibility attribute should be applied only to declarations that
would otherwise have external linkage.  The attribute should be applied
consistently, so that the same entity should not be declared with
different settings of the attribute.

In C++, the visibility attribute applies to types as well as functions
and objects, because in C++ types have linkage.  A class must not have
greater visibility than its non-static data member types and bases,
and class members default to the visibility of their class.  Also, a
declaration without explicit visibility is limited to the visibility
of its type.

In C++, you can mark member functions and static member variables of a
class with the visibility attribute.  This is useful if you know a
particular method or static member variable should only be used from
one shared object; then you can mark it hidden while the rest of the
class has default visibility.  Care must be taken to avoid breaking
the One Definition Rule; for example, it is usually not useful to mark
an inline method as hidden without marking the whole class as hidden.

A C++ namespace declaration can also have the visibility attribute.

@smallexample
namespace nspace1 __attribute__ ((visibility ("protected")))
@{ /* @r{Do something.} */; @}
@end smallexample

This attribute applies only to the particular namespace body, not to
other definitions of the same namespace; it is equivalent to using
@samp{#pragma GCC visibility} before and after the namespace
definition (@pxref{Visibility Pragmas}).

In C++, if a template argument has limited visibility, this
restriction is implicitly propagated to the template instantiation.
Otherwise, template instantiations and specializations default to the
visibility of their template.

If both the template and enclosing class have explicit visibility, the
visibility from the template is used.

@item warn_unused_result
@cindex @code{warn_unused_result} function attribute
The @code{warn_unused_result} attribute causes a warning to be emitted
if a caller of the function with this attribute does not use its
return value.  This is useful for functions where not checking
the result is either a security problem or always a bug, such as
@code{realloc}.

@smallexample
int fn () __attribute__ ((warn_unused_result));
int foo ()
@{
  if (fn () < 0) return -1;
  fn ();
  return 0;
@}
@end smallexample

@noindent
results in warning on line 5.

@item weak
@cindex @code{weak} function attribute
The @code{weak} attribute causes the declaration to be emitted as a weak
symbol rather than a global.  This is primarily useful in defining
library functions that can be overridden in user code, though it can
also be used with non-function declarations.  Weak symbols are supported
for ELF targets, and also for a.out targets when using the GNU assembler
and linker.

@item weakref
@itemx weakref ("@var{target}")
@cindex @code{weakref} function attribute
The @code{weakref} attribute marks a declaration as a weak reference.
Without arguments, it should be accompanied by an @code{alias} attribute
naming the target symbol.  Optionally, the @var{target} may be given as
an argument to @code{weakref} itself.  In either case, @code{weakref}
implicitly marks the declaration as @code{weak}.  Without a
@var{target}, given as an argument to @code{weakref} or to @code{alias},
@code{weakref} is equivalent to @code{weak}.

@smallexample
static int x() __attribute__ ((weakref ("y")));
/* is equivalent to... */
static int x() __attribute__ ((weak, weakref, alias ("y")));
/* and to... */
static int x() __attribute__ ((weakref));
static int x() __attribute__ ((alias ("y")));
@end smallexample

A weak reference is an alias that does not by itself require a
definition to be given for the target symbol.  If the target symbol is
only referenced through weak references, then it becomes a @code{weak}
undefined symbol.  If it is directly referenced, however, then such
strong references prevail, and a definition is required for the
symbol, not necessarily in the same translation unit.

The effect is equivalent to moving all references to the alias to a
separate translation unit, renaming the alias to the aliased symbol,
declaring it as weak, compiling the two separate translation units and
performing a link with relocatable output (ie: @code{ld -r}) on them.

At present, a declaration to which @code{weakref} is attached can
only be @code{static}.

@end table

@c This is the end of the target-independent attribute table

@node AArch64 Function Attributes
@subsection AArch64 Function Attributes

The following target-specific function attributes are available for the
AArch64 target.  For the most part, these options mirror the behavior of
similar command-line options (@pxref{AArch64 Options}), but on a
per-function basis.

@table @code
@item general-regs-only
@cindex @code{general-regs-only} function attribute, AArch64
Indicates that no floating-point or Advanced SIMD registers should be
used when generating code for this function.  If the function explicitly
uses floating-point code, then the compiler gives an error.  This is
the same behavior as that of the command-line option
@option{-mgeneral-regs-only}.

@item fix-cortex-a53-835769
@cindex @code{fix-cortex-a53-835769} function attribute, AArch64
Indicates that the workaround for the Cortex-A53 erratum 835769 should be
applied to this function.  To explicitly disable the workaround for this
function specify the negated form: @code{no-fix-cortex-a53-835769}.
This corresponds to the behavior of the command line options
@option{-mfix-cortex-a53-835769} and @option{-mno-fix-cortex-a53-835769}.

@item cmodel=
@cindex @code{cmodel=} function attribute, AArch64
Indicates that code should be generated for a particular code model for
this function.  The behavior and permissible arguments are the same as
for the command line option @option{-mcmodel=}.

@item strict-align
@itemx no-strict-align
@cindex @code{strict-align} function attribute, AArch64
@code{strict-align} indicates that the compiler should not assume that unaligned
memory references are handled by the system.  To allow the compiler to assume
that aligned memory references are handled by the system, the inverse attribute
@code{no-strict-align} can be specified.  The behavior is same as for the
command-line option @option{-mstrict-align} and @option{-mno-strict-align}.

@item omit-leaf-frame-pointer
@cindex @code{omit-leaf-frame-pointer} function attribute, AArch64
Indicates that the frame pointer should be omitted for a leaf function call.
To keep the frame pointer, the inverse attribute
@code{no-omit-leaf-frame-pointer} can be specified.  These attributes have
the same behavior as the command-line options @option{-momit-leaf-frame-pointer}
and @option{-mno-omit-leaf-frame-pointer}.

@item tls-dialect=
@cindex @code{tls-dialect=} function attribute, AArch64
Specifies the TLS dialect to use for this function.  The behavior and
permissible arguments are the same as for the command-line option
@option{-mtls-dialect=}.

@item arch=
@cindex @code{arch=} function attribute, AArch64
Specifies the architecture version and architectural extensions to use
for this function.  The behavior and permissible arguments are the same as
for the @option{-march=} command-line option.

@item tune=
@cindex @code{tune=} function attribute, AArch64
Specifies the core for which to tune the performance of this function.
The behavior and permissible arguments are the same as for the @option{-mtune=}
command-line option.

@item cpu=
@cindex @code{cpu=} function attribute, AArch64
Specifies the core for which to tune the performance of this function and also
whose architectural features to use.  The behavior and valid arguments are the
same as for the @option{-mcpu=} command-line option.

@item sign-return-address
@cindex @code{sign-return-address} function attribute, AArch64
Select the function scope on which return address signing will be applied.  The
behavior and permissible arguments are the same as for the command-line option
@option{-msign-return-address=}.  The default value is @code{none}.  This
attribute is deprecated.  The @code{branch-protection} attribute should
be used instead.

@item branch-protection
@cindex @code{branch-protection} function attribute, AArch64
Select the function scope on which branch protection will be applied.  The
behavior and permissible arguments are the same as for the command-line option
@option{-mbranch-protection=}.  The default value is @code{none}.

@end table

The above target attributes can be specified as follows:

@smallexample
__attribute__((target("@var{attr-string}")))
int
f (int a)
@{
  return a + 5;
@}
@end smallexample

where @code{@var{attr-string}} is one of the attribute strings specified above.

Additionally, the architectural extension string may be specified on its
own.  This can be used to turn on and off particular architectural extensions
without having to specify a particular architecture version or core.  Example:

@smallexample
__attribute__((target("+crc+nocrypto")))
int
foo (int a)
@{
  return a + 5;
@}
@end smallexample

In this example @code{target("+crc+nocrypto")} enables the @code{crc}
extension and disables the @code{crypto} extension for the function @code{foo}
without modifying an existing @option{-march=} or @option{-mcpu} option.

Multiple target function attributes can be specified by separating them with
a comma.  For example:
@smallexample
__attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53")))
int
foo (int a)
@{
  return a + 5;
@}
@end smallexample

is valid and compiles function @code{foo} for ARMv8-A with @code{crc}
and @code{crypto} extensions and tunes it for @code{cortex-a53}.

@subsubsection Inlining rules
Specifying target attributes on individual functions or performing link-time
optimization across translation units compiled with different target options
can affect function inlining rules:

In particular, a caller function can inline a callee function only if the
architectural features available to the callee are a subset of the features
available to the caller.
For example: A function @code{foo} compiled with @option{-march=armv8-a+crc},
or tagged with the equivalent @code{arch=armv8-a+crc} attribute,
can inline a function @code{bar} compiled with @option{-march=armv8-a+nocrc}
because the all the architectural features that function @code{bar} requires
are available to function @code{foo}.  Conversely, function @code{bar} cannot
inline function @code{foo}.

Additionally inlining a function compiled with @option{-mstrict-align} into a
function compiled without @code{-mstrict-align} is not allowed.
However, inlining a function compiled without @option{-mstrict-align} into a
function compiled with @option{-mstrict-align} is allowed.

Note that CPU tuning options and attributes such as the @option{-mcpu=},
@option{-mtune=} do not inhibit inlining unless the CPU specified by the
@option{-mcpu=} option or the @code{cpu=} attribute conflicts with the
architectural feature rules specified above.

@node AMD GCN Function Attributes
@subsection AMD GCN Function Attributes

These function attributes are supported by the AMD GCN back end:

@table @code
@item amdgpu_hsa_kernel
@cindex @code{amdgpu_hsa_kernel} function attribute, AMD GCN
This attribute indicates that the corresponding function should be compiled as
a kernel function, that is an entry point that can be invoked from the host
via the HSA runtime library.  By default functions are only callable only from
other GCN functions.

This attribute is implicitly applied to any function named @code{main}, using
default parameters.

Kernel functions may return an integer value, which will be written to a
conventional place within the HSA "kernargs" region.

The attribute parameters configure what values are passed into the kernel
function by the GPU drivers, via the initial register state.  Some values are
used by the compiler, and therefore forced on.  Enabling other options may
break assumptions in the compiler and/or run-time libraries.

@table @code
@item private_segment_buffer
Set @code{enable_sgpr_private_segment_buffer} flag.  Always on (required to
locate the stack).

@item dispatch_ptr
Set @code{enable_sgpr_dispatch_ptr} flag.  Always on (required to locate the
launch dimensions).

@item queue_ptr
Set @code{enable_sgpr_queue_ptr} flag.  Always on (required to convert address
spaces).

@item kernarg_segment_ptr
Set @code{enable_sgpr_kernarg_segment_ptr} flag.  Always on (required to
locate the kernel arguments, "kernargs").

@item dispatch_id
Set @code{enable_sgpr_dispatch_id} flag.

@item flat_scratch_init
Set @code{enable_sgpr_flat_scratch_init} flag.

@item private_segment_size
Set @code{enable_sgpr_private_segment_size} flag.

@item grid_workgroup_count_X
Set @code{enable_sgpr_grid_workgroup_count_x} flag.  Always on (required to
use OpenACC/OpenMP).

@item grid_workgroup_count_Y
Set @code{enable_sgpr_grid_workgroup_count_y} flag.

@item grid_workgroup_count_Z
Set @code{enable_sgpr_grid_workgroup_count_z} flag.

@item workgroup_id_X
Set @code{enable_sgpr_workgroup_id_x} flag.

@item workgroup_id_Y
Set @code{enable_sgpr_workgroup_id_y} flag.

@item workgroup_id_Z
Set @code{enable_sgpr_workgroup_id_z} flag.

@item workgroup_info
Set @code{enable_sgpr_workgroup_info} flag.

@item private_segment_wave_offset
Set @code{enable_sgpr_private_segment_wave_byte_offset} flag.  Always on
(required to locate the stack).

@item work_item_id_X
Set @code{enable_vgpr_workitem_id} parameter.  Always on (can't be disabled).

@item work_item_id_Y
Set @code{enable_vgpr_workitem_id} parameter.  Always on (required to enable
vectorization.)

@item work_item_id_Z
Set @code{enable_vgpr_workitem_id} parameter.  Always on (required to use
OpenACC/OpenMP).

@end table
@end table

@node ARC Function Attributes
@subsection ARC Function Attributes

These function attributes are supported by the ARC back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, ARC
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

On the ARC, you must specify the kind of interrupt to be handled
in a parameter to the interrupt attribute like this:

@smallexample
void f () __attribute__ ((interrupt ("ilink1")));
@end smallexample

Permissible values for this parameter are: @w{@code{ilink1}} and
@w{@code{ilink2}} for ARCv1 architecture, and @w{@code{ilink}} and
@w{@code{firq}} for ARCv2 architecture.

@item long_call
@itemx medium_call
@itemx short_call
@cindex @code{long_call} function attribute, ARC
@cindex @code{medium_call} function attribute, ARC
@cindex @code{short_call} function attribute, ARC
@cindex indirect calls, ARC
These attributes specify how a particular function is called.
These attributes override the
@option{-mlong-calls} and @option{-mmedium-calls} (@pxref{ARC Options})
command-line switches and @code{#pragma long_calls} settings.

For ARC, a function marked with the @code{long_call} attribute is
always called using register-indirect jump-and-link instructions,
thereby enabling the called function to be placed anywhere within the
32-bit address space.  A function marked with the @code{medium_call}
attribute will always be close enough to be called with an unconditional
branch-and-link instruction, which has a 25-bit offset from
the call site.  A function marked with the @code{short_call}
attribute will always be close enough to be called with a conditional
branch-and-link instruction, which has a 21-bit offset from
the call site.

@item jli_always
@cindex @code{jli_always} function attribute, ARC
Forces a particular function to be called using @code{jli}
instruction.  The @code{jli} instruction makes use of a table stored
into @code{.jlitab} section, which holds the location of the functions
which are addressed using this instruction.

@item jli_fixed
@cindex @code{jli_fixed} function attribute, ARC
Identical like the above one, but the location of the function in the
@code{jli} table is known and given as an attribute parameter.

@item secure_call
@cindex @code{secure_call} function attribute, ARC
This attribute allows one to mark secure-code functions that are
callable from normal mode.  The location of the secure call function
into the @code{sjli} table needs to be passed as argument.

@item naked
@cindex @code{naked} function attribute, ARC
This attribute allows the compiler to construct the requisite function
declaration, while allowing the body of the function to be assembly
code.  The specified function will not have prologue/epilogue
sequences generated by the compiler.  Only basic @code{asm} statements
can safely be included in naked functions (@pxref{Basic Asm}).  While
using extended @code{asm} or a mixture of basic @code{asm} and C code
may appear to work, they cannot be depended upon to work reliably and
are not supported.

@end table

@node ARM Function Attributes
@subsection ARM Function Attributes

These function attributes are supported for ARM targets:

@table @code

@item general-regs-only
@cindex @code{general-regs-only} function attribute, ARM
Indicates that no floating-point or Advanced SIMD registers should be
used when generating code for this function.  If the function explicitly
uses floating-point code, then the compiler gives an error.  This is
the same behavior as that of the command-line option
@option{-mgeneral-regs-only}.

@item interrupt
@cindex @code{interrupt} function attribute, ARM
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

You can specify the kind of interrupt to be handled by
adding an optional parameter to the interrupt attribute like this:

@smallexample
void f () __attribute__ ((interrupt ("IRQ")));
@end smallexample

@noindent
Permissible values for this parameter are: @code{IRQ}, @code{FIQ},
@code{SWI}, @code{ABORT} and @code{UNDEF}.

On ARMv7-M the interrupt type is ignored, and the attribute means the function
may be called with a word-aligned stack pointer.

@item isr
@cindex @code{isr} function attribute, ARM
Use this attribute on ARM to write Interrupt Service Routines. This is an
alias to the @code{interrupt} attribute above.

@item long_call
@itemx short_call
@cindex @code{long_call} function attribute, ARM
@cindex @code{short_call} function attribute, ARM
@cindex indirect calls, ARM
These attributes specify how a particular function is called.
These attributes override the
@option{-mlong-calls} (@pxref{ARM Options})
command-line switch and @code{#pragma long_calls} settings.  For ARM, the
@code{long_call} attribute indicates that the function might be far
away from the call site and require a different (more expensive)
calling sequence.   The @code{short_call} attribute always places
the offset to the function from the call site into the @samp{BL}
instruction directly.

@item naked
@cindex @code{naked} function attribute, ARM
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item pcs
@cindex @code{pcs} function attribute, ARM

The @code{pcs} attribute can be used to control the calling convention
used for a function on ARM.  The attribute takes an argument that specifies
the calling convention to use.

When compiling using the AAPCS ABI (or a variant of it) then valid
values for the argument are @code{"aapcs"} and @code{"aapcs-vfp"}.  In
order to use a variant other than @code{"aapcs"} then the compiler must
be permitted to use the appropriate co-processor registers (i.e., the
VFP registers must be available in order to use @code{"aapcs-vfp"}).
For example,

@smallexample
/* Argument passed in r0, and result returned in r0+r1.  */
double f2d (float) __attribute__((pcs("aapcs")));
@end smallexample

Variadic functions always use the @code{"aapcs"} calling convention and
the compiler rejects attempts to specify an alternative.

@item target (@var{options})
@cindex @code{target} function attribute
As discussed in @ref{Common Function Attributes}, this attribute 
allows specification of target-specific compilation options.

On ARM, the following options are allowed:

@table @samp
@item thumb
@cindex @code{target("thumb")} function attribute, ARM
Force code generation in the Thumb (T16/T32) ISA, depending on the
architecture level.

@item arm
@cindex @code{target("arm")} function attribute, ARM
Force code generation in the ARM (A32) ISA.

Functions from different modes can be inlined in the caller's mode.

@item fpu=
@cindex @code{target("fpu=")} function attribute, ARM
Specifies the fpu for which to tune the performance of this function.
The behavior and permissible arguments are the same as for the @option{-mfpu=}
command-line option.

@item arch=
@cindex @code{arch=} function attribute, ARM
Specifies the architecture version and architectural extensions to use
for this function.  The behavior and permissible arguments are the same as
for the @option{-march=} command-line option.

The above target attributes can be specified as follows:

@smallexample
__attribute__((target("arch=armv8-a+crc")))
int
f (int a)
@{
  return a + 5;
@}
@end smallexample

Additionally, the architectural extension string may be specified on its
own.  This can be used to turn on and off particular architectural extensions
without having to specify a particular architecture version or core.  Example:

@smallexample
__attribute__((target("+crc+nocrypto")))
int
foo (int a)
@{
  return a + 5;
@}
@end smallexample

In this example @code{target("+crc+nocrypto")} enables the @code{crc}
extension and disables the @code{crypto} extension for the function @code{foo}
without modifying an existing @option{-march=} or @option{-mcpu} option.

@end table

@end table

@node AVR Function Attributes
@subsection AVR Function Attributes

These function attributes are supported by the AVR back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, AVR
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

On the AVR, the hardware globally disables interrupts when an
interrupt is executed.  The first instruction of an interrupt handler
declared with this attribute is a @code{SEI} instruction to
re-enable interrupts.  See also the @code{signal} function attribute
that does not insert a @code{SEI} instruction.  If both @code{signal} and
@code{interrupt} are specified for the same function, @code{signal}
is silently ignored.

@item naked
@cindex @code{naked} function attribute, AVR
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item no_gccisr
@cindex @code{no_gccisr} function attribute, AVR
Do not use @code{__gcc_isr} pseudo instructions in a function with
the @code{interrupt} or @code{signal} attribute aka. interrupt
service routine (ISR).
Use this attribute if the preamble of the ISR prologue should always read
@example
push  __zero_reg__
push  __tmp_reg__
in    __tmp_reg__, __SREG__
push  __tmp_reg__
clr   __zero_reg__
@end example
and accordingly for the postamble of the epilogue --- no matter whether
the mentioned registers are actually used in the ISR or not.
Situations where you might want to use this attribute include:
@itemize @bullet
@item
Code that (effectively) clobbers bits of @code{SREG} other than the
@code{I}-flag by writing to the memory location of @code{SREG}.
@item
Code that uses inline assembler to jump to a different function which
expects (parts of) the prologue code as outlined above to be present.
@end itemize
To disable @code{__gcc_isr} generation for the whole compilation unit,
there is option @option{-mno-gas-isr-prologues}, @pxref{AVR Options}.

@item OS_main
@itemx OS_task
@cindex @code{OS_main} function attribute, AVR
@cindex @code{OS_task} function attribute, AVR
On AVR, functions with the @code{OS_main} or @code{OS_task} attribute
do not save/restore any call-saved register in their prologue/epilogue.

The @code{OS_main} attribute can be used when there @emph{is
guarantee} that interrupts are disabled at the time when the function
is entered.  This saves resources when the stack pointer has to be
changed to set up a frame for local variables.

The @code{OS_task} attribute can be used when there is @emph{no
guarantee} that interrupts are disabled at that time when the function
is entered like for, e@.g@. task functions in a multi-threading operating
system. In that case, changing the stack pointer register is
guarded by save/clear/restore of the global interrupt enable flag.

The differences to the @code{naked} function attribute are:
@itemize @bullet
@item @code{naked} functions do not have a return instruction whereas 
@code{OS_main} and @code{OS_task} functions have a @code{RET} or
@code{RETI} return instruction.
@item @code{naked} functions do not set up a frame for local variables
or a frame pointer whereas @code{OS_main} and @code{OS_task} do this
as needed.
@end itemize

@item signal
@cindex @code{signal} function attribute, AVR
Use this attribute on the AVR to indicate that the specified
function is an interrupt handler.  The compiler generates function
entry and exit sequences suitable for use in an interrupt handler when this
attribute is present.

See also the @code{interrupt} function attribute. 

The AVR hardware globally disables interrupts when an interrupt is executed.
Interrupt handler functions defined with the @code{signal} attribute
do not re-enable interrupts.  It is save to enable interrupts in a
@code{signal} handler.  This ``save'' only applies to the code
generated by the compiler and not to the IRQ layout of the
application which is responsibility of the application.

If both @code{signal} and @code{interrupt} are specified for the same
function, @code{signal} is silently ignored.
@end table

@node Blackfin Function Attributes
@subsection Blackfin Function Attributes

These function attributes are supported by the Blackfin back end:

@table @code

@item exception_handler
@cindex @code{exception_handler} function attribute
@cindex exception handler functions, Blackfin
Use this attribute on the Blackfin to indicate that the specified function
is an exception handler.  The compiler generates function entry and
exit sequences suitable for use in an exception handler when this
attribute is present.

@item interrupt_handler
@cindex @code{interrupt_handler} function attribute, Blackfin
Use this attribute to
indicate that the specified function is an interrupt handler.  The compiler
generates function entry and exit sequences suitable for use in an
interrupt handler when this attribute is present.

@item kspisusp
@cindex @code{kspisusp} function attribute, Blackfin
@cindex User stack pointer in interrupts on the Blackfin
When used together with @code{interrupt_handler}, @code{exception_handler}
or @code{nmi_handler}, code is generated to load the stack pointer
from the USP register in the function prologue.

@item l1_text
@cindex @code{l1_text} function attribute, Blackfin
This attribute specifies a function to be placed into L1 Instruction
SRAM@. The function is put into a specific section named @code{.l1.text}.
With @option{-mfdpic}, function calls with a such function as the callee
or caller uses inlined PLT.

@item l2
@cindex @code{l2} function attribute, Blackfin
This attribute specifies a function to be placed into L2
SRAM. The function is put into a specific section named
@code{.l2.text}. With @option{-mfdpic}, callers of such functions use
an inlined PLT.

@item longcall
@itemx shortcall
@cindex indirect calls, Blackfin
@cindex @code{longcall} function attribute, Blackfin
@cindex @code{shortcall} function attribute, Blackfin
The @code{longcall} attribute
indicates that the function might be far away from the call site and
require a different (more expensive) calling sequence.  The
@code{shortcall} attribute indicates that the function is always close
enough for the shorter calling sequence to be used.  These attributes
override the @option{-mlongcall} switch.

@item nesting
@cindex @code{nesting} function attribute, Blackfin
@cindex Allow nesting in an interrupt handler on the Blackfin processor
Use this attribute together with @code{interrupt_handler},
@code{exception_handler} or @code{nmi_handler} to indicate that the function
entry code should enable nested interrupts or exceptions.

@item nmi_handler
@cindex @code{nmi_handler} function attribute, Blackfin
@cindex NMI handler functions on the Blackfin processor
Use this attribute on the Blackfin to indicate that the specified function
is an NMI handler.  The compiler generates function entry and
exit sequences suitable for use in an NMI handler when this
attribute is present.

@item saveall
@cindex @code{saveall} function attribute, Blackfin
@cindex save all registers on the Blackfin
Use this attribute to indicate that
all registers except the stack pointer should be saved in the prologue
regardless of whether they are used or not.
@end table

@node CR16 Function Attributes
@subsection CR16 Function Attributes

These function attributes are supported by the CR16 back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, CR16
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.
@end table

@node C-SKY Function Attributes
@subsection C-SKY Function Attributes

These function attributes are supported by the C-SKY back end:

@table @code
@item interrupt
@itemx isr
@cindex @code{interrupt} function attribute, C-SKY
@cindex @code{isr} function attribute, C-SKY
Use these attributes to indicate that the specified function
is an interrupt handler.
The compiler generates function entry and exit sequences suitable for
use in an interrupt handler when either of these attributes are present.

Use of these options requires the @option{-mistack} command-line option
to enable support for the necessary interrupt stack instructions.  They
are ignored with a warning otherwise.  @xref{C-SKY Options}.

@item naked
@cindex @code{naked} function attribute, C-SKY
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.
@end table


@node Epiphany Function Attributes
@subsection Epiphany Function Attributes

These function attributes are supported by the Epiphany back end:

@table @code
@item disinterrupt
@cindex @code{disinterrupt} function attribute, Epiphany
This attribute causes the compiler to emit
instructions to disable interrupts for the duration of the given
function.

@item forwarder_section
@cindex @code{forwarder_section} function attribute, Epiphany
This attribute modifies the behavior of an interrupt handler.
The interrupt handler may be in external memory which cannot be
reached by a branch instruction, so generate a local memory trampoline
to transfer control.  The single parameter identifies the section where
the trampoline is placed.

@item interrupt
@cindex @code{interrupt} function attribute, Epiphany
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.  It may also generate
a special section with code to initialize the interrupt vector table.

On Epiphany targets one or more optional parameters can be added like this:

@smallexample
void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler ();
@end smallexample

Permissible values for these parameters are: @w{@code{reset}},
@w{@code{software_exception}}, @w{@code{page_miss}},
@w{@code{timer0}}, @w{@code{timer1}}, @w{@code{message}},
@w{@code{dma0}}, @w{@code{dma1}}, @w{@code{wand}} and @w{@code{swi}}.
Multiple parameters indicate that multiple entries in the interrupt
vector table should be initialized for this function, i.e.@: for each
parameter @w{@var{name}}, a jump to the function is emitted in
the section @w{ivt_entry_@var{name}}.  The parameter(s) may be omitted
entirely, in which case no interrupt vector table entry is provided.

Note that interrupts are enabled inside the function
unless the @code{disinterrupt} attribute is also specified.

The following examples are all valid uses of these attributes on
Epiphany targets:
@smallexample
void __attribute__ ((interrupt)) universal_handler ();
void __attribute__ ((interrupt ("dma1"))) dma1_handler ();
void __attribute__ ((interrupt ("dma0, dma1"))) 
  universal_dma_handler ();
void __attribute__ ((interrupt ("timer0"), disinterrupt))
  fast_timer_handler ();
void __attribute__ ((interrupt ("dma0, dma1"), 
                     forwarder_section ("tramp")))
  external_dma_handler ();
@end smallexample

@item long_call
@itemx short_call
@cindex @code{long_call} function attribute, Epiphany
@cindex @code{short_call} function attribute, Epiphany
@cindex indirect calls, Epiphany
These attributes specify how a particular function is called.
These attributes override the
@option{-mlong-calls} (@pxref{Adapteva Epiphany Options})
command-line switch and @code{#pragma long_calls} settings.
@end table


@node H8/300 Function Attributes
@subsection H8/300 Function Attributes

These function attributes are available for H8/300 targets:

@table @code
@item function_vector
@cindex @code{function_vector} function attribute, H8/300
Use this attribute on the H8/300, H8/300H, and H8S to indicate 
that the specified function should be called through the function vector.
Calling a function through the function vector reduces code size; however,
the function vector has a limited size (maximum 128 entries on the H8/300
and 64 entries on the H8/300H and H8S)
and shares space with the interrupt vector.

@item interrupt_handler
@cindex @code{interrupt_handler} function attribute, H8/300
Use this attribute on the H8/300, H8/300H, and H8S to
indicate that the specified function is an interrupt handler.  The compiler
generates function entry and exit sequences suitable for use in an
interrupt handler when this attribute is present.

@item saveall
@cindex @code{saveall} function attribute, H8/300
@cindex save all registers on the H8/300, H8/300H, and H8S
Use this attribute on the H8/300, H8/300H, and H8S to indicate that
all registers except the stack pointer should be saved in the prologue
regardless of whether they are used or not.
@end table

@node IA-64 Function Attributes
@subsection IA-64 Function Attributes

These function attributes are supported on IA-64 targets:

@table @code
@item syscall_linkage
@cindex @code{syscall_linkage} function attribute, IA-64
This attribute is used to modify the IA-64 calling convention by marking
all input registers as live at all function exits.  This makes it possible
to restart a system call after an interrupt without having to save/restore
the input registers.  This also prevents kernel data from leaking into
application code.

@item version_id
@cindex @code{version_id} function attribute, IA-64
This IA-64 HP-UX attribute, attached to a global variable or function, renames a
symbol to contain a version string, thus allowing for function level
versioning.  HP-UX system header files may use function level versioning
for some system calls.

@smallexample
extern int foo () __attribute__((version_id ("20040821")));
@end smallexample

@noindent
Calls to @code{foo} are mapped to calls to @code{foo@{20040821@}}.
@end table

@node M32C Function Attributes
@subsection M32C Function Attributes

These function attributes are supported by the M32C back end:

@table @code
@item bank_switch
@cindex @code{bank_switch} function attribute, M32C
When added to an interrupt handler with the M32C port, causes the
prologue and epilogue to use bank switching to preserve the registers
rather than saving them on the stack.

@item fast_interrupt
@cindex @code{fast_interrupt} function attribute, M32C
Use this attribute on the M32C port to indicate that the specified
function is a fast interrupt handler.  This is just like the
@code{interrupt} attribute, except that @code{freit} is used to return
instead of @code{reit}.

@item function_vector
@cindex @code{function_vector} function attribute, M16C/M32C
On M16C/M32C targets, the @code{function_vector} attribute declares a
special page subroutine call function. Use of this attribute reduces
the code size by 2 bytes for each call generated to the
subroutine. The argument to the attribute is the vector number entry
from the special page vector table which contains the 16 low-order
bits of the subroutine's entry address. Each vector table has special
page number (18 to 255) that is used in @code{jsrs} instructions.
Jump addresses of the routines are generated by adding 0x0F0000 (in
case of M16C targets) or 0xFF0000 (in case of M32C targets), to the
2-byte addresses set in the vector table. Therefore you need to ensure
that all the special page vector routines should get mapped within the
address range 0x0F0000 to 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF
(for M32C).

In the following example 2 bytes are saved for each call to
function @code{foo}.

@smallexample
void foo (void) __attribute__((function_vector(0x18)));
void foo (void)
@{
@}

void bar (void)
@{
    foo();
@}
@end smallexample

If functions are defined in one file and are called in another file,
then be sure to write this declaration in both files.

This attribute is ignored for R8C target.

@item interrupt
@cindex @code{interrupt} function attribute, M32C
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.
@end table

@node M32R/D Function Attributes
@subsection M32R/D Function Attributes

These function attributes are supported by the M32R/D back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, M32R/D
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

@item model (@var{model-name})
@cindex @code{model} function attribute, M32R/D
@cindex function addressability on the M32R/D

On the M32R/D, use this attribute to set the addressability of an
object, and of the code generated for a function.  The identifier
@var{model-name} is one of @code{small}, @code{medium}, or
@code{large}, representing each of the code models.

Small model objects live in the lower 16MB of memory (so that their
addresses can be loaded with the @code{ld24} instruction), and are
callable with the @code{bl} instruction.

Medium model objects may live anywhere in the 32-bit address space (the
compiler generates @code{seth/add3} instructions to load their addresses),
and are callable with the @code{bl} instruction.

Large model objects may live anywhere in the 32-bit address space (the
compiler generates @code{seth/add3} instructions to load their addresses),
and may not be reachable with the @code{bl} instruction (the compiler
generates the much slower @code{seth/add3/jl} instruction sequence).
@end table

@node m68k Function Attributes
@subsection m68k Function Attributes

These function attributes are supported by the m68k back end:

@table @code
@item interrupt
@itemx interrupt_handler
@cindex @code{interrupt} function attribute, m68k
@cindex @code{interrupt_handler} function attribute, m68k
Use this attribute to
indicate that the specified function is an interrupt handler.  The compiler
generates function entry and exit sequences suitable for use in an
interrupt handler when this attribute is present.  Either name may be used.

@item interrupt_thread
@cindex @code{interrupt_thread} function attribute, fido
Use this attribute on fido, a subarchitecture of the m68k, to indicate
that the specified function is an interrupt handler that is designed
to run as a thread.  The compiler omits generate prologue/epilogue
sequences and replaces the return instruction with a @code{sleep}
instruction.  This attribute is available only on fido.
@end table

@node MCORE Function Attributes
@subsection MCORE Function Attributes

These function attributes are supported by the MCORE back end:

@table @code
@item naked
@cindex @code{naked} function attribute, MCORE
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.
@end table

@node MeP Function Attributes
@subsection MeP Function Attributes

These function attributes are supported by the MeP back end:

@table @code
@item disinterrupt
@cindex @code{disinterrupt} function attribute, MeP
On MeP targets, this attribute causes the compiler to emit
instructions to disable interrupts for the duration of the given
function.

@item interrupt
@cindex @code{interrupt} function attribute, MeP
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

@item near
@cindex @code{near} function attribute, MeP
This attribute causes the compiler to assume the called
function is close enough to use the normal calling convention,
overriding the @option{-mtf} command-line option.

@item far
@cindex @code{far} function attribute, MeP
On MeP targets this causes the compiler to use a calling convention
that assumes the called function is too far away for the built-in
addressing modes.

@item vliw
@cindex @code{vliw} function attribute, MeP
The @code{vliw} attribute tells the compiler to emit
instructions in VLIW mode instead of core mode.  Note that this
attribute is not allowed unless a VLIW coprocessor has been configured
and enabled through command-line options.
@end table

@node MicroBlaze Function Attributes
@subsection MicroBlaze Function Attributes

These function attributes are supported on MicroBlaze targets:

@table @code
@item save_volatiles
@cindex @code{save_volatiles} function attribute, MicroBlaze
Use this attribute to indicate that the function is
an interrupt handler.  All volatile registers (in addition to non-volatile
registers) are saved in the function prologue.  If the function is a leaf
function, only volatiles used by the function are saved.  A normal function
return is generated instead of a return from interrupt.

@item break_handler
@cindex @code{break_handler} function attribute, MicroBlaze
@cindex break handler functions
Use this attribute to indicate that
the specified function is a break handler.  The compiler generates function
entry and exit sequences suitable for use in an break handler when this
attribute is present. The return from @code{break_handler} is done through
the @code{rtbd} instead of @code{rtsd}.

@smallexample
void f () __attribute__ ((break_handler));
@end smallexample

@item interrupt_handler
@itemx fast_interrupt 
@cindex @code{interrupt_handler} function attribute, MicroBlaze
@cindex @code{fast_interrupt} function attribute, MicroBlaze
These attributes indicate that the specified function is an interrupt
handler.  Use the @code{fast_interrupt} attribute to indicate handlers
used in low-latency interrupt mode, and @code{interrupt_handler} for
interrupts that do not use low-latency handlers.  In both cases, GCC
emits appropriate prologue code and generates a return from the handler
using @code{rtid} instead of @code{rtsd}.
@end table

@node Microsoft Windows Function Attributes
@subsection Microsoft Windows Function Attributes

The following attributes are available on Microsoft Windows and Symbian OS
targets.

@table @code
@item dllexport
@cindex @code{dllexport} function attribute
@cindex @code{__declspec(dllexport)}
On Microsoft Windows targets and Symbian OS targets the
@code{dllexport} attribute causes the compiler to provide a global
pointer to a pointer in a DLL, so that it can be referenced with the
@code{dllimport} attribute.  On Microsoft Windows targets, the pointer
name is formed by combining @code{_imp__} and the function or variable
name.

You can use @code{__declspec(dllexport)} as a synonym for
@code{__attribute__ ((dllexport))} for compatibility with other
compilers.

On systems that support the @code{visibility} attribute, this
attribute also implies ``default'' visibility.  It is an error to
explicitly specify any other visibility.

GCC's default behavior is to emit all inline functions with the
@code{dllexport} attribute.  Since this can cause object file-size bloat,
you can use @option{-fno-keep-inline-dllexport}, which tells GCC to
ignore the attribute for inlined functions unless the 
@option{-fkeep-inline-functions} flag is used instead.

The attribute is ignored for undefined symbols.

When applied to C++ classes, the attribute marks defined non-inlined
member functions and static data members as exports.  Static consts
initialized in-class are not marked unless they are also defined
out-of-class.

For Microsoft Windows targets there are alternative methods for
including the symbol in the DLL's export table such as using a
@file{.def} file with an @code{EXPORTS} section or, with GNU ld, using
the @option{--export-all} linker flag.

@item dllimport
@cindex @code{dllimport} function attribute
@cindex @code{__declspec(dllimport)}
On Microsoft Windows and Symbian OS targets, the @code{dllimport}
attribute causes the compiler to reference a function or variable via
a global pointer to a pointer that is set up by the DLL exporting the
symbol.  The attribute implies @code{extern}.  On Microsoft Windows
targets, the pointer name is formed by combining @code{_imp__} and the
function or variable name.

You can use @code{__declspec(dllimport)} as a synonym for
@code{__attribute__ ((dllimport))} for compatibility with other
compilers.

On systems that support the @code{visibility} attribute, this
attribute also implies ``default'' visibility.  It is an error to
explicitly specify any other visibility.

Currently, the attribute is ignored for inlined functions.  If the
attribute is applied to a symbol @emph{definition}, an error is reported.
If a symbol previously declared @code{dllimport} is later defined, the
attribute is ignored in subsequent references, and a warning is emitted.
The attribute is also overridden by a subsequent declaration as
@code{dllexport}.

When applied to C++ classes, the attribute marks non-inlined
member functions and static data members as imports.  However, the
attribute is ignored for virtual methods to allow creation of vtables
using thunks.

On the SH Symbian OS target the @code{dllimport} attribute also has
another affect---it can cause the vtable and run-time type information
for a class to be exported.  This happens when the class has a
dllimported constructor or a non-inline, non-pure virtual function
and, for either of those two conditions, the class also has an inline
constructor or destructor and has a key function that is defined in
the current translation unit.

For Microsoft Windows targets the use of the @code{dllimport}
attribute on functions is not necessary, but provides a small
performance benefit by eliminating a thunk in the DLL@.  The use of the
@code{dllimport} attribute on imported variables can be avoided by passing the
@option{--enable-auto-import} switch to the GNU linker.  As with
functions, using the attribute for a variable eliminates a thunk in
the DLL@.

One drawback to using this attribute is that a pointer to a
@emph{variable} marked as @code{dllimport} cannot be used as a constant
address. However, a pointer to a @emph{function} with the
@code{dllimport} attribute can be used as a constant initializer; in
this case, the address of a stub function in the import lib is
referenced.  On Microsoft Windows targets, the attribute can be disabled
for functions by setting the @option{-mnop-fun-dllimport} flag.
@end table

@node MIPS Function Attributes
@subsection MIPS Function Attributes

These function attributes are supported by the MIPS back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, MIPS
Use this attribute to indicate that the specified function is an interrupt
handler.  The compiler generates function entry and exit sequences suitable
for use in an interrupt handler when this attribute is present.
An optional argument is supported for the interrupt attribute which allows
the interrupt mode to be described.  By default GCC assumes the external
interrupt controller (EIC) mode is in use, this can be explicitly set using
@code{eic}.  When interrupts are non-masked then the requested Interrupt
Priority Level (IPL) is copied to the current IPL which has the effect of only
enabling higher priority interrupts.  To use vectored interrupt mode use
the argument @code{vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]}, this will change
the behavior of the non-masked interrupt support and GCC will arrange to mask
all interrupts from sw0 up to and including the specified interrupt vector.

You can use the following attributes to modify the behavior
of an interrupt handler:
@table @code
@item use_shadow_register_set
@cindex @code{use_shadow_register_set} function attribute, MIPS
Assume that the handler uses a shadow register set, instead of
the main general-purpose registers.  An optional argument @code{intstack} is
supported to indicate that the shadow register set contains a valid stack
pointer.

@item keep_interrupts_masked
@cindex @code{keep_interrupts_masked} function attribute, MIPS
Keep interrupts masked for the whole function.  Without this attribute,
GCC tries to reenable interrupts for as much of the function as it can.

@item use_debug_exception_return
@cindex @code{use_debug_exception_return} function attribute, MIPS
Return using the @code{deret} instruction.  Interrupt handlers that don't
have this attribute return using @code{eret} instead.
@end table

You can use any combination of these attributes, as shown below:
@smallexample
void __attribute__ ((interrupt)) v0 ();
void __attribute__ ((interrupt, use_shadow_register_set)) v1 ();
void __attribute__ ((interrupt, keep_interrupts_masked)) v2 ();
void __attribute__ ((interrupt, use_debug_exception_return)) v3 ();
void __attribute__ ((interrupt, use_shadow_register_set,
                     keep_interrupts_masked)) v4 ();
void __attribute__ ((interrupt, use_shadow_register_set,
                     use_debug_exception_return)) v5 ();
void __attribute__ ((interrupt, keep_interrupts_masked,
                     use_debug_exception_return)) v6 ();
void __attribute__ ((interrupt, use_shadow_register_set,
                     keep_interrupts_masked,
                     use_debug_exception_return)) v7 ();
void __attribute__ ((interrupt("eic"))) v8 ();
void __attribute__ ((interrupt("vector=hw3"))) v9 ();
@end smallexample

@item long_call
@itemx short_call
@itemx near
@itemx far
@cindex indirect calls, MIPS
@cindex @code{long_call} function attribute, MIPS
@cindex @code{short_call} function attribute, MIPS
@cindex @code{near} function attribute, MIPS
@cindex @code{far} function attribute, MIPS
These attributes specify how a particular function is called on MIPS@.
The attributes override the @option{-mlong-calls} (@pxref{MIPS Options})
command-line switch.  The @code{long_call} and @code{far} attributes are
synonyms, and cause the compiler to always call
the function by first loading its address into a register, and then using
the contents of that register.  The @code{short_call} and @code{near}
attributes are synonyms, and have the opposite
effect; they specify that non-PIC calls should be made using the more
efficient @code{jal} instruction.

@item mips16
@itemx nomips16
@cindex @code{mips16} function attribute, MIPS
@cindex @code{nomips16} function attribute, MIPS

On MIPS targets, you can use the @code{mips16} and @code{nomips16}
function attributes to locally select or turn off MIPS16 code generation.
A function with the @code{mips16} attribute is emitted as MIPS16 code,
while MIPS16 code generation is disabled for functions with the
@code{nomips16} attribute.  These attributes override the
@option{-mips16} and @option{-mno-mips16} options on the command line
(@pxref{MIPS Options}).

When compiling files containing mixed MIPS16 and non-MIPS16 code, the
preprocessor symbol @code{__mips16} reflects the setting on the command line,
not that within individual functions.  Mixed MIPS16 and non-MIPS16 code
may interact badly with some GCC extensions such as @code{__builtin_apply}
(@pxref{Constructing Calls}).

@item micromips, MIPS
@itemx nomicromips, MIPS
@cindex @code{micromips} function attribute
@cindex @code{nomicromips} function attribute

On MIPS targets, you can use the @code{micromips} and @code{nomicromips}
function attributes to locally select or turn off microMIPS code generation.
A function with the @code{micromips} attribute is emitted as microMIPS code,
while microMIPS code generation is disabled for functions with the
@code{nomicromips} attribute.  These attributes override the
@option{-mmicromips} and @option{-mno-micromips} options on the command line
(@pxref{MIPS Options}).

When compiling files containing mixed microMIPS and non-microMIPS code, the
preprocessor symbol @code{__mips_micromips} reflects the setting on the
command line,
not that within individual functions.  Mixed microMIPS and non-microMIPS code
may interact badly with some GCC extensions such as @code{__builtin_apply}
(@pxref{Constructing Calls}).

@item nocompression
@cindex @code{nocompression} function attribute, MIPS
On MIPS targets, you can use the @code{nocompression} function attribute
to locally turn off MIPS16 and microMIPS code generation.  This attribute
overrides the @option{-mips16} and @option{-mmicromips} options on the
command line (@pxref{MIPS Options}).
@end table

@node MSP430 Function Attributes
@subsection MSP430 Function Attributes

These function attributes are supported by the MSP430 back end:

@table @code
@item critical
@cindex @code{critical} function attribute, MSP430
Critical functions disable interrupts upon entry and restore the
previous interrupt state upon exit.  Critical functions cannot also
have the @code{naked}, @code{reentrant} or @code{interrupt} attributes.

The MSP430 hardware ensures that interrupts are disabled on entry to
@code{interrupt} functions, and restores the previous interrupt state
on exit. The @code{critical} attribute is therefore redundant on
@code{interrupt} functions.

@item interrupt
@cindex @code{interrupt} function attribute, MSP430
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

You can provide an argument to the interrupt
attribute which specifies a name or number.  If the argument is a
number it indicates the slot in the interrupt vector table (0 - 31) to
which this handler should be assigned.  If the argument is a name it
is treated as a symbolic name for the vector slot.  These names should
match up with appropriate entries in the linker script.  By default
the names @code{watchdog} for vector 26, @code{nmi} for vector 30 and
@code{reset} for vector 31 are recognized.

@item naked
@cindex @code{naked} function attribute, MSP430
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item reentrant
@cindex @code{reentrant} function attribute, MSP430
Reentrant functions disable interrupts upon entry and enable them
upon exit.  Reentrant functions cannot also have the @code{naked}
or @code{critical} attributes.  They can have the @code{interrupt}
attribute.

@item wakeup
@cindex @code{wakeup} function attribute, MSP430
This attribute only applies to interrupt functions.  It is silently
ignored if applied to a non-interrupt function.  A wakeup interrupt
function will rouse the processor from any low-power state that it
might be in when the function exits.

@item lower
@itemx upper
@itemx either
@cindex @code{lower} function attribute, MSP430
@cindex @code{upper} function attribute, MSP430
@cindex @code{either} function attribute, MSP430
On the MSP430 target these attributes can be used to specify whether
the function or variable should be placed into low memory, high
memory, or the placement should be left to the linker to decide.  The
attributes are only significant if compiling for the MSP430X
architecture in the large memory model.

The attributes work in conjunction with a linker script that has been
augmented to specify where to place sections with a @code{.lower} and
a @code{.upper} prefix.  So, for example, as well as placing the
@code{.data} section, the script also specifies the placement of a
@code{.lower.data} and a @code{.upper.data} section.  The intention
is that @code{lower} sections are placed into a small but easier to
access memory region and the upper sections are placed into a larger, but
slower to access, region.

The @code{either} attribute is special.  It tells the linker to place
the object into the corresponding @code{lower} section if there is
room for it.  If there is insufficient room then the object is placed
into the corresponding @code{upper} section instead.  Note that the
placement algorithm is not very sophisticated.  It does not attempt to
find an optimal packing of the @code{lower} sections.  It just makes
one pass over the objects and does the best that it can.  Using the
@option{-ffunction-sections} and @option{-fdata-sections} command-line
options can help the packing, however, since they produce smaller,
easier to pack regions.
@end table

@node NDS32 Function Attributes
@subsection NDS32 Function Attributes

These function attributes are supported by the NDS32 back end:

@table @code
@item exception
@cindex @code{exception} function attribute
@cindex exception handler functions, NDS32
Use this attribute on the NDS32 target to indicate that the specified function
is an exception handler.  The compiler will generate corresponding sections
for use in an exception handler.

@item interrupt
@cindex @code{interrupt} function attribute, NDS32
On NDS32 target, this attribute indicates that the specified function
is an interrupt handler.  The compiler generates corresponding sections
for use in an interrupt handler.  You can use the following attributes
to modify the behavior:
@table @code
@item nested
@cindex @code{nested} function attribute, NDS32
This interrupt service routine is interruptible.
@item not_nested
@cindex @code{not_nested} function attribute, NDS32
This interrupt service routine is not interruptible.
@item nested_ready
@cindex @code{nested_ready} function attribute, NDS32
This interrupt service routine is interruptible after @code{PSW.GIE}
(global interrupt enable) is set.  This allows interrupt service routine to
finish some short critical code before enabling interrupts.
@item save_all
@cindex @code{save_all} function attribute, NDS32
The system will help save all registers into stack before entering
interrupt handler.
@item partial_save
@cindex @code{partial_save} function attribute, NDS32
The system will help save caller registers into stack before entering
interrupt handler.
@end table

@item naked
@cindex @code{naked} function attribute, NDS32
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item reset
@cindex @code{reset} function attribute, NDS32
@cindex reset handler functions
Use this attribute on the NDS32 target to indicate that the specified function
is a reset handler.  The compiler will generate corresponding sections
for use in a reset handler.  You can use the following attributes
to provide extra exception handling:
@table @code
@item nmi
@cindex @code{nmi} function attribute, NDS32
Provide a user-defined function to handle NMI exception.
@item warm
@cindex @code{warm} function attribute, NDS32
Provide a user-defined function to handle warm reset exception.
@end table
@end table

@node Nios II Function Attributes
@subsection Nios II Function Attributes

These function attributes are supported by the Nios II back end:

@table @code
@item target (@var{options})
@cindex @code{target} function attribute
As discussed in @ref{Common Function Attributes}, this attribute 
allows specification of target-specific compilation options.

When compiling for Nios II, the following options are allowed:

@table @samp
@item custom-@var{insn}=@var{N}
@itemx no-custom-@var{insn}
@cindex @code{target("custom-@var{insn}=@var{N}")} function attribute, Nios II
@cindex @code{target("no-custom-@var{insn}")} function attribute, Nios II
Each @samp{custom-@var{insn}=@var{N}} attribute locally enables use of a
custom instruction with encoding @var{N} when generating code that uses 
@var{insn}.  Similarly, @samp{no-custom-@var{insn}} locally inhibits use of
the custom instruction @var{insn}.
These target attributes correspond to the
@option{-mcustom-@var{insn}=@var{N}} and @option{-mno-custom-@var{insn}}
command-line options, and support the same set of @var{insn} keywords.
@xref{Nios II Options}, for more information.

@item custom-fpu-cfg=@var{name}
@cindex @code{target("custom-fpu-cfg=@var{name}")} function attribute, Nios II
This attribute corresponds to the @option{-mcustom-fpu-cfg=@var{name}}
command-line option, to select a predefined set of custom instructions
named @var{name}.
@xref{Nios II Options}, for more information.
@end table
@end table

@node Nvidia PTX Function Attributes
@subsection Nvidia PTX Function Attributes

These function attributes are supported by the Nvidia PTX back end:

@table @code
@item kernel
@cindex @code{kernel} attribute, Nvidia PTX
This attribute indicates that the corresponding function should be compiled
as a kernel function, which can be invoked from the host via the CUDA RT 
library.
By default functions are only callable only from other PTX functions.

Kernel functions must have @code{void} return type.
@end table

@node PowerPC Function Attributes
@subsection PowerPC Function Attributes

These function attributes are supported by the PowerPC back end:

@table @code
@item longcall
@itemx shortcall
@cindex indirect calls, PowerPC
@cindex @code{longcall} function attribute, PowerPC
@cindex @code{shortcall} function attribute, PowerPC
The @code{longcall} attribute
indicates that the function might be far away from the call site and
require a different (more expensive) calling sequence.  The
@code{shortcall} attribute indicates that the function is always close
enough for the shorter calling sequence to be used.  These attributes
override both the @option{-mlongcall} switch and
the @code{#pragma longcall} setting.

@xref{RS/6000 and PowerPC Options}, for more information on whether long
calls are necessary.

@item target (@var{options})
@cindex @code{target} function attribute
As discussed in @ref{Common Function Attributes}, this attribute 
allows specification of target-specific compilation options.

On the PowerPC, the following options are allowed:

@table @samp
@item altivec
@itemx no-altivec
@cindex @code{target("altivec")} function attribute, PowerPC
Generate code that uses (does not use) AltiVec instructions.  In
32-bit code, you cannot enable AltiVec instructions unless
@option{-mabi=altivec} is used on the command line.

@item cmpb
@itemx no-cmpb
@cindex @code{target("cmpb")} function attribute, PowerPC
Generate code that uses (does not use) the compare bytes instruction
implemented on the POWER6 processor and other processors that support
the PowerPC V2.05 architecture.

@item dlmzb
@itemx no-dlmzb
@cindex @code{target("dlmzb")} function attribute, PowerPC
Generate code that uses (does not use) the string-search @samp{dlmzb}
instruction on the IBM 405, 440, 464 and 476 processors.  This instruction is
generated by default when targeting those processors.

@item fprnd
@itemx no-fprnd
@cindex @code{target("fprnd")} function attribute, PowerPC
Generate code that uses (does not use) the FP round to integer
instructions implemented on the POWER5+ processor and other processors
that support the PowerPC V2.03 architecture.

@item hard-dfp
@itemx no-hard-dfp
@cindex @code{target("hard-dfp")} function attribute, PowerPC
Generate code that uses (does not use) the decimal floating-point
instructions implemented on some POWER processors.

@item isel
@itemx no-isel
@cindex @code{target("isel")} function attribute, PowerPC
Generate code that uses (does not use) ISEL instruction.

@item mfcrf
@itemx no-mfcrf
@cindex @code{target("mfcrf")} function attribute, PowerPC
Generate code that uses (does not use) the move from condition
register field instruction implemented on the POWER4 processor and
other processors that support the PowerPC V2.01 architecture.

@item mulhw
@itemx no-mulhw
@cindex @code{target("mulhw")} function attribute, PowerPC
Generate code that uses (does not use) the half-word multiply and
multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
These instructions are generated by default when targeting those
processors.

@item multiple
@itemx no-multiple
@cindex @code{target("multiple")} function attribute, PowerPC
Generate code that uses (does not use) the load multiple word
instructions and the store multiple word instructions.

@item update
@itemx no-update
@cindex @code{target("update")} function attribute, PowerPC
Generate code that uses (does not use) the load or store instructions
that update the base register to the address of the calculated memory
location.

@item popcntb
@itemx no-popcntb
@cindex @code{target("popcntb")} function attribute, PowerPC
Generate code that uses (does not use) the popcount and double-precision
FP reciprocal estimate instruction implemented on the POWER5
processor and other processors that support the PowerPC V2.02
architecture.

@item popcntd
@itemx no-popcntd
@cindex @code{target("popcntd")} function attribute, PowerPC
Generate code that uses (does not use) the popcount instruction
implemented on the POWER7 processor and other processors that support
the PowerPC V2.06 architecture.

@item powerpc-gfxopt
@itemx no-powerpc-gfxopt
@cindex @code{target("powerpc-gfxopt")} function attribute, PowerPC
Generate code that uses (does not use) the optional PowerPC
architecture instructions in the Graphics group, including
floating-point select.

@item powerpc-gpopt
@itemx no-powerpc-gpopt
@cindex @code{target("powerpc-gpopt")} function attribute, PowerPC
Generate code that uses (does not use) the optional PowerPC
architecture instructions in the General Purpose group, including
floating-point square root.

@item recip-precision
@itemx no-recip-precision
@cindex @code{target("recip-precision")} function attribute, PowerPC
Assume (do not assume) that the reciprocal estimate instructions
provide higher-precision estimates than is mandated by the PowerPC
ABI.

@item string
@itemx no-string
@cindex @code{target("string")} function attribute, PowerPC
Generate code that uses (does not use) the load string instructions
and the store string word instructions to save multiple registers and
do small block moves.

@item vsx
@itemx no-vsx
@cindex @code{target("vsx")} function attribute, PowerPC
Generate code that uses (does not use) vector/scalar (VSX)
instructions, and also enable the use of built-in functions that allow
more direct access to the VSX instruction set.  In 32-bit code, you
cannot enable VSX or AltiVec instructions unless
@option{-mabi=altivec} is used on the command line.

@item friz
@itemx no-friz
@cindex @code{target("friz")} function attribute, PowerPC
Generate (do not generate) the @code{friz} instruction when the
@option{-funsafe-math-optimizations} option is used to optimize
rounding a floating-point value to 64-bit integer and back to floating
point.  The @code{friz} instruction does not return the same value if
the floating-point number is too large to fit in an integer.

@item avoid-indexed-addresses
@itemx no-avoid-indexed-addresses
@cindex @code{target("avoid-indexed-addresses")} function attribute, PowerPC
Generate code that tries to avoid (not avoid) the use of indexed load
or store instructions.

@item paired
@itemx no-paired
@cindex @code{target("paired")} function attribute, PowerPC
Generate code that uses (does not use) the generation of PAIRED simd
instructions.

@item longcall
@itemx no-longcall
@cindex @code{target("longcall")} function attribute, PowerPC
Generate code that assumes (does not assume) that all calls are far
away so that a longer more expensive calling sequence is required.

@item cpu=@var{CPU}
@cindex @code{target("cpu=@var{CPU}")} function attribute, PowerPC
Specify the architecture to generate code for when compiling the
function.  If you select the @code{target("cpu=power7")} attribute when
generating 32-bit code, VSX and AltiVec instructions are not generated
unless you use the @option{-mabi=altivec} option on the command line.

@item tune=@var{TUNE}
@cindex @code{target("tune=@var{TUNE}")} function attribute, PowerPC
Specify the architecture to tune for when compiling the function.  If
you do not specify the @code{target("tune=@var{TUNE}")} attribute and
you do specify the @code{target("cpu=@var{CPU}")} attribute,
compilation tunes for the @var{CPU} architecture, and not the
default tuning specified on the command line.
@end table

On the PowerPC, the inliner does not inline a
function that has different target options than the caller, unless the
callee has a subset of the target options of the caller.
@end table

@node RISC-V Function Attributes
@subsection RISC-V Function Attributes

These function attributes are supported by the RISC-V back end:

@table @code
@item naked
@cindex @code{naked} function attribute, RISC-V
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item interrupt
@cindex @code{interrupt} function attribute, RISC-V
Use this attribute to indicate that the specified function is an interrupt
handler.  The compiler generates function entry and exit sequences suitable
for use in an interrupt handler when this attribute is present.

You can specify the kind of interrupt to be handled by adding an optional
parameter to the interrupt attribute like this:

@smallexample
void f (void) __attribute__ ((interrupt ("user")));
@end smallexample

Permissible values for this parameter are @code{user}, @code{supervisor},
and @code{machine}.  If there is no parameter, then it defaults to
@code{machine}.
@end table

@node RL78 Function Attributes
@subsection RL78 Function Attributes

These function attributes are supported by the RL78 back end:

@table @code
@item interrupt
@itemx brk_interrupt
@cindex @code{interrupt} function attribute, RL78
@cindex @code{brk_interrupt} function attribute, RL78
These attributes indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

Use @code{brk_interrupt} instead of @code{interrupt} for
handlers intended to be used with the @code{BRK} opcode (i.e.@: those
that must end with @code{RETB} instead of @code{RETI}).

@item naked
@cindex @code{naked} function attribute, RL78
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.
@end table

@node RX Function Attributes
@subsection RX Function Attributes

These function attributes are supported by the RX back end:

@table @code
@item fast_interrupt
@cindex @code{fast_interrupt} function attribute, RX
Use this attribute on the RX port to indicate that the specified
function is a fast interrupt handler.  This is just like the
@code{interrupt} attribute, except that @code{freit} is used to return
instead of @code{reit}.

@item interrupt
@cindex @code{interrupt} function attribute, RX
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.

On RX and RL78 targets, you may specify one or more vector numbers as arguments
to the attribute, as well as naming an alternate table name.
Parameters are handled sequentially, so one handler can be assigned to
multiple entries in multiple tables.  One may also pass the magic
string @code{"$default"} which causes the function to be used for any
unfilled slots in the current table.

This example shows a simple assignment of a function to one vector in
the default table (note that preprocessor macros may be used for
chip-specific symbolic vector names):
@smallexample
void __attribute__ ((interrupt (5))) txd1_handler ();
@end smallexample

This example assigns a function to two slots in the default table
(using preprocessor macros defined elsewhere) and makes it the default
for the @code{dct} table:
@smallexample
void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default")))
	txd1_handler ();
@end smallexample

@item naked
@cindex @code{naked} function attribute, RX
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item vector
@cindex @code{vector} function attribute, RX
This RX attribute is similar to the @code{interrupt} attribute, including its
parameters, but does not make the function an interrupt-handler type
function (i.e.@: it retains the normal C function calling ABI).  See the
@code{interrupt} attribute for a description of its arguments.
@end table

@node S/390 Function Attributes
@subsection S/390 Function Attributes

These function attributes are supported on the S/390:

@table @code
@item hotpatch (@var{halfwords-before-function-label},@var{halfwords-after-function-label})
@cindex @code{hotpatch} function attribute, S/390

On S/390 System z targets, you can use this function attribute to
make GCC generate a ``hot-patching'' function prologue.  If the
@option{-mhotpatch=} command-line option is used at the same time,
the @code{hotpatch} attribute takes precedence.  The first of the
two arguments specifies the number of halfwords to be added before
the function label.  A second argument can be used to specify the
number of halfwords to be added after the function label.  For
both arguments the maximum allowed value is 1000000.

If both arguments are zero, hotpatching is disabled.

@item target (@var{options})
@cindex @code{target} function attribute
As discussed in @ref{Common Function Attributes}, this attribute
allows specification of target-specific compilation options.

On S/390, the following options are supported:

@table @samp
@item arch=
@item tune=
@item stack-guard=
@item stack-size=
@item branch-cost=
@item warn-framesize=
@item backchain
@itemx no-backchain
@item hard-dfp
@itemx no-hard-dfp
@item hard-float
@itemx soft-float
@item htm
@itemx no-htm
@item vx
@itemx no-vx
@item packed-stack
@itemx no-packed-stack
@item small-exec
@itemx no-small-exec
@item mvcle
@itemx no-mvcle
@item warn-dynamicstack
@itemx no-warn-dynamicstack
@end table

The options work exactly like the S/390 specific command line
options (without the prefix @option{-m}) except that they do not
change any feature macros.  For example,

@smallexample
@code{target("no-vx")}
@end smallexample

does not undefine the @code{__VEC__} macro.
@end table

@node SH Function Attributes
@subsection SH Function Attributes

These function attributes are supported on the SH family of processors:

@table @code
@item function_vector
@cindex @code{function_vector} function attribute, SH
@cindex calling functions through the function vector on SH2A
On SH2A targets, this attribute declares a function to be called using the
TBR relative addressing mode.  The argument to this attribute is the entry
number of the same function in a vector table containing all the TBR
relative addressable functions.  For correct operation the TBR must be setup
accordingly to point to the start of the vector table before any functions with
this attribute are invoked.  Usually a good place to do the initialization is
the startup routine.  The TBR relative vector table can have at max 256 function
entries.  The jumps to these functions are generated using a SH2A specific,
non delayed branch instruction JSR/N @@(disp8,TBR).  You must use GAS and GLD
from GNU binutils version 2.7 or later for this attribute to work correctly.

In an application, for a function being called once, this attribute
saves at least 8 bytes of code; and if other successive calls are being
made to the same function, it saves 2 bytes of code per each of these
calls.

@item interrupt_handler
@cindex @code{interrupt_handler} function attribute, SH
Use this attribute to
indicate that the specified function is an interrupt handler.  The compiler
generates function entry and exit sequences suitable for use in an
interrupt handler when this attribute is present.

@item nosave_low_regs
@cindex @code{nosave_low_regs} function attribute, SH
Use this attribute on SH targets to indicate that an @code{interrupt_handler}
function should not save and restore registers R0..R7.  This can be used on SH3*
and SH4* targets that have a second R0..R7 register bank for non-reentrant
interrupt handlers.

@item renesas
@cindex @code{renesas} function attribute, SH
On SH targets this attribute specifies that the function or struct follows the
Renesas ABI.

@item resbank
@cindex @code{resbank} function attribute, SH
On the SH2A target, this attribute enables the high-speed register
saving and restoration using a register bank for @code{interrupt_handler}
routines.  Saving to the bank is performed automatically after the CPU
accepts an interrupt that uses a register bank.

The nineteen 32-bit registers comprising general register R0 to R14,
control register GBR, and system registers MACH, MACL, and PR and the
vector table address offset are saved into a register bank.  Register
banks are stacked in first-in last-out (FILO) sequence.  Restoration
from the bank is executed by issuing a RESBANK instruction.

@item sp_switch
@cindex @code{sp_switch} function attribute, SH
Use this attribute on the SH to indicate an @code{interrupt_handler}
function should switch to an alternate stack.  It expects a string
argument that names a global variable holding the address of the
alternate stack.

@smallexample
void *alt_stack;
void f () __attribute__ ((interrupt_handler,
                          sp_switch ("alt_stack")));
@end smallexample

@item trap_exit
@cindex @code{trap_exit} function attribute, SH
Use this attribute on the SH for an @code{interrupt_handler} to return using
@code{trapa} instead of @code{rte}.  This attribute expects an integer
argument specifying the trap number to be used.

@item trapa_handler
@cindex @code{trapa_handler} function attribute, SH
On SH targets this function attribute is similar to @code{interrupt_handler}
but it does not save and restore all registers.
@end table

@node Symbian OS Function Attributes
@subsection Symbian OS Function Attributes

@xref{Microsoft Windows Function Attributes}, for discussion of the
@code{dllexport} and @code{dllimport} attributes.

@node V850 Function Attributes
@subsection V850 Function Attributes

The V850 back end supports these function attributes:

@table @code
@item interrupt
@itemx interrupt_handler
@cindex @code{interrupt} function attribute, V850
@cindex @code{interrupt_handler} function attribute, V850
Use these attributes to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when either attribute is present.
@end table

@node Visium Function Attributes
@subsection Visium Function Attributes

These function attributes are supported by the Visium back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, Visium
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.
@end table

@node x86 Function Attributes
@subsection x86 Function Attributes

These function attributes are supported by the x86 back end:

@table @code
@item cdecl
@cindex @code{cdecl} function attribute, x86-32
@cindex functions that pop the argument stack on x86-32
@opindex mrtd
On the x86-32 targets, the @code{cdecl} attribute causes the compiler to
assume that the calling function pops off the stack space used to
pass arguments.  This is
useful to override the effects of the @option{-mrtd} switch.

@item fastcall
@cindex @code{fastcall} function attribute, x86-32
@cindex functions that pop the argument stack on x86-32
On x86-32 targets, the @code{fastcall} attribute causes the compiler to
pass the first argument (if of integral type) in the register ECX and
the second argument (if of integral type) in the register EDX@.  Subsequent
and other typed arguments are passed on the stack.  The called function
pops the arguments off the stack.  If the number of arguments is variable all
arguments are pushed on the stack.

@item thiscall
@cindex @code{thiscall} function attribute, x86-32
@cindex functions that pop the argument stack on x86-32
On x86-32 targets, the @code{thiscall} attribute causes the compiler to
pass the first argument (if of integral type) in the register ECX.
Subsequent and other typed arguments are passed on the stack. The called
function pops the arguments off the stack.
If the number of arguments is variable all arguments are pushed on the
stack.
The @code{thiscall} attribute is intended for C++ non-static member functions.
As a GCC extension, this calling convention can be used for C functions
and for static member methods.

@item ms_abi
@itemx sysv_abi
@cindex @code{ms_abi} function attribute, x86
@cindex @code{sysv_abi} function attribute, x86

On 32-bit and 64-bit x86 targets, you can use an ABI attribute
to indicate which calling convention should be used for a function.  The
@code{ms_abi} attribute tells the compiler to use the Microsoft ABI,
while the @code{sysv_abi} attribute tells the compiler to use the ABI
used on GNU/Linux and other systems.  The default is to use the Microsoft ABI
when targeting Windows.  On all other systems, the default is the x86/AMD ABI.

Note, the @code{ms_abi} attribute for Microsoft Windows 64-bit targets currently
requires the @option{-maccumulate-outgoing-args} option.

@item callee_pop_aggregate_return (@var{number})
@cindex @code{callee_pop_aggregate_return} function attribute, x86

On x86-32 targets, you can use this attribute to control how
aggregates are returned in memory.  If the caller is responsible for
popping the hidden pointer together with the rest of the arguments, specify
@var{number} equal to zero.  If callee is responsible for popping the
hidden pointer, specify @var{number} equal to one.  

The default x86-32 ABI assumes that the callee pops the
stack for hidden pointer.  However, on x86-32 Microsoft Windows targets,
the compiler assumes that the
caller pops the stack for hidden pointer.

@item ms_hook_prologue
@cindex @code{ms_hook_prologue} function attribute, x86

On 32-bit and 64-bit x86 targets, you can use
this function attribute to make GCC generate the ``hot-patching'' function
prologue used in Win32 API functions in Microsoft Windows XP Service Pack 2
and newer.

@item naked
@cindex @code{naked} function attribute, x86
This attribute allows the compiler to construct the
requisite function declaration, while allowing the body of the
function to be assembly code. The specified function will not have
prologue/epilogue sequences generated by the compiler. Only basic
@code{asm} statements can safely be included in naked functions
(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
basic @code{asm} and C code may appear to work, they cannot be
depended upon to work reliably and are not supported.

@item regparm (@var{number})
@cindex @code{regparm} function attribute, x86
@cindex functions that are passed arguments in registers on x86-32
On x86-32 targets, the @code{regparm} attribute causes the compiler to
pass arguments number one to @var{number} if they are of integral type
in registers EAX, EDX, and ECX instead of on the stack.  Functions that
take a variable number of arguments continue to be passed all of their
arguments on the stack.

Beware that on some ELF systems this attribute is unsuitable for
global functions in shared libraries with lazy binding (which is the
default).  Lazy binding sends the first call via resolving code in
the loader, which might assume EAX, EDX and ECX can be clobbered, as
per the standard calling conventions.  Solaris 8 is affected by this.
Systems with the GNU C Library version 2.1 or higher
and FreeBSD are believed to be
safe since the loaders there save EAX, EDX and ECX.  (Lazy binding can be
disabled with the linker or the loader if desired, to avoid the
problem.)

@item sseregparm
@cindex @code{sseregparm} function attribute, x86
On x86-32 targets with SSE support, the @code{sseregparm} attribute
causes the compiler to pass up to 3 floating-point arguments in
SSE registers instead of on the stack.  Functions that take a
variable number of arguments continue to pass all of their
floating-point arguments on the stack.

@item force_align_arg_pointer
@cindex @code{force_align_arg_pointer} function attribute, x86
On x86 targets, the @code{force_align_arg_pointer} attribute may be
applied to individual function definitions, generating an alternate
prologue and epilogue that realigns the run-time stack if necessary.
This supports mixing legacy codes that run with a 4-byte aligned stack
with modern codes that keep a 16-byte stack for SSE compatibility.

@item stdcall
@cindex @code{stdcall} function attribute, x86-32
@cindex functions that pop the argument stack on x86-32
On x86-32 targets, the @code{stdcall} attribute causes the compiler to
assume that the called function pops off the stack space used to
pass arguments, unless it takes a variable number of arguments.

@item no_caller_saved_registers
@cindex @code{no_caller_saved_registers} function attribute, x86
Use this attribute to indicate that the specified function has no
caller-saved registers. That is, all registers are callee-saved. For
example, this attribute can be used for a function called from an
interrupt handler. The compiler generates proper function entry and
exit sequences to save and restore any modified registers, except for
the EFLAGS register.  Since GCC doesn't preserve SSE, MMX nor x87
states, the GCC option @option{-mgeneral-regs-only} should be used to
compile functions with @code{no_caller_saved_registers} attribute.

@item interrupt
@cindex @code{interrupt} function attribute, x86
Use this attribute to indicate that the specified function is an
interrupt handler or an exception handler (depending on parameters passed
to the function, explained further).  The compiler generates function
entry and exit sequences suitable for use in an interrupt handler when
this attribute is present.  The @code{IRET} instruction, instead of the
@code{RET} instruction, is used to return from interrupt handlers.  All
registers, except for the EFLAGS register which is restored by the
@code{IRET} instruction, are preserved by the compiler.  Since GCC
doesn't preserve SSE, MMX nor x87 states, the GCC option
@option{-mgeneral-regs-only} should be used to compile interrupt and
exception handlers.

Any interruptible-without-stack-switch code must be compiled with
@option{-mno-red-zone} since interrupt handlers can and will, because
of the hardware design, touch the red zone.

An interrupt handler must be declared with a mandatory pointer
argument:

@smallexample
struct interrupt_frame;

__attribute__ ((interrupt))
void
f (struct interrupt_frame *frame)
@{
@}
@end smallexample

@noindent
and you must define @code{struct interrupt_frame} as described in the
processor's manual.

Exception handlers differ from interrupt handlers because the system
pushes an error code on the stack.  An exception handler declaration is
similar to that for an interrupt handler, but with a different mandatory
function signature.  The compiler arranges to pop the error code off the
stack before the @code{IRET} instruction.

@smallexample
#ifdef __x86_64__
typedef unsigned long long int uword_t;
#else
typedef unsigned int uword_t;
#endif

struct interrupt_frame;

__attribute__ ((interrupt))
void
f (struct interrupt_frame *frame, uword_t error_code)
@{
  ...
@}
@end smallexample

Exception handlers should only be used for exceptions that push an error
code; you should use an interrupt handler in other cases.  The system
will crash if the wrong kind of handler is used.

@item target (@var{options})
@cindex @code{target} function attribute
As discussed in @ref{Common Function Attributes}, this attribute 
allows specification of target-specific compilation options.

On the x86, the following options are allowed:
@table @samp
@item 3dnow
@itemx no-3dnow
@cindex @code{target("3dnow")} function attribute, x86
Enable/disable the generation of the 3DNow!@: instructions.

@item 3dnowa
@itemx no-3dnowa
@cindex @code{target("3dnowa")} function attribute, x86
Enable/disable the generation of the enhanced 3DNow!@: instructions.

@item abm
@itemx no-abm
@cindex @code{target("abm")} function attribute, x86
Enable/disable the generation of the advanced bit instructions.

@item adx
@itemx no-adx
@cindex @code{target("adx")} function attribute, x86
Enable/disable the generation of the ADX instructions.

@item aes
@itemx no-aes
@cindex @code{target("aes")} function attribute, x86
Enable/disable the generation of the AES instructions.

@item avx
@itemx no-avx
@cindex @code{target("avx")} function attribute, x86
Enable/disable the generation of the AVX instructions.

@item avx2
@itemx no-avx2
@cindex @code{target("avx2")} function attribute, x86
Enable/disable the generation of the AVX2 instructions.

@item avx5124fmaps
@itemx no-avx5124fmaps
@cindex @code{target("avx5124fmaps")} function attribute, x86
Enable/disable the generation of the AVX5124FMAPS instructions.

@item avx5124vnniw
@itemx no-avx5124vnniw
@cindex @code{target("avx5124vnniw")} function attribute, x86
Enable/disable the generation of the AVX5124VNNIW instructions.

@item avx512bitalg
@itemx no-avx512bitalg
@cindex @code{target("avx512bitalg")} function attribute, x86
Enable/disable the generation of the AVX512BITALG instructions.

@item avx512bw
@itemx no-avx512bw
@cindex @code{target("avx512bw")} function attribute, x86
Enable/disable the generation of the AVX512BW instructions.

@item avx512cd
@itemx no-avx512cd
@cindex @code{target("avx512cd")} function attribute, x86
Enable/disable the generation of the AVX512CD instructions.

@item avx512dq
@itemx no-avx512dq
@cindex @code{target("avx512dq")} function attribute, x86
Enable/disable the generation of the AVX512DQ instructions.

@item avx512er
@itemx no-avx512er
@cindex @code{target("avx512er")} function attribute, x86
Enable/disable the generation of the AVX512ER instructions.

@item avx512f
@itemx no-avx512f
@cindex @code{target("avx512f")} function attribute, x86
Enable/disable the generation of the AVX512F instructions.

@item avx512ifma
@itemx no-avx512ifma
@cindex @code{target("avx512ifma")} function attribute, x86
Enable/disable the generation of the AVX512IFMA instructions.

@item avx512pf
@itemx no-avx512pf
@cindex @code{target("avx512pf")} function attribute, x86
Enable/disable the generation of the AVX512PF instructions.

@item avx512vbmi
@itemx no-avx512vbmi
@cindex @code{target("avx512vbmi")} function attribute, x86
Enable/disable the generation of the AVX512VBMI instructions.

@item avx512vbmi2
@itemx no-avx512vbmi2
@cindex @code{target("avx512vbmi2")} function attribute, x86
Enable/disable the generation of the AVX512VBMI2 instructions.

@item avx512vl
@itemx no-avx512vl
@cindex @code{target("avx512vl")} function attribute, x86
Enable/disable the generation of the AVX512VL instructions.

@item avx512vnni
@itemx no-avx512vnni
@cindex @code{target("avx512vnni")} function attribute, x86
Enable/disable the generation of the AVX512VNNI instructions.

@item avx512vpopcntdq
@itemx no-avx512vpopcntdq
@cindex @code{target("avx512vpopcntdq")} function attribute, x86
Enable/disable the generation of the AVX512VPOPCNTDQ instructions.

@item bmi
@itemx no-bmi
@cindex @code{target("bmi")} function attribute, x86
Enable/disable the generation of the BMI instructions.

@item bmi2
@itemx no-bmi2
@cindex @code{target("bmi2")} function attribute, x86
Enable/disable the generation of the BMI2 instructions.

@item cldemote
@itemx no-cldemote
@cindex @code{target("cldemote")} function attribute, x86
Enable/disable the generation of the CLDEMOTE instructions.

@item clflushopt
@itemx no-clflushopt
@cindex @code{target("clflushopt")} function attribute, x86
Enable/disable the generation of the CLFLUSHOPT instructions.

@item clwb
@itemx no-clwb
@cindex @code{target("clwb")} function attribute, x86
Enable/disable the generation of the CLWB instructions.

@item clzero
@itemx no-clzero
@cindex @code{target("clzero")} function attribute, x86
Enable/disable the generation of the CLZERO instructions.

@item crc32
@itemx no-crc32
@cindex @code{target("crc32")} function attribute, x86
Enable/disable the generation of the CRC32 instructions.

@item cx16
@itemx no-cx16
@cindex @code{target("cx16")} function attribute, x86
Enable/disable the generation of the CMPXCHG16B instructions.

@item default
@cindex @code{target("default")} function attribute, x86
@xref{Function Multiversioning}, where it is used to specify the
default function version.

@item f16c
@itemx no-f16c
@cindex @code{target("f16c")} function attribute, x86
Enable/disable the generation of the F16C instructions.

@item fma
@itemx no-fma
@cindex @code{target("fma")} function attribute, x86
Enable/disable the generation of the FMA instructions.

@item fma4
@itemx no-fma4
@cindex @code{target("fma4")} function attribute, x86
Enable/disable the generation of the FMA4 instructions.

@item fsgsbase
@itemx no-fsgsbase
@cindex @code{target("fsgsbase")} function attribute, x86
Enable/disable the generation of the FSGSBASE instructions.

@item fxsr
@itemx no-fxsr
@cindex @code{target("fxsr")} function attribute, x86
Enable/disable the generation of the FXSR instructions.

@item gfni
@itemx no-gfni
@cindex @code{target("gfni")} function attribute, x86
Enable/disable the generation of the GFNI instructions.

@item hle
@itemx no-hle
@cindex @code{target("hle")} function attribute, x86
Enable/disable the generation of the HLE instruction prefixes.

@item lwp
@itemx no-lwp
@cindex @code{target("lwp")} function attribute, x86
Enable/disable the generation of the LWP instructions.

@item lzcnt
@itemx no-lzcnt
@cindex @code{target("lzcnt")} function attribute, x86
Enable/disable the generation of the LZCNT instructions.

@item mmx
@itemx no-mmx
@cindex @code{target("mmx")} function attribute, x86
Enable/disable the generation of the MMX instructions.

@item movbe
@itemx no-movbe
@cindex @code{target("movbe")} function attribute, x86
Enable/disable the generation of the MOVBE instructions.

@item movdir64b
@itemx no-movdir64b
@cindex @code{target("movdir64b")} function attribute, x86
Enable/disable the generation of the MOVDIR64B instructions.

@item movdiri
@itemx no-movdiri
@cindex @code{target("movdiri")} function attribute, x86
Enable/disable the generation of the MOVDIRI instructions.

@item mwaitx
@itemx no-mwaitx
@cindex @code{target("mwaitx")} function attribute, x86
Enable/disable the generation of the MWAITX instructions.

@item pclmul
@itemx no-pclmul
@cindex @code{target("pclmul")} function attribute, x86
Enable/disable the generation of the PCLMUL instructions.

@item pconfig
@itemx no-pconfig
@cindex @code{target("pconfig")} function attribute, x86
Enable/disable the generation of the PCONFIG instructions.

@item pku
@itemx no-pku
@cindex @code{target("pku")} function attribute, x86
Enable/disable the generation of the PKU instructions.

@item popcnt
@itemx no-popcnt
@cindex @code{target("popcnt")} function attribute, x86
Enable/disable the generation of the POPCNT instruction.

@item prefetchwt1
@itemx no-prefetchwt1
@cindex @code{target("prefetchwt1")} function attribute, x86
Enable/disable the generation of the PREFETCHWT1 instructions.

@item prfchw
@itemx no-prfchw
@cindex @code{target("prfchw")} function attribute, x86
Enable/disable the generation of the PREFETCHW instruction.

@item ptwrite
@itemx no-ptwrite
@cindex @code{target("ptwrite")} function attribute, x86
Enable/disable the generation of the PTWRITE instructions.

@item rdpid
@itemx no-rdpid
@cindex @code{target("rdpid")} function attribute, x86
Enable/disable the generation of the RDPID instructions.

@item rdrnd
@itemx no-rdrnd
@cindex @code{target("rdrnd")} function attribute, x86
Enable/disable the generation of the RDRND instructions.

@item rdseed
@itemx no-rdseed
@cindex @code{target("rdseed")} function attribute, x86
Enable/disable the generation of the RDSEED instructions.

@item rtm
@itemx no-rtm
@cindex @code{target("rtm")} function attribute, x86
Enable/disable the generation of the RTM instructions.

@item sahf
@itemx no-sahf
@cindex @code{target("sahf")} function attribute, x86
Enable/disable the generation of the SAHF instructions.

@item sgx
@itemx no-sgx
@cindex @code{target("sgx")} function attribute, x86
Enable/disable the generation of the SGX instructions.

@item sha
@itemx no-sha
@cindex @code{target("sha")} function attribute, x86
Enable/disable the generation of the SHA instructions.

@item shstk
@itemx no-shstk
@cindex @code{target("shstk")} function attribute, x86
Enable/disable the shadow stack built-in functions from CET.

@item sse
@itemx no-sse
@cindex @code{target("sse")} function attribute, x86
Enable/disable the generation of the SSE instructions.

@item sse2
@itemx no-sse2
@cindex @code{target("sse2")} function attribute, x86
Enable/disable the generation of the SSE2 instructions.

@item sse3
@itemx no-sse3
@cindex @code{target("sse3")} function attribute, x86
Enable/disable the generation of the SSE3 instructions.

@item sse4
@itemx no-sse4
@cindex @code{target("sse4")} function attribute, x86
Enable/disable the generation of the SSE4 instructions (both SSE4.1
and SSE4.2).

@item sse4.1
@itemx no-sse4.1
@cindex @code{target("sse4.1")} function attribute, x86
Enable/disable the generation of the sse4.1 instructions.

@item sse4.2
@itemx no-sse4.2
@cindex @code{target("sse4.2")} function attribute, x86
Enable/disable the generation of the sse4.2 instructions.

@item sse4a
@itemx no-sse4a
@cindex @code{target("sse4a")} function attribute, x86
Enable/disable the generation of the SSE4A instructions.

@item ssse3
@itemx no-ssse3
@cindex @code{target("ssse3")} function attribute, x86
Enable/disable the generation of the SSSE3 instructions.

@item tbm
@itemx no-tbm
@cindex @code{target("tbm")} function attribute, x86
Enable/disable the generation of the TBM instructions.

@item vaes
@itemx no-vaes
@cindex @code{target("vaes")} function attribute, x86
Enable/disable the generation of the VAES instructions.

@item vpclmulqdq
@itemx no-vpclmulqdq
@cindex @code{target("vpclmulqdq")} function attribute, x86
Enable/disable the generation of the VPCLMULQDQ instructions.

@item waitpkg
@itemx no-waitpkg
@cindex @code{target("waitpkg")} function attribute, x86
Enable/disable the generation of the WAITPKG instructions.

@item wbnoinvd
@itemx no-wbnoinvd
@cindex @code{target("wbnoinvd")} function attribute, x86
Enable/disable the generation of the WBNOINVD instructions.

@item xop
@itemx no-xop
@cindex @code{target("xop")} function attribute, x86
Enable/disable the generation of the XOP instructions.

@item xsave
@itemx no-xsave
@cindex @code{target("xsave")} function attribute, x86
Enable/disable the generation of the XSAVE instructions.

@item xsavec
@itemx no-xsavec
@cindex @code{target("xsavec")} function attribute, x86
Enable/disable the generation of the XSAVEC instructions.

@item xsaveopt
@itemx no-xsaveopt
@cindex @code{target("xsaveopt")} function attribute, x86
Enable/disable the generation of the XSAVEOPT instructions.

@item xsaves
@itemx no-xsaves
@cindex @code{target("xsaves")} function attribute, x86
Enable/disable the generation of the XSAVES instructions.

@item cld
@itemx no-cld
@cindex @code{target("cld")} function attribute, x86
Enable/disable the generation of the CLD before string moves.

@item fancy-math-387
@itemx no-fancy-math-387
@cindex @code{target("fancy-math-387")} function attribute, x86
Enable/disable the generation of the @code{sin}, @code{cos}, and
@code{sqrt} instructions on the 387 floating-point unit.

@item ieee-fp
@itemx no-ieee-fp
@cindex @code{target("ieee-fp")} function attribute, x86
Enable/disable the generation of floating point that depends on IEEE arithmetic.

@item inline-all-stringops
@itemx no-inline-all-stringops
@cindex @code{target("inline-all-stringops")} function attribute, x86
Enable/disable inlining of string operations.

@item inline-stringops-dynamically
@itemx no-inline-stringops-dynamically
@cindex @code{target("inline-stringops-dynamically")} function attribute, x86
Enable/disable the generation of the inline code to do small string
operations and calling the library routines for large operations.

@item align-stringops
@itemx no-align-stringops
@cindex @code{target("align-stringops")} function attribute, x86
Do/do not align destination of inlined string operations.

@item recip
@itemx no-recip
@cindex @code{target("recip")} function attribute, x86
Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and RSQRTPS
instructions followed an additional Newton-Raphson step instead of
doing a floating-point division.

@item arch=@var{ARCH}
@cindex @code{target("arch=@var{ARCH}")} function attribute, x86
Specify the architecture to generate code for in compiling the function.

@item tune=@var{TUNE}
@cindex @code{target("tune=@var{TUNE}")} function attribute, x86
Specify the architecture to tune for in compiling the function.

@item fpmath=@var{FPMATH}
@cindex @code{target("fpmath=@var{FPMATH}")} function attribute, x86
Specify which floating-point unit to use.  You must specify the
@code{target("fpmath=sse,387")} option as
@code{target("fpmath=sse+387")} because the comma would separate
different options.

@item indirect_branch("@var{choice}")
@cindex @code{indirect_branch} function attribute, x86
On x86 targets, the @code{indirect_branch} attribute causes the compiler
to convert indirect call and jump with @var{choice}.  @samp{keep}
keeps indirect call and jump unmodified.  @samp{thunk} converts indirect
call and jump to call and return thunk.  @samp{thunk-inline} converts
indirect call and jump to inlined call and return thunk.
@samp{thunk-extern} converts indirect call and jump to external call
and return thunk provided in a separate object file.

@item function_return("@var{choice}")
@cindex @code{function_return} function attribute, x86
On x86 targets, the @code{function_return} attribute causes the compiler
to convert function return with @var{choice}.  @samp{keep} keeps function
return unmodified.  @samp{thunk} converts function return to call and
return thunk.  @samp{thunk-inline} converts function return to inlined
call and return thunk.  @samp{thunk-extern} converts function return to
external call and return thunk provided in a separate object file.

@item nocf_check
@cindex @code{nocf_check} function attribute
The @code{nocf_check} attribute on a function is used to inform the
compiler that the function's prologue should not be instrumented when
compiled with the @option{-fcf-protection=branch} option.  The
compiler assumes that the function's address is a valid target for a
control-flow transfer.

The @code{nocf_check} attribute on a type of pointer to function is
used to inform the compiler that a call through the pointer should
not be instrumented when compiled with the
@option{-fcf-protection=branch} option.  The compiler assumes
that the function's address from the pointer is a valid target for
a control-flow transfer.  A direct function call through a function
name is assumed to be a safe call thus direct calls are not
instrumented by the compiler.

The @code{nocf_check} attribute is applied to an object's type.
In case of assignment of a function address or a function pointer to
another pointer, the attribute is not carried over from the right-hand
object's type; the type of left-hand object stays unchanged.  The
compiler checks for @code{nocf_check} attribute mismatch and reports
a warning in case of mismatch.

@smallexample
@{
int foo (void) __attribute__(nocf_check);
void (*foo1)(void) __attribute__(nocf_check);
void (*foo2)(void);

/* foo's address is assumed to be valid.  */
int
foo (void) 

  /* This call site is not checked for control-flow 
     validity.  */
  (*foo1)();

  /* A warning is issued about attribute mismatch.  */
  foo1 = foo2; 

  /* This call site is still not checked.  */
  (*foo1)();

  /* This call site is checked.  */
  (*foo2)();

  /* A warning is issued about attribute mismatch.  */
  foo2 = foo1; 

  /* This call site is still checked.  */
  (*foo2)();

  return 0;
@}
@end smallexample

@item cf_check
@cindex @code{cf_check} function attribute, x86

The @code{cf_check} attribute on a function is used to inform the
compiler that ENDBR instruction should be placed at the function
entry when @option{-fcf-protection=branch} is enabled.

@item indirect_return
@cindex @code{indirect_return} function attribute, x86

The @code{indirect_return} attribute can be applied to a function,
as well as variable or type of function pointer to inform the
compiler that the function may return via indirect branch.

@item fentry_name("@var{name}")
@cindex @code{fentry_name} function attribute, x86
On x86 targets, the @code{fentry_name} attribute sets the function to
call on function entry when function instrumentation is enabled
with @option{-pg -mfentry}. When @var{name} is nop then a 5 byte
nop sequence is generated.

@item fentry_section("@var{name}")
@cindex @code{fentry_section} function attribute, x86
On x86 targets, the @code{fentry_section} attribute sets the name
of the section to record function entry instrumentation calls in when
enabled with @option{-pg -mrecord-mcount}

@end table

On the x86, the inliner does not inline a
function that has different target options than the caller, unless the
callee has a subset of the target options of the caller.  For example
a function declared with @code{target("sse3")} can inline a function
with @code{target("sse2")}, since @code{-msse3} implies @code{-msse2}.
@end table

@node Xstormy16 Function Attributes
@subsection Xstormy16 Function Attributes

These function attributes are supported by the Xstormy16 back end:

@table @code
@item interrupt
@cindex @code{interrupt} function attribute, Xstormy16
Use this attribute to indicate
that the specified function is an interrupt handler.  The compiler generates
function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.
@end table

@node Variable Attributes
@section Specifying Attributes of Variables
@cindex attribute of variables
@cindex variable attributes

The keyword @code{__attribute__} allows you to specify special properties
of variables, function parameters, or structure, union, and, in C++, class
members.  This @code{__attribute__} keyword is followed by an attribute
specification enclosed in double parentheses.  Some attributes are currently
defined generically for variables.  Other attributes are defined for
variables on particular target systems.  Other attributes are available
for functions (@pxref{Function Attributes}), labels (@pxref{Label Attributes}),
enumerators (@pxref{Enumerator Attributes}), statements
(@pxref{Statement Attributes}), and for types (@pxref{Type Attributes}).
Other front ends might define more attributes
(@pxref{C++ Extensions,,Extensions to the C++ Language}).

@xref{Attribute Syntax}, for details of the exact syntax for using
attributes.

@menu
* Common Variable Attributes::
* ARC Variable Attributes::
* AVR Variable Attributes::
* Blackfin Variable Attributes::
* H8/300 Variable Attributes::
* IA-64 Variable Attributes::
* M32R/D Variable Attributes::
* MeP Variable Attributes::
* Microsoft Windows Variable Attributes::
* MSP430 Variable Attributes::
* Nvidia PTX Variable Attributes::
* PowerPC Variable Attributes::
* RL78 Variable Attributes::
* V850 Variable Attributes::
* x86 Variable Attributes::
* Xstormy16 Variable Attributes::
@end menu

@node Common Variable Attributes
@subsection Common Variable Attributes

The following attributes are supported on most targets.

@table @code

@item alias ("@var{target}")
@cindex @code{alias} variable attribute
The @code{alias} variable attribute causes the declaration to be emitted
as an alias for another symbol known as an @dfn{alias target}.  Except
for top-level qualifiers the alias target must have the same type as
the alias.  For instance, the following

@smallexample
int var_target;
extern int __attribute__ ((alias ("var_target"))) var_alias;
@end smallexample

@noindent
defines @code{var_alias} to be an alias for the @code{var_target} variable.

It is an error if the alias target is not defined in the same translation
unit as the alias.

Note that in the absence of the attribute GCC assumes that distinct
declarations with external linkage denote distinct objects.  Using both
the alias and the alias target to access the same object is undefined
in a translation unit without a declaration of the alias with the attribute.

This attribute requires assembler and object file support, and may not be
available on all targets.

@cindex @code{aligned} variable attribute
@item aligned
@itemx aligned (@var{alignment})
The @code{aligned} attribute specifies a minimum alignment for the variable
or structure field, measured in bytes.  When specified, @var{alignment} must
be an integer constant power of 2.  Specifying no @var{alignment} argument
implies the maximum alignment for the target, which is often, but by no
means always, 8 or 16 bytes.

For example, the declaration:

@smallexample
int x __attribute__ ((aligned (16))) = 0;
@end smallexample

@noindent
causes the compiler to allocate the global variable @code{x} on a
16-byte boundary.  On a 68040, this could be used in conjunction with
an @code{asm} expression to access the @code{move16} instruction which
requires 16-byte aligned operands.

You can also specify the alignment of structure fields.  For example, to
create a double-word aligned @code{int} pair, you could write:

@smallexample
struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
@end smallexample

@noindent
This is an alternative to creating a union with a @code{double} member,
which forces the union to be double-word aligned.

As in the preceding examples, you can explicitly specify the alignment
(in bytes) that you wish the compiler to use for a given variable or
structure field.  Alternatively, you can leave out the alignment factor
and just ask the compiler to align a variable or field to the
default alignment for the target architecture you are compiling for.
The default alignment is sufficient for all scalar types, but may not be
enough for all vector types on a target that supports vector operations.
The default alignment is fixed for a particular target ABI.

GCC also provides a target specific macro @code{__BIGGEST_ALIGNMENT__},
which is the largest alignment ever used for any data type on the
target machine you are compiling for.  For example, you could write:

@smallexample
short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__)));
@end smallexample

The compiler automatically sets the alignment for the declared
variable or field to @code{__BIGGEST_ALIGNMENT__}.  Doing this can
often make copy operations more efficient, because the compiler can
use whatever instructions copy the biggest chunks of memory when
performing copies to or from the variables or fields that you have
aligned this way.  Note that the value of @code{__BIGGEST_ALIGNMENT__}
may change depending on command-line options.

When used on a struct, or struct member, the @code{aligned} attribute can
only increase the alignment; in order to decrease it, the @code{packed}
attribute must be specified as well.  When used as part of a typedef, the
@code{aligned} attribute can both increase and decrease alignment, and
specifying the @code{packed} attribute generates a warning.

Note that the effectiveness of @code{aligned} attributes for static
variables may be limited by inherent limitations in the system linker
and/or object file format.  On some systems, the linker is
only able to arrange for variables to be aligned up to a certain maximum
alignment.  (For some linkers, the maximum supported alignment may
be very very small.)  If your linker is only able to align variables
up to a maximum of 8-byte alignment, then specifying @code{aligned(16)}
in an @code{__attribute__} still only provides you with 8-byte
alignment.  See your linker documentation for further information.

Stack variables are not affected by linker restrictions; GCC can properly
align them on any target.

The @code{aligned} attribute can also be used for functions
(@pxref{Common Function Attributes}.)

@cindex @code{warn_if_not_aligned} variable attribute
@item warn_if_not_aligned (@var{alignment})
This attribute specifies a threshold for the structure field, measured
in bytes.  If the structure field is aligned below the threshold, a
warning will be issued.  For example, the declaration:

@smallexample
struct foo
@{
  int i1;
  int i2;
  unsigned long long x __attribute__ ((warn_if_not_aligned (16)));
@};
@end smallexample

@noindent
causes the compiler to issue an warning on @code{struct foo}, like
@samp{warning: alignment 8 of 'struct foo' is less than 16}.
The compiler also issues a warning, like @samp{warning: 'x' offset
8 in 'struct foo' isn't aligned to 16}, when the structure field has
the misaligned offset:

@smallexample
struct __attribute__ ((aligned (16))) foo
@{
  int i1;
  int i2;
  unsigned long long x __attribute__ ((warn_if_not_aligned (16)));
@};
@end smallexample

This warning can be disabled by @option{-Wno-if-not-aligned}.
The @code{warn_if_not_aligned} attribute can also be used for types
(@pxref{Common Type Attributes}.)

@item alloc_size (@var{position})
@itemx alloc_size (@var{position-1}, @var{position-2})
@cindex @code{alloc_size} variable attribute
The @code{alloc_size} variable attribute may be applied to the declaration
of a pointer to a function that returns a pointer and takes at least one
argument of an integer type.  It indicates that the returned pointer points
to an object whose size is given by the function argument at @var{position-1},
or by the product of the arguments at @var{position-1} and @var{position-2}.
Meaningful sizes are positive values less than @code{PTRDIFF_MAX}.  Other
sizes are disagnosed when detected.  GCC uses this information to improve
the results of @code{__builtin_object_size}.

For instance, the following declarations

@smallexample
typedef __attribute__ ((alloc_size (1, 2))) void*
  (*calloc_ptr) (size_t, size_t);
typedef __attribute__ ((alloc_size (1))) void*
  (*malloc_ptr) (size_t);
@end smallexample

@noindent
specify that @code{calloc_ptr} is a pointer of a function that, like
the standard C function @code{calloc}, returns an object whose size
is given by the product of arguments 1 and 2, and similarly, that
@code{malloc_ptr}, like the standard C function @code{malloc},
returns an object whose size is given by argument 1 to the function.

@item cleanup (@var{cleanup_function})
@cindex @code{cleanup} variable attribute
The @code{cleanup} attribute runs a function when the variable goes
out of scope.  This attribute can only be applied to auto function
scope variables; it may not be applied to parameters or variables
with static storage duration.  The function must take one parameter,
a pointer to a type compatible with the variable.  The return value
of the function (if any) is ignored.

If @option{-fexceptions} is enabled, then @var{cleanup_function}
is run during the stack unwinding that happens during the
processing of the exception.  Note that the @code{cleanup} attribute
does not allow the exception to be caught, only to perform an action.
It is undefined what happens if @var{cleanup_function} does not
return normally.

@item common
@itemx nocommon
@cindex @code{common} variable attribute
@cindex @code{nocommon} variable attribute
@opindex fcommon
@opindex fno-common
The @code{common} attribute requests GCC to place a variable in
``common'' storage.  The @code{nocommon} attribute requests the
opposite---to allocate space for it directly.

These attributes override the default chosen by the
@option{-fno-common} and @option{-fcommon} flags respectively.

@item copy
@itemx copy (@var{variable})
@cindex @code{copy} variable attribute
The @code{copy} attribute applies the set of attributes with which
@var{variable} has been declared to the declaration of the variable
to which the attribute is applied.  The attribute is designed for
libraries that define aliases that are expected to specify the same
set of attributes as the aliased symbols.  The @code{copy} attribute
can be used with variables, functions or types.  However, the kind
of symbol to which the attribute is applied (either varible or
function) must match the kind of symbol to which the argument refers.
The @code{copy} attribute copies only syntactic and semantic attributes
but not attributes that affect a symbol's linkage or visibility such as
@code{alias}, @code{visibility}, or @code{weak}.  The @code{deprecated}
attribute is also not copied.  @xref{Common Function Attributes}.
@xref{Common Type Attributes}.

@item deprecated
@itemx deprecated (@var{msg})
@cindex @code{deprecated} variable attribute
The @code{deprecated} attribute results in a warning if the variable
is used anywhere in the source file.  This is useful when identifying
variables that are expected to be removed in a future version of a
program.  The warning also includes the location of the declaration
of the deprecated variable, to enable users to easily find further
information about why the variable is deprecated, or what they should
do instead.  Note that the warning only occurs for uses:

@smallexample
extern int old_var __attribute__ ((deprecated));
extern int old_var;
int new_fn () @{ return old_var; @}
@end smallexample

@noindent
results in a warning on line 3 but not line 2.  The optional @var{msg}
argument, which must be a string, is printed in the warning if
present.

The @code{deprecated} attribute can also be used for functions and
types (@pxref{Common Function Attributes},
@pxref{Common Type Attributes}).

The message attached to the attribute is affected by the setting of
the @option{-fmessage-length} option.

@item mode (@var{mode})
@cindex @code{mode} variable attribute
This attribute specifies the data type for the declaration---whichever
type corresponds to the mode @var{mode}.  This in effect lets you
request an integer or floating-point type according to its width.

@xref{Machine Modes,,, gccint, GNU Compiler Collection (GCC) Internals},
for a list of the possible keywords for @var{mode}.
You may also specify a mode of @code{byte} or @code{__byte__} to
indicate the mode corresponding to a one-byte integer, @code{word} or
@code{__word__} for the mode of a one-word integer, and @code{pointer}
or @code{__pointer__} for the mode used to represent pointers.

@item nonstring
@cindex @code{nonstring} variable attribute
The @code{nonstring} variable attribute specifies that an object or member
declaration with type array of @code{char}, @code{signed char}, or
@code{unsigned char}, or pointer to such a type is intended to store
character arrays that do not necessarily contain a terminating @code{NUL}.
This is useful in detecting uses of such arrays or pointers with functions
that expect @code{NUL}-terminated strings, and to avoid warnings when such
an array or pointer is used as an argument to a bounded string manipulation
function such as @code{strncpy}.  For example, without the attribute, GCC
will issue a warning for the @code{strncpy} call below because it may
truncate the copy without appending the terminating @code{NUL} character.
Using the attribute makes it possible to suppress the warning.  However,
when the array is declared with the attribute the call to @code{strlen} is
diagnosed because when the array doesn't contain a @code{NUL}-terminated
string the call is undefined.  To copy, compare, of search non-string
character arrays use the @code{memcpy}, @code{memcmp}, @code{memchr},
and other functions that operate on arrays of bytes.  In addition,
calling @code{strnlen} and @code{strndup} with such arrays is safe
provided a suitable bound is specified, and not diagnosed.

@smallexample
struct Data
@{
  char name [32] __attribute__ ((nonstring));
@};

int f (struct Data *pd, const char *s)
@{
  strncpy (pd->name, s, sizeof pd->name);
  @dots{}
  return strlen (pd->name);   // unsafe, gets a warning
@}
@end smallexample

@item packed
@cindex @code{packed} variable attribute
The @code{packed} attribute specifies that a structure member should have
the smallest possible alignment---one bit for a bit-field and one byte
otherwise, unless a larger value is specified with the @code{aligned}
attribute.  The attribute does not apply to non-member objects.

For example in the structure below, the member array @code{x} is packed
so that it immediately follows @code{a} with no intervening padding:

@smallexample
struct foo
@{
  char a;
  int x[2] __attribute__ ((packed));
@};
@end smallexample

@emph{Note:} The 4.1, 4.2 and 4.3 series of GCC ignore the
@code{packed} attribute on bit-fields of type @code{char}.  This has
been fixed in GCC 4.4 but the change can lead to differences in the
structure layout.  See the documentation of
@option{-Wpacked-bitfield-compat} for more information.

@item section ("@var{section-name}")
@cindex @code{section} variable attribute
Normally, the compiler places the objects it generates in sections like
@code{data} and @code{bss}.  Sometimes, however, you need additional sections,
or you need certain particular variables to appear in special sections,
for example to map to special hardware.  The @code{section}
attribute specifies that a variable (or function) lives in a particular
section.  For example, this small program uses several specific section names:

@smallexample
struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
int init_data __attribute__ ((section ("INITDATA")));

main()
@{
  /* @r{Initialize stack pointer} */
  init_sp (stack + sizeof (stack));

  /* @r{Initialize initialized data} */
  memcpy (&init_data, &data, &edata - &data);

  /* @r{Turn on the serial ports} */
  init_duart (&a);
  init_duart (&b);
@}
@end smallexample

@noindent
Use the @code{section} attribute with
@emph{global} variables and not @emph{local} variables,
as shown in the example.

You may use the @code{section} attribute with initialized or
uninitialized global variables but the linker requires
each object be defined once, with the exception that uninitialized
variables tentatively go in the @code{common} (or @code{bss}) section
and can be multiply ``defined''.  Using the @code{section} attribute
changes what section the variable goes into and may cause the
linker to issue an error if an uninitialized variable has multiple
definitions.  You can force a variable to be initialized with the
@option{-fno-common} flag or the @code{nocommon} attribute.

Some file formats do not support arbitrary sections so the @code{section}
attribute is not available on all platforms.
If you need to map the entire contents of a module to a particular
section, consider using the facilities of the linker instead.

@item tls_model ("@var{tls_model}")
@cindex @code{tls_model} variable attribute
The @code{tls_model} attribute sets thread-local storage model
(@pxref{Thread-Local}) of a particular @code{__thread} variable,
overriding @option{-ftls-model=} command-line switch on a per-variable
basis.
The @var{tls_model} argument should be one of @code{global-dynamic},
@code{local-dynamic}, @code{initial-exec} or @code{local-exec}.

Not all targets support this attribute.

@item unused
@cindex @code{unused} variable attribute
This attribute, attached to a variable, means that the variable is meant
to be possibly unused.  GCC does not produce a warning for this
variable.

@item used
@cindex @code{used} variable attribute
This attribute, attached to a variable with static storage, means that
the variable must be emitted even if it appears that the variable is not
referenced.

When applied to a static data member of a C++ class template, the
attribute also means that the member is instantiated if the
class itself is instantiated.

@item vector_size (@var{bytes})
@cindex @code{vector_size} variable attribute
This attribute specifies the vector size for the type of the declared
variable, measured in bytes.  The type to which it applies is known as
the @dfn{base type}.  The @var{bytes} argument must be a positive
power-of-two multiple of the base type size.  For example, the declaration:

@smallexample
int foo __attribute__ ((vector_size (16)));
@end smallexample

@noindent
causes the compiler to set the mode for @code{foo}, to be 16 bytes,
divided into @code{int} sized units.  Assuming a 32-bit @code{int},
@code{foo}'s type is a vector of four units of four bytes each, and
the corresponding mode of @code{foo} is @code{V4SI}.
@xref{Vector Extensions}, for details of manipulating vector variables.

This attribute is only applicable to integral and floating scalars,
although arrays, pointers, and function return values are allowed in
conjunction with this construct.

Aggregates with this attribute are invalid, even if they are of the same
size as a corresponding scalar.  For example, the declaration:

@smallexample
struct S @{ int a; @};
struct S  __attribute__ ((vector_size (16))) foo;
@end smallexample

@noindent
is invalid even if the size of the structure is the same as the size of
the @code{int}.

@item visibility ("@var{visibility_type}")
@cindex @code{visibility} variable attribute
This attribute affects the linkage of the declaration to which it is attached.
The @code{visibility} attribute is described in
@ref{Common Function Attributes}.

@item weak
@cindex @code{weak} variable attribute
The @code{weak} attribute is described in
@ref{Common Function Attributes}.

@item noinit
@cindex @code{noinit} variable attribute
Any data with the @code{noinit} attribute will not be initialized by
the C runtime startup code, or the program loader.  Not initializing
data in this way can reduce program startup times.  This attribute is
specific to ELF targets and relies on the linker to place such data in
the right location

@end table

@node ARC Variable Attributes
@subsection ARC Variable Attributes

@table @code
@item aux
@cindex @code{aux} variable attribute, ARC
The @code{aux} attribute is used to directly access the ARC's
auxiliary register space from C.  The auxilirary register number is
given via attribute argument.

@end table

@node AVR Variable Attributes
@subsection AVR Variable Attributes

@table @code
@item progmem
@cindex @code{progmem} variable attribute, AVR
The @code{progmem} attribute is used on the AVR to place read-only
data in the non-volatile program memory (flash). The @code{progmem}
attribute accomplishes this by putting respective variables into a
section whose name starts with @code{.progmem}.

This attribute works similar to the @code{section} attribute
but adds additional checking.

@table @asis
@item @bullet{}@tie{} Ordinary AVR cores with 32 general purpose registers:
@code{progmem} affects the location
of the data but not how this data is accessed.
In order to read data located with the @code{progmem} attribute
(inline) assembler must be used.
@smallexample
/* Use custom macros from @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC}} */
#include <avr/pgmspace.h> 

/* Locate var in flash memory */
const int var[2] PROGMEM = @{ 1, 2 @};

int read_var (int i)
@{
    /* Access var[] by accessor macro from avr/pgmspace.h */
    return (int) pgm_read_word (& var[i]);
@}
@end smallexample

AVR is a Harvard architecture processor and data and read-only data
normally resides in the data memory (RAM).

See also the @ref{AVR Named Address Spaces} section for
an alternate way to locate and access data in flash memory.

@item @bullet{}@tie{} AVR cores with flash memory visible in the RAM address range:
On such devices, there is no need for attribute @code{progmem} or
@ref{AVR Named Address Spaces,,@code{__flash}} qualifier at all.
Just use standard C / C++.  The compiler will generate @code{LD*}
instructions.  As flash memory is visible in the RAM address range,
and the default linker script does @emph{not} locate @code{.rodata} in
RAM, no special features are needed in order not to waste RAM for
read-only data or to read from flash.  You might even get slightly better
performance by
avoiding @code{progmem} and @code{__flash}.  This applies to devices from
families @code{avrtiny} and @code{avrxmega3}, see @ref{AVR Options} for
an overview.

@item @bullet{}@tie{}Reduced AVR Tiny cores like ATtiny40:
The compiler adds @code{0x4000}
to the addresses of objects and declarations in @code{progmem} and locates
the objects in flash memory, namely in section @code{.progmem.data}.
The offset is needed because the flash memory is visible in the RAM
address space starting at address @code{0x4000}.

Data in @code{progmem} can be accessed by means of ordinary C@tie{}code,
no special functions or macros are needed.

@smallexample
/* var is located in flash memory */
extern const int var[2] __attribute__((progmem));

int read_var (int i)
@{
    return var[i];
@}
@end smallexample

Please notice that on these devices, there is no need for @code{progmem}
at all.

@end table

@item io
@itemx io (@var{addr})
@cindex @code{io} variable attribute, AVR
Variables with the @code{io} attribute are used to address
memory-mapped peripherals in the io address range.
If an address is specified, the variable
is assigned that address, and the value is interpreted as an
address in the data address space.
Example:

@smallexample
volatile int porta __attribute__((io (0x22)));
@end smallexample

The address specified in the address in the data address range.

Otherwise, the variable it is not assigned an address, but the
compiler will still use in/out instructions where applicable,
assuming some other module assigns an address in the io address range.
Example:

@smallexample
extern volatile int porta __attribute__((io));
@end smallexample

@item io_low
@itemx io_low (@var{addr})
@cindex @code{io_low} variable attribute, AVR
This is like the @code{io} attribute, but additionally it informs the
compiler that the object lies in the lower half of the I/O area,
allowing the use of @code{cbi}, @code{sbi}, @code{sbic} and @code{sbis}
instructions.

@item address
@itemx address (@var{addr})
@cindex @code{address} variable attribute, AVR
Variables with the @code{address} attribute are used to address
memory-mapped peripherals that may lie outside the io address range.

@smallexample
volatile int porta __attribute__((address (0x600)));
@end smallexample

@item absdata
@cindex @code{absdata} variable attribute, AVR
Variables in static storage and with the @code{absdata} attribute can
be accessed by the @code{LDS} and @code{STS} instructions which take
absolute addresses.

@itemize @bullet
@item
This attribute is only supported for the reduced AVR Tiny core
like ATtiny40.

@item
You must make sure that respective data is located in the
address range @code{0x40}@dots{}@code{0xbf} accessible by
@code{LDS} and @code{STS}.  One way to achieve this as an
appropriate linker description file.

@item
If the location does not fit the address range of @code{LDS}
and @code{STS}, there is currently (Binutils 2.26) just an unspecific
warning like
@quotation
@code{module.c:(.text+0x1c): warning: internal error: out of range error}
@end quotation

@end itemize

See also the @option{-mabsdata} @ref{AVR Options,command-line option}.

@end table

@node Blackfin Variable Attributes
@subsection Blackfin Variable Attributes

Three attributes are currently defined for the Blackfin.

@table @code
@item l1_data
@itemx l1_data_A
@itemx l1_data_B
@cindex @code{l1_data} variable attribute, Blackfin
@cindex @code{l1_data_A} variable attribute, Blackfin
@cindex @code{l1_data_B} variable attribute, Blackfin
Use these attributes on the Blackfin to place the variable into L1 Data SRAM.
Variables with @code{l1_data} attribute are put into the specific section
named @code{.l1.data}. Those with @code{l1_data_A} attribute are put into
the specific section named @code{.l1.data.A}. Those with @code{l1_data_B}
attribute are put into the specific section named @code{.l1.data.B}.

@item l2
@cindex @code{l2} variable attribute, Blackfin
Use this attribute on the Blackfin to place the variable into L2 SRAM.
Variables with @code{l2} attribute are put into the specific section
named @code{.l2.data}.
@end table

@node H8/300 Variable Attributes
@subsection H8/300 Variable Attributes

These variable attributes are available for H8/300 targets:

@table @code
@item eightbit_data
@cindex @code{eightbit_data} variable attribute, H8/300
@cindex eight-bit data on the H8/300, H8/300H, and H8S
Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
variable should be placed into the eight-bit data section.
The compiler generates more efficient code for certain operations
on data in the eight-bit data area.  Note the eight-bit data area is limited to
256 bytes of data.

You must use GAS and GLD from GNU binutils version 2.7 or later for
this attribute to work correctly.

@item tiny_data
@cindex @code{tiny_data} variable attribute, H8/300
@cindex tiny data section on the H8/300H and H8S
Use this attribute on the H8/300H and H8S to indicate that the specified
variable should be placed into the tiny data section.
The compiler generates more efficient code for loads and stores
on data in the tiny data section.  Note the tiny data area is limited to
slightly under 32KB of data.

@end table

@node IA-64 Variable Attributes
@subsection IA-64 Variable Attributes

The IA-64 back end supports the following variable attribute:

@table @code
@item model (@var{model-name})
@cindex @code{model} variable attribute, IA-64

On IA-64, use this attribute to set the addressability of an object.
At present, the only supported identifier for @var{model-name} is
@code{small}, indicating addressability via ``small'' (22-bit)
addresses (so that their addresses can be loaded with the @code{addl}
instruction).  Caveat: such addressing is by definition not position
independent and hence this attribute must not be used for objects
defined by shared libraries.

@end table

@node M32R/D Variable Attributes
@subsection M32R/D Variable Attributes

One attribute is currently defined for the M32R/D@.

@table @code
@item model (@var{model-name})
@cindex @code{model-name} variable attribute, M32R/D
@cindex variable addressability on the M32R/D
Use this attribute on the M32R/D to set the addressability of an object.
The identifier @var{model-name} is one of @code{small}, @code{medium},
or @code{large}, representing each of the code models.

Small model objects live in the lower 16MB of memory (so that their
addresses can be loaded with the @code{ld24} instruction).

Medium and large model objects may live anywhere in the 32-bit address space
(the compiler generates @code{seth/add3} instructions to load their
addresses).
@end table

@node MeP Variable Attributes
@subsection MeP Variable Attributes

The MeP target has a number of addressing modes and busses.  The
@code{near} space spans the standard memory space's first 16 megabytes
(24 bits).  The @code{far} space spans the entire 32-bit memory space.
The @code{based} space is a 128-byte region in the memory space that
is addressed relative to the @code{$tp} register.  The @code{tiny}
space is a 65536-byte region relative to the @code{$gp} register.  In
addition to these memory regions, the MeP target has a separate 16-bit
control bus which is specified with @code{cb} attributes.

@table @code

@item based
@cindex @code{based} variable attribute, MeP
Any variable with the @code{based} attribute is assigned to the
@code{.based} section, and is accessed with relative to the
@code{$tp} register.

@item tiny
@cindex @code{tiny} variable attribute, MeP
Likewise, the @code{tiny} attribute assigned variables to the
@code{.tiny} section, relative to the @code{$gp} register.

@item near
@cindex @code{near} variable attribute, MeP
Variables with the @code{near} attribute are assumed to have addresses
that fit in a 24-bit addressing mode.  This is the default for large
variables (@code{-mtiny=4} is the default) but this attribute can
override @code{-mtiny=} for small variables, or override @code{-ml}.

@item far
@cindex @code{far} variable attribute, MeP
Variables with the @code{far} attribute are addressed using a full
32-bit address.  Since this covers the entire memory space, this
allows modules to make no assumptions about where variables might be
stored.

@item io
@cindex @code{io} variable attribute, MeP
@itemx io (@var{addr})
Variables with the @code{io} attribute are used to address
memory-mapped peripherals.  If an address is specified, the variable
is assigned that address, else it is not assigned an address (it is
assumed some other module assigns an address).  Example:

@smallexample
int timer_count __attribute__((io(0x123)));
@end smallexample

@item cb
@itemx cb (@var{addr})
@cindex @code{cb} variable attribute, MeP
Variables with the @code{cb} attribute are used to access the control
bus, using special instructions.  @code{addr} indicates the control bus
address.  Example:

@smallexample
int cpu_clock __attribute__((cb(0x123)));
@end smallexample

@end table

@node Microsoft Windows Variable Attributes
@subsection Microsoft Windows Variable Attributes

You can use these attributes on Microsoft Windows targets.
@ref{x86 Variable Attributes} for additional Windows compatibility
attributes available on all x86 targets.

@table @code
@item dllimport
@itemx dllexport
@cindex @code{dllimport} variable attribute
@cindex @code{dllexport} variable attribute
The @code{dllimport} and @code{dllexport} attributes are described in
@ref{Microsoft Windows Function Attributes}.

@item selectany
@cindex @code{selectany} variable attribute
The @code{selectany} attribute causes an initialized global variable to
have link-once semantics.  When multiple definitions of the variable are
encountered by the linker, the first is selected and the remainder are
discarded.  Following usage by the Microsoft compiler, the linker is told
@emph{not} to warn about size or content differences of the multiple
definitions.

Although the primary usage of this attribute is for POD types, the
attribute can also be applied to global C++ objects that are initialized
by a constructor.  In this case, the static initialization and destruction
code for the object is emitted in each translation defining the object,
but the calls to the constructor and destructor are protected by a
link-once guard variable.

The @code{selectany} attribute is only available on Microsoft Windows
targets.  You can use @code{__declspec (selectany)} as a synonym for
@code{__attribute__ ((selectany))} for compatibility with other
compilers.

@item shared
@cindex @code{shared} variable attribute
On Microsoft Windows, in addition to putting variable definitions in a named
section, the section can also be shared among all running copies of an
executable or DLL@.  For example, this small program defines shared data
by putting it in a named section @code{shared} and marking the section
shareable:

@smallexample
int foo __attribute__((section ("shared"), shared)) = 0;

int
main()
@{
  /* @r{Read and write foo.  All running
     copies see the same value.}  */
  return 0;
@}
@end smallexample

@noindent
You may only use the @code{shared} attribute along with @code{section}
attribute with a fully-initialized global definition because of the way
linkers work.  See @code{section} attribute for more information.

The @code{shared} attribute is only available on Microsoft Windows@.

@end table

@node MSP430 Variable Attributes
@subsection MSP430 Variable Attributes

@table @code
@item noinit
@cindex @code{noinit} variable attribute, MSP430 
Any data with the @code{noinit} attribute will not be initialised by
the C runtime startup code, or the program loader.  Not initialising
data in this way can reduce program startup times.

@item persistent
@cindex @code{persistent} variable attribute, MSP430 
Any variable with the @code{persistent} attribute will not be
initialised by the C runtime startup code.  Instead its value will be
set once, when the application is loaded, and then never initialised
again, even if the processor is reset or the program restarts.
Persistent data is intended to be placed into FLASH RAM, where its
value will be retained across resets.  The linker script being used to
create the application should ensure that persistent data is correctly
placed.

@item upper
@itemx either
@cindex @code{upper} variable attribute, MSP430 
@cindex @code{either} variable attribute, MSP430 
These attributes are the same as the MSP430 function attributes of the
same name (@pxref{MSP430 Function Attributes}).  

@item lower
@cindex @code{lower} variable attribute, MSP430
This option behaves mostly the same as the MSP430 function attribute of the
same name (@pxref{MSP430 Function Attributes}), but it has some additional
functionality.

If @option{-mdata-region=}@{@code{upper,either,none}@} has been passed, or
the @code{section} attribute is applied to a variable, the compiler will
generate 430X instructions to handle it.  This is because the compiler has
to assume that the variable could get placed in the upper memory region
(above address 0xFFFF).  Marking the variable with the @code{lower} attribute
informs the compiler that the variable will be placed in lower memory so it
is safe to use 430 instructions to handle it.

In the case of the @code{section} attribute, the section name given
will be used, and the @code{.lower} prefix will not be added.

@end table

@node Nvidia PTX Variable Attributes
@subsection Nvidia PTX Variable Attributes

These variable attributes are supported by the Nvidia PTX back end:

@table @code
@item shared
@cindex @code{shared} attribute, Nvidia PTX
Use this attribute to place a variable in the @code{.shared} memory space.
This memory space is private to each cooperative thread array; only threads
within one thread block refer to the same instance of the variable.
The runtime does not initialize variables in this memory space.
@end table

@node PowerPC Variable Attributes
@subsection PowerPC Variable Attributes

Three attributes currently are defined for PowerPC configurations:
@code{altivec}, @code{ms_struct} and @code{gcc_struct}.

@cindex @code{ms_struct} variable attribute, PowerPC
@cindex @code{gcc_struct} variable attribute, PowerPC
For full documentation of the struct attributes please see the
documentation in @ref{x86 Variable Attributes}.

@cindex @code{altivec} variable attribute, PowerPC
For documentation of @code{altivec} attribute please see the
documentation in @ref{PowerPC Type Attributes}.

@node RL78 Variable Attributes
@subsection RL78 Variable Attributes

@cindex @code{saddr} variable attribute, RL78
The RL78 back end supports the @code{saddr} variable attribute.  This
specifies placement of the corresponding variable in the SADDR area,
which can be accessed more efficiently than the default memory region.

@node V850 Variable Attributes
@subsection V850 Variable Attributes

These variable attributes are supported by the V850 back end:

@table @code

@item sda
@cindex @code{sda} variable attribute, V850
Use this attribute to explicitly place a variable in the small data area,
which can hold up to 64 kilobytes.

@item tda
@cindex @code{tda} variable attribute, V850
Use this attribute to explicitly place a variable in the tiny data area,
which can hold up to 256 bytes in total.

@item zda
@cindex @code{zda} variable attribute, V850
Use this attribute to explicitly place a variable in the first 32 kilobytes
of memory.
@end table

@node x86 Variable Attributes
@subsection x86 Variable Attributes

Two attributes are currently defined for x86 configurations:
@code{ms_struct} and @code{gcc_struct}.

@table @code
@item ms_struct
@itemx gcc_struct
@cindex @code{ms_struct} variable attribute, x86
@cindex @code{gcc_struct} variable attribute, x86

If @code{packed} is used on a structure, or if bit-fields are used,
it may be that the Microsoft ABI lays out the structure differently
than the way GCC normally does.  Particularly when moving packed
data between functions compiled with GCC and the native Microsoft compiler
(either via function call or as data in a file), it may be necessary to access
either format.

The @code{ms_struct} and @code{gcc_struct} attributes correspond
to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
command-line options, respectively;
see @ref{x86 Options}, for details of how structure layout is affected.
@xref{x86 Type Attributes}, for information about the corresponding
attributes on types.

@end table

@node Xstormy16 Variable Attributes
@subsection Xstormy16 Variable Attributes

One attribute is currently defined for xstormy16 configurations:
@code{below100}.

@table @code
@item below100
@cindex @code{below100} variable attribute, Xstormy16

If a variable has the @code{below100} attribute (@code{BELOW100} is
allowed also), GCC places the variable in the first 0x100 bytes of
memory and use special opcodes to access it.  Such variables are
placed in either the @code{.bss_below100} section or the
@code{.data_below100} section.

@end table

@node Type Attributes
@section Specifying Attributes of Types
@cindex attribute of types
@cindex type attributes

The keyword @code{__attribute__} allows you to specify various special
properties of types.  Some type attributes apply only to structure and
union types, and in C++, also class types, while others can apply to
any type defined via a @code{typedef} declaration.  Unless otherwise
specified, the same restrictions and effects apply to attributes regardless
of whether a type is a trivial structure or a C++ class with user-defined
constructors, destructors, or a copy assignment.

Other attributes are defined for functions (@pxref{Function Attributes}),
labels (@pxref{Label  Attributes}), enumerators (@pxref{Enumerator
Attributes}), statements (@pxref{Statement Attributes}), and for variables
(@pxref{Variable Attributes}).

The @code{__attribute__} keyword is followed by an attribute specification
enclosed in double parentheses.

You may specify type attributes in an enum, struct or union type
declaration or definition by placing them immediately after the
@code{struct}, @code{union} or @code{enum} keyword.  You can also place
them just past the closing curly brace of the definition, but this is less
preferred because logically the type should be fully defined at 
the closing brace.

You can also include type attributes in a @code{typedef} declaration.
@xref{Attribute Syntax}, for details of the exact syntax for using
attributes.

@menu
* Common Type Attributes::
* ARC Type Attributes::
* ARM Type Attributes::
* MeP Type Attributes::
* PowerPC Type Attributes::
* x86 Type Attributes::
@end menu

@node Common Type Attributes
@subsection Common Type Attributes

The following type attributes are supported on most targets.

@table @code
@cindex @code{aligned} type attribute
@item aligned
@itemx aligned (@var{alignment})
The @code{aligned} attribute specifies a minimum alignment (in bytes) for
variables of the specified type.  When specified, @var{alignment} must be
a power of 2.  Specifying no @var{alignment} argument implies the maximum
alignment for the target, which is often, but by no means always, 8 or 16
bytes.  For example, the declarations:

@smallexample
struct __attribute__ ((aligned (8))) S @{ short f[3]; @};
typedef int more_aligned_int __attribute__ ((aligned (8)));
@end smallexample

@noindent
force the compiler to ensure (as far as it can) that each variable whose
type is @code{struct S} or @code{more_aligned_int} is allocated and
aligned @emph{at least} on a 8-byte boundary.  On a SPARC, having all
variables of type @code{struct S} aligned to 8-byte boundaries allows
the compiler to use the @code{ldd} and @code{std} (doubleword load and
store) instructions when copying one variable of type @code{struct S} to
another, thus improving run-time efficiency.

Note that the alignment of any given @code{struct} or @code{union} type
is required by the ISO C standard to be at least a perfect multiple of
the lowest common multiple of the alignments of all of the members of
the @code{struct} or @code{union} in question.  This means that you @emph{can}
effectively adjust the alignment of a @code{struct} or @code{union}
type by attaching an @code{aligned} attribute to any one of the members
of such a type, but the notation illustrated in the example above is a
more obvious, intuitive, and readable way to request the compiler to
adjust the alignment of an entire @code{struct} or @code{union} type.

As in the preceding example, you can explicitly specify the alignment
(in bytes) that you wish the compiler to use for a given @code{struct}
or @code{union} type.  Alternatively, you can leave out the alignment factor
and just ask the compiler to align a type to the maximum
useful alignment for the target machine you are compiling for.  For
example, you could write:

@smallexample
struct __attribute__ ((aligned)) S @{ short f[3]; @};
@end smallexample

Whenever you leave out the alignment factor in an @code{aligned}
attribute specification, the compiler automatically sets the alignment
for the type to the largest alignment that is ever used for any data
type on the target machine you are compiling for.  Doing this can often
make copy operations more efficient, because the compiler can use
whatever instructions copy the biggest chunks of memory when performing
copies to or from the variables that have types that you have aligned
this way.

In the example above, if the size of each @code{short} is 2 bytes, then
the size of the entire @code{struct S} type is 6 bytes.  The smallest
power of two that is greater than or equal to that is 8, so the
compiler sets the alignment for the entire @code{struct S} type to 8
bytes.

Note that although you can ask the compiler to select a time-efficient
alignment for a given type and then declare only individual stand-alone
objects of that type, the compiler's ability to select a time-efficient
alignment is primarily useful only when you plan to create arrays of
variables having the relevant (efficiently aligned) type.  If you
declare or use arrays of variables of an efficiently-aligned type, then
it is likely that your program also does pointer arithmetic (or
subscripting, which amounts to the same thing) on pointers to the
relevant type, and the code that the compiler generates for these
pointer arithmetic operations is often more efficient for
efficiently-aligned types than for other types.

Note that the effectiveness of @code{aligned} attributes may be limited
by inherent limitations in your linker.  On many systems, the linker is
only able to arrange for variables to be aligned up to a certain maximum
alignment.  (For some linkers, the maximum supported alignment may
be very very small.)  If your linker is only able to align variables
up to a maximum of 8-byte alignment, then specifying @code{aligned (16)}
in an @code{__attribute__} still only provides you with 8-byte
alignment.  See your linker documentation for further information.

When used on a struct, or struct member, the @code{aligned} attribute can
only increase the alignment; in order to decrease it, the @code{packed}
attribute must be specified as well.  When used as part of a typedef, the
@code{aligned} attribute can both increase and decrease alignment, and
specifying the @code{packed} attribute generates a warning.

@cindex @code{warn_if_not_aligned} type attribute
@item warn_if_not_aligned (@var{alignment})
This attribute specifies a threshold for the structure field, measured
in bytes.  If the structure field is aligned below the threshold, a
warning will be issued.  For example, the declaration:

@smallexample
typedef unsigned long long __u64
   __attribute__((aligned (4), warn_if_not_aligned (8)));

struct foo
@{
  int i1;
  int i2;
  __u64 x;
@};
@end smallexample

@noindent
causes the compiler to issue an warning on @code{struct foo}, like
@samp{warning: alignment 4 of 'struct foo' is less than 8}.
It is used to define @code{struct foo} in such a way that
@code{struct foo} has the same layout and the structure field @code{x}
has the same alignment when @code{__u64} is aligned at either 4 or
8 bytes.  Align @code{struct foo} to 8 bytes:

@smallexample
struct __attribute__ ((aligned (8))) foo
@{
  int i1;
  int i2;
  __u64 x;
@};
@end smallexample

@noindent
silences the warning.  The compiler also issues a warning, like
@samp{warning: 'x' offset 12 in 'struct foo' isn't aligned to 8},
when the structure field has the misaligned offset:

@smallexample
struct __attribute__ ((aligned (8))) foo
@{
  int i1;
  int i2;
  int i3;
  __u64 x;
@};
@end smallexample

This warning can be disabled by @option{-Wno-if-not-aligned}.

@item alloc_size (@var{position})
@itemx alloc_size (@var{position-1}, @var{position-2})
@cindex @code{alloc_size} type attribute
The @code{alloc_size} type attribute may be applied to the definition
of a type of a function that returns a pointer and takes at least one
argument of an integer type.  It indicates that the returned pointer
points to an object whose size is given by the function argument at
@var{position-1}, or by the product of the arguments at @var{position-1}
and @var{position-2}.  Meaningful sizes are positive values less than
@code{PTRDIFF_MAX}.  Other sizes are disagnosed when detected.  GCC uses
this information to improve the results of @code{__builtin_object_size}.

For instance, the following declarations

@smallexample
typedef __attribute__ ((alloc_size (1, 2))) void*
  calloc_type (size_t, size_t);
typedef __attribute__ ((alloc_size (1))) void*
  malloc_type (size_t);
@end smallexample

@noindent
specify that @code{calloc_type} is a type of a function that, like
the standard C function @code{calloc}, returns an object whose size
is given by the product of arguments 1 and 2, and that
@code{malloc_type}, like the standard C function @code{malloc},
returns an object whose size is given by argument 1 to the function.

@item copy
@itemx copy (@var{expression})
@cindex @code{copy} type attribute
The @code{copy} attribute applies the set of attributes with which
the type of the @var{expression} has been declared to the declaration
of the type to which the attribute is applied.  The attribute is
designed for libraries that define aliases that are expected to
specify the same set of attributes as the aliased symbols.
The @code{copy} attribute can be used with types, variables, or
functions.  However, the kind of symbol to which the attribute is
applied (either varible or function) must match the kind of symbol
to which the argument refers.
The @code{copy} attribute copies only syntactic and semantic attributes
but not attributes that affect a symbol's linkage or visibility such as
@code{alias}, @code{visibility}, or @code{weak}.  The @code{deprecated}
attribute is also not copied.  @xref{Common Function Attributes}.
@xref{Common Variable Attributes}.

For example, suppose @code{struct A} below is defined in some third
party library header to have the alignment requirement @code{N} and
to force a warning whenever a variable of the type is not so aligned
due to attribute @code{packed}.  Specifying the @code{copy} attribute
on the definition on the unrelated @code{struct B} has the effect of
copying all relevant attributes from the type referenced by the pointer
expression to @code{struct B}.

@smallexample
struct __attribute__ ((aligned (N), warn_if_not_aligned (N)))
A @{ /* @r{@dots{}} */ @};
struct __attribute__ ((copy ( (struct A *)0)) B @{ /* @r{@dots{}} */ @};
@end smallexample

@item deprecated
@itemx deprecated (@var{msg})
@cindex @code{deprecated} type attribute
The @code{deprecated} attribute results in a warning if the type
is used anywhere in the source file.  This is useful when identifying
types that are expected to be removed in a future version of a program.
If possible, the warning also includes the location of the declaration
of the deprecated type, to enable users to easily find further
information about why the type is deprecated, or what they should do
instead.  Note that the warnings only occur for uses and then only
if the type is being applied to an identifier that itself is not being
declared as deprecated.

@smallexample
typedef int T1 __attribute__ ((deprecated));
T1 x;
typedef T1 T2;
T2 y;
typedef T1 T3 __attribute__ ((deprecated));
T3 z __attribute__ ((deprecated));
@end smallexample

@noindent
results in a warning on line 2 and 3 but not lines 4, 5, or 6.  No
warning is issued for line 4 because T2 is not explicitly
deprecated.  Line 5 has no warning because T3 is explicitly
deprecated.  Similarly for line 6.  The optional @var{msg}
argument, which must be a string, is printed in the warning if
present.  Control characters in the string will be replaced with
escape sequences, and if the @option{-fmessage-length} option is set
to 0 (its default value) then any newline characters will be ignored.

The @code{deprecated} attribute can also be used for functions and
variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)

The message attached to the attribute is affected by the setting of
the @option{-fmessage-length} option.

@item designated_init
@cindex @code{designated_init} type attribute
This attribute may only be applied to structure types.  It indicates
that any initialization of an object of this type must use designated
initializers rather than positional initializers.  The intent of this
attribute is to allow the programmer to indicate that a structure's
layout may change, and that therefore relying on positional
initialization will result in future breakage.

GCC emits warnings based on this attribute by default; use
@option{-Wno-designated-init} to suppress them.

@item may_alias
@cindex @code{may_alias} type attribute
Accesses through pointers to types with this attribute are not subject
to type-based alias analysis, but are instead assumed to be able to alias
any other type of objects.
In the context of section 6.5 paragraph 7 of the C99 standard,
an lvalue expression
dereferencing such a pointer is treated like having a character type.
See @option{-fstrict-aliasing} for more information on aliasing issues.
This extension exists to support some vector APIs, in which pointers to
one vector type are permitted to alias pointers to a different vector type.

Note that an object of a type with this attribute does not have any
special semantics.

Example of use:

@smallexample
typedef short __attribute__ ((__may_alias__)) short_a;

int
main (void)
@{
  int a = 0x12345678;
  short_a *b = (short_a *) &a;

  b[1] = 0;

  if (a == 0x12345678)
    abort();

  exit(0);
@}
@end smallexample

@noindent
If you replaced @code{short_a} with @code{short} in the variable
declaration, the above program would abort when compiled with
@option{-fstrict-aliasing}, which is on by default at @option{-O2} or
above.

@item mode (@var{mode})
@cindex @code{mode} type attribute
This attribute specifies the data type for the declaration---whichever
type corresponds to the mode @var{mode}.  This in effect lets you
request an integer or floating-point type according to its width.

@xref{Machine Modes,,, gccint, GNU Compiler Collection (GCC) Internals},
for a list of the possible keywords for @var{mode}.
You may also specify a mode of @code{byte} or @code{__byte__} to
indicate the mode corresponding to a one-byte integer, @code{word} or
@code{__word__} for the mode of a one-word integer, and @code{pointer}
or @code{__pointer__} for the mode used to represent pointers.

@item packed
@cindex @code{packed} type attribute
This attribute, attached to a @code{struct}, @code{union}, or C++ @code{class}
type definition, specifies that each of its members (other than zero-width
bit-fields) is placed to minimize the memory required.  This is equivalent
to specifying the @code{packed} attribute on each of the members.

@opindex fshort-enums
When attached to an @code{enum} definition, the @code{packed} attribute
indicates that the smallest integral type should be used.
Specifying the @option{-fshort-enums} flag on the command line
is equivalent to specifying the @code{packed}
attribute on all @code{enum} definitions.

In the following example @code{struct my_packed_struct}'s members are
packed closely together, but the internal layout of its @code{s} member
is not packed---to do that, @code{struct my_unpacked_struct} needs to
be packed too.

@smallexample
struct my_unpacked_struct
 @{
    char c;
    int i;
 @};

struct __attribute__ ((__packed__)) my_packed_struct
  @{
     char c;
     int  i;
     struct my_unpacked_struct s;
  @};
@end smallexample

You may only specify the @code{packed} attribute on the definition
of an @code{enum}, @code{struct}, @code{union}, or @code{class}, 
not on a @code{typedef} that does not also define the enumerated type,
structure, union, or class.

@item scalar_storage_order ("@var{endianness}")
@cindex @code{scalar_storage_order} type attribute
When attached to a @code{union} or a @code{struct}, this attribute sets
the storage order, aka endianness, of the scalar fields of the type, as
well as the array fields whose component is scalar.  The supported
endiannesses are @code{big-endian} and @code{little-endian}.  The attribute
has no effects on fields which are themselves a @code{union}, a @code{struct}
or an array whose component is a @code{union} or a @code{struct}, and it is
possible for these fields to have a different scalar storage order than the
enclosing type.

This attribute is supported only for targets that use a uniform default
scalar storage order (fortunately, most of them), i.e.@: targets that store
the scalars either all in big-endian or all in little-endian.

Additional restrictions are enforced for types with the reverse scalar
storage order with regard to the scalar storage order of the target:

@itemize
@item Taking the address of a scalar field of a @code{union} or a
@code{struct} with reverse scalar storage order is not permitted and yields
an error.
@item Taking the address of an array field, whose component is scalar, of
a @code{union} or a @code{struct} with reverse scalar storage order is
permitted but yields a warning, unless @option{-Wno-scalar-storage-order}
is specified.
@item Taking the address of a @code{union} or a @code{struct} with reverse
scalar storage order is permitted.
@end itemize

These restrictions exist because the storage order attribute is lost when
the address of a scalar or the address of an array with scalar component is
taken, so storing indirectly through this address generally does not work.
The second case is nevertheless allowed to be able to perform a block copy
from or to the array.

Moreover, the use of type punning or aliasing to toggle the storage order
is not supported; that is to say, a given scalar object cannot be accessed
through distinct types that assign a different storage order to it.

@item transparent_union
@cindex @code{transparent_union} type attribute

This attribute, attached to a @code{union} type definition, indicates
that any function parameter having that union type causes calls to that
function to be treated in a special way.

First, the argument corresponding to a transparent union type can be of
any type in the union; no cast is required.  Also, if the union contains
a pointer type, the corresponding argument can be a null pointer
constant or a void pointer expression; and if the union contains a void
pointer type, the corresponding argument can be any pointer expression.
If the union member type is a pointer, qualifiers like @code{const} on
the referenced type must be respected, just as with normal pointer
conversions.

Second, the argument is passed to the function using the calling
conventions of the first member of the transparent union, not the calling
conventions of the union itself.  All members of the union must have the
same machine representation; this is necessary for this argument passing
to work properly.

Transparent unions are designed for library functions that have multiple
interfaces for compatibility reasons.  For example, suppose the
@code{wait} function must accept either a value of type @code{int *} to
comply with POSIX, or a value of type @code{union wait *} to comply with
the 4.1BSD interface.  If @code{wait}'s parameter were @code{void *},
@code{wait} would accept both kinds of arguments, but it would also
accept any other pointer type and this would make argument type checking
less useful.  Instead, @code{<sys/wait.h>} might define the interface
as follows:

@smallexample
typedef union __attribute__ ((__transparent_union__))
  @{
    int *__ip;
    union wait *__up;
  @} wait_status_ptr_t;

pid_t wait (wait_status_ptr_t);
@end smallexample

@noindent
This interface allows either @code{int *} or @code{union wait *}
arguments to be passed, using the @code{int *} calling convention.
The program can call @code{wait} with arguments of either type:

@smallexample
int w1 () @{ int w; return wait (&w); @}
int w2 () @{ union wait w; return wait (&w); @}
@end smallexample

@noindent
With this interface, @code{wait}'s implementation might look like this:

@smallexample
pid_t wait (wait_status_ptr_t p)
@{
  return waitpid (-1, p.__ip, 0);
@}
@end smallexample

@item unused
@cindex @code{unused} type attribute
When attached to a type (including a @code{union} or a @code{struct}),
this attribute means that variables of that type are meant to appear
possibly unused.  GCC does not produce a warning for any variables of
that type, even if the variable appears to do nothing.  This is often
the case with lock or thread classes, which are usually defined and then
not referenced, but contain constructors and destructors that have
nontrivial bookkeeping functions.

@item vector_size (@var{bytes})
@cindex @code{vector_size} type attribute
This attribute specifies the vector size for the type, measured in bytes.
The type to which it applies is known as the @dfn{base type}.  The @var{bytes}
argument must be a positive power-of-two multiple of the base type size.  For
example, the following declarations:

@smallexample
typedef __attribute__ ((vector_size (32))) int int_vec32_t ;
typedef __attribute__ ((vector_size (32))) int* int_vec32_ptr_t;
typedef __attribute__ ((vector_size (32))) int int_vec32_arr3_t[3];
@end smallexample

@noindent
define @code{int_vec32_t} to be a 32-byte vector type composed of @code{int}
sized units.  With @code{int} having a size of 4 bytes, the type defines
a vector of eight units, four bytes each.  The mode of variables of type
@code{int_vec32_t} is @code{V8SI}.  @code{int_vec32_ptr_t} is then defined
to be a pointer to such a vector type, and @code{int_vec32_arr3_t} to be
an array of three such vectors.  @xref{Vector Extensions}, for details of
manipulating objects of vector types.

This attribute is only applicable to integral and floating scalar types.
In function declarations the attribute applies to the function return
type.

For example, the following:
@smallexample
__attribute__ ((vector_size (16))) float get_flt_vec16 (void);
@end smallexample
declares @code{get_flt_vec16} to be a function returning a 16-byte vector
with the base type @code{float}.

@item visibility
@cindex @code{visibility} type attribute
In C++, attribute visibility (@pxref{Function Attributes}) can also be
applied to class, struct, union and enum types.  Unlike other type
attributes, the attribute must appear between the initial keyword and
the name of the type; it cannot appear after the body of the type.

Note that the type visibility is applied to vague linkage entities
associated with the class (vtable, typeinfo node, etc.).  In
particular, if a class is thrown as an exception in one shared object
and caught in another, the class must have default visibility.
Otherwise the two shared objects are unable to use the same
typeinfo node and exception handling will break.

@end table

To specify multiple attributes, separate them by commas within the
double parentheses: for example, @samp{__attribute__ ((aligned (16),
packed))}.

@node ARC Type Attributes
@subsection ARC Type Attributes

@cindex @code{uncached} type attribute, ARC
Declaring objects with @code{uncached} allows you to exclude
data-cache participation in load and store operations on those objects
without involving the additional semantic implications of
@code{volatile}.  The @code{.di} instruction suffix is used for all
loads and stores of data declared @code{uncached}.

@node ARM Type Attributes
@subsection ARM Type Attributes

@cindex @code{notshared} type attribute, ARM
On those ARM targets that support @code{dllimport} (such as Symbian
OS), you can use the @code{notshared} attribute to indicate that the
virtual table and other similar data for a class should not be
exported from a DLL@.  For example:

@smallexample
class __declspec(notshared) C @{
public:
  __declspec(dllimport) C();
  virtual void f();
@}

__declspec(dllexport)
C::C() @{@}
@end smallexample

@noindent
In this code, @code{C::C} is exported from the current DLL, but the
virtual table for @code{C} is not exported.  (You can use
@code{__attribute__} instead of @code{__declspec} if you prefer, but
most Symbian OS code uses @code{__declspec}.)

@node MeP Type Attributes
@subsection MeP Type Attributes

@cindex @code{based} type attribute, MeP
@cindex @code{tiny} type attribute, MeP
@cindex @code{near} type attribute, MeP
@cindex @code{far} type attribute, MeP
Many of the MeP variable attributes may be applied to types as well.
Specifically, the @code{based}, @code{tiny}, @code{near}, and
@code{far} attributes may be applied to either.  The @code{io} and
@code{cb} attributes may not be applied to types.

@node PowerPC Type Attributes
@subsection PowerPC Type Attributes

Three attributes currently are defined for PowerPC configurations:
@code{altivec}, @code{ms_struct} and @code{gcc_struct}.

@cindex @code{ms_struct} type attribute, PowerPC
@cindex @code{gcc_struct} type attribute, PowerPC
For full documentation of the @code{ms_struct} and @code{gcc_struct}
attributes please see the documentation in @ref{x86 Type Attributes}.

@cindex @code{altivec} type attribute, PowerPC
The @code{altivec} attribute allows one to declare AltiVec vector data
types supported by the AltiVec Programming Interface Manual.  The
attribute requires an argument to specify one of three vector types:
@code{vector__}, @code{pixel__} (always followed by unsigned short),
and @code{bool__} (always followed by unsigned).

@smallexample
__attribute__((altivec(vector__)))
__attribute__((altivec(pixel__))) unsigned short
__attribute__((altivec(bool__))) unsigned
@end smallexample

These attributes mainly are intended to support the @code{__vector},
@code{__pixel}, and @code{__bool} AltiVec keywords.

@node x86 Type Attributes
@subsection x86 Type Attributes

Two attributes are currently defined for x86 configurations:
@code{ms_struct} and @code{gcc_struct}.

@table @code

@item ms_struct
@itemx gcc_struct
@cindex @code{ms_struct} type attribute, x86
@cindex @code{gcc_struct} type attribute, x86

If @code{packed} is used on a structure, or if bit-fields are used
it may be that the Microsoft ABI packs them differently
than GCC normally packs them.  Particularly when moving packed
data between functions compiled with GCC and the native Microsoft compiler
(either via function call or as data in a file), it may be necessary to access
either format.

The @code{ms_struct} and @code{gcc_struct} attributes correspond
to the @option{-mms-bitfields} and @option{-mno-ms-bitfields}
command-line options, respectively;
see @ref{x86 Options}, for details of how structure layout is affected.
@xref{x86 Variable Attributes}, for information about the corresponding
attributes on variables.

@end table

@node Label Attributes
@section Label Attributes
@cindex Label Attributes

GCC allows attributes to be set on C labels.  @xref{Attribute Syntax}, for 
details of the exact syntax for using attributes.  Other attributes are 
available for functions (@pxref{Function Attributes}), variables 
(@pxref{Variable Attributes}), enumerators (@pxref{Enumerator Attributes}),
statements (@pxref{Statement Attributes}), and for types
(@pxref{Type Attributes}).

This example uses the @code{cold} label attribute to indicate the 
@code{ErrorHandling} branch is unlikely to be taken and that the
@code{ErrorHandling} label is unused:

@smallexample

   asm goto ("some asm" : : : : NoError);

/* This branch (the fall-through from the asm) is less commonly used */
ErrorHandling: 
   __attribute__((cold, unused)); /* Semi-colon is required here */
   printf("error\n");
   return 0;

NoError:
   printf("no error\n");
   return 1;
@end smallexample

@table @code
@item unused
@cindex @code{unused} label attribute
This feature is intended for program-generated code that may contain 
unused labels, but which is compiled with @option{-Wall}.  It is
not normally appropriate to use in it human-written code, though it
could be useful in cases where the code that jumps to the label is
contained within an @code{#ifdef} conditional.

@item hot
@cindex @code{hot} label attribute
The @code{hot} attribute on a label is used to inform the compiler that
the path following the label is more likely than paths that are not so
annotated.  This attribute is used in cases where @code{__builtin_expect}
cannot be used, for instance with computed goto or @code{asm goto}.

@item cold
@cindex @code{cold} label attribute
The @code{cold} attribute on labels is used to inform the compiler that
the path following the label is unlikely to be executed.  This attribute
is used in cases where @code{__builtin_expect} cannot be used, for instance
with computed goto or @code{asm goto}.

@end table

@node Enumerator Attributes
@section Enumerator Attributes
@cindex Enumerator Attributes

GCC allows attributes to be set on enumerators.  @xref{Attribute Syntax}, for
details of the exact syntax for using attributes.  Other attributes are
available for functions (@pxref{Function Attributes}), variables
(@pxref{Variable Attributes}), labels (@pxref{Label Attributes}), statements
(@pxref{Statement Attributes}), and for types (@pxref{Type Attributes}).

This example uses the @code{deprecated} enumerator attribute to indicate the
@code{oldval} enumerator is deprecated:

@smallexample
enum E @{
  oldval __attribute__((deprecated)),
  newval
@};

int
fn (void)
@{
  return oldval;
@}
@end smallexample

@table @code
@item deprecated
@cindex @code{deprecated} enumerator attribute
The @code{deprecated} attribute results in a warning if the enumerator
is used anywhere in the source file.  This is useful when identifying
enumerators that are expected to be removed in a future version of a
program.  The warning also includes the location of the declaration
of the deprecated enumerator, to enable users to easily find further
information about why the enumerator is deprecated, or what they should
do instead.  Note that the warnings only occurs for uses.

@end table

@node Statement Attributes
@section Statement Attributes
@cindex Statement Attributes

GCC allows attributes to be set on null statements.  @xref{Attribute Syntax},
for details of the exact syntax for using attributes.  Other attributes are
available for functions (@pxref{Function Attributes}), variables
(@pxref{Variable Attributes}), labels (@pxref{Label Attributes}), enumerators
(@pxref{Enumerator Attributes}), and for types (@pxref{Type Attributes}).

This example uses the @code{fallthrough} statement attribute to indicate that
the @option{-Wimplicit-fallthrough} warning should not be emitted:

@smallexample
switch (cond)
  @{
  case 1:
    bar (1);
    __attribute__((fallthrough));
  case 2:
    @dots{}
  @}
@end smallexample

@table @code
@item fallthrough
@cindex @code{fallthrough} statement attribute
The @code{fallthrough} attribute with a null statement serves as a
fallthrough statement.  It hints to the compiler that a statement
that falls through to another case label, or user-defined label
in a switch statement is intentional and thus the
@option{-Wimplicit-fallthrough} warning must not trigger.  The
fallthrough attribute may appear at most once in each attribute
list, and may not be mixed with other attributes.  It can only
be used in a switch statement (the compiler will issue an error
otherwise), after a preceding statement and before a logically
succeeding case label, or user-defined label.

@end table

@node Attribute Syntax
@section Attribute Syntax
@cindex attribute syntax

This section describes the syntax with which @code{__attribute__} may be
used, and the constructs to which attribute specifiers bind, for the C
language.  Some details may vary for C++ and Objective-C@.  Because of
infelicities in the grammar for attributes, some forms described here
may not be successfully parsed in all cases.

There are some problems with the semantics of attributes in C++.  For
example, there are no manglings for attributes, although they may affect
code generation, so problems may arise when attributed types are used in
conjunction with templates or overloading.  Similarly, @code{typeid}
does not distinguish between types with different attributes.  Support
for attributes in C++ may be restricted in future to attributes on
declarations only, but not on nested declarators.

@xref{Function Attributes}, for details of the semantics of attributes
applying to functions.  @xref{Variable Attributes}, for details of the
semantics of attributes applying to variables.  @xref{Type Attributes},
for details of the semantics of attributes applying to structure, union
and enumerated types.
@xref{Label Attributes}, for details of the semantics of attributes 
applying to labels.
@xref{Enumerator Attributes}, for details of the semantics of attributes
applying to enumerators.
@xref{Statement Attributes}, for details of the semantics of attributes
applying to statements.

An @dfn{attribute specifier} is of the form
@code{__attribute__ ((@var{attribute-list}))}.  An @dfn{attribute list}
is a possibly empty comma-separated sequence of @dfn{attributes}, where
each attribute is one of the following:

@itemize @bullet
@item
Empty.  Empty attributes are ignored.

@item
An attribute name
(which may be an identifier such as @code{unused}, or a reserved
word such as @code{const}).

@item
An attribute name followed by a parenthesized list of
parameters for the attribute.
These parameters take one of the following forms:

@itemize @bullet
@item
An identifier.  For example, @code{mode} attributes use this form.

@item
An identifier followed by a comma and a non-empty comma-separated list
of expressions.  For example, @code{format} attributes use this form.

@item
A possibly empty comma-separated list of expressions.  For example,
@code{format_arg} attributes use this form with the list being a single
integer constant expression, and @code{alias} attributes use this form
with the list being a single string constant.
@end itemize
@end itemize

An @dfn{attribute specifier list} is a sequence of one or more attribute
specifiers, not separated by any other tokens.

You may optionally specify attribute names with @samp{__}
preceding and following the name.
This allows you to use them in header files without
being concerned about a possible macro of the same name.  For example,
you may use the attribute name @code{__noreturn__} instead of @code{noreturn}.


@subsubheading Label Attributes

In GNU C, an attribute specifier list may appear after the colon following a
label, other than a @code{case} or @code{default} label.  GNU C++ only permits
attributes on labels if the attribute specifier is immediately
followed by a semicolon (i.e., the label applies to an empty
statement).  If the semicolon is missing, C++ label attributes are
ambiguous, as it is permissible for a declaration, which could begin
with an attribute list, to be labelled in C++.  Declarations cannot be
labelled in C90 or C99, so the ambiguity does not arise there.

@subsubheading Enumerator Attributes

In GNU C, an attribute specifier list may appear as part of an enumerator.
The attribute goes after the enumeration constant, before @code{=}, if
present.  The optional attribute in the enumerator appertains to the
enumeration constant.  It is not possible to place the attribute after
the constant expression, if present.

@subsubheading Statement Attributes
In GNU C, an attribute specifier list may appear as part of a null
statement.  The attribute goes before the semicolon.

@subsubheading Type Attributes

An attribute specifier list may appear as part of a @code{struct},
@code{union} or @code{enum} specifier.  It may go either immediately
after the @code{struct}, @code{union} or @code{enum} keyword, or after
the closing brace.  The former syntax is preferred.
Where attribute specifiers follow the closing brace, they are considered
to relate to the structure, union or enumerated type defined, not to any
enclosing declaration the type specifier appears in, and the type
defined is not complete until after the attribute specifiers.
@c Otherwise, there would be the following problems: a shift/reduce
@c conflict between attributes binding the struct/union/enum and
@c binding to the list of specifiers/qualifiers; and "aligned"
@c attributes could use sizeof for the structure, but the size could be
@c changed later by "packed" attributes.


@subsubheading All other attributes

Otherwise, an attribute specifier appears as part of a declaration,
counting declarations of unnamed parameters and type names, and relates
to that declaration (which may be nested in another declaration, for
example in the case of a parameter declaration), or to a particular declarator
within a declaration.  Where an
attribute specifier is applied to a parameter declared as a function or
an array, it should apply to the function or array rather than the
pointer to which the parameter is implicitly converted, but this is not
yet correctly implemented.

Any list of specifiers and qualifiers at the start of a declaration may
contain attribute specifiers, whether or not such a list may in that
context contain storage class specifiers.  (Some attributes, however,
are essentially in the nature of storage class specifiers, and only make
sense where storage class specifiers may be used; for example,
@code{section}.)  There is one necessary limitation to this syntax: the
first old-style parameter declaration in a function definition cannot
begin with an attribute specifier, because such an attribute applies to
the function instead by syntax described below (which, however, is not
yet implemented in this case).  In some other cases, attribute
specifiers are permitted by this grammar but not yet supported by the
compiler.  All attribute specifiers in this place relate to the
declaration as a whole.  In the obsolescent usage where a type of
@code{int} is implied by the absence of type specifiers, such a list of
specifiers and qualifiers may be an attribute specifier list with no
other specifiers or qualifiers.

At present, the first parameter in a function prototype must have some
type specifier that is not an attribute specifier; this resolves an
ambiguity in the interpretation of @code{void f(int
(__attribute__((foo)) x))}, but is subject to change.  At present, if
the parentheses of a function declarator contain only attributes then
those attributes are ignored, rather than yielding an error or warning
or implying a single parameter of type int, but this is subject to
change.

An attribute specifier list may appear immediately before a declarator
(other than the first) in a comma-separated list of declarators in a
declaration of more than one identifier using a single list of
specifiers and qualifiers.  Such attribute specifiers apply
only to the identifier before whose declarator they appear.  For
example, in

@smallexample
__attribute__((noreturn)) void d0 (void),
    __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
     d2 (void);
@end smallexample

@noindent
the @code{noreturn} attribute applies to all the functions
declared; the @code{format} attribute only applies to @code{d1}.

An attribute specifier list may appear immediately before the comma,
@code{=} or semicolon terminating the declaration of an identifier other
than a function definition.  Such attribute specifiers apply
to the declared object or function.  Where an
assembler name for an object or function is specified (@pxref{Asm
Labels}), the attribute must follow the @code{asm}
specification.

An attribute specifier list may, in future, be permitted to appear after
the declarator in a function definition (before any old-style parameter
declarations or the function body).

Attribute specifiers may be mixed with type qualifiers appearing inside
the @code{[]} of a parameter array declarator, in the C99 construct by
which such qualifiers are applied to the pointer to which the array is
implicitly converted.  Such attribute specifiers apply to the pointer,
not to the array, but at present this is not implemented and they are
ignored.

An attribute specifier list may appear at the start of a nested
declarator.  At present, there are some limitations in this usage: the
attributes correctly apply to the declarator, but for most individual
attributes the semantics this implies are not implemented.
When attribute specifiers follow the @code{*} of a pointer
declarator, they may be mixed with any type qualifiers present.
The following describes the formal semantics of this syntax.  It makes the
most sense if you are familiar with the formal specification of
declarators in the ISO C standard.

Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
D1}, where @code{T} contains declaration specifiers that specify a type
@var{Type} (such as @code{int}) and @code{D1} is a declarator that
contains an identifier @var{ident}.  The type specified for @var{ident}
for derived declarators whose type does not include an attribute
specifier is as in the ISO C standard.

If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
and the declaration @code{T D} specifies the type
``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
@code{T D1} specifies the type ``@var{derived-declarator-type-list}
@var{attribute-specifier-list} @var{Type}'' for @var{ident}.

If @code{D1} has the form @code{*
@var{type-qualifier-and-attribute-specifier-list} D}, and the
declaration @code{T D} specifies the type
``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
@code{T D1} specifies the type ``@var{derived-declarator-type-list}
@var{type-qualifier-and-attribute-specifier-list} pointer to @var{Type}'' for
@var{ident}.

For example,

@smallexample
void (__attribute__((noreturn)) ****f) (void);
@end smallexample

@noindent
specifies the type ``pointer to pointer to pointer to pointer to
non-returning function returning @code{void}''.  As another example,

@smallexample
char *__attribute__((aligned(8))) *f;
@end smallexample

@noindent
specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
Note again that this does not work with most attributes; for example,
the usage of @samp{aligned} and @samp{noreturn} attributes given above
is not yet supported.

For compatibility with existing code written for compiler versions that
did not implement attributes on nested declarators, some laxity is
allowed in the placing of attributes.  If an attribute that only applies
to types is applied to a declaration, it is treated as applying to
the type of that declaration.  If an attribute that only applies to
declarations is applied to the type of a declaration, it is treated
as applying to that declaration; and, for compatibility with code
placing the attributes immediately before the identifier declared, such
an attribute applied to a function return type is treated as
applying to the function type, and such an attribute applied to an array
element type is treated as applying to the array type.  If an
attribute that only applies to function types is applied to a
pointer-to-function type, it is treated as applying to the pointer
target type; if such an attribute is applied to a function return type
that is not a pointer-to-function type, it is treated as applying
to the function type.

@node Function Prototypes
@section Prototypes and Old-Style Function Definitions
@cindex function prototype declarations
@cindex old-style function definitions
@cindex promotion of formal parameters

GNU C extends ISO C to allow a function prototype to override a later
old-style non-prototype definition.  Consider the following example:

@smallexample
/* @r{Use prototypes unless the compiler is old-fashioned.}  */
#ifdef __STDC__
#define P(x) x
#else
#define P(x) ()
#endif

/* @r{Prototype function declaration.}  */
int isroot P((uid_t));

/* @r{Old-style function definition.}  */
int
isroot (x)   /* @r{??? lossage here ???} */
     uid_t x;
@{
  return x == 0;
@}
@end smallexample

Suppose the type @code{uid_t} happens to be @code{short}.  ISO C does
not allow this example, because subword arguments in old-style
non-prototype definitions are promoted.  Therefore in this example the
function definition's argument is really an @code{int}, which does not
match the prototype argument type of @code{short}.

This restriction of ISO C makes it hard to write code that is portable
to traditional C compilers, because the programmer does not know
whether the @code{uid_t} type is @code{short}, @code{int}, or
@code{long}.  Therefore, in cases like these GNU C allows a prototype
to override a later old-style definition.  More precisely, in GNU C, a
function prototype argument type overrides the argument type specified
by a later old-style definition if the former type is the same as the
latter type before promotion.  Thus in GNU C the above example is
equivalent to the following:

@smallexample
int isroot (uid_t);

int
isroot (uid_t x)
@{
  return x == 0;
@}
@end smallexample

@noindent
GNU C++ does not support old-style function definitions, so this
extension is irrelevant.

@node C++ Comments
@section C++ Style Comments
@cindex @code{//}
@cindex C++ comments
@cindex comments, C++ style

In GNU C, you may use C++ style comments, which start with @samp{//} and
continue until the end of the line.  Many other C implementations allow
such comments, and they are included in the 1999 C standard.  However,
C++ style comments are not recognized if you specify an @option{-std}
option specifying a version of ISO C before C99, or @option{-ansi}
(equivalent to @option{-std=c90}).

@node Dollar Signs
@section Dollar Signs in Identifier Names
@cindex $
@cindex dollar signs in identifier names
@cindex identifier names, dollar signs in

In GNU C, you may normally use dollar signs in identifier names.
This is because many traditional C implementations allow such identifiers.
However, dollar signs in identifiers are not supported on a few target
machines, typically because the target assembler does not allow them.

@node Character Escapes
@section The Character @key{ESC} in Constants

You can use the sequence @samp{\e} in a string or character constant to
stand for the ASCII character @key{ESC}.

@node Alignment
@section Determining the Alignment of Functions, Types or Variables
@cindex alignment
@cindex type alignment
@cindex variable alignment

The keyword @code{__alignof__} determines the alignment requirement of
a function, object, or a type, or the minimum alignment usually required
by a type.  Its syntax is just like @code{sizeof} and C11 @code{_Alignof}.

For example, if the target machine requires a @code{double} value to be
aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
This is true on many RISC machines.  On more traditional machine
designs, @code{__alignof__ (double)} is 4 or even 2.

Some machines never actually require alignment; they allow references to any
data type even at an odd address.  For these machines, @code{__alignof__}
reports the smallest alignment that GCC gives the data type, usually as
mandated by the target ABI.

If the operand of @code{__alignof__} is an lvalue rather than a type,
its value is the required alignment for its type, taking into account
any minimum alignment specified by attribute @code{aligned}
(@pxref{Common Variable Attributes}).  For example, after this
declaration:

@smallexample
struct foo @{ int x; char y; @} foo1;
@end smallexample

@noindent
the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
It is an error to ask for the alignment of an incomplete type other
than @code{void}.

If the operand of the @code{__alignof__} expression is a function,
the expression evaluates to the alignment of the function which may
be specified by attribute @code{aligned} (@pxref{Common Function Attributes}).

@node Inline
@section An Inline Function is As Fast As a Macro
@cindex inline functions
@cindex integrating function code
@cindex open coding
@cindex macros, inline alternative

By declaring a function inline, you can direct GCC to make
calls to that function faster.  One way GCC can achieve this is to
integrate that function's code into the code for its callers.  This
makes execution faster by eliminating the function-call overhead; in
addition, if any of the actual argument values are constant, their
known values may permit simplifications at compile time so that not
all of the inline function's code needs to be included.  The effect on
code size is less predictable; object code may be larger or smaller
with function inlining, depending on the particular case.  You can
also direct GCC to try to integrate all ``simple enough'' functions
into their callers with the option @option{-finline-functions}.

GCC implements three different semantics of declaring a function
inline.  One is available with @option{-std=gnu89} or
@option{-fgnu89-inline} or when @code{gnu_inline} attribute is present
on all inline declarations, another when
@option{-std=c99},
@option{-std=gnu99} or an option for a later C version is used
(without @option{-fgnu89-inline}), and the third
is used when compiling C++.

To declare a function inline, use the @code{inline} keyword in its
declaration, like this:

@smallexample
static inline int
inc (int *a)
@{
  return (*a)++;
@}
@end smallexample

If you are writing a header file to be included in ISO C90 programs, write
@code{__inline__} instead of @code{inline}.  @xref{Alternate Keywords}.

The three types of inlining behave similarly in two important cases:
when the @code{inline} keyword is used on a @code{static} function,
like the example above, and when a function is first declared without
using the @code{inline} keyword and then is defined with
@code{inline}, like this:

@smallexample
extern int inc (int *a);
inline int
inc (int *a)
@{
  return (*a)++;
@}
@end smallexample

In both of these common cases, the program behaves the same as if you
had not used the @code{inline} keyword, except for its speed.

@cindex inline functions, omission of
@opindex fkeep-inline-functions
When a function is both inline and @code{static}, if all calls to the
function are integrated into the caller, and the function's address is
never used, then the function's own assembler code is never referenced.
In this case, GCC does not actually output assembler code for the
function, unless you specify the option @option{-fkeep-inline-functions}.
If there is a nonintegrated call, then the function is compiled to
assembler code as usual.  The function must also be compiled as usual if
the program refers to its address, because that cannot be inlined.

@opindex Winline
Note that certain usages in a function definition can make it unsuitable
for inline substitution.  Among these usages are: variadic functions,
use of @code{alloca}, use of computed goto (@pxref{Labels as Values}),
use of nonlocal goto, use of nested functions, use of @code{setjmp}, use
of @code{__builtin_longjmp} and use of @code{__builtin_return} or
@code{__builtin_apply_args}.  Using @option{-Winline} warns when a
function marked @code{inline} could not be substituted, and gives the
reason for the failure.

@cindex automatic @code{inline} for C++ member fns
@cindex @code{inline} automatic for C++ member fns
@cindex member fns, automatically @code{inline}
@cindex C++ member fns, automatically @code{inline}
@opindex fno-default-inline
As required by ISO C++, GCC considers member functions defined within
the body of a class to be marked inline even if they are
not explicitly declared with the @code{inline} keyword.  You can
override this with @option{-fno-default-inline}; @pxref{C++ Dialect
Options,,Options Controlling C++ Dialect}.

GCC does not inline any functions when not optimizing unless you specify
the @samp{always_inline} attribute for the function, like this:

@smallexample
/* @r{Prototype.}  */
inline void foo (const char) __attribute__((always_inline));
@end smallexample

The remainder of this section is specific to GNU C90 inlining.

@cindex non-static inline function
When an inline function is not @code{static}, then the compiler must assume
that there may be calls from other source files; since a global symbol can
be defined only once in any program, the function must not be defined in
the other source files, so the calls therein cannot be integrated.
Therefore, a non-@code{static} inline function is always compiled on its
own in the usual fashion.

If you specify both @code{inline} and @code{extern} in the function
definition, then the definition is used only for inlining.  In no case
is the function compiled on its own, not even if you refer to its
address explicitly.  Such an address becomes an external reference, as
if you had only declared the function, and had not defined it.

This combination of @code{inline} and @code{extern} has almost the
effect of a macro.  The way to use it is to put a function definition in
a header file with these keywords, and put another copy of the
definition (lacking @code{inline} and @code{extern}) in a library file.
The definition in the header file causes most calls to the function
to be inlined.  If any uses of the function remain, they refer to
the single copy in the library.

@node Volatiles
@section When is a Volatile Object Accessed?
@cindex accessing volatiles
@cindex volatile read
@cindex volatile write
@cindex volatile access

C has the concept of volatile objects.  These are normally accessed by
pointers and used for accessing hardware or inter-thread
communication.  The standard encourages compilers to refrain from
optimizations concerning accesses to volatile objects, but leaves it
implementation defined as to what constitutes a volatile access.  The
minimum requirement is that at a sequence point all previous accesses
to volatile objects have stabilized and no subsequent accesses have
occurred.  Thus an implementation is free to reorder and combine
volatile accesses that occur between sequence points, but cannot do
so for accesses across a sequence point.  The use of volatile does
not allow you to violate the restriction on updating objects multiple
times between two sequence points.

Accesses to non-volatile objects are not ordered with respect to
volatile accesses.  You cannot use a volatile object as a memory
barrier to order a sequence of writes to non-volatile memory.  For
instance:

@smallexample
int *ptr = @var{something};
volatile int vobj;
*ptr = @var{something};
vobj = 1;
@end smallexample

@noindent
Unless @var{*ptr} and @var{vobj} can be aliased, it is not guaranteed
that the write to @var{*ptr} occurs by the time the update
of @var{vobj} happens.  If you need this guarantee, you must use
a stronger memory barrier such as:

@smallexample
int *ptr = @var{something};
volatile int vobj;
*ptr = @var{something};
asm volatile ("" : : : "memory");
vobj = 1;
@end smallexample

A scalar volatile object is read when it is accessed in a void context:

@smallexample
volatile int *src = @var{somevalue};
*src;
@end smallexample

Such expressions are rvalues, and GCC implements this as a
read of the volatile object being pointed to.

Assignments are also expressions and have an rvalue.  However when
assigning to a scalar volatile, the volatile object is not reread,
regardless of whether the assignment expression's rvalue is used or
not.  If the assignment's rvalue is used, the value is that assigned
to the volatile object.  For instance, there is no read of @var{vobj}
in all the following cases:

@smallexample
int obj;
volatile int vobj;
vobj = @var{something};
obj = vobj = @var{something};
obj ? vobj = @var{onething} : vobj = @var{anotherthing};
obj = (@var{something}, vobj = @var{anotherthing});
@end smallexample

If you need to read the volatile object after an assignment has
occurred, you must use a separate expression with an intervening
sequence point.

As bit-fields are not individually addressable, volatile bit-fields may
be implicitly read when written to, or when adjacent bit-fields are
accessed.  Bit-field operations may be optimized such that adjacent
bit-fields are only partially accessed, if they straddle a storage unit
boundary.  For these reasons it is unwise to use volatile bit-fields to
access hardware.

@node Using Assembly Language with C
@section How to Use Inline Assembly Language in C Code
@cindex @code{asm} keyword
@cindex assembly language in C
@cindex inline assembly language
@cindex mixing assembly language and C

The @code{asm} keyword allows you to embed assembler instructions
within C code.  GCC provides two forms of inline @code{asm}
statements.  A @dfn{basic @code{asm}} statement is one with no
operands (@pxref{Basic Asm}), while an @dfn{extended @code{asm}}
statement (@pxref{Extended Asm}) includes one or more operands.  
The extended form is preferred for mixing C and assembly language
within a function, but to include assembly language at
top level you must use basic @code{asm}.

You can also use the @code{asm} keyword to override the assembler name
for a C symbol, or to place a C variable in a specific register.

@menu
* Basic Asm::          Inline assembler without operands.
* Extended Asm::       Inline assembler with operands.
* Constraints::        Constraints for @code{asm} operands
* Asm Labels::         Specifying the assembler name to use for a C symbol.
* Explicit Register Variables::  Defining variables residing in specified 
                       registers.
* Size of an asm::     How GCC calculates the size of an @code{asm} block.
@end menu

@node Basic Asm
@subsection Basic Asm --- Assembler Instructions Without Operands
@cindex basic @code{asm}
@cindex assembly language in C, basic

A basic @code{asm} statement has the following syntax:

@example
asm @var{asm-qualifiers} ( @var{AssemblerInstructions} )
@end example

The @code{asm} keyword is a GNU extension.
When writing code that can be compiled with @option{-ansi} and the
various @option{-std} options, use @code{__asm__} instead of 
@code{asm} (@pxref{Alternate Keywords}).

@subsubheading Qualifiers
@table @code
@item volatile
The optional @code{volatile} qualifier has no effect. 
All basic @code{asm} blocks are implicitly volatile.

@item inline
If you use the @code{inline} qualifier, then for inlining purposes the size
of the @code{asm} statement is taken as the smallest size possible (@pxref{Size
of an asm}).
@end table

@subsubheading Parameters
@table @var

@item AssemblerInstructions
This is a literal string that specifies the assembler code. The string can 
contain any instructions recognized by the assembler, including directives. 
GCC does not parse the assembler instructions themselves and 
does not know what they mean or even whether they are valid assembler input. 

You may place multiple assembler instructions together in a single @code{asm} 
string, separated by the characters normally used in assembly code for the 
system. A combination that works in most places is a newline to break the 
line, plus a tab character (written as @samp{\n\t}).
Some assemblers allow semicolons as a line separator. However, 
note that some assembler dialects use semicolons to start a comment. 
@end table

@subsubheading Remarks
Using extended @code{asm} (@pxref{Extended Asm}) typically produces
smaller, safer, and more efficient code, and in most cases it is a
better solution than basic @code{asm}.  However, there are two
situations where only basic @code{asm} can be used:

@itemize @bullet
@item
Extended @code{asm} statements have to be inside a C
function, so to write inline assembly language at file scope (``top-level''),
outside of C functions, you must use basic @code{asm}.
You can use this technique to emit assembler directives,
define assembly language macros that can be invoked elsewhere in the file,
or write entire functions in assembly language.
Basic @code{asm} statements outside of functions may not use any
qualifiers.

@item
Functions declared
with the @code{naked} attribute also require basic @code{asm}
(@pxref{Function Attributes}).
@end itemize

Safely accessing C data and calling functions from basic @code{asm} is more 
complex than it may appear. To access C data, it is better to use extended 
@code{asm}.

Do not expect a sequence of @code{asm} statements to remain perfectly 
consecutive after compilation. If certain instructions need to remain 
consecutive in the output, put them in a single multi-instruction @code{asm}
statement. Note that GCC's optimizers can move @code{asm} statements 
relative to other code, including across jumps.

@code{asm} statements may not perform jumps into other @code{asm} statements. 
GCC does not know about these jumps, and therefore cannot take 
account of them when deciding how to optimize. Jumps from @code{asm} to C 
labels are only supported in extended @code{asm}.

Under certain circumstances, GCC may duplicate (or remove duplicates of) your 
assembly code when optimizing. This can lead to unexpected duplicate 
symbol errors during compilation if your assembly code defines symbols or 
labels.

@strong{Warning:} The C standards do not specify semantics for @code{asm},
making it a potential source of incompatibilities between compilers.  These
incompatibilities may not produce compiler warnings/errors.

GCC does not parse basic @code{asm}'s @var{AssemblerInstructions}, which
means there is no way to communicate to the compiler what is happening
inside them.  GCC has no visibility of symbols in the @code{asm} and may
discard them as unreferenced.  It also does not know about side effects of
the assembler code, such as modifications to memory or registers.  Unlike
some compilers, GCC assumes that no changes to general purpose registers
occur.  This assumption may change in a future release.

To avoid complications from future changes to the semantics and the
compatibility issues between compilers, consider replacing basic @code{asm}
with extended @code{asm}.  See
@uref{https://gcc.gnu.org/wiki/ConvertBasicAsmToExtended, How to convert
from basic asm to extended asm} for information about how to perform this
conversion.

The compiler copies the assembler instructions in a basic @code{asm} 
verbatim to the assembly language output file, without 
processing dialects or any of the @samp{%} operators that are available with
extended @code{asm}. This results in minor differences between basic 
@code{asm} strings and extended @code{asm} templates. For example, to refer to 
registers you might use @samp{%eax} in basic @code{asm} and
@samp{%%eax} in extended @code{asm}.

On targets such as x86 that support multiple assembler dialects,
all basic @code{asm} blocks use the assembler dialect specified by the 
@option{-masm} command-line option (@pxref{x86 Options}).  
Basic @code{asm} provides no
mechanism to provide different assembler strings for different dialects.

For basic @code{asm} with non-empty assembler string GCC assumes
the assembler block does not change any general purpose registers,
but it may read or write any globally accessible variable.

Here is an example of basic @code{asm} for i386:

@example
/* Note that this code will not compile with -masm=intel */
#define DebugBreak() asm("int $3")
@end example

@node Extended Asm
@subsection Extended Asm - Assembler Instructions with C Expression Operands
@cindex extended @code{asm}
@cindex assembly language in C, extended

With extended @code{asm} you can read and write C variables from 
assembler and perform jumps from assembler code to C labels.  
Extended @code{asm} syntax uses colons (@samp{:}) to delimit
the operand parameters after the assembler template:

@example
asm @var{asm-qualifiers} ( @var{AssemblerTemplate} 
                 : @var{OutputOperands} 
                 @r{[} : @var{InputOperands}
                 @r{[} : @var{Clobbers} @r{]} @r{]})

asm @var{asm-qualifiers} ( @var{AssemblerTemplate} 
                      : 
                      : @var{InputOperands}
                      : @var{Clobbers}
                      : @var{GotoLabels})
@end example
where in the last form, @var{asm-qualifiers} contains @code{goto} (and in the
first form, not).

The @code{asm} keyword is a GNU extension.
When writing code that can be compiled with @option{-ansi} and the
various @option{-std} options, use @code{__asm__} instead of 
@code{asm} (@pxref{Alternate Keywords}).

@subsubheading Qualifiers
@table @code

@item volatile
The typical use of extended @code{asm} statements is to manipulate input 
values to produce output values. However, your @code{asm} statements may 
also produce side effects. If so, you may need to use the @code{volatile} 
qualifier to disable certain optimizations. @xref{Volatile}.

@item inline
If you use the @code{inline} qualifier, then for inlining purposes the size
of the @code{asm} statement is taken as the smallest size possible
(@pxref{Size of an asm}).

@item goto
This qualifier informs the compiler that the @code{asm} statement may 
perform a jump to one of the labels listed in the @var{GotoLabels}.
@xref{GotoLabels}.
@end table

@subsubheading Parameters
@table @var
@item AssemblerTemplate
This is a literal string that is the template for the assembler code. It is a 
combination of fixed text and tokens that refer to the input, output, 
and goto parameters. @xref{AssemblerTemplate}.

@item OutputOperands
A comma-separated list of the C variables modified by the instructions in the 
@var{AssemblerTemplate}.  An empty list is permitted.  @xref{OutputOperands}.

@item InputOperands
A comma-separated list of C expressions read by the instructions in the 
@var{AssemblerTemplate}.  An empty list is permitted.  @xref{InputOperands}.

@item Clobbers
A comma-separated list of registers or other values changed by the 
@var{AssemblerTemplate}, beyond those listed as outputs.
An empty list is permitted.  @xref{Clobbers and Scratch Registers}.

@item GotoLabels
When you are using the @code{goto} form of @code{asm}, this section contains 
the list of all C labels to which the code in the 
@var{AssemblerTemplate} may jump. 
@xref{GotoLabels}.

@code{asm} statements may not perform jumps into other @code{asm} statements,
only to the listed @var{GotoLabels}.
GCC's optimizers do not know about other jumps; therefore they cannot take 
account of them when deciding how to optimize.
@end table

The total number of input + output + goto operands is limited to 30.

@subsubheading Remarks
The @code{asm} statement allows you to include assembly instructions directly 
within C code. This may help you to maximize performance in time-sensitive 
code or to access assembly instructions that are not readily available to C 
programs.

Note that extended @code{asm} statements must be inside a function. Only 
basic @code{asm} may be outside functions (@pxref{Basic Asm}).
Functions declared with the @code{naked} attribute also require basic 
@code{asm} (@pxref{Function Attributes}).

While the uses of @code{asm} are many and varied, it may help to think of an 
@code{asm} statement as a series of low-level instructions that convert input 
parameters to output parameters. So a simple (if not particularly useful) 
example for i386 using @code{asm} might look like this:

@example
int src = 1;
int dst;   

asm ("mov %1, %0\n\t"
    "add $1, %0"
    : "=r" (dst) 
    : "r" (src));

printf("%d\n", dst);
@end example

This code copies @code{src} to @code{dst} and add 1 to @code{dst}.

@anchor{Volatile}
@subsubsection Volatile
@cindex volatile @code{asm}
@cindex @code{asm} volatile

GCC's optimizers sometimes discard @code{asm} statements if they determine 
there is no need for the output variables. Also, the optimizers may move 
code out of loops if they believe that the code will always return the same 
result (i.e.@: none of its input values change between calls). Using the 
@code{volatile} qualifier disables these optimizations. @code{asm} statements 
that have no output operands, including @code{asm goto} statements, 
are implicitly volatile.

This i386 code demonstrates a case that does not use (or require) the 
@code{volatile} qualifier. If it is performing assertion checking, this code 
uses @code{asm} to perform the validation. Otherwise, @code{dwRes} is 
unreferenced by any code. As a result, the optimizers can discard the 
@code{asm} statement, which in turn removes the need for the entire 
@code{DoCheck} routine. By omitting the @code{volatile} qualifier when it 
isn't needed you allow the optimizers to produce the most efficient code 
possible.

@example
void DoCheck(uint32_t dwSomeValue)
@{
   uint32_t dwRes;

   // Assumes dwSomeValue is not zero.
   asm ("bsfl %1,%0"
     : "=r" (dwRes)
     : "r" (dwSomeValue)
     : "cc");

   assert(dwRes > 3);
@}
@end example

The next example shows a case where the optimizers can recognize that the input 
(@code{dwSomeValue}) never changes during the execution of the function and can 
therefore move the @code{asm} outside the loop to produce more efficient code. 
Again, using the @code{volatile} qualifier disables this type of optimization.

@example
void do_print(uint32_t dwSomeValue)
@{
   uint32_t dwRes;

   for (uint32_t x=0; x < 5; x++)
   @{
      // Assumes dwSomeValue is not zero.
      asm ("bsfl %1,%0"
        : "=r" (dwRes)
        : "r" (dwSomeValue)
        : "cc");

      printf("%u: %u %u\n", x, dwSomeValue, dwRes);
   @}
@}
@end example

The following example demonstrates a case where you need to use the 
@code{volatile} qualifier. 
It uses the x86 @code{rdtsc} instruction, which reads 
the computer's time-stamp counter. Without the @code{volatile} qualifier, 
the optimizers might assume that the @code{asm} block will always return the 
same value and therefore optimize away the second call.

@example
uint64_t msr;

asm volatile ( "rdtsc\n\t"    // Returns the time in EDX:EAX.
        "shl $32, %%rdx\n\t"  // Shift the upper bits left.
        "or %%rdx, %0"        // 'Or' in the lower bits.
        : "=a" (msr)
        : 
        : "rdx");

printf("msr: %llx\n", msr);

// Do other work...

// Reprint the timestamp
asm volatile ( "rdtsc\n\t"    // Returns the time in EDX:EAX.
        "shl $32, %%rdx\n\t"  // Shift the upper bits left.
        "or %%rdx, %0"        // 'Or' in the lower bits.
        : "=a" (msr)
        : 
        : "rdx");

printf("msr: %llx\n", msr);
@end example

GCC's optimizers do not treat this code like the non-volatile code in the 
earlier examples. They do not move it out of loops or omit it on the 
assumption that the result from a previous call is still valid.

Note that the compiler can move even @code{volatile asm} instructions relative
to other code, including across jump instructions. For example, on many 
targets there is a system register that controls the rounding mode of 
floating-point operations. Setting it with a @code{volatile asm} statement,
as in the following PowerPC example, does not work reliably.

@example
asm volatile("mtfsf 255, %0" : : "f" (fpenv));
sum = x + y;
@end example

The compiler may move the addition back before the @code{volatile asm}
statement. To make it work as expected, add an artificial dependency to
the @code{asm} by referencing a variable in the subsequent code, for
example:

@example
asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv));
sum = x + y;
@end example

Under certain circumstances, GCC may duplicate (or remove duplicates of) your 
assembly code when optimizing. This can lead to unexpected duplicate symbol 
errors during compilation if your @code{asm} code defines symbols or labels. 
Using @samp{%=} 
(@pxref{AssemblerTemplate}) may help resolve this problem.

@anchor{AssemblerTemplate}
@subsubsection Assembler Template
@cindex @code{asm} assembler template

An assembler template is a literal string containing assembler instructions.
The compiler replaces tokens in the template that refer 
to inputs, outputs, and goto labels,
and then outputs the resulting string to the assembler. The 
string can contain any instructions recognized by the assembler, including 
directives. GCC does not parse the assembler instructions 
themselves and does not know what they mean or even whether they are valid 
assembler input. However, it does count the statements 
(@pxref{Size of an asm}).

You may place multiple assembler instructions together in a single @code{asm} 
string, separated by the characters normally used in assembly code for the 
system. A combination that works in most places is a newline to break the 
line, plus a tab character to move to the instruction field (written as 
@samp{\n\t}). 
Some assemblers allow semicolons as a line separator. However, note 
that some assembler dialects use semicolons to start a comment. 

Do not expect a sequence of @code{asm} statements to remain perfectly 
consecutive after compilation, even when you are using the @code{volatile} 
qualifier. If certain instructions need to remain consecutive in the output, 
put them in a single multi-instruction @code{asm} statement.

Accessing data from C programs without using input/output operands (such as 
by using global symbols directly from the assembler template) may not work as 
expected. Similarly, calling functions directly from an assembler template 
requires a detailed understanding of the target assembler and ABI.

Since GCC does not parse the assembler template,
it has no visibility of any 
symbols it references. This may result in GCC discarding those symbols as 
unreferenced unless they are also listed as input, output, or goto operands.

@subsubheading Special format strings

In addition to the tokens described by the input, output, and goto operands, 
these tokens have special meanings in the assembler template:

@table @samp
@item %% 
Outputs a single @samp{%} into the assembler code.

@item %= 
Outputs a number that is unique to each instance of the @code{asm} 
statement in the entire compilation. This option is useful when creating local 
labels and referring to them multiple times in a single template that 
generates multiple assembler instructions. 

@item %@{
@itemx %|
@itemx %@}
Outputs @samp{@{}, @samp{|}, and @samp{@}} characters (respectively)
into the assembler code.  When unescaped, these characters have special
meaning to indicate multiple assembler dialects, as described below.
@end table

@subsubheading Multiple assembler dialects in @code{asm} templates

On targets such as x86, GCC supports multiple assembler dialects.
The @option{-masm} option controls which dialect GCC uses as its 
default for inline assembler. The target-specific documentation for the 
@option{-masm} option contains the list of supported dialects, as well as the 
default dialect if the option is not specified. This information may be 
important to understand, since assembler code that works correctly when 
compiled using one dialect will likely fail if compiled using another.
@xref{x86 Options}.

If your code needs to support multiple assembler dialects (for example, if 
you are writing public headers that need to support a variety of compilation 
options), use constructs of this form:

@example
@{ dialect0 | dialect1 | dialect2... @}
@end example

This construct outputs @code{dialect0} 
when using dialect #0 to compile the code, 
@code{dialect1} for dialect #1, etc. If there are fewer alternatives within the 
braces than the number of dialects the compiler supports, the construct 
outputs nothing.

For example, if an x86 compiler supports two dialects
(@samp{att}, @samp{intel}), an 
assembler template such as this:

@example
"bt@{l %[Offset],%[Base] | %[Base],%[Offset]@}; jc %l2"
@end example

@noindent
is equivalent to one of

@example
"btl %[Offset],%[Base] ; jc %l2"   @r{/* att dialect */}
"bt %[Base],%[Offset]; jc %l2"     @r{/* intel dialect */}
@end example

Using that same compiler, this code:

@example
"xchg@{l@}\t@{%%@}ebx, %1"
@end example

@noindent
corresponds to either

@example
"xchgl\t%%ebx, %1"                 @r{/* att dialect */}
"xchg\tebx, %1"                    @r{/* intel dialect */}
@end example

There is no support for nesting dialect alternatives.

@anchor{OutputOperands}
@subsubsection Output Operands
@cindex @code{asm} output operands

An @code{asm} statement has zero or more output operands indicating the names
of C variables modified by the assembler code.

In this i386 example, @code{old} (referred to in the template string as 
@code{%0}) and @code{*Base} (as @code{%1}) are outputs and @code{Offset} 
(@code{%2}) is an input:

@example
bool old;

__asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base.
         "sbb %0,%0"      // Use the CF to calculate old.
   : "=r" (old), "+rm" (*Base)
   : "Ir" (Offset)
   : "cc");

return old;
@end example

Operands are separated by commas.  Each operand has this format:

@example
@r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cvariablename})
@end example

@table @var
@item asmSymbolicName
Specifies a symbolic name for the operand.
Reference the name in the assembler template 
by enclosing it in square brackets 
(i.e.@: @samp{%[Value]}). The scope of the name is the @code{asm} statement 
that contains the definition. Any valid C variable name is acceptable, 
including names already defined in the surrounding code. No two operands 
within the same @code{asm} statement can use the same symbolic name.

When not using an @var{asmSymbolicName}, use the (zero-based) position
of the operand 
in the list of operands in the assembler template. For example if there are 
three output operands, use @samp{%0} in the template to refer to the first, 
@samp{%1} for the second, and @samp{%2} for the third. 

@item constraint
A string constant specifying constraints on the placement of the operand; 
@xref{Constraints}, for details.

Output constraints must begin with either @samp{=} (a variable overwriting an 
existing value) or @samp{+} (when reading and writing). When using 
@samp{=}, do not assume the location contains the existing value
on entry to the @code{asm}, except 
when the operand is tied to an input; @pxref{InputOperands,,Input Operands}.

After the prefix, there must be one or more additional constraints 
(@pxref{Constraints}) that describe where the value resides. Common 
constraints include @samp{r} for register and @samp{m} for memory. 
When you list more than one possible location (for example, @code{"=rm"}),
the compiler chooses the most efficient one based on the current context. 
If you list as many alternates as the @code{asm} statement allows, you permit 
the optimizers to produce the best possible code. 
If you must use a specific register, but your Machine Constraints do not
provide sufficient control to select the specific register you want, 
local register variables may provide a solution (@pxref{Local Register 
Variables}).

@item cvariablename
Specifies a C lvalue expression to hold the output, typically a variable name.
The enclosing parentheses are a required part of the syntax.

@end table

When the compiler selects the registers to use to 
represent the output operands, it does not use any of the clobbered registers 
(@pxref{Clobbers and Scratch Registers}).

Output operand expressions must be lvalues. The compiler cannot check whether 
the operands have data types that are reasonable for the instruction being 
executed. For output expressions that are not directly addressable (for 
example a bit-field), the constraint must allow a register. In that case, GCC 
uses the register as the output of the @code{asm}, and then stores that 
register into the output. 

Operands using the @samp{+} constraint modifier count as two operands 
(that is, both as input and output) towards the total maximum of 30 operands
per @code{asm} statement.

Use the @samp{&} constraint modifier (@pxref{Modifiers}) on all output
operands that must not overlap an input.  Otherwise, 
GCC may allocate the output operand in the same register as an unrelated 
input operand, on the assumption that the assembler code consumes its 
inputs before producing outputs. This assumption may be false if the assembler 
code actually consists of more than one instruction.

The same problem can occur if one output parameter (@var{a}) allows a register 
constraint and another output parameter (@var{b}) allows a memory constraint.
The code generated by GCC to access the memory address in @var{b} can contain
registers which @emph{might} be shared by @var{a}, and GCC considers those 
registers to be inputs to the asm. As above, GCC assumes that such input
registers are consumed before any outputs are written. This assumption may 
result in incorrect behavior if the @code{asm} statement writes to @var{a}
before using
@var{b}. Combining the @samp{&} modifier with the register constraint on @var{a}
ensures that modifying @var{a} does not affect the address referenced by 
@var{b}. Otherwise, the location of @var{b} 
is undefined if @var{a} is modified before using @var{b}.

@code{asm} supports operand modifiers on operands (for example @samp{%k2} 
instead of simply @samp{%2}). Typically these qualifiers are hardware 
dependent. The list of supported modifiers for x86 is found at 
@ref{x86Operandmodifiers,x86 Operand modifiers}.

If the C code that follows the @code{asm} makes no use of any of the output 
operands, use @code{volatile} for the @code{asm} statement to prevent the 
optimizers from discarding the @code{asm} statement as unneeded 
(see @ref{Volatile}).

This code makes no use of the optional @var{asmSymbolicName}. Therefore it 
references the first output operand as @code{%0} (were there a second, it 
would be @code{%1}, etc). The number of the first input operand is one greater 
than that of the last output operand. In this i386 example, that makes 
@code{Mask} referenced as @code{%1}:

@example
uint32_t Mask = 1234;
uint32_t Index;

  asm ("bsfl %1, %0"
     : "=r" (Index)
     : "r" (Mask)
     : "cc");
@end example

That code overwrites the variable @code{Index} (@samp{=}),
placing the value in a register (@samp{r}).
Using the generic @samp{r} constraint instead of a constraint for a specific 
register allows the compiler to pick the register to use, which can result 
in more efficient code. This may not be possible if an assembler instruction 
requires a specific register.

The following i386 example uses the @var{asmSymbolicName} syntax.
It produces the 
same result as the code above, but some may consider it more readable or more 
maintainable since reordering index numbers is not necessary when adding or 
removing operands. The names @code{aIndex} and @code{aMask}
are only used in this example to emphasize which 
names get used where.
It is acceptable to reuse the names @code{Index} and @code{Mask}.

@example
uint32_t Mask = 1234;
uint32_t Index;

  asm ("bsfl %[aMask], %[aIndex]"
     : [aIndex] "=r" (Index)
     : [aMask] "r" (Mask)
     : "cc");
@end example

Here are some more examples of output operands.

@example
uint32_t c = 1;
uint32_t d;
uint32_t *e = &c;

asm ("mov %[e], %[d]"
   : [d] "=rm" (d)
   : [e] "rm" (*e));
@end example

Here, @code{d} may either be in a register or in memory. Since the compiler 
might already have the current value of the @code{uint32_t} location
pointed to by @code{e}
in a register, you can enable it to choose the best location
for @code{d} by specifying both constraints.

@anchor{FlagOutputOperands}
@subsubsection Flag Output Operands
@cindex @code{asm} flag output operands

Some targets have a special register that holds the ``flags'' for the
result of an operation or comparison.  Normally, the contents of that
register are either unmodifed by the asm, or the @code{asm} statement is
considered to clobber the contents.

On some targets, a special form of output operand exists by which
conditions in the flags register may be outputs of the asm.  The set of
conditions supported are target specific, but the general rule is that
the output variable must be a scalar integer, and the value is boolean.
When supported, the target defines the preprocessor symbol
@code{__GCC_ASM_FLAG_OUTPUTS__}.

Because of the special nature of the flag output operands, the constraint
may not include alternatives.

Most often, the target has only one flags register, and thus is an implied
operand of many instructions.  In this case, the operand should not be
referenced within the assembler template via @code{%0} etc, as there's
no corresponding text in the assembly language.

@table @asis
@item ARM
@itemx AArch64
The flag output constraints for the ARM family are of the form
@samp{=@@cc@var{cond}} where @var{cond} is one of the standard
conditions defined in the ARM ARM for @code{ConditionHolds}.

@table @code
@item eq
Z flag set, or equal
@item ne
Z flag clear or not equal
@item cs
@itemx hs
C flag set or unsigned greater than equal
@item cc
@itemx lo
C flag clear or unsigned less than
@item mi
N flag set or ``minus''
@item pl
N flag clear or ``plus''
@item vs
V flag set or signed overflow
@item vc
V flag clear
@item hi
unsigned greater than
@item ls
unsigned less than equal
@item ge
signed greater than equal
@item lt
signed less than
@item gt
signed greater than
@item le
signed less than equal
@end table

The flag output constraints are not supported in thumb1 mode.

@item x86 family
The flag output constraints for the x86 family are of the form
@samp{=@@cc@var{cond}} where @var{cond} is one of the standard
conditions defined in the ISA manual for @code{j@var{cc}} or
@code{set@var{cc}}.

@table @code
@item a
``above'' or unsigned greater than
@item ae
``above or equal'' or unsigned greater than or equal
@item b
``below'' or unsigned less than
@item be
``below or equal'' or unsigned less than or equal
@item c
carry flag set
@item e
@itemx z
``equal'' or zero flag set
@item g
signed greater than
@item ge
signed greater than or equal
@item l
signed less than
@item le
signed less than or equal
@item o
overflow flag set
@item p
parity flag set
@item s
sign flag set
@item na
@itemx nae
@itemx nb
@itemx nbe
@itemx nc
@itemx ne
@itemx ng
@itemx nge
@itemx nl
@itemx nle
@itemx no
@itemx np
@itemx ns
@itemx nz
``not'' @var{flag}, or inverted versions of those above
@end table

@end table

@anchor{InputOperands}
@subsubsection Input Operands
@cindex @code{asm} input operands
@cindex @code{asm} expressions

Input operands make values from C variables and expressions available to the 
assembly code.

Operands are separated by commas.  Each operand has this format:

@example
@r{[} [@var{asmSymbolicName}] @r{]} @var{constraint} (@var{cexpression})
@end example

@table @var
@item asmSymbolicName
Specifies a symbolic name for the operand.
Reference the name in the assembler template 
by enclosing it in square brackets 
(i.e.@: @samp{%[Value]}). The scope of the name is the @code{asm} statement 
that contains the definition. Any valid C variable name is acceptable, 
including names already defined in the surrounding code. No two operands 
within the same @code{asm} statement can use the same symbolic name.

When not using an @var{asmSymbolicName}, use the (zero-based) position
of the operand 
in the list of operands in the assembler template. For example if there are
two output operands and three inputs,
use @samp{%2} in the template to refer to the first input operand,
@samp{%3} for the second, and @samp{%4} for the third. 

@item constraint
A string constant specifying constraints on the placement of the operand; 
@xref{Constraints}, for details.

Input constraint strings may not begin with either @samp{=} or @samp{+}.
When you list more than one possible location (for example, @samp{"irm"}), 
the compiler chooses the most efficient one based on the current context.
If you must use a specific register, but your Machine Constraints do not
provide sufficient control to select the specific register you want, 
local register variables may provide a solution (@pxref{Local Register 
Variables}).

Input constraints can also be digits (for example, @code{"0"}). This indicates 
that the specified input must be in the same place as the output constraint 
at the (zero-based) index in the output constraint list. 
When using @var{asmSymbolicName} syntax for the output operands,
you may use these names (enclosed in brackets @samp{[]}) instead of digits.

@item cexpression
This is the C variable or expression being passed to the @code{asm} statement 
as input.  The enclosing parentheses are a required part of the syntax.

@end table

When the compiler selects the registers to use to represent the input 
operands, it does not use any of the clobbered registers
(@pxref{Clobbers and Scratch Registers}).

If there are no output operands but there are input operands, place two 
consecutive colons where the output operands would go:

@example
__asm__ ("some instructions"
   : /* No outputs. */
   : "r" (Offset / 8));
@end example

@strong{Warning:} Do @emph{not} modify the contents of input-only operands 
(except for inputs tied to outputs). The compiler assumes that on exit from 
the @code{asm} statement these operands contain the same values as they 
had before executing the statement. 
It is @emph{not} possible to use clobbers
to inform the compiler that the values in these inputs are changing. One 
common work-around is to tie the changing input variable to an output variable 
that never gets used. Note, however, that if the code that follows the 
@code{asm} statement makes no use of any of the output operands, the GCC 
optimizers may discard the @code{asm} statement as unneeded 
(see @ref{Volatile}).

@code{asm} supports operand modifiers on operands (for example @samp{%k2} 
instead of simply @samp{%2}). Typically these qualifiers are hardware 
dependent. The list of supported modifiers for x86 is found at 
@ref{x86Operandmodifiers,x86 Operand modifiers}.

In this example using the fictitious @code{combine} instruction, the 
constraint @code{"0"} for input operand 1 says that it must occupy the same 
location as output operand 0. Only input operands may use numbers in 
constraints, and they must each refer to an output operand. Only a number (or 
the symbolic assembler name) in the constraint can guarantee that one operand 
is in the same place as another. The mere fact that @code{foo} is the value of 
both operands is not enough to guarantee that they are in the same place in 
the generated assembler code.

@example
asm ("combine %2, %0" 
   : "=r" (foo) 
   : "0" (foo), "g" (bar));
@end example

Here is an example using symbolic names.

@example
asm ("cmoveq %1, %2, %[result]" 
   : [result] "=r"(result) 
   : "r" (test), "r" (new), "[result]" (old));
@end example

@anchor{Clobbers and Scratch Registers}
@subsubsection Clobbers and Scratch Registers
@cindex @code{asm} clobbers
@cindex @code{asm} scratch registers

While the compiler is aware of changes to entries listed in the output 
operands, the inline @code{asm} code may modify more than just the outputs. For 
example, calculations may require additional registers, or the processor may 
overwrite a register as a side effect of a particular assembler instruction. 
In order to inform the compiler of these changes, list them in the clobber 
list. Clobber list items are either register names or the special clobbers 
(listed below). Each clobber list item is a string constant 
enclosed in double quotes and separated by commas.

Clobber descriptions may not in any way overlap with an input or output 
operand. For example, you may not have an operand describing a register class 
with one member when listing that register in the clobber list. Variables 
declared to live in specific registers (@pxref{Explicit Register 
Variables}) and used 
as @code{asm} input or output operands must have no part mentioned in the 
clobber description. In particular, there is no way to specify that input 
operands get modified without also specifying them as output operands.

When the compiler selects which registers to use to represent input and output 
operands, it does not use any of the clobbered registers. As a result, 
clobbered registers are available for any use in the assembler code.

Another restriction is that the clobber list should not contain the
stack pointer register.  This is because the compiler requires the
value of the stack pointer to be the same after an @code{asm}
statement as it was on entry to the statement.  However, previous
versions of GCC did not enforce this rule and allowed the stack
pointer to appear in the list, with unclear semantics.  This behavior
is deprecated and listing the stack pointer may become an error in
future versions of GCC@.

Here is a realistic example for the VAX showing the use of clobbered 
registers: 

@example
asm volatile ("movc3 %0, %1, %2"
                   : /* No outputs. */
                   : "g" (from), "g" (to), "g" (count)
                   : "r0", "r1", "r2", "r3", "r4", "r5", "memory");
@end example

Also, there are two special clobber arguments:

@table @code
@item "cc"
The @code{"cc"} clobber indicates that the assembler code modifies the flags 
register. On some machines, GCC represents the condition codes as a specific 
hardware register; @code{"cc"} serves to name this register.
On other machines, condition code handling is different, 
and specifying @code{"cc"} has no effect. But 
it is valid no matter what the target.

@item "memory"
The @code{"memory"} clobber tells the compiler that the assembly code
performs memory 
reads or writes to items other than those listed in the input and output 
operands (for example, accessing the memory pointed to by one of the input 
parameters). To ensure memory contains correct values, GCC may need to flush 
specific register values to memory before executing the @code{asm}. Further, 
the compiler does not assume that any values read from memory before an 
@code{asm} remain unchanged after that @code{asm}; it reloads them as 
needed.  
Using the @code{"memory"} clobber effectively forms a read/write
memory barrier for the compiler.

Note that this clobber does not prevent the @emph{processor} from doing 
speculative reads past the @code{asm} statement. To prevent that, you need 
processor-specific fence instructions.

@end table

Flushing registers to memory has performance implications and may be
an issue for time-sensitive code.  You can provide better information
to GCC to avoid this, as shown in the following examples.  At a
minimum, aliasing rules allow GCC to know what memory @emph{doesn't}
need to be flushed.

Here is a fictitious sum of squares instruction, that takes two
pointers to floating point values in memory and produces a floating
point register output.
Notice that @code{x}, and @code{y} both appear twice in the @code{asm}
parameters, once to specify memory accessed, and once to specify a
base register used by the @code{asm}.  You won't normally be wasting a
register by doing this as GCC can use the same register for both
purposes.  However, it would be foolish to use both @code{%1} and
@code{%3} for @code{x} in this @code{asm} and expect them to be the
same.  In fact, @code{%3} may well not be a register.  It might be a
symbolic memory reference to the object pointed to by @code{x}.

@smallexample
asm ("sumsq %0, %1, %2"
     : "+f" (result)
     : "r" (x), "r" (y), "m" (*x), "m" (*y));
@end smallexample

Here is a fictitious @code{*z++ = *x++ * *y++} instruction.
Notice that the @code{x}, @code{y} and @code{z} pointer registers
must be specified as input/output because the @code{asm} modifies
them.

@smallexample
asm ("vecmul %0, %1, %2"
     : "+r" (z), "+r" (x), "+r" (y), "=m" (*z)
     : "m" (*x), "m" (*y));
@end smallexample

An x86 example where the string memory argument is of unknown length.

@smallexample
asm("repne scasb"
    : "=c" (count), "+D" (p)
    : "m" (*(const char (*)[]) p), "0" (-1), "a" (0));
@end smallexample

If you know the above will only be reading a ten byte array then you
could instead use a memory input like:
@code{"m" (*(const char (*)[10]) p)}.

Here is an example of a PowerPC vector scale implemented in assembly,
complete with vector and condition code clobbers, and some initialized
offset registers that are unchanged by the @code{asm}.

@smallexample
void
dscal (size_t n, double *x, double alpha)
@{
  asm ("/* lots of asm here */"
       : "+m" (*(double (*)[n]) x), "+&r" (n), "+b" (x)
       : "d" (alpha), "b" (32), "b" (48), "b" (64),
         "b" (80), "b" (96), "b" (112)
       : "cr0",
         "vs32","vs33","vs34","vs35","vs36","vs37","vs38","vs39",
         "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47");
@}
@end smallexample

Rather than allocating fixed registers via clobbers to provide scratch
registers for an @code{asm} statement, an alternative is to define a
variable and make it an early-clobber output as with @code{a2} and
@code{a3} in the example below.  This gives the compiler register
allocator more freedom.  You can also define a variable and make it an
output tied to an input as with @code{a0} and @code{a1}, tied
respectively to @code{ap} and @code{lda}.  Of course, with tied
outputs your @code{asm} can't use the input value after modifying the
output register since they are one and the same register.  What's
more, if you omit the early-clobber on the output, it is possible that
GCC might allocate the same register to another of the inputs if GCC
could prove they had the same value on entry to the @code{asm}.  This
is why @code{a1} has an early-clobber.  Its tied input, @code{lda}
might conceivably be known to have the value 16 and without an
early-clobber share the same register as @code{%11}.  On the other
hand, @code{ap} can't be the same as any of the other inputs, so an
early-clobber on @code{a0} is not needed.  It is also not desirable in
this case.  An early-clobber on @code{a0} would cause GCC to allocate
a separate register for the @code{"m" (*(const double (*)[]) ap)}
input.  Note that tying an input to an output is the way to set up an
initialized temporary register modified by an @code{asm} statement.
An input not tied to an output is assumed by GCC to be unchanged, for
example @code{"b" (16)} below sets up @code{%11} to 16, and GCC might
use that register in following code if the value 16 happened to be
needed.  You can even use a normal @code{asm} output for a scratch if
all inputs that might share the same register are consumed before the
scratch is used.  The VSX registers clobbered by the @code{asm}
statement could have used this technique except for GCC's limit on the
number of @code{asm} parameters.

@smallexample
static void
dgemv_kernel_4x4 (long n, const double *ap, long lda,
                  const double *x, double *y, double alpha)
@{
  double *a0;
  double *a1;
  double *a2;
  double *a3;

  __asm__
    (
     /* lots of asm here */
     "#n=%1 ap=%8=%12 lda=%13 x=%7=%10 y=%0=%2 alpha=%9 o16=%11\n"
     "#a0=%3 a1=%4 a2=%5 a3=%6"
     :
       "+m" (*(double (*)[n]) y),
       "+&r" (n),	// 1
       "+b" (y),	// 2
       "=b" (a0),	// 3
       "=&b" (a1),	// 4
       "=&b" (a2),	// 5
       "=&b" (a3)	// 6
     :
       "m" (*(const double (*)[n]) x),
       "m" (*(const double (*)[]) ap),
       "d" (alpha),	// 9
       "r" (x),		// 10
       "b" (16),	// 11
       "3" (ap),	// 12
       "4" (lda)	// 13
     :
       "cr0",
       "vs32","vs33","vs34","vs35","vs36","vs37",
       "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47"
     );
@}
@end smallexample

@anchor{GotoLabels}
@subsubsection Goto Labels
@cindex @code{asm} goto labels

@code{asm goto} allows assembly code to jump to one or more C labels.  The
@var{GotoLabels} section in an @code{asm goto} statement contains 
a comma-separated 
list of all C labels to which the assembler code may jump. GCC assumes that 
@code{asm} execution falls through to the next statement (if this is not the 
case, consider using the @code{__builtin_unreachable} intrinsic after the 
@code{asm} statement). Optimization of @code{asm goto} may be improved by 
using the @code{hot} and @code{cold} label attributes (@pxref{Label 
Attributes}).

An @code{asm goto} statement cannot have outputs.
This is due to an internal restriction of 
the compiler: control transfer instructions cannot have outputs. 
If the assembler code does modify anything, use the @code{"memory"} clobber 
to force the 
optimizers to flush all register values to memory and reload them if 
necessary after the @code{asm} statement.

Also note that an @code{asm goto} statement is always implicitly
considered volatile.

To reference a label in the assembler template,
prefix it with @samp{%l} (lowercase @samp{L}) followed 
by its (zero-based) position in @var{GotoLabels} plus the number of input 
operands.  For example, if the @code{asm} has three inputs and references two 
labels, refer to the first label as @samp{%l3} and the second as @samp{%l4}).

Alternately, you can reference labels using the actual C label name enclosed
in brackets.  For example, to reference a label named @code{carry}, you can
use @samp{%l[carry]}.  The label must still be listed in the @var{GotoLabels}
section when using this approach.

Here is an example of @code{asm goto} for i386:

@example
asm goto (
    "btl %1, %0\n\t"
    "jc %l2"
    : /* No outputs. */
    : "r" (p1), "r" (p2) 
    : "cc" 
    : carry);

return 0;

carry:
return 1;
@end example

The following example shows an @code{asm goto} that uses a memory clobber.

@example
int frob(int x)
@{
  int y;
  asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5"
            : /* No outputs. */
            : "r"(x), "r"(&y)
            : "r5", "memory" 
            : error);
  return y;
error:
  return -1;
@}
@end example

@anchor{x86Operandmodifiers}
@subsubsection x86 Operand Modifiers

References to input, output, and goto operands in the assembler template
of extended @code{asm} statements can use 
modifiers to affect the way the operands are formatted in 
the code output to the assembler. For example, the 
following code uses the @samp{h} and @samp{b} modifiers for x86:

@example
uint16_t  num;
asm volatile ("xchg %h0, %b0" : "+a" (num) );
@end example

@noindent
These modifiers generate this assembler code:

@example
xchg %ah, %al
@end example

The rest of this discussion uses the following code for illustrative purposes.

@example
int main()
@{
   int iInt = 1;

top:

   asm volatile goto ("some assembler instructions here"
   : /* No outputs. */
   : "q" (iInt), "X" (sizeof(unsigned char) + 1), "i" (42)
   : /* No clobbers. */
   : top);
@}
@end example

With no modifiers, this is what the output from the operands would be
for the @samp{att} and @samp{intel} dialects of assembler:

@multitable {Operand} {$.L2} {OFFSET FLAT:.L2}
@headitem Operand @tab @samp{att} @tab @samp{intel}
@item @code{%0}
@tab @code{%eax}
@tab @code{eax}
@item @code{%1}
@tab @code{$2}
@tab @code{2}
@item @code{%3}
@tab @code{$.L3}
@tab @code{OFFSET FLAT:.L3}
@end multitable

The table below shows the list of supported modifiers and their effects.

@multitable {Modifier} {Print the opcode suffix for the size of th} {Operand} {@samp{att}} {@samp{intel}}
@headitem Modifier @tab Description @tab Operand @tab @samp{att} @tab @samp{intel}
@item @code{a}
@tab Print an absolute memory reference.
@tab @code{%A0}
@tab @code{*%rax}
@tab @code{rax}
@item @code{b}
@tab Print the QImode name of the register.
@tab @code{%b0}
@tab @code{%al}
@tab @code{al}
@item @code{c}
@tab Require a constant operand and print the constant expression with no punctuation.
@tab @code{%c1}
@tab @code{2}
@tab @code{2}
@item @code{E}
@tab Print the address in Double Integer (DImode) mode (8 bytes) when the target is 64-bit.
Otherwise mode is unspecified (VOIDmode).
@tab @code{%E1}
@tab @code{%(rax)}
@tab @code{[rax]}
@item @code{h}
@tab Print the QImode name for a ``high'' register.
@tab @code{%h0}
@tab @code{%ah}
@tab @code{ah}
@item @code{H}
@tab Add 8 bytes to an offsettable memory reference. Useful when accessing the
high 8 bytes of SSE values. For a memref in (%rax), it generates
@tab @code{%H0}
@tab @code{8(%rax)}
@tab @code{8[rax]}
@item @code{k}
@tab Print the SImode name of the register.
@tab @code{%k0}
@tab @code{%eax}
@tab @code{eax}
@item @code{l}
@tab Print the label name with no punctuation.
@tab @code{%l3}
@tab @code{.L3}
@tab @code{.L3}
@item @code{p}
@tab Print raw symbol name (without syntax-specific prefixes).
@tab @code{%p2}
@tab @code{42}
@tab @code{42}
@item @code{P}
@tab If used for a function, print the PLT suffix and generate PIC code.
For example, emit @code{foo@@PLT} instead of 'foo' for the function
foo(). If used for a constant, drop all syntax-specific prefixes and
issue the bare constant. See @code{p} above.
@item @code{q}
@tab Print the DImode name of the register.
@tab @code{%q0}
@tab @code{%rax}
@tab @code{rax}
@item @code{w}
@tab Print the HImode name of the register.
@tab @code{%w0}
@tab @code{%ax}
@tab @code{ax}
@item @code{z}
@tab Print the opcode suffix for the size of the current integer operand (one of @code{b}/@code{w}/@code{l}/@code{q}).
@tab @code{%z0}
@tab @code{l}
@tab 
@end multitable

@code{V} is a special modifier which prints the name of the full integer
register without @code{%}.

@anchor{x86floatingpointasmoperands}
@subsubsection x86 Floating-Point @code{asm} Operands

On x86 targets, there are several rules on the usage of stack-like registers
in the operands of an @code{asm}.  These rules apply only to the operands
that are stack-like registers:

@enumerate
@item
Given a set of input registers that die in an @code{asm}, it is
necessary to know which are implicitly popped by the @code{asm}, and
which must be explicitly popped by GCC@.

An input register that is implicitly popped by the @code{asm} must be
explicitly clobbered, unless it is constrained to match an
output operand.

@item
For any input register that is implicitly popped by an @code{asm}, it is
necessary to know how to adjust the stack to compensate for the pop.
If any non-popped input is closer to the top of the reg-stack than
the implicitly popped register, it would not be possible to know what the
stack looked like---it's not clear how the rest of the stack ``slides
up''.

All implicitly popped input registers must be closer to the top of
the reg-stack than any input that is not implicitly popped.

It is possible that if an input dies in an @code{asm}, the compiler might
use the input register for an output reload.  Consider this example:

@smallexample
asm ("foo" : "=t" (a) : "f" (b));
@end smallexample

@noindent
This code says that input @code{b} is not popped by the @code{asm}, and that
the @code{asm} pushes a result onto the reg-stack, i.e., the stack is one
deeper after the @code{asm} than it was before.  But, it is possible that
reload may think that it can use the same register for both the input and
the output.

To prevent this from happening,
if any input operand uses the @samp{f} constraint, all output register
constraints must use the @samp{&} early-clobber modifier.

The example above is correctly written as:

@smallexample
asm ("foo" : "=&t" (a) : "f" (b));
@end smallexample

@item
Some operands need to be in particular places on the stack.  All
output operands fall in this category---GCC has no other way to
know which registers the outputs appear in unless you indicate
this in the constraints.

Output operands must specifically indicate which register an output
appears in after an @code{asm}.  @samp{=f} is not allowed: the operand
constraints must select a class with a single register.

@item
Output operands may not be ``inserted'' between existing stack registers.
Since no 387 opcode uses a read/write operand, all output operands
are dead before the @code{asm}, and are pushed by the @code{asm}.
It makes no sense to push anywhere but the top of the reg-stack.

Output operands must start at the top of the reg-stack: output
operands may not ``skip'' a register.

@item
Some @code{asm} statements may need extra stack space for internal
calculations.  This can be guaranteed by clobbering stack registers
unrelated to the inputs and outputs.

@end enumerate

This @code{asm}
takes one input, which is internally popped, and produces two outputs.

@smallexample
asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
@end smallexample

@noindent
This @code{asm} takes two inputs, which are popped by the @code{fyl2xp1} opcode,
and replaces them with one output.  The @code{st(1)} clobber is necessary 
for the compiler to know that @code{fyl2xp1} pops both inputs.

@smallexample
asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
@end smallexample

@lowersections
@include md.texi
@raisesections

@node Asm Labels
@subsection Controlling Names Used in Assembler Code
@cindex assembler names for identifiers
@cindex names used in assembler code
@cindex identifiers, names in assembler code

You can specify the name to be used in the assembler code for a C
function or variable by writing the @code{asm} (or @code{__asm__})
keyword after the declarator.
It is up to you to make sure that the assembler names you choose do not
conflict with any other assembler symbols, or reference registers.

@subsubheading Assembler names for data:

This sample shows how to specify the assembler name for data:

@smallexample
int foo asm ("myfoo") = 2;
@end smallexample

@noindent
This specifies that the name to be used for the variable @code{foo} in
the assembler code should be @samp{myfoo} rather than the usual
@samp{_foo}.

On systems where an underscore is normally prepended to the name of a C
variable, this feature allows you to define names for the
linker that do not start with an underscore.

GCC does not support using this feature with a non-static local variable 
since such variables do not have assembler names.  If you are
trying to put the variable in a particular register, see 
@ref{Explicit Register Variables}.

@subsubheading Assembler names for functions:

To specify the assembler name for functions, write a declaration for the 
function before its definition and put @code{asm} there, like this:

@smallexample
int func (int x, int y) asm ("MYFUNC");
     
int func (int x, int y)
@{
   /* @r{@dots{}} */
@end smallexample

@noindent
This specifies that the name to be used for the function @code{func} in
the assembler code should be @code{MYFUNC}.

@node Explicit Register Variables
@subsection Variables in Specified Registers
@anchor{Explicit Reg Vars}
@cindex explicit register variables
@cindex variables in specified registers
@cindex specified registers

GNU C allows you to associate specific hardware registers with C 
variables.  In almost all cases, allowing the compiler to assign
registers produces the best code.  However under certain unusual
circumstances, more precise control over the variable storage is 
required.

Both global and local variables can be associated with a register.  The
consequences of performing this association are very different between
the two, as explained in the sections below.

@menu
* Global Register Variables::   Variables declared at global scope.
* Local Register Variables::    Variables declared within a function.
@end menu

@node Global Register Variables
@subsubsection Defining Global Register Variables
@anchor{Global Reg Vars}
@cindex global register variables
@cindex registers, global variables in
@cindex registers, global allocation

You can define a global register variable and associate it with a specified 
register like this:

@smallexample
register int *foo asm ("r12");
@end smallexample

@noindent
Here @code{r12} is the name of the register that should be used. Note that 
this is the same syntax used for defining local register variables, but for 
a global variable the declaration appears outside a function. The 
@code{register} keyword is required, and cannot be combined with 
@code{static}. The register name must be a valid register name for the
target platform.

Do not use type qualifiers such as @code{const} and @code{volatile}, as
the outcome may be contrary to expectations.  In  particular, using the
@code{volatile} qualifier does not fully prevent the compiler from
optimizing accesses to the register.

Registers are a scarce resource on most systems and allowing the 
compiler to manage their usage usually results in the best code. However, 
under special circumstances it can make sense to reserve some globally.
For example this may be useful in programs such as programming language 
interpreters that have a couple of global variables that are accessed 
very often.

After defining a global register variable, for the current compilation
unit:

@itemize @bullet
@item If the register is a call-saved register, call ABI is affected:
the register will not be restored in function epilogue sequences after
the variable has been assigned.  Therefore, functions cannot safely
return to callers that assume standard ABI.
@item Conversely, if the register is a call-clobbered register, making
calls to functions that use standard ABI may lose contents of the variable.
Such calls may be created by the compiler even if none are evident in
the original program, for example when libgcc functions are used to
make up for unavailable instructions.
@item Accesses to the variable may be optimized as usual and the register
remains available for allocation and use in any computations, provided that
observable values of the variable are not affected.
@item If the variable is referenced in inline assembly, the type of access
must be provided to the compiler via constraints (@pxref{Constraints}).
Accesses from basic asms are not supported.
@end itemize

Note that these points @emph{only} apply to code that is compiled with the
definition. The behavior of code that is merely linked in (for example 
code from libraries) is not affected.

If you want to recompile source files that do not actually use your global 
register variable so they do not use the specified register for any other 
purpose, you need not actually add the global register declaration to 
their source code. It suffices to specify the compiler option 
@option{-ffixed-@var{reg}} (@pxref{Code Gen Options}) to reserve the 
register.

@subsubheading Declaring the variable

Global register variables cannot have initial values, because an
executable file has no means to supply initial contents for a register.

When selecting a register, choose one that is normally saved and 
restored by function calls on your machine. This ensures that code
which is unaware of this reservation (such as library routines) will 
restore it before returning.

On machines with register windows, be sure to choose a global
register that is not affected magically by the function call mechanism.

@subsubheading Using the variable

@cindex @code{qsort}, and global register variables
When calling routines that are not aware of the reservation, be 
cautious if those routines call back into code which uses them. As an 
example, if you call the system library version of @code{qsort}, it may 
clobber your registers during execution, but (if you have selected 
appropriate registers) it will restore them before returning. However 
it will @emph{not} restore them before calling @code{qsort}'s comparison 
function. As a result, global values will not reliably be available to 
the comparison function unless the @code{qsort} function itself is rebuilt.

Similarly, it is not safe to access the global register variables from signal
handlers or from more than one thread of control. Unless you recompile 
them specially for the task at hand, the system library routines may 
temporarily use the register for other things.  Furthermore, since the register
is not reserved exclusively for the variable, accessing it from handlers of
asynchronous signals may observe unrelated temporary values residing in the
register.

@cindex register variable after @code{longjmp}
@cindex global register after @code{longjmp}
@cindex value after @code{longjmp}
@findex longjmp
@findex setjmp
On most machines, @code{longjmp} restores to each global register
variable the value it had at the time of the @code{setjmp}. On some
machines, however, @code{longjmp} does not change the value of global
register variables. To be portable, the function that called @code{setjmp}
should make other arrangements to save the values of the global register
variables, and to restore them in a @code{longjmp}. This way, the same
thing happens regardless of what @code{longjmp} does.

@node Local Register Variables
@subsubsection Specifying Registers for Local Variables
@anchor{Local Reg Vars}
@cindex local variables, specifying registers
@cindex specifying registers for local variables
@cindex registers for local variables

You can define a local register variable and associate it with a specified 
register like this:

@smallexample
register int *foo asm ("r12");
@end smallexample

@noindent
Here @code{r12} is the name of the register that should be used.  Note
that this is the same syntax used for defining global register variables, 
but for a local variable the declaration appears within a function.  The 
@code{register} keyword is required, and cannot be combined with 
@code{static}.  The register name must be a valid register name for the
target platform.

Do not use type qualifiers such as @code{const} and @code{volatile}, as
the outcome may be contrary to expectations. In particular, when the
@code{const} qualifier is used, the compiler may substitute the
variable with its initializer in @code{asm} statements, which may cause
the corresponding operand to appear in a different register.

As with global register variables, it is recommended that you choose 
a register that is normally saved and restored by function calls on your 
machine, so that calls to library routines will not clobber it.

The only supported use for this feature is to specify registers
for input and output operands when calling Extended @code{asm} 
(@pxref{Extended Asm}).  This may be necessary if the constraints for a 
particular machine don't provide sufficient control to select the desired 
register.  To force an operand into a register, create a local variable 
and specify the register name after the variable's declaration.  Then use 
the local variable for the @code{asm} operand and specify any constraint 
letter that matches the register:

@smallexample
register int *p1 asm ("r0") = @dots{};
register int *p2 asm ("r1") = @dots{};
register int *result asm ("r0");
asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
@end smallexample

@emph{Warning:} In the above example, be aware that a register (for example 
@code{r0}) can be call-clobbered by subsequent code, including function 
calls and library calls for arithmetic operators on other variables (for 
example the initialization of @code{p2}).  In this case, use temporary 
variables for expressions between the register assignments:

@smallexample
int t1 = @dots{};
register int *p1 asm ("r0") = @dots{};
register int *p2 asm ("r1") = t1;
register int *result asm ("r0");
asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
@end smallexample

Defining a register variable does not reserve the register.  Other than
when invoking the Extended @code{asm}, the contents of the specified 
register are not guaranteed.  For this reason, the following uses 
are explicitly @emph{not} supported.  If they appear to work, it is only 
happenstance, and may stop working as intended due to (seemingly) 
unrelated changes in surrounding code, or even minor changes in the 
optimization of a future version of gcc:

@itemize @bullet
@item Passing parameters to or from Basic @code{asm}
@item Passing parameters to or from Extended @code{asm} without using input 
or output operands.
@item Passing parameters to or from routines written in assembler (or
other languages) using non-standard calling conventions.
@end itemize

Some developers use Local Register Variables in an attempt to improve 
gcc's allocation of registers, especially in large functions.  In this 
case the register name is essentially a hint to the register allocator.
While in some instances this can generate better code, improvements are
subject to the whims of the allocator/optimizers.  Since there are no
guarantees that your improvements won't be lost, this usage of Local
Register Variables is discouraged.

On the MIPS platform, there is related use for local register variables 
with slightly different characteristics (@pxref{MIPS Coprocessors,, 
Defining coprocessor specifics for MIPS targets, gccint, 
GNU Compiler Collection (GCC) Internals}).

@node Size of an asm
@subsection Size of an @code{asm}

Some targets require that GCC track the size of each instruction used
in order to generate correct code.  Because the final length of the
code produced by an @code{asm} statement is only known by the
assembler, GCC must make an estimate as to how big it will be.  It
does this by counting the number of instructions in the pattern of the
@code{asm} and multiplying that by the length of the longest
instruction supported by that processor.  (When working out the number
of instructions, it assumes that any occurrence of a newline or of
whatever statement separator character is supported by the assembler ---
typically @samp{;} --- indicates the end of an instruction.)

Normally, GCC's estimate is adequate to ensure that correct
code is generated, but it is possible to confuse the compiler if you use
pseudo instructions or assembler macros that expand into multiple real
instructions, or if you use assembler directives that expand to more
space in the object file than is needed for a single instruction.
If this happens then the assembler may produce a diagnostic saying that
a label is unreachable.

@cindex @code{asm inline}
This size is also used for inlining decisions.  If you use @code{asm inline}
instead of just @code{asm}, then for inlining purposes the size of the asm
is taken as the minimum size, ignoring how many instructions GCC thinks it is.

@node Alternate Keywords
@section Alternate Keywords
@cindex alternate keywords
@cindex keywords, alternate

@option{-ansi} and the various @option{-std} options disable certain
keywords.  This causes trouble when you want to use GNU C extensions, or
a general-purpose header file that should be usable by all programs,
including ISO C programs.  The keywords @code{asm}, @code{typeof} and
@code{inline} are not available in programs compiled with
@option{-ansi} or @option{-std} (although @code{inline} can be used in a
program compiled with @option{-std=c99} or a later standard).  The
ISO C99 keyword
@code{restrict} is only available when @option{-std=gnu99} (which will
eventually be the default) or @option{-std=c99} (or the equivalent
@option{-std=iso9899:1999}), or an option for a later standard
version, is used.

The way to solve these problems is to put @samp{__} at the beginning and
end of each problematical keyword.  For example, use @code{__asm__}
instead of @code{asm}, and @code{__inline__} instead of @code{inline}.

Other C compilers won't accept these alternative keywords; if you want to
compile with another compiler, you can define the alternate keywords as
macros to replace them with the customary keywords.  It looks like this:

@smallexample
#ifndef __GNUC__
#define __asm__ asm
#endif
@end smallexample

@findex __extension__
@opindex pedantic
@option{-pedantic} and other options cause warnings for many GNU C extensions.
You can
prevent such warnings within one expression by writing
@code{__extension__} before the expression.  @code{__extension__} has no
effect aside from this.

@node Incomplete Enums
@section Incomplete @code{enum} Types

You can define an @code{enum} tag without specifying its possible values.
This results in an incomplete type, much like what you get if you write
@code{struct foo} without describing the elements.  A later declaration
that does specify the possible values completes the type.

You cannot allocate variables or storage using the type while it is
incomplete.  However, you can work with pointers to that type.

This extension may not be very useful, but it makes the handling of
@code{enum} more consistent with the way @code{struct} and @code{union}
are handled.

This extension is not supported by GNU C++.

@node Function Names
@section Function Names as Strings
@cindex @code{__func__} identifier
@cindex @code{__FUNCTION__} identifier
@cindex @code{__PRETTY_FUNCTION__} identifier

GCC provides three magic constants that hold the name of the current
function as a string.  In C++11 and later modes, all three are treated
as constant expressions and can be used in @code{constexpr} constexts.
The first of these constants is @code{__func__}, which is part of
the C99 standard:

The identifier @code{__func__} is implicitly declared by the translator
as if, immediately following the opening brace of each function
definition, the declaration

@smallexample
static const char __func__[] = "function-name";
@end smallexample

@noindent
appeared, where function-name is the name of the lexically-enclosing
function.  This name is the unadorned name of the function.  As an
extension, at file (or, in C++, namespace scope), @code{__func__}
evaluates to the empty string.

@code{__FUNCTION__} is another name for @code{__func__}, provided for
backward compatibility with old versions of GCC.

In C, @code{__PRETTY_FUNCTION__} is yet another name for
@code{__func__}, except that at file scope (or, in C++, namespace scope),
it evaluates to the string @code{"top level"}.  In addition, in C++,
@code{__PRETTY_FUNCTION__} contains the signature of the function as
well as its bare name.  For example, this program:

@smallexample
extern "C" int printf (const char *, ...);

class a @{
 public:
  void sub (int i)
    @{
      printf ("__FUNCTION__ = %s\n", __FUNCTION__);
      printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
    @}
@};

int
main (void)
@{
  a ax;
  ax.sub (0);
  return 0;
@}
@end smallexample

@noindent
gives this output:

@smallexample
__FUNCTION__ = sub
__PRETTY_FUNCTION__ = void a::sub(int)
@end smallexample

These identifiers are variables, not preprocessor macros, and may not
be used to initialize @code{char} arrays or be concatenated with string
literals.

@node Return Address
@section Getting the Return or Frame Address of a Function

These functions may be used to get information about the callers of a
function.

@deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
This function returns the return address of the current function, or of
one of its callers.  The @var{level} argument is number of frames to
scan up the call stack.  A value of @code{0} yields the return address
of the current function, a value of @code{1} yields the return address
of the caller of the current function, and so forth.  When inlining
the expected behavior is that the function returns the address of
the function that is returned to.  To work around this behavior use
the @code{noinline} function attribute.

The @var{level} argument must be a constant integer.

On some machines it may be impossible to determine the return address of
any function other than the current one; in such cases, or when the top
of the stack has been reached, this function returns @code{0} or a
random value.  In addition, @code{__builtin_frame_address} may be used
to determine if the top of the stack has been reached.

Additional post-processing of the returned value may be needed, see
@code{__builtin_extract_return_addr}.

Calling this function with a nonzero argument can have unpredictable
effects, including crashing the calling program.  As a result, calls
that are considered unsafe are diagnosed when the @option{-Wframe-address}
option is in effect.  Such calls should only be made in debugging
situations.
@end deftypefn

@deftypefn {Built-in Function} {void *} __builtin_extract_return_addr (void *@var{addr})
The address as returned by @code{__builtin_return_address} may have to be fed
through this function to get the actual encoded address.  For example, on the
31-bit S/390 platform the highest bit has to be masked out, or on SPARC
platforms an offset has to be added for the true next instruction to be
executed.

If no fixup is needed, this function simply passes through @var{addr}.
@end deftypefn

@deftypefn {Built-in Function} {void *} __builtin_frob_return_addr (void *@var{addr})
This function does the reverse of @code{__builtin_extract_return_addr}.
@end deftypefn

@deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
This function is similar to @code{__builtin_return_address}, but it
returns the address of the function frame rather than the return address
of the function.  Calling @code{__builtin_frame_address} with a value of
@code{0} yields the frame address of the current function, a value of
@code{1} yields the frame address of the caller of the current function,
and so forth.

The frame is the area on the stack that holds local variables and saved
registers.  The frame address is normally the address of the first word
pushed on to the stack by the function.  However, the exact definition
depends upon the processor and the calling convention.  If the processor
has a dedicated frame pointer register, and the function has a frame,
then @code{__builtin_frame_address} returns the value of the frame
pointer register.

On some machines it may be impossible to determine the frame address of
any function other than the current one; in such cases, or when the top
of the stack has been reached, this function returns @code{0} if
the first frame pointer is properly initialized by the startup code.

Calling this function with a nonzero argument can have unpredictable
effects, including crashing the calling program.  As a result, calls
that are considered unsafe are diagnosed when the @option{-Wframe-address}
option is in effect.  Such calls should only be made in debugging
situations.
@end deftypefn

@node Vector Extensions
@section Using Vector Instructions through Built-in Functions

On some targets, the instruction set contains SIMD vector instructions which
operate on multiple values contained in one large register at the same time.
For example, on the x86 the MMX, 3DNow!@: and SSE extensions can be used
this way.

The first step in using these extensions is to provide the necessary data
types.  This should be done using an appropriate @code{typedef}:

@smallexample
typedef int v4si __attribute__ ((vector_size (16)));
@end smallexample

@noindent
The @code{int} type specifies the @dfn{base type}, while the attribute specifies
the vector size for the variable, measured in bytes.  For example, the
declaration above causes the compiler to set the mode for the @code{v4si}
type to be 16 bytes wide and divided into @code{int} sized units.  For
a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the
corresponding mode of @code{foo} is @acronym{V4SI}.

The @code{vector_size} attribute is only applicable to integral and
floating scalars, although arrays, pointers, and function return values
are allowed in conjunction with this construct. Only sizes that are
positive power-of-two multiples of the base type size are currently allowed.

All the basic integer types can be used as base types, both as signed
and as unsigned: @code{char}, @code{short}, @code{int}, @code{long},
@code{long long}.  In addition, @code{float} and @code{double} can be
used to build floating-point vector types.

Specifying a combination that is not valid for the current architecture
causes GCC to synthesize the instructions using a narrower mode.
For example, if you specify a variable of type @code{V4SI} and your
architecture does not allow for this specific SIMD type, GCC
produces code that uses 4 @code{SIs}.

The types defined in this manner can be used with a subset of normal C
operations.  Currently, GCC allows using the following operators
on these types: @code{+, -, *, /, unary minus, ^, |, &, ~, %}@.

The operations behave like C++ @code{valarrays}.  Addition is defined as
the addition of the corresponding elements of the operands.  For
example, in the code below, each of the 4 elements in @var{a} is
added to the corresponding 4 elements in @var{b} and the resulting
vector is stored in @var{c}.

@smallexample
typedef int v4si __attribute__ ((vector_size (16)));

v4si a, b, c;

c = a + b;
@end smallexample

Subtraction, multiplication, division, and the logical operations
operate in a similar manner.  Likewise, the result of using the unary
minus or complement operators on a vector type is a vector whose
elements are the negative or complemented values of the corresponding
elements in the operand.

It is possible to use shifting operators @code{<<}, @code{>>} on
integer-type vectors. The operation is defined as following: @code{@{a0,
a1, @dots{}, an@} >> @{b0, b1, @dots{}, bn@} == @{a0 >> b0, a1 >> b1,
@dots{}, an >> bn@}}@. Vector operands must have the same number of
elements. 

For convenience, it is allowed to use a binary vector operation
where one operand is a scalar. In that case the compiler transforms
the scalar operand into a vector where each element is the scalar from
the operation. The transformation happens only if the scalar could be
safely converted to the vector-element type.
Consider the following code.

@smallexample
typedef int v4si __attribute__ ((vector_size (16)));

v4si a, b, c;
long l;

a = b + 1;    /* a = b + @{1,1,1,1@}; */
a = 2 * b;    /* a = @{2,2,2,2@} * b; */

a = l + a;    /* Error, cannot convert long to int. */
@end smallexample

Vectors can be subscripted as if the vector were an array with
the same number of elements and base type.  Out of bound accesses
invoke undefined behavior at run time.  Warnings for out of bound
accesses for vector subscription can be enabled with
@option{-Warray-bounds}.

Vector comparison is supported with standard comparison
operators: @code{==, !=, <, <=, >, >=}. Comparison operands can be
vector expressions of integer-type or real-type. Comparison between
integer-type vectors and real-type vectors are not supported.  The
result of the comparison is a vector of the same width and number of
elements as the comparison operands with a signed integral element
type.

Vectors are compared element-wise producing 0 when comparison is false
and -1 (constant of the appropriate type where all bits are set)
otherwise. Consider the following example.

@smallexample
typedef int v4si __attribute__ ((vector_size (16)));

v4si a = @{1,2,3,4@};
v4si b = @{3,2,1,4@};
v4si c;

c = a >  b;     /* The result would be @{0, 0,-1, 0@}  */
c = a == b;     /* The result would be @{0,-1, 0,-1@}  */
@end smallexample

In C++, the ternary operator @code{?:} is available. @code{a?b:c}, where
@code{b} and @code{c} are vectors of the same type and @code{a} is an
integer vector with the same number of elements of the same size as @code{b}
and @code{c}, computes all three arguments and creates a vector
@code{@{a[0]?b[0]:c[0], a[1]?b[1]:c[1], @dots{}@}}.  Note that unlike in
OpenCL, @code{a} is thus interpreted as @code{a != 0} and not @code{a < 0}.
As in the case of binary operations, this syntax is also accepted when
one of @code{b} or @code{c} is a scalar that is then transformed into a
vector. If both @code{b} and @code{c} are scalars and the type of
@code{true?b:c} has the same size as the element type of @code{a}, then
@code{b} and @code{c} are converted to a vector type whose elements have
this type and with the same number of elements as @code{a}.

In C++, the logic operators @code{!, &&, ||} are available for vectors.
@code{!v} is equivalent to @code{v == 0}, @code{a && b} is equivalent to
@code{a!=0 & b!=0} and @code{a || b} is equivalent to @code{a!=0 | b!=0}.
For mixed operations between a scalar @code{s} and a vector @code{v},
@code{s && v} is equivalent to @code{s?v!=0:0} (the evaluation is
short-circuit) and @code{v && s} is equivalent to @code{v!=0 & (s?-1:0)}.

@findex __builtin_shuffle
Vector shuffling is available using functions
@code{__builtin_shuffle (vec, mask)} and
@code{__builtin_shuffle (vec0, vec1, mask)}.
Both functions construct a permutation of elements from one or two
vectors and return a vector of the same type as the input vector(s).
The @var{mask} is an integral vector with the same width (@var{W})
and element count (@var{N}) as the output vector.

The elements of the input vectors are numbered in memory ordering of
@var{vec0} beginning at 0 and @var{vec1} beginning at @var{N}.  The
elements of @var{mask} are considered modulo @var{N} in the single-operand
case and modulo @math{2*@var{N}} in the two-operand case.

Consider the following example,

@smallexample
typedef int v4si __attribute__ ((vector_size (16)));

v4si a = @{1,2,3,4@};
v4si b = @{5,6,7,8@};
v4si mask1 = @{0,1,1,3@};
v4si mask2 = @{0,4,2,5@};
v4si res;

res = __builtin_shuffle (a, mask1);       /* res is @{1,2,2,4@}  */
res = __builtin_shuffle (a, b, mask2);    /* res is @{1,5,3,6@}  */
@end smallexample

Note that @code{__builtin_shuffle} is intentionally semantically
compatible with the OpenCL @code{shuffle} and @code{shuffle2} functions.

You can declare variables and use them in function calls and returns, as
well as in assignments and some casts.  You can specify a vector type as
a return type for a function.  Vector types can also be used as function
arguments.  It is possible to cast from one vector type to another,
provided they are of the same size (in fact, you can also cast vectors
to and from other datatypes of the same size).

You cannot operate between vectors of different lengths or different
signedness without a cast.

@findex __builtin_convertvector
Vector conversion is available using the
@code{__builtin_convertvector (vec, vectype)}
function.  @var{vec} must be an expression with integral or floating
vector type and @var{vectype} an integral or floating vector type with the
same number of elements.  The result has @var{vectype} type and value of
a C cast of every element of @var{vec} to the element type of @var{vectype}.

Consider the following example,
@smallexample
typedef int v4si __attribute__ ((vector_size (16)));
typedef float v4sf __attribute__ ((vector_size (16)));
typedef double v4df __attribute__ ((vector_size (32)));
typedef unsigned long long v4di __attribute__ ((vector_size (32)));

v4si a = @{1,-2,3,-4@};
v4sf b = @{1.5f,-2.5f,3.f,7.f@};
v4di c = @{1ULL,5ULL,0ULL,10ULL@};
v4sf d = __builtin_convertvector (a, v4sf); /* d is @{1.f,-2.f,3.f,-4.f@} */
/* Equivalent of:
   v4sf d = @{ (float)a[0], (float)a[1], (float)a[2], (float)a[3] @}; */
v4df e = __builtin_convertvector (a, v4df); /* e is @{1.,-2.,3.,-4.@} */
v4df f = __builtin_convertvector (b, v4df); /* f is @{1.5,-2.5,3.,7.@} */
v4si g = __builtin_convertvector (f, v4si); /* g is @{1,-2,3,7@} */
v4si h = __builtin_convertvector (c, v4si); /* h is @{1,5,0,10@} */
@end smallexample

@cindex vector types, using with x86 intrinsics
Sometimes it is desirable to write code using a mix of generic vector
operations (for clarity) and machine-specific vector intrinsics (to
access vector instructions that are not exposed via generic built-ins).
On x86, intrinsic functions for integer vectors typically use the same
vector type @code{__m128i} irrespective of how they interpret the vector,
making it necessary to cast their arguments and return values from/to
other vector types.  In C, you can make use of a @code{union} type:
@c In C++ such type punning via a union is not allowed by the language
@smallexample
#include <immintrin.h>

typedef unsigned char u8x16 __attribute__ ((vector_size (16)));
typedef unsigned int  u32x4 __attribute__ ((vector_size (16)));

typedef union @{
        __m128i mm;
        u8x16   u8;
        u32x4   u32;
@} v128;
@end smallexample

@noindent
for variables that can be used with both built-in operators and x86
intrinsics:

@smallexample
v128 x, y = @{ 0 @};
memcpy (&x, ptr, sizeof x);
y.u8  += 0x80;
x.mm  = _mm_adds_epu8 (x.mm, y.mm);
x.u32 &= 0xffffff;

/* Instead of a variable, a compound literal may be used to pass the
   return value of an intrinsic call to a function expecting the union: */
v128 foo (v128);
x = foo ((v128) @{_mm_adds_epu8 (x.mm, y.mm)@});
@c This could be done implicitly with __attribute__((transparent_union)),
@c but GCC does not accept it for unions of vector types (PR 88955).
@end smallexample

@node Offsetof
@section Support for @code{offsetof}
@findex __builtin_offsetof

GCC implements for both C and C++ a syntactic extension to implement
the @code{offsetof} macro.

@smallexample
primary:
        "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")"

offsetof_member_designator:
          @code{identifier}
        | offsetof_member_designator "." @code{identifier}
        | offsetof_member_designator "[" @code{expr} "]"
@end smallexample

This extension is sufficient such that

@smallexample
#define offsetof(@var{type}, @var{member})  __builtin_offsetof (@var{type}, @var{member})
@end smallexample

@noindent
is a suitable definition of the @code{offsetof} macro.  In C++, @var{type}
may be dependent.  In either case, @var{member} may consist of a single
identifier, or a sequence of member accesses and array references.

@node __sync Builtins
@section Legacy @code{__sync} Built-in Functions for Atomic Memory Access

The following built-in functions
are intended to be compatible with those described
in the @cite{Intel Itanium Processor-specific Application Binary Interface},
section 7.4.  As such, they depart from normal GCC practice by not using
the @samp{__builtin_} prefix and also by being overloaded so that they
work on multiple types.

The definition given in the Intel documentation allows only for the use of
the types @code{int}, @code{long}, @code{long long} or their unsigned
counterparts.  GCC allows any scalar type that is 1, 2, 4 or 8 bytes in
size other than the C type @code{_Bool} or the C++ type @code{bool}.
Operations on pointer arguments are performed as if the operands were
of the @code{uintptr_t} type.  That is, they are not scaled by the size
of the type to which the pointer points.

These functions are implemented in terms of the @samp{__atomic}
builtins (@pxref{__atomic Builtins}).  They should not be used for new
code which should use the @samp{__atomic} builtins instead.

Not all operations are supported by all target processors.  If a particular
operation cannot be implemented on the target processor, a warning is
generated and a call to an external function is generated.  The external
function carries the same name as the built-in version,
with an additional suffix
@samp{_@var{n}} where @var{n} is the size of the data type.

@c ??? Should we have a mechanism to suppress this warning?  This is almost
@c useful for implementing the operation under the control of an external
@c mutex.

In most cases, these built-in functions are considered a @dfn{full barrier}.
That is,
no memory operand is moved across the operation, either forward or
backward.  Further, instructions are issued as necessary to prevent the
processor from speculating loads across the operation and from queuing stores
after the operation.

All of the routines are described in the Intel documentation to take
``an optional list of variables protected by the memory barrier''.  It's
not clear what is meant by that; it could mean that @emph{only} the
listed variables are protected, or it could mean a list of additional
variables to be protected.  The list is ignored by GCC which treats it as
empty.  GCC interprets an empty list as meaning that all globally
accessible variables should be protected.

@table @code
@item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...)
@findex __sync_fetch_and_add
@findex __sync_fetch_and_sub
@findex __sync_fetch_and_or
@findex __sync_fetch_and_and
@findex __sync_fetch_and_xor
@findex __sync_fetch_and_nand
These built-in functions perform the operation suggested by the name, and
returns the value that had previously been in memory.  That is, operations
on integer operands have the following semantics.  Operations on pointer
arguments are performed as if the operands were of the @code{uintptr_t}
type.  That is, they are not scaled by the size of the type to which
the pointer points.

@smallexample
@{ tmp = *ptr; *ptr @var{op}= value; return tmp; @}
@{ tmp = *ptr; *ptr = ~(tmp & value); return tmp; @}   // nand
@end smallexample

The object pointed to by the first argument must be of integer or pointer
type.  It must not be a boolean type.

@emph{Note:} GCC 4.4 and later implement @code{__sync_fetch_and_nand}
as @code{*ptr = ~(tmp & value)} instead of @code{*ptr = ~tmp & value}.

@item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...)
@itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...)
@findex __sync_add_and_fetch
@findex __sync_sub_and_fetch
@findex __sync_or_and_fetch
@findex __sync_and_and_fetch
@findex __sync_xor_and_fetch
@findex __sync_nand_and_fetch
These built-in functions perform the operation suggested by the name, and
return the new value.  That is, operations on integer operands have
the following semantics.  Operations on pointer operands are performed as
if the operand's type were @code{uintptr_t}.

@smallexample
@{ *ptr @var{op}= value; return *ptr; @}
@{ *ptr = ~(*ptr & value); return *ptr; @}   // nand
@end smallexample

The same constraints on arguments apply as for the corresponding
@code{__sync_op_and_fetch} built-in functions.

@emph{Note:} GCC 4.4 and later implement @code{__sync_nand_and_fetch}
as @code{*ptr = ~(*ptr & value)} instead of
@code{*ptr = ~*ptr & value}.

@item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
@itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval, @var{type} newval, ...)
@findex __sync_bool_compare_and_swap
@findex __sync_val_compare_and_swap
These built-in functions perform an atomic compare and swap.
That is, if the current
value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into
@code{*@var{ptr}}.

The ``bool'' version returns @code{true} if the comparison is successful and
@var{newval} is written.  The ``val'' version returns the contents
of @code{*@var{ptr}} before the operation.

@item __sync_synchronize (...)
@findex __sync_synchronize
This built-in function issues a full memory barrier.

@item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...)
@findex __sync_lock_test_and_set
This built-in function, as described by Intel, is not a traditional test-and-set
operation, but rather an atomic exchange operation.  It writes @var{value}
into @code{*@var{ptr}}, and returns the previous contents of
@code{*@var{ptr}}.

Many targets have only minimal support for such locks, and do not support
a full exchange operation.  In this case, a target may support reduced
functionality here by which the @emph{only} valid value to store is the
immediate constant 1.  The exact value actually stored in @code{*@var{ptr}}
is implementation defined.

This built-in function is not a full barrier,
but rather an @dfn{acquire barrier}.
This means that references after the operation cannot move to (or be
speculated to) before the operation, but previous memory stores may not
be globally visible yet, and previous memory loads may not yet be
satisfied.

@item void __sync_lock_release (@var{type} *ptr, ...)
@findex __sync_lock_release
This built-in function releases the lock acquired by
@code{__sync_lock_test_and_set}.
Normally this means writing the constant 0 to @code{*@var{ptr}}.

This built-in function is not a full barrier,
but rather a @dfn{release barrier}.
This means that all previous memory stores are globally visible, and all
previous memory loads have been satisfied, but following memory reads
are not prevented from being speculated to before the barrier.
@end table

@node __atomic Builtins
@section Built-in Functions for Memory Model Aware Atomic Operations

The following built-in functions approximately match the requirements
for the C++11 memory model.  They are all
identified by being prefixed with @samp{__atomic} and most are
overloaded so that they work with multiple types.

These functions are intended to replace the legacy @samp{__sync}
builtins.  The main difference is that the memory order that is requested
is a parameter to the functions.  New code should always use the
@samp{__atomic} builtins rather than the @samp{__sync} builtins.

Note that the @samp{__atomic} builtins assume that programs will
conform to the C++11 memory model.  In particular, they assume
that programs are free of data races.  See the C++11 standard for
detailed requirements.

The @samp{__atomic} builtins can be used with any integral scalar or
pointer type that is 1, 2, 4, or 8 bytes in length.  16-byte integral
types are also allowed if @samp{__int128} (@pxref{__int128}) is
supported by the architecture.

The four non-arithmetic functions (load, store, exchange, and 
compare_exchange) all have a generic version as well.  This generic
version works on any data type.  It uses the lock-free built-in function
if the specific data type size makes that possible; otherwise, an
external call is left to be resolved at run time.  This external call is
the same format with the addition of a @samp{size_t} parameter inserted
as the first parameter indicating the size of the object being pointed to.
All objects must be the same size.

There are 6 different memory orders that can be specified.  These map
to the C++11 memory orders with the same names, see the C++11 standard
or the @uref{http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync,GCC wiki
on atomic synchronization} for detailed definitions.  Individual
targets may also support additional memory orders for use on specific
architectures.  Refer to the target documentation for details of
these.

An atomic operation can both constrain code motion and
be mapped to hardware instructions for synchronization between threads
(e.g., a fence).  To which extent this happens is controlled by the
memory orders, which are listed here in approximately ascending order of
strength.  The description of each memory order is only meant to roughly
illustrate the effects and is not a specification; see the C++11
memory model for precise semantics.

@table  @code
@item __ATOMIC_RELAXED
Implies no inter-thread ordering constraints.
@item __ATOMIC_CONSUME
This is currently implemented using the stronger @code{__ATOMIC_ACQUIRE}
memory order because of a deficiency in C++11's semantics for
@code{memory_order_consume}.
@item __ATOMIC_ACQUIRE
Creates an inter-thread happens-before constraint from the release (or
stronger) semantic store to this acquire load.  Can prevent hoisting
of code to before the operation.
@item __ATOMIC_RELEASE
Creates an inter-thread happens-before constraint to acquire (or stronger)
semantic loads that read from this release store.  Can prevent sinking
of code to after the operation.
@item __ATOMIC_ACQ_REL
Combines the effects of both @code{__ATOMIC_ACQUIRE} and
@code{__ATOMIC_RELEASE}.
@item __ATOMIC_SEQ_CST
Enforces total ordering with all other @code{__ATOMIC_SEQ_CST} operations.
@end table

Note that in the C++11 memory model, @emph{fences} (e.g.,
@samp{__atomic_thread_fence}) take effect in combination with other
atomic operations on specific memory locations (e.g., atomic loads);
operations on specific memory locations do not necessarily affect other
operations in the same way.

Target architectures are encouraged to provide their own patterns for
each of the atomic built-in functions.  If no target is provided, the original
non-memory model set of @samp{__sync} atomic built-in functions are
used, along with any required synchronization fences surrounding it in
order to achieve the proper behavior.  Execution in this case is subject
to the same restrictions as those built-in functions.

If there is no pattern or mechanism to provide a lock-free instruction
sequence, a call is made to an external routine with the same parameters
to be resolved at run time.

When implementing patterns for these built-in functions, the memory order
parameter can be ignored as long as the pattern implements the most
restrictive @code{__ATOMIC_SEQ_CST} memory order.  Any of the other memory
orders execute correctly with this memory order but they may not execute as
efficiently as they could with a more appropriate implementation of the
relaxed requirements.

Note that the C++11 standard allows for the memory order parameter to be
determined at run time rather than at compile time.  These built-in
functions map any run-time value to @code{__ATOMIC_SEQ_CST} rather
than invoke a runtime library call or inline a switch statement.  This is
standard compliant, safe, and the simplest approach for now.

The memory order parameter is a signed int, but only the lower 16 bits are
reserved for the memory order.  The remainder of the signed int is reserved
for target use and should be 0.  Use of the predefined atomic values
ensures proper usage.

@deftypefn {Built-in Function} @var{type} __atomic_load_n (@var{type} *ptr, int memorder)
This built-in function implements an atomic load operation.  It returns the
contents of @code{*@var{ptr}}.

The valid memory order variants are
@code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
and @code{__ATOMIC_CONSUME}.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_load (@var{type} *ptr, @var{type} *ret, int memorder)
This is the generic version of an atomic load.  It returns the
contents of @code{*@var{ptr}} in @code{*@var{ret}}.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_store_n (@var{type} *ptr, @var{type} val, int memorder)
This built-in function implements an atomic store operation.  It writes 
@code{@var{val}} into @code{*@var{ptr}}.  

The valid memory order variants are
@code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and @code{__ATOMIC_RELEASE}.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_store (@var{type} *ptr, @var{type} *val, int memorder)
This is the generic version of an atomic store.  It stores the value
of @code{*@var{val}} into @code{*@var{ptr}}.

@end deftypefn

@deftypefn {Built-in Function} @var{type} __atomic_exchange_n (@var{type} *ptr, @var{type} val, int memorder)
This built-in function implements an atomic exchange operation.  It writes
@var{val} into @code{*@var{ptr}}, and returns the previous contents of
@code{*@var{ptr}}.

The valid memory order variants are
@code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, @code{__ATOMIC_ACQUIRE},
@code{__ATOMIC_RELEASE}, and @code{__ATOMIC_ACQ_REL}.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_exchange (@var{type} *ptr, @var{type} *val, @var{type} *ret, int memorder)
This is the generic version of an atomic exchange.  It stores the
contents of @code{*@var{val}} into @code{*@var{ptr}}. The original value
of @code{*@var{ptr}} is copied into @code{*@var{ret}}.

@end deftypefn

@deftypefn {Built-in Function} bool __atomic_compare_exchange_n (@var{type} *ptr, @var{type} *expected, @var{type} desired, bool weak, int success_memorder, int failure_memorder)
This built-in function implements an atomic compare and exchange operation.
This compares the contents of @code{*@var{ptr}} with the contents of
@code{*@var{expected}}. If equal, the operation is a @emph{read-modify-write}
operation that writes @var{desired} into @code{*@var{ptr}}.  If they are not
equal, the operation is a @emph{read} and the current contents of
@code{*@var{ptr}} are written into @code{*@var{expected}}.  @var{weak} is @code{true}
for weak compare_exchange, which may fail spuriously, and @code{false} for
the strong variation, which never fails spuriously.  Many targets
only offer the strong variation and ignore the parameter.  When in doubt, use
the strong variation.

If @var{desired} is written into @code{*@var{ptr}} then @code{true} is returned
and memory is affected according to the
memory order specified by @var{success_memorder}.  There are no
restrictions on what memory order can be used here.

Otherwise, @code{false} is returned and memory is affected according
to @var{failure_memorder}. This memory order cannot be
@code{__ATOMIC_RELEASE} nor @code{__ATOMIC_ACQ_REL}.  It also cannot be a
stronger order than that specified by @var{success_memorder}.

@end deftypefn

@deftypefn {Built-in Function} bool __atomic_compare_exchange (@var{type} *ptr, @var{type} *expected, @var{type} *desired, bool weak, int success_memorder, int failure_memorder)
This built-in function implements the generic version of
@code{__atomic_compare_exchange}.  The function is virtually identical to
@code{__atomic_compare_exchange_n}, except the desired value is also a
pointer.

@end deftypefn

@deftypefn {Built-in Function} @var{type} __atomic_add_fetch (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_sub_fetch (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_and_fetch (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_xor_fetch (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_or_fetch (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_nand_fetch (@var{type} *ptr, @var{type} val, int memorder)
These built-in functions perform the operation suggested by the name, and
return the result of the operation.  Operations on pointer arguments are
performed as if the operands were of the @code{uintptr_t} type.  That is,
they are not scaled by the size of the type to which the pointer points.

@smallexample
@{ *ptr @var{op}= val; return *ptr; @}
@{ *ptr = ~(*ptr & val); return *ptr; @} // nand
@end smallexample

The object pointed to by the first argument must be of integer or pointer
type.  It must not be a boolean type.  All memory orders are valid.

@end deftypefn

@deftypefn {Built-in Function} @var{type} __atomic_fetch_add (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_fetch_sub (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_fetch_and (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_fetch_xor (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_fetch_or (@var{type} *ptr, @var{type} val, int memorder)
@deftypefnx {Built-in Function} @var{type} __atomic_fetch_nand (@var{type} *ptr, @var{type} val, int memorder)
These built-in functions perform the operation suggested by the name, and
return the value that had previously been in @code{*@var{ptr}}.  Operations
on pointer arguments are performed as if the operands were of
the @code{uintptr_t} type.  That is, they are not scaled by the size of
the type to which the pointer points.

@smallexample
@{ tmp = *ptr; *ptr @var{op}= val; return tmp; @}
@{ tmp = *ptr; *ptr = ~(*ptr & val); return tmp; @} // nand
@end smallexample

The same constraints on arguments apply as for the corresponding
@code{__atomic_op_fetch} built-in functions.  All memory orders are valid.

@end deftypefn

@deftypefn {Built-in Function} bool __atomic_test_and_set (void *ptr, int memorder)

This built-in function performs an atomic test-and-set operation on
the byte at @code{*@var{ptr}}.  The byte is set to some implementation
defined nonzero ``set'' value and the return value is @code{true} if and only
if the previous contents were ``set''.
It should be only used for operands of type @code{bool} or @code{char}. For 
other types only part of the value may be set.

All memory orders are valid.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_clear (bool *ptr, int memorder)

This built-in function performs an atomic clear operation on
@code{*@var{ptr}}.  After the operation, @code{*@var{ptr}} contains 0.
It should be only used for operands of type @code{bool} or @code{char} and 
in conjunction with @code{__atomic_test_and_set}.
For other types it may only clear partially. If the type is not @code{bool}
prefer using @code{__atomic_store}.

The valid memory order variants are
@code{__ATOMIC_RELAXED}, @code{__ATOMIC_SEQ_CST}, and
@code{__ATOMIC_RELEASE}.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_thread_fence (int memorder)

This built-in function acts as a synchronization fence between threads
based on the specified memory order.

All memory orders are valid.

@end deftypefn

@deftypefn {Built-in Function} void __atomic_signal_fence (int memorder)

This built-in function acts as a synchronization fence between a thread
and signal handlers based in the same thread.

All memory orders are valid.

@end deftypefn

@deftypefn {Built-in Function} bool __atomic_always_lock_free (size_t size,  void *ptr)

This built-in function returns @code{true} if objects of @var{size} bytes always
generate lock-free atomic instructions for the target architecture.
@var{size} must resolve to a compile-time constant and the result also
resolves to a compile-time constant.

@var{ptr} is an optional pointer to the object that may be used to determine
alignment.  A value of 0 indicates typical alignment should be used.  The 
compiler may also ignore this parameter.

@smallexample
if (__atomic_always_lock_free (sizeof (long long), 0))
@end smallexample

@end deftypefn

@deftypefn {Built-in Function} bool __atomic_is_lock_free (size_t size, void *ptr)

This built-in function returns @code{true} if objects of @var{size} bytes always
generate lock-free atomic instructions for the target architecture.  If
the built-in function is not known to be lock-free, a call is made to a
runtime routine named @code{__atomic_is_lock_free}.

@var{ptr} is an optional pointer to the object that may be used to determine
alignment.  A value of 0 indicates typical alignment should be used.  The 
compiler may also ignore this parameter.
@end deftypefn

@node Integer Overflow Builtins
@section Built-in Functions to Perform Arithmetic with Overflow Checking

The following built-in functions allow performing simple arithmetic operations
together with checking whether the operations overflowed.

@deftypefn {Built-in Function} bool __builtin_add_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
@deftypefnx {Built-in Function} bool __builtin_sadd_overflow (int a, int b, int *res)
@deftypefnx {Built-in Function} bool __builtin_saddl_overflow (long int a, long int b, long int *res)
@deftypefnx {Built-in Function} bool __builtin_saddll_overflow (long long int a, long long int b, long long int *res)
@deftypefnx {Built-in Function} bool __builtin_uadd_overflow (unsigned int a, unsigned int b, unsigned int *res)
@deftypefnx {Built-in Function} bool __builtin_uaddl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
@deftypefnx {Built-in Function} bool __builtin_uaddll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)

These built-in functions promote the first two operands into infinite precision signed
type and perform addition on those promoted operands.  The result is then
cast to the type the third pointer argument points to and stored there.
If the stored result is equal to the infinite precision result, the built-in
functions return @code{false}, otherwise they return @code{true}.  As the addition is
performed in infinite signed precision, these built-in functions have fully defined
behavior for all argument values.

The first built-in function allows arbitrary integral types for operands and
the result type must be pointer to some integral type other than enumerated or
boolean type, the rest of the built-in functions have explicit integer types.

The compiler will attempt to use hardware instructions to implement
these built-in functions where possible, like conditional jump on overflow
after addition, conditional jump on carry etc.

@end deftypefn

@deftypefn {Built-in Function} bool __builtin_sub_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
@deftypefnx {Built-in Function} bool __builtin_ssub_overflow (int a, int b, int *res)
@deftypefnx {Built-in Function} bool __builtin_ssubl_overflow (long int a, long int b, long int *res)
@deftypefnx {Built-in Function} bool __builtin_ssubll_overflow (long long int a, long long int b, long long int *res)
@deftypefnx {Built-in Function} bool __builtin_usub_overflow (unsigned int a, unsigned int b, unsigned int *res)
@deftypefnx {Built-in Function} bool __builtin_usubl_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
@deftypefnx {Built-in Function} bool __builtin_usubll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)

These built-in functions are similar to the add overflow checking built-in
functions above, except they perform subtraction, subtract the second argument
from the first one, instead of addition.

@end deftypefn

@deftypefn {Built-in Function} bool __builtin_mul_overflow (@var{type1} a, @var{type2} b, @var{type3} *res)
@deftypefnx {Built-in Function} bool __builtin_smul_overflow (int a, int b, int *res)
@deftypefnx {Built-in Function} bool __builtin_smull_overflow (long int a, long int b, long int *res)
@deftypefnx {Built-in Function} bool __builtin_smulll_overflow (long long int a, long long int b, long long int *res)
@deftypefnx {Built-in Function} bool __builtin_umul_overflow (unsigned int a, unsigned int b, unsigned int *res)
@deftypefnx {Built-in Function} bool __builtin_umull_overflow (unsigned long int a, unsigned long int b, unsigned long int *res)
@deftypefnx {Built-in Function} bool __builtin_umulll_overflow (unsigned long long int a, unsigned long long int b, unsigned long long int *res)

These built-in functions are similar to the add overflow checking built-in
functions above, except they perform multiplication, instead of addition.

@end deftypefn

The following built-in functions allow checking if simple arithmetic operation
would overflow.

@deftypefn {Built-in Function} bool __builtin_add_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
@deftypefnx {Built-in Function} bool __builtin_sub_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)
@deftypefnx {Built-in Function} bool __builtin_mul_overflow_p (@var{type1} a, @var{type2} b, @var{type3} c)

These built-in functions are similar to @code{__builtin_add_overflow},
@code{__builtin_sub_overflow}, or @code{__builtin_mul_overflow}, except that
they don't store the result of the arithmetic operation anywhere and the
last argument is not a pointer, but some expression with integral type other
than enumerated or boolean type.

The built-in functions promote the first two operands into infinite precision signed type
and perform addition on those promoted operands. The result is then
cast to the type of the third argument.  If the cast result is equal to the infinite
precision result, the built-in functions return @code{false}, otherwise they return @code{true}.
The value of the third argument is ignored, just the side effects in the third argument
are evaluated, and no integral argument promotions are performed on the last argument.
If the third argument is a bit-field, the type used for the result cast has the
precision and signedness of the given bit-field, rather than precision and signedness
of the underlying type.

For example, the following macro can be used to portably check, at
compile-time, whether or not adding two constant integers will overflow,
and perform the addition only when it is known to be safe and not to trigger
a @option{-Woverflow} warning.

@smallexample
#define INT_ADD_OVERFLOW_P(a, b) \
   __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0)

enum @{
    A = INT_MAX, B = 3,
    C = INT_ADD_OVERFLOW_P (A, B) ? 0 : A + B,
    D = __builtin_add_overflow_p (1, SCHAR_MAX, (signed char) 0)
@};
@end smallexample

The compiler will attempt to use hardware instructions to implement
these built-in functions where possible, like conditional jump on overflow
after addition, conditional jump on carry etc.
 
@end deftypefn

@node x86 specific memory model extensions for transactional memory
@section x86-Specific Memory Model Extensions for Transactional Memory

The x86 architecture supports additional memory ordering flags
to mark critical sections for hardware lock elision. 
These must be specified in addition to an existing memory order to
atomic intrinsics.

@table @code
@item __ATOMIC_HLE_ACQUIRE
Start lock elision on a lock variable.
Memory order must be @code{__ATOMIC_ACQUIRE} or stronger.
@item __ATOMIC_HLE_RELEASE
End lock elision on a lock variable.
Memory order must be @code{__ATOMIC_RELEASE} or stronger.
@end table

When a lock acquire fails, it is required for good performance to abort
the transaction quickly. This can be done with a @code{_mm_pause}.

@smallexample
#include <immintrin.h> // For _mm_pause

int lockvar;

/* Acquire lock with lock elision */
while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE))
    _mm_pause(); /* Abort failed transaction */
...
/* Free lock with lock elision */
__atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE);
@end smallexample

@node Object Size Checking
@section Object Size Checking Built-in Functions
@findex __builtin_object_size
@findex __builtin___memcpy_chk
@findex __builtin___mempcpy_chk
@findex __builtin___memmove_chk
@findex __builtin___memset_chk
@findex __builtin___strcpy_chk
@findex __builtin___stpcpy_chk
@findex __builtin___strncpy_chk
@findex __builtin___strcat_chk
@findex __builtin___strncat_chk
@findex __builtin___sprintf_chk
@findex __builtin___snprintf_chk
@findex __builtin___vsprintf_chk
@findex __builtin___vsnprintf_chk
@findex __builtin___printf_chk
@findex __builtin___vprintf_chk
@findex __builtin___fprintf_chk
@findex __builtin___vfprintf_chk

GCC implements a limited buffer overflow protection mechanism that can
prevent some buffer overflow attacks by determining the sizes of objects
into which data is about to be written and preventing the writes when
the size isn't sufficient.  The built-in functions described below yield
the best results when used together and when optimization is enabled.
For example, to detect object sizes across function boundaries or to
follow pointer assignments through non-trivial control flow they rely
on various optimization passes enabled with @option{-O2}.  However, to
a limited extent, they can be used without optimization as well.

@deftypefn {Built-in Function} {size_t} __builtin_object_size (const void * @var{ptr}, int @var{type})
is a built-in construct that returns a constant number of bytes from
@var{ptr} to the end of the object @var{ptr} pointer points to
(if known at compile time).  To determine the sizes of dynamically allocated
objects the function relies on the allocation functions called to obtain
the storage to be declared with the @code{alloc_size} attribute (@pxref{Common
Function Attributes}).  @code{__builtin_object_size} never evaluates
its arguments for side effects.  If there are any side effects in them, it
returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
for @var{type} 2 or 3.  If there are multiple objects @var{ptr} can
point to and all of them are known at compile time, the returned number
is the maximum of remaining byte counts in those objects if @var{type} & 2 is
0 and minimum if nonzero.  If it is not possible to determine which objects
@var{ptr} points to at compile time, @code{__builtin_object_size} should
return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0}
for @var{type} 2 or 3.

@var{type} is an integer constant from 0 to 3.  If the least significant
bit is clear, objects are whole variables, if it is set, a closest
surrounding subobject is considered the object a pointer points to.
The second bit determines if maximum or minimum of remaining bytes
is computed.

@smallexample
struct V @{ char buf1[10]; int b; char buf2[10]; @} var;
char *p = &var.buf1[1], *q = &var.b;

/* Here the object p points to is var.  */
assert (__builtin_object_size (p, 0) == sizeof (var) - 1);
/* The subobject p points to is var.buf1.  */
assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1);
/* The object q points to is var.  */
assert (__builtin_object_size (q, 0)
        == (char *) (&var + 1) - (char *) &var.b);
/* The subobject q points to is var.b.  */
assert (__builtin_object_size (q, 1) == sizeof (var.b));
@end smallexample
@end deftypefn

There are built-in functions added for many common string operation
functions, e.g., for @code{memcpy} @code{__builtin___memcpy_chk}
built-in is provided.  This built-in has an additional last argument,
which is the number of bytes remaining in the object the @var{dest}
argument points to or @code{(size_t) -1} if the size is not known.

The built-in functions are optimized into the normal string functions
like @code{memcpy} if the last argument is @code{(size_t) -1} or if
it is known at compile time that the destination object will not
be overflowed.  If the compiler can determine at compile time that the
object will always be overflowed, it issues a warning.

The intended use can be e.g.@:

@smallexample
#undef memcpy
#define bos0(dest) __builtin_object_size (dest, 0)
#define memcpy(dest, src, n) \
  __builtin___memcpy_chk (dest, src, n, bos0 (dest))

char *volatile p;
char buf[10];
/* It is unknown what object p points to, so this is optimized
   into plain memcpy - no checking is possible.  */
memcpy (p, "abcde", n);
/* Destination is known and length too.  It is known at compile
   time there will be no overflow.  */
memcpy (&buf[5], "abcde", 5);
/* Destination is known, but the length is not known at compile time.
   This will result in __memcpy_chk call that can check for overflow
   at run time.  */
memcpy (&buf[5], "abcde", n);
/* Destination is known and it is known at compile time there will
   be overflow.  There will be a warning and __memcpy_chk call that
   will abort the program at run time.  */
memcpy (&buf[6], "abcde", 5);
@end smallexample

Such built-in functions are provided for @code{memcpy}, @code{mempcpy},
@code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy},
@code{strcat} and @code{strncat}.

There are also checking built-in functions for formatted output functions.
@smallexample
int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...);
int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os,
                              const char *fmt, ...);
int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt,
                              va_list ap);
int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os,
                               const char *fmt, va_list ap);
@end smallexample

The added @var{flag} argument is passed unchanged to @code{__sprintf_chk}
etc.@: functions and can contain implementation specific flags on what
additional security measures the checking function might take, such as
handling @code{%n} differently.

The @var{os} argument is the object size @var{s} points to, like in the
other built-in functions.  There is a small difference in the behavior
though, if @var{os} is @code{(size_t) -1}, the built-in functions are
optimized into the non-checking functions only if @var{flag} is 0, otherwise
the checking function is called with @var{os} argument set to
@code{(size_t) -1}.

In addition to this, there are checking built-in functions
@code{__builtin___printf_chk}, @code{__builtin___vprintf_chk},
@code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}.
These have just one additional argument, @var{flag}, right before
format string @var{fmt}.  If the compiler is able to optimize them to
@code{fputc} etc.@: functions, it does, otherwise the checking function
is called and the @var{flag} argument passed to it.

@node Other Builtins
@section Other Built-in Functions Provided by GCC
@cindex built-in functions
@findex __builtin_alloca
@findex __builtin_alloca_with_align
@findex __builtin_alloca_with_align_and_max
@findex __builtin_call_with_static_chain
@findex __builtin_extend_pointer
@findex __builtin_fpclassify
@findex __builtin_has_attribute
@findex __builtin_isfinite
@findex __builtin_isnormal
@findex __builtin_isgreater
@findex __builtin_isgreaterequal
@findex __builtin_isinf_sign
@findex __builtin_isless
@findex __builtin_islessequal
@findex __builtin_islessgreater
@findex __builtin_isunordered
@findex __builtin_object_size
@findex __builtin_powi
@findex __builtin_powif
@findex __builtin_powil
@findex __builtin_speculation_safe_value
@findex _Exit
@findex _exit
@findex abort
@findex abs
@findex acos
@findex acosf
@findex acosh
@findex acoshf
@findex acoshl
@findex acosl
@findex alloca
@findex asin
@findex asinf
@findex asinh
@findex asinhf
@findex asinhl
@findex asinl
@findex atan
@findex atan2
@findex atan2f
@findex atan2l
@findex atanf
@findex atanh
@findex atanhf
@findex atanhl
@findex atanl
@findex bcmp
@findex bzero
@findex cabs
@findex cabsf
@findex cabsl
@findex cacos
@findex cacosf
@findex cacosh
@findex cacoshf
@findex cacoshl
@findex cacosl
@findex calloc
@findex carg
@findex cargf
@findex cargl
@findex casin
@findex casinf
@findex casinh
@findex casinhf
@findex casinhl
@findex casinl
@findex catan
@findex catanf
@findex catanh
@findex catanhf
@findex catanhl
@findex catanl
@findex cbrt
@findex cbrtf
@findex cbrtl
@findex ccos
@findex ccosf
@findex ccosh
@findex ccoshf
@findex ccoshl
@findex ccosl
@findex ceil
@findex ceilf
@findex ceill
@findex cexp
@findex cexpf
@findex cexpl
@findex cimag
@findex cimagf
@findex cimagl
@findex clog
@findex clogf
@findex clogl
@findex clog10
@findex clog10f
@findex clog10l
@findex conj
@findex conjf
@findex conjl
@findex copysign
@findex copysignf
@findex copysignl
@findex cos
@findex cosf
@findex cosh
@findex coshf
@findex coshl
@findex cosl
@findex cpow
@findex cpowf
@findex cpowl
@findex cproj
@findex cprojf
@findex cprojl
@findex creal
@findex crealf
@findex creall
@findex csin
@findex csinf
@findex csinh
@findex csinhf
@findex csinhl
@findex csinl
@findex csqrt
@findex csqrtf
@findex csqrtl
@findex ctan
@findex ctanf
@findex ctanh
@findex ctanhf
@findex ctanhl
@findex ctanl
@findex dcgettext
@findex dgettext
@findex drem
@findex dremf
@findex dreml
@findex erf
@findex erfc
@findex erfcf
@findex erfcl
@findex erff
@findex erfl
@findex exit
@findex exp
@findex exp10
@findex exp10f
@findex exp10l
@findex exp2
@findex exp2f
@findex exp2l
@findex expf
@findex expl
@findex expm1
@findex expm1f
@findex expm1l
@findex fabs
@findex fabsf
@findex fabsl
@findex fdim
@findex fdimf
@findex fdiml
@findex ffs
@findex floor
@findex floorf
@findex floorl
@findex fma
@findex fmaf
@findex fmal
@findex fmax
@findex fmaxf
@findex fmaxl
@findex fmin
@findex fminf
@findex fminl
@findex fmod
@findex fmodf
@findex fmodl