1@c Copyright (C) 1988,1989,1992,1993,1994,1996,1998,1999,2000,2001,2002,2003,2004
2@c Free Software Foundation, Inc.
3@c This is part of the GCC manual.
4@c For copying conditions, see the file gcc.texi.
5
6@node C Implementation
7@chapter C Implementation-defined behavior
8@cindex implementation-defined behavior, C language
9
10A conforming implementation of ISO C is required to document its
11choice of behavior in each of the areas that are designated
12``implementation defined.''  The following lists all such areas,
13along with the section number from the ISO/IEC 9899:1999 standard.
14
15@menu
16* Translation implementation::
17* Environment implementation::
18* Identifiers implementation::
19* Characters implementation::
20* Integers implementation::
21* Floating point implementation::
22* Arrays and pointers implementation::
23* Hints implementation::
24* Structures unions enumerations and bit-fields implementation::
25* Qualifiers implementation::
26* Preprocessing directives implementation::
27* Library functions implementation::
28* Architecture implementation::
29* Locale-specific behavior implementation::
30@end menu
31
32@node Translation implementation
33@section Translation
34
35@itemize @bullet
36@item
37@cite{How a diagnostic is identified (3.10, 5.1.1.3).}
38
39Diagnostics consist of all the output sent to stderr by GCC.
40
41@item
42@cite{Whether each nonempty sequence of white-space characters other than
43new-line is retained or replaced by one space character in translation
44phase 3 (5.1.1.2).}
45@end itemize
46
47@node Environment implementation
48@section Environment
49
50The behavior of these points are dependent on the implementation
51of the C library, and are not defined by GCC itself.
52
53@node Identifiers implementation
54@section Identifiers
55
56@itemize @bullet
57@item
58@cite{Which additional multibyte characters may appear in identifiers
59and their correspondence to universal character names (6.4.2).}
60
61@item
62@cite{The number of significant initial characters in an identifier
63(5.2.4.1, 6.4.2).}
64
65For internal names, all characters are significant.  For external names,
66the number of significant characters are defined by the linker; for
67almost all targets, all characters are significant.
68
69@end itemize
70
71@node Characters implementation
72@section Characters
73
74@itemize @bullet
75@item
76@cite{The number of bits in a byte (3.6).}
77
78@item
79@cite{The values of the members of the execution character set (5.2.1).}
80
81@item
82@cite{The unique value of the member of the execution character set produced
83for each of the standard alphabetic escape sequences (5.2.2).}
84
85@item
86@cite{The value of a @code{char} object into which has been stored any
87character other than a member of the basic execution character set (6.2.5).}
88
89@item
90@cite{Which of @code{signed char} or @code{unsigned char} has the same range,
91representation, and behavior as ``plain'' @code{char} (6.2.5, 6.3.1.1).}
92
93@item
94@cite{The mapping of members of the source character set (in character
95constants and string literals) to members of the execution character
96set (6.4.4.4, 5.1.1.2).}
97
98@item
99@cite{The value of an integer character constant containing more than one
100character or containing a character or escape sequence that does not map
101to a single-byte execution character (6.4.4.4).}
102
103@item
104@cite{The value of a wide character constant containing more than one
105multibyte character, or containing a multibyte character or escape
106sequence not represented in the extended execution character set (6.4.4.4).}
107
108@item
109@cite{The current locale used to convert a wide character constant consisting
110of a single multibyte character that maps to a member of the extended
111execution character set into a corresponding wide character code (6.4.4.4).}
112
113@item
114@cite{The current locale used to convert a wide string literal into
115corresponding wide character codes (6.4.5).}
116
117@item
118@cite{The value of a string literal containing a multibyte character or escape
119sequence not represented in the execution character set (6.4.5).}
120@end itemize
121
122@node Integers implementation
123@section Integers
124
125@itemize @bullet
126@item
127@cite{Any extended integer types that exist in the implementation (6.2.5).}
128
129@item
130@cite{Whether signed integer types are represented using sign and magnitude,
131two's complement, or one's complement, and whether the extraordinary value
132is a trap representation or an ordinary value (6.2.6.2).}
133
134GCC supports only two's complement integer types, and all bit patterns
135are ordinary values.
136
137@item
138@cite{The rank of any extended integer type relative to another extended
139integer type with the same precision (6.3.1.1).}
140
141@item
142@cite{The result of, or the signal raised by, converting an integer to a
143signed integer type when the value cannot be represented in an object of
144that type (6.3.1.3).}
145
146@item
147@cite{The results of some bitwise operations on signed integers (6.5).}
148@end itemize
149
150@node Floating point implementation
151@section Floating point
152
153@itemize @bullet
154@item
155@cite{The accuracy of the floating-point operations and of the library
156functions in @code{<math.h>} and @code{<complex.h>} that return floating-point
157results (5.2.4.2.2).}
158
159@item
160@cite{The rounding behaviors characterized by non-standard values
161of @code{FLT_ROUNDS} @gol
162(5.2.4.2.2).}
163
164@item
165@cite{The evaluation methods characterized by non-standard negative
166values of @code{FLT_EVAL_METHOD} (5.2.4.2.2).}
167
168@item
169@cite{The direction of rounding when an integer is converted to a
170floating-point number that cannot exactly represent the original
171value (6.3.1.4).}
172
173@item
174@cite{The direction of rounding when a floating-point number is
175converted to a narrower floating-point number (6.3.1.5).}
176
177@item
178@cite{How the nearest representable value or the larger or smaller
179representable value immediately adjacent to the nearest representable
180value is chosen for certain floating constants (6.4.4.2).}
181
182@item
183@cite{Whether and how floating expressions are contracted when not
184disallowed by the @code{FP_CONTRACT} pragma (6.5).}
185
186@item
187@cite{The default state for the @code{FENV_ACCESS} pragma (7.6.1).}
188
189@item
190@cite{Additional floating-point exceptions, rounding modes, environments,
191and classifications, and their macro names (7.6, 7.12).}
192
193@item
194@cite{The default state for the @code{FP_CONTRACT} pragma (7.12.2).}
195
196@item
197@cite{Whether the ``inexact'' floating-point exception can be raised
198when the rounded result actually does equal the mathematical result
199in an IEC 60559 conformant implementation (F.9).}
200
201@item
202@cite{Whether the ``underflow'' (and ``inexact'') floating-point
203exception can be raised when a result is tiny but not inexact in an
204IEC 60559 conformant implementation (F.9).}
205
206@end itemize
207
208@node Arrays and pointers implementation
209@section Arrays and pointers
210
211@itemize @bullet
212@item
213@cite{The result of converting a pointer to an integer or
214vice versa (6.3.2.3).}
215
216A cast from pointer to integer discards most-significant bits if the
217pointer representation is larger than the integer type,
218sign-extends@footnote{Future versions of GCC may zero-extend, or use
219a target-defined @code{ptr_extend} pattern.  Do not rely on sign extension.}
220if the pointer representation is smaller than the integer type, otherwise
221the bits are unchanged.
222@c ??? We've always claimed that pointers were unsigned entities.
223@c Shouldn't we therefore be doing zero-extension?  If so, the bug
224@c is in convert_to_integer, where we call type_for_size and request
225@c a signed integral type.  On the other hand, it might be most useful
226@c for the target if we extend according to POINTERS_EXTEND_UNSIGNED.
227
228A cast from integer to pointer discards most-significant bits if the
229pointer representation is smaller than the integer type, extends according
230to the signedness of the integer type if the pointer representation
231is larger than the integer type, otherwise the bits are unchanged.
232
233When casting from pointer to integer and back again, the resulting
234pointer must reference the same object as the original pointer, otherwise
235the behavior is undefined.  That is, one may not use integer arithmetic to
236avoid the undefined behavior of pointer arithmetic as proscribed in 6.5.6/8.
237
238@item
239@cite{The size of the result of subtracting two pointers to elements
240of the same array (6.5.6).}
241
242@end itemize
243
244@node Hints implementation
245@section Hints
246
247@itemize @bullet
248@item
249@cite{The extent to which suggestions made by using the @code{register}
250storage-class specifier are effective (6.7.1).}
251
252The @code{register} specifier affects code generation only in these ways:
253
254@itemize @bullet
255@item
256When used as part of the register variable extension, see
257@ref{Explicit Reg Vars}.
258
259@item
260When @option{-O0} is in use, the compiler allocates distinct stack
261memory for all variables that do not have the @code{register}
262storage-class specifier; if @code{register} is specified, the variable
263may have a shorter lifespan than the code would indicate and may never
264be placed in memory.
265
266@item
267On some rare x86 targets, @code{setjmp} doesn't save the registers in
268all circumstances.  In those cases, GCC doesn't allocate any variables
269in registers unless they are marked @code{register}.
270
271@end itemize
272
273@item
274@cite{The extent to which suggestions made by using the inline function
275specifier are effective (6.7.4).}
276
277GCC will not inline any functions if the @option{-fno-inline} option is
278used or if @option{-O0} is used.  Otherwise, GCC may still be unable to
279inline a function for many reasons; the @option{-Winline} option may be
280used to determine if a function has not been inlined and why not.
281
282@end itemize
283
284@node Structures unions enumerations and bit-fields implementation
285@section Structures, unions, enumerations, and bit-fields
286
287@itemize @bullet
288@item
289@cite{Whether a ``plain'' int bit-field is treated as a @code{signed int}
290bit-field or as an @code{unsigned int} bit-field (6.7.2, 6.7.2.1).}
291
292@item
293@cite{Allowable bit-field types other than @code{_Bool}, @code{signed int},
294and @code{unsigned int} (6.7.2.1).}
295
296@item
297@cite{Whether a bit-field can straddle a storage-unit boundary (6.7.2.1).}
298
299@item
300@cite{The order of allocation of bit-fields within a unit (6.7.2.1).}
301
302@item
303@cite{The alignment of non-bit-field members of structures (6.7.2.1).}
304
305@item
306@cite{The integer type compatible with each enumerated type (6.7.2.2).}
307
308@end itemize
309
310@node Qualifiers implementation
311@section Qualifiers
312
313@itemize @bullet
314@item
315@cite{What constitutes an access to an object that has volatile-qualified
316type (6.7.3).}
317
318@end itemize
319
320@node Preprocessing directives implementation
321@section Preprocessing directives
322
323@itemize @bullet
324@item
325@cite{How sequences in both forms of header names are mapped to headers
326or external source file names (6.4.7).}
327
328@item
329@cite{Whether the value of a character constant in a constant expression
330that controls conditional inclusion matches the value of the same character
331constant in the execution character set (6.10.1).}
332
333@item
334@cite{Whether the value of a single-character character constant in a
335constant expression that controls conditional inclusion may have a
336negative value (6.10.1).}
337
338@item
339@cite{The places that are searched for an included @samp{<>} delimited
340header, and how the places are specified or the header is
341identified (6.10.2).}
342
343@item
344@cite{How the named source file is searched for in an included @samp{""}
345delimited header (6.10.2).}
346
347@item
348@cite{The method by which preprocessing tokens (possibly resulting from
349macro expansion) in a @code{#include} directive are combined into a header
350name (6.10.2).}
351
352@item
353@cite{The nesting limit for @code{#include} processing (6.10.2).}
354
355GCC imposes a limit of 200 nested @code{#include}s.
356
357@item
358@cite{Whether the @samp{#} operator inserts a @samp{\} character before
359the @samp{\} character that begins a universal character name in a
360character constant or string literal (6.10.3.2).}
361
362@item
363@cite{The behavior on each recognized non-@code{STDC #pragma}
364directive (6.10.6).}
365
366@item
367@cite{The definitions for @code{__DATE__} and @code{__TIME__} when
368respectively, the date and time of translation are not available (6.10.8).}
369
370If the date and time are not available, @code{__DATE__} expands to
371@code{@w{"??? ?? ????"}} and @code{__TIME__} expands to
372@code{"??:??:??"}.
373
374@end itemize
375
376@node Library functions implementation
377@section Library functions
378
379The behavior of these points are dependent on the implementation
380of the C library, and are not defined by GCC itself.
381
382@node Architecture implementation
383@section Architecture
384
385@itemize @bullet
386@item
387@cite{The values or expressions assigned to the macros specified in the
388headers @code{<float.h>}, @code{<limits.h>}, and @code{<stdint.h>}
389(5.2.4.2, 7.18.2, 7.18.3).}
390
391@item
392@cite{The number, order, and encoding of bytes in any object
393(when not explicitly specified in this International Standard) (6.2.6.1).}
394
395@item
396@cite{The value of the result of the sizeof operator (6.5.3.4).}
397
398@end itemize
399
400@node Locale-specific behavior implementation
401@section Locale-specific behavior
402
403The behavior of these points are dependent on the implementation
404of the C library, and are not defined by GCC itself.
405
406@node C Extensions
407@chapter Extensions to the C Language Family
408@cindex extensions, C language
409@cindex C language extensions
410
411@opindex pedantic
412GNU C provides several language features not found in ISO standard C@.
413(The @option{-pedantic} option directs GCC to print a warning message if
414any of these features is used.)  To test for the availability of these
415features in conditional compilation, check for a predefined macro
416@code{__GNUC__}, which is always defined under GCC@.
417
418These extensions are available in C and Objective-C@.  Most of them are
419also available in C++.  @xref{C++ Extensions,,Extensions to the
420C++ Language}, for extensions that apply @emph{only} to C++.
421
422Some features that are in ISO C99 but not C89 or C++ are also, as
423extensions, accepted by GCC in C89 mode and in C++.
424
425@menu
426* Statement Exprs::     Putting statements and declarations inside expressions.
427* Local Labels::        Labels local to a block.
428* Labels as Values::    Getting pointers to labels, and computed gotos.
429* Nested Functions::    As in Algol and Pascal, lexical scoping of functions.
430* Constructing Calls::	Dispatching a call to another function.
431* Typeof::              @code{typeof}: referring to the type of an expression.
432* Lvalues::             Using @samp{?:}, @samp{,} and casts in lvalues.
433* Conditionals::        Omitting the middle operand of a @samp{?:} expression.
434* Long Long::		Double-word integers---@code{long long int}.
435* Complex::             Data types for complex numbers.
436* Hex Floats::          Hexadecimal floating-point constants.
437* Zero Length::         Zero-length arrays.
438* Variable Length::     Arrays whose length is computed at run time.
439* Empty Structures::    Structures with no members.
440* Variadic Macros::	Macros with a variable number of arguments.
441* Escaped Newlines::    Slightly looser rules for escaped newlines.
442* Subscripting::        Any array can be subscripted, even if not an lvalue.
443* Pointer Arith::       Arithmetic on @code{void}-pointers and function pointers.
444* Initializers::        Non-constant initializers.
445* Compound Literals::   Compound literals give structures, unions
446                         or arrays as values.
447* Designated Inits::	Labeling elements of initializers.
448* Cast to Union::       Casting to union type from any member of the union.
449* Case Ranges::		`case 1 ... 9' and such.
450* Mixed Declarations::	Mixing declarations and code.
451* Function Attributes:: Declaring that functions have no side effects,
452                         or that they can never return.
453* Attribute Syntax::    Formal syntax for attributes.
454* Function Prototypes:: Prototype declarations and old-style definitions.
455* C++ Comments::        C++ comments are recognized.
456* Dollar Signs::        Dollar sign is allowed in identifiers.
457* Character Escapes::   @samp{\e} stands for the character @key{ESC}.
458* Variable Attributes::	Specifying attributes of variables.
459* Type Attributes::	Specifying attributes of types.
460* Alignment::           Inquiring about the alignment of a type or variable.
461* Inline::              Defining inline functions (as fast as macros).
462* Extended Asm::        Assembler instructions with C expressions as operands.
463                         (With them you can define ``built-in'' functions.)
464* Constraints::         Constraints for asm operands
465* Asm Labels::          Specifying the assembler name to use for a C symbol.
466* Explicit Reg Vars::   Defining variables residing in specified registers.
467* Alternate Keywords::  @code{__const__}, @code{__asm__}, etc., for header files.
468* Incomplete Enums::    @code{enum foo;}, with details to follow.
469* Function Names::	Printable strings which are the name of the current
470			 function.
471* Return Address::      Getting the return or frame address of a function.
472* Vector Extensions::   Using vector instructions through built-in functions.
473* Other Builtins::      Other built-in functions.
474* Target Builtins::     Built-in functions specific to particular targets.
475* Pragmas::             Pragmas accepted by GCC.
476* Unnamed Fields::      Unnamed struct/union fields within structs/unions.
477* Thread-Local::        Per-thread variables.
478@end menu
479
480@node Statement Exprs
481@section Statements and Declarations in Expressions
482@cindex statements inside expressions
483@cindex declarations inside expressions
484@cindex expressions containing statements
485@cindex macros, statements in expressions
486
487@c the above section title wrapped and causes an underfull hbox.. i
488@c changed it from "within" to "in". --mew 4feb93
489A compound statement enclosed in parentheses may appear as an expression
490in GNU C@.  This allows you to use loops, switches, and local variables
491within an expression.
492
493Recall that a compound statement is a sequence of statements surrounded
494by braces; in this construct, parentheses go around the braces.  For
495example:
496
497@smallexample
498(@{ int y = foo (); int z;
499   if (y > 0) z = y;
500   else z = - y;
501   z; @})
502@end smallexample
503
504@noindent
505is a valid (though slightly more complex than necessary) expression
506for the absolute value of @code{foo ()}.
507
508The last thing in the compound statement should be an expression
509followed by a semicolon; the value of this subexpression serves as the
510value of the entire construct.  (If you use some other kind of statement
511last within the braces, the construct has type @code{void}, and thus
512effectively no value.)
513
514This feature is especially useful in making macro definitions ``safe'' (so
515that they evaluate each operand exactly once).  For example, the
516``maximum'' function is commonly defined as a macro in standard C as
517follows:
518
519@smallexample
520#define max(a,b) ((a) > (b) ? (a) : (b))
521@end smallexample
522
523@noindent
524@cindex side effects, macro argument
525But this definition computes either @var{a} or @var{b} twice, with bad
526results if the operand has side effects.  In GNU C, if you know the
527type of the operands (here taken as @code{int}), you can define
528the macro safely as follows:
529
530@smallexample
531#define maxint(a,b) \
532  (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @})
533@end smallexample
534
535Embedded statements are not allowed in constant expressions, such as
536the value of an enumeration constant, the width of a bit-field, or
537the initial value of a static variable.
538
539If you don't know the type of the operand, you can still do this, but you
540must use @code{typeof} (@pxref{Typeof}).
541
542In G++, the result value of a statement expression undergoes array and
543function pointer decay, and is returned by value to the enclosing
544expression. For instance, if @code{A} is a class, then
545
546@smallexample
547        A a;
548
549        (@{a;@}).Foo ()
550@end smallexample
551
552@noindent
553will construct a temporary @code{A} object to hold the result of the
554statement expression, and that will be used to invoke @code{Foo}.
555Therefore the @code{this} pointer observed by @code{Foo} will not be the
556address of @code{a}.
557
558Any temporaries created within a statement within a statement expression
559will be destroyed at the statement's end.  This makes statement
560expressions inside macros slightly different from function calls.  In
561the latter case temporaries introduced during argument evaluation will
562be destroyed at the end of the statement that includes the function
563call.  In the statement expression case they will be destroyed during
564the statement expression.  For instance,
565
566@smallexample
567#define macro(a)  (@{__typeof__(a) b = (a); b + 3; @})
568template<typename T> T function(T a) @{ T b = a; return b + 3; @}
569
570void foo ()
571@{
572  macro (X ());
573  function (X ());
574@}
575@end smallexample
576
577@noindent
578will have different places where temporaries are destroyed.  For the
579@code{macro} case, the temporary @code{X} will be destroyed just after
580the initialization of @code{b}.  In the @code{function} case that
581temporary will be destroyed when the function returns.
582
583These considerations mean that it is probably a bad idea to use
584statement-expressions of this form in header files that are designed to
585work with C++.  (Note that some versions of the GNU C Library contained
586header files using statement-expression that lead to precisely this
587bug.)
588
589@node Local Labels
590@section Locally Declared Labels
591@cindex local labels
592@cindex macros, local labels
593
594GCC allows you to declare @dfn{local labels} in any nested block
595scope. A local label is just like an ordinary label, but you can
596only reference it (with a @code{goto} statement, or by taking its
597address) within the block in which it was declared.
598
599A local label declaration looks like this:
600
601@smallexample
602__label__ @var{label};
603@end smallexample
604
605@noindent
606or
607
608@smallexample
609__label__ @var{label1}, @var{label2}, /* @r{@dots{}} */;
610@end smallexample
611
612Local label declarations must come at the beginning of the block,
613before any ordinary declarations or statements.
614
615The label declaration defines the label @emph{name}, but does not define
616the label itself.  You must do this in the usual way, with
617@code{@var{label}:}, within the statements of the statement expression.
618
619The local label feature is useful for complex macros.  If a macro
620contains nested loops, a @code{goto} can be useful for breaking out of
621them.  However, an ordinary label whose scope is the whole function
622cannot be used: if the macro can be expanded several times in one
623function, the label will be multiply defined in that function.  A
624local label avoids this problem.  For example:
625
626@smallexample
627#define SEARCH(value, array, target)              \
628do @{                                              \
629  __label__ found;                                \
630  typeof (target) _SEARCH_target = (target);      \
631  typeof (*(array)) *_SEARCH_array = (array);     \
632  int i, j;                                       \
633  int value;                                      \
634  for (i = 0; i < max; i++)                       \
635    for (j = 0; j < max; j++)                     \
636      if (_SEARCH_array[i][j] == _SEARCH_target)  \
637        @{ (value) = i; goto found; @}              \
638  (value) = -1;                                   \
639 found:;                                          \
640@} while (0)
641@end smallexample
642
643This could also be written using a statement-expression:
644
645@smallexample
646#define SEARCH(array, target)                     \
647(@{                                                \
648  __label__ found;                                \
649  typeof (target) _SEARCH_target = (target);      \
650  typeof (*(array)) *_SEARCH_array = (array);     \
651  int i, j;                                       \
652  int value;                                      \
653  for (i = 0; i < max; i++)                       \
654    for (j = 0; j < max; j++)                     \
655      if (_SEARCH_array[i][j] == _SEARCH_target)  \
656        @{ value = i; goto found; @}                \
657  value = -1;                                     \
658 found:                                           \
659  value;                                          \
660@})
661@end smallexample
662
663Local label declarations also make the labels they declare visible to
664nested functions, if there are any.  @xref{Nested Functions}, for details.
665
666@node Labels as Values
667@section Labels as Values
668@cindex labels as values
669@cindex computed gotos
670@cindex goto with computed label
671@cindex address of a label
672
673You can get the address of a label defined in the current function
674(or a containing function) with the unary operator @samp{&&}.  The
675value has type @code{void *}.  This value is a constant and can be used
676wherever a constant of that type is valid.  For example:
677
678@smallexample
679void *ptr;
680/* @r{@dots{}} */
681ptr = &&foo;
682@end smallexample
683
684To use these values, you need to be able to jump to one.  This is done
685with the computed goto statement@footnote{The analogous feature in
686Fortran is called an assigned goto, but that name seems inappropriate in
687C, where one can do more than simply store label addresses in label
688variables.}, @code{goto *@var{exp};}.  For example,
689
690@smallexample
691goto *ptr;
692@end smallexample
693
694@noindent
695Any expression of type @code{void *} is allowed.
696
697One way of using these constants is in initializing a static array that
698will serve as a jump table:
699
700@smallexample
701static void *array[] = @{ &&foo, &&bar, &&hack @};
702@end smallexample
703
704Then you can select a label with indexing, like this:
705
706@smallexample
707goto *array[i];
708@end smallexample
709
710@noindent
711Note that this does not check whether the subscript is in bounds---array
712indexing in C never does that.
713
714Such an array of label values serves a purpose much like that of the
715@code{switch} statement.  The @code{switch} statement is cleaner, so
716use that rather than an array unless the problem does not fit a
717@code{switch} statement very well.
718
719Another use of label values is in an interpreter for threaded code.
720The labels within the interpreter function can be stored in the
721threaded code for super-fast dispatching.
722
723You may not use this mechanism to jump to code in a different function.
724If you do that, totally unpredictable things will happen.  The best way to
725avoid this is to store the label address only in automatic variables and
726never pass it as an argument.
727
728An alternate way to write the above example is
729
730@smallexample
731static const int array[] = @{ &&foo - &&foo, &&bar - &&foo,
732                             &&hack - &&foo @};
733goto *(&&foo + array[i]);
734@end smallexample
735
736@noindent
737This is more friendly to code living in shared libraries, as it reduces
738the number of dynamic relocations that are needed, and by consequence,
739allows the data to be read-only.
740
741@node Nested Functions
742@section Nested Functions
743@cindex nested functions
744@cindex downward funargs
745@cindex thunks
746
747A @dfn{nested function} is a function defined inside another function.
748(Nested functions are not supported for GNU C++.)  The nested function's
749name is local to the block where it is defined.  For example, here we
750define a nested function named @code{square}, and call it twice:
751
752@smallexample
753@group
754foo (double a, double b)
755@{
756  double square (double z) @{ return z * z; @}
757
758  return square (a) + square (b);
759@}
760@end group
761@end smallexample
762
763The nested function can access all the variables of the containing
764function that are visible at the point of its definition.  This is
765called @dfn{lexical scoping}.  For example, here we show a nested
766function which uses an inherited variable named @code{offset}:
767
768@smallexample
769@group
770bar (int *array, int offset, int size)
771@{
772  int access (int *array, int index)
773    @{ return array[index + offset]; @}
774  int i;
775  /* @r{@dots{}} */
776  for (i = 0; i < size; i++)
777    /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
778@}
779@end group
780@end smallexample
781
782Nested function definitions are permitted within functions in the places
783where variable definitions are allowed; that is, in any block, before
784the first statement in the block.
785
786It is possible to call the nested function from outside the scope of its
787name by storing its address or passing the address to another function:
788
789@smallexample
790hack (int *array, int size)
791@{
792  void store (int index, int value)
793    @{ array[index] = value; @}
794
795  intermediate (store, size);
796@}
797@end smallexample
798
799Here, the function @code{intermediate} receives the address of
800@code{store} as an argument.  If @code{intermediate} calls @code{store},
801the arguments given to @code{store} are used to store into @code{array}.
802But this technique works only so long as the containing function
803(@code{hack}, in this example) does not exit.
804
805If you try to call the nested function through its address after the
806containing function has exited, all hell will break loose.  If you try
807to call it after a containing scope level has exited, and if it refers
808to some of the variables that are no longer in scope, you may be lucky,
809but it's not wise to take the risk.  If, however, the nested function
810does not refer to anything that has gone out of scope, you should be
811safe.
812
813GCC implements taking the address of a nested function using a technique
814called @dfn{trampolines}.  A paper describing them is available as
815
816@noindent
817@uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}.
818
819A nested function can jump to a label inherited from a containing
820function, provided the label was explicitly declared in the containing
821function (@pxref{Local Labels}).  Such a jump returns instantly to the
822containing function, exiting the nested function which did the
823@code{goto} and any intermediate functions as well.  Here is an example:
824
825@smallexample
826@group
827bar (int *array, int offset, int size)
828@{
829  __label__ failure;
830  int access (int *array, int index)
831    @{
832      if (index > size)
833        goto failure;
834      return array[index + offset];
835    @}
836  int i;
837  /* @r{@dots{}} */
838  for (i = 0; i < size; i++)
839    /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */
840  /* @r{@dots{}} */
841  return 0;
842
843 /* @r{Control comes here from @code{access}
844    if it detects an error.}  */
845 failure:
846  return -1;
847@}
848@end group
849@end smallexample
850
851A nested function always has internal linkage.  Declaring one with
852@code{extern} is erroneous.  If you need to declare the nested function
853before its definition, use @code{auto} (which is otherwise meaningless
854for function declarations).
855
856@smallexample
857bar (int *array, int offset, int size)
858@{
859  __label__ failure;
860  auto int access (int *, int);
861  /* @r{@dots{}} */
862  int access (int *array, int index)
863    @{
864      if (index > size)
865        goto failure;
866      return array[index + offset];
867    @}
868  /* @r{@dots{}} */
869@}
870@end smallexample
871
872@node Constructing Calls
873@section Constructing Function Calls
874@cindex constructing calls
875@cindex forwarding calls
876
877Using the built-in functions described below, you can record
878the arguments a function received, and call another function
879with the same arguments, without knowing the number or types
880of the arguments.
881
882You can also record the return value of that function call,
883and later return that value, without knowing what data type
884the function tried to return (as long as your caller expects
885that data type).
886
887However, these built-in functions may interact badly with some
888sophisticated features or other extensions of the language.  It
889is, therefore, not recommended to use them outside very simple
890functions acting as mere forwarders for their arguments.
891
892@deftypefn {Built-in Function} {void *} __builtin_apply_args ()
893This built-in function returns a pointer to data
894describing how to perform a call with the same arguments as were passed
895to the current function.
896
897The function saves the arg pointer register, structure value address,
898and all registers that might be used to pass arguments to a function
899into a block of memory allocated on the stack.  Then it returns the
900address of that block.
901@end deftypefn
902
903@deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size})
904This built-in function invokes @var{function}
905with a copy of the parameters described by @var{arguments}
906and @var{size}.
907
908The value of @var{arguments} should be the value returned by
909@code{__builtin_apply_args}.  The argument @var{size} specifies the size
910of the stack argument data, in bytes.
911
912This function returns a pointer to data describing
913how to return whatever value was returned by @var{function}.  The data
914is saved in a block of memory allocated on the stack.
915
916It is not always simple to compute the proper value for @var{size}.  The
917value is used by @code{__builtin_apply} to compute the amount of data
918that should be pushed on the stack and copied from the incoming argument
919area.
920@end deftypefn
921
922@deftypefn {Built-in Function} {void} __builtin_return (void *@var{result})
923This built-in function returns the value described by @var{result} from
924the containing function.  You should specify, for @var{result}, a value
925returned by @code{__builtin_apply}.
926@end deftypefn
927
928@node Typeof
929@section Referring to a Type with @code{typeof}
930@findex typeof
931@findex sizeof
932@cindex macros, types of arguments
933
934Another way to refer to the type of an expression is with @code{typeof}.
935The syntax of using of this keyword looks like @code{sizeof}, but the
936construct acts semantically like a type name defined with @code{typedef}.
937
938There are two ways of writing the argument to @code{typeof}: with an
939expression or with a type.  Here is an example with an expression:
940
941@smallexample
942typeof (x[0](1))
943@end smallexample
944
945@noindent
946This assumes that @code{x} is an array of pointers to functions;
947the type described is that of the values of the functions.
948
949Here is an example with a typename as the argument:
950
951@smallexample
952typeof (int *)
953@end smallexample
954
955@noindent
956Here the type described is that of pointers to @code{int}.
957
958If you are writing a header file that must work when included in ISO C
959programs, write @code{__typeof__} instead of @code{typeof}.
960@xref{Alternate Keywords}.
961
962A @code{typeof}-construct can be used anywhere a typedef name could be
963used.  For example, you can use it in a declaration, in a cast, or inside
964of @code{sizeof} or @code{typeof}.
965
966@code{typeof} is often useful in conjunction with the
967statements-within-expressions feature.  Here is how the two together can
968be used to define a safe ``maximum'' macro that operates on any
969arithmetic type and evaluates each of its arguments exactly once:
970
971@smallexample
972#define max(a,b) \
973  (@{ typeof (a) _a = (a); \
974      typeof (b) _b = (b); \
975    _a > _b ? _a : _b; @})
976@end smallexample
977
978@cindex underscores in variables in macros
979@cindex @samp{_} in variables in macros
980@cindex local variables in macros
981@cindex variables, local, in macros
982@cindex macros, local variables in
983
984The reason for using names that start with underscores for the local
985variables is to avoid conflicts with variable names that occur within the
986expressions that are substituted for @code{a} and @code{b}.  Eventually we
987hope to design a new form of declaration syntax that allows you to declare
988variables whose scopes start only after their initializers; this will be a
989more reliable way to prevent such conflicts.
990
991@noindent
992Some more examples of the use of @code{typeof}:
993
994@itemize @bullet
995@item
996This declares @code{y} with the type of what @code{x} points to.
997
998@smallexample
999typeof (*x) y;
1000@end smallexample
1001
1002@item
1003This declares @code{y} as an array of such values.
1004
1005@smallexample
1006typeof (*x) y[4];
1007@end smallexample
1008
1009@item
1010This declares @code{y} as an array of pointers to characters:
1011
1012@smallexample
1013typeof (typeof (char *)[4]) y;
1014@end smallexample
1015
1016@noindent
1017It is equivalent to the following traditional C declaration:
1018
1019@smallexample
1020char *y[4];
1021@end smallexample
1022
1023To see the meaning of the declaration using @code{typeof}, and why it
1024might be a useful way to write, rewrite it with these macros:
1025
1026@smallexample
1027#define pointer(T)  typeof(T *)
1028#define array(T, N) typeof(T [N])
1029@end smallexample
1030
1031@noindent
1032Now the declaration can be rewritten this way:
1033
1034@smallexample
1035array (pointer (char), 4) y;
1036@end smallexample
1037
1038@noindent
1039Thus, @code{array (pointer (char), 4)} is the type of arrays of 4
1040pointers to @code{char}.
1041@end itemize
1042
1043@emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported
1044a more limited extension which permitted one to write
1045
1046@smallexample
1047typedef @var{T} = @var{expr};
1048@end smallexample
1049
1050@noindent
1051with the effect of declaring @var{T} to have the type of the expression
1052@var{expr}.  This extension does not work with GCC 3 (versions between
10533.0 and 3.2 will crash; 3.2.1 and later give an error).  Code which
1054relies on it should be rewritten to use @code{typeof}:
1055
1056@smallexample
1057typedef typeof(@var{expr}) @var{T};
1058@end smallexample
1059
1060@noindent
1061This will work with all versions of GCC@.
1062
1063@node Lvalues
1064@section Generalized Lvalues
1065@cindex compound expressions as lvalues
1066@cindex expressions, compound, as lvalues
1067@cindex conditional expressions as lvalues
1068@cindex expressions, conditional, as lvalues
1069@cindex casts as lvalues
1070@cindex generalized lvalues
1071@cindex lvalues, generalized
1072@cindex extensions, @code{?:}
1073@cindex @code{?:} extensions
1074
1075Compound expressions, conditional expressions and casts are allowed as
1076lvalues provided their operands are lvalues.  This means that you can take
1077their addresses or store values into them.  All these extensions are
1078deprecated.
1079
1080Standard C++ allows compound expressions and conditional expressions
1081as lvalues, and permits casts to reference type, so use of this
1082extension is not supported for C++ code.
1083
1084For example, a compound expression can be assigned, provided the last
1085expression in the sequence is an lvalue.  These two expressions are
1086equivalent:
1087
1088@smallexample
1089(a, b) += 5
1090a, (b += 5)
1091@end smallexample
1092
1093Similarly, the address of the compound expression can be taken.  These two
1094expressions are equivalent:
1095
1096@smallexample
1097&(a, b)
1098a, &b
1099@end smallexample
1100
1101A conditional expression is a valid lvalue if its type is not void and the
1102true and false branches are both valid lvalues.  For example, these two
1103expressions are equivalent:
1104
1105@smallexample
1106(a ? b : c) = 5
1107(a ? b = 5 : (c = 5))
1108@end smallexample
1109
1110A cast is a valid lvalue if its operand is an lvalue.  This extension
1111is deprecated.  A simple
1112assignment whose left-hand side is a cast works by converting the
1113right-hand side first to the specified type, then to the type of the
1114inner left-hand side expression.  After this is stored, the value is
1115converted back to the specified type to become the value of the
1116assignment.  Thus, if @code{a} has type @code{char *}, the following two
1117expressions are equivalent:
1118
1119@smallexample
1120(int)a = 5
1121(int)(a = (char *)(int)5)
1122@end smallexample
1123
1124An assignment-with-arithmetic operation such as @samp{+=} applied to a cast
1125performs the arithmetic using the type resulting from the cast, and then
1126continues as in the previous case.  Therefore, these two expressions are
1127equivalent:
1128
1129@smallexample
1130(int)a += 5
1131(int)(a = (char *)(int) ((int)a + 5))
1132@end smallexample
1133
1134You cannot take the address of an lvalue cast, because the use of its
1135address would not work out coherently.  Suppose that @code{&(int)f} were
1136permitted, where @code{f} has type @code{float}.  Then the following
1137statement would try to store an integer bit-pattern where a floating
1138point number belongs:
1139
1140@smallexample
1141*&(int)f = 1;
1142@end smallexample
1143
1144This is quite different from what @code{(int)f = 1} would do---that
1145would convert 1 to floating point and store it.  Rather than cause this
1146inconsistency, we think it is better to prohibit use of @samp{&} on a cast.
1147
1148If you really do want an @code{int *} pointer with the address of
1149@code{f}, you can simply write @code{(int *)&f}.
1150
1151@node Conditionals
1152@section Conditionals with Omitted Operands
1153@cindex conditional expressions, extensions
1154@cindex omitted middle-operands
1155@cindex middle-operands, omitted
1156@cindex extensions, @code{?:}
1157@cindex @code{?:} extensions
1158
1159The middle operand in a conditional expression may be omitted.  Then
1160if the first operand is nonzero, its value is the value of the conditional
1161expression.
1162
1163Therefore, the expression
1164
1165@smallexample
1166x ? : y
1167@end smallexample
1168
1169@noindent
1170has the value of @code{x} if that is nonzero; otherwise, the value of
1171@code{y}.
1172
1173This example is perfectly equivalent to
1174
1175@smallexample
1176x ? x : y
1177@end smallexample
1178
1179@cindex side effect in ?:
1180@cindex ?: side effect
1181@noindent
1182In this simple case, the ability to omit the middle operand is not
1183especially useful.  When it becomes useful is when the first operand does,
1184or may (if it is a macro argument), contain a side effect.  Then repeating
1185the operand in the middle would perform the side effect twice.  Omitting
1186the middle operand uses the value already computed without the undesirable
1187effects of recomputing it.
1188
1189@node Long Long
1190@section Double-Word Integers
1191@cindex @code{long long} data types
1192@cindex double-word arithmetic
1193@cindex multiprecision arithmetic
1194@cindex @code{LL} integer suffix
1195@cindex @code{ULL} integer suffix
1196
1197ISO C99 supports data types for integers that are at least 64 bits wide,
1198and as an extension GCC supports them in C89 mode and in C++.
1199Simply write @code{long long int} for a signed integer, or
1200@code{unsigned long long int} for an unsigned integer.  To make an
1201integer constant of type @code{long long int}, add the suffix @samp{LL}
1202to the integer.  To make an integer constant of type @code{unsigned long
1203long int}, add the suffix @samp{ULL} to the integer.
1204
1205You can use these types in arithmetic like any other integer types.
1206Addition, subtraction, and bitwise boolean operations on these types
1207are open-coded on all types of machines.  Multiplication is open-coded
1208if the machine supports fullword-to-doubleword a widening multiply
1209instruction.  Division and shifts are open-coded only on machines that
1210provide special support.  The operations that are not open-coded use
1211special library routines that come with GCC@.
1212
1213There may be pitfalls when you use @code{long long} types for function
1214arguments, unless you declare function prototypes.  If a function
1215expects type @code{int} for its argument, and you pass a value of type
1216@code{long long int}, confusion will result because the caller and the
1217subroutine will disagree about the number of bytes for the argument.
1218Likewise, if the function expects @code{long long int} and you pass
1219@code{int}.  The best way to avoid such problems is to use prototypes.
1220
1221@node Complex
1222@section Complex Numbers
1223@cindex complex numbers
1224@cindex @code{_Complex} keyword
1225@cindex @code{__complex__} keyword
1226
1227ISO C99 supports complex floating data types, and as an extension GCC
1228supports them in C89 mode and in C++, and supports complex integer data
1229types which are not part of ISO C99.  You can declare complex types
1230using the keyword @code{_Complex}.  As an extension, the older GNU
1231keyword @code{__complex__} is also supported.
1232
1233For example, @samp{_Complex double x;} declares @code{x} as a
1234variable whose real part and imaginary part are both of type
1235@code{double}.  @samp{_Complex short int y;} declares @code{y} to
1236have real and imaginary parts of type @code{short int}; this is not
1237likely to be useful, but it shows that the set of complex types is
1238complete.
1239
1240To write a constant with a complex data type, use the suffix @samp{i} or
1241@samp{j} (either one; they are equivalent).  For example, @code{2.5fi}
1242has type @code{_Complex float} and @code{3i} has type
1243@code{_Complex int}.  Such a constant always has a pure imaginary
1244value, but you can form any complex value you like by adding one to a
1245real constant.  This is a GNU extension; if you have an ISO C99
1246conforming C library (such as GNU libc), and want to construct complex
1247constants of floating type, you should include @code{<complex.h>} and
1248use the macros @code{I} or @code{_Complex_I} instead.
1249
1250@cindex @code{__real__} keyword
1251@cindex @code{__imag__} keyword
1252To extract the real part of a complex-valued expression @var{exp}, write
1253@code{__real__ @var{exp}}.  Likewise, use @code{__imag__} to
1254extract the imaginary part.  This is a GNU extension; for values of
1255floating type, you should use the ISO C99 functions @code{crealf},
1256@code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and
1257@code{cimagl}, declared in @code{<complex.h>} and also provided as
1258built-in functions by GCC@.
1259
1260@cindex complex conjugation
1261The operator @samp{~} performs complex conjugation when used on a value
1262with a complex type.  This is a GNU extension; for values of
1263floating type, you should use the ISO C99 functions @code{conjf},
1264@code{conj} and @code{conjl}, declared in @code{<complex.h>} and also
1265provided as built-in functions by GCC@.
1266
1267GCC can allocate complex automatic variables in a noncontiguous
1268fashion; it's even possible for the real part to be in a register while
1269the imaginary part is on the stack (or vice-versa).  Only the DWARF2
1270debug info format can represent this, so use of DWARF2 is recommended.
1271If you are using the stabs debug info format, GCC describes a noncontiguous
1272complex variable as if it were two separate variables of noncomplex type.
1273If the variable's actual name is @code{foo}, the two fictitious
1274variables are named @code{foo$real} and @code{foo$imag}.  You can
1275examine and set these two fictitious variables with your debugger.
1276
1277@node Hex Floats
1278@section Hex Floats
1279@cindex hex floats
1280
1281ISO C99 supports floating-point numbers written not only in the usual
1282decimal notation, such as @code{1.55e1}, but also numbers such as
1283@code{0x1.fp3} written in hexadecimal format.  As a GNU extension, GCC
1284supports this in C89 mode (except in some cases when strictly
1285conforming) and in C++.  In that format the
1286@samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are
1287mandatory.  The exponent is a decimal number that indicates the power of
12882 by which the significant part will be multiplied.  Thus @samp{0x1.f} is
1289@tex
1290$1 {15\over16}$,
1291@end tex
1292@ifnottex
12931 15/16,
1294@end ifnottex
1295@samp{p3} multiplies it by 8, and the value of @code{0x1.fp3}
1296is the same as @code{1.55e1}.
1297
1298Unlike for floating-point numbers in the decimal notation the exponent
1299is always required in the hexadecimal notation.  Otherwise the compiler
1300would not be able to resolve the ambiguity of, e.g., @code{0x1.f}.  This
1301could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the
1302extension for floating-point constants of type @code{float}.
1303
1304@node Zero Length
1305@section Arrays of Length Zero
1306@cindex arrays of length zero
1307@cindex zero-length arrays
1308@cindex length-zero arrays
1309@cindex flexible array members
1310
1311Zero-length arrays are allowed in GNU C@.  They are very useful as the
1312last element of a structure which is really a header for a variable-length
1313object:
1314
1315@smallexample
1316struct line @{
1317  int length;
1318  char contents[0];
1319@};
1320
1321struct line *thisline = (struct line *)
1322  malloc (sizeof (struct line) + this_length);
1323thisline->length = this_length;
1324@end smallexample
1325
1326In ISO C90, you would have to give @code{contents} a length of 1, which
1327means either you waste space or complicate the argument to @code{malloc}.
1328
1329In ISO C99, you would use a @dfn{flexible array member}, which is
1330slightly different in syntax and semantics:
1331
1332@itemize @bullet
1333@item
1334Flexible array members are written as @code{contents[]} without
1335the @code{0}.
1336
1337@item
1338Flexible array members have incomplete type, and so the @code{sizeof}
1339operator may not be applied.  As a quirk of the original implementation
1340of zero-length arrays, @code{sizeof} evaluates to zero.
1341
1342@item
1343Flexible array members may only appear as the last member of a
1344@code{struct} that is otherwise non-empty.
1345
1346@item
1347A structure containing a flexible array member, or a union containing
1348such a structure (possibly recursively), may not be a member of a
1349structure or an element of an array.  (However, these uses are
1350permitted by GCC as extensions.)
1351@end itemize
1352
1353GCC versions before 3.0 allowed zero-length arrays to be statically
1354initialized, as if they were flexible arrays.  In addition to those
1355cases that were useful, it also allowed initializations in situations
1356that would corrupt later data.  Non-empty initialization of zero-length
1357arrays is now treated like any case where there are more initializer
1358elements than the array holds, in that a suitable warning about "excess
1359elements in array" is given, and the excess elements (all of them, in
1360this case) are ignored.
1361
1362Instead GCC allows static initialization of flexible array members.
1363This is equivalent to defining a new structure containing the original
1364structure followed by an array of sufficient size to contain the data.
1365I.e.@: in the following, @code{f1} is constructed as if it were declared
1366like @code{f2}.
1367
1368@smallexample
1369struct f1 @{
1370  int x; int y[];
1371@} f1 = @{ 1, @{ 2, 3, 4 @} @};
1372
1373struct f2 @{
1374  struct f1 f1; int data[3];
1375@} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @};
1376@end smallexample
1377
1378@noindent
1379The convenience of this extension is that @code{f1} has the desired
1380type, eliminating the need to consistently refer to @code{f2.f1}.
1381
1382This has symmetry with normal static arrays, in that an array of
1383unknown size is also written with @code{[]}.
1384
1385Of course, this extension only makes sense if the extra data comes at
1386the end of a top-level object, as otherwise we would be overwriting
1387data at subsequent offsets.  To avoid undue complication and confusion
1388with initialization of deeply nested arrays, we simply disallow any
1389non-empty initialization except when the structure is the top-level
1390object.  For example:
1391
1392@smallexample
1393struct foo @{ int x; int y[]; @};
1394struct bar @{ struct foo z; @};
1395
1396struct foo a = @{ 1, @{ 2, 3, 4 @} @};        // @r{Valid.}
1397struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @};    // @r{Invalid.}
1398struct bar c = @{ @{ 1, @{ @} @} @};            // @r{Valid.}
1399struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @};  // @r{Invalid.}
1400@end smallexample
1401
1402@node Empty Structures
1403@section Structures With No Members
1404@cindex empty structures
1405@cindex zero-size structures
1406
1407GCC permits a C structure to have no members:
1408
1409@smallexample
1410struct empty @{
1411@};
1412@end smallexample
1413
1414The structure will have size zero.  In C++, empty structures are part
1415of the language.  G++ treats empty structures as if they had a single
1416member of type @code{char}.
1417
1418@node Variable Length
1419@section Arrays of Variable Length
1420@cindex variable-length arrays
1421@cindex arrays of variable length
1422@cindex VLAs
1423
1424Variable-length automatic arrays are allowed in ISO C99, and as an
1425extension GCC accepts them in C89 mode and in C++.  (However, GCC's
1426implementation of variable-length arrays does not yet conform in detail
1427to the ISO C99 standard.)  These arrays are
1428declared like any other automatic arrays, but with a length that is not
1429a constant expression.  The storage is allocated at the point of
1430declaration and deallocated when the brace-level is exited.  For
1431example:
1432
1433@smallexample
1434FILE *
1435concat_fopen (char *s1, char *s2, char *mode)
1436@{
1437  char str[strlen (s1) + strlen (s2) + 1];
1438  strcpy (str, s1);
1439  strcat (str, s2);
1440  return fopen (str, mode);
1441@}
1442@end smallexample
1443
1444@cindex scope of a variable length array
1445@cindex variable-length array scope
1446@cindex deallocating variable length arrays
1447Jumping or breaking out of the scope of the array name deallocates the
1448storage.  Jumping into the scope is not allowed; you get an error
1449message for it.
1450
1451@cindex @code{alloca} vs variable-length arrays
1452You can use the function @code{alloca} to get an effect much like
1453variable-length arrays.  The function @code{alloca} is available in
1454many other C implementations (but not in all).  On the other hand,
1455variable-length arrays are more elegant.
1456
1457There are other differences between these two methods.  Space allocated
1458with @code{alloca} exists until the containing @emph{function} returns.
1459The space for a variable-length array is deallocated as soon as the array
1460name's scope ends.  (If you use both variable-length arrays and
1461@code{alloca} in the same function, deallocation of a variable-length array
1462will also deallocate anything more recently allocated with @code{alloca}.)
1463
1464You can also use variable-length arrays as arguments to functions:
1465
1466@smallexample
1467struct entry
1468tester (int len, char data[len][len])
1469@{
1470  /* @r{@dots{}} */
1471@}
1472@end smallexample
1473
1474The length of an array is computed once when the storage is allocated
1475and is remembered for the scope of the array in case you access it with
1476@code{sizeof}.
1477
1478If you want to pass the array first and the length afterward, you can
1479use a forward declaration in the parameter list---another GNU extension.
1480
1481@smallexample
1482struct entry
1483tester (int len; char data[len][len], int len)
1484@{
1485  /* @r{@dots{}} */
1486@}
1487@end smallexample
1488
1489@cindex parameter forward declaration
1490The @samp{int len} before the semicolon is a @dfn{parameter forward
1491declaration}, and it serves the purpose of making the name @code{len}
1492known when the declaration of @code{data} is parsed.
1493
1494You can write any number of such parameter forward declarations in the
1495parameter list.  They can be separated by commas or semicolons, but the
1496last one must end with a semicolon, which is followed by the ``real''
1497parameter declarations.  Each forward declaration must match a ``real''
1498declaration in parameter name and data type.  ISO C99 does not support
1499parameter forward declarations.
1500
1501@node Variadic Macros
1502@section Macros with a Variable Number of Arguments.
1503@cindex variable number of arguments
1504@cindex macro with variable arguments
1505@cindex rest argument (in macro)
1506@cindex variadic macros
1507
1508In the ISO C standard of 1999, a macro can be declared to accept a
1509variable number of arguments much as a function can.  The syntax for
1510defining the macro is similar to that of a function.  Here is an
1511example:
1512
1513@smallexample
1514#define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
1515@end smallexample
1516
1517Here @samp{@dots{}} is a @dfn{variable argument}.  In the invocation of
1518such a macro, it represents the zero or more tokens until the closing
1519parenthesis that ends the invocation, including any commas.  This set of
1520tokens replaces the identifier @code{__VA_ARGS__} in the macro body
1521wherever it appears.  See the CPP manual for more information.
1522
1523GCC has long supported variadic macros, and used a different syntax that
1524allowed you to give a name to the variable arguments just like any other
1525argument.  Here is an example:
1526
1527@smallexample
1528#define debug(format, args...) fprintf (stderr, format, args)
1529@end smallexample
1530
1531This is in all ways equivalent to the ISO C example above, but arguably
1532more readable and descriptive.
1533
1534GNU CPP has two further variadic macro extensions, and permits them to
1535be used with either of the above forms of macro definition.
1536
1537In standard C, you are not allowed to leave the variable argument out
1538entirely; but you are allowed to pass an empty argument.  For example,
1539this invocation is invalid in ISO C, because there is no comma after
1540the string:
1541
1542@smallexample
1543debug ("A message")
1544@end smallexample
1545
1546GNU CPP permits you to completely omit the variable arguments in this
1547way.  In the above examples, the compiler would complain, though since
1548the expansion of the macro still has the extra comma after the format
1549string.
1550
1551To help solve this problem, CPP behaves specially for variable arguments
1552used with the token paste operator, @samp{##}.  If instead you write
1553
1554@smallexample
1555#define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
1556@end smallexample
1557
1558and if the variable arguments are omitted or empty, the @samp{##}
1559operator causes the preprocessor to remove the comma before it.  If you
1560do provide some variable arguments in your macro invocation, GNU CPP
1561does not complain about the paste operation and instead places the
1562variable arguments after the comma.  Just like any other pasted macro
1563argument, these arguments are not macro expanded.
1564
1565@node Escaped Newlines
1566@section Slightly Looser Rules for Escaped Newlines
1567@cindex escaped newlines
1568@cindex newlines (escaped)
1569
1570Recently, the preprocessor has relaxed its treatment of escaped
1571newlines.  Previously, the newline had to immediately follow a
1572backslash.  The current implementation allows whitespace in the form
1573of spaces, horizontal and vertical tabs, and form feeds between the
1574backslash and the subsequent newline.  The preprocessor issues a
1575warning, but treats it as a valid escaped newline and combines the two
1576lines to form a single logical line.  This works within comments and
1577tokens, as well as between tokens.  Comments are @emph{not} treated as
1578whitespace for the purposes of this relaxation, since they have not
1579yet been replaced with spaces.
1580
1581@node Subscripting
1582@section Non-Lvalue Arrays May Have Subscripts
1583@cindex subscripting
1584@cindex arrays, non-lvalue
1585
1586@cindex subscripting and function values
1587In ISO C99, arrays that are not lvalues still decay to pointers, and
1588may be subscripted, although they may not be modified or used after
1589the next sequence point and the unary @samp{&} operator may not be
1590applied to them.  As an extension, GCC allows such arrays to be
1591subscripted in C89 mode, though otherwise they do not decay to
1592pointers outside C99 mode.  For example,
1593this is valid in GNU C though not valid in C89:
1594
1595@smallexample
1596@group
1597struct foo @{int a[4];@};
1598
1599struct foo f();
1600
1601bar (int index)
1602@{
1603  return f().a[index];
1604@}
1605@end group
1606@end smallexample
1607
1608@node Pointer Arith
1609@section Arithmetic on @code{void}- and Function-Pointers
1610@cindex void pointers, arithmetic
1611@cindex void, size of pointer to
1612@cindex function pointers, arithmetic
1613@cindex function, size of pointer to
1614
1615In GNU C, addition and subtraction operations are supported on pointers to
1616@code{void} and on pointers to functions.  This is done by treating the
1617size of a @code{void} or of a function as 1.
1618
1619A consequence of this is that @code{sizeof} is also allowed on @code{void}
1620and on function types, and returns 1.
1621
1622@opindex Wpointer-arith
1623The option @option{-Wpointer-arith} requests a warning if these extensions
1624are used.
1625
1626@node Initializers
1627@section Non-Constant Initializers
1628@cindex initializers, non-constant
1629@cindex non-constant initializers
1630
1631As in standard C++ and ISO C99, the elements of an aggregate initializer for an
1632automatic variable are not required to be constant expressions in GNU C@.
1633Here is an example of an initializer with run-time varying elements:
1634
1635@smallexample
1636foo (float f, float g)
1637@{
1638  float beat_freqs[2] = @{ f-g, f+g @};
1639  /* @r{@dots{}} */
1640@}
1641@end smallexample
1642
1643@node Compound Literals
1644@section Compound Literals
1645@cindex constructor expressions
1646@cindex initializations in expressions
1647@cindex structures, constructor expression
1648@cindex expressions, constructor
1649@cindex compound literals
1650@c The GNU C name for what C99 calls compound literals was "constructor expressions".
1651
1652ISO C99 supports compound literals.  A compound literal looks like
1653a cast containing an initializer.  Its value is an object of the
1654type specified in the cast, containing the elements specified in
1655the initializer; it is an lvalue.  As an extension, GCC supports
1656compound literals in C89 mode and in C++.
1657
1658Usually, the specified type is a structure.  Assume that
1659@code{struct foo} and @code{structure} are declared as shown:
1660
1661@smallexample
1662struct foo @{int a; char b[2];@} structure;
1663@end smallexample
1664
1665@noindent
1666Here is an example of constructing a @code{struct foo} with a compound literal:
1667
1668@smallexample
1669structure = ((struct foo) @{x + y, 'a', 0@});
1670@end smallexample
1671
1672@noindent
1673This is equivalent to writing the following:
1674
1675@smallexample
1676@{
1677  struct foo temp = @{x + y, 'a', 0@};
1678  structure = temp;
1679@}
1680@end smallexample
1681
1682You can also construct an array.  If all the elements of the compound literal
1683are (made up of) simple constant expressions, suitable for use in
1684initializers of objects of static storage duration, then the compound
1685literal can be coerced to a pointer to its first element and used in
1686such an initializer, as shown here:
1687
1688@smallexample
1689char **foo = (char *[]) @{ "x", "y", "z" @};
1690@end smallexample
1691
1692Compound literals for scalar types and union types are is
1693also allowed, but then the compound literal is equivalent
1694to a cast.
1695
1696As a GNU extension, GCC allows initialization of objects with static storage
1697duration by compound literals (which is not possible in ISO C99, because
1698the initializer is not a constant).
1699It is handled as if the object was initialized only with the bracket
1700enclosed list if compound literal's and object types match.
1701The initializer list of the compound literal must be constant.
1702If the object being initialized has array type of unknown size, the size is
1703determined by compound literal size.
1704
1705@smallexample
1706static struct foo x = (struct foo) @{1, 'a', 'b'@};
1707static int y[] = (int []) @{1, 2, 3@};
1708static int z[] = (int [3]) @{1@};
1709@end smallexample
1710
1711@noindent
1712The above lines are equivalent to the following:
1713@smallexample
1714static struct foo x = @{1, 'a', 'b'@};
1715static int y[] = @{1, 2, 3@};
1716static int z[] = @{1, 0, 0@};
1717@end smallexample
1718
1719@node Designated Inits
1720@section Designated Initializers
1721@cindex initializers with labeled elements
1722@cindex labeled elements in initializers
1723@cindex case labels in initializers
1724@cindex designated initializers
1725
1726Standard C89 requires the elements of an initializer to appear in a fixed
1727order, the same as the order of the elements in the array or structure
1728being initialized.
1729
1730In ISO C99 you can give the elements in any order, specifying the array
1731indices or structure field names they apply to, and GNU C allows this as
1732an extension in C89 mode as well.  This extension is not
1733implemented in GNU C++.
1734
1735To specify an array index, write
1736@samp{[@var{index}] =} before the element value.  For example,
1737
1738@smallexample
1739int a[6] = @{ [4] = 29, [2] = 15 @};
1740@end smallexample
1741
1742@noindent
1743is equivalent to
1744
1745@smallexample
1746int a[6] = @{ 0, 0, 15, 0, 29, 0 @};
1747@end smallexample
1748
1749@noindent
1750The index values must be constant expressions, even if the array being
1751initialized is automatic.
1752
1753An alternative syntax for this which has been obsolete since GCC 2.5 but
1754GCC still accepts is to write @samp{[@var{index}]} before the element
1755value, with no @samp{=}.
1756
1757To initialize a range of elements to the same value, write
1758@samp{[@var{first} ... @var{last}] = @var{value}}.  This is a GNU
1759extension.  For example,
1760
1761@smallexample
1762int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @};
1763@end smallexample
1764
1765@noindent
1766If the value in it has side-effects, the side-effects will happen only once,
1767not for each initialized field by the range initializer.
1768
1769@noindent
1770Note that the length of the array is the highest value specified
1771plus one.
1772
1773In a structure initializer, specify the name of a field to initialize
1774with @samp{.@var{fieldname} =} before the element value.  For example,
1775given the following structure,
1776
1777@smallexample
1778struct point @{ int x, y; @};
1779@end smallexample
1780
1781@noindent
1782the following initialization
1783
1784@smallexample
1785struct point p = @{ .y = yvalue, .x = xvalue @};
1786@end smallexample
1787
1788@noindent
1789is equivalent to
1790
1791@smallexample
1792struct point p = @{ xvalue, yvalue @};
1793@end smallexample
1794
1795Another syntax which has the same meaning, obsolete since GCC 2.5, is
1796@samp{@var{fieldname}:}, as shown here:
1797
1798@smallexample
1799struct point p = @{ y: yvalue, x: xvalue @};
1800@end smallexample
1801
1802@cindex designators
1803The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a
1804@dfn{designator}.  You can also use a designator (or the obsolete colon
1805syntax) when initializing a union, to specify which element of the union
1806should be used.  For example,
1807
1808@smallexample
1809union foo @{ int i; double d; @};
1810
1811union foo f = @{ .d = 4 @};
1812@end smallexample
1813
1814@noindent
1815will convert 4 to a @code{double} to store it in the union using
1816the second element.  By contrast, casting 4 to type @code{union foo}
1817would store it into the union as the integer @code{i}, since it is
1818an integer.  (@xref{Cast to Union}.)
1819
1820You can combine this technique of naming elements with ordinary C
1821initialization of successive elements.  Each initializer element that
1822does not have a designator applies to the next consecutive element of the
1823array or structure.  For example,
1824
1825@smallexample
1826int a[6] = @{ [1] = v1, v2, [4] = v4 @};
1827@end smallexample
1828
1829@noindent
1830is equivalent to
1831
1832@smallexample
1833int a[6] = @{ 0, v1, v2, 0, v4, 0 @};
1834@end smallexample
1835
1836Labeling the elements of an array initializer is especially useful
1837when the indices are characters or belong to an @code{enum} type.
1838For example:
1839
1840@smallexample
1841int whitespace[256]
1842  = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1,
1843      ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @};
1844@end smallexample
1845
1846@cindex designator lists
1847You can also write a series of @samp{.@var{fieldname}} and
1848@samp{[@var{index}]} designators before an @samp{=} to specify a
1849nested subobject to initialize; the list is taken relative to the
1850subobject corresponding to the closest surrounding brace pair.  For
1851example, with the @samp{struct point} declaration above:
1852
1853@smallexample
1854struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @};
1855@end smallexample
1856
1857@noindent
1858If the same field is initialized multiple times, it will have value from
1859the last initialization.  If any such overridden initialization has
1860side-effect, it is unspecified whether the side-effect happens or not.
1861Currently, GCC will discard them and issue a warning.
1862
1863@node Case Ranges
1864@section Case Ranges
1865@cindex case ranges
1866@cindex ranges in case statements
1867
1868You can specify a range of consecutive values in a single @code{case} label,
1869like this:
1870
1871@smallexample
1872case @var{low} ... @var{high}:
1873@end smallexample
1874
1875@noindent
1876This has the same effect as the proper number of individual @code{case}
1877labels, one for each integer value from @var{low} to @var{high}, inclusive.
1878
1879This feature is especially useful for ranges of ASCII character codes:
1880
1881@smallexample
1882case 'A' ... 'Z':
1883@end smallexample
1884
1885@strong{Be careful:} Write spaces around the @code{...}, for otherwise
1886it may be parsed wrong when you use it with integer values.  For example,
1887write this:
1888
1889@smallexample
1890case 1 ... 5:
1891@end smallexample
1892
1893@noindent
1894rather than this:
1895
1896@smallexample
1897case 1...5:
1898@end smallexample
1899
1900@node Cast to Union
1901@section Cast to a Union Type
1902@cindex cast to a union
1903@cindex union, casting to a
1904
1905A cast to union type is similar to other casts, except that the type
1906specified is a union type.  You can specify the type either with
1907@code{union @var{tag}} or with a typedef name.  A cast to union is actually
1908a constructor though, not a cast, and hence does not yield an lvalue like
1909normal casts.  (@xref{Compound Literals}.)
1910
1911The types that may be cast to the union type are those of the members
1912of the union.  Thus, given the following union and variables:
1913
1914@smallexample
1915union foo @{ int i; double d; @};
1916int x;
1917double y;
1918@end smallexample
1919
1920@noindent
1921both @code{x} and @code{y} can be cast to type @code{union foo}.
1922
1923Using the cast as the right-hand side of an assignment to a variable of
1924union type is equivalent to storing in a member of the union:
1925
1926@smallexample
1927union foo u;
1928/* @r{@dots{}} */
1929u = (union foo) x  @equiv{}  u.i = x
1930u = (union foo) y  @equiv{}  u.d = y
1931@end smallexample
1932
1933You can also use the union cast as a function argument:
1934
1935@smallexample
1936void hack (union foo);
1937/* @r{@dots{}} */
1938hack ((union foo) x);
1939@end smallexample
1940
1941@node Mixed Declarations
1942@section Mixed Declarations and Code
1943@cindex mixed declarations and code
1944@cindex declarations, mixed with code
1945@cindex code, mixed with declarations
1946
1947ISO C99 and ISO C++ allow declarations and code to be freely mixed
1948within compound statements.  As an extension, GCC also allows this in
1949C89 mode.  For example, you could do:
1950
1951@smallexample
1952int i;
1953/* @r{@dots{}} */
1954i++;
1955int j = i + 2;
1956@end smallexample
1957
1958Each identifier is visible from where it is declared until the end of
1959the enclosing block.
1960
1961@node Function Attributes
1962@section Declaring Attributes of Functions
1963@cindex function attributes
1964@cindex declaring attributes of functions
1965@cindex functions that never return
1966@cindex functions that have no side effects
1967@cindex functions in arbitrary sections
1968@cindex functions that behave like malloc
1969@cindex @code{volatile} applied to function
1970@cindex @code{const} applied to function
1971@cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments
1972@cindex functions with non-null pointer arguments
1973@cindex functions that are passed arguments in registers on the 386
1974@cindex functions that pop the argument stack on the 386
1975@cindex functions that do not pop the argument stack on the 386
1976
1977In GNU C, you declare certain things about functions called in your program
1978which help the compiler optimize function calls and check your code more
1979carefully.
1980
1981The keyword @code{__attribute__} allows you to specify special
1982attributes when making a declaration.  This keyword is followed by an
1983attribute specification inside double parentheses.  The following
1984attributes are currently defined for functions on all targets:
1985@code{noreturn}, @code{noinline}, @code{always_inline},
1986@code{pure}, @code{const}, @code{nothrow},
1987@code{format}, @code{format_arg}, @code{no_instrument_function},
1988@code{section}, @code{constructor}, @code{destructor}, @code{used},
1989@code{unused}, @code{deprecated}, @code{weak}, @code{malloc},
1990@code{alias}, @code{warn_unused_result} and @code{nonnull}.  Several other
1991attributes are defined for functions on particular target systems.  Other
1992attributes, including @code{section} are supported for variables declarations
1993(@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}).
1994
1995You may also specify attributes with @samp{__} preceding and following
1996each keyword.  This allows you to use them in header files without
1997being concerned about a possible macro of the same name.  For example,
1998you may use @code{__noreturn__} instead of @code{noreturn}.
1999
2000@xref{Attribute Syntax}, for details of the exact syntax for using
2001attributes.
2002
2003@table @code
2004@cindex @code{noreturn} function attribute
2005@item noreturn
2006A few standard library functions, such as @code{abort} and @code{exit},
2007cannot return.  GCC knows this automatically.  Some programs define
2008their own functions that never return.  You can declare them
2009@code{noreturn} to tell the compiler this fact.  For example,
2010
2011@smallexample
2012@group
2013void fatal () __attribute__ ((noreturn));
2014
2015void
2016fatal (/* @r{@dots{}} */)
2017@{
2018  /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */
2019  exit (1);
2020@}
2021@end group
2022@end smallexample
2023
2024The @code{noreturn} keyword tells the compiler to assume that
2025@code{fatal} cannot return.  It can then optimize without regard to what
2026would happen if @code{fatal} ever did return.  This makes slightly
2027better code.  More importantly, it helps avoid spurious warnings of
2028uninitialized variables.
2029
2030The @code{noreturn} keyword does not affect the exceptional path when that
2031applies: a @code{noreturn}-marked function may still return to the caller
2032by throwing an exception.
2033
2034Do not assume that registers saved by the calling function are
2035restored before calling the @code{noreturn} function.
2036
2037It does not make sense for a @code{noreturn} function to have a return
2038type other than @code{void}.
2039
2040The attribute @code{noreturn} is not implemented in GCC versions
2041earlier than 2.5.  An alternative way to declare that a function does
2042not return, which works in the current version and in some older
2043versions, is as follows:
2044
2045@smallexample
2046typedef void voidfn ();
2047
2048volatile voidfn fatal;
2049@end smallexample
2050
2051@cindex @code{noinline} function attribute
2052@item noinline
2053This function attribute prevents a function from being considered for
2054inlining.
2055
2056@cindex @code{always_inline} function attribute
2057@item always_inline
2058Generally, functions are not inlined unless optimization is specified.
2059For functions declared inline, this attribute inlines the function even
2060if no optimization level was specified.
2061
2062@cindex @code{pure} function attribute
2063@item pure
2064Many functions have no effects except the return value and their
2065return value depends only on the parameters and/or global variables.
2066Such a function can be subject
2067to common subexpression elimination and loop optimization just as an
2068arithmetic operator would be.  These functions should be declared
2069with the attribute @code{pure}.  For example,
2070
2071@smallexample
2072int square (int) __attribute__ ((pure));
2073@end smallexample
2074
2075@noindent
2076says that the hypothetical function @code{square} is safe to call
2077fewer times than the program says.
2078
2079Some of common examples of pure functions are @code{strlen} or @code{memcmp}.
2080Interesting non-pure functions are functions with infinite loops or those
2081depending on volatile memory or other system resource, that may change between
2082two consecutive calls (such as @code{feof} in a multithreading environment).
2083
2084The attribute @code{pure} is not implemented in GCC versions earlier
2085than 2.96.
2086@cindex @code{const} function attribute
2087@item const
2088Many functions do not examine any values except their arguments, and
2089have no effects except the return value.  Basically this is just slightly
2090more strict class than the @code{pure} attribute above, since function is not
2091allowed to read global memory.
2092
2093@cindex pointer arguments
2094Note that a function that has pointer arguments and examines the data
2095pointed to must @emph{not} be declared @code{const}.  Likewise, a
2096function that calls a non-@code{const} function usually must not be
2097@code{const}.  It does not make sense for a @code{const} function to
2098return @code{void}.
2099
2100The attribute @code{const} is not implemented in GCC versions earlier
2101than 2.5.  An alternative way to declare that a function has no side
2102effects, which works in the current version and in some older versions,
2103is as follows:
2104
2105@smallexample
2106typedef int intfn ();
2107
2108extern const intfn square;
2109@end smallexample
2110
2111This approach does not work in GNU C++ from 2.6.0 on, since the language
2112specifies that the @samp{const} must be attached to the return value.
2113
2114@cindex @code{nothrow} function attribute
2115@item nothrow
2116The @code{nothrow} attribute is used to inform the compiler that a
2117function cannot throw an exception.  For example, most functions in
2118the standard C library can be guaranteed not to throw an exception
2119with the notable exceptions of @code{qsort} and @code{bsearch} that
2120take function pointer arguments.  The @code{nothrow} attribute is not
2121implemented in GCC versions earlier than 3.2.
2122
2123@item format (@var{archetype}, @var{string-index}, @var{first-to-check})
2124@cindex @code{format} function attribute
2125@opindex Wformat
2126The @code{format} attribute specifies that a function takes @code{printf},
2127@code{scanf}, @code{strftime} or @code{strfmon} style arguments which
2128should be type-checked against a format string.  For example, the
2129declaration:
2130
2131@smallexample
2132extern int
2133my_printf (void *my_object, const char *my_format, ...)
2134      __attribute__ ((format (printf, 2, 3)));
2135@end smallexample
2136
2137@noindent
2138causes the compiler to check the arguments in calls to @code{my_printf}
2139for consistency with the @code{printf} style format string argument
2140@code{my_format}.
2141
2142The parameter @var{archetype} determines how the format string is
2143interpreted, and should be @code{printf}, @code{scanf}, @code{strftime}
2144or @code{strfmon}.  (You can also use @code{__printf__},
2145@code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.)  The
2146parameter @var{string-index} specifies which argument is the format
2147string argument (starting from 1), while @var{first-to-check} is the
2148number of the first argument to check against the format string.  For
2149functions where the arguments are not available to be checked (such as
2150@code{vprintf}), specify the third parameter as zero.  In this case the
2151compiler only checks the format string for consistency.  For
2152@code{strftime} formats, the third parameter is required to be zero.
2153Since non-static C++ methods have an implicit @code{this} argument, the
2154arguments of such methods should be counted from two, not one, when
2155giving values for @var{string-index} and @var{first-to-check}.
2156
2157In the example above, the format string (@code{my_format}) is the second
2158argument of the function @code{my_print}, and the arguments to check
2159start with the third argument, so the correct parameters for the format
2160attribute are 2 and 3.
2161
2162@opindex ffreestanding
2163The @code{format} attribute allows you to identify your own functions
2164which take format strings as arguments, so that GCC can check the
2165calls to these functions for errors.  The compiler always (unless
2166@option{-ffreestanding} is used) checks formats
2167for the standard library functions @code{printf}, @code{fprintf},
2168@code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime},
2169@code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such
2170warnings are requested (using @option{-Wformat}), so there is no need to
2171modify the header file @file{stdio.h}.  In C99 mode, the functions
2172@code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and
2173@code{vsscanf} are also checked.  Except in strictly conforming C
2174standard modes, the X/Open function @code{strfmon} is also checked as
2175are @code{printf_unlocked} and @code{fprintf_unlocked}.
2176@xref{C Dialect Options,,Options Controlling C Dialect}.
2177
2178@item format_arg (@var{string-index})
2179@cindex @code{format_arg} function attribute
2180@opindex Wformat-nonliteral
2181The @code{format_arg} attribute specifies that a function takes a format
2182string for a @code{printf}, @code{scanf}, @code{strftime} or
2183@code{strfmon} style function and modifies it (for example, to translate
2184it into another language), so the result can be passed to a
2185@code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style
2186function (with the remaining arguments to the format function the same
2187as they would have been for the unmodified string).  For example, the
2188declaration:
2189
2190@smallexample
2191extern char *
2192my_dgettext (char *my_domain, const char *my_format)
2193      __attribute__ ((format_arg (2)));
2194@end smallexample
2195
2196@noindent
2197causes the compiler to check the arguments in calls to a @code{printf},
2198@code{scanf}, @code{strftime} or @code{strfmon} type function, whose
2199format string argument is a call to the @code{my_dgettext} function, for
2200consistency with the format string argument @code{my_format}.  If the
2201@code{format_arg} attribute had not been specified, all the compiler
2202could tell in such calls to format functions would be that the format
2203string argument is not constant; this would generate a warning when
2204@option{-Wformat-nonliteral} is used, but the calls could not be checked
2205without the attribute.
2206
2207The parameter @var{string-index} specifies which argument is the format
2208string argument (starting from one).  Since non-static C++ methods have
2209an implicit @code{this} argument, the arguments of such methods should
2210be counted from two.
2211
2212The @code{format-arg} attribute allows you to identify your own
2213functions which modify format strings, so that GCC can check the
2214calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon}
2215type function whose operands are a call to one of your own function.
2216The compiler always treats @code{gettext}, @code{dgettext}, and
2217@code{dcgettext} in this manner except when strict ISO C support is
2218requested by @option{-ansi} or an appropriate @option{-std} option, or
2219@option{-ffreestanding} is used.  @xref{C Dialect Options,,Options
2220Controlling C Dialect}.
2221
2222@item nonnull (@var{arg-index}, @dots{})
2223@cindex @code{nonnull} function attribute
2224The @code{nonnull} attribute specifies that some function parameters should
2225be non-null pointers.  For instance, the declaration:
2226
2227@smallexample
2228extern void *
2229my_memcpy (void *dest, const void *src, size_t len)
2230	__attribute__((nonnull (1, 2)));
2231@end smallexample
2232
2233@noindent
2234causes the compiler to check that, in calls to @code{my_memcpy},
2235arguments @var{dest} and @var{src} are non-null.  If the compiler
2236determines that a null pointer is passed in an argument slot marked
2237as non-null, and the @option{-Wnonnull} option is enabled, a warning
2238is issued.  The compiler may also choose to make optimizations based
2239on the knowledge that certain function arguments will not be null.
2240
2241If no argument index list is given to the @code{nonnull} attribute,
2242all pointer arguments are marked as non-null.  To illustrate, the
2243following declaration is equivalent to the previous example:
2244
2245@smallexample
2246extern void *
2247my_memcpy (void *dest, const void *src, size_t len)
2248	__attribute__((nonnull));
2249@end smallexample
2250
2251@item no_instrument_function
2252@cindex @code{no_instrument_function} function attribute
2253@opindex finstrument-functions
2254If @option{-finstrument-functions} is given, profiling function calls will
2255be generated at entry and exit of most user-compiled functions.
2256Functions with this attribute will not be so instrumented.
2257
2258@item section ("@var{section-name}")
2259@cindex @code{section} function attribute
2260Normally, the compiler places the code it generates in the @code{text} section.
2261Sometimes, however, you need additional sections, or you need certain
2262particular functions to appear in special sections.  The @code{section}
2263attribute specifies that a function lives in a particular section.
2264For example, the declaration:
2265
2266@smallexample
2267extern void foobar (void) __attribute__ ((section ("bar")));
2268@end smallexample
2269
2270@noindent
2271puts the function @code{foobar} in the @code{bar} section.
2272
2273Some file formats do not support arbitrary sections so the @code{section}
2274attribute is not available on all platforms.
2275If you need to map the entire contents of a module to a particular
2276section, consider using the facilities of the linker instead.
2277
2278@item constructor
2279@itemx destructor
2280@cindex @code{constructor} function attribute
2281@cindex @code{destructor} function attribute
2282The @code{constructor} attribute causes the function to be called
2283automatically before execution enters @code{main ()}.  Similarly, the
2284@code{destructor} attribute causes the function to be called
2285automatically after @code{main ()} has completed or @code{exit ()} has
2286been called.  Functions with these attributes are useful for
2287initializing data that will be used implicitly during the execution of
2288the program.
2289
2290These attributes are not currently implemented for Objective-C@.
2291
2292@cindex @code{unused} attribute.
2293@item unused
2294This attribute, attached to a function, means that the function is meant
2295to be possibly unused.  GCC will not produce a warning for this
2296function.
2297
2298@cindex @code{used} attribute.
2299@item used
2300This attribute, attached to a function, means that code must be emitted
2301for the function even if it appears that the function is not referenced.
2302This is useful, for example, when the function is referenced only in
2303inline assembly.
2304
2305@cindex @code{deprecated} attribute.
2306@item deprecated
2307The @code{deprecated} attribute results in a warning if the function
2308is used anywhere in the source file.  This is useful when identifying
2309functions that are expected to be removed in a future version of a
2310program.  The warning also includes the location of the declaration
2311of the deprecated function, to enable users to easily find further
2312information about why the function is deprecated, or what they should
2313do instead.  Note that the warnings only occurs for uses:
2314
2315@smallexample
2316int old_fn () __attribute__ ((deprecated));
2317int old_fn ();
2318int (*fn_ptr)() = old_fn;
2319@end smallexample
2320
2321results in a warning on line 3 but not line 2.
2322
2323The @code{deprecated} attribute can also be used for variables and
2324types (@pxref{Variable Attributes}, @pxref{Type Attributes}.)
2325
2326@item warn_unused_result
2327@cindex @code{warn_unused_result} attribute
2328The @code{warn_unused_result} attribute causes a warning to be emitted
2329if a caller of the function with this attribute does not use its
2330return value.  This is useful for functions where not checking
2331the result is either a security problem or always a bug, such as
2332@code{realloc}.
2333
2334@smallexample
2335int fn () __attribute__ ((warn_unused_result));
2336int foo ()
2337@{
2338  if (fn () < 0) return -1;
2339  fn ();
2340  return 0;
2341@}
2342@end smallexample
2343
2344results in warning on line 5.
2345
2346@item weak
2347@cindex @code{weak} attribute
2348The @code{weak} attribute causes the declaration to be emitted as a weak
2349symbol rather than a global.  This is primarily useful in defining
2350library functions which can be overridden in user code, though it can
2351also be used with non-function declarations.  Weak symbols are supported
2352for ELF targets, and also for a.out targets when using the GNU assembler
2353and linker.
2354
2355@item malloc
2356@cindex @code{malloc} attribute
2357The @code{malloc} attribute is used to tell the compiler that a function
2358may be treated as if any non-@code{NULL} pointer it returns cannot
2359alias any other pointer valid when the function returns.
2360This will often improve optimization.
2361Standard functions with this property include @code{malloc} and
2362@code{calloc}.  @code{realloc}-like functions have this property as
2363long as the old pointer is never referred to (including comparing it
2364to the new pointer) after the function returns a non-@code{NULL}
2365value.
2366
2367@item alias ("@var{target}")
2368@cindex @code{alias} attribute
2369The @code{alias} attribute causes the declaration to be emitted as an
2370alias for another symbol, which must be specified.  For instance,
2371
2372@smallexample
2373void __f () @{ /* @r{Do something.} */; @}
2374void f () __attribute__ ((weak, alias ("__f")));
2375@end smallexample
2376
2377declares @samp{f} to be a weak alias for @samp{__f}.  In C++, the
2378mangled name for the target must be used.
2379
2380Not all target machines support this attribute.
2381
2382@item visibility ("@var{visibility_type}")
2383@cindex @code{visibility} attribute
2384The @code{visibility} attribute on ELF targets causes the declaration
2385to be emitted with default, hidden, protected or internal visibility.
2386
2387@smallexample
2388void __attribute__ ((visibility ("protected")))
2389f () @{ /* @r{Do something.} */; @}
2390int i __attribute__ ((visibility ("hidden")));
2391@end smallexample
2392
2393See the ELF gABI for complete details, but the short story is:
2394
2395@table @dfn
2396@item default
2397Default visibility is the normal case for ELF.  This value is
2398available for the visibility attribute to override other options
2399that may change the assumed visibility of symbols.
2400
2401@item hidden
2402Hidden visibility indicates that the symbol will not be placed into
2403the dynamic symbol table, so no other @dfn{module} (executable or
2404shared library) can reference it directly.
2405
2406@item protected
2407Protected visibility indicates that the symbol will be placed in the
2408dynamic symbol table, but that references within the defining module
2409will bind to the local symbol.  That is, the symbol cannot be overridden
2410by another module.
2411
2412@item internal
2413Internal visibility is like hidden visibility, but with additional
2414processor specific semantics.  Unless otherwise specified by the psABI,
2415GCC defines internal visibility to mean that the function is @emph{never}
2416called from another module.  Note that hidden symbols, while they cannot
2417be referenced directly by other modules, can be referenced indirectly via
2418function pointers.  By indicating that a symbol cannot be called from
2419outside the module, GCC may for instance omit the load of a PIC register
2420since it is known that the calling function loaded the correct value.
2421@end table
2422
2423Not all ELF targets support this attribute.
2424
2425@item regparm (@var{number})
2426@cindex @code{regparm} attribute
2427@cindex functions that are passed arguments in registers on the 386
2428On the Intel 386, the @code{regparm} attribute causes the compiler to
2429pass up to @var{number} integer arguments in registers EAX,
2430EDX, and ECX instead of on the stack.  Functions that take a
2431variable number of arguments will continue to be passed all of their
2432arguments on the stack.
2433
2434Beware that on some ELF systems this attribute is unsuitable for
2435global functions in shared libraries with lazy binding (which is the
2436default).  Lazy binding will send the first call via resolving code in
2437the loader, which might assume EAX, EDX and ECX can be clobbered, as
2438per the standard calling conventions.  Solaris 8 is affected by this.
2439GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be
2440safe since the loaders there save all registers.  (Lazy binding can be
2441disabled with the linker or the loader if desired, to avoid the
2442problem.)
2443
2444@item stdcall
2445@cindex functions that pop the argument stack on the 386
2446On the Intel 386, the @code{stdcall} attribute causes the compiler to
2447assume that the called function will pop off the stack space used to
2448pass arguments, unless it takes a variable number of arguments.
2449
2450@item fastcall
2451@cindex functions that pop the argument stack on the 386
2452On the Intel 386, the @code{fastcall} attribute causes the compiler to
2453pass the first two arguments in the registers ECX and EDX. Subsequent
2454arguments are passed on the stack. The called function will pop the
2455arguments off the stack. If the number of arguments is variable all
2456arguments are pushed on the stack.
2457
2458@item cdecl
2459@cindex functions that do pop the argument stack on the 386
2460@opindex mrtd
2461On the Intel 386, the @code{cdecl} attribute causes the compiler to
2462assume that the calling function will pop off the stack space used to
2463pass arguments.  This is
2464useful to override the effects of the @option{-mrtd} switch.
2465
2466@item longcall/shortcall
2467@cindex functions called via pointer on the RS/6000 and PowerPC
2468On the RS/6000 and PowerPC, the @code{longcall} attribute causes the
2469compiler to always call this function via a pointer, just as it would if
2470the @option{-mlongcall} option had been specified.  The @code{shortcall}
2471attribute causes the compiler not to do this.  These attributes override
2472both the @option{-mlongcall} switch and the @code{#pragma longcall}
2473setting.
2474
2475@xref{RS/6000 and PowerPC Options}, for more information on whether long
2476calls are necessary.
2477
2478@item long_call/short_call
2479@cindex indirect calls on ARM
2480This attribute specifies how a particular function is called on
2481ARM@.  Both attributes override the @option{-mlong-calls} (@pxref{ARM Options})
2482command line switch and @code{#pragma long_calls} settings.  The
2483@code{long_call} attribute causes the compiler to always call the
2484function by first loading its address into a register and then using the
2485contents of that register.   The @code{short_call} attribute always places
2486the offset to the function from the call site into the @samp{BL}
2487instruction directly.
2488
2489@item function_vector
2490@cindex calling functions through the function vector on the H8/300 processors
2491Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2492function should be called through the function vector.  Calling a
2493function through the function vector will reduce code size, however;
2494the function vector has a limited size (maximum 128 entries on the H8/300
2495and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector.
2496
2497You must use GAS and GLD from GNU binutils version 2.7 or later for
2498this attribute to work correctly.
2499
2500@item interrupt
2501@cindex interrupt handler functions
2502Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16 ports to indicate
2503that the specified function is an interrupt handler.  The compiler will
2504generate function entry and exit sequences suitable for use in an
2505interrupt handler when this attribute is present.
2506
2507Note, interrupt handlers for the m68k, H8/300, H8/300H, H8S, and SH processors
2508can be specified via the @code{interrupt_handler} attribute.
2509
2510Note, on the AVR, interrupts will be enabled inside the function.
2511
2512Note, for the ARM, you can specify the kind of interrupt to be handled by
2513adding an optional parameter to the interrupt attribute like this:
2514
2515@smallexample
2516void f () __attribute__ ((interrupt ("IRQ")));
2517@end smallexample
2518
2519Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@.
2520
2521@item interrupt_handler
2522@cindex interrupt handler functions on the m68k, H8/300 and SH processors
2523Use this attribute on the m68k, H8/300, H8/300H, H8S, and SH to indicate that
2524the specified function is an interrupt handler.  The compiler will generate
2525function entry and exit sequences suitable for use in an interrupt
2526handler when this attribute is present.
2527
2528@item sp_switch
2529Use this attribute on the SH to indicate an @code{interrupt_handler}
2530function should switch to an alternate stack.  It expects a string
2531argument that names a global variable holding the address of the
2532alternate stack.
2533
2534@smallexample
2535void *alt_stack;
2536void f () __attribute__ ((interrupt_handler,
2537                          sp_switch ("alt_stack")));
2538@end smallexample
2539
2540@item trap_exit
2541Use this attribute on the SH for an @code{interrupt_handler} to return using
2542@code{trapa} instead of @code{rte}.  This attribute expects an integer
2543argument specifying the trap number to be used.
2544
2545@item eightbit_data
2546@cindex eight bit data on the H8/300, H8/300H, and H8S
2547Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified
2548variable should be placed into the eight bit data section.
2549The compiler will generate more efficient code for certain operations
2550on data in the eight bit data area.  Note the eight bit data area is limited to
2551256 bytes of data.
2552
2553You must use GAS and GLD from GNU binutils version 2.7 or later for
2554this attribute to work correctly.
2555
2556@item tiny_data
2557@cindex tiny data section on the H8/300H and H8S
2558Use this attribute on the H8/300H and H8S to indicate that the specified
2559variable should be placed into the tiny data section.
2560The compiler will generate more efficient code for loads and stores
2561on data in the tiny data section.  Note the tiny data area is limited to
2562slightly under 32kbytes of data.
2563
2564@item saveall
2565@cindex save all registers on the H8/300, H8/300H, and H8S
2566Use this attribute on the H8/300, H8/300H, and H8S to indicate that
2567all registers except the stack pointer should be saved in the prologue
2568regardless of whether they are used or not.
2569
2570@item signal
2571@cindex signal handler functions on the AVR processors
2572Use this attribute on the AVR to indicate that the specified
2573function is a signal handler.  The compiler will generate function
2574entry and exit sequences suitable for use in a signal handler when this
2575attribute is present.  Interrupts will be disabled inside the function.
2576
2577@item naked
2578@cindex function without a prologue/epilogue code
2579Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the
2580specified function does not need prologue/epilogue sequences generated by
2581the compiler.  It is up to the programmer to provide these sequences.
2582
2583@item model (@var{model-name})
2584@cindex function addressability on the M32R/D
2585@cindex variable addressability on the IA-64
2586
2587On the M32R/D, use this attribute to set the addressability of an
2588object, and of the code generated for a function.  The identifier
2589@var{model-name} is one of @code{small}, @code{medium}, or
2590@code{large}, representing each of the code models.
2591
2592Small model objects live in the lower 16MB of memory (so that their
2593addresses can be loaded with the @code{ld24} instruction), and are
2594callable with the @code{bl} instruction.
2595
2596Medium model objects may live anywhere in the 32-bit address space (the
2597compiler will generate @code{seth/add3} instructions to load their addresses),
2598and are callable with the @code{bl} instruction.
2599
2600Large model objects may live anywhere in the 32-bit address space (the
2601compiler will generate @code{seth/add3} instructions to load their addresses),
2602and may not be reachable with the @code{bl} instruction (the compiler will
2603generate the much slower @code{seth/add3/jl} instruction sequence).
2604
2605On IA-64, use this attribute to set the addressability of an object.
2606At present, the only supported identifier for @var{model-name} is
2607@code{small}, indicating addressability via ``small'' (22-bit)
2608addresses (so that their addresses can be loaded with the @code{addl}
2609instruction).  Caveat: such addressing is by definition not position
2610independent and hence this attribute must not be used for objects
2611defined by shared libraries.
2612
2613@item far
2614@cindex functions which handle memory bank switching
2615On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to
2616use a calling convention that takes care of switching memory banks when
2617entering and leaving a function.  This calling convention is also the
2618default when using the @option{-mlong-calls} option.
2619
2620On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions
2621to call and return from a function.
2622
2623On 68HC11 the compiler will generate a sequence of instructions
2624to invoke a board-specific routine to switch the memory bank and call the
2625real function. The board-specific routine simulates a @code{call}.
2626At the end of a function, it will jump to a board-specific routine
2627instead of using @code{rts}. The board-specific return routine simulates
2628the @code{rtc}.
2629
2630@item near
2631@cindex functions which do not handle memory bank switching on 68HC11/68HC12
2632On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to
2633use the normal calling convention based on @code{jsr} and @code{rts}.
2634This attribute can be used to cancel the effect of the @option{-mlong-calls}
2635option.
2636
2637@item dllimport
2638@cindex @code{__declspec(dllimport)}
2639On Microsoft Windows targets, the @code{dllimport} attribute causes the compiler
2640to reference a function or variable via a global pointer to a pointer
2641that is set up by the Microsoft Windows dll library. The pointer name is formed by
2642combining @code{_imp__} and the function or variable name. The attribute
2643implies @code{extern} storage.
2644
2645Currently, the attribute is ignored for inlined functions. If the
2646attribute is applied to a symbol @emph{definition}, an error is reported.
2647If a symbol previously declared @code{dllimport} is later defined, the
2648attribute is ignored in subsequent references, and a warning is emitted.
2649The attribute is also overridden by a subsequent declaration as
2650@code{dllexport}.
2651
2652When applied to C++ classes, the attribute marks non-inlined
2653member functions and static data members as imports.  However, the
2654attribute is ignored for virtual methods to allow creation of vtables
2655using thunks.
2656
2657On cygwin, mingw and arm-pe targets, @code{__declspec(dllimport)} is
2658recognized as a synonym for @code{__attribute__ ((dllimport))} for
2659compatibility with other Microsoft Windows compilers.
2660
2661The use of the @code{dllimport} attribute on functions is not necessary,
2662but provides a small performance benefit by eliminating a thunk in the
2663dll. The use of the @code{dllimport} attribute on imported variables was
2664required on older versions of GNU ld, but can now be avoided by passing
2665the @option{--enable-auto-import} switch to ld. As with functions, using
2666the attribute for a variable eliminates a thunk in the dll.
2667
2668One drawback to using this attribute is that a pointer to a function or
2669variable marked as dllimport cannot be used as a constant address. The
2670attribute can be disabled for functions by setting the
2671@option{-mnop-fun-dllimport} flag.
2672
2673@item dllexport
2674@cindex @code{__declspec(dllexport)}
2675On Microsoft Windows targets the @code{dllexport} attribute causes the compiler to
2676provide a global pointer to a pointer in a dll, so that it can be
2677referenced with the @code{dllimport} attribute. The pointer name is
2678formed by combining @code{_imp__} and the function or variable name.
2679
2680Currently, the @code{dllexport}attribute is ignored for inlined
2681functions, but export can be forced by using the
2682@option{-fkeep-inline-functions} flag. The attribute is also ignored for
2683undefined symbols.
2684
2685When applied to C++ classes. the attribute marks defined non-inlined
2686member functions and static data members as exports. Static consts
2687initialized in-class are not marked unless they are also defined
2688out-of-class.
2689
2690On cygwin, mingw and arm-pe targets, @code{__declspec(dllexport)} is
2691recognized as a synonym for @code{__attribute__ ((dllexport))} for
2692compatibility with other Microsoft Windows compilers.
2693
2694Alternative methods for including the symbol in the dll's export table
2695are to use a .def file with an @code{EXPORTS} section or, with GNU ld,
2696using the @option{--export-all} linker flag.
2697
2698@end table
2699
2700You can specify multiple attributes in a declaration by separating them
2701by commas within the double parentheses or by immediately following an
2702attribute declaration with another attribute declaration.
2703
2704@cindex @code{#pragma}, reason for not using
2705@cindex pragma, reason for not using
2706Some people object to the @code{__attribute__} feature, suggesting that
2707ISO C's @code{#pragma} should be used instead.  At the time
2708@code{__attribute__} was designed, there were two reasons for not doing
2709this.
2710
2711@enumerate
2712@item
2713It is impossible to generate @code{#pragma} commands from a macro.
2714
2715@item
2716There is no telling what the same @code{#pragma} might mean in another
2717compiler.
2718@end enumerate
2719
2720These two reasons applied to almost any application that might have been
2721proposed for @code{#pragma}.  It was basically a mistake to use
2722@code{#pragma} for @emph{anything}.
2723
2724The ISO C99 standard includes @code{_Pragma}, which now allows pragmas
2725to be generated from macros.  In addition, a @code{#pragma GCC}
2726namespace is now in use for GCC-specific pragmas.  However, it has been
2727found convenient to use @code{__attribute__} to achieve a natural
2728attachment of attributes to their corresponding declarations, whereas
2729@code{#pragma GCC} is of use for constructs that do not naturally form
2730part of the grammar.  @xref{Other Directives,,Miscellaneous
2731Preprocessing Directives, cpp, The GNU C Preprocessor}.
2732
2733@node Attribute Syntax
2734@section Attribute Syntax
2735@cindex attribute syntax
2736
2737This section describes the syntax with which @code{__attribute__} may be
2738used, and the constructs to which attribute specifiers bind, for the C
2739language.  Some details may vary for C++ and Objective-C@.  Because of
2740infelicities in the grammar for attributes, some forms described here
2741may not be successfully parsed in all cases.
2742
2743There are some problems with the semantics of attributes in C++.  For
2744example, there are no manglings for attributes, although they may affect
2745code generation, so problems may arise when attributed types are used in
2746conjunction with templates or overloading.  Similarly, @code{typeid}
2747does not distinguish between types with different attributes.  Support
2748for attributes in C++ may be restricted in future to attributes on
2749declarations only, but not on nested declarators.
2750
2751@xref{Function Attributes}, for details of the semantics of attributes
2752applying to functions.  @xref{Variable Attributes}, for details of the
2753semantics of attributes applying to variables.  @xref{Type Attributes},
2754for details of the semantics of attributes applying to structure, union
2755and enumerated types.
2756
2757An @dfn{attribute specifier} is of the form
2758@code{__attribute__ ((@var{attribute-list}))}.  An @dfn{attribute list}
2759is a possibly empty comma-separated sequence of @dfn{attributes}, where
2760each attribute is one of the following:
2761
2762@itemize @bullet
2763@item
2764Empty.  Empty attributes are ignored.
2765
2766@item
2767A word (which may be an identifier such as @code{unused}, or a reserved
2768word such as @code{const}).
2769
2770@item
2771A word, followed by, in parentheses, parameters for the attribute.
2772These parameters take one of the following forms:
2773
2774@itemize @bullet
2775@item
2776An identifier.  For example, @code{mode} attributes use this form.
2777
2778@item
2779An identifier followed by a comma and a non-empty comma-separated list
2780of expressions.  For example, @code{format} attributes use this form.
2781
2782@item
2783A possibly empty comma-separated list of expressions.  For example,
2784@code{format_arg} attributes use this form with the list being a single
2785integer constant expression, and @code{alias} attributes use this form
2786with the list being a single string constant.
2787@end itemize
2788@end itemize
2789
2790An @dfn{attribute specifier list} is a sequence of one or more attribute
2791specifiers, not separated by any other tokens.
2792
2793In GNU C, an attribute specifier list may appear after the colon following a
2794label, other than a @code{case} or @code{default} label.  The only
2795attribute it makes sense to use after a label is @code{unused}.  This
2796feature is intended for code generated by programs which contains labels
2797that may be unused but which is compiled with @option{-Wall}.  It would
2798not normally be appropriate to use in it human-written code, though it
2799could be useful in cases where the code that jumps to the label is
2800contained within an @code{#ifdef} conditional. GNU C++ does not permit
2801such placement of attribute lists, as it is permissible for a
2802declaration, which could begin with an attribute list, to be labelled in
2803C++. Declarations cannot be labelled in C90 or C99, so the ambiguity
2804does not arise there.
2805
2806An attribute specifier list may appear as part of a @code{struct},
2807@code{union} or @code{enum} specifier.  It may go either immediately
2808after the @code{struct}, @code{union} or @code{enum} keyword, or after
2809the closing brace.  It is ignored if the content of the structure, union
2810or enumerated type is not defined in the specifier in which the
2811attribute specifier list is used---that is, in usages such as
2812@code{struct __attribute__((foo)) bar} with no following opening brace.
2813Where attribute specifiers follow the closing brace, they are considered
2814to relate to the structure, union or enumerated type defined, not to any
2815enclosing declaration the type specifier appears in, and the type
2816defined is not complete until after the attribute specifiers.
2817@c Otherwise, there would be the following problems: a shift/reduce
2818@c conflict between attributes binding the struct/union/enum and
2819@c binding to the list of specifiers/qualifiers; and "aligned"
2820@c attributes could use sizeof for the structure, but the size could be
2821@c changed later by "packed" attributes.
2822
2823Otherwise, an attribute specifier appears as part of a declaration,
2824counting declarations of unnamed parameters and type names, and relates
2825to that declaration (which may be nested in another declaration, for
2826example in the case of a parameter declaration), or to a particular declarator
2827within a declaration.  Where an
2828attribute specifier is applied to a parameter declared as a function or
2829an array, it should apply to the function or array rather than the
2830pointer to which the parameter is implicitly converted, but this is not
2831yet correctly implemented.
2832
2833Any list of specifiers and qualifiers at the start of a declaration may
2834contain attribute specifiers, whether or not such a list may in that
2835context contain storage class specifiers.  (Some attributes, however,
2836are essentially in the nature of storage class specifiers, and only make
2837sense where storage class specifiers may be used; for example,
2838@code{section}.)  There is one necessary limitation to this syntax: the
2839first old-style parameter declaration in a function definition cannot
2840begin with an attribute specifier, because such an attribute applies to
2841the function instead by syntax described below (which, however, is not
2842yet implemented in this case).  In some other cases, attribute
2843specifiers are permitted by this grammar but not yet supported by the
2844compiler.  All attribute specifiers in this place relate to the
2845declaration as a whole.  In the obsolescent usage where a type of
2846@code{int} is implied by the absence of type specifiers, such a list of
2847specifiers and qualifiers may be an attribute specifier list with no
2848other specifiers or qualifiers.
2849
2850An attribute specifier list may appear immediately before a declarator
2851(other than the first) in a comma-separated list of declarators in a
2852declaration of more than one identifier using a single list of
2853specifiers and qualifiers.  Such attribute specifiers apply
2854only to the identifier before whose declarator they appear.  For
2855example, in
2856
2857@smallexample
2858__attribute__((noreturn)) void d0 (void),
2859    __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
2860     d2 (void)
2861@end smallexample
2862
2863@noindent
2864the @code{noreturn} attribute applies to all the functions
2865declared; the @code{format} attribute only applies to @code{d1}.
2866
2867An attribute specifier list may appear immediately before the comma,
2868@code{=} or semicolon terminating the declaration of an identifier other
2869than a function definition.  At present, such attribute specifiers apply
2870to the declared object or function, but in future they may attach to the
2871outermost adjacent declarator.  In simple cases there is no difference,
2872but, for example, in
2873
2874@smallexample
2875void (****f)(void) __attribute__((noreturn));
2876@end smallexample
2877
2878@noindent
2879at present the @code{noreturn} attribute applies to @code{f}, which
2880causes a warning since @code{f} is not a function, but in future it may
2881apply to the function @code{****f}.  The precise semantics of what
2882attributes in such cases will apply to are not yet specified.  Where an
2883assembler name for an object or function is specified (@pxref{Asm
2884Labels}), at present the attribute must follow the @code{asm}
2885specification; in future, attributes before the @code{asm} specification
2886may apply to the adjacent declarator, and those after it to the declared
2887object or function.
2888
2889An attribute specifier list may, in future, be permitted to appear after
2890the declarator in a function definition (before any old-style parameter
2891declarations or the function body).
2892
2893Attribute specifiers may be mixed with type qualifiers appearing inside
2894the @code{[]} of a parameter array declarator, in the C99 construct by
2895which such qualifiers are applied to the pointer to which the array is
2896implicitly converted.  Such attribute specifiers apply to the pointer,
2897not to the array, but at present this is not implemented and they are
2898ignored.
2899
2900An attribute specifier list may appear at the start of a nested
2901declarator.  At present, there are some limitations in this usage: the
2902attributes correctly apply to the declarator, but for most individual
2903attributes the semantics this implies are not implemented.
2904When attribute specifiers follow the @code{*} of a pointer
2905declarator, they may be mixed with any type qualifiers present.
2906The following describes the formal semantics of this syntax.  It will make the
2907most sense if you are familiar with the formal specification of
2908declarators in the ISO C standard.
2909
2910Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T
2911D1}, where @code{T} contains declaration specifiers that specify a type
2912@var{Type} (such as @code{int}) and @code{D1} is a declarator that
2913contains an identifier @var{ident}.  The type specified for @var{ident}
2914for derived declarators whose type does not include an attribute
2915specifier is as in the ISO C standard.
2916
2917If @code{D1} has the form @code{( @var{attribute-specifier-list} D )},
2918and the declaration @code{T D} specifies the type
2919``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2920@code{T D1} specifies the type ``@var{derived-declarator-type-list}
2921@var{attribute-specifier-list} @var{Type}'' for @var{ident}.
2922
2923If @code{D1} has the form @code{*
2924@var{type-qualifier-and-attribute-specifier-list} D}, and the
2925declaration @code{T D} specifies the type
2926``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then
2927@code{T D1} specifies the type ``@var{derived-declarator-type-list}
2928@var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for
2929@var{ident}.
2930
2931For example,
2932
2933@smallexample
2934void (__attribute__((noreturn)) ****f) (void);
2935@end smallexample
2936
2937@noindent
2938specifies the type ``pointer to pointer to pointer to pointer to
2939non-returning function returning @code{void}''.  As another example,
2940
2941@smallexample
2942char *__attribute__((aligned(8))) *f;
2943@end smallexample
2944
2945@noindent
2946specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''.
2947Note again that this does not work with most attributes; for example,
2948the usage of @samp{aligned} and @samp{noreturn} attributes given above
2949is not yet supported.
2950
2951For compatibility with existing code written for compiler versions that
2952did not implement attributes on nested declarators, some laxity is
2953allowed in the placing of attributes.  If an attribute that only applies
2954to types is applied to a declaration, it will be treated as applying to
2955the type of that declaration.  If an attribute that only applies to
2956declarations is applied to the type of a declaration, it will be treated
2957as applying to that declaration; and, for compatibility with code
2958placing the attributes immediately before the identifier declared, such
2959an attribute applied to a function return type will be treated as
2960applying to the function type, and such an attribute applied to an array
2961element type will be treated as applying to the array type.  If an
2962attribute that only applies to function types is applied to a
2963pointer-to-function type, it will be treated as applying to the pointer
2964target type; if such an attribute is applied to a function return type
2965that is not a pointer-to-function type, it will be treated as applying
2966to the function type.
2967
2968@node Function Prototypes
2969@section Prototypes and Old-Style Function Definitions
2970@cindex function prototype declarations
2971@cindex old-style function definitions
2972@cindex promotion of formal parameters
2973
2974GNU C extends ISO C to allow a function prototype to override a later
2975old-style non-prototype definition.  Consider the following example:
2976
2977@smallexample
2978/* @r{Use prototypes unless the compiler is old-fashioned.}  */
2979#ifdef __STDC__
2980#define P(x) x
2981#else
2982#define P(x) ()
2983#endif
2984
2985/* @r{Prototype function declaration.}  */
2986int isroot P((uid_t));
2987
2988/* @r{Old-style function definition.}  */
2989int
2990isroot (x)   /* ??? lossage here ??? */
2991     uid_t x;
2992@{
2993  return x == 0;
2994@}
2995@end smallexample
2996
2997Suppose the type @code{uid_t} happens to be @code{short}.  ISO C does
2998not allow this example, because subword arguments in old-style
2999non-prototype definitions are promoted.  Therefore in this example the
3000function definition's argument is really an @code{int}, which does not
3001match the prototype argument type of @code{short}.
3002
3003This restriction of ISO C makes it hard to write code that is portable
3004to traditional C compilers, because the programmer does not know
3005whether the @code{uid_t} type is @code{short}, @code{int}, or
3006@code{long}.  Therefore, in cases like these GNU C allows a prototype
3007to override a later old-style definition.  More precisely, in GNU C, a
3008function prototype argument type overrides the argument type specified
3009by a later old-style definition if the former type is the same as the
3010latter type before promotion.  Thus in GNU C the above example is
3011equivalent to the following:
3012
3013@smallexample
3014int isroot (uid_t);
3015
3016int
3017isroot (uid_t x)
3018@{
3019  return x == 0;
3020@}
3021@end smallexample
3022
3023@noindent
3024GNU C++ does not support old-style function definitions, so this
3025extension is irrelevant.
3026
3027@node C++ Comments
3028@section C++ Style Comments
3029@cindex //
3030@cindex C++ comments
3031@cindex comments, C++ style
3032
3033In GNU C, you may use C++ style comments, which start with @samp{//} and
3034continue until the end of the line.  Many other C implementations allow
3035such comments, and they are included in the 1999 C standard.  However,
3036C++ style comments are not recognized if you specify an @option{-std}
3037option specifying a version of ISO C before C99, or @option{-ansi}
3038(equivalent to @option{-std=c89}).
3039
3040@node Dollar Signs
3041@section Dollar Signs in Identifier Names
3042@cindex $
3043@cindex dollar signs in identifier names
3044@cindex identifier names, dollar signs in
3045
3046In GNU C, you may normally use dollar signs in identifier names.
3047This is because many traditional C implementations allow such identifiers.
3048However, dollar signs in identifiers are not supported on a few target
3049machines, typically because the target assembler does not allow them.
3050
3051@node Character Escapes
3052@section The Character @key{ESC} in Constants
3053
3054You can use the sequence @samp{\e} in a string or character constant to
3055stand for the ASCII character @key{ESC}.
3056
3057@node Alignment
3058@section Inquiring on Alignment of Types or Variables
3059@cindex alignment
3060@cindex type alignment
3061@cindex variable alignment
3062
3063The keyword @code{__alignof__} allows you to inquire about how an object
3064is aligned, or the minimum alignment usually required by a type.  Its
3065syntax is just like @code{sizeof}.
3066
3067For example, if the target machine requires a @code{double} value to be
3068aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8.
3069This is true on many RISC machines.  On more traditional machine
3070designs, @code{__alignof__ (double)} is 4 or even 2.
3071
3072Some machines never actually require alignment; they allow reference to any
3073data type even at an odd address.  For these machines, @code{__alignof__}
3074reports the @emph{recommended} alignment of a type.
3075
3076If the operand of @code{__alignof__} is an lvalue rather than a type,
3077its value is the required alignment for its type, taking into account
3078any minimum alignment specified with GCC's @code{__attribute__}
3079extension (@pxref{Variable Attributes}).  For example, after this
3080declaration:
3081
3082@smallexample
3083struct foo @{ int x; char y; @} foo1;
3084@end smallexample
3085
3086@noindent
3087the value of @code{__alignof__ (foo1.y)} is 1, even though its actual
3088alignment is probably 2 or 4, the same as @code{__alignof__ (int)}.
3089
3090It is an error to ask for the alignment of an incomplete type.
3091
3092@node Variable Attributes
3093@section Specifying Attributes of Variables
3094@cindex attribute of variables
3095@cindex variable attributes
3096
3097The keyword @code{__attribute__} allows you to specify special
3098attributes of variables or structure fields.  This keyword is followed
3099by an attribute specification inside double parentheses.  Some
3100attributes are currently defined generically for variables.
3101Other attributes are defined for variables on particular target
3102systems.  Other attributes are available for functions
3103(@pxref{Function Attributes}) and for types (@pxref{Type Attributes}).
3104Other front ends might define more attributes
3105(@pxref{C++ Extensions,,Extensions to the C++ Language}).
3106
3107You may also specify attributes with @samp{__} preceding and following
3108each keyword.  This allows you to use them in header files without
3109being concerned about a possible macro of the same name.  For example,
3110you may use @code{__aligned__} instead of @code{aligned}.
3111
3112@xref{Attribute Syntax}, for details of the exact syntax for using
3113attributes.
3114
3115@table @code
3116@cindex @code{aligned} attribute
3117@item aligned (@var{alignment})
3118This attribute specifies a minimum alignment for the variable or
3119structure field, measured in bytes.  For example, the declaration:
3120
3121@smallexample
3122int x __attribute__ ((aligned (16))) = 0;
3123@end smallexample
3124
3125@noindent
3126causes the compiler to allocate the global variable @code{x} on a
312716-byte boundary.  On a 68040, this could be used in conjunction with
3128an @code{asm} expression to access the @code{move16} instruction which
3129requires 16-byte aligned operands.
3130
3131You can also specify the alignment of structure fields.  For example, to
3132create a double-word aligned @code{int} pair, you could write:
3133
3134@smallexample
3135struct foo @{ int x[2] __attribute__ ((aligned (8))); @};
3136@end smallexample
3137
3138@noindent
3139This is an alternative to creating a union with a @code{double} member
3140that forces the union to be double-word aligned.
3141
3142As in the preceding examples, you can explicitly specify the alignment
3143(in bytes) that you wish the compiler to use for a given variable or
3144structure field.  Alternatively, you can leave out the alignment factor
3145and just ask the compiler to align a variable or field to the maximum
3146useful alignment for the target machine you are compiling for.  For
3147example, you could write:
3148
3149@smallexample
3150short array[3] __attribute__ ((aligned));
3151@end smallexample
3152
3153Whenever you leave out the alignment factor in an @code{aligned} attribute
3154specification, the compiler automatically sets the alignment for the declared
3155variable or field to the largest alignment which is ever used for any data
3156type on the target machine you are compiling for.  Doing this can often make
3157copy operations more efficient, because the compiler can use whatever
3158instructions copy the biggest chunks of memory when performing copies to
3159or from the variables or fields that you have aligned this way.
3160
3161The @code{aligned} attribute can only increase the alignment; but you
3162can decrease it by specifying @code{packed} as well.  See below.
3163
3164Note that the effectiveness of @code{aligned} attributes may be limited
3165by inherent limitations in your linker.  On many systems, the linker is
3166only able to arrange for variables to be aligned up to a certain maximum
3167alignment.  (For some linkers, the maximum supported alignment may
3168be very very small.)  If your linker is only able to align variables
3169up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3170in an @code{__attribute__} will still only provide you with 8 byte
3171alignment.  See your linker documentation for further information.
3172
3173@item cleanup (@var{cleanup_function})
3174@cindex @code{cleanup} attribute
3175The @code{cleanup} attribute runs a function when the variable goes
3176out of scope.  This attribute can only be applied to auto function
3177scope variables; it may not be applied to parameters or variables
3178with static storage duration.  The function must take one parameter,
3179a pointer to a type compatible with the variable.  The return value
3180of the function (if any) is ignored.
3181
3182If @option{-fexceptions} is enabled, then @var{cleanup_function}
3183will be run during the stack unwinding that happens during the
3184processing of the exception.  Note that the @code{cleanup} attribute
3185does not allow the exception to be caught, only to perform an action.
3186It is undefined what happens if @var{cleanup_function} does not
3187return normally.
3188
3189@item common
3190@itemx nocommon
3191@cindex @code{common} attribute
3192@cindex @code{nocommon} attribute
3193@opindex fcommon
3194@opindex fno-common
3195The @code{common} attribute requests GCC to place a variable in
3196``common'' storage.  The @code{nocommon} attribute requests the
3197opposite -- to allocate space for it directly.
3198
3199These attributes override the default chosen by the
3200@option{-fno-common} and @option{-fcommon} flags respectively.
3201
3202@item deprecated
3203@cindex @code{deprecated} attribute
3204The @code{deprecated} attribute results in a warning if the variable
3205is used anywhere in the source file.  This is useful when identifying
3206variables that are expected to be removed in a future version of a
3207program.  The warning also includes the location of the declaration
3208of the deprecated variable, to enable users to easily find further
3209information about why the variable is deprecated, or what they should
3210do instead.  Note that the warning only occurs for uses:
3211
3212@smallexample
3213extern int old_var __attribute__ ((deprecated));
3214extern int old_var;
3215int new_fn () @{ return old_var; @}
3216@end smallexample
3217
3218results in a warning on line 3 but not line 2.
3219
3220The @code{deprecated} attribute can also be used for functions and
3221types (@pxref{Function Attributes}, @pxref{Type Attributes}.)
3222
3223@item mode (@var{mode})
3224@cindex @code{mode} attribute
3225This attribute specifies the data type for the declaration---whichever
3226type corresponds to the mode @var{mode}.  This in effect lets you
3227request an integer or floating point type according to its width.
3228
3229You may also specify a mode of @samp{byte} or @samp{__byte__} to
3230indicate the mode corresponding to a one-byte integer, @samp{word} or
3231@samp{__word__} for the mode of a one-word integer, and @samp{pointer}
3232or @samp{__pointer__} for the mode used to represent pointers.
3233
3234@item packed
3235@cindex @code{packed} attribute
3236The @code{packed} attribute specifies that a variable or structure field
3237should have the smallest possible alignment---one byte for a variable,
3238and one bit for a field, unless you specify a larger value with the
3239@code{aligned} attribute.
3240
3241Here is a structure in which the field @code{x} is packed, so that it
3242immediately follows @code{a}:
3243
3244@smallexample
3245struct foo
3246@{
3247  char a;
3248  int x[2] __attribute__ ((packed));
3249@};
3250@end smallexample
3251
3252@item section ("@var{section-name}")
3253@cindex @code{section} variable attribute
3254Normally, the compiler places the objects it generates in sections like
3255@code{data} and @code{bss}.  Sometimes, however, you need additional sections,
3256or you need certain particular variables to appear in special sections,
3257for example to map to special hardware.  The @code{section}
3258attribute specifies that a variable (or function) lives in a particular
3259section.  For example, this small program uses several specific section names:
3260
3261@smallexample
3262struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @};
3263struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @};
3264char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @};
3265int init_data __attribute__ ((section ("INITDATA"))) = 0;
3266
3267main()
3268@{
3269  /* Initialize stack pointer */
3270  init_sp (stack + sizeof (stack));
3271
3272  /* Initialize initialized data */
3273  memcpy (&init_data, &data, &edata - &data);
3274
3275  /* Turn on the serial ports */
3276  init_duart (&a);
3277  init_duart (&b);
3278@}
3279@end smallexample
3280
3281@noindent
3282Use the @code{section} attribute with an @emph{initialized} definition
3283of a @emph{global} variable, as shown in the example.  GCC issues
3284a warning and otherwise ignores the @code{section} attribute in
3285uninitialized variable declarations.
3286
3287You may only use the @code{section} attribute with a fully initialized
3288global definition because of the way linkers work.  The linker requires
3289each object be defined once, with the exception that uninitialized
3290variables tentatively go in the @code{common} (or @code{bss}) section
3291and can be multiply ``defined''.  You can force a variable to be
3292initialized with the @option{-fno-common} flag or the @code{nocommon}
3293attribute.
3294
3295Some file formats do not support arbitrary sections so the @code{section}
3296attribute is not available on all platforms.
3297If you need to map the entire contents of a module to a particular
3298section, consider using the facilities of the linker instead.
3299
3300@item shared
3301@cindex @code{shared} variable attribute
3302On Microsoft Windows, in addition to putting variable definitions in a named
3303section, the section can also be shared among all running copies of an
3304executable or DLL@.  For example, this small program defines shared data
3305by putting it in a named section @code{shared} and marking the section
3306shareable:
3307
3308@smallexample
3309int foo __attribute__((section ("shared"), shared)) = 0;
3310
3311int
3312main()
3313@{
3314  /* Read and write foo.  All running
3315     copies see the same value.  */
3316  return 0;
3317@}
3318@end smallexample
3319
3320@noindent
3321You may only use the @code{shared} attribute along with @code{section}
3322attribute with a fully initialized global definition because of the way
3323linkers work.  See @code{section} attribute for more information.
3324
3325The @code{shared} attribute is only available on Microsoft Windows@.
3326
3327@item tls_model ("@var{tls_model}")
3328@cindex @code{tls_model} attribute
3329The @code{tls_model} attribute sets thread-local storage model
3330(@pxref{Thread-Local}) of a particular @code{__thread} variable,
3331overriding @code{-ftls-model=} command line switch on a per-variable
3332basis.
3333The @var{tls_model} argument should be one of @code{global-dynamic},
3334@code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
3335
3336Not all targets support this attribute.
3337
3338@item transparent_union
3339This attribute, attached to a function parameter which is a union, means
3340that the corresponding argument may have the type of any union member,
3341but the argument is passed as if its type were that of the first union
3342member.  For more details see @xref{Type Attributes}.  You can also use
3343this attribute on a @code{typedef} for a union data type; then it
3344applies to all function parameters with that type.
3345
3346@item unused
3347This attribute, attached to a variable, means that the variable is meant
3348to be possibly unused.  GCC will not produce a warning for this
3349variable.
3350
3351@item vector_size (@var{bytes})
3352This attribute specifies the vector size for the variable, measured in
3353bytes.  For example, the declaration:
3354
3355@smallexample
3356int foo __attribute__ ((vector_size (16)));
3357@end smallexample
3358
3359@noindent
3360causes the compiler to set the mode for @code{foo}, to be 16 bytes,
3361divided into @code{int} sized units.  Assuming a 32-bit int (a vector of
33624 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@.
3363
3364This attribute is only applicable to integral and float scalars,
3365although arrays, pointers, and function return values are allowed in
3366conjunction with this construct.
3367
3368Aggregates with this attribute are invalid, even if they are of the same
3369size as a corresponding scalar.  For example, the declaration:
3370
3371@smallexample
3372struct S @{ int a; @};
3373struct S  __attribute__ ((vector_size (16))) foo;
3374@end smallexample
3375
3376@noindent
3377is invalid even if the size of the structure is the same as the size of
3378the @code{int}.
3379
3380@item weak
3381The @code{weak} attribute is described in @xref{Function Attributes}.
3382
3383@item dllimport
3384The @code{dllimport} attribute is described in @xref{Function Attributes}.
3385
3386@item dlexport
3387The @code{dllexport} attribute is described in @xref{Function Attributes}.
3388
3389@end table
3390
3391@subsection M32R/D Variable Attributes
3392
3393One attribute is currently defined for the M32R/D.
3394
3395@table @code
3396@item model (@var{model-name})
3397@cindex variable addressability on the M32R/D
3398Use this attribute on the M32R/D to set the addressability of an object.
3399The identifier @var{model-name} is one of @code{small}, @code{medium},
3400or @code{large}, representing each of the code models.
3401
3402Small model objects live in the lower 16MB of memory (so that their
3403addresses can be loaded with the @code{ld24} instruction).
3404
3405Medium and large model objects may live anywhere in the 32-bit address space
3406(the compiler will generate @code{seth/add3} instructions to load their
3407addresses).
3408@end table
3409
3410@subsection i386 Variable Attributes
3411
3412Two attributes are currently defined for i386 configurations:
3413@code{ms_struct} and @code{gcc_struct}
3414
3415@table @code
3416@item ms_struct
3417@itemx gcc_struct
3418@cindex @code{ms_struct} attribute
3419@cindex @code{gcc_struct} attribute
3420
3421If @code{packed} is used on a structure, or if bit-fields are used
3422it may be that the Microsoft ABI packs them differently
3423than GCC would normally pack them.  Particularly when moving packed
3424data between functions compiled with GCC and the native Microsoft compiler
3425(either via function call or as data in a file), it may be necessary to access
3426either format.
3427
3428Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3429compilers to match the native Microsoft compiler.
3430@end table
3431
3432@node Type Attributes
3433@section Specifying Attributes of Types
3434@cindex attribute of types
3435@cindex type attributes
3436
3437The keyword @code{__attribute__} allows you to specify special
3438attributes of @code{struct} and @code{union} types when you define such
3439types.  This keyword is followed by an attribute specification inside
3440double parentheses.  Six attributes are currently defined for types:
3441@code{aligned}, @code{packed}, @code{transparent_union}, @code{unused},
3442@code{deprecated} and @code{may_alias}.  Other attributes are defined for
3443functions (@pxref{Function Attributes}) and for variables
3444(@pxref{Variable Attributes}).
3445
3446You may also specify any one of these attributes with @samp{__}
3447preceding and following its keyword.  This allows you to use these
3448attributes in header files without being concerned about a possible
3449macro of the same name.  For example, you may use @code{__aligned__}
3450instead of @code{aligned}.
3451
3452You may specify the @code{aligned} and @code{transparent_union}
3453attributes either in a @code{typedef} declaration or just past the
3454closing curly brace of a complete enum, struct or union type
3455@emph{definition} and the @code{packed} attribute only past the closing
3456brace of a definition.
3457
3458You may also specify attributes between the enum, struct or union
3459tag and the name of the type rather than after the closing brace.
3460
3461@xref{Attribute Syntax}, for details of the exact syntax for using
3462attributes.
3463
3464@table @code
3465@cindex @code{aligned} attribute
3466@item aligned (@var{alignment})
3467This attribute specifies a minimum alignment (in bytes) for variables
3468of the specified type.  For example, the declarations:
3469
3470@smallexample
3471struct S @{ short f[3]; @} __attribute__ ((aligned (8)));
3472typedef int more_aligned_int __attribute__ ((aligned (8)));
3473@end smallexample
3474
3475@noindent
3476force the compiler to insure (as far as it can) that each variable whose
3477type is @code{struct S} or @code{more_aligned_int} will be allocated and
3478aligned @emph{at least} on a 8-byte boundary.  On a SPARC, having all
3479variables of type @code{struct S} aligned to 8-byte boundaries allows
3480the compiler to use the @code{ldd} and @code{std} (doubleword load and
3481store) instructions when copying one variable of type @code{struct S} to
3482another, thus improving run-time efficiency.
3483
3484Note that the alignment of any given @code{struct} or @code{union} type
3485is required by the ISO C standard to be at least a perfect multiple of
3486the lowest common multiple of the alignments of all of the members of
3487the @code{struct} or @code{union} in question.  This means that you @emph{can}
3488effectively adjust the alignment of a @code{struct} or @code{union}
3489type by attaching an @code{aligned} attribute to any one of the members
3490of such a type, but the notation illustrated in the example above is a
3491more obvious, intuitive, and readable way to request the compiler to
3492adjust the alignment of an entire @code{struct} or @code{union} type.
3493
3494As in the preceding example, you can explicitly specify the alignment
3495(in bytes) that you wish the compiler to use for a given @code{struct}
3496or @code{union} type.  Alternatively, you can leave out the alignment factor
3497and just ask the compiler to align a type to the maximum
3498useful alignment for the target machine you are compiling for.  For
3499example, you could write:
3500
3501@smallexample
3502struct S @{ short f[3]; @} __attribute__ ((aligned));
3503@end smallexample
3504
3505Whenever you leave out the alignment factor in an @code{aligned}
3506attribute specification, the compiler automatically sets the alignment
3507for the type to the largest alignment which is ever used for any data
3508type on the target machine you are compiling for.  Doing this can often
3509make copy operations more efficient, because the compiler can use
3510whatever instructions copy the biggest chunks of memory when performing
3511copies to or from the variables which have types that you have aligned
3512this way.
3513
3514In the example above, if the size of each @code{short} is 2 bytes, then
3515the size of the entire @code{struct S} type is 6 bytes.  The smallest
3516power of two which is greater than or equal to that is 8, so the
3517compiler sets the alignment for the entire @code{struct S} type to 8
3518bytes.
3519
3520Note that although you can ask the compiler to select a time-efficient
3521alignment for a given type and then declare only individual stand-alone
3522objects of that type, the compiler's ability to select a time-efficient
3523alignment is primarily useful only when you plan to create arrays of
3524variables having the relevant (efficiently aligned) type.  If you
3525declare or use arrays of variables of an efficiently-aligned type, then
3526it is likely that your program will also be doing pointer arithmetic (or
3527subscripting, which amounts to the same thing) on pointers to the
3528relevant type, and the code that the compiler generates for these
3529pointer arithmetic operations will often be more efficient for
3530efficiently-aligned types than for other types.
3531
3532The @code{aligned} attribute can only increase the alignment; but you
3533can decrease it by specifying @code{packed} as well.  See below.
3534
3535Note that the effectiveness of @code{aligned} attributes may be limited
3536by inherent limitations in your linker.  On many systems, the linker is
3537only able to arrange for variables to be aligned up to a certain maximum
3538alignment.  (For some linkers, the maximum supported alignment may
3539be very very small.)  If your linker is only able to align variables
3540up to a maximum of 8 byte alignment, then specifying @code{aligned(16)}
3541in an @code{__attribute__} will still only provide you with 8 byte
3542alignment.  See your linker documentation for further information.
3543
3544@item packed
3545This attribute, attached to @code{struct} or @code{union} type
3546definition, specifies that each member of the structure or union is
3547placed to minimize the memory required. When attached to an @code{enum}
3548definition, it indicates that the smallest integral type should be used.
3549
3550@opindex fshort-enums
3551Specifying this attribute for @code{struct} and @code{union} types is
3552equivalent to specifying the @code{packed} attribute on each of the
3553structure or union members.  Specifying the @option{-fshort-enums}
3554flag on the line is equivalent to specifying the @code{packed}
3555attribute on all @code{enum} definitions.
3556
3557In the following example @code{struct my_packed_struct}'s members are
3558packed closely together, but the internal layout of its @code{s} member
3559is not packed -- to do that, @code{struct my_unpacked_struct} would need to
3560be packed too.
3561
3562@smallexample
3563struct my_unpacked_struct
3564 @{
3565    char c;
3566    int i;
3567 @};
3568
3569struct my_packed_struct __attribute__ ((__packed__))
3570  @{
3571     char c;
3572     int  i;
3573     struct my_unpacked_struct s;
3574  @};
3575@end smallexample
3576
3577You may only specify this attribute on the definition of a @code{enum},
3578@code{struct} or @code{union}, not on a @code{typedef} which does not
3579also define the enumerated type, structure or union.
3580
3581@item transparent_union
3582This attribute, attached to a @code{union} type definition, indicates
3583that any function parameter having that union type causes calls to that
3584function to be treated in a special way.
3585
3586First, the argument corresponding to a transparent union type can be of
3587any type in the union; no cast is required.  Also, if the union contains
3588a pointer type, the corresponding argument can be a null pointer
3589constant or a void pointer expression; and if the union contains a void
3590pointer type, the corresponding argument can be any pointer expression.
3591If the union member type is a pointer, qualifiers like @code{const} on
3592the referenced type must be respected, just as with normal pointer
3593conversions.
3594
3595Second, the argument is passed to the function using the calling
3596conventions of the first member of the transparent union, not the calling
3597conventions of the union itself.  All members of the union must have the
3598same machine representation; this is necessary for this argument passing
3599to work properly.
3600
3601Transparent unions are designed for library functions that have multiple
3602interfaces for compatibility reasons.  For example, suppose the
3603@code{wait} function must accept either a value of type @code{int *} to
3604comply with Posix, or a value of type @code{union wait *} to comply with
3605the 4.1BSD interface.  If @code{wait}'s parameter were @code{void *},
3606@code{wait} would accept both kinds of arguments, but it would also
3607accept any other pointer type and this would make argument type checking
3608less useful.  Instead, @code{<sys/wait.h>} might define the interface
3609as follows:
3610
3611@smallexample
3612typedef union
3613  @{
3614    int *__ip;
3615    union wait *__up;
3616  @} wait_status_ptr_t __attribute__ ((__transparent_union__));
3617
3618pid_t wait (wait_status_ptr_t);
3619@end smallexample
3620
3621This interface allows either @code{int *} or @code{union wait *}
3622arguments to be passed, using the @code{int *} calling convention.
3623The program can call @code{wait} with arguments of either type:
3624
3625@smallexample
3626int w1 () @{ int w; return wait (&w); @}
3627int w2 () @{ union wait w; return wait (&w); @}
3628@end smallexample
3629
3630With this interface, @code{wait}'s implementation might look like this:
3631
3632@smallexample
3633pid_t wait (wait_status_ptr_t p)
3634@{
3635  return waitpid (-1, p.__ip, 0);
3636@}
3637@end smallexample
3638
3639@item unused
3640When attached to a type (including a @code{union} or a @code{struct}),
3641this attribute means that variables of that type are meant to appear
3642possibly unused.  GCC will not produce a warning for any variables of
3643that type, even if the variable appears to do nothing.  This is often
3644the case with lock or thread classes, which are usually defined and then
3645not referenced, but contain constructors and destructors that have
3646nontrivial bookkeeping functions.
3647
3648@item deprecated
3649The @code{deprecated} attribute results in a warning if the type
3650is used anywhere in the source file.  This is useful when identifying
3651types that are expected to be removed in a future version of a program.
3652If possible, the warning also includes the location of the declaration
3653of the deprecated type, to enable users to easily find further
3654information about why the type is deprecated, or what they should do
3655instead.  Note that the warnings only occur for uses and then only
3656if the type is being applied to an identifier that itself is not being
3657declared as deprecated.
3658
3659@smallexample
3660typedef int T1 __attribute__ ((deprecated));
3661T1 x;
3662typedef T1 T2;
3663T2 y;
3664typedef T1 T3 __attribute__ ((deprecated));
3665T3 z __attribute__ ((deprecated));
3666@end smallexample
3667
3668results in a warning on line 2 and 3 but not lines 4, 5, or 6.  No
3669warning is issued for line 4 because T2 is not explicitly
3670deprecated.  Line 5 has no warning because T3 is explicitly
3671deprecated.  Similarly for line 6.
3672
3673The @code{deprecated} attribute can also be used for functions and
3674variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.)
3675
3676@item may_alias
3677Accesses to objects with types with this attribute are not subjected to
3678type-based alias analysis, but are instead assumed to be able to alias
3679any other type of objects, just like the @code{char} type.  See
3680@option{-fstrict-aliasing} for more information on aliasing issues.
3681
3682Example of use:
3683
3684@smallexample
3685typedef short __attribute__((__may_alias__)) short_a;
3686
3687int
3688main (void)
3689@{
3690  int a = 0x12345678;
3691  short_a *b = (short_a *) &a;
3692
3693  b[1] = 0;
3694
3695  if (a == 0x12345678)
3696    abort();
3697
3698  exit(0);
3699@}
3700@end smallexample
3701
3702If you replaced @code{short_a} with @code{short} in the variable
3703declaration, the above program would abort when compiled with
3704@option{-fstrict-aliasing}, which is on by default at @option{-O2} or
3705above in recent GCC versions.
3706
3707@subsection i386 Type Attributes
3708
3709Two attributes are currently defined for i386 configurations:
3710@code{ms_struct} and @code{gcc_struct}
3711
3712@item ms_struct
3713@itemx gcc_struct
3714@cindex @code{ms_struct}
3715@cindex @code{gcc_struct}
3716
3717If @code{packed} is used on a structure, or if bit-fields are used
3718it may be that the Microsoft ABI packs them differently
3719than GCC would normally pack them.  Particularly when moving packed
3720data between functions compiled with GCC and the native Microsoft compiler
3721(either via function call or as data in a file), it may be necessary to access
3722either format.
3723
3724Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86
3725compilers to match the native Microsoft compiler.
3726@end table
3727
3728To specify multiple attributes, separate them by commas within the
3729double parentheses: for example, @samp{__attribute__ ((aligned (16),
3730packed))}.
3731
3732@node Inline
3733@section An Inline Function is As Fast As a Macro
3734@cindex inline functions
3735@cindex integrating function code
3736@cindex open coding
3737@cindex macros, inline alternative
3738
3739By declaring a function @code{inline}, you can direct GCC to
3740integrate that function's code into the code for its callers.  This
3741makes execution faster by eliminating the function-call overhead; in
3742addition, if any of the actual argument values are constant, their known
3743values may permit simplifications at compile time so that not all of the
3744inline function's code needs to be included.  The effect on code size is
3745less predictable; object code may be larger or smaller with function
3746inlining, depending on the particular case.  Inlining of functions is an
3747optimization and it really ``works'' only in optimizing compilation.  If
3748you don't use @option{-O}, no function is really inline.
3749
3750Inline functions are included in the ISO C99 standard, but there are
3751currently substantial differences between what GCC implements and what
3752the ISO C99 standard requires.
3753
3754To declare a function inline, use the @code{inline} keyword in its
3755declaration, like this:
3756
3757@smallexample
3758inline int
3759inc (int *a)
3760@{
3761  (*a)++;
3762@}
3763@end smallexample
3764
3765(If you are writing a header file to be included in ISO C programs, write
3766@code{__inline__} instead of @code{inline}.  @xref{Alternate Keywords}.)
3767You can also make all ``simple enough'' functions inline with the option
3768@option{-finline-functions}.
3769
3770@opindex Winline
3771Note that certain usages in a function definition can make it unsuitable
3772for inline substitution.  Among these usages are: use of varargs, use of
3773alloca, use of variable sized data types (@pxref{Variable Length}),
3774use of computed goto (@pxref{Labels as Values}), use of nonlocal goto,
3775and nested functions (@pxref{Nested Functions}).  Using @option{-Winline}
3776will warn when a function marked @code{inline} could not be substituted,
3777and will give the reason for the failure.
3778
3779Note that in C and Objective-C, unlike C++, the @code{inline} keyword
3780does not affect the linkage of the function.
3781
3782@cindex automatic @code{inline} for C++ member fns
3783@cindex @code{inline} automatic for C++ member fns
3784@cindex member fns, automatically @code{inline}
3785@cindex C++ member fns, automatically @code{inline}
3786@opindex fno-default-inline
3787GCC automatically inlines member functions defined within the class
3788body of C++ programs even if they are not explicitly declared
3789@code{inline}.  (You can override this with @option{-fno-default-inline};
3790@pxref{C++ Dialect Options,,Options Controlling C++ Dialect}.)
3791
3792@cindex inline functions, omission of
3793@opindex fkeep-inline-functions
3794When a function is both inline and @code{static}, if all calls to the
3795function are integrated into the caller, and the function's address is
3796never used, then the function's own assembler code is never referenced.
3797In this case, GCC does not actually output assembler code for the
3798function, unless you specify the option @option{-fkeep-inline-functions}.
3799Some calls cannot be integrated for various reasons (in particular,
3800calls that precede the function's definition cannot be integrated, and
3801neither can recursive calls within the definition).  If there is a
3802nonintegrated call, then the function is compiled to assembler code as
3803usual.  The function must also be compiled as usual if the program
3804refers to its address, because that can't be inlined.
3805
3806@cindex non-static inline function
3807When an inline function is not @code{static}, then the compiler must assume
3808that there may be calls from other source files; since a global symbol can
3809be defined only once in any program, the function must not be defined in
3810the other source files, so the calls therein cannot be integrated.
3811Therefore, a non-@code{static} inline function is always compiled on its
3812own in the usual fashion.
3813
3814If you specify both @code{inline} and @code{extern} in the function
3815definition, then the definition is used only for inlining.  In no case
3816is the function compiled on its own, not even if you refer to its
3817address explicitly.  Such an address becomes an external reference, as
3818if you had only declared the function, and had not defined it.
3819
3820This combination of @code{inline} and @code{extern} has almost the
3821effect of a macro.  The way to use it is to put a function definition in
3822a header file with these keywords, and put another copy of the
3823definition (lacking @code{inline} and @code{extern}) in a library file.
3824The definition in the header file will cause most calls to the function
3825to be inlined.  If any uses of the function remain, they will refer to
3826the single copy in the library.
3827
3828Since GCC eventually will implement ISO C99 semantics for
3829inline functions, it is best to use @code{static inline} only
3830to guarantee compatibility.  (The
3831existing semantics will remain available when @option{-std=gnu89} is
3832specified, but eventually the default will be @option{-std=gnu99} and
3833that will implement the C99 semantics, though it does not do so yet.)
3834
3835GCC does not inline any functions when not optimizing unless you specify
3836the @samp{always_inline} attribute for the function, like this:
3837
3838@smallexample
3839/* Prototype.  */
3840inline void foo (const char) __attribute__((always_inline));
3841@end smallexample
3842
3843@node Extended Asm
3844@section Assembler Instructions with C Expression Operands
3845@cindex extended @code{asm}
3846@cindex @code{asm} expressions
3847@cindex assembler instructions
3848@cindex registers
3849
3850In an assembler instruction using @code{asm}, you can specify the
3851operands of the instruction using C expressions.  This means you need not
3852guess which registers or memory locations will contain the data you want
3853to use.
3854
3855You must specify an assembler instruction template much like what
3856appears in a machine description, plus an operand constraint string for
3857each operand.
3858
3859For example, here is how to use the 68881's @code{fsinx} instruction:
3860
3861@smallexample
3862asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
3863@end smallexample
3864
3865@noindent
3866Here @code{angle} is the C expression for the input operand while
3867@code{result} is that of the output operand.  Each has @samp{"f"} as its
3868operand constraint, saying that a floating point register is required.
3869The @samp{=} in @samp{=f} indicates that the operand is an output; all
3870output operands' constraints must use @samp{=}.  The constraints use the
3871same language used in the machine description (@pxref{Constraints}).
3872
3873Each operand is described by an operand-constraint string followed by
3874the C expression in parentheses.  A colon separates the assembler
3875template from the first output operand and another separates the last
3876output operand from the first input, if any.  Commas separate the
3877operands within each group.  The total number of operands is currently
3878limited to 30; this limitation may be lifted in some future version of
3879GCC.
3880
3881If there are no output operands but there are input operands, you must
3882place two consecutive colons surrounding the place where the output
3883operands would go.
3884
3885As of GCC version 3.1, it is also possible to specify input and output
3886operands using symbolic names which can be referenced within the
3887assembler code.  These names are specified inside square brackets
3888preceding the constraint string, and can be referenced inside the
3889assembler code using @code{%[@var{name}]} instead of a percentage sign
3890followed by the operand number.  Using named operands the above example
3891could look like:
3892
3893@smallexample
3894asm ("fsinx %[angle],%[output]"
3895     : [output] "=f" (result)
3896     : [angle] "f" (angle));
3897@end smallexample
3898
3899@noindent
3900Note that the symbolic operand names have no relation whatsoever to
3901other C identifiers.  You may use any name you like, even those of
3902existing C symbols, but you must ensure that no two operands within the same
3903assembler construct use the same symbolic name.
3904
3905Output operand expressions must be lvalues; the compiler can check this.
3906The input operands need not be lvalues.  The compiler cannot check
3907whether the operands have data types that are reasonable for the
3908instruction being executed.  It does not parse the assembler instruction
3909template and does not know what it means or even whether it is valid
3910assembler input.  The extended @code{asm} feature is most often used for
3911machine instructions the compiler itself does not know exist.  If
3912the output expression cannot be directly addressed (for example, it is a
3913bit-field), your constraint must allow a register.  In that case, GCC
3914will use the register as the output of the @code{asm}, and then store
3915that register into the output.
3916
3917The ordinary output operands must be write-only; GCC will assume that
3918the values in these operands before the instruction are dead and need
3919not be generated.  Extended asm supports input-output or read-write
3920operands.  Use the constraint character @samp{+} to indicate such an
3921operand and list it with the output operands.  You should only use
3922read-write operands when the constraints for the operand (or the
3923operand in which only some of the bits are to be changed) allow a
3924register.
3925
3926You may, as an alternative, logically split its function into two
3927separate operands, one input operand and one write-only output
3928operand.  The connection between them is expressed by constraints
3929which say they need to be in the same location when the instruction
3930executes.  You can use the same C expression for both operands, or
3931different expressions.  For example, here we write the (fictitious)
3932@samp{combine} instruction with @code{bar} as its read-only source
3933operand and @code{foo} as its read-write destination:
3934
3935@smallexample
3936asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
3937@end smallexample
3938
3939@noindent
3940The constraint @samp{"0"} for operand 1 says that it must occupy the
3941same location as operand 0.  A number in constraint is allowed only in
3942an input operand and it must refer to an output operand.
3943
3944Only a number in the constraint can guarantee that one operand will be in
3945the same place as another.  The mere fact that @code{foo} is the value
3946of both operands is not enough to guarantee that they will be in the
3947same place in the generated assembler code.  The following would not
3948work reliably:
3949
3950@smallexample
3951asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
3952@end smallexample
3953
3954Various optimizations or reloading could cause operands 0 and 1 to be in
3955different registers; GCC knows no reason not to do so.  For example, the
3956compiler might find a copy of the value of @code{foo} in one register and
3957use it for operand 1, but generate the output operand 0 in a different
3958register (copying it afterward to @code{foo}'s own address).  Of course,
3959since the register for operand 1 is not even mentioned in the assembler
3960code, the result will not work, but GCC can't tell that.
3961
3962As of GCC version 3.1, one may write @code{[@var{name}]} instead of
3963the operand number for a matching constraint.  For example:
3964
3965@smallexample
3966asm ("cmoveq %1,%2,%[result]"
3967     : [result] "=r"(result)
3968     : "r" (test), "r"(new), "[result]"(old));
3969@end smallexample
3970
3971Some instructions clobber specific hard registers.  To describe this,
3972write a third colon after the input operands, followed by the names of
3973the clobbered hard registers (given as strings).  Here is a realistic
3974example for the VAX:
3975
3976@smallexample
3977asm volatile ("movc3 %0,%1,%2"
3978              : /* no outputs */
3979              : "g" (from), "g" (to), "g" (count)
3980              : "r0", "r1", "r2", "r3", "r4", "r5");
3981@end smallexample
3982
3983You may not write a clobber description in a way that overlaps with an
3984input or output operand.  For example, you may not have an operand
3985describing a register class with one member if you mention that register
3986in the clobber list.  Variables declared to live in specific registers
3987(@pxref{Explicit Reg Vars}), and used as asm input or output operands must
3988have no part mentioned in the clobber description.
3989There is no way for you to specify that an input
3990operand is modified without also specifying it as an output
3991operand.  Note that if all the output operands you specify are for this
3992purpose (and hence unused), you will then also need to specify
3993@code{volatile} for the @code{asm} construct, as described below, to
3994prevent GCC from deleting the @code{asm} statement as unused.
3995
3996If you refer to a particular hardware register from the assembler code,
3997you will probably have to list the register after the third colon to
3998tell the compiler the register's value is modified.  In some assemblers,
3999the register names begin with @samp{%}; to produce one @samp{%} in the
4000assembler code, you must write @samp{%%} in the input.
4001
4002If your assembler instruction can alter the condition code register, add
4003@samp{cc} to the list of clobbered registers.  GCC on some machines
4004represents the condition codes as a specific hardware register;
4005@samp{cc} serves to name this register.  On other machines, the
4006condition code is handled differently, and specifying @samp{cc} has no
4007effect.  But it is valid no matter what the machine.
4008
4009If your assembler instructions access memory in an unpredictable
4010fashion, add @samp{memory} to the list of clobbered registers.  This
4011will cause GCC to not keep memory values cached in registers across the
4012assembler instruction and not optimize stores or loads to that memory.
4013You will also want to add the @code{volatile} keyword if the memory
4014affected is not listed in the inputs or outputs of the @code{asm}, as
4015the @samp{memory} clobber does not count as a side-effect of the
4016@code{asm}.  If you know how large the accessed memory is, you can add
4017it as input or output but if this is not known, you should add
4018@samp{memory}.  As an example, if you access ten bytes of a string, you
4019can use a memory input like:
4020
4021@example
4022@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
4023@end example
4024
4025Note that in the following example the memory input is necessary,
4026otherwise GCC might optimize the store to @code{x} away:
4027@example
4028int foo ()
4029@{
4030  int x = 42;
4031  int *y = &x;
4032  int result;
4033  asm ("magic stuff accessing an 'int' pointed to by '%1'"
4034        "=&d" (r) : "a" (y), "m" (*y));
4035  return result;
4036@}
4037@end example
4038
4039You can put multiple assembler instructions together in a single
4040@code{asm} template, separated by the characters normally used in assembly
4041code for the system.  A combination that works in most places is a newline
4042to break the line, plus a tab character to move to the instruction field
4043(written as @samp{\n\t}).  Sometimes semicolons can be used, if the
4044assembler allows semicolons as a line-breaking character.  Note that some
4045assembler dialects use semicolons to start a comment.
4046The input operands are guaranteed not to use any of the clobbered
4047registers, and neither will the output operands' addresses, so you can
4048read and write the clobbered registers as many times as you like.  Here
4049is an example of multiple instructions in a template; it assumes the
4050subroutine @code{_foo} accepts arguments in registers 9 and 10:
4051
4052@smallexample
4053asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
4054     : /* no outputs */
4055     : "g" (from), "g" (to)
4056     : "r9", "r10");
4057@end smallexample
4058
4059Unless an output operand has the @samp{&} constraint modifier, GCC
4060may allocate it in the same register as an unrelated input operand, on
4061the assumption the inputs are consumed before the outputs are produced.
4062This assumption may be false if the assembler code actually consists of
4063more than one instruction.  In such a case, use @samp{&} for each output
4064operand that may not overlap an input.  @xref{Modifiers}.
4065
4066If you want to test the condition code produced by an assembler
4067instruction, you must include a branch and a label in the @code{asm}
4068construct, as follows:
4069
4070@smallexample
4071asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
4072     : "g" (result)
4073     : "g" (input));
4074@end smallexample
4075
4076@noindent
4077This assumes your assembler supports local labels, as the GNU assembler
4078and most Unix assemblers do.
4079
4080Speaking of labels, jumps from one @code{asm} to another are not
4081supported.  The compiler's optimizers do not know about these jumps, and
4082therefore they cannot take account of them when deciding how to
4083optimize.
4084
4085@cindex macros containing @code{asm}
4086Usually the most convenient way to use these @code{asm} instructions is to
4087encapsulate them in macros that look like functions.  For example,
4088
4089@smallexample
4090#define sin(x)       \
4091(@{ double __value, __arg = (x);   \
4092   asm ("fsinx %1,%0": "=f" (__value): "f" (__arg));  \
4093   __value; @})
4094@end smallexample
4095
4096@noindent
4097Here the variable @code{__arg} is used to make sure that the instruction
4098operates on a proper @code{double} value, and to accept only those
4099arguments @code{x} which can convert automatically to a @code{double}.
4100
4101Another way to make sure the instruction operates on the correct data
4102type is to use a cast in the @code{asm}.  This is different from using a
4103variable @code{__arg} in that it converts more different types.  For
4104example, if the desired type were @code{int}, casting the argument to
4105@code{int} would accept a pointer with no complaint, while assigning the
4106argument to an @code{int} variable named @code{__arg} would warn about
4107using a pointer unless the caller explicitly casts it.
4108
4109If an @code{asm} has output operands, GCC assumes for optimization
4110purposes the instruction has no side effects except to change the output
4111operands.  This does not mean instructions with a side effect cannot be
4112used, but you must be careful, because the compiler may eliminate them
4113if the output operands aren't used, or move them out of loops, or
4114replace two with one if they constitute a common subexpression.  Also,
4115if your instruction does have a side effect on a variable that otherwise
4116appears not to change, the old value of the variable may be reused later
4117if it happens to be found in a register.
4118
4119You can prevent an @code{asm} instruction from being deleted, moved
4120significantly, or combined, by writing the keyword @code{volatile} after
4121the @code{asm}.  For example:
4122
4123@smallexample
4124#define get_and_set_priority(new)              \
4125(@{ int __old;                                  \
4126   asm volatile ("get_and_set_priority %0, %1" \
4127                 : "=g" (__old) : "g" (new));  \
4128   __old; @})
4129@end smallexample
4130
4131@noindent
4132If you write an @code{asm} instruction with no outputs, GCC will know
4133the instruction has side-effects and will not delete the instruction or
4134move it outside of loops.
4135
4136The @code{volatile} keyword indicates that the instruction has
4137important side-effects.  GCC will not delete a volatile @code{asm} if
4138it is reachable.  (The instruction can still be deleted if GCC can
4139prove that control-flow will never reach the location of the
4140instruction.)  In addition, GCC will not reschedule instructions
4141across a volatile @code{asm} instruction.  For example:
4142
4143@smallexample
4144*(volatile int *)addr = foo;
4145asm volatile ("eieio" : : );
4146@end smallexample
4147
4148@noindent
4149Assume @code{addr} contains the address of a memory mapped device
4150register.  The PowerPC @code{eieio} instruction (Enforce In-order
4151Execution of I/O) tells the CPU to make sure that the store to that
4152device register happens before it issues any other I/O@.
4153
4154Note that even a volatile @code{asm} instruction can be moved in ways
4155that appear insignificant to the compiler, such as across jump
4156instructions.  You can't expect a sequence of volatile @code{asm}
4157instructions to remain perfectly consecutive.  If you want consecutive
4158output, use a single @code{asm}.  Also, GCC will perform some
4159optimizations across a volatile @code{asm} instruction; GCC does not
4160``forget everything'' when it encounters a volatile @code{asm}
4161instruction the way some other compilers do.
4162
4163An @code{asm} instruction without any operands or clobbers (an ``old
4164style'' @code{asm}) will be treated identically to a volatile
4165@code{asm} instruction.
4166
4167It is a natural idea to look for a way to give access to the condition
4168code left by the assembler instruction.  However, when we attempted to
4169implement this, we found no way to make it work reliably.  The problem
4170is that output operands might need reloading, which would result in
4171additional following ``store'' instructions.  On most machines, these
4172instructions would alter the condition code before there was time to
4173test it.  This problem doesn't arise for ordinary ``test'' and
4174``compare'' instructions because they don't have any output operands.
4175
4176For reasons similar to those described above, it is not possible to give
4177an assembler instruction access to the condition code left by previous
4178instructions.
4179
4180If you are writing a header file that should be includable in ISO C
4181programs, write @code{__asm__} instead of @code{asm}.  @xref{Alternate
4182Keywords}.
4183
4184@subsection Size of an @code{asm}
4185
4186Some targets require that GCC track the size of each instruction used in
4187order to generate correct code.  Because the final length of an
4188@code{asm} is only known by the assembler, GCC must make an estimate as
4189to how big it will be.  The estimate is formed by counting the number of
4190statements in the pattern of the @code{asm} and multiplying that by the
4191length of the longest instruction on that processor.  Statements in the
4192@code{asm} are identified by newline characters and whatever statement
4193separator characters are supported by the assembler; on most processors
4194this is the `@code{;}' character.
4195
4196Normally, GCC's estimate is perfectly adequate to ensure that correct
4197code is generated, but it is possible to confuse the compiler if you use
4198pseudo instructions or assembler macros that expand into multiple real
4199instructions or if you use assembler directives that expand to more
4200space in the object file than would be needed for a single instruction.
4201If this happens then the assembler will produce a diagnostic saying that
4202a label is unreachable.
4203
4204@subsection i386 floating point asm operands
4205
4206There are several rules on the usage of stack-like regs in
4207asm_operands insns.  These rules apply only to the operands that are
4208stack-like regs:
4209
4210@enumerate
4211@item
4212Given a set of input regs that die in an asm_operands, it is
4213necessary to know which are implicitly popped by the asm, and
4214which must be explicitly popped by gcc.
4215
4216An input reg that is implicitly popped by the asm must be
4217explicitly clobbered, unless it is constrained to match an
4218output operand.
4219
4220@item
4221For any input reg that is implicitly popped by an asm, it is
4222necessary to know how to adjust the stack to compensate for the pop.
4223If any non-popped input is closer to the top of the reg-stack than
4224the implicitly popped reg, it would not be possible to know what the
4225stack looked like---it's not clear how the rest of the stack ``slides
4226up''.
4227
4228All implicitly popped input regs must be closer to the top of
4229the reg-stack than any input that is not implicitly popped.
4230
4231It is possible that if an input dies in an insn, reload might
4232use the input reg for an output reload.  Consider this example:
4233
4234@smallexample
4235asm ("foo" : "=t" (a) : "f" (b));
4236@end smallexample
4237
4238This asm says that input B is not popped by the asm, and that
4239the asm pushes a result onto the reg-stack, i.e., the stack is one
4240deeper after the asm than it was before.  But, it is possible that
4241reload will think that it can use the same reg for both the input and
4242the output, if input B dies in this insn.
4243
4244If any input operand uses the @code{f} constraint, all output reg
4245constraints must use the @code{&} earlyclobber.
4246
4247The asm above would be written as
4248
4249@smallexample
4250asm ("foo" : "=&t" (a) : "f" (b));
4251@end smallexample
4252
4253@item
4254Some operands need to be in particular places on the stack.  All
4255output operands fall in this category---there is no other way to
4256know which regs the outputs appear in unless the user indicates
4257this in the constraints.
4258
4259Output operands must specifically indicate which reg an output
4260appears in after an asm.  @code{=f} is not allowed: the operand
4261constraints must select a class with a single reg.
4262
4263@item
4264Output operands may not be ``inserted'' between existing stack regs.
4265Since no 387 opcode uses a read/write operand, all output operands
4266are dead before the asm_operands, and are pushed by the asm_operands.
4267It makes no sense to push anywhere but the top of the reg-stack.
4268
4269Output operands must start at the top of the reg-stack: output
4270operands may not ``skip'' a reg.
4271
4272@item
4273Some asm statements may need extra stack space for internal
4274calculations.  This can be guaranteed by clobbering stack registers
4275unrelated to the inputs and outputs.
4276
4277@end enumerate
4278
4279Here are a couple of reasonable asms to want to write.  This asm
4280takes one input, which is internally popped, and produces two outputs.
4281
4282@smallexample
4283asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
4284@end smallexample
4285
4286This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode,
4287and replaces them with one output.  The user must code the @code{st(1)}
4288clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs.
4289
4290@smallexample
4291asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
4292@end smallexample
4293
4294@include md.texi
4295
4296@node Asm Labels
4297@section Controlling Names Used in Assembler Code
4298@cindex assembler names for identifiers
4299@cindex names used in assembler code
4300@cindex identifiers, names in assembler code
4301
4302You can specify the name to be used in the assembler code for a C
4303function or variable by writing the @code{asm} (or @code{__asm__})
4304keyword after the declarator as follows:
4305
4306@smallexample
4307int foo asm ("myfoo") = 2;
4308@end smallexample
4309
4310@noindent
4311This specifies that the name to be used for the variable @code{foo} in
4312the assembler code should be @samp{myfoo} rather than the usual
4313@samp{_foo}.
4314
4315On systems where an underscore is normally prepended to the name of a C
4316function or variable, this feature allows you to define names for the
4317linker that do not start with an underscore.
4318
4319It does not make sense to use this feature with a non-static local
4320variable since such variables do not have assembler names.  If you are
4321trying to put the variable in a particular register, see @ref{Explicit
4322Reg Vars}.  GCC presently accepts such code with a warning, but will
4323probably be changed to issue an error, rather than a warning, in the
4324future.
4325
4326You cannot use @code{asm} in this way in a function @emph{definition}; but
4327you can get the same effect by writing a declaration for the function
4328before its definition and putting @code{asm} there, like this:
4329
4330@smallexample
4331extern func () asm ("FUNC");
4332
4333func (x, y)
4334     int x, y;
4335/* @r{@dots{}} */
4336@end smallexample
4337
4338It is up to you to make sure that the assembler names you choose do not
4339conflict with any other assembler symbols.  Also, you must not use a
4340register name; that would produce completely invalid assembler code.  GCC
4341does not as yet have the ability to store static variables in registers.
4342Perhaps that will be added.
4343
4344@node Explicit Reg Vars
4345@section Variables in Specified Registers
4346@cindex explicit register variables
4347@cindex variables in specified registers
4348@cindex specified registers
4349@cindex registers, global allocation
4350
4351GNU C allows you to put a few global variables into specified hardware
4352registers.  You can also specify the register in which an ordinary
4353register variable should be allocated.
4354
4355@itemize @bullet
4356@item
4357Global register variables reserve registers throughout the program.
4358This may be useful in programs such as programming language
4359interpreters which have a couple of global variables that are accessed
4360very often.
4361
4362@item
4363Local register variables in specific registers do not reserve the
4364registers.  The compiler's data flow analysis is capable of determining
4365where the specified registers contain live values, and where they are
4366available for other uses.  Stores into local register variables may be deleted
4367when they appear to be dead according to dataflow analysis.  References
4368to local register variables may be deleted or moved or simplified.
4369
4370These local variables are sometimes convenient for use with the extended
4371@code{asm} feature (@pxref{Extended Asm}), if you want to write one
4372output of the assembler instruction directly into a particular register.
4373(This will work provided the register you specify fits the constraints
4374specified for that operand in the @code{asm}.)
4375@end itemize
4376
4377@menu
4378* Global Reg Vars::
4379* Local Reg Vars::
4380@end menu
4381
4382@node Global Reg Vars
4383@subsection Defining Global Register Variables
4384@cindex global register variables
4385@cindex registers, global variables in
4386
4387You can define a global register variable in GNU C like this:
4388
4389@smallexample
4390register int *foo asm ("a5");
4391@end smallexample
4392
4393@noindent
4394Here @code{a5} is the name of the register which should be used.  Choose a
4395register which is normally saved and restored by function calls on your
4396machine, so that library routines will not clobber it.
4397
4398Naturally the register name is cpu-dependent, so you would need to
4399conditionalize your program according to cpu type.  The register
4400@code{a5} would be a good choice on a 68000 for a variable of pointer
4401type.  On machines with register windows, be sure to choose a ``global''
4402register that is not affected magically by the function call mechanism.
4403
4404In addition, operating systems on one type of cpu may differ in how they
4405name the registers; then you would need additional conditionals.  For
4406example, some 68000 operating systems call this register @code{%a5}.
4407
4408Eventually there may be a way of asking the compiler to choose a register
4409automatically, but first we need to figure out how it should choose and
4410how to enable you to guide the choice.  No solution is evident.
4411
4412Defining a global register variable in a certain register reserves that
4413register entirely for this use, at least within the current compilation.
4414The register will not be allocated for any other purpose in the functions
4415in the current compilation.  The register will not be saved and restored by
4416these functions.  Stores into this register are never deleted even if they
4417would appear to be dead, but references may be deleted or moved or
4418simplified.
4419
4420It is not safe to access the global register variables from signal
4421handlers, or from more than one thread of control, because the system
4422library routines may temporarily use the register for other things (unless
4423you recompile them specially for the task at hand).
4424
4425@cindex @code{qsort}, and global register variables
4426It is not safe for one function that uses a global register variable to
4427call another such function @code{foo} by way of a third function
4428@code{lose} that was compiled without knowledge of this variable (i.e.@: in a
4429different source file in which the variable wasn't declared).  This is
4430because @code{lose} might save the register and put some other value there.
4431For example, you can't expect a global register variable to be available in
4432the comparison-function that you pass to @code{qsort}, since @code{qsort}
4433might have put something else in that register.  (If you are prepared to
4434recompile @code{qsort} with the same global register variable, you can
4435solve this problem.)
4436
4437If you want to recompile @code{qsort} or other source files which do not
4438actually use your global register variable, so that they will not use that
4439register for any other purpose, then it suffices to specify the compiler
4440option @option{-ffixed-@var{reg}}.  You need not actually add a global
4441register declaration to their source code.
4442
4443A function which can alter the value of a global register variable cannot
4444safely be called from a function compiled without this variable, because it
4445could clobber the value the caller expects to find there on return.
4446Therefore, the function which is the entry point into the part of the
4447program that uses the global register variable must explicitly save and
4448restore the value which belongs to its caller.
4449
4450@cindex register variable after @code{longjmp}
4451@cindex global register after @code{longjmp}
4452@cindex value after @code{longjmp}
4453@findex longjmp
4454@findex setjmp
4455On most machines, @code{longjmp} will restore to each global register
4456variable the value it had at the time of the @code{setjmp}.  On some
4457machines, however, @code{longjmp} will not change the value of global
4458register variables.  To be portable, the function that called @code{setjmp}
4459should make other arrangements to save the values of the global register
4460variables, and to restore them in a @code{longjmp}.  This way, the same
4461thing will happen regardless of what @code{longjmp} does.
4462
4463All global register variable declarations must precede all function
4464definitions.  If such a declaration could appear after function
4465definitions, the declaration would be too late to prevent the register from
4466being used for other purposes in the preceding functions.
4467
4468Global register variables may not have initial values, because an
4469executable file has no means to supply initial contents for a register.
4470
4471On the SPARC, there are reports that g3 @dots{} g7 are suitable
4472registers, but certain library functions, such as @code{getwd}, as well
4473as the subroutines for division and remainder, modify g3 and g4.  g1 and
4474g2 are local temporaries.
4475
4476On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7.
4477Of course, it will not do to use more than a few of those.
4478
4479@node Local Reg Vars
4480@subsection Specifying Registers for Local Variables
4481@cindex local variables, specifying registers
4482@cindex specifying registers for local variables
4483@cindex registers for local variables
4484
4485You can define a local register variable with a specified register
4486like this:
4487
4488@smallexample
4489register int *foo asm ("a5");
4490@end smallexample
4491
4492@noindent
4493Here @code{a5} is the name of the register which should be used.  Note
4494that this is the same syntax used for defining global register
4495variables, but for a local variable it would appear within a function.
4496
4497Naturally the register name is cpu-dependent, but this is not a
4498problem, since specific registers are most often useful with explicit
4499assembler instructions (@pxref{Extended Asm}).  Both of these things
4500generally require that you conditionalize your program according to
4501cpu type.
4502
4503In addition, operating systems on one type of cpu may differ in how they
4504name the registers; then you would need additional conditionals.  For
4505example, some 68000 operating systems call this register @code{%a5}.
4506
4507Defining such a register variable does not reserve the register; it
4508remains available for other uses in places where flow control determines
4509the variable's value is not live.  However, these registers are made
4510unavailable for use in the reload pass; excessive use of this feature
4511leaves the compiler too few available registers to compile certain
4512functions.
4513
4514This option does not guarantee that GCC will generate code that has
4515this variable in the register you specify at all times.  You may not
4516code an explicit reference to this register in an @code{asm} statement
4517and assume it will always refer to this variable.
4518
4519Stores into local register variables may be deleted when they appear to be dead
4520according to dataflow analysis.  References to local register variables may
4521be deleted or moved or simplified.
4522
4523@node Alternate Keywords
4524@section Alternate Keywords
4525@cindex alternate keywords
4526@cindex keywords, alternate
4527
4528@option{-ansi} and the various @option{-std} options disable certain
4529keywords.  This causes trouble when you want to use GNU C extensions, or
4530a general-purpose header file that should be usable by all programs,
4531including ISO C programs.  The keywords @code{asm}, @code{typeof} and
4532@code{inline} are not available in programs compiled with
4533@option{-ansi} or @option{-std} (although @code{inline} can be used in a
4534program compiled with @option{-std=c99}).  The ISO C99 keyword
4535@code{restrict} is only available when @option{-std=gnu99} (which will
4536eventually be the default) or @option{-std=c99} (or the equivalent
4537@option{-std=iso9899:1999}) is used.
4538
4539The way to solve these problems is to put @samp{__} at the beginning and
4540end of each problematical keyword.  For example, use @code{__asm__}
4541instead of @code{asm}, and @code{__inline__} instead of @code{inline}.
4542
4543Other C compilers won't accept these alternative keywords; if you want to
4544compile with another compiler, you can define the alternate keywords as
4545macros to replace them with the customary keywords.  It looks like this:
4546
4547@smallexample
4548#ifndef __GNUC__
4549#define __asm__ asm
4550#endif
4551@end smallexample
4552
4553@findex __extension__
4554@opindex pedantic
4555@option{-pedantic} and other options cause warnings for many GNU C extensions.
4556You can
4557prevent such warnings within one expression by writing
4558@code{__extension__} before the expression.  @code{__extension__} has no
4559effect aside from this.
4560
4561@node Incomplete Enums
4562@section Incomplete @code{enum} Types
4563
4564You can define an @code{enum} tag without specifying its possible values.
4565This results in an incomplete type, much like what you get if you write
4566@code{struct foo} without describing the elements.  A later declaration
4567which does specify the possible values completes the type.
4568
4569You can't allocate variables or storage using the type while it is
4570incomplete.  However, you can work with pointers to that type.
4571
4572This extension may not be very useful, but it makes the handling of
4573@code{enum} more consistent with the way @code{struct} and @code{union}
4574are handled.
4575
4576This extension is not supported by GNU C++.
4577
4578@node Function Names
4579@section Function Names as Strings
4580@cindex @code{__func__} identifier
4581@cindex @code{__FUNCTION__} identifier
4582@cindex @code{__PRETTY_FUNCTION__} identifier
4583
4584GCC provides three magic variables which hold the name of the current
4585function, as a string.  The first of these is @code{__func__}, which
4586is part of the C99 standard:
4587
4588@display
4589The identifier @code{__func__} is implicitly declared by the translator
4590as if, immediately following the opening brace of each function
4591definition, the declaration
4592
4593@smallexample
4594static const char __func__[] = "function-name";
4595@end smallexample
4596
4597appeared, where function-name is the name of the lexically-enclosing
4598function.  This name is the unadorned name of the function.
4599@end display
4600
4601@code{__FUNCTION__} is another name for @code{__func__}.  Older
4602versions of GCC recognize only this name.  However, it is not
4603standardized.  For maximum portability, we recommend you use
4604@code{__func__}, but provide a fallback definition with the
4605preprocessor:
4606
4607@smallexample
4608#if __STDC_VERSION__ < 199901L
4609# if __GNUC__ >= 2
4610#  define __func__ __FUNCTION__
4611# else
4612#  define __func__ "<unknown>"
4613# endif
4614#endif
4615@end smallexample
4616
4617In C, @code{__PRETTY_FUNCTION__} is yet another name for
4618@code{__func__}.  However, in C++, @code{__PRETTY_FUNCTION__} contains
4619the type signature of the function as well as its bare name.  For
4620example, this program:
4621
4622@smallexample
4623extern "C" @{
4624extern int printf (char *, ...);
4625@}
4626
4627class a @{
4628 public:
4629  void sub (int i)
4630    @{
4631      printf ("__FUNCTION__ = %s\n", __FUNCTION__);
4632      printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
4633    @}
4634@};
4635
4636int
4637main (void)
4638@{
4639  a ax;
4640  ax.sub (0);
4641  return 0;
4642@}
4643@end smallexample
4644
4645@noindent
4646gives this output:
4647
4648@smallexample
4649__FUNCTION__ = sub
4650__PRETTY_FUNCTION__ = void a::sub(int)
4651@end smallexample
4652
4653These identifiers are not preprocessor macros.  In GCC 3.3 and
4654earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__}
4655were treated as string literals; they could be used to initialize
4656@code{char} arrays, and they could be concatenated with other string
4657literals.  GCC 3.4 and later treat them as variables, like
4658@code{__func__}.  In C++, @code{__FUNCTION__} and
4659@code{__PRETTY_FUNCTION__} have always been variables.
4660
4661@node Return Address
4662@section Getting the Return or Frame Address of a Function
4663
4664These functions may be used to get information about the callers of a
4665function.
4666
4667@deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level})
4668This function returns the return address of the current function, or of
4669one of its callers.  The @var{level} argument is number of frames to
4670scan up the call stack.  A value of @code{0} yields the return address
4671of the current function, a value of @code{1} yields the return address
4672of the caller of the current function, and so forth. When inlining
4673the expected behavior is that the function will return the address of
4674the function that will be returned to.  To work around this behavior use
4675the @code{noinline} function attribute.
4676
4677The @var{level} argument must be a constant integer.
4678
4679On some machines it may be impossible to determine the return address of
4680any function other than the current one; in such cases, or when the top
4681of the stack has been reached, this function will return @code{0} or a
4682random value. In addition, @code{__builtin_frame_address} may be used
4683to determine if the top of the stack has been reached.
4684
4685This function should only be used with a nonzero argument for debugging
4686purposes.
4687@end deftypefn
4688
4689@deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level})
4690This function is similar to @code{__builtin_return_address}, but it
4691returns the address of the function frame rather than the return address
4692of the function.  Calling @code{__builtin_frame_address} with a value of
4693@code{0} yields the frame address of the current function, a value of
4694@code{1} yields the frame address of the caller of the current function,
4695and so forth.
4696
4697The frame is the area on the stack which holds local variables and saved
4698registers.  The frame address is normally the address of the first word
4699pushed on to the stack by the function.  However, the exact definition
4700depends upon the processor and the calling convention.  If the processor
4701has a dedicated frame pointer register, and the function has a frame,
4702then @code{__builtin_frame_address} will return the value of the frame
4703pointer register.
4704
4705On some machines it may be impossible to determine the frame address of
4706any function other than the current one; in such cases, or when the top
4707of the stack has been reached, this function will return @code{0} if
4708the first frame pointer is properly initialized by the startup code.
4709
4710This function should only be used with a nonzero argument for debugging
4711purposes.
4712@end deftypefn
4713
4714@node Vector Extensions
4715@section Using vector instructions through built-in functions
4716
4717On some targets, the instruction set contains SIMD vector instructions that
4718operate on multiple values contained in one large register at the same time.
4719For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used
4720this way.
4721
4722The first step in using these extensions is to provide the necessary data
4723types.  This should be done using an appropriate @code{typedef}:
4724
4725@smallexample
4726typedef int v4si __attribute__ ((mode(V4SI)));
4727@end smallexample
4728
4729The base type @code{int} is effectively ignored by the compiler, the
4730actual properties of the new type @code{v4si} are defined by the
4731@code{__attribute__}.  It defines the machine mode to be used; for vector
4732types these have the form @code{V@var{n}@var{B}}; @var{n} should be the
4733number of elements in the vector, and @var{B} should be the base mode of the
4734individual elements.  The following can be used as base modes:
4735
4736@table @code
4737@item QI
4738An integer that is as wide as the smallest addressable unit, usually 8 bits.
4739@item HI
4740An integer, twice as wide as a QI mode integer, usually 16 bits.
4741@item SI
4742An integer, four times as wide as a QI mode integer, usually 32 bits.
4743@item DI
4744An integer, eight times as wide as a QI mode integer, usually 64 bits.
4745@item SF
4746A floating point value, as wide as a SI mode integer, usually 32 bits.
4747@item DF
4748A floating point value, as wide as a DI mode integer, usually 64 bits.
4749@end table
4750
4751Specifying a combination that is not valid for the current architecture
4752will cause GCC to synthesize the instructions using a narrower mode.
4753For example, if you specify a variable of type @code{V4SI} and your
4754architecture does not allow for this specific SIMD type, GCC will
4755produce code that uses 4 @code{SIs}.
4756
4757The types defined in this manner can be used with a subset of normal C
4758operations.  Currently, GCC will allow using the following operators
4759on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@.
4760
4761The operations behave like C++ @code{valarrays}.  Addition is defined as
4762the addition of the corresponding elements of the operands.  For
4763example, in the code below, each of the 4 elements in @var{a} will be
4764added to the corresponding 4 elements in @var{b} and the resulting
4765vector will be stored in @var{c}.
4766
4767@smallexample
4768typedef int v4si __attribute__ ((mode(V4SI)));
4769
4770v4si a, b, c;
4771
4772c = a + b;
4773@end smallexample
4774
4775Subtraction, multiplication, division, and the logical operations
4776operate in a similar manner.  Likewise, the result of using the unary
4777minus or complement operators on a vector type is a vector whose
4778elements are the negative or complemented values of the corresponding
4779elements in the operand.
4780
4781You can declare variables and use them in function calls and returns, as
4782well as in assignments and some casts.  You can specify a vector type as
4783a return type for a function.  Vector types can also be used as function
4784arguments.  It is possible to cast from one vector type to another,
4785provided they are of the same size (in fact, you can also cast vectors
4786to and from other datatypes of the same size).
4787
4788You cannot operate between vectors of different lengths or different
4789signedness without a cast.
4790
4791A port that supports hardware vector operations, usually provides a set
4792of built-in functions that can be used to operate on vectors.  For
4793example, a function to add two vectors and multiply the result by a
4794third could look like this:
4795
4796@smallexample
4797v4si f (v4si a, v4si b, v4si c)
4798@{
4799  v4si tmp = __builtin_addv4si (a, b);
4800  return __builtin_mulv4si (tmp, c);
4801@}
4802
4803@end smallexample
4804
4805@node Other Builtins
4806@section Other built-in functions provided by GCC
4807@cindex built-in functions
4808@findex __builtin_isgreater
4809@findex __builtin_isgreaterequal
4810@findex __builtin_isless
4811@findex __builtin_islessequal
4812@findex __builtin_islessgreater
4813@findex __builtin_isunordered
4814@findex _Exit
4815@findex _exit
4816@findex abort
4817@findex abs
4818@findex acos
4819@findex acosf
4820@findex acosh
4821@findex acoshf
4822@findex acoshl
4823@findex acosl
4824@findex alloca
4825@findex asin
4826@findex asinf
4827@findex asinh
4828@findex asinhf
4829@findex asinhl
4830@findex asinl
4831@findex atan
4832@findex atan2
4833@findex atan2f
4834@findex atan2l
4835@findex atanf
4836@findex atanh
4837@findex atanhf
4838@findex atanhl
4839@findex atanl
4840@findex bcmp
4841@findex bzero
4842@findex cabs
4843@findex cabsf
4844@findex cabsl
4845@findex cacos
4846@findex cacosf
4847@findex cacosh
4848@findex cacoshf
4849@findex cacoshl
4850@findex cacosl
4851@findex calloc
4852@findex carg
4853@findex cargf
4854@findex cargl
4855@findex casin
4856@findex casinf
4857@findex casinh
4858@findex casinhf
4859@findex casinhl
4860@findex casinl
4861@findex catan
4862@findex catanf
4863@findex catanh
4864@findex catanhf
4865@findex catanhl
4866@findex catanl
4867@findex cbrt
4868@findex cbrtf
4869@findex cbrtl
4870@findex ccos
4871@findex ccosf
4872@findex ccosh
4873@findex ccoshf
4874@findex ccoshl
4875@findex ccosl
4876@findex ceil
4877@findex ceilf
4878@findex ceill
4879@findex cexp
4880@findex cexpf
4881@findex cexpl
4882@findex cimag
4883@findex cimagf
4884@findex cimagl
4885@findex conj
4886@findex conjf
4887@findex conjl
4888@findex copysign
4889@findex copysignf
4890@findex copysignl
4891@findex cos
4892@findex cosf
4893@findex cosh
4894@findex coshf
4895@findex coshl
4896@findex cosl
4897@findex cpow
4898@findex cpowf
4899@findex cpowl
4900@findex cproj
4901@findex cprojf
4902@findex cprojl
4903@findex creal
4904@findex crealf
4905@findex creall
4906@findex csin
4907@findex csinf
4908@findex csinh
4909@findex csinhf
4910@findex csinhl
4911@findex csinl
4912@findex csqrt
4913@findex csqrtf
4914@findex csqrtl
4915@findex ctan
4916@findex ctanf
4917@findex ctanh
4918@findex ctanhf
4919@findex ctanhl
4920@findex ctanl
4921@findex dcgettext
4922@findex dgettext
4923@findex drem
4924@findex dremf
4925@findex dreml
4926@findex erf
4927@findex erfc
4928@findex erfcf
4929@findex erfcl
4930@findex erff
4931@findex erfl
4932@findex exit
4933@findex exp
4934@findex exp10
4935@findex exp10f
4936@findex exp10l
4937@findex exp2
4938@findex exp2f
4939@findex exp2l
4940@findex expf
4941@findex expl
4942@findex expm1
4943@findex expm1f
4944@findex expm1l
4945@findex fabs
4946@findex fabsf
4947@findex fabsl
4948@findex fdim
4949@findex fdimf
4950@findex fdiml
4951@findex ffs
4952@findex floor
4953@findex floorf
4954@findex floorl
4955@findex fma
4956@findex fmaf
4957@findex fmal
4958@findex fmax
4959@findex fmaxf
4960@findex fmaxl
4961@findex fmin
4962@findex fminf
4963@findex fminl
4964@findex fmod
4965@findex fmodf
4966@findex fmodl
4967@findex fprintf
4968@findex fprintf_unlocked
4969@findex fputs
4970@findex fputs_unlocked
4971@findex frexp
4972@findex frexpf
4973@findex frexpl
4974@findex fscanf
4975@findex gamma
4976@findex gammaf
4977@findex gammal
4978@findex gettext
4979@findex hypot
4980@findex hypotf
4981@findex hypotl
4982@findex ilogb
4983@findex ilogbf
4984@findex ilogbl
4985@findex imaxabs
4986@findex index
4987@findex j0
4988@findex j0f
4989@findex j0l
4990@findex j1
4991@findex j1f
4992@findex j1l
4993@findex jn
4994@findex jnf
4995@findex jnl
4996@findex labs
4997@findex ldexp
4998@findex ldexpf
4999@findex ldexpl
5000@findex lgamma
5001@findex lgammaf
5002@findex lgammal
5003@findex llabs
5004@findex llrint
5005@findex llrintf
5006@findex llrintl
5007@findex llround
5008@findex llroundf
5009@findex llroundl
5010@findex log
5011@findex log10
5012@findex log10f
5013@findex log10l
5014@findex log1p
5015@findex log1pf
5016@findex log1pl
5017@findex log2
5018@findex log2f
5019@findex log2l
5020@findex logb
5021@findex logbf
5022@findex logbl
5023@findex logf
5024@findex logl
5025@findex lrint
5026@findex lrintf
5027@findex lrintl
5028@findex lround
5029@findex lroundf
5030@findex lroundl
5031@findex malloc
5032@findex memcmp
5033@findex memcpy
5034@findex mempcpy
5035@findex memset
5036@findex modf
5037@findex modff
5038@findex modfl
5039@findex nearbyint
5040@findex nearbyintf
5041@findex nearbyintl
5042@findex nextafter
5043@findex nextafterf
5044@findex nextafterl
5045@findex nexttoward
5046@findex nexttowardf
5047@findex nexttowardl
5048@findex pow
5049@findex pow10
5050@findex pow10f
5051@findex pow10l
5052@findex powf
5053@findex powl
5054@findex printf
5055@findex printf_unlocked
5056@findex putchar
5057@findex puts
5058@findex remainder
5059@findex remainderf
5060@findex remainderl
5061@findex remquo
5062@findex remquof
5063@findex remquol
5064@findex rindex
5065@findex rint
5066@findex rintf
5067@findex rintl
5068@findex round
5069@findex roundf
5070@findex roundl
5071@findex scalb
5072@findex scalbf
5073@findex scalbl
5074@findex scalbln
5075@findex scalblnf
5076@findex scalblnf
5077@findex scalbn
5078@findex scalbnf
5079@findex scanfnl
5080@findex significand
5081@findex significandf
5082@findex significandl
5083@findex sin
5084@findex sincos
5085@findex sincosf
5086@findex sincosl
5087@findex sinf
5088@findex sinh
5089@findex sinhf
5090@findex sinhl
5091@findex sinl
5092@findex snprintf
5093@findex sprintf
5094@findex sqrt
5095@findex sqrtf
5096@findex sqrtl
5097@findex sscanf
5098@findex stpcpy
5099@findex strcat
5100@findex strchr
5101@findex strcmp
5102@findex strcpy
5103@findex strcspn
5104@findex strdup
5105@findex strfmon
5106@findex strftime
5107@findex strlen
5108@findex strncat
5109@findex strncmp
5110@findex strncpy
5111@findex strpbrk
5112@findex strrchr
5113@findex strspn
5114@findex strstr
5115@findex tan
5116@findex tanf
5117@findex tanh
5118@findex tanhf
5119@findex tanhl
5120@findex tanl
5121@findex tgamma
5122@findex tgammaf
5123@findex tgammal
5124@findex trunc
5125@findex truncf
5126@findex truncl
5127@findex vfprintf
5128@findex vfscanf
5129@findex vprintf
5130@findex vscanf
5131@findex vsnprintf
5132@findex vsprintf
5133@findex vsscanf
5134@findex y0
5135@findex y0f
5136@findex y0l
5137@findex y1
5138@findex y1f
5139@findex y1l
5140@findex yn
5141@findex ynf
5142@findex ynl
5143
5144GCC provides a large number of built-in functions other than the ones
5145mentioned above.  Some of these are for internal use in the processing
5146of exceptions or variable-length argument lists and will not be
5147documented here because they may change from time to time; we do not
5148recommend general use of these functions.
5149
5150The remaining functions are provided for optimization purposes.
5151
5152@opindex fno-builtin
5153GCC includes built-in versions of many of the functions in the standard
5154C library.  The versions prefixed with @code{__builtin_} will always be
5155treated as having the same meaning as the C library function even if you
5156specify the @option{-fno-builtin} option. (@pxref{C Dialect Options})
5157Many of these functions are only optimized in certain cases; if they are
5158not optimized in a particular case, a call to the library function will
5159be emitted.
5160
5161@opindex ansi
5162@opindex std
5163Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or
5164@option{-std=c99}), the functions
5165@code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero},
5166@code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml},
5167@code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll},
5168@code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked},
5169@code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext},
5170@code{index}, @code{j0f}, @code{j0l}, @code{j0}, @code{j1f}, @code{j1l},
5171@code{j1}, @code{jnf}, @code{jnl}, @code{jn}, @code{mempcpy},
5172@code{pow10f}, @code{pow10l}, @code{pow10}, @code{printf_unlocked},
5173@code{rindex}, @code{scalbf}, @code{scalbl}, @code{scalb},
5174@code{significandf}, @code{significandl}, @code{significand},
5175@code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy},
5176@code{strdup}, @code{strfmon}, @code{y0f}, @code{y0l}, @code{y0},
5177@code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, @code{ynl} and @code{yn}
5178may be handled as built-in functions.
5179All these functions have corresponding versions
5180prefixed with @code{__builtin_}, which may be used even in strict C89
5181mode.
5182
5183The ISO C99 functions
5184@code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf},
5185@code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh},
5186@code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf},
5187@code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos},
5188@code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf},
5189@code{casinhl}, @code{casinh}, @code{casinl}, @code{casin},
5190@code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh},
5191@code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt},
5192@code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl},
5193@code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf},
5194@code{cimagl}, @code{cimag},
5195@code{conjf}, @code{conjl}, @code{conj}, @code{copysignf},
5196@code{copysignl}, @code{copysign}, @code{cpowf}, @code{cpowl},
5197@code{cpow}, @code{cprojf}, @code{cprojl}, @code{cproj}, @code{crealf},
5198@code{creall}, @code{creal}, @code{csinf}, @code{csinhf}, @code{csinhl},
5199@code{csinh}, @code{csinl}, @code{csin}, @code{csqrtf}, @code{csqrtl},
5200@code{csqrt}, @code{ctanf}, @code{ctanhf}, @code{ctanhl}, @code{ctanh},
5201@code{ctanl}, @code{ctan}, @code{erfcf}, @code{erfcl}, @code{erfc},
5202@code{erff}, @code{erfl}, @code{erf}, @code{exp2f}, @code{exp2l},
5203@code{exp2}, @code{expm1f}, @code{expm1l}, @code{expm1}, @code{fdimf},
5204@code{fdiml}, @code{fdim}, @code{fmaf}, @code{fmal}, @code{fmaxf},
5205@code{fmaxl}, @code{fmax}, @code{fma}, @code{fminf}, @code{fminl},
5206@code{fmin}, @code{hypotf}, @code{hypotl}, @code{hypot}, @code{ilogbf},
5207@code{ilogbl}, @code{ilogb}, @code{imaxabs}, @code{lgammaf},
5208@code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf},
5209@code{llrintl}, @code{llrint}, @code{llroundf}, @code{llroundl},
5210@code{llround}, @code{log1pf}, @code{log1pl}, @code{log1p},
5211@code{log2f}, @code{log2l}, @code{log2}, @code{logbf}, @code{logbl},
5212@code{logb}, @code{lrintf}, @code{lrintl}, @code{lrint}, @code{lroundf},
5213@code{lroundl}, @code{lround}, @code{nearbyintf}, @code{nearbyintl},
5214@code{nearbyint}, @code{nextafterf}, @code{nextafterl},
5215@code{nextafter}, @code{nexttowardf}, @code{nexttowardl},
5216@code{nexttoward}, @code{remainderf}, @code{remainderl},
5217@code{remainder}, @code{remquof}, @code{remquol}, @code{remquo},
5218@code{rintf}, @code{rintl}, @code{rint}, @code{roundf}, @code{roundl},
5219@code{round}, @code{scalblnf}, @code{scalblnl}, @code{scalbln},
5220@code{scalbnf}, @code{scalbnl}, @code{scalbn}, @code{snprintf},
5221@code{tgammaf}, @code{tgammal}, @code{tgamma}, @code{truncf},
5222@code{truncl}, @code{trunc}, @code{vfscanf}, @code{vscanf},
5223@code{vsnprintf} and @code{vsscanf}
5224are handled as built-in functions
5225except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}).
5226
5227There are also built-in versions of the ISO C99 functions
5228@code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f},
5229@code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill},
5230@code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf},
5231@code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl},
5232@code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf},
5233@code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl},
5234@code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf},
5235@code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl},
5236@code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl}
5237that are recognized in any mode since ISO C90 reserves these names for
5238the purpose to which ISO C99 puts them.  All these functions have
5239corresponding versions prefixed with @code{__builtin_}.
5240
5241The ISO C90 functions
5242@code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2},
5243@code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos},
5244@code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod},
5245@code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf}, @code{labs},
5246@code{ldexp}, @code{log10}, @code{log}, @code{malloc}, @code{memcmp},
5247@code{memcpy}, @code{memset}, @code{modf}, @code{pow}, @code{printf},
5248@code{putchar}, @code{puts}, @code{scanf}, @code{sinh}, @code{sin},
5249@code{snprintf}, @code{sprintf}, @code{sqrt}, @code{sscanf},
5250@code{strcat}, @code{strchr}, @code{strcmp}, @code{strcpy},
5251@code{strcspn}, @code{strlen}, @code{strncat}, @code{strncmp},
5252@code{strncpy}, @code{strpbrk}, @code{strrchr}, @code{strspn},
5253@code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf}, @code{vprintf}
5254and @code{vsprintf}
5255are all recognized as built-in functions unless
5256@option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}}
5257is specified for an individual function).  All of these functions have
5258corresponding versions prefixed with @code{__builtin_}.
5259
5260GCC provides built-in versions of the ISO C99 floating point comparison
5261macros that avoid raising exceptions for unordered operands.  They have
5262the same names as the standard macros ( @code{isgreater},
5263@code{isgreaterequal}, @code{isless}, @code{islessequal},
5264@code{islessgreater}, and @code{isunordered}) , with @code{__builtin_}
5265prefixed.  We intend for a library implementor to be able to simply
5266@code{#define} each standard macro to its built-in equivalent.
5267
5268@deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2})
5269
5270You can use the built-in function @code{__builtin_types_compatible_p} to
5271determine whether two types are the same.
5272
5273This built-in function returns 1 if the unqualified versions of the
5274types @var{type1} and @var{type2} (which are types, not expressions) are
5275compatible, 0 otherwise.  The result of this built-in function can be
5276used in integer constant expressions.
5277
5278This built-in function ignores top level qualifiers (e.g., @code{const},
5279@code{volatile}).  For example, @code{int} is equivalent to @code{const
5280int}.
5281
5282The type @code{int[]} and @code{int[5]} are compatible.  On the other
5283hand, @code{int} and @code{char *} are not compatible, even if the size
5284of their types, on the particular architecture are the same.  Also, the
5285amount of pointer indirection is taken into account when determining
5286similarity.  Consequently, @code{short *} is not similar to
5287@code{short **}.  Furthermore, two types that are typedefed are
5288considered compatible if their underlying types are compatible.
5289
5290An @code{enum} type is not considered to be compatible with another
5291@code{enum} type even if both are compatible with the same integer
5292type; this is what the C standard specifies.
5293For example, @code{enum @{foo, bar@}} is not similar to
5294@code{enum @{hot, dog@}}.
5295
5296You would typically use this function in code whose execution varies
5297depending on the arguments' types.  For example:
5298
5299@smallexample
5300#define foo(x)                                                  \
5301  (@{                                                           \
5302    typeof (x) tmp;                                             \
5303    if (__builtin_types_compatible_p (typeof (x), long double)) \
5304      tmp = foo_long_double (tmp);                              \
5305    else if (__builtin_types_compatible_p (typeof (x), double)) \
5306      tmp = foo_double (tmp);                                   \
5307    else if (__builtin_types_compatible_p (typeof (x), float))  \
5308      tmp = foo_float (tmp);                                    \
5309    else                                                        \
5310      abort ();                                                 \
5311    tmp;                                                        \
5312  @})
5313@end smallexample
5314
5315@emph{Note:} This construct is only available for C.
5316
5317@end deftypefn
5318
5319@deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2})
5320
5321You can use the built-in function @code{__builtin_choose_expr} to
5322evaluate code depending on the value of a constant expression.  This
5323built-in function returns @var{exp1} if @var{const_exp}, which is a
5324constant expression that must be able to be determined at compile time,
5325is nonzero.  Otherwise it returns 0.
5326
5327This built-in function is analogous to the @samp{? :} operator in C,
5328except that the expression returned has its type unaltered by promotion
5329rules.  Also, the built-in function does not evaluate the expression
5330that was not chosen.  For example, if @var{const_exp} evaluates to true,
5331@var{exp2} is not evaluated even if it has side-effects.
5332
5333This built-in function can return an lvalue if the chosen argument is an
5334lvalue.
5335
5336If @var{exp1} is returned, the return type is the same as @var{exp1}'s
5337type.  Similarly, if @var{exp2} is returned, its return type is the same
5338as @var{exp2}.
5339
5340Example:
5341
5342@smallexample
5343#define foo(x)                                                    \
5344  __builtin_choose_expr (                                         \
5345    __builtin_types_compatible_p (typeof (x), double),            \
5346    foo_double (x),                                               \
5347    __builtin_choose_expr (                                       \
5348      __builtin_types_compatible_p (typeof (x), float),           \
5349      foo_float (x),                                              \
5350      /* @r{The void expression results in a compile-time error}  \
5351         @r{when assigning the result to something.}  */          \
5352      (void)0))
5353@end smallexample
5354
5355@emph{Note:} This construct is only available for C.  Furthermore, the
5356unused expression (@var{exp1} or @var{exp2} depending on the value of
5357@var{const_exp}) may still generate syntax errors.  This may change in
5358future revisions.
5359
5360@end deftypefn
5361
5362@deftypefn {Built-in Function} int __builtin_constant_p (@var{exp})
5363You can use the built-in function @code{__builtin_constant_p} to
5364determine if a value is known to be constant at compile-time and hence
5365that GCC can perform constant-folding on expressions involving that
5366value.  The argument of the function is the value to test.  The function
5367returns the integer 1 if the argument is known to be a compile-time
5368constant and 0 if it is not known to be a compile-time constant.  A
5369return of 0 does not indicate that the value is @emph{not} a constant,
5370but merely that GCC cannot prove it is a constant with the specified
5371value of the @option{-O} option.
5372
5373You would typically use this function in an embedded application where
5374memory was a critical resource.  If you have some complex calculation,
5375you may want it to be folded if it involves constants, but need to call
5376a function if it does not.  For example:
5377
5378@smallexample
5379#define Scale_Value(X)      \
5380  (__builtin_constant_p (X) \
5381  ? ((X) * SCALE + OFFSET) : Scale (X))
5382@end smallexample
5383
5384You may use this built-in function in either a macro or an inline
5385function.  However, if you use it in an inlined function and pass an
5386argument of the function as the argument to the built-in, GCC will
5387never return 1 when you call the inline function with a string constant
5388or compound literal (@pxref{Compound Literals}) and will not return 1
5389when you pass a constant numeric value to the inline function unless you
5390specify the @option{-O} option.
5391
5392You may also use @code{__builtin_constant_p} in initializers for static
5393data.  For instance, you can write
5394
5395@smallexample
5396static const int table[] = @{
5397   __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
5398   /* @r{@dots{}} */
5399@};
5400@end smallexample
5401
5402@noindent
5403This is an acceptable initializer even if @var{EXPRESSION} is not a
5404constant expression.  GCC must be more conservative about evaluating the
5405built-in in this case, because it has no opportunity to perform
5406optimization.
5407
5408Previous versions of GCC did not accept this built-in in data
5409initializers.  The earliest version where it is completely safe is
54103.0.1.
5411@end deftypefn
5412
5413@deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c})
5414@opindex fprofile-arcs
5415You may use @code{__builtin_expect} to provide the compiler with
5416branch prediction information.  In general, you should prefer to
5417use actual profile feedback for this (@option{-fprofile-arcs}), as
5418programmers are notoriously bad at predicting how their programs
5419actually perform.  However, there are applications in which this
5420data is hard to collect.
5421
5422The return value is the value of @var{exp}, which should be an
5423integral expression.  The value of @var{c} must be a compile-time
5424constant.  The semantics of the built-in are that it is expected
5425that @var{exp} == @var{c}.  For example:
5426
5427@smallexample
5428if (__builtin_expect (x, 0))
5429  foo ();
5430@end smallexample
5431
5432@noindent
5433would indicate that we do not expect to call @code{foo}, since
5434we expect @code{x} to be zero.  Since you are limited to integral
5435expressions for @var{exp}, you should use constructions such as
5436
5437@smallexample
5438if (__builtin_expect (ptr != NULL, 1))
5439  error ();
5440@end smallexample
5441
5442@noindent
5443when testing pointer or floating-point values.
5444@end deftypefn
5445
5446@deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...)
5447This function is used to minimize cache-miss latency by moving data into
5448a cache before it is accessed.
5449You can insert calls to @code{__builtin_prefetch} into code for which
5450you know addresses of data in memory that is likely to be accessed soon.
5451If the target supports them, data prefetch instructions will be generated.
5452If the prefetch is done early enough before the access then the data will
5453be in the cache by the time it is accessed.
5454
5455The value of @var{addr} is the address of the memory to prefetch.
5456There are two optional arguments, @var{rw} and @var{locality}.
5457The value of @var{rw} is a compile-time constant one or zero; one
5458means that the prefetch is preparing for a write to the memory address
5459and zero, the default, means that the prefetch is preparing for a read.
5460The value @var{locality} must be a compile-time constant integer between
5461zero and three.  A value of zero means that the data has no temporal
5462locality, so it need not be left in the cache after the access.  A value
5463of three means that the data has a high degree of temporal locality and
5464should be left in all levels of cache possible.  Values of one and two
5465mean, respectively, a low or moderate degree of temporal locality.  The
5466default is three.
5467
5468@smallexample
5469for (i = 0; i < n; i++)
5470  @{
5471    a[i] = a[i] + b[i];
5472    __builtin_prefetch (&a[i+j], 1, 1);
5473    __builtin_prefetch (&b[i+j], 0, 1);
5474    /* @r{@dots{}} */
5475  @}
5476@end smallexample
5477
5478Data prefetch does not generate faults if @var{addr} is invalid, but
5479the address expression itself must be valid.  For example, a prefetch
5480of @code{p->next} will not fault if @code{p->next} is not a valid
5481address, but evaluation will fault if @code{p} is not a valid address.
5482
5483If the target does not support data prefetch, the address expression
5484is evaluated if it includes side effects but no other code is generated
5485and GCC does not issue a warning.
5486@end deftypefn
5487
5488@deftypefn {Built-in Function} double __builtin_huge_val (void)
5489Returns a positive infinity, if supported by the floating-point format,
5490else @code{DBL_MAX}.  This function is suitable for implementing the
5491ISO C macro @code{HUGE_VAL}.
5492@end deftypefn
5493
5494@deftypefn {Built-in Function} float __builtin_huge_valf (void)
5495Similar to @code{__builtin_huge_val}, except the return type is @code{float}.
5496@end deftypefn
5497
5498@deftypefn {Built-in Function} {long double} __builtin_huge_vall (void)
5499Similar to @code{__builtin_huge_val}, except the return
5500type is @code{long double}.
5501@end deftypefn
5502
5503@deftypefn {Built-in Function} double __builtin_inf (void)
5504Similar to @code{__builtin_huge_val}, except a warning is generated
5505if the target floating-point format does not support infinities.
5506This function is suitable for implementing the ISO C99 macro @code{INFINITY}.
5507@end deftypefn
5508
5509@deftypefn {Built-in Function} float __builtin_inff (void)
5510Similar to @code{__builtin_inf}, except the return type is @code{float}.
5511@end deftypefn
5512
5513@deftypefn {Built-in Function} {long double} __builtin_infl (void)
5514Similar to @code{__builtin_inf}, except the return
5515type is @code{long double}.
5516@end deftypefn
5517
5518@deftypefn {Built-in Function} double __builtin_nan (const char *str)
5519This is an implementation of the ISO C99 function @code{nan}.
5520
5521Since ISO C99 defines this function in terms of @code{strtod}, which we
5522do not implement, a description of the parsing is in order.  The string
5523is parsed as by @code{strtol}; that is, the base is recognized by
5524leading @samp{0} or @samp{0x} prefixes.  The number parsed is placed
5525in the significand such that the least significant bit of the number
5526is at the least significant bit of the significand.  The number is
5527truncated to fit the significand field provided.  The significand is
5528forced to be a quiet NaN.
5529
5530This function, if given a string literal, is evaluated early enough
5531that it is considered a compile-time constant.
5532@end deftypefn
5533
5534@deftypefn {Built-in Function} float __builtin_nanf (const char *str)
5535Similar to @code{__builtin_nan}, except the return type is @code{float}.
5536@end deftypefn
5537
5538@deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str)
5539Similar to @code{__builtin_nan}, except the return type is @code{long double}.
5540@end deftypefn
5541
5542@deftypefn {Built-in Function} double __builtin_nans (const char *str)
5543Similar to @code{__builtin_nan}, except the significand is forced
5544to be a signaling NaN.  The @code{nans} function is proposed by
5545@uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}.
5546@end deftypefn
5547
5548@deftypefn {Built-in Function} float __builtin_nansf (const char *str)
5549Similar to @code{__builtin_nans}, except the return type is @code{float}.
5550@end deftypefn
5551
5552@deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str)
5553Similar to @code{__builtin_nans}, except the return type is @code{long double}.
5554@end deftypefn
5555
5556@deftypefn {Built-in Function} int __builtin_ffs (unsigned int x)
5557Returns one plus the index of the least significant 1-bit of @var{x}, or
5558if @var{x} is zero, returns zero.
5559@end deftypefn
5560
5561@deftypefn {Built-in Function} int __builtin_clz (unsigned int x)
5562Returns the number of leading 0-bits in @var{x}, starting at the most
5563significant bit position.  If @var{x} is 0, the result is undefined.
5564@end deftypefn
5565
5566@deftypefn {Built-in Function} int __builtin_ctz (unsigned int x)
5567Returns the number of trailing 0-bits in @var{x}, starting at the least
5568significant bit position.  If @var{x} is 0, the result is undefined.
5569@end deftypefn
5570
5571@deftypefn {Built-in Function} int __builtin_popcount (unsigned int x)
5572Returns the number of 1-bits in @var{x}.
5573@end deftypefn
5574
5575@deftypefn {Built-in Function} int __builtin_parity (unsigned int x)
5576Returns the parity of @var{x}, i.@:e. the number of 1-bits in @var{x}
5577modulo 2.
5578@end deftypefn
5579
5580@deftypefn {Built-in Function} int __builtin_ffsl (unsigned long)
5581Similar to @code{__builtin_ffs}, except the argument type is
5582@code{unsigned long}.
5583@end deftypefn
5584
5585@deftypefn {Built-in Function} int __builtin_clzl (unsigned long)
5586Similar to @code{__builtin_clz}, except the argument type is
5587@code{unsigned long}.
5588@end deftypefn
5589
5590@deftypefn {Built-in Function} int __builtin_ctzl (unsigned long)
5591Similar to @code{__builtin_ctz}, except the argument type is
5592@code{unsigned long}.
5593@end deftypefn
5594
5595@deftypefn {Built-in Function} int __builtin_popcountl (unsigned long)
5596Similar to @code{__builtin_popcount}, except the argument type is
5597@code{unsigned long}.
5598@end deftypefn
5599
5600@deftypefn {Built-in Function} int __builtin_parityl (unsigned long)
5601Similar to @code{__builtin_parity}, except the argument type is
5602@code{unsigned long}.
5603@end deftypefn
5604
5605@deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long)
5606Similar to @code{__builtin_ffs}, except the argument type is
5607@code{unsigned long long}.
5608@end deftypefn
5609
5610@deftypefn {Built-in Function} int __builtin_clzll (unsigned long long)
5611Similar to @code{__builtin_clz}, except the argument type is
5612@code{unsigned long long}.
5613@end deftypefn
5614
5615@deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long)
5616Similar to @code{__builtin_ctz}, except the argument type is
5617@code{unsigned long long}.
5618@end deftypefn
5619
5620@deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long)
5621Similar to @code{__builtin_popcount}, except the argument type is
5622@code{unsigned long long}.
5623@end deftypefn
5624
5625@deftypefn {Built-in Function} int __builtin_parityll (unsigned long long)
5626Similar to @code{__builtin_parity}, except the argument type is
5627@code{unsigned long long}.
5628@end deftypefn
5629
5630
5631@node Target Builtins
5632@section Built-in Functions Specific to Particular Target Machines
5633
5634On some target machines, GCC supports many built-in functions specific
5635to those machines.  Generally these generate calls to specific machine
5636instructions, but allow the compiler to schedule those calls.
5637
5638@menu
5639* Alpha Built-in Functions::
5640* ARM Built-in Functions::
5641* X86 Built-in Functions::
5642* PowerPC AltiVec Built-in Functions::
5643@end menu
5644
5645@node Alpha Built-in Functions
5646@subsection Alpha Built-in Functions
5647
5648These built-in functions are available for the Alpha family of
5649processors, depending on the command-line switches used.
5650
5651The following built-in functions are always available.  They
5652all generate the machine instruction that is part of the name.
5653
5654@smallexample
5655long __builtin_alpha_implver (void)
5656long __builtin_alpha_rpcc (void)
5657long __builtin_alpha_amask (long)
5658long __builtin_alpha_cmpbge (long, long)
5659long __builtin_alpha_extbl (long, long)
5660long __builtin_alpha_extwl (long, long)
5661long __builtin_alpha_extll (long, long)
5662long __builtin_alpha_extql (long, long)
5663long __builtin_alpha_extwh (long, long)
5664long __builtin_alpha_extlh (long, long)
5665long __builtin_alpha_extqh (long, long)
5666long __builtin_alpha_insbl (long, long)
5667long __builtin_alpha_inswl (long, long)
5668long __builtin_alpha_insll (long, long)
5669long __builtin_alpha_insql (long, long)
5670long __builtin_alpha_inswh (long, long)
5671long __builtin_alpha_inslh (long, long)
5672long __builtin_alpha_insqh (long, long)
5673long __builtin_alpha_mskbl (long, long)
5674long __builtin_alpha_mskwl (long, long)
5675long __builtin_alpha_mskll (long, long)
5676long __builtin_alpha_mskql (long, long)
5677long __builtin_alpha_mskwh (long, long)
5678long __builtin_alpha_msklh (long, long)
5679long __builtin_alpha_mskqh (long, long)
5680long __builtin_alpha_umulh (long, long)
5681long __builtin_alpha_zap (long, long)
5682long __builtin_alpha_zapnot (long, long)
5683@end smallexample
5684
5685The following built-in functions are always with @option{-mmax}
5686or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or
5687later.  They all generate the machine instruction that is part
5688of the name.
5689
5690@smallexample
5691long __builtin_alpha_pklb (long)
5692long __builtin_alpha_pkwb (long)
5693long __builtin_alpha_unpkbl (long)
5694long __builtin_alpha_unpkbw (long)
5695long __builtin_alpha_minub8 (long, long)
5696long __builtin_alpha_minsb8 (long, long)
5697long __builtin_alpha_minuw4 (long, long)
5698long __builtin_alpha_minsw4 (long, long)
5699long __builtin_alpha_maxub8 (long, long)
5700long __builtin_alpha_maxsb8 (long, long)
5701long __builtin_alpha_maxuw4 (long, long)
5702long __builtin_alpha_maxsw4 (long, long)
5703long __builtin_alpha_perr (long, long)
5704@end smallexample
5705
5706The following built-in functions are always with @option{-mcix}
5707or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or
5708later.  They all generate the machine instruction that is part
5709of the name.
5710
5711@smallexample
5712long __builtin_alpha_cttz (long)
5713long __builtin_alpha_ctlz (long)
5714long __builtin_alpha_ctpop (long)
5715@end smallexample
5716
5717The following builtins are available on systems that use the OSF/1
5718PALcode.  Normally they invoke the @code{rduniq} and @code{wruniq}
5719PAL calls, but when invoked with @option{-mtls-kernel}, they invoke
5720@code{rdval} and @code{wrval}.
5721
5722@smallexample
5723void *__builtin_thread_pointer (void)
5724void __builtin_set_thread_pointer (void *)
5725@end smallexample
5726
5727@node ARM Built-in Functions
5728@subsection ARM Built-in Functions
5729
5730These built-in functions are available for the ARM family of
5731processors, when the @option{-mcpu=iwmmxt} switch is used:
5732
5733@smallexample
5734typedef int v2si __attribute__ ((vector_size (8)));
5735typedef short v4hi __attribute__ ((vector_size (8)));
5736typedef char v8qi __attribute__ ((vector_size (8)));
5737
5738int __builtin_arm_getwcx (int)
5739void __builtin_arm_setwcx (int, int)
5740int __builtin_arm_textrmsb (v8qi, int)
5741int __builtin_arm_textrmsh (v4hi, int)
5742int __builtin_arm_textrmsw (v2si, int)
5743int __builtin_arm_textrmub (v8qi, int)
5744int __builtin_arm_textrmuh (v4hi, int)
5745int __builtin_arm_textrmuw (v2si, int)
5746v8qi __builtin_arm_tinsrb (v8qi, int)
5747v4hi __builtin_arm_tinsrh (v4hi, int)
5748v2si __builtin_arm_tinsrw (v2si, int)
5749long long __builtin_arm_tmia (long long, int, int)
5750long long __builtin_arm_tmiabb (long long, int, int)
5751long long __builtin_arm_tmiabt (long long, int, int)
5752long long __builtin_arm_tmiaph (long long, int, int)
5753long long __builtin_arm_tmiatb (long long, int, int)
5754long long __builtin_arm_tmiatt (long long, int, int)
5755int __builtin_arm_tmovmskb (v8qi)
5756int __builtin_arm_tmovmskh (v4hi)
5757int __builtin_arm_tmovmskw (v2si)
5758long long __builtin_arm_waccb (v8qi)
5759long long __builtin_arm_wacch (v4hi)
5760long long __builtin_arm_waccw (v2si)
5761v8qi __builtin_arm_waddb (v8qi, v8qi)
5762v8qi __builtin_arm_waddbss (v8qi, v8qi)
5763v8qi __builtin_arm_waddbus (v8qi, v8qi)
5764v4hi __builtin_arm_waddh (v4hi, v4hi)
5765v4hi __builtin_arm_waddhss (v4hi, v4hi)
5766v4hi __builtin_arm_waddhus (v4hi, v4hi)
5767v2si __builtin_arm_waddw (v2si, v2si)
5768v2si __builtin_arm_waddwss (v2si, v2si)
5769v2si __builtin_arm_waddwus (v2si, v2si)
5770v8qi __builtin_arm_walign (v8qi, v8qi, int)
5771long long __builtin_arm_wand(long long, long long)
5772long long __builtin_arm_wandn (long long, long long)
5773v8qi __builtin_arm_wavg2b (v8qi, v8qi)
5774v8qi __builtin_arm_wavg2br (v8qi, v8qi)
5775v4hi __builtin_arm_wavg2h (v4hi, v4hi)
5776v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
5777v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
5778v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
5779v2si __builtin_arm_wcmpeqw (v2si, v2si)
5780v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
5781v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
5782v2si __builtin_arm_wcmpgtsw (v2si, v2si)
5783v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
5784v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
5785v2si __builtin_arm_wcmpgtuw (v2si, v2si)
5786long long __builtin_arm_wmacs (long long, v4hi, v4hi)
5787long long __builtin_arm_wmacsz (v4hi, v4hi)
5788long long __builtin_arm_wmacu (long long, v4hi, v4hi)
5789long long __builtin_arm_wmacuz (v4hi, v4hi)
5790v4hi __builtin_arm_wmadds (v4hi, v4hi)
5791v4hi __builtin_arm_wmaddu (v4hi, v4hi)
5792v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
5793v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
5794v2si __builtin_arm_wmaxsw (v2si, v2si)
5795v8qi __builtin_arm_wmaxub (v8qi, v8qi)
5796v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
5797v2si __builtin_arm_wmaxuw (v2si, v2si)
5798v8qi __builtin_arm_wminsb (v8qi, v8qi)
5799v4hi __builtin_arm_wminsh (v4hi, v4hi)
5800v2si __builtin_arm_wminsw (v2si, v2si)
5801v8qi __builtin_arm_wminub (v8qi, v8qi)
5802v4hi __builtin_arm_wminuh (v4hi, v4hi)
5803v2si __builtin_arm_wminuw (v2si, v2si)
5804v4hi __builtin_arm_wmulsm (v4hi, v4hi)
5805v4hi __builtin_arm_wmulul (v4hi, v4hi)
5806v4hi __builtin_arm_wmulum (v4hi, v4hi)
5807long long __builtin_arm_wor (long long, long long)
5808v2si __builtin_arm_wpackdss (long long, long long)
5809v2si __builtin_arm_wpackdus (long long, long long)
5810v8qi __builtin_arm_wpackhss (v4hi, v4hi)
5811v8qi __builtin_arm_wpackhus (v4hi, v4hi)
5812v4hi __builtin_arm_wpackwss (v2si, v2si)
5813v4hi __builtin_arm_wpackwus (v2si, v2si)
5814long long __builtin_arm_wrord (long long, long long)
5815long long __builtin_arm_wrordi (long long, int)
5816v4hi __builtin_arm_wrorh (v4hi, long long)
5817v4hi __builtin_arm_wrorhi (v4hi, int)
5818v2si __builtin_arm_wrorw (v2si, long long)
5819v2si __builtin_arm_wrorwi (v2si, int)
5820v2si __builtin_arm_wsadb (v8qi, v8qi)
5821v2si __builtin_arm_wsadbz (v8qi, v8qi)
5822v2si __builtin_arm_wsadh (v4hi, v4hi)
5823v2si __builtin_arm_wsadhz (v4hi, v4hi)
5824v4hi __builtin_arm_wshufh (v4hi, int)
5825long long __builtin_arm_wslld (long long, long long)
5826long long __builtin_arm_wslldi (long long, int)
5827v4hi __builtin_arm_wsllh (v4hi, long long)
5828v4hi __builtin_arm_wsllhi (v4hi, int)
5829v2si __builtin_arm_wsllw (v2si, long long)
5830v2si __builtin_arm_wsllwi (v2si, int)
5831long long __builtin_arm_wsrad (long long, long long)
5832long long __builtin_arm_wsradi (long long, int)
5833v4hi __builtin_arm_wsrah (v4hi, long long)
5834v4hi __builtin_arm_wsrahi (v4hi, int)
5835v2si __builtin_arm_wsraw (v2si, long long)
5836v2si __builtin_arm_wsrawi (v2si, int)
5837long long __builtin_arm_wsrld (long long, long long)
5838long long __builtin_arm_wsrldi (long long, int)
5839v4hi __builtin_arm_wsrlh (v4hi, long long)
5840v4hi __builtin_arm_wsrlhi (v4hi, int)
5841v2si __builtin_arm_wsrlw (v2si, long long)
5842v2si __builtin_arm_wsrlwi (v2si, int)
5843v8qi __builtin_arm_wsubb (v8qi, v8qi)
5844v8qi __builtin_arm_wsubbss (v8qi, v8qi)
5845v8qi __builtin_arm_wsubbus (v8qi, v8qi)
5846v4hi __builtin_arm_wsubh (v4hi, v4hi)
5847v4hi __builtin_arm_wsubhss (v4hi, v4hi)
5848v4hi __builtin_arm_wsubhus (v4hi, v4hi)
5849v2si __builtin_arm_wsubw (v2si, v2si)
5850v2si __builtin_arm_wsubwss (v2si, v2si)
5851v2si __builtin_arm_wsubwus (v2si, v2si)
5852v4hi __builtin_arm_wunpckehsb (v8qi)
5853v2si __builtin_arm_wunpckehsh (v4hi)
5854long long __builtin_arm_wunpckehsw (v2si)
5855v4hi __builtin_arm_wunpckehub (v8qi)
5856v2si __builtin_arm_wunpckehuh (v4hi)
5857long long __builtin_arm_wunpckehuw (v2si)
5858v4hi __builtin_arm_wunpckelsb (v8qi)
5859v2si __builtin_arm_wunpckelsh (v4hi)
5860long long __builtin_arm_wunpckelsw (v2si)
5861v4hi __builtin_arm_wunpckelub (v8qi)
5862v2si __builtin_arm_wunpckeluh (v4hi)
5863long long __builtin_arm_wunpckeluw (v2si)
5864v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
5865v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
5866v2si __builtin_arm_wunpckihw (v2si, v2si)
5867v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
5868v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
5869v2si __builtin_arm_wunpckilw (v2si, v2si)
5870long long __builtin_arm_wxor (long long, long long)
5871long long __builtin_arm_wzero ()
5872@end smallexample
5873
5874@node X86 Built-in Functions
5875@subsection X86 Built-in Functions
5876
5877These built-in functions are available for the i386 and x86-64 family
5878of computers, depending on the command-line switches used.
5879
5880The following machine modes are available for use with MMX built-in functions
5881(@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers,
5882@code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a
5883vector of eight 8-bit integers.  Some of the built-in functions operate on
5884MMX registers as a whole 64-bit entity, these use @code{DI} as their mode.
5885
5886If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector
5887of two 32-bit floating point values.
5888
5889If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit
5890floating point values.  Some instructions use a vector of four 32-bit
5891integers, these use @code{V4SI}.  Finally, some instructions operate on an
5892entire vector register, interpreting it as a 128-bit integer, these use mode
5893@code{TI}.
5894
5895The following built-in functions are made available by @option{-mmmx}.
5896All of them generate the machine instruction that is part of the name.
5897
5898@smallexample
5899v8qi __builtin_ia32_paddb (v8qi, v8qi)
5900v4hi __builtin_ia32_paddw (v4hi, v4hi)
5901v2si __builtin_ia32_paddd (v2si, v2si)
5902v8qi __builtin_ia32_psubb (v8qi, v8qi)
5903v4hi __builtin_ia32_psubw (v4hi, v4hi)
5904v2si __builtin_ia32_psubd (v2si, v2si)
5905v8qi __builtin_ia32_paddsb (v8qi, v8qi)
5906v4hi __builtin_ia32_paddsw (v4hi, v4hi)
5907v8qi __builtin_ia32_psubsb (v8qi, v8qi)
5908v4hi __builtin_ia32_psubsw (v4hi, v4hi)
5909v8qi __builtin_ia32_paddusb (v8qi, v8qi)
5910v4hi __builtin_ia32_paddusw (v4hi, v4hi)
5911v8qi __builtin_ia32_psubusb (v8qi, v8qi)
5912v4hi __builtin_ia32_psubusw (v4hi, v4hi)
5913v4hi __builtin_ia32_pmullw (v4hi, v4hi)
5914v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
5915di __builtin_ia32_pand (di, di)
5916di __builtin_ia32_pandn (di,di)
5917di __builtin_ia32_por (di, di)
5918di __builtin_ia32_pxor (di, di)
5919v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
5920v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
5921v2si __builtin_ia32_pcmpeqd (v2si, v2si)
5922v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
5923v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
5924v2si __builtin_ia32_pcmpgtd (v2si, v2si)
5925v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
5926v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
5927v2si __builtin_ia32_punpckhdq (v2si, v2si)
5928v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
5929v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
5930v2si __builtin_ia32_punpckldq (v2si, v2si)
5931v8qi __builtin_ia32_packsswb (v4hi, v4hi)
5932v4hi __builtin_ia32_packssdw (v2si, v2si)
5933v8qi __builtin_ia32_packuswb (v4hi, v4hi)
5934@end smallexample
5935
5936The following built-in functions are made available either with
5937@option{-msse}, or with a combination of @option{-m3dnow} and
5938@option{-march=athlon}.  All of them generate the machine
5939instruction that is part of the name.
5940
5941@smallexample
5942v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
5943v8qi __builtin_ia32_pavgb (v8qi, v8qi)
5944v4hi __builtin_ia32_pavgw (v4hi, v4hi)
5945v4hi __builtin_ia32_psadbw (v8qi, v8qi)
5946v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
5947v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
5948v8qi __builtin_ia32_pminub (v8qi, v8qi)
5949v4hi __builtin_ia32_pminsw (v4hi, v4hi)
5950int __builtin_ia32_pextrw (v4hi, int)
5951v4hi __builtin_ia32_pinsrw (v4hi, int, int)
5952int __builtin_ia32_pmovmskb (v8qi)
5953void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
5954void __builtin_ia32_movntq (di *, di)
5955void __builtin_ia32_sfence (void)
5956@end smallexample
5957
5958The following built-in functions are available when @option{-msse} is used.
5959All of them generate the machine instruction that is part of the name.
5960
5961@smallexample
5962int __builtin_ia32_comieq (v4sf, v4sf)
5963int __builtin_ia32_comineq (v4sf, v4sf)
5964int __builtin_ia32_comilt (v4sf, v4sf)
5965int __builtin_ia32_comile (v4sf, v4sf)
5966int __builtin_ia32_comigt (v4sf, v4sf)
5967int __builtin_ia32_comige (v4sf, v4sf)
5968int __builtin_ia32_ucomieq (v4sf, v4sf)
5969int __builtin_ia32_ucomineq (v4sf, v4sf)
5970int __builtin_ia32_ucomilt (v4sf, v4sf)
5971int __builtin_ia32_ucomile (v4sf, v4sf)
5972int __builtin_ia32_ucomigt (v4sf, v4sf)
5973int __builtin_ia32_ucomige (v4sf, v4sf)
5974v4sf __builtin_ia32_addps (v4sf, v4sf)
5975v4sf __builtin_ia32_subps (v4sf, v4sf)
5976v4sf __builtin_ia32_mulps (v4sf, v4sf)
5977v4sf __builtin_ia32_divps (v4sf, v4sf)
5978v4sf __builtin_ia32_addss (v4sf, v4sf)
5979v4sf __builtin_ia32_subss (v4sf, v4sf)
5980v4sf __builtin_ia32_mulss (v4sf, v4sf)
5981v4sf __builtin_ia32_divss (v4sf, v4sf)
5982v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
5983v4si __builtin_ia32_cmpltps (v4sf, v4sf)
5984v4si __builtin_ia32_cmpleps (v4sf, v4sf)
5985v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
5986v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
5987v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
5988v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
5989v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
5990v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
5991v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
5992v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
5993v4si __builtin_ia32_cmpordps (v4sf, v4sf)
5994v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
5995v4si __builtin_ia32_cmpltss (v4sf, v4sf)
5996v4si __builtin_ia32_cmpless (v4sf, v4sf)
5997v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
5998v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
5999v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
6000v4si __builtin_ia32_cmpnless (v4sf, v4sf)
6001v4si __builtin_ia32_cmpordss (v4sf, v4sf)
6002v4sf __builtin_ia32_maxps (v4sf, v4sf)
6003v4sf __builtin_ia32_maxss (v4sf, v4sf)
6004v4sf __builtin_ia32_minps (v4sf, v4sf)
6005v4sf __builtin_ia32_minss (v4sf, v4sf)
6006v4sf __builtin_ia32_andps (v4sf, v4sf)
6007v4sf __builtin_ia32_andnps (v4sf, v4sf)
6008v4sf __builtin_ia32_orps (v4sf, v4sf)
6009v4sf __builtin_ia32_xorps (v4sf, v4sf)
6010v4sf __builtin_ia32_movss (v4sf, v4sf)
6011v4sf __builtin_ia32_movhlps (v4sf, v4sf)
6012v4sf __builtin_ia32_movlhps (v4sf, v4sf)
6013v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
6014v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
6015v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
6016v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
6017v2si __builtin_ia32_cvtps2pi (v4sf)
6018int __builtin_ia32_cvtss2si (v4sf)
6019v2si __builtin_ia32_cvttps2pi (v4sf)
6020int __builtin_ia32_cvttss2si (v4sf)
6021v4sf __builtin_ia32_rcpps (v4sf)
6022v4sf __builtin_ia32_rsqrtps (v4sf)
6023v4sf __builtin_ia32_sqrtps (v4sf)
6024v4sf __builtin_ia32_rcpss (v4sf)
6025v4sf __builtin_ia32_rsqrtss (v4sf)
6026v4sf __builtin_ia32_sqrtss (v4sf)
6027v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
6028void __builtin_ia32_movntps (float *, v4sf)
6029int __builtin_ia32_movmskps (v4sf)
6030@end smallexample
6031
6032The following built-in functions are available when @option{-msse} is used.
6033
6034@table @code
6035@item v4sf __builtin_ia32_loadaps (float *)
6036Generates the @code{movaps} machine instruction as a load from memory.
6037@item void __builtin_ia32_storeaps (float *, v4sf)
6038Generates the @code{movaps} machine instruction as a store to memory.
6039@item v4sf __builtin_ia32_loadups (float *)
6040Generates the @code{movups} machine instruction as a load from memory.
6041@item void __builtin_ia32_storeups (float *, v4sf)
6042Generates the @code{movups} machine instruction as a store to memory.
6043@item v4sf __builtin_ia32_loadsss (float *)
6044Generates the @code{movss} machine instruction as a load from memory.
6045@item void __builtin_ia32_storess (float *, v4sf)
6046Generates the @code{movss} machine instruction as a store to memory.
6047@item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
6048Generates the @code{movhps} machine instruction as a load from memory.
6049@item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
6050Generates the @code{movlps} machine instruction as a load from memory
6051@item void __builtin_ia32_storehps (v4sf, v2si *)
6052Generates the @code{movhps} machine instruction as a store to memory.
6053@item void __builtin_ia32_storelps (v4sf, v2si *)
6054Generates the @code{movlps} machine instruction as a store to memory.
6055@end table
6056
6057The following built-in functions are available when @option{-msse3} is used.
6058All of them generate the machine instruction that is part of the name.
6059
6060@smallexample
6061v2df __builtin_ia32_addsubpd (v2df, v2df)
6062v2df __builtin_ia32_addsubps (v2df, v2df)
6063v2df __builtin_ia32_haddpd (v2df, v2df)
6064v2df __builtin_ia32_haddps (v2df, v2df)
6065v2df __builtin_ia32_hsubpd (v2df, v2df)
6066v2df __builtin_ia32_hsubps (v2df, v2df)
6067v16qi __builtin_ia32_lddqu (char const *)
6068void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
6069v2df __builtin_ia32_movddup (v2df)
6070v4sf __builtin_ia32_movshdup (v4sf)
6071v4sf __builtin_ia32_movsldup (v4sf)
6072void __builtin_ia32_mwait (unsigned int, unsigned int)
6073@end smallexample
6074
6075The following built-in functions are available when @option{-msse3} is used.
6076
6077@table @code
6078@item v2df __builtin_ia32_loadddup (double const *)
6079Generates the @code{movddup} machine instruction as a load from memory.
6080@end table
6081
6082The following built-in functions are available when @option{-m3dnow} is used.
6083All of them generate the machine instruction that is part of the name.
6084
6085@smallexample
6086void __builtin_ia32_femms (void)
6087v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
6088v2si __builtin_ia32_pf2id (v2sf)
6089v2sf __builtin_ia32_pfacc (v2sf, v2sf)
6090v2sf __builtin_ia32_pfadd (v2sf, v2sf)
6091v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
6092v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
6093v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
6094v2sf __builtin_ia32_pfmax (v2sf, v2sf)
6095v2sf __builtin_ia32_pfmin (v2sf, v2sf)
6096v2sf __builtin_ia32_pfmul (v2sf, v2sf)
6097v2sf __builtin_ia32_pfrcp (v2sf)
6098v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
6099v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
6100v2sf __builtin_ia32_pfrsqrt (v2sf)
6101v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
6102v2sf __builtin_ia32_pfsub (v2sf, v2sf)
6103v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
6104v2sf __builtin_ia32_pi2fd (v2si)
6105v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
6106@end smallexample
6107
6108The following built-in functions are available when both @option{-m3dnow}
6109and @option{-march=athlon} are used.  All of them generate the machine
6110instruction that is part of the name.
6111
6112@smallexample
6113v2si __builtin_ia32_pf2iw (v2sf)
6114v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
6115v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
6116v2sf __builtin_ia32_pi2fw (v2si)
6117v2sf __builtin_ia32_pswapdsf (v2sf)
6118v2si __builtin_ia32_pswapdsi (v2si)
6119@end smallexample
6120
6121@node PowerPC AltiVec Built-in Functions
6122@subsection PowerPC AltiVec Built-in Functions
6123
6124GCC provides an interface for the PowerPC family of processors to access
6125the AltiVec operations described in Motorola's AltiVec Programming
6126Interface Manual.  The interface is made available by including
6127@code{<altivec.h>} and using @option{-maltivec} and
6128@option{-mabi=altivec}.  The interface supports the following vector
6129types.
6130
6131@smallexample
6132vector unsigned char
6133vector signed char
6134vector bool char
6135
6136vector unsigned short
6137vector signed short
6138vector bool short
6139vector pixel
6140
6141vector unsigned int
6142vector signed int
6143vector bool int
6144vector float
6145@end smallexample
6146
6147GCC's implementation of the high-level language interface available from
6148C and C++ code differs from Motorola's documentation in several ways.
6149
6150@itemize @bullet
6151
6152@item
6153A vector constant is a list of constant expressions within curly braces.
6154
6155@item
6156A vector initializer requires no cast if the vector constant is of the
6157same type as the variable it is initializing.
6158
6159@item
6160If @code{signed} or @code{unsigned} is omitted, the vector type defaults
6161to @code{signed} for @code{vector int} or @code{vector short} and to
6162@code{unsigned} for @code{vector char}.
6163
6164@item
6165Compiling with @option{-maltivec} adds keywords @code{__vector},
6166@code{__pixel}, and @code{__bool}.  Macros @option{vector},
6167@code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can
6168be undefined.
6169
6170@item
6171GCC allows using a @code{typedef} name as the type specifier for a
6172vector type.
6173
6174@item
6175For C, overloaded functions are implemented with macros so the following
6176does not work:
6177
6178@smallexample
6179  vec_add ((vector signed int)@{1, 2, 3, 4@}, foo);
6180@end smallexample
6181
6182Since @code{vec_add} is a macro, the vector constant in the example
6183is treated as four separate arguments.  Wrap the entire argument in
6184parentheses for this to work.
6185@end itemize
6186
6187@emph{Note:} Only the @code{<altivec.h>} interface is supported.
6188Internally, GCC uses built-in functions to achieve the functionality in
6189the aforementioned header file, but they are not supported and are
6190subject to change without notice.
6191
6192The following interfaces are supported for the generic and specific
6193AltiVec operations and the AltiVec predicates.  In cases where there
6194is a direct mapping between generic and specific operations, only the
6195generic names are shown here, although the specific operations can also
6196be used.
6197
6198Arguments that are documented as @code{const int} require literal
6199integral values within the range required for that operation.
6200
6201@smallexample
6202vector signed char vec_abs (vector signed char);
6203vector signed short vec_abs (vector signed short);
6204vector signed int vec_abs (vector signed int);
6205vector float vec_abs (vector float);
6206
6207vector signed char vec_abss (vector signed char);
6208vector signed short vec_abss (vector signed short);
6209vector signed int vec_abss (vector signed int);
6210
6211vector signed char vec_add (vector bool char, vector signed char);
6212vector signed char vec_add (vector signed char, vector bool char);
6213vector signed char vec_add (vector signed char, vector signed char);
6214vector unsigned char vec_add (vector bool char, vector unsigned char);
6215vector unsigned char vec_add (vector unsigned char, vector bool char);
6216vector unsigned char vec_add (vector unsigned char,
6217                              vector unsigned char);
6218vector signed short vec_add (vector bool short, vector signed short);
6219vector signed short vec_add (vector signed short, vector bool short);
6220vector signed short vec_add (vector signed short, vector signed short);
6221vector unsigned short vec_add (vector bool short,
6222                               vector unsigned short);
6223vector unsigned short vec_add (vector unsigned short,
6224                               vector bool short);
6225vector unsigned short vec_add (vector unsigned short,
6226                               vector unsigned short);
6227vector signed int vec_add (vector bool int, vector signed int);
6228vector signed int vec_add (vector signed int, vector bool int);
6229vector signed int vec_add (vector signed int, vector signed int);
6230vector unsigned int vec_add (vector bool int, vector unsigned int);
6231vector unsigned int vec_add (vector unsigned int, vector bool int);
6232vector unsigned int vec_add (vector unsigned int, vector unsigned int);
6233vector float vec_add (vector float, vector float);
6234
6235vector float vec_vaddfp (vector float, vector float);
6236
6237vector signed int vec_vadduwm (vector bool int, vector signed int);
6238vector signed int vec_vadduwm (vector signed int, vector bool int);
6239vector signed int vec_vadduwm (vector signed int, vector signed int);
6240vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
6241vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
6242vector unsigned int vec_vadduwm (vector unsigned int,
6243                                 vector unsigned int);
6244
6245vector signed short vec_vadduhm (vector bool short,
6246                                 vector signed short);
6247vector signed short vec_vadduhm (vector signed short,
6248                                 vector bool short);
6249vector signed short vec_vadduhm (vector signed short,
6250                                 vector signed short);
6251vector unsigned short vec_vadduhm (vector bool short,
6252                                   vector unsigned short);
6253vector unsigned short vec_vadduhm (vector unsigned short,
6254                                   vector bool short);
6255vector unsigned short vec_vadduhm (vector unsigned short,
6256                                   vector unsigned short);
6257
6258vector signed char vec_vaddubm (vector bool char, vector signed char);
6259vector signed char vec_vaddubm (vector signed char, vector bool char);
6260vector signed char vec_vaddubm (vector signed char, vector signed char);
6261vector unsigned char vec_vaddubm (vector bool char,
6262                                  vector unsigned char);
6263vector unsigned char vec_vaddubm (vector unsigned char,
6264                                  vector bool char);
6265vector unsigned char vec_vaddubm (vector unsigned char,
6266                                  vector unsigned char);
6267
6268vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
6269
6270vector unsigned char vec_adds (vector bool char, vector unsigned char);
6271vector unsigned char vec_adds (vector unsigned char, vector bool char);
6272vector unsigned char vec_adds (vector unsigned char,
6273                               vector unsigned char);
6274vector signed char vec_adds (vector bool char, vector signed char);
6275vector signed char vec_adds (vector signed char, vector bool char);
6276vector signed char vec_adds (vector signed char, vector signed char);
6277vector unsigned short vec_adds (vector bool short,
6278                                vector unsigned short);
6279vector unsigned short vec_adds (vector unsigned short,
6280                                vector bool short);
6281vector unsigned short vec_adds (vector unsigned short,
6282                                vector unsigned short);
6283vector signed short vec_adds (vector bool short, vector signed short);
6284vector signed short vec_adds (vector signed short, vector bool short);
6285vector signed short vec_adds (vector signed short, vector signed short);
6286vector unsigned int vec_adds (vector bool int, vector unsigned int);
6287vector unsigned int vec_adds (vector unsigned int, vector bool int);
6288vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
6289vector signed int vec_adds (vector bool int, vector signed int);
6290vector signed int vec_adds (vector signed int, vector bool int);
6291vector signed int vec_adds (vector signed int, vector signed int);
6292
6293vector signed int vec_vaddsws (vector bool int, vector signed int);
6294vector signed int vec_vaddsws (vector signed int, vector bool int);
6295vector signed int vec_vaddsws (vector signed int, vector signed int);
6296
6297vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
6298vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
6299vector unsigned int vec_vadduws (vector unsigned int,
6300                                 vector unsigned int);
6301
6302vector signed short vec_vaddshs (vector bool short,
6303                                 vector signed short);
6304vector signed short vec_vaddshs (vector signed short,
6305                                 vector bool short);
6306vector signed short vec_vaddshs (vector signed short,
6307                                 vector signed short);
6308
6309vector unsigned short vec_vadduhs (vector bool short,
6310                                   vector unsigned short);
6311vector unsigned short vec_vadduhs (vector unsigned short,
6312                                   vector bool short);
6313vector unsigned short vec_vadduhs (vector unsigned short,
6314                                   vector unsigned short);
6315
6316vector signed char vec_vaddsbs (vector bool char, vector signed char);
6317vector signed char vec_vaddsbs (vector signed char, vector bool char);
6318vector signed char vec_vaddsbs (vector signed char, vector signed char);
6319
6320vector unsigned char vec_vaddubs (vector bool char,
6321                                  vector unsigned char);
6322vector unsigned char vec_vaddubs (vector unsigned char,
6323                                  vector bool char);
6324vector unsigned char vec_vaddubs (vector unsigned char,
6325                                  vector unsigned char);
6326
6327vector float vec_and (vector float, vector float);
6328vector float vec_and (vector float, vector bool int);
6329vector float vec_and (vector bool int, vector float);
6330vector bool int vec_and (vector bool int, vector bool int);
6331vector signed int vec_and (vector bool int, vector signed int);
6332vector signed int vec_and (vector signed int, vector bool int);
6333vector signed int vec_and (vector signed int, vector signed int);
6334vector unsigned int vec_and (vector bool int, vector unsigned int);
6335vector unsigned int vec_and (vector unsigned int, vector bool int);
6336vector unsigned int vec_and (vector unsigned int, vector unsigned int);
6337vector bool short vec_and (vector bool short, vector bool short);
6338vector signed short vec_and (vector bool short, vector signed short);
6339vector signed short vec_and (vector signed short, vector bool short);
6340vector signed short vec_and (vector signed short, vector signed short);
6341vector unsigned short vec_and (vector bool short,
6342                               vector unsigned short);
6343vector unsigned short vec_and (vector unsigned short,
6344                               vector bool short);
6345vector unsigned short vec_and (vector unsigned short,
6346                               vector unsigned short);
6347vector signed char vec_and (vector bool char, vector signed char);
6348vector bool char vec_and (vector bool char, vector bool char);
6349vector signed char vec_and (vector signed char, vector bool char);
6350vector signed char vec_and (vector signed char, vector signed char);
6351vector unsigned char vec_and (vector bool char, vector unsigned char);
6352vector unsigned char vec_and (vector unsigned char, vector bool char);
6353vector unsigned char vec_and (vector unsigned char,
6354                              vector unsigned char);
6355
6356vector float vec_andc (vector float, vector float);
6357vector float vec_andc (vector float, vector bool int);
6358vector float vec_andc (vector bool int, vector float);
6359vector bool int vec_andc (vector bool int, vector bool int);
6360vector signed int vec_andc (vector bool int, vector signed int);
6361vector signed int vec_andc (vector signed int, vector bool int);
6362vector signed int vec_andc (vector signed int, vector signed int);
6363vector unsigned int vec_andc (vector bool int, vector unsigned int);
6364vector unsigned int vec_andc (vector unsigned int, vector bool int);
6365vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
6366vector bool short vec_andc (vector bool short, vector bool short);
6367vector signed short vec_andc (vector bool short, vector signed short);
6368vector signed short vec_andc (vector signed short, vector bool short);
6369vector signed short vec_andc (vector signed short, vector signed short);
6370vector unsigned short vec_andc (vector bool short,
6371                                vector unsigned short);
6372vector unsigned short vec_andc (vector unsigned short,
6373                                vector bool short);
6374vector unsigned short vec_andc (vector unsigned short,
6375                                vector unsigned short);
6376vector signed char vec_andc (vector bool char, vector signed char);
6377vector bool char vec_andc (vector bool char, vector bool char);
6378vector signed char vec_andc (vector signed char, vector bool char);
6379vector signed char vec_andc (vector signed char, vector signed char);
6380vector unsigned char vec_andc (vector bool char, vector unsigned char);
6381vector unsigned char vec_andc (vector unsigned char, vector bool char);
6382vector unsigned char vec_andc (vector unsigned char,
6383                               vector unsigned char);
6384
6385vector unsigned char vec_avg (vector unsigned char,
6386                              vector unsigned char);
6387vector signed char vec_avg (vector signed char, vector signed char);
6388vector unsigned short vec_avg (vector unsigned short,
6389                               vector unsigned short);
6390vector signed short vec_avg (vector signed short, vector signed short);
6391vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
6392vector signed int vec_avg (vector signed int, vector signed int);
6393
6394vector signed int vec_vavgsw (vector signed int, vector signed int);
6395
6396vector unsigned int vec_vavguw (vector unsigned int,
6397                                vector unsigned int);
6398
6399vector signed short vec_vavgsh (vector signed short,
6400                                vector signed short);
6401
6402vector unsigned short vec_vavguh (vector unsigned short,
6403                                  vector unsigned short);
6404
6405vector signed char vec_vavgsb (vector signed char, vector signed char);
6406
6407vector unsigned char vec_vavgub (vector unsigned char,
6408                                 vector unsigned char);
6409
6410vector float vec_ceil (vector float);
6411
6412vector signed int vec_cmpb (vector float, vector float);
6413
6414vector bool char vec_cmpeq (vector signed char, vector signed char);
6415vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
6416vector bool short vec_cmpeq (vector signed short, vector signed short);
6417vector bool short vec_cmpeq (vector unsigned short,
6418                             vector unsigned short);
6419vector bool int vec_cmpeq (vector signed int, vector signed int);
6420vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
6421vector bool int vec_cmpeq (vector float, vector float);
6422
6423vector bool int vec_vcmpeqfp (vector float, vector float);
6424
6425vector bool int vec_vcmpequw (vector signed int, vector signed int);
6426vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
6427
6428vector bool short vec_vcmpequh (vector signed short,
6429                                vector signed short);
6430vector bool short vec_vcmpequh (vector unsigned short,
6431                                vector unsigned short);
6432
6433vector bool char vec_vcmpequb (vector signed char, vector signed char);
6434vector bool char vec_vcmpequb (vector unsigned char,
6435                               vector unsigned char);
6436
6437vector bool int vec_cmpge (vector float, vector float);
6438
6439vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
6440vector bool char vec_cmpgt (vector signed char, vector signed char);
6441vector bool short vec_cmpgt (vector unsigned short,
6442                             vector unsigned short);
6443vector bool short vec_cmpgt (vector signed short, vector signed short);
6444vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
6445vector bool int vec_cmpgt (vector signed int, vector signed int);
6446vector bool int vec_cmpgt (vector float, vector float);
6447
6448vector bool int vec_vcmpgtfp (vector float, vector float);
6449
6450vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
6451
6452vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
6453
6454vector bool short vec_vcmpgtsh (vector signed short,
6455                                vector signed short);
6456
6457vector bool short vec_vcmpgtuh (vector unsigned short,
6458                                vector unsigned short);
6459
6460vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
6461
6462vector bool char vec_vcmpgtub (vector unsigned char,
6463                               vector unsigned char);
6464
6465vector bool int vec_cmple (vector float, vector float);
6466
6467vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
6468vector bool char vec_cmplt (vector signed char, vector signed char);
6469vector bool short vec_cmplt (vector unsigned short,
6470                             vector unsigned short);
6471vector bool short vec_cmplt (vector signed short, vector signed short);
6472vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
6473vector bool int vec_cmplt (vector signed int, vector signed int);
6474vector bool int vec_cmplt (vector float, vector float);
6475
6476vector float vec_ctf (vector unsigned int, const int);
6477vector float vec_ctf (vector signed int, const int);
6478
6479vector float vec_vcfsx (vector signed int, const int);
6480
6481vector float vec_vcfux (vector unsigned int, const int);
6482
6483vector signed int vec_cts (vector float, const int);
6484
6485vector unsigned int vec_ctu (vector float, const int);
6486
6487void vec_dss (const int);
6488
6489void vec_dssall (void);
6490
6491void vec_dst (const vector unsigned char *, int, const int);
6492void vec_dst (const vector signed char *, int, const int);
6493void vec_dst (const vector bool char *, int, const int);
6494void vec_dst (const vector unsigned short *, int, const int);
6495void vec_dst (const vector signed short *, int, const int);
6496void vec_dst (const vector bool short *, int, const int);
6497void vec_dst (const vector pixel *, int, const int);
6498void vec_dst (const vector unsigned int *, int, const int);
6499void vec_dst (const vector signed int *, int, const int);
6500void vec_dst (const vector bool int *, int, const int);
6501void vec_dst (const vector float *, int, const int);
6502void vec_dst (const unsigned char *, int, const int);
6503void vec_dst (const signed char *, int, const int);
6504void vec_dst (const unsigned short *, int, const int);
6505void vec_dst (const short *, int, const int);
6506void vec_dst (const unsigned int *, int, const int);
6507void vec_dst (const int *, int, const int);
6508void vec_dst (const unsigned long *, int, const int);
6509void vec_dst (const long *, int, const int);
6510void vec_dst (const float *, int, const int);
6511
6512void vec_dstst (const vector unsigned char *, int, const int);
6513void vec_dstst (const vector signed char *, int, const int);
6514void vec_dstst (const vector bool char *, int, const int);
6515void vec_dstst (const vector unsigned short *, int, const int);
6516void vec_dstst (const vector signed short *, int, const int);
6517void vec_dstst (const vector bool short *, int, const int);
6518void vec_dstst (const vector pixel *, int, const int);
6519void vec_dstst (const vector unsigned int *, int, const int);
6520void vec_dstst (const vector signed int *, int, const int);
6521void vec_dstst (const vector bool int *, int, const int);
6522void vec_dstst (const vector float *, int, const int);
6523void vec_dstst (const unsigned char *, int, const int);
6524void vec_dstst (const signed char *, int, const int);
6525void vec_dstst (const unsigned short *, int, const int);
6526void vec_dstst (const short *, int, const int);
6527void vec_dstst (const unsigned int *, int, const int);
6528void vec_dstst (const int *, int, const int);
6529void vec_dstst (const unsigned long *, int, const int);
6530void vec_dstst (const long *, int, const int);
6531void vec_dstst (const float *, int, const int);
6532
6533void vec_dststt (const vector unsigned char *, int, const int);
6534void vec_dststt (const vector signed char *, int, const int);
6535void vec_dststt (const vector bool char *, int, const int);
6536void vec_dststt (const vector unsigned short *, int, const int);
6537void vec_dststt (const vector signed short *, int, const int);
6538void vec_dststt (const vector bool short *, int, const int);
6539void vec_dststt (const vector pixel *, int, const int);
6540void vec_dststt (const vector unsigned int *, int, const int);
6541void vec_dststt (const vector signed int *, int, const int);
6542void vec_dststt (const vector bool int *, int, const int);
6543void vec_dststt (const vector float *, int, const int);
6544void vec_dststt (const unsigned char *, int, const int);
6545void vec_dststt (const signed char *, int, const int);
6546void vec_dststt (const unsigned short *, int, const int);
6547void vec_dststt (const short *, int, const int);
6548void vec_dststt (const unsigned int *, int, const int);
6549void vec_dststt (const int *, int, const int);
6550void vec_dststt (const unsigned long *, int, const int);
6551void vec_dststt (const long *, int, const int);
6552void vec_dststt (const float *, int, const int);
6553
6554void vec_dstt (const vector unsigned char *, int, const int);
6555void vec_dstt (const vector signed char *, int, const int);
6556void vec_dstt (const vector bool char *, int, const int);
6557void vec_dstt (const vector unsigned short *, int, const int);
6558void vec_dstt (const vector signed short *, int, const int);
6559void vec_dstt (const vector bool short *, int, const int);
6560void vec_dstt (const vector pixel *, int, const int);
6561void vec_dstt (const vector unsigned int *, int, const int);
6562void vec_dstt (const vector signed int *, int, const int);
6563void vec_dstt (const vector bool int *, int, const int);
6564void vec_dstt (const vector float *, int, const int);
6565void vec_dstt (const unsigned char *, int, const int);
6566void vec_dstt (const signed char *, int, const int);
6567void vec_dstt (const unsigned short *, int, const int);
6568void vec_dstt (const short *, int, const int);
6569void vec_dstt (const unsigned int *, int, const int);
6570void vec_dstt (const int *, int, const int);
6571void vec_dstt (const unsigned long *, int, const int);
6572void vec_dstt (const long *, int, const int);
6573void vec_dstt (const float *, int, const int);
6574
6575vector float vec_expte (vector float);
6576
6577vector float vec_floor (vector float);
6578
6579vector float vec_ld (int, const vector float *);
6580vector float vec_ld (int, const float *);
6581vector bool int vec_ld (int, const vector bool int *);
6582vector signed int vec_ld (int, const vector signed int *);
6583vector signed int vec_ld (int, const int *);
6584vector signed int vec_ld (int, const long *);
6585vector unsigned int vec_ld (int, const vector unsigned int *);
6586vector unsigned int vec_ld (int, const unsigned int *);
6587vector unsigned int vec_ld (int, const unsigned long *);
6588vector bool short vec_ld (int, const vector bool short *);
6589vector pixel vec_ld (int, const vector pixel *);
6590vector signed short vec_ld (int, const vector signed short *);
6591vector signed short vec_ld (int, const short *);
6592vector unsigned short vec_ld (int, const vector unsigned short *);
6593vector unsigned short vec_ld (int, const unsigned short *);
6594vector bool char vec_ld (int, const vector bool char *);
6595vector signed char vec_ld (int, const vector signed char *);
6596vector signed char vec_ld (int, const signed char *);
6597vector unsigned char vec_ld (int, const vector unsigned char *);
6598vector unsigned char vec_ld (int, const unsigned char *);
6599
6600vector signed char vec_lde (int, const signed char *);
6601vector unsigned char vec_lde (int, const unsigned char *);
6602vector signed short vec_lde (int, const short *);
6603vector unsigned short vec_lde (int, const unsigned short *);
6604vector float vec_lde (int, const float *);
6605vector signed int vec_lde (int, const int *);
6606vector unsigned int vec_lde (int, const unsigned int *);
6607vector signed int vec_lde (int, const long *);
6608vector unsigned int vec_lde (int, const unsigned long *);
6609
6610vector float vec_lvewx (int, float *);
6611vector signed int vec_lvewx (int, int *);
6612vector unsigned int vec_lvewx (int, unsigned int *);
6613vector signed int vec_lvewx (int, long *);
6614vector unsigned int vec_lvewx (int, unsigned long *);
6615
6616vector signed short vec_lvehx (int, short *);
6617vector unsigned short vec_lvehx (int, unsigned short *);
6618
6619vector signed char vec_lvebx (int, char *);
6620vector unsigned char vec_lvebx (int, unsigned char *);
6621
6622vector float vec_ldl (int, const vector float *);
6623vector float vec_ldl (int, const float *);
6624vector bool int vec_ldl (int, const vector bool int *);
6625vector signed int vec_ldl (int, const vector signed int *);
6626vector signed int vec_ldl (int, const int *);
6627vector signed int vec_ldl (int, const long *);
6628vector unsigned int vec_ldl (int, const vector unsigned int *);
6629vector unsigned int vec_ldl (int, const unsigned int *);
6630vector unsigned int vec_ldl (int, const unsigned long *);
6631vector bool short vec_ldl (int, const vector bool short *);
6632vector pixel vec_ldl (int, const vector pixel *);
6633vector signed short vec_ldl (int, const vector signed short *);
6634vector signed short vec_ldl (int, const short *);
6635vector unsigned short vec_ldl (int, const vector unsigned short *);
6636vector unsigned short vec_ldl (int, const unsigned short *);
6637vector bool char vec_ldl (int, const vector bool char *);
6638vector signed char vec_ldl (int, const vector signed char *);
6639vector signed char vec_ldl (int, const signed char *);
6640vector unsigned char vec_ldl (int, const vector unsigned char *);
6641vector unsigned char vec_ldl (int, const unsigned char *);
6642
6643vector float vec_loge (vector float);
6644
6645vector unsigned char vec_lvsl (int, const volatile unsigned char *);
6646vector unsigned char vec_lvsl (int, const volatile signed char *);
6647vector unsigned char vec_lvsl (int, const volatile unsigned short *);
6648vector unsigned char vec_lvsl (int, const volatile short *);
6649vector unsigned char vec_lvsl (int, const volatile unsigned int *);
6650vector unsigned char vec_lvsl (int, const volatile int *);
6651vector unsigned char vec_lvsl (int, const volatile unsigned long *);
6652vector unsigned char vec_lvsl (int, const volatile long *);
6653vector unsigned char vec_lvsl (int, const volatile float *);
6654
6655vector unsigned char vec_lvsr (int, const volatile unsigned char *);
6656vector unsigned char vec_lvsr (int, const volatile signed char *);
6657vector unsigned char vec_lvsr (int, const volatile unsigned short *);
6658vector unsigned char vec_lvsr (int, const volatile short *);
6659vector unsigned char vec_lvsr (int, const volatile unsigned int *);
6660vector unsigned char vec_lvsr (int, const volatile int *);
6661vector unsigned char vec_lvsr (int, const volatile unsigned long *);
6662vector unsigned char vec_lvsr (int, const volatile long *);
6663vector unsigned char vec_lvsr (int, const volatile float *);
6664
6665vector float vec_madd (vector float, vector float, vector float);
6666
6667vector signed short vec_madds (vector signed short,
6668                               vector signed short,
6669                               vector signed short);
6670
6671vector unsigned char vec_max (vector bool char, vector unsigned char);
6672vector unsigned char vec_max (vector unsigned char, vector bool char);
6673vector unsigned char vec_max (vector unsigned char,
6674                              vector unsigned char);
6675vector signed char vec_max (vector bool char, vector signed char);
6676vector signed char vec_max (vector signed char, vector bool char);
6677vector signed char vec_max (vector signed char, vector signed char);
6678vector unsigned short vec_max (vector bool short,
6679                               vector unsigned short);
6680vector unsigned short vec_max (vector unsigned short,
6681                               vector bool short);
6682vector unsigned short vec_max (vector unsigned short,
6683                               vector unsigned short);
6684vector signed short vec_max (vector bool short, vector signed short);
6685vector signed short vec_max (vector signed short, vector bool short);
6686vector signed short vec_max (vector signed short, vector signed short);
6687vector unsigned int vec_max (vector bool int, vector unsigned int);
6688vector unsigned int vec_max (vector unsigned int, vector bool int);
6689vector unsigned int vec_max (vector unsigned int, vector unsigned int);
6690vector signed int vec_max (vector bool int, vector signed int);
6691vector signed int vec_max (vector signed int, vector bool int);
6692vector signed int vec_max (vector signed int, vector signed int);
6693vector float vec_max (vector float, vector float);
6694
6695vector float vec_vmaxfp (vector float, vector float);
6696
6697vector signed int vec_vmaxsw (vector bool int, vector signed int);
6698vector signed int vec_vmaxsw (vector signed int, vector bool int);
6699vector signed int vec_vmaxsw (vector signed int, vector signed int);
6700
6701vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
6702vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
6703vector unsigned int vec_vmaxuw (vector unsigned int,
6704                                vector unsigned int);
6705
6706vector signed short vec_vmaxsh (vector bool short, vector signed short);
6707vector signed short vec_vmaxsh (vector signed short, vector bool short);
6708vector signed short vec_vmaxsh (vector signed short,
6709                                vector signed short);
6710
6711vector unsigned short vec_vmaxuh (vector bool short,
6712                                  vector unsigned short);
6713vector unsigned short vec_vmaxuh (vector unsigned short,
6714                                  vector bool short);
6715vector unsigned short vec_vmaxuh (vector unsigned short,
6716                                  vector unsigned short);
6717
6718vector signed char vec_vmaxsb (vector bool char, vector signed char);
6719vector signed char vec_vmaxsb (vector signed char, vector bool char);
6720vector signed char vec_vmaxsb (vector signed char, vector signed char);
6721
6722vector unsigned char vec_vmaxub (vector bool char,
6723                                 vector unsigned char);
6724vector unsigned char vec_vmaxub (vector unsigned char,
6725                                 vector bool char);
6726vector unsigned char vec_vmaxub (vector unsigned char,
6727                                 vector unsigned char);
6728
6729vector bool char vec_mergeh (vector bool char, vector bool char);
6730vector signed char vec_mergeh (vector signed char, vector signed char);
6731vector unsigned char vec_mergeh (vector unsigned char,
6732                                 vector unsigned char);
6733vector bool short vec_mergeh (vector bool short, vector bool short);
6734vector pixel vec_mergeh (vector pixel, vector pixel);
6735vector signed short vec_mergeh (vector signed short,
6736                                vector signed short);
6737vector unsigned short vec_mergeh (vector unsigned short,
6738                                  vector unsigned short);
6739vector float vec_mergeh (vector float, vector float);
6740vector bool int vec_mergeh (vector bool int, vector bool int);
6741vector signed int vec_mergeh (vector signed int, vector signed int);
6742vector unsigned int vec_mergeh (vector unsigned int,
6743                                vector unsigned int);
6744
6745vector float vec_vmrghw (vector float, vector float);
6746vector bool int vec_vmrghw (vector bool int, vector bool int);
6747vector signed int vec_vmrghw (vector signed int, vector signed int);
6748vector unsigned int vec_vmrghw (vector unsigned int,
6749                                vector unsigned int);
6750
6751vector bool short vec_vmrghh (vector bool short, vector bool short);
6752vector signed short vec_vmrghh (vector signed short,
6753                                vector signed short);
6754vector unsigned short vec_vmrghh (vector unsigned short,
6755                                  vector unsigned short);
6756vector pixel vec_vmrghh (vector pixel, vector pixel);
6757
6758vector bool char vec_vmrghb (vector bool char, vector bool char);
6759vector signed char vec_vmrghb (vector signed char, vector signed char);
6760vector unsigned char vec_vmrghb (vector unsigned char,
6761                                 vector unsigned char);
6762
6763vector bool char vec_mergel (vector bool char, vector bool char);
6764vector signed char vec_mergel (vector signed char, vector signed char);
6765vector unsigned char vec_mergel (vector unsigned char,
6766                                 vector unsigned char);
6767vector bool short vec_mergel (vector bool short, vector bool short);
6768vector pixel vec_mergel (vector pixel, vector pixel);
6769vector signed short vec_mergel (vector signed short,
6770                                vector signed short);
6771vector unsigned short vec_mergel (vector unsigned short,
6772                                  vector unsigned short);
6773vector float vec_mergel (vector float, vector float);
6774vector bool int vec_mergel (vector bool int, vector bool int);
6775vector signed int vec_mergel (vector signed int, vector signed int);
6776vector unsigned int vec_mergel (vector unsigned int,
6777                                vector unsigned int);
6778
6779vector float vec_vmrglw (vector float, vector float);
6780vector signed int vec_vmrglw (vector signed int, vector signed int);
6781vector unsigned int vec_vmrglw (vector unsigned int,
6782                                vector unsigned int);
6783vector bool int vec_vmrglw (vector bool int, vector bool int);
6784
6785vector bool short vec_vmrglh (vector bool short, vector bool short);
6786vector signed short vec_vmrglh (vector signed short,
6787                                vector signed short);
6788vector unsigned short vec_vmrglh (vector unsigned short,
6789                                  vector unsigned short);
6790vector pixel vec_vmrglh (vector pixel, vector pixel);
6791
6792vector bool char vec_vmrglb (vector bool char, vector bool char);
6793vector signed char vec_vmrglb (vector signed char, vector signed char);
6794vector unsigned char vec_vmrglb (vector unsigned char,
6795                                 vector unsigned char);
6796
6797vector unsigned short vec_mfvscr (void);
6798
6799vector unsigned char vec_min (vector bool char, vector unsigned char);
6800vector unsigned char vec_min (vector unsigned char, vector bool char);
6801vector unsigned char vec_min (vector unsigned char,
6802                              vector unsigned char);
6803vector signed char vec_min (vector bool char, vector signed char);
6804vector signed char vec_min (vector signed char, vector bool char);
6805vector signed char vec_min (vector signed char, vector signed char);
6806vector unsigned short vec_min (vector bool short,
6807                               vector unsigned short);
6808vector unsigned short vec_min (vector unsigned short,
6809                               vector bool short);
6810vector unsigned short vec_min (vector unsigned short,
6811                               vector unsigned short);
6812vector signed short vec_min (vector bool short, vector signed short);
6813vector signed short vec_min (vector signed short, vector bool short);
6814vector signed short vec_min (vector signed short, vector signed short);
6815vector unsigned int vec_min (vector bool int, vector unsigned int);
6816vector unsigned int vec_min (vector unsigned int, vector bool int);
6817vector unsigned int vec_min (vector unsigned int, vector unsigned int);
6818vector signed int vec_min (vector bool int, vector signed int);
6819vector signed int vec_min (vector signed int, vector bool int);
6820vector signed int vec_min (vector signed int, vector signed int);
6821vector float vec_min (vector float, vector float);
6822
6823vector float vec_vminfp (vector float, vector float);
6824
6825vector signed int vec_vminsw (vector bool int, vector signed int);
6826vector signed int vec_vminsw (vector signed int, vector bool int);
6827vector signed int vec_vminsw (vector signed int, vector signed int);
6828
6829vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
6830vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
6831vector unsigned int vec_vminuw (vector unsigned int,
6832                                vector unsigned int);
6833
6834vector signed short vec_vminsh (vector bool short, vector signed short);
6835vector signed short vec_vminsh (vector signed short, vector bool short);
6836vector signed short vec_vminsh (vector signed short,
6837                                vector signed short);
6838
6839vector unsigned short vec_vminuh (vector bool short,
6840                                  vector unsigned short);
6841vector unsigned short vec_vminuh (vector unsigned short,
6842                                  vector bool short);
6843vector unsigned short vec_vminuh (vector unsigned short,
6844                                  vector unsigned short);
6845
6846vector signed char vec_vminsb (vector bool char, vector signed char);
6847vector signed char vec_vminsb (vector signed char, vector bool char);
6848vector signed char vec_vminsb (vector signed char, vector signed char);
6849
6850vector unsigned char vec_vminub (vector bool char,
6851                                 vector unsigned char);
6852vector unsigned char vec_vminub (vector unsigned char,
6853                                 vector bool char);
6854vector unsigned char vec_vminub (vector unsigned char,
6855                                 vector unsigned char);
6856
6857vector signed short vec_mladd (vector signed short,
6858                               vector signed short,
6859                               vector signed short);
6860vector signed short vec_mladd (vector signed short,
6861                               vector unsigned short,
6862                               vector unsigned short);
6863vector signed short vec_mladd (vector unsigned short,
6864                               vector signed short,
6865                               vector signed short);
6866vector unsigned short vec_mladd (vector unsigned short,
6867                                 vector unsigned short,
6868                                 vector unsigned short);
6869
6870vector signed short vec_mradds (vector signed short,
6871                                vector signed short,
6872                                vector signed short);
6873
6874vector unsigned int vec_msum (vector unsigned char,
6875                              vector unsigned char,
6876                              vector unsigned int);
6877vector signed int vec_msum (vector signed char,
6878                            vector unsigned char,
6879                            vector signed int);
6880vector unsigned int vec_msum (vector unsigned short,
6881                              vector unsigned short,
6882                              vector unsigned int);
6883vector signed int vec_msum (vector signed short,
6884                            vector signed short,
6885                            vector signed int);
6886
6887vector signed int vec_vmsumshm (vector signed short,
6888                                vector signed short,
6889                                vector signed int);
6890
6891vector unsigned int vec_vmsumuhm (vector unsigned short,
6892                                  vector unsigned short,
6893                                  vector unsigned int);
6894
6895vector signed int vec_vmsummbm (vector signed char,
6896                                vector unsigned char,
6897                                vector signed int);
6898
6899vector unsigned int vec_vmsumubm (vector unsigned char,
6900                                  vector unsigned char,
6901                                  vector unsigned int);
6902
6903vector unsigned int vec_msums (vector unsigned short,
6904                               vector unsigned short,
6905                               vector unsigned int);
6906vector signed int vec_msums (vector signed short,
6907                             vector signed short,
6908                             vector signed int);
6909
6910vector signed int vec_vmsumshs (vector signed short,
6911                                vector signed short,
6912                                vector signed int);
6913
6914vector unsigned int vec_vmsumuhs (vector unsigned short,
6915                                  vector unsigned short,
6916                                  vector unsigned int);
6917
6918void vec_mtvscr (vector signed int);
6919void vec_mtvscr (vector unsigned int);
6920void vec_mtvscr (vector bool int);
6921void vec_mtvscr (vector signed short);
6922void vec_mtvscr (vector unsigned short);
6923void vec_mtvscr (vector bool short);
6924void vec_mtvscr (vector pixel);
6925void vec_mtvscr (vector signed char);
6926void vec_mtvscr (vector unsigned char);
6927void vec_mtvscr (vector bool char);
6928
6929vector unsigned short vec_mule (vector unsigned char,
6930                                vector unsigned char);
6931vector signed short vec_mule (vector signed char,
6932                              vector signed char);
6933vector unsigned int vec_mule (vector unsigned short,
6934                              vector unsigned short);
6935vector signed int vec_mule (vector signed short, vector signed short);
6936
6937vector signed int vec_vmulesh (vector signed short,
6938                               vector signed short);
6939
6940vector unsigned int vec_vmuleuh (vector unsigned short,
6941                                 vector unsigned short);
6942
6943vector signed short vec_vmulesb (vector signed char,
6944                                 vector signed char);
6945
6946vector unsigned short vec_vmuleub (vector unsigned char,
6947                                  vector unsigned char);
6948
6949vector unsigned short vec_mulo (vector unsigned char,
6950                                vector unsigned char);
6951vector signed short vec_mulo (vector signed char, vector signed char);
6952vector unsigned int vec_mulo (vector unsigned short,
6953                              vector unsigned short);
6954vector signed int vec_mulo (vector signed short, vector signed short);
6955
6956vector signed int vec_vmulosh (vector signed short,
6957                               vector signed short);
6958
6959vector unsigned int vec_vmulouh (vector unsigned short,
6960                                 vector unsigned short);
6961
6962vector signed short vec_vmulosb (vector signed char,
6963                                 vector signed char);
6964
6965vector unsigned short vec_vmuloub (vector unsigned char,
6966                                   vector unsigned char);
6967
6968vector float vec_nmsub (vector float, vector float, vector float);
6969
6970vector float vec_nor (vector float, vector float);
6971vector signed int vec_nor (vector signed int, vector signed int);
6972vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
6973vector bool int vec_nor (vector bool int, vector bool int);
6974vector signed short vec_nor (vector signed short, vector signed short);
6975vector unsigned short vec_nor (vector unsigned short,
6976                               vector unsigned short);
6977vector bool short vec_nor (vector bool short, vector bool short);
6978vector signed char vec_nor (vector signed char, vector signed char);
6979vector unsigned char vec_nor (vector unsigned char,
6980                              vector unsigned char);
6981vector bool char vec_nor (vector bool char, vector bool char);
6982
6983vector float vec_or (vector float, vector float);
6984vector float vec_or (vector float, vector bool int);
6985vector float vec_or (vector bool int, vector float);
6986vector bool int vec_or (vector bool int, vector bool int);
6987vector signed int vec_or (vector bool int, vector signed int);
6988vector signed int vec_or (vector signed int, vector bool int);
6989vector signed int vec_or (vector signed int, vector signed int);
6990vector unsigned int vec_or (vector bool int, vector unsigned int);
6991vector unsigned int vec_or (vector unsigned int, vector bool int);
6992vector unsigned int vec_or (vector unsigned int, vector unsigned int);
6993vector bool short vec_or (vector bool short, vector bool short);
6994vector signed short vec_or (vector bool short, vector signed short);
6995vector signed short vec_or (vector signed short, vector bool short);
6996vector signed short vec_or (vector signed short, vector signed short);
6997vector unsigned short vec_or (vector bool short, vector unsigned short);
6998vector unsigned short vec_or (vector unsigned short, vector bool short);
6999vector unsigned short vec_or (vector unsigned short,
7000                              vector unsigned short);
7001vector signed char vec_or (vector bool char, vector signed char);
7002vector bool char vec_or (vector bool char, vector bool char);
7003vector signed char vec_or (vector signed char, vector bool char);
7004vector signed char vec_or (vector signed char, vector signed char);
7005vector unsigned char vec_or (vector bool char, vector unsigned char);
7006vector unsigned char vec_or (vector unsigned char, vector bool char);
7007vector unsigned char vec_or (vector unsigned char,
7008                             vector unsigned char);
7009
7010vector signed char vec_pack (vector signed short, vector signed short);
7011vector unsigned char vec_pack (vector unsigned short,
7012                               vector unsigned short);
7013vector bool char vec_pack (vector bool short, vector bool short);
7014vector signed short vec_pack (vector signed int, vector signed int);
7015vector unsigned short vec_pack (vector unsigned int,
7016                                vector unsigned int);
7017vector bool short vec_pack (vector bool int, vector bool int);
7018
7019vector bool short vec_vpkuwum (vector bool int, vector bool int);
7020vector signed short vec_vpkuwum (vector signed int, vector signed int);
7021vector unsigned short vec_vpkuwum (vector unsigned int,
7022                                   vector unsigned int);
7023
7024vector bool char vec_vpkuhum (vector bool short, vector bool short);
7025vector signed char vec_vpkuhum (vector signed short,
7026                                vector signed short);
7027vector unsigned char vec_vpkuhum (vector unsigned short,
7028                                  vector unsigned short);
7029
7030vector pixel vec_packpx (vector unsigned int, vector unsigned int);
7031
7032vector unsigned char vec_packs (vector unsigned short,
7033                                vector unsigned short);
7034vector signed char vec_packs (vector signed short, vector signed short);
7035vector unsigned short vec_packs (vector unsigned int,
7036                                 vector unsigned int);
7037vector signed short vec_packs (vector signed int, vector signed int);
7038
7039vector signed short vec_vpkswss (vector signed int, vector signed int);
7040
7041vector unsigned short vec_vpkuwus (vector unsigned int,
7042                                   vector unsigned int);
7043
7044vector signed char vec_vpkshss (vector signed short,
7045                                vector signed short);
7046
7047vector unsigned char vec_vpkuhus (vector unsigned short,
7048                                  vector unsigned short);
7049
7050vector unsigned char vec_packsu (vector unsigned short,
7051                                 vector unsigned short);
7052vector unsigned char vec_packsu (vector signed short,
7053                                 vector signed short);
7054vector unsigned short vec_packsu (vector unsigned int,
7055                                  vector unsigned int);
7056vector unsigned short vec_packsu (vector signed int, vector signed int);
7057
7058vector unsigned short vec_vpkswus (vector signed int,
7059                                   vector signed int);
7060
7061vector unsigned char vec_vpkshus (vector signed short,
7062                                  vector signed short);
7063
7064vector float vec_perm (vector float,
7065                       vector float,
7066                       vector unsigned char);
7067vector signed int vec_perm (vector signed int,
7068                            vector signed int,
7069                            vector unsigned char);
7070vector unsigned int vec_perm (vector unsigned int,
7071                              vector unsigned int,
7072                              vector unsigned char);
7073vector bool int vec_perm (vector bool int,
7074                          vector bool int,
7075                          vector unsigned char);
7076vector signed short vec_perm (vector signed short,
7077                              vector signed short,
7078                              vector unsigned char);
7079vector unsigned short vec_perm (vector unsigned short,
7080                                vector unsigned short,
7081                                vector unsigned char);
7082vector bool short vec_perm (vector bool short,
7083                            vector bool short,
7084                            vector unsigned char);
7085vector pixel vec_perm (vector pixel,
7086                       vector pixel,
7087                       vector unsigned char);
7088vector signed char vec_perm (vector signed char,
7089                             vector signed char,
7090                             vector unsigned char);
7091vector unsigned char vec_perm (vector unsigned char,
7092                               vector unsigned char,
7093                               vector unsigned char);
7094vector bool char vec_perm (vector bool char,
7095                           vector bool char,
7096                           vector unsigned char);
7097
7098vector float vec_re (vector float);
7099
7100vector signed char vec_rl (vector signed char,
7101                           vector unsigned char);
7102vector unsigned char vec_rl (vector unsigned char,
7103                             vector unsigned char);
7104vector signed short vec_rl (vector signed short, vector unsigned short);
7105vector unsigned short vec_rl (vector unsigned short,
7106                              vector unsigned short);
7107vector signed int vec_rl (vector signed int, vector unsigned int);
7108vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
7109
7110vector signed int vec_vrlw (vector signed int, vector unsigned int);
7111vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
7112
7113vector signed short vec_vrlh (vector signed short,
7114                              vector unsigned short);
7115vector unsigned short vec_vrlh (vector unsigned short,
7116                                vector unsigned short);
7117
7118vector signed char vec_vrlb (vector signed char, vector unsigned char);
7119vector unsigned char vec_vrlb (vector unsigned char,
7120                               vector unsigned char);
7121
7122vector float vec_round (vector float);
7123
7124vector float vec_rsqrte (vector float);
7125
7126vector float vec_sel (vector float, vector float, vector bool int);
7127vector float vec_sel (vector float, vector float, vector unsigned int);
7128vector signed int vec_sel (vector signed int,
7129                           vector signed int,
7130                           vector bool int);
7131vector signed int vec_sel (vector signed int,
7132                           vector signed int,
7133                           vector unsigned int);
7134vector unsigned int vec_sel (vector unsigned int,
7135                             vector unsigned int,
7136                             vector bool int);
7137vector unsigned int vec_sel (vector unsigned int,
7138                             vector unsigned int,
7139                             vector unsigned int);
7140vector bool int vec_sel (vector bool int,
7141                         vector bool int,
7142                         vector bool int);
7143vector bool int vec_sel (vector bool int,
7144                         vector bool int,
7145                         vector unsigned int);
7146vector signed short vec_sel (vector signed short,
7147                             vector signed short,
7148                             vector bool short);
7149vector signed short vec_sel (vector signed short,
7150                             vector signed short,
7151                             vector unsigned short);
7152vector unsigned short vec_sel (vector unsigned short,
7153                               vector unsigned short,
7154                               vector bool short);
7155vector unsigned short vec_sel (vector unsigned short,
7156                               vector unsigned short,
7157                               vector unsigned short);
7158vector bool short vec_sel (vector bool short,
7159                           vector bool short,
7160                           vector bool short);
7161vector bool short vec_sel (vector bool short,
7162                           vector bool short,
7163                           vector unsigned short);
7164vector signed char vec_sel (vector signed char,
7165                            vector signed char,
7166                            vector bool char);
7167vector signed char vec_sel (vector signed char,
7168                            vector signed char,
7169                            vector unsigned char);
7170vector unsigned char vec_sel (vector unsigned char,
7171                              vector unsigned char,
7172                              vector bool char);
7173vector unsigned char vec_sel (vector unsigned char,
7174                              vector unsigned char,
7175                              vector unsigned char);
7176vector bool char vec_sel (vector bool char,
7177                          vector bool char,
7178                          vector bool char);
7179vector bool char vec_sel (vector bool char,
7180                          vector bool char,
7181                          vector unsigned char);
7182
7183vector signed char vec_sl (vector signed char,
7184                           vector unsigned char);
7185vector unsigned char vec_sl (vector unsigned char,
7186                             vector unsigned char);
7187vector signed short vec_sl (vector signed short, vector unsigned short);
7188vector unsigned short vec_sl (vector unsigned short,
7189                              vector unsigned short);
7190vector signed int vec_sl (vector signed int, vector unsigned int);
7191vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
7192
7193vector signed int vec_vslw (vector signed int, vector unsigned int);
7194vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
7195
7196vector signed short vec_vslh (vector signed short,
7197                              vector unsigned short);
7198vector unsigned short vec_vslh (vector unsigned short,
7199                                vector unsigned short);
7200
7201vector signed char vec_vslb (vector signed char, vector unsigned char);
7202vector unsigned char vec_vslb (vector unsigned char,
7203                               vector unsigned char);
7204
7205vector float vec_sld (vector float, vector float, const int);
7206vector signed int vec_sld (vector signed int,
7207                           vector signed int,
7208                           const int);
7209vector unsigned int vec_sld (vector unsigned int,
7210                             vector unsigned int,
7211                             const int);
7212vector bool int vec_sld (vector bool int,
7213                         vector bool int,
7214                         const int);
7215vector signed short vec_sld (vector signed short,
7216                             vector signed short,
7217                             const int);
7218vector unsigned short vec_sld (vector unsigned short,
7219                               vector unsigned short,
7220                               const int);
7221vector bool short vec_sld (vector bool short,
7222                           vector bool short,
7223                           const int);
7224vector pixel vec_sld (vector pixel,
7225                      vector pixel,
7226                      const int);
7227vector signed char vec_sld (vector signed char,
7228                            vector signed char,
7229                            const int);
7230vector unsigned char vec_sld (vector unsigned char,
7231                              vector unsigned char,
7232                              const int);
7233vector bool char vec_sld (vector bool char,
7234                          vector bool char,
7235                          const int);
7236
7237vector signed int vec_sll (vector signed int,
7238                           vector unsigned int);
7239vector signed int vec_sll (vector signed int,
7240                           vector unsigned short);
7241vector signed int vec_sll (vector signed int,
7242                           vector unsigned char);
7243vector unsigned int vec_sll (vector unsigned int,
7244                             vector unsigned int);
7245vector unsigned int vec_sll (vector unsigned int,
7246                             vector unsigned short);
7247vector unsigned int vec_sll (vector unsigned int,
7248                             vector unsigned char);
7249vector bool int vec_sll (vector bool int,
7250                         vector unsigned int);
7251vector bool int vec_sll (vector bool int,
7252                         vector unsigned short);
7253vector bool int vec_sll (vector bool int,
7254                         vector unsigned char);
7255vector signed short vec_sll (vector signed short,
7256                             vector unsigned int);
7257vector signed short vec_sll (vector signed short,
7258                             vector unsigned short);
7259vector signed short vec_sll (vector signed short,
7260                             vector unsigned char);
7261vector unsigned short vec_sll (vector unsigned short,
7262                               vector unsigned int);
7263vector unsigned short vec_sll (vector unsigned short,
7264                               vector unsigned short);
7265vector unsigned short vec_sll (vector unsigned short,
7266                               vector unsigned char);
7267vector bool short vec_sll (vector bool short, vector unsigned int);
7268vector bool short vec_sll (vector bool short, vector unsigned short);
7269vector bool short vec_sll (vector bool short, vector unsigned char);
7270vector pixel vec_sll (vector pixel, vector unsigned int);
7271vector pixel vec_sll (vector pixel, vector unsigned short);
7272vector pixel vec_sll (vector pixel, vector unsigned char);
7273vector signed char vec_sll (vector signed char, vector unsigned int);
7274vector signed char vec_sll (vector signed char, vector unsigned short);
7275vector signed char vec_sll (vector signed char, vector unsigned char);
7276vector unsigned char vec_sll (vector unsigned char,
7277                              vector unsigned int);
7278vector unsigned char vec_sll (vector unsigned char,
7279                              vector unsigned short);
7280vector unsigned char vec_sll (vector unsigned char,
7281                              vector unsigned char);
7282vector bool char vec_sll (vector bool char, vector unsigned int);
7283vector bool char vec_sll (vector bool char, vector unsigned short);
7284vector bool char vec_sll (vector bool char, vector unsigned char);
7285
7286vector float vec_slo (vector float, vector signed char);
7287vector float vec_slo (vector float, vector unsigned char);
7288vector signed int vec_slo (vector signed int, vector signed char);
7289vector signed int vec_slo (vector signed int, vector unsigned char);
7290vector unsigned int vec_slo (vector unsigned int, vector signed char);
7291vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
7292vector signed short vec_slo (vector signed short, vector signed char);
7293vector signed short vec_slo (vector signed short, vector unsigned char);
7294vector unsigned short vec_slo (vector unsigned short,
7295                               vector signed char);
7296vector unsigned short vec_slo (vector unsigned short,
7297                               vector unsigned char);
7298vector pixel vec_slo (vector pixel, vector signed char);
7299vector pixel vec_slo (vector pixel, vector unsigned char);
7300vector signed char vec_slo (vector signed char, vector signed char);
7301vector signed char vec_slo (vector signed char, vector unsigned char);
7302vector unsigned char vec_slo (vector unsigned char, vector signed char);
7303vector unsigned char vec_slo (vector unsigned char,
7304                              vector unsigned char);
7305
7306vector signed char vec_splat (vector signed char, const int);
7307vector unsigned char vec_splat (vector unsigned char, const int);
7308vector bool char vec_splat (vector bool char, const int);
7309vector signed short vec_splat (vector signed short, const int);
7310vector unsigned short vec_splat (vector unsigned short, const int);
7311vector bool short vec_splat (vector bool short, const int);
7312vector pixel vec_splat (vector pixel, const int);
7313vector float vec_splat (vector float, const int);
7314vector signed int vec_splat (vector signed int, const int);
7315vector unsigned int vec_splat (vector unsigned int, const int);
7316vector bool int vec_splat (vector bool int, const int);
7317
7318vector float vec_vspltw (vector float, const int);
7319vector signed int vec_vspltw (vector signed int, const int);
7320vector unsigned int vec_vspltw (vector unsigned int, const int);
7321vector bool int vec_vspltw (vector bool int, const int);
7322
7323vector bool short vec_vsplth (vector bool short, const int);
7324vector signed short vec_vsplth (vector signed short, const int);
7325vector unsigned short vec_vsplth (vector unsigned short, const int);
7326vector pixel vec_vsplth (vector pixel, const int);
7327
7328vector signed char vec_vspltb (vector signed char, const int);
7329vector unsigned char vec_vspltb (vector unsigned char, const int);
7330vector bool char vec_vspltb (vector bool char, const int);
7331
7332vector signed char vec_splat_s8 (const int);
7333
7334vector signed short vec_splat_s16 (const int);
7335
7336vector signed int vec_splat_s32 (const int);
7337
7338vector unsigned char vec_splat_u8 (const int);
7339
7340vector unsigned short vec_splat_u16 (const int);
7341
7342vector unsigned int vec_splat_u32 (const int);
7343
7344vector signed char vec_sr (vector signed char, vector unsigned char);
7345vector unsigned char vec_sr (vector unsigned char,
7346                             vector unsigned char);
7347vector signed short vec_sr (vector signed short,
7348                            vector unsigned short);
7349vector unsigned short vec_sr (vector unsigned short,
7350                              vector unsigned short);
7351vector signed int vec_sr (vector signed int, vector unsigned int);
7352vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
7353
7354vector signed int vec_vsrw (vector signed int, vector unsigned int);
7355vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
7356
7357vector signed short vec_vsrh (vector signed short,
7358                              vector unsigned short);
7359vector unsigned short vec_vsrh (vector unsigned short,
7360                                vector unsigned short);
7361
7362vector signed char vec_vsrb (vector signed char, vector unsigned char);
7363vector unsigned char vec_vsrb (vector unsigned char,
7364                               vector unsigned char);
7365
7366vector signed char vec_sra (vector signed char, vector unsigned char);
7367vector unsigned char vec_sra (vector unsigned char,
7368                              vector unsigned char);
7369vector signed short vec_sra (vector signed short,
7370                             vector unsigned short);
7371vector unsigned short vec_sra (vector unsigned short,
7372                               vector unsigned short);
7373vector signed int vec_sra (vector signed int, vector unsigned int);
7374vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
7375
7376vector signed int vec_vsraw (vector signed int, vector unsigned int);
7377vector unsigned int vec_vsraw (vector unsigned int,
7378                               vector unsigned int);
7379
7380vector signed short vec_vsrah (vector signed short,
7381                               vector unsigned short);
7382vector unsigned short vec_vsrah (vector unsigned short,
7383                                 vector unsigned short);
7384
7385vector signed char vec_vsrab (vector signed char, vector unsigned char);
7386vector unsigned char vec_vsrab (vector unsigned char,
7387                                vector unsigned char);
7388
7389vector signed int vec_srl (vector signed int, vector unsigned int);
7390vector signed int vec_srl (vector signed int, vector unsigned short);
7391vector signed int vec_srl (vector signed int, vector unsigned char);
7392vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
7393vector unsigned int vec_srl (vector unsigned int,
7394                             vector unsigned short);
7395vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
7396vector bool int vec_srl (vector bool int, vector unsigned int);
7397vector bool int vec_srl (vector bool int, vector unsigned short);
7398vector bool int vec_srl (vector bool int, vector unsigned char);
7399vector signed short vec_srl (vector signed short, vector unsigned int);
7400vector signed short vec_srl (vector signed short,
7401                             vector unsigned short);
7402vector signed short vec_srl (vector signed short, vector unsigned char);
7403vector unsigned short vec_srl (vector unsigned short,
7404                               vector unsigned int);
7405vector unsigned short vec_srl (vector unsigned short,
7406                               vector unsigned short);
7407vector unsigned short vec_srl (vector unsigned short,
7408                               vector unsigned char);
7409vector bool short vec_srl (vector bool short, vector unsigned int);
7410vector bool short vec_srl (vector bool short, vector unsigned short);
7411vector bool short vec_srl (vector bool short, vector unsigned char);
7412vector pixel vec_srl (vector pixel, vector unsigned int);
7413vector pixel vec_srl (vector pixel, vector unsigned short);
7414vector pixel vec_srl (vector pixel, vector unsigned char);
7415vector signed char vec_srl (vector signed char, vector unsigned int);
7416vector signed char vec_srl (vector signed char, vector unsigned short);
7417vector signed char vec_srl (vector signed char, vector unsigned char);
7418vector unsigned char vec_srl (vector unsigned char,
7419                              vector unsigned int);
7420vector unsigned char vec_srl (vector unsigned char,
7421                              vector unsigned short);
7422vector unsigned char vec_srl (vector unsigned char,
7423                              vector unsigned char);
7424vector bool char vec_srl (vector bool char, vector unsigned int);
7425vector bool char vec_srl (vector bool char, vector unsigned short);
7426vector bool char vec_srl (vector bool char, vector unsigned char);
7427
7428vector float vec_sro (vector float, vector signed char);
7429vector float vec_sro (vector float, vector unsigned char);
7430vector signed int vec_sro (vector signed int, vector signed char);
7431vector signed int vec_sro (vector signed int, vector unsigned char);
7432vector unsigned int vec_sro (vector unsigned int, vector signed char);
7433vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
7434vector signed short vec_sro (vector signed short, vector signed char);
7435vector signed short vec_sro (vector signed short, vector unsigned char);
7436vector unsigned short vec_sro (vector unsigned short,
7437                               vector signed char);
7438vector unsigned short vec_sro (vector unsigned short,
7439                               vector unsigned char);
7440vector pixel vec_sro (vector pixel, vector signed char);
7441vector pixel vec_sro (vector pixel, vector unsigned char);
7442vector signed char vec_sro (vector signed char, vector signed char);
7443vector signed char vec_sro (vector signed char, vector unsigned char);
7444vector unsigned char vec_sro (vector unsigned char, vector signed char);
7445vector unsigned char vec_sro (vector unsigned char,
7446                              vector unsigned char);
7447
7448void vec_st (vector float, int, vector float *);
7449void vec_st (vector float, int, float *);
7450void vec_st (vector signed int, int, vector signed int *);
7451void vec_st (vector signed int, int, int *);
7452void vec_st (vector unsigned int, int, vector unsigned int *);
7453void vec_st (vector unsigned int, int, unsigned int *);
7454void vec_st (vector bool int, int, vector bool int *);
7455void vec_st (vector bool int, int, unsigned int *);
7456void vec_st (vector bool int, int, int *);
7457void vec_st (vector signed short, int, vector signed short *);
7458void vec_st (vector signed short, int, short *);
7459void vec_st (vector unsigned short, int, vector unsigned short *);
7460void vec_st (vector unsigned short, int, unsigned short *);
7461void vec_st (vector bool short, int, vector bool short *);
7462void vec_st (vector bool short, int, unsigned short *);
7463void vec_st (vector pixel, int, vector pixel *);
7464void vec_st (vector pixel, int, unsigned short *);
7465void vec_st (vector pixel, int, short *);
7466void vec_st (vector bool short, int, short *);
7467void vec_st (vector signed char, int, vector signed char *);
7468void vec_st (vector signed char, int, signed char *);
7469void vec_st (vector unsigned char, int, vector unsigned char *);
7470void vec_st (vector unsigned char, int, unsigned char *);
7471void vec_st (vector bool char, int, vector bool char *);
7472void vec_st (vector bool char, int, unsigned char *);
7473void vec_st (vector bool char, int, signed char *);
7474
7475void vec_ste (vector signed char, int, signed char *);
7476void vec_ste (vector unsigned char, int, unsigned char *);
7477void vec_ste (vector bool char, int, signed char *);
7478void vec_ste (vector bool char, int, unsigned char *);
7479void vec_ste (vector signed short, int, short *);
7480void vec_ste (vector unsigned short, int, unsigned short *);
7481void vec_ste (vector bool short, int, short *);
7482void vec_ste (vector bool short, int, unsigned short *);
7483void vec_ste (vector pixel, int, short *);
7484void vec_ste (vector pixel, int, unsigned short *);
7485void vec_ste (vector float, int, float *);
7486void vec_ste (vector signed int, int, int *);
7487void vec_ste (vector unsigned int, int, unsigned int *);
7488void vec_ste (vector bool int, int, int *);
7489void vec_ste (vector bool int, int, unsigned int *);
7490
7491void vec_stvewx (vector float, int, float *);
7492void vec_stvewx (vector signed int, int, int *);
7493void vec_stvewx (vector unsigned int, int, unsigned int *);
7494void vec_stvewx (vector bool int, int, int *);
7495void vec_stvewx (vector bool int, int, unsigned int *);
7496
7497void vec_stvehx (vector signed short, int, short *);
7498void vec_stvehx (vector unsigned short, int, unsigned short *);
7499void vec_stvehx (vector bool short, int, short *);
7500void vec_stvehx (vector bool short, int, unsigned short *);
7501void vec_stvehx (vector pixel, int, short *);
7502void vec_stvehx (vector pixel, int, unsigned short *);
7503
7504void vec_stvebx (vector signed char, int, signed char *);
7505void vec_stvebx (vector unsigned char, int, unsigned char *);
7506void vec_stvebx (vector bool char, int, signed char *);
7507void vec_stvebx (vector bool char, int, unsigned char *);
7508
7509void vec_stl (vector float, int, vector float *);
7510void vec_stl (vector float, int, float *);
7511void vec_stl (vector signed int, int, vector signed int *);
7512void vec_stl (vector signed int, int, int *);
7513void vec_stl (vector unsigned int, int, vector unsigned int *);
7514void vec_stl (vector unsigned int, int, unsigned int *);
7515void vec_stl (vector bool int, int, vector bool int *);
7516void vec_stl (vector bool int, int, unsigned int *);
7517void vec_stl (vector bool int, int, int *);
7518void vec_stl (vector signed short, int, vector signed short *);
7519void vec_stl (vector signed short, int, short *);
7520void vec_stl (vector unsigned short, int, vector unsigned short *);
7521void vec_stl (vector unsigned short, int, unsigned short *);
7522void vec_stl (vector bool short, int, vector bool short *);
7523void vec_stl (vector bool short, int, unsigned short *);
7524void vec_stl (vector bool short, int, short *);
7525void vec_stl (vector pixel, int, vector pixel *);
7526void vec_stl (vector pixel, int, unsigned short *);
7527void vec_stl (vector pixel, int, short *);
7528void vec_stl (vector signed char, int, vector signed char *);
7529void vec_stl (vector signed char, int, signed char *);
7530void vec_stl (vector unsigned char, int, vector unsigned char *);
7531void vec_stl (vector unsigned char, int, unsigned char *);
7532void vec_stl (vector bool char, int, vector bool char *);
7533void vec_stl (vector bool char, int, unsigned char *);
7534void vec_stl (vector bool char, int, signed char *);
7535
7536vector signed char vec_sub (vector bool char, vector signed char);
7537vector signed char vec_sub (vector signed char, vector bool char);
7538vector signed char vec_sub (vector signed char, vector signed char);
7539vector unsigned char vec_sub (vector bool char, vector unsigned char);
7540vector unsigned char vec_sub (vector unsigned char, vector bool char);
7541vector unsigned char vec_sub (vector unsigned char,
7542                              vector unsigned char);
7543vector signed short vec_sub (vector bool short, vector signed short);
7544vector signed short vec_sub (vector signed short, vector bool short);
7545vector signed short vec_sub (vector signed short, vector signed short);
7546vector unsigned short vec_sub (vector bool short,
7547                               vector unsigned short);
7548vector unsigned short vec_sub (vector unsigned short,
7549                               vector bool short);
7550vector unsigned short vec_sub (vector unsigned short,
7551                               vector unsigned short);
7552vector signed int vec_sub (vector bool int, vector signed int);
7553vector signed int vec_sub (vector signed int, vector bool int);
7554vector signed int vec_sub (vector signed int, vector signed int);
7555vector unsigned int vec_sub (vector bool int, vector unsigned int);
7556vector unsigned int vec_sub (vector unsigned int, vector bool int);
7557vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
7558vector float vec_sub (vector float, vector float);
7559
7560vector float vec_vsubfp (vector float, vector float);
7561
7562vector signed int vec_vsubuwm (vector bool int, vector signed int);
7563vector signed int vec_vsubuwm (vector signed int, vector bool int);
7564vector signed int vec_vsubuwm (vector signed int, vector signed int);
7565vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
7566vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
7567vector unsigned int vec_vsubuwm (vector unsigned int,
7568                                 vector unsigned int);
7569
7570vector signed short vec_vsubuhm (vector bool short,
7571                                 vector signed short);
7572vector signed short vec_vsubuhm (vector signed short,
7573                                 vector bool short);
7574vector signed short vec_vsubuhm (vector signed short,
7575                                 vector signed short);
7576vector unsigned short vec_vsubuhm (vector bool short,
7577                                   vector unsigned short);
7578vector unsigned short vec_vsubuhm (vector unsigned short,
7579                                   vector bool short);
7580vector unsigned short vec_vsubuhm (vector unsigned short,
7581                                   vector unsigned short);
7582
7583vector signed char vec_vsububm (vector bool char, vector signed char);
7584vector signed char vec_vsububm (vector signed char, vector bool char);
7585vector signed char vec_vsububm (vector signed char, vector signed char);
7586vector unsigned char vec_vsububm (vector bool char,
7587                                  vector unsigned char);
7588vector unsigned char vec_vsububm (vector unsigned char,
7589                                  vector bool char);
7590vector unsigned char vec_vsububm (vector unsigned char,
7591                                  vector unsigned char);
7592
7593vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
7594
7595vector unsigned char vec_subs (vector bool char, vector unsigned char);
7596vector unsigned char vec_subs (vector unsigned char, vector bool char);
7597vector unsigned char vec_subs (vector unsigned char,
7598                               vector unsigned char);
7599vector signed char vec_subs (vector bool char, vector signed char);
7600vector signed char vec_subs (vector signed char, vector bool char);
7601vector signed char vec_subs (vector signed char, vector signed char);
7602vector unsigned short vec_subs (vector bool short,
7603                                vector unsigned short);
7604vector unsigned short vec_subs (vector unsigned short,
7605                                vector bool short);
7606vector unsigned short vec_subs (vector unsigned short,
7607                                vector unsigned short);
7608vector signed short vec_subs (vector bool short, vector signed short);
7609vector signed short vec_subs (vector signed short, vector bool short);
7610vector signed short vec_subs (vector signed short, vector signed short);
7611vector unsigned int vec_subs (vector bool int, vector unsigned int);
7612vector unsigned int vec_subs (vector unsigned int, vector bool int);
7613vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
7614vector signed int vec_subs (vector bool int, vector signed int);
7615vector signed int vec_subs (vector signed int, vector bool int);
7616vector signed int vec_subs (vector signed int, vector signed int);
7617
7618vector signed int vec_vsubsws (vector bool int, vector signed int);
7619vector signed int vec_vsubsws (vector signed int, vector bool int);
7620vector signed int vec_vsubsws (vector signed int, vector signed int);
7621
7622vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
7623vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
7624vector unsigned int vec_vsubuws (vector unsigned int,
7625                                 vector unsigned int);
7626
7627vector signed short vec_vsubshs (vector bool short,
7628                                 vector signed short);
7629vector signed short vec_vsubshs (vector signed short,
7630                                 vector bool short);
7631vector signed short vec_vsubshs (vector signed short,
7632                                 vector signed short);
7633
7634vector unsigned short vec_vsubuhs (vector bool short,
7635                                   vector unsigned short);
7636vector unsigned short vec_vsubuhs (vector unsigned short,
7637                                   vector bool short);
7638vector unsigned short vec_vsubuhs (vector unsigned short,
7639                                   vector unsigned short);
7640
7641vector signed char vec_vsubsbs (vector bool char, vector signed char);
7642vector signed char vec_vsubsbs (vector signed char, vector bool char);
7643vector signed char vec_vsubsbs (vector signed char, vector signed char);
7644
7645vector unsigned char vec_vsububs (vector bool char,
7646                                  vector unsigned char);
7647vector unsigned char vec_vsububs (vector unsigned char,
7648                                  vector bool char);
7649vector unsigned char vec_vsububs (vector unsigned char,
7650                                  vector unsigned char);
7651
7652vector unsigned int vec_sum4s (vector unsigned char,
7653                               vector unsigned int);
7654vector signed int vec_sum4s (vector signed char, vector signed int);
7655vector signed int vec_sum4s (vector signed short, vector signed int);
7656
7657vector signed int vec_vsum4shs (vector signed short, vector signed int);
7658
7659vector signed int vec_vsum4sbs (vector signed char, vector signed int);
7660
7661vector unsigned int vec_vsum4ubs (vector unsigned char,
7662                                  vector unsigned int);
7663
7664vector signed int vec_sum2s (vector signed int, vector signed int);
7665
7666vector signed int vec_sums (vector signed int, vector signed int);
7667
7668vector float vec_trunc (vector float);
7669
7670vector signed short vec_unpackh (vector signed char);
7671vector bool short vec_unpackh (vector bool char);
7672vector signed int vec_unpackh (vector signed short);
7673vector bool int vec_unpackh (vector bool short);
7674vector unsigned int vec_unpackh (vector pixel);
7675
7676vector bool int vec_vupkhsh (vector bool short);
7677vector signed int vec_vupkhsh (vector signed short);
7678
7679vector unsigned int vec_vupkhpx (vector pixel);
7680
7681vector bool short vec_vupkhsb (vector bool char);
7682vector signed short vec_vupkhsb (vector signed char);
7683
7684vector signed short vec_unpackl (vector signed char);
7685vector bool short vec_unpackl (vector bool char);
7686vector unsigned int vec_unpackl (vector pixel);
7687vector signed int vec_unpackl (vector signed short);
7688vector bool int vec_unpackl (vector bool short);
7689
7690vector unsigned int vec_vupklpx (vector pixel);
7691
7692vector bool int vec_vupklsh (vector bool short);
7693vector signed int vec_vupklsh (vector signed short);
7694
7695vector bool short vec_vupklsb (vector bool char);
7696vector signed short vec_vupklsb (vector signed char);
7697
7698vector float vec_xor (vector float, vector float);
7699vector float vec_xor (vector float, vector bool int);
7700vector float vec_xor (vector bool int, vector float);
7701vector bool int vec_xor (vector bool int, vector bool int);
7702vector signed int vec_xor (vector bool int, vector signed int);
7703vector signed int vec_xor (vector signed int, vector bool int);
7704vector signed int vec_xor (vector signed int, vector signed int);
7705vector unsigned int vec_xor (vector bool int, vector unsigned int);
7706vector unsigned int vec_xor (vector unsigned int, vector bool int);
7707vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
7708vector bool short vec_xor (vector bool short, vector bool short);
7709vector signed short vec_xor (vector bool short, vector signed short);
7710vector signed short vec_xor (vector signed short, vector bool short);
7711vector signed short vec_xor (vector signed short, vector signed short);
7712vector unsigned short vec_xor (vector bool short,
7713                               vector unsigned short);
7714vector unsigned short vec_xor (vector unsigned short,
7715                               vector bool short);
7716vector unsigned short vec_xor (vector unsigned short,
7717                               vector unsigned short);
7718vector signed char vec_xor (vector bool char, vector signed char);
7719vector bool char vec_xor (vector bool char, vector bool char);
7720vector signed char vec_xor (vector signed char, vector bool char);
7721vector signed char vec_xor (vector signed char, vector signed char);
7722vector unsigned char vec_xor (vector bool char, vector unsigned char);
7723vector unsigned char vec_xor (vector unsigned char, vector bool char);
7724vector unsigned char vec_xor (vector unsigned char,
7725                              vector unsigned char);
7726
7727int vec_all_eq (vector signed char, vector bool char);
7728int vec_all_eq (vector signed char, vector signed char);
7729int vec_all_eq (vector unsigned char, vector bool char);
7730int vec_all_eq (vector unsigned char, vector unsigned char);
7731int vec_all_eq (vector bool char, vector bool char);
7732int vec_all_eq (vector bool char, vector unsigned char);
7733int vec_all_eq (vector bool char, vector signed char);
7734int vec_all_eq (vector signed short, vector bool short);
7735int vec_all_eq (vector signed short, vector signed short);
7736int vec_all_eq (vector unsigned short, vector bool short);
7737int vec_all_eq (vector unsigned short, vector unsigned short);
7738int vec_all_eq (vector bool short, vector bool short);
7739int vec_all_eq (vector bool short, vector unsigned short);
7740int vec_all_eq (vector bool short, vector signed short);
7741int vec_all_eq (vector pixel, vector pixel);
7742int vec_all_eq (vector signed int, vector bool int);
7743int vec_all_eq (vector signed int, vector signed int);
7744int vec_all_eq (vector unsigned int, vector bool int);
7745int vec_all_eq (vector unsigned int, vector unsigned int);
7746int vec_all_eq (vector bool int, vector bool int);
7747int vec_all_eq (vector bool int, vector unsigned int);
7748int vec_all_eq (vector bool int, vector signed int);
7749int vec_all_eq (vector float, vector float);
7750
7751int vec_all_ge (vector bool char, vector unsigned char);
7752int vec_all_ge (vector unsigned char, vector bool char);
7753int vec_all_ge (vector unsigned char, vector unsigned char);
7754int vec_all_ge (vector bool char, vector signed char);
7755int vec_all_ge (vector signed char, vector bool char);
7756int vec_all_ge (vector signed char, vector signed char);
7757int vec_all_ge (vector bool short, vector unsigned short);
7758int vec_all_ge (vector unsigned short, vector bool short);
7759int vec_all_ge (vector unsigned short, vector unsigned short);
7760int vec_all_ge (vector signed short, vector signed short);
7761int vec_all_ge (vector bool short, vector signed short);
7762int vec_all_ge (vector signed short, vector bool short);
7763int vec_all_ge (vector bool int, vector unsigned int);
7764int vec_all_ge (vector unsigned int, vector bool int);
7765int vec_all_ge (vector unsigned int, vector unsigned int);
7766int vec_all_ge (vector bool int, vector signed int);
7767int vec_all_ge (vector signed int, vector bool int);
7768int vec_all_ge (vector signed int, vector signed int);
7769int vec_all_ge (vector float, vector float);
7770
7771int vec_all_gt (vector bool char, vector unsigned char);
7772int vec_all_gt (vector unsigned char, vector bool char);
7773int vec_all_gt (vector unsigned char, vector unsigned char);
7774int vec_all_gt (vector bool char, vector signed char);
7775int vec_all_gt (vector signed char, vector bool char);
7776int vec_all_gt (vector signed char, vector signed char);
7777int vec_all_gt (vector bool short, vector unsigned short);
7778int vec_all_gt (vector unsigned short, vector bool short);
7779int vec_all_gt (vector unsigned short, vector unsigned short);
7780int vec_all_gt (vector bool short, vector signed short);
7781int vec_all_gt (vector signed short, vector bool short);
7782int vec_all_gt (vector signed short, vector signed short);
7783int vec_all_gt (vector bool int, vector unsigned int);
7784int vec_all_gt (vector unsigned int, vector bool int);
7785int vec_all_gt (vector unsigned int, vector unsigned int);
7786int vec_all_gt (vector bool int, vector signed int);
7787int vec_all_gt (vector signed int, vector bool int);
7788int vec_all_gt (vector signed int, vector signed int);
7789int vec_all_gt (vector float, vector float);
7790
7791int vec_all_in (vector float, vector float);
7792
7793int vec_all_le (vector bool char, vector unsigned char);
7794int vec_all_le (vector unsigned char, vector bool char);
7795int vec_all_le (vector unsigned char, vector unsigned char);
7796int vec_all_le (vector bool char, vector signed char);
7797int vec_all_le (vector signed char, vector bool char);
7798int vec_all_le (vector signed char, vector signed char);
7799int vec_all_le (vector bool short, vector unsigned short);
7800int vec_all_le (vector unsigned short, vector bool short);
7801int vec_all_le (vector unsigned short, vector unsigned short);
7802int vec_all_le (vector bool short, vector signed short);
7803int vec_all_le (vector signed short, vector bool short);
7804int vec_all_le (vector signed short, vector signed short);
7805int vec_all_le (vector bool int, vector unsigned int);
7806int vec_all_le (vector unsigned int, vector bool int);
7807int vec_all_le (vector unsigned int, vector unsigned int);
7808int vec_all_le (vector bool int, vector signed int);
7809int vec_all_le (vector signed int, vector bool int);
7810int vec_all_le (vector signed int, vector signed int);
7811int vec_all_le (vector float, vector float);
7812
7813int vec_all_lt (vector bool char, vector unsigned char);
7814int vec_all_lt (vector unsigned char, vector bool char);
7815int vec_all_lt (vector unsigned char, vector unsigned char);
7816int vec_all_lt (vector bool char, vector signed char);
7817int vec_all_lt (vector signed char, vector bool char);
7818int vec_all_lt (vector signed char, vector signed char);
7819int vec_all_lt (vector bool short, vector unsigned short);
7820int vec_all_lt (vector unsigned short, vector bool short);
7821int vec_all_lt (vector unsigned short, vector unsigned short);
7822int vec_all_lt (vector bool short, vector signed short);
7823int vec_all_lt (vector signed short, vector bool short);
7824int vec_all_lt (vector signed short, vector signed short);
7825int vec_all_lt (vector bool int, vector unsigned int);
7826int vec_all_lt (vector unsigned int, vector bool int);
7827int vec_all_lt (vector unsigned int, vector unsigned int);
7828int vec_all_lt (vector bool int, vector signed int);
7829int vec_all_lt (vector signed int, vector bool int);
7830int vec_all_lt (vector signed int, vector signed int);
7831int vec_all_lt (vector float, vector float);
7832
7833int vec_all_nan (vector float);
7834
7835int vec_all_ne (vector signed char, vector bool char);
7836int vec_all_ne (vector signed char, vector signed char);
7837int vec_all_ne (vector unsigned char, vector bool char);
7838int vec_all_ne (vector unsigned char, vector unsigned char);
7839int vec_all_ne (vector bool char, vector bool char);
7840int vec_all_ne (vector bool char, vector unsigned char);
7841int vec_all_ne (vector bool char, vector signed char);
7842int vec_all_ne (vector signed short, vector bool short);
7843int vec_all_ne (vector signed short, vector signed short);
7844int vec_all_ne (vector unsigned short, vector bool short);
7845int vec_all_ne (vector unsigned short, vector unsigned short);
7846int vec_all_ne (vector bool short, vector bool short);
7847int vec_all_ne (vector bool short, vector unsigned short);
7848int vec_all_ne (vector bool short, vector signed short);
7849int vec_all_ne (vector pixel, vector pixel);
7850int vec_all_ne (vector signed int, vector bool int);
7851int vec_all_ne (vector signed int, vector signed int);
7852int vec_all_ne (vector unsigned int, vector bool int);
7853int vec_all_ne (vector unsigned int, vector unsigned int);
7854int vec_all_ne (vector bool int, vector bool int);
7855int vec_all_ne (vector bool int, vector unsigned int);
7856int vec_all_ne (vector bool int, vector signed int);
7857int vec_all_ne (vector float, vector float);
7858
7859int vec_all_nge (vector float, vector float);
7860
7861int vec_all_ngt (vector float, vector float);
7862
7863int vec_all_nle (vector float, vector float);
7864
7865int vec_all_nlt (vector float, vector float);
7866
7867int vec_all_numeric (vector float);
7868
7869int vec_any_eq (vector signed char, vector bool char);
7870int vec_any_eq (vector signed char, vector signed char);
7871int vec_any_eq (vector unsigned char, vector bool char);
7872int vec_any_eq (vector unsigned char, vector unsigned char);
7873int vec_any_eq (vector bool char, vector bool char);
7874int vec_any_eq (vector bool char, vector unsigned char);
7875int vec_any_eq (vector bool char, vector signed char);
7876int vec_any_eq (vector signed short, vector bool short);
7877int vec_any_eq (vector signed short, vector signed short);
7878int vec_any_eq (vector unsigned short, vector bool short);
7879int vec_any_eq (vector unsigned short, vector unsigned short);
7880int vec_any_eq (vector bool short, vector bool short);
7881int vec_any_eq (vector bool short, vector unsigned short);
7882int vec_any_eq (vector bool short, vector signed short);
7883int vec_any_eq (vector pixel, vector pixel);
7884int vec_any_eq (vector signed int, vector bool int);
7885int vec_any_eq (vector signed int, vector signed int);
7886int vec_any_eq (vector unsigned int, vector bool int);
7887int vec_any_eq (vector unsigned int, vector unsigned int);
7888int vec_any_eq (vector bool int, vector bool int);
7889int vec_any_eq (vector bool int, vector unsigned int);
7890int vec_any_eq (vector bool int, vector signed int);
7891int vec_any_eq (vector float, vector float);
7892
7893int vec_any_ge (vector signed char, vector bool char);
7894int vec_any_ge (vector unsigned char, vector bool char);
7895int vec_any_ge (vector unsigned char, vector unsigned char);
7896int vec_any_ge (vector signed char, vector signed char);
7897int vec_any_ge (vector bool char, vector unsigned char);
7898int vec_any_ge (vector bool char, vector signed char);
7899int vec_any_ge (vector unsigned short, vector bool short);
7900int vec_any_ge (vector unsigned short, vector unsigned short);
7901int vec_any_ge (vector signed short, vector signed short);
7902int vec_any_ge (vector signed short, vector bool short);
7903int vec_any_ge (vector bool short, vector unsigned short);
7904int vec_any_ge (vector bool short, vector signed short);
7905int vec_any_ge (vector signed int, vector bool int);
7906int vec_any_ge (vector unsigned int, vector bool int);
7907int vec_any_ge (vector unsigned int, vector unsigned int);
7908int vec_any_ge (vector signed int, vector signed int);
7909int vec_any_ge (vector bool int, vector unsigned int);
7910int vec_any_ge (vector bool int, vector signed int);
7911int vec_any_ge (vector float, vector float);
7912
7913int vec_any_gt (vector bool char, vector unsigned char);
7914int vec_any_gt (vector unsigned char, vector bool char);
7915int vec_any_gt (vector unsigned char, vector unsigned char);
7916int vec_any_gt (vector bool char, vector signed char);
7917int vec_any_gt (vector signed char, vector bool char);
7918int vec_any_gt (vector signed char, vector signed char);
7919int vec_any_gt (vector bool short, vector unsigned short);
7920int vec_any_gt (vector unsigned short, vector bool short);
7921int vec_any_gt (vector unsigned short, vector unsigned short);
7922int vec_any_gt (vector bool short, vector signed short);
7923int vec_any_gt (vector signed short, vector bool short);
7924int vec_any_gt (vector signed short, vector signed short);
7925int vec_any_gt (vector bool int, vector unsigned int);
7926int vec_any_gt (vector unsigned int, vector bool int);
7927int vec_any_gt (vector unsigned int, vector unsigned int);
7928int vec_any_gt (vector bool int, vector signed int);
7929int vec_any_gt (vector signed int, vector bool int);
7930int vec_any_gt (vector signed int, vector signed int);
7931int vec_any_gt (vector float, vector float);
7932
7933int vec_any_le (vector bool char, vector unsigned char);
7934int vec_any_le (vector unsigned char, vector bool char);
7935int vec_any_le (vector unsigned char, vector unsigned char);
7936int vec_any_le (vector bool char, vector signed char);
7937int vec_any_le (vector signed char, vector bool char);
7938int vec_any_le (vector signed char, vector signed char);
7939int vec_any_le (vector bool short, vector unsigned short);
7940int vec_any_le (vector unsigned short, vector bool short);
7941int vec_any_le (vector unsigned short, vector unsigned short);
7942int vec_any_le (vector bool short, vector signed short);
7943int vec_any_le (vector signed short, vector bool short);
7944int vec_any_le (vector signed short, vector signed short);
7945int vec_any_le (vector bool int, vector unsigned int);
7946int vec_any_le (vector unsigned int, vector bool int);
7947int vec_any_le (vector unsigned int, vector unsigned int);
7948int vec_any_le (vector bool int, vector signed int);
7949int vec_any_le (vector signed int, vector bool int);
7950int vec_any_le (vector signed int, vector signed int);
7951int vec_any_le (vector float, vector float);
7952
7953int vec_any_lt (vector bool char, vector unsigned char);
7954int vec_any_lt (vector unsigned char, vector bool char);
7955int vec_any_lt (vector unsigned char, vector unsigned char);
7956int vec_any_lt (vector bool char, vector signed char);
7957int vec_any_lt (vector signed char, vector bool char);
7958int vec_any_lt (vector signed char, vector signed char);
7959int vec_any_lt (vector bool short, vector unsigned short);
7960int vec_any_lt (vector unsigned short, vector bool short);
7961int vec_any_lt (vector unsigned short, vector unsigned short);
7962int vec_any_lt (vector bool short, vector signed short);
7963int vec_any_lt (vector signed short, vector bool short);
7964int vec_any_lt (vector signed short, vector signed short);
7965int vec_any_lt (vector bool int, vector unsigned int);
7966int vec_any_lt (vector unsigned int, vector bool int);
7967int vec_any_lt (vector unsigned int, vector unsigned int);
7968int vec_any_lt (vector bool int, vector signed int);
7969int vec_any_lt (vector signed int, vector bool int);
7970int vec_any_lt (vector signed int, vector signed int);
7971int vec_any_lt (vector float, vector float);
7972
7973int vec_any_nan (vector float);
7974
7975int vec_any_ne (vector signed char, vector bool char);
7976int vec_any_ne (vector signed char, vector signed char);
7977int vec_any_ne (vector unsigned char, vector bool char);
7978int vec_any_ne (vector unsigned char, vector unsigned char);
7979int vec_any_ne (vector bool char, vector bool char);
7980int vec_any_ne (vector bool char, vector unsigned char);
7981int vec_any_ne (vector bool char, vector signed char);
7982int vec_any_ne (vector signed short, vector bool short);
7983int vec_any_ne (vector signed short, vector signed short);
7984int vec_any_ne (vector unsigned short, vector bool short);
7985int vec_any_ne (vector unsigned short, vector unsigned short);
7986int vec_any_ne (vector bool short, vector bool short);
7987int vec_any_ne (vector bool short, vector unsigned short);
7988int vec_any_ne (vector bool short, vector signed short);
7989int vec_any_ne (vector pixel, vector pixel);
7990int vec_any_ne (vector signed int, vector bool int);
7991int vec_any_ne (vector signed int, vector signed int);
7992int vec_any_ne (vector unsigned int, vector bool int);
7993int vec_any_ne (vector unsigned int, vector unsigned int);
7994int vec_any_ne (vector bool int, vector bool int);
7995int vec_any_ne (vector bool int, vector unsigned int);
7996int vec_any_ne (vector bool int, vector signed int);
7997int vec_any_ne (vector float, vector float);
7998
7999int vec_any_nge (vector float, vector float);
8000
8001int vec_any_ngt (vector float, vector float);
8002
8003int vec_any_nle (vector float, vector float);
8004
8005int vec_any_nlt (vector float, vector float);
8006
8007int vec_any_numeric (vector float);
8008
8009int vec_any_out (vector float, vector float);
8010@end smallexample
8011
8012@node Pragmas
8013@section Pragmas Accepted by GCC
8014@cindex pragmas
8015@cindex #pragma
8016
8017GCC supports several types of pragmas, primarily in order to compile
8018code originally written for other compilers.  Note that in general
8019we do not recommend the use of pragmas; @xref{Function Attributes},
8020for further explanation.
8021
8022@menu
8023* ARM Pragmas::
8024* RS/6000 and PowerPC Pragmas::
8025* Darwin Pragmas::
8026* Solaris Pragmas::
8027* Tru64 Pragmas::
8028@end menu
8029
8030@node ARM Pragmas
8031@subsection ARM Pragmas
8032
8033The ARM target defines pragmas for controlling the default addition of
8034@code{long_call} and @code{short_call} attributes to functions.
8035@xref{Function Attributes}, for information about the effects of these
8036attributes.
8037
8038@table @code
8039@item long_calls
8040@cindex pragma, long_calls
8041Set all subsequent functions to have the @code{long_call} attribute.
8042
8043@item no_long_calls
8044@cindex pragma, no_long_calls
8045Set all subsequent functions to have the @code{short_call} attribute.
8046
8047@item long_calls_off
8048@cindex pragma, long_calls_off
8049Do not affect the @code{long_call} or @code{short_call} attributes of
8050subsequent functions.
8051@end table
8052
8053@node RS/6000 and PowerPC Pragmas
8054@subsection RS/6000 and PowerPC Pragmas
8055
8056The RS/6000 and PowerPC targets define one pragma for controlling
8057whether or not the @code{longcall} attribute is added to function
8058declarations by default.  This pragma overrides the @option{-mlongcall}
8059option, but not the @code{longcall} and @code{shortcall} attributes.
8060@xref{RS/6000 and PowerPC Options}, for more information about when long
8061calls are and are not necessary.
8062
8063@table @code
8064@item longcall (1)
8065@cindex pragma, longcall
8066Apply the @code{longcall} attribute to all subsequent function
8067declarations.
8068
8069@item longcall (0)
8070Do not apply the @code{longcall} attribute to subsequent function
8071declarations.
8072@end table
8073
8074@c Describe c4x pragmas here.
8075@c Describe h8300 pragmas here.
8076@c Describe i370 pragmas here.
8077@c Describe i960 pragmas here.
8078@c Describe sh pragmas here.
8079@c Describe v850 pragmas here.
8080
8081@node Darwin Pragmas
8082@subsection Darwin Pragmas
8083
8084The following pragmas are available for all architectures running the
8085Darwin operating system.  These are useful for compatibility with other
8086Mac OS compilers.
8087
8088@table @code
8089@item mark @var{tokens}@dots{}
8090@cindex pragma, mark
8091This pragma is accepted, but has no effect.
8092
8093@item options align=@var{alignment}
8094@cindex pragma, options align
8095This pragma sets the alignment of fields in structures.  The values of
8096@var{alignment} may be @code{mac68k}, to emulate m68k alignment, or
8097@code{power}, to emulate PowerPC alignment.  Uses of this pragma nest
8098properly; to restore the previous setting, use @code{reset} for the
8099@var{alignment}.
8100
8101@item segment @var{tokens}@dots{}
8102@cindex pragma, segment
8103This pragma is accepted, but has no effect.
8104
8105@item unused (@var{var} [, @var{var}]@dots{})
8106@cindex pragma, unused
8107This pragma declares variables to be possibly unused.  GCC will not
8108produce warnings for the listed variables.  The effect is similar to
8109that of the @code{unused} attribute, except that this pragma may appear
8110anywhere within the variables' scopes.
8111@end table
8112
8113@node Solaris Pragmas
8114@subsection Solaris Pragmas
8115
8116For compatibility with the SunPRO compiler, the following pragma
8117is supported.
8118
8119@table @code
8120@item redefine_extname @var{oldname} @var{newname}
8121@cindex pragma, redefine_extname
8122
8123This pragma gives the C function @var{oldname} the assembler label
8124@var{newname}.  The pragma must appear before the function declaration.
8125This pragma is equivalent to the asm labels extension (@pxref{Asm
8126Labels}).  The preprocessor defines @code{__PRAGMA_REDEFINE_EXTNAME}
8127if the pragma is available.
8128@end table
8129
8130@node Tru64 Pragmas
8131@subsection Tru64 Pragmas
8132
8133For compatibility with the Compaq C compiler, the following pragma
8134is supported.
8135
8136@table @code
8137@item extern_prefix @var{string}
8138@cindex pragma, extern_prefix
8139
8140This pragma renames all subsequent function and variable declarations
8141such that @var{string} is prepended to the name.  This effect may be
8142terminated by using another @code{extern_prefix} pragma with the
8143empty string.
8144
8145This pragma is similar in intent to to the asm labels extension
8146(@pxref{Asm Labels}) in that the system programmer wants to change
8147the assembly-level ABI without changing the source-level API.  The
8148preprocessor defines @code{__PRAGMA_EXTERN_PREFIX} if the pragma is
8149available.
8150@end table
8151
8152@node Unnamed Fields
8153@section Unnamed struct/union fields within structs/unions.
8154@cindex struct
8155@cindex union
8156
8157For compatibility with other compilers, GCC allows you to define
8158a structure or union that contains, as fields, structures and unions
8159without names.  For example:
8160
8161@smallexample
8162struct @{
8163  int a;
8164  union @{
8165    int b;
8166    float c;
8167  @};
8168  int d;
8169@} foo;
8170@end smallexample
8171
8172In this example, the user would be able to access members of the unnamed
8173union with code like @samp{foo.b}.  Note that only unnamed structs and
8174unions are allowed, you may not have, for example, an unnamed
8175@code{int}.
8176
8177You must never create such structures that cause ambiguous field definitions.
8178For example, this structure:
8179
8180@smallexample
8181struct @{
8182  int a;
8183  struct @{
8184    int a;
8185  @};
8186@} foo;
8187@end smallexample
8188
8189It is ambiguous which @code{a} is being referred to with @samp{foo.a}.
8190Such constructs are not supported and must be avoided.  In the future,
8191such constructs may be detected and treated as compilation errors.
8192
8193@node Thread-Local
8194@section Thread-Local Storage
8195@cindex Thread-Local Storage
8196@cindex @acronym{TLS}
8197@cindex __thread
8198
8199Thread-local storage (@acronym{TLS}) is a mechanism by which variables
8200are allocated such that there is one instance of the variable per extant
8201thread.  The run-time model GCC uses to implement this originates
8202in the IA-64 processor-specific ABI, but has since been migrated
8203to other processors as well.  It requires significant support from
8204the linker (@command{ld}), dynamic linker (@command{ld.so}), and
8205system libraries (@file{libc.so} and @file{libpthread.so}), so it
8206is not available everywhere.
8207
8208At the user level, the extension is visible with a new storage
8209class keyword: @code{__thread}.  For example:
8210
8211@smallexample
8212__thread int i;
8213extern __thread struct state s;
8214static __thread char *p;
8215@end smallexample
8216
8217The @code{__thread} specifier may be used alone, with the @code{extern}
8218or @code{static} specifiers, but with no other storage class specifier.
8219When used with @code{extern} or @code{static}, @code{__thread} must appear
8220immediately after the other storage class specifier.
8221
8222The @code{__thread} specifier may be applied to any global, file-scoped
8223static, function-scoped static, or static data member of a class.  It may
8224not be applied to block-scoped automatic or non-static data member.
8225
8226When the address-of operator is applied to a thread-local variable, it is
8227evaluated at run-time and returns the address of the current thread's
8228instance of that variable.  An address so obtained may be used by any
8229thread.  When a thread terminates, any pointers to thread-local variables
8230in that thread become invalid.
8231
8232No static initialization may refer to the address of a thread-local variable.
8233
8234In C++, if an initializer is present for a thread-local variable, it must
8235be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++
8236standard.
8237
8238See @uref{http://people.redhat.com/drepper/tls.pdf,
8239ELF Handling For Thread-Local Storage} for a detailed explanation of
8240the four thread-local storage addressing models, and how the run-time
8241is expected to function.
8242
8243@menu
8244* C99 Thread-Local Edits::
8245* C++98 Thread-Local Edits::
8246@end menu
8247
8248@node C99 Thread-Local Edits
8249@subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage
8250
8251The following are a set of changes to ISO/IEC 9899:1999 (aka C99)
8252that document the exact semantics of the language extension.
8253
8254@itemize @bullet
8255@item
8256@cite{5.1.2  Execution environments}
8257
8258Add new text after paragraph 1
8259
8260@quotation
8261Within either execution environment, a @dfn{thread} is a flow of
8262control within a program.  It is implementation defined whether
8263or not there may be more than one thread associated with a program.
8264It is implementation defined how threads beyond the first are
8265created, the name and type of the function called at thread
8266startup, and how threads may be terminated.  However, objects
8267with thread storage duration shall be initialized before thread
8268startup.
8269@end quotation
8270
8271@item
8272@cite{6.2.4  Storage durations of objects}
8273
8274Add new text before paragraph 3
8275
8276@quotation
8277An object whose identifier is declared with the storage-class
8278specifier @w{@code{__thread}} has @dfn{thread storage duration}.
8279Its lifetime is the entire execution of the thread, and its
8280stored value is initialized only once, prior to thread startup.
8281@end quotation
8282
8283@item
8284@cite{6.4.1  Keywords}
8285
8286Add @code{__thread}.
8287
8288@item
8289@cite{6.7.1  Storage-class specifiers}
8290
8291Add @code{__thread} to the list of storage class specifiers in
8292paragraph 1.
8293
8294Change paragraph 2 to
8295
8296@quotation
8297With the exception of @code{__thread}, at most one storage-class
8298specifier may be given [@dots{}].  The @code{__thread} specifier may
8299be used alone, or immediately following @code{extern} or
8300@code{static}.
8301@end quotation
8302
8303Add new text after paragraph 6
8304
8305@quotation
8306The declaration of an identifier for a variable that has
8307block scope that specifies @code{__thread} shall also
8308specify either @code{extern} or @code{static}.
8309
8310The @code{__thread} specifier shall be used only with
8311variables.
8312@end quotation
8313@end itemize
8314
8315@node C++98 Thread-Local Edits
8316@subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage
8317
8318The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
8319that document the exact semantics of the language extension.
8320
8321@itemize @bullet
8322@item
8323@b{[intro.execution]}
8324
8325New text after paragraph 4
8326
8327@quotation
8328A @dfn{thread} is a flow of control within the abstract machine.
8329It is implementation defined whether or not there may be more than
8330one thread.
8331@end quotation
8332
8333New text after paragraph 7
8334
8335@quotation
8336It is unspecified whether additional action must be taken to
8337ensure when and whether side effects are visible to other threads.
8338@end quotation
8339
8340@item
8341@b{[lex.key]}
8342
8343Add @code{__thread}.
8344
8345@item
8346@b{[basic.start.main]}
8347
8348Add after paragraph 5
8349
8350@quotation
8351The thread that begins execution at the @code{main} function is called
8352the @dfn{main thread}.  It is implementation defined how functions
8353beginning threads other than the main thread are designated or typed.
8354A function so designated, as well as the @code{main} function, is called
8355a @dfn{thread startup function}.  It is implementation defined what
8356happens if a thread startup function returns.  It is implementation
8357defined what happens to other threads when any thread calls @code{exit}.
8358@end quotation
8359
8360@item
8361@b{[basic.start.init]}
8362
8363Add after paragraph 4
8364
8365@quotation
8366The storage for an object of thread storage duration shall be
8367statically initialized before the first statement of the thread startup
8368function.  An object of thread storage duration shall not require
8369dynamic initialization.
8370@end quotation
8371
8372@item
8373@b{[basic.start.term]}
8374
8375Add after paragraph 3
8376
8377@quotation
8378The type of an object with thread storage duration shall not have a
8379non-trivial destructor, nor shall it be an array type whose elements
8380(directly or indirectly) have non-trivial destructors.
8381@end quotation
8382
8383@item
8384@b{[basic.stc]}
8385
8386Add ``thread storage duration'' to the list in paragraph 1.
8387
8388Change paragraph 2
8389
8390@quotation
8391Thread, static, and automatic storage durations are associated with
8392objects introduced by declarations [@dots{}].
8393@end quotation
8394
8395Add @code{__thread} to the list of specifiers in paragraph 3.
8396
8397@item
8398@b{[basic.stc.thread]}
8399
8400New section before @b{[basic.stc.static]}
8401
8402@quotation
8403The keyword @code{__thread} applied to a non-local object gives the
8404object thread storage duration.
8405
8406A local variable or class data member declared both @code{static}
8407and @code{__thread} gives the variable or member thread storage
8408duration.
8409@end quotation
8410
8411@item
8412@b{[basic.stc.static]}
8413
8414Change paragraph 1
8415
8416@quotation
8417All objects which have neither thread storage duration, dynamic
8418storage duration nor are local [@dots{}].
8419@end quotation
8420
8421@item
8422@b{[dcl.stc]}
8423
8424Add @code{__thread} to the list in paragraph 1.
8425
8426Change paragraph 1
8427
8428@quotation
8429With the exception of @code{__thread}, at most one
8430@var{storage-class-specifier} shall appear in a given
8431@var{decl-specifier-seq}.  The @code{__thread} specifier may
8432be used alone, or immediately following the @code{extern} or
8433@code{static} specifiers.  [@dots{}]
8434@end quotation
8435
8436Add after paragraph 5
8437
8438@quotation
8439The @code{__thread} specifier can be applied only to the names of objects
8440and to anonymous unions.
8441@end quotation
8442
8443@item
8444@b{[class.mem]}
8445
8446Add after paragraph 6
8447
8448@quotation
8449Non-@code{static} members shall not be @code{__thread}.
8450@end quotation
8451@end itemize
8452
8453@node C++ Extensions
8454@chapter Extensions to the C++ Language
8455@cindex extensions, C++ language
8456@cindex C++ language extensions
8457
8458The GNU compiler provides these extensions to the C++ language (and you
8459can also use most of the C language extensions in your C++ programs).  If you
8460want to write code that checks whether these features are available, you can
8461test for the GNU compiler the same way as for C programs: check for a
8462predefined macro @code{__GNUC__}.  You can also use @code{__GNUG__} to
8463test specifically for GNU C++ (@pxref{Common Predefined Macros,,
8464Predefined Macros,cpp,The GNU C Preprocessor}).
8465
8466@menu
8467* Min and Max::		C++ Minimum and maximum operators.
8468* Volatiles::		What constitutes an access to a volatile object.
8469* Restricted Pointers:: C99 restricted pointers and references.
8470* Vague Linkage::       Where G++ puts inlines, vtables and such.
8471* C++ Interface::       You can use a single C++ header file for both
8472                        declarations and definitions.
8473* Template Instantiation:: Methods for ensuring that exactly one copy of
8474                        each needed template instantiation is emitted.
8475* Bound member functions:: You can extract a function pointer to the
8476                        method denoted by a @samp{->*} or @samp{.*} expression.
8477* C++ Attributes::      Variable, function, and type attributes for C++ only.
8478* Strong Using::      Strong using-directives for namespace composition.
8479* Offsetof::            Special syntax for implementing @code{offsetof}.
8480* Java Exceptions::     Tweaking exception handling to work with Java.
8481* Deprecated Features:: Things will disappear from g++.
8482* Backwards Compatibility:: Compatibilities with earlier definitions of C++.
8483@end menu
8484
8485@node Min and Max
8486@section Minimum and Maximum Operators in C++
8487
8488It is very convenient to have operators which return the ``minimum'' or the
8489``maximum'' of two arguments.  In GNU C++ (but not in GNU C),
8490
8491@table @code
8492@item @var{a} <? @var{b}
8493@findex <?
8494@cindex minimum operator
8495is the @dfn{minimum}, returning the smaller of the numeric values
8496@var{a} and @var{b};
8497
8498@item @var{a} >? @var{b}
8499@findex >?
8500@cindex maximum operator
8501is the @dfn{maximum}, returning the larger of the numeric values @var{a}
8502and @var{b}.
8503@end table
8504
8505These operations are not primitive in ordinary C++, since you can
8506use a macro to return the minimum of two things in C++, as in the
8507following example.
8508
8509@smallexample
8510#define MIN(X,Y) ((X) < (Y) ? : (X) : (Y))
8511@end smallexample
8512
8513@noindent
8514You might then use @w{@samp{int min = MIN (i, j);}} to set @var{min} to
8515the minimum value of variables @var{i} and @var{j}.
8516
8517However, side effects in @code{X} or @code{Y} may cause unintended
8518behavior.  For example, @code{MIN (i++, j++)} will fail, incrementing
8519the smaller counter twice.  The GNU C @code{typeof} extension allows you
8520to write safe macros that avoid this kind of problem (@pxref{Typeof}).
8521However, writing @code{MIN} and @code{MAX} as macros also forces you to
8522use function-call notation for a fundamental arithmetic operation.
8523Using GNU C++ extensions, you can write @w{@samp{int min = i <? j;}}
8524instead.
8525
8526Since @code{<?} and @code{>?} are built into the compiler, they properly
8527handle expressions with side-effects;  @w{@samp{int min = i++ <? j++;}}
8528works correctly.
8529
8530@node Volatiles
8531@section When is a Volatile Object Accessed?
8532@cindex accessing volatiles
8533@cindex volatile read
8534@cindex volatile write
8535@cindex volatile access
8536
8537Both the C and C++ standard have the concept of volatile objects.  These
8538are normally accessed by pointers and used for accessing hardware.  The
8539standards encourage compilers to refrain from optimizations
8540concerning accesses to volatile objects that it might perform on
8541non-volatile objects.  The C standard leaves it implementation defined
8542as to what constitutes a volatile access.  The C++ standard omits to
8543specify this, except to say that C++ should behave in a similar manner
8544to C with respect to volatiles, where possible.  The minimum either
8545standard specifies is that at a sequence point all previous accesses to
8546volatile objects have stabilized and no subsequent accesses have
8547occurred.  Thus an implementation is free to reorder and combine
8548volatile accesses which occur between sequence points, but cannot do so
8549for accesses across a sequence point.  The use of volatiles does not
8550allow you to violate the restriction on updating objects multiple times
8551within a sequence point.
8552
8553In most expressions, it is intuitively obvious what is a read and what is
8554a write.  For instance
8555
8556@smallexample
8557volatile int *dst = @var{somevalue};
8558volatile int *src = @var{someothervalue};
8559*dst = *src;
8560@end smallexample
8561
8562@noindent
8563will cause a read of the volatile object pointed to by @var{src} and stores the
8564value into the volatile object pointed to by @var{dst}.  There is no
8565guarantee that these reads and writes are atomic, especially for objects
8566larger than @code{int}.
8567
8568Less obvious expressions are where something which looks like an access
8569is used in a void context.  An example would be,
8570
8571@smallexample
8572volatile int *src = @var{somevalue};
8573*src;
8574@end smallexample
8575
8576With C, such expressions are rvalues, and as rvalues cause a read of
8577the object, GCC interprets this as a read of the volatile being pointed
8578to.  The C++ standard specifies that such expressions do not undergo
8579lvalue to rvalue conversion, and that the type of the dereferenced
8580object may be incomplete.  The C++ standard does not specify explicitly
8581that it is this lvalue to rvalue conversion which is responsible for
8582causing an access.  However, there is reason to believe that it is,
8583because otherwise certain simple expressions become undefined.  However,
8584because it would surprise most programmers, G++ treats dereferencing a
8585pointer to volatile object of complete type in a void context as a read
8586of the object.  When the object has incomplete type, G++ issues a
8587warning.
8588
8589@smallexample
8590struct S;
8591struct T @{int m;@};
8592volatile S *ptr1 = @var{somevalue};
8593volatile T *ptr2 = @var{somevalue};
8594*ptr1;
8595*ptr2;
8596@end smallexample
8597
8598In this example, a warning is issued for @code{*ptr1}, and @code{*ptr2}
8599causes a read of the object pointed to.  If you wish to force an error on
8600the first case, you must force a conversion to rvalue with, for instance
8601a static cast, @code{static_cast<S>(*ptr1)}.
8602
8603When using a reference to volatile, G++ does not treat equivalent
8604expressions as accesses to volatiles, but instead issues a warning that
8605no volatile is accessed.  The rationale for this is that otherwise it
8606becomes difficult to determine where volatile access occur, and not
8607possible to ignore the return value from functions returning volatile
8608references.  Again, if you wish to force a read, cast the reference to
8609an rvalue.
8610
8611@node Restricted Pointers
8612@section Restricting Pointer Aliasing
8613@cindex restricted pointers
8614@cindex restricted references
8615@cindex restricted this pointer
8616
8617As with the C front end, G++ understands the C99 feature of restricted pointers,
8618specified with the @code{__restrict__}, or @code{__restrict} type
8619qualifier.  Because you cannot compile C++ by specifying the @option{-std=c99}
8620language flag, @code{restrict} is not a keyword in C++.
8621
8622In addition to allowing restricted pointers, you can specify restricted
8623references, which indicate that the reference is not aliased in the local
8624context.
8625
8626@smallexample
8627void fn (int *__restrict__ rptr, int &__restrict__ rref)
8628@{
8629  /* @r{@dots{}} */
8630@}
8631@end smallexample
8632
8633@noindent
8634In the body of @code{fn}, @var{rptr} points to an unaliased integer and
8635@var{rref} refers to a (different) unaliased integer.
8636
8637You may also specify whether a member function's @var{this} pointer is
8638unaliased by using @code{__restrict__} as a member function qualifier.
8639
8640@smallexample
8641void T::fn () __restrict__
8642@{
8643  /* @r{@dots{}} */
8644@}
8645@end smallexample
8646
8647@noindent
8648Within the body of @code{T::fn}, @var{this} will have the effective
8649definition @code{T *__restrict__ const this}.  Notice that the
8650interpretation of a @code{__restrict__} member function qualifier is
8651different to that of @code{const} or @code{volatile} qualifier, in that it
8652is applied to the pointer rather than the object.  This is consistent with
8653other compilers which implement restricted pointers.
8654
8655As with all outermost parameter qualifiers, @code{__restrict__} is
8656ignored in function definition matching.  This means you only need to
8657specify @code{__restrict__} in a function definition, rather than
8658in a function prototype as well.
8659
8660@node Vague Linkage
8661@section Vague Linkage
8662@cindex vague linkage
8663
8664There are several constructs in C++ which require space in the object
8665file but are not clearly tied to a single translation unit.  We say that
8666these constructs have ``vague linkage''.  Typically such constructs are
8667emitted wherever they are needed, though sometimes we can be more
8668clever.
8669
8670@table @asis
8671@item Inline Functions
8672Inline functions are typically defined in a header file which can be
8673included in many different compilations.  Hopefully they can usually be
8674inlined, but sometimes an out-of-line copy is necessary, if the address
8675of the function is taken or if inlining fails.  In general, we emit an
8676out-of-line copy in all translation units where one is needed.  As an
8677exception, we only emit inline virtual functions with the vtable, since
8678it will always require a copy.
8679
8680Local static variables and string constants used in an inline function
8681are also considered to have vague linkage, since they must be shared
8682between all inlined and out-of-line instances of the function.
8683
8684@item VTables
8685@cindex vtable
8686C++ virtual functions are implemented in most compilers using a lookup
8687table, known as a vtable.  The vtable contains pointers to the virtual
8688functions provided by a class, and each object of the class contains a
8689pointer to its vtable (or vtables, in some multiple-inheritance
8690situations).  If the class declares any non-inline, non-pure virtual
8691functions, the first one is chosen as the ``key method'' for the class,
8692and the vtable is only emitted in the translation unit where the key
8693method is defined.
8694
8695@emph{Note:} If the chosen key method is later defined as inline, the
8696vtable will still be emitted in every translation unit which defines it.
8697Make sure that any inline virtuals are declared inline in the class
8698body, even if they are not defined there.
8699
8700@item type_info objects
8701@cindex type_info
8702@cindex RTTI
8703C++ requires information about types to be written out in order to
8704implement @samp{dynamic_cast}, @samp{typeid} and exception handling.
8705For polymorphic classes (classes with virtual functions), the type_info
8706object is written out along with the vtable so that @samp{dynamic_cast}
8707can determine the dynamic type of a class object at runtime.  For all
8708other types, we write out the type_info object when it is used: when
8709applying @samp{typeid} to an expression, throwing an object, or
8710referring to a type in a catch clause or exception specification.
8711
8712@item Template Instantiations
8713Most everything in this section also applies to template instantiations,
8714but there are other options as well.
8715@xref{Template Instantiation,,Where's the Template?}.
8716
8717@end table
8718
8719When used with GNU ld version 2.8 or later on an ELF system such as
8720GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
8721these constructs will be discarded at link time.  This is known as
8722COMDAT support.
8723
8724On targets that don't support COMDAT, but do support weak symbols, GCC
8725will use them.  This way one copy will override all the others, but
8726the unused copies will still take up space in the executable.
8727
8728For targets which do not support either COMDAT or weak symbols,
8729most entities with vague linkage will be emitted as local symbols to
8730avoid duplicate definition errors from the linker.  This will not happen
8731for local statics in inlines, however, as having multiple copies will
8732almost certainly break things.
8733
8734@xref{C++ Interface,,Declarations and Definitions in One Header}, for
8735another way to control placement of these constructs.
8736
8737@node C++ Interface
8738@section #pragma interface and implementation
8739
8740@cindex interface and implementation headers, C++
8741@cindex C++ interface and implementation headers
8742@cindex pragmas, interface and implementation
8743
8744@code{#pragma interface} and @code{#pragma implementation} provide the
8745user with a way of explicitly directing the compiler to emit entities
8746with vague linkage (and debugging information) in a particular
8747translation unit.
8748
8749@emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in
8750most cases, because of COMDAT support and the ``key method'' heuristic
8751mentioned in @ref{Vague Linkage}.  Using them can actually cause your
8752program to grow due to unnecesary out-of-line copies of inline
8753functions.  Currently the only benefit of these @code{#pragma}s is
8754reduced duplication of debugging information, and that should be
8755addressed soon on DWARF 2 targets with the use of COMDAT sections.
8756
8757@table @code
8758@item #pragma interface
8759@itemx #pragma interface "@var{subdir}/@var{objects}.h"
8760@kindex #pragma interface
8761Use this directive in @emph{header files} that define object classes, to save
8762space in most of the object files that use those classes.  Normally,
8763local copies of certain information (backup copies of inline member
8764functions, debugging information, and the internal tables that implement
8765virtual functions) must be kept in each object file that includes class
8766definitions.  You can use this pragma to avoid such duplication.  When a
8767header file containing @samp{#pragma interface} is included in a
8768compilation, this auxiliary information will not be generated (unless
8769the main input source file itself uses @samp{#pragma implementation}).
8770Instead, the object files will contain references to be resolved at link
8771time.
8772
8773The second form of this directive is useful for the case where you have
8774multiple headers with the same name in different directories.  If you
8775use this form, you must specify the same string to @samp{#pragma
8776implementation}.
8777
8778@item #pragma implementation
8779@itemx #pragma implementation "@var{objects}.h"
8780@kindex #pragma implementation
8781Use this pragma in a @emph{main input file}, when you want full output from
8782included header files to be generated (and made globally visible).  The
8783included header file, in turn, should use @samp{#pragma interface}.
8784Backup copies of inline member functions, debugging information, and the
8785internal tables used to implement virtual functions are all generated in
8786implementation files.
8787
8788@cindex implied @code{#pragma implementation}
8789@cindex @code{#pragma implementation}, implied
8790@cindex naming convention, implementation headers
8791If you use @samp{#pragma implementation} with no argument, it applies to
8792an include file with the same basename@footnote{A file's @dfn{basename}
8793was the name stripped of all leading path information and of trailing
8794suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source
8795file.  For example, in @file{allclass.cc}, giving just
8796@samp{#pragma implementation}
8797by itself is equivalent to @samp{#pragma implementation "allclass.h"}.
8798
8799In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as
8800an implementation file whenever you would include it from
8801@file{allclass.cc} even if you never specified @samp{#pragma
8802implementation}.  This was deemed to be more trouble than it was worth,
8803however, and disabled.
8804
8805Use the string argument if you want a single implementation file to
8806include code from multiple header files.  (You must also use
8807@samp{#include} to include the header file; @samp{#pragma
8808implementation} only specifies how to use the file---it doesn't actually
8809include it.)
8810
8811There is no way to split up the contents of a single header file into
8812multiple implementation files.
8813@end table
8814
8815@cindex inlining and C++ pragmas
8816@cindex C++ pragmas, effect on inlining
8817@cindex pragmas in C++, effect on inlining
8818@samp{#pragma implementation} and @samp{#pragma interface} also have an
8819effect on function inlining.
8820
8821If you define a class in a header file marked with @samp{#pragma
8822interface}, the effect on an inline function defined in that class is
8823similar to an explicit @code{extern} declaration---the compiler emits
8824no code at all to define an independent version of the function.  Its
8825definition is used only for inlining with its callers.
8826
8827@opindex fno-implement-inlines
8828Conversely, when you include the same header file in a main source file
8829that declares it as @samp{#pragma implementation}, the compiler emits
8830code for the function itself; this defines a version of the function
8831that can be found via pointers (or by callers compiled without
8832inlining).  If all calls to the function can be inlined, you can avoid
8833emitting the function by compiling with @option{-fno-implement-inlines}.
8834If any calls were not inlined, you will get linker errors.
8835
8836@node Template Instantiation
8837@section Where's the Template?
8838@cindex template instantiation
8839
8840C++ templates are the first language feature to require more
8841intelligence from the environment than one usually finds on a UNIX
8842system.  Somehow the compiler and linker have to make sure that each
8843template instance occurs exactly once in the executable if it is needed,
8844and not at all otherwise.  There are two basic approaches to this
8845problem, which are referred to as the Borland model and the Cfront model.
8846
8847@table @asis
8848@item Borland model
8849Borland C++ solved the template instantiation problem by adding the code
8850equivalent of common blocks to their linker; the compiler emits template
8851instances in each translation unit that uses them, and the linker
8852collapses them together.  The advantage of this model is that the linker
8853only has to consider the object files themselves; there is no external
8854complexity to worry about.  This disadvantage is that compilation time
8855is increased because the template code is being compiled repeatedly.
8856Code written for this model tends to include definitions of all
8857templates in the header file, since they must be seen to be
8858instantiated.
8859
8860@item Cfront model
8861The AT&T C++ translator, Cfront, solved the template instantiation
8862problem by creating the notion of a template repository, an
8863automatically maintained place where template instances are stored.  A
8864more modern version of the repository works as follows: As individual
8865object files are built, the compiler places any template definitions and
8866instantiations encountered in the repository.  At link time, the link
8867wrapper adds in the objects in the repository and compiles any needed
8868instances that were not previously emitted.  The advantages of this
8869model are more optimal compilation speed and the ability to use the
8870system linker; to implement the Borland model a compiler vendor also
8871needs to replace the linker.  The disadvantages are vastly increased
8872complexity, and thus potential for error; for some code this can be
8873just as transparent, but in practice it can been very difficult to build
8874multiple programs in one directory and one program in multiple
8875directories.  Code written for this model tends to separate definitions
8876of non-inline member templates into a separate file, which should be
8877compiled separately.
8878@end table
8879
8880When used with GNU ld version 2.8 or later on an ELF system such as
8881GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
8882Borland model.  On other systems, G++ implements neither automatic
8883model.
8884
8885A future version of G++ will support a hybrid model whereby the compiler
8886will emit any instantiations for which the template definition is
8887included in the compile, and store template definitions and
8888instantiation context information into the object file for the rest.
8889The link wrapper will extract that information as necessary and invoke
8890the compiler to produce the remaining instantiations.  The linker will
8891then combine duplicate instantiations.
8892
8893In the mean time, you have the following options for dealing with
8894template instantiations:
8895
8896@enumerate
8897@item
8898@opindex frepo
8899Compile your template-using code with @option{-frepo}.  The compiler will
8900generate files with the extension @samp{.rpo} listing all of the
8901template instantiations used in the corresponding object files which
8902could be instantiated there; the link wrapper, @samp{collect2}, will
8903then update the @samp{.rpo} files to tell the compiler where to place
8904those instantiations and rebuild any affected object files.  The
8905link-time overhead is negligible after the first pass, as the compiler
8906will continue to place the instantiations in the same files.
8907
8908This is your best option for application code written for the Borland
8909model, as it will just work.  Code written for the Cfront model will
8910need to be modified so that the template definitions are available at
8911one or more points of instantiation; usually this is as simple as adding
8912@code{#include <tmethods.cc>} to the end of each template header.
8913
8914For library code, if you want the library to provide all of the template
8915instantiations it needs, just try to link all of its object files
8916together; the link will fail, but cause the instantiations to be
8917generated as a side effect.  Be warned, however, that this may cause
8918conflicts if multiple libraries try to provide the same instantiations.
8919For greater control, use explicit instantiation as described in the next
8920option.
8921
8922@item
8923@opindex fno-implicit-templates
8924Compile your code with @option{-fno-implicit-templates} to disable the
8925implicit generation of template instances, and explicitly instantiate
8926all the ones you use.  This approach requires more knowledge of exactly
8927which instances you need than do the others, but it's less
8928mysterious and allows greater control.  You can scatter the explicit
8929instantiations throughout your program, perhaps putting them in the
8930translation units where the instances are used or the translation units
8931that define the templates themselves; you can put all of the explicit
8932instantiations you need into one big file; or you can create small files
8933like
8934
8935@smallexample
8936#include "Foo.h"
8937#include "Foo.cc"
8938
8939template class Foo<int>;
8940template ostream& operator <<
8941                (ostream&, const Foo<int>&);
8942@end smallexample
8943
8944for each of the instances you need, and create a template instantiation
8945library from those.
8946
8947If you are using Cfront-model code, you can probably get away with not
8948using @option{-fno-implicit-templates} when compiling files that don't
8949@samp{#include} the member template definitions.
8950
8951If you use one big file to do the instantiations, you may want to
8952compile it without @option{-fno-implicit-templates} so you get all of the
8953instances required by your explicit instantiations (but not by any
8954other files) without having to specify them as well.
8955
8956G++ has extended the template instantiation syntax given in the ISO
8957standard to allow forward declaration of explicit instantiations
8958(with @code{extern}), instantiation of the compiler support data for a
8959template class (i.e.@: the vtable) without instantiating any of its
8960members (with @code{inline}), and instantiation of only the static data
8961members of a template class, without the support data or member
8962functions (with (@code{static}):
8963
8964@smallexample
8965extern template int max (int, int);
8966inline template class Foo<int>;
8967static template class Foo<int>;
8968@end smallexample
8969
8970@item
8971Do nothing.  Pretend G++ does implement automatic instantiation
8972management.  Code written for the Borland model will work fine, but
8973each translation unit will contain instances of each of the templates it
8974uses.  In a large program, this can lead to an unacceptable amount of code
8975duplication.
8976@end enumerate
8977
8978@node Bound member functions
8979@section Extracting the function pointer from a bound pointer to member function
8980@cindex pmf
8981@cindex pointer to member function
8982@cindex bound pointer to member function
8983
8984In C++, pointer to member functions (PMFs) are implemented using a wide
8985pointer of sorts to handle all the possible call mechanisms; the PMF
8986needs to store information about how to adjust the @samp{this} pointer,
8987and if the function pointed to is virtual, where to find the vtable, and
8988where in the vtable to look for the member function.  If you are using
8989PMFs in an inner loop, you should really reconsider that decision.  If
8990that is not an option, you can extract the pointer to the function that
8991would be called for a given object/PMF pair and call it directly inside
8992the inner loop, to save a bit of time.
8993
8994Note that you will still be paying the penalty for the call through a
8995function pointer; on most modern architectures, such a call defeats the
8996branch prediction features of the CPU@.  This is also true of normal
8997virtual function calls.
8998
8999The syntax for this extension is
9000
9001@smallexample
9002extern A a;
9003extern int (A::*fp)();
9004typedef int (*fptr)(A *);
9005
9006fptr p = (fptr)(a.*fp);
9007@end smallexample
9008
9009For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}),
9010no object is needed to obtain the address of the function.  They can be
9011converted to function pointers directly:
9012
9013@smallexample
9014fptr p1 = (fptr)(&A::foo);
9015@end smallexample
9016
9017@opindex Wno-pmf-conversions
9018You must specify @option{-Wno-pmf-conversions} to use this extension.
9019
9020@node C++ Attributes
9021@section C++-Specific Variable, Function, and Type Attributes
9022
9023Some attributes only make sense for C++ programs.
9024
9025@table @code
9026@item init_priority (@var{priority})
9027@cindex init_priority attribute
9028
9029
9030In Standard C++, objects defined at namespace scope are guaranteed to be
9031initialized in an order in strict accordance with that of their definitions
9032@emph{in a given translation unit}.  No guarantee is made for initializations
9033across translation units.  However, GNU C++ allows users to control the
9034order of initialization of objects defined at namespace scope with the
9035@code{init_priority} attribute by specifying a relative @var{priority},
9036a constant integral expression currently bounded between 101 and 65535
9037inclusive.  Lower numbers indicate a higher priority.
9038
9039In the following example, @code{A} would normally be created before
9040@code{B}, but the @code{init_priority} attribute has reversed that order:
9041
9042@smallexample
9043Some_Class  A  __attribute__ ((init_priority (2000)));
9044Some_Class  B  __attribute__ ((init_priority (543)));
9045@end smallexample
9046
9047@noindent
9048Note that the particular values of @var{priority} do not matter; only their
9049relative ordering.
9050
9051@item java_interface
9052@cindex java_interface attribute
9053
9054This type attribute informs C++ that the class is a Java interface.  It may
9055only be applied to classes declared within an @code{extern "Java"} block.
9056Calls to methods declared in this interface will be dispatched using GCJ's
9057interface table mechanism, instead of regular virtual table dispatch.
9058
9059@end table
9060
9061See also @xref{Strong Using}.
9062
9063@node Strong Using
9064@section Strong Using
9065
9066@strong{Caution:} The semantics of this extension are not fully
9067defined.  Users should refrain from using this extension as its
9068semantics may change subtly over time.  It is possible that this
9069extension wil be removed in future versions of G++.
9070
9071A using-directive with @code{__attribute ((strong))} is stronger
9072than a normal using-directive in two ways:
9073
9074@itemize @bullet
9075@item
9076Templates from the used namespace can be specialized as though they were members of the using namespace.
9077
9078@item
9079The using namespace is considered an associated namespace of all
9080templates in the used namespace for purposes of argument-dependent
9081name lookup.
9082@end itemize
9083
9084This is useful for composing a namespace transparently from
9085implementation namespaces.  For example:
9086
9087@smallexample
9088namespace std @{
9089  namespace debug @{
9090    template <class T> struct A @{ @};
9091  @}
9092  using namespace debug __attribute ((__strong__));
9093  template <> struct A<int> @{ @};   // ok to specialize
9094
9095  template <class T> void f (A<T>);
9096@}
9097
9098int main()
9099@{
9100  f (std::A<float>());             // lookup finds std::f
9101  f (std::A<int>());
9102@}
9103@end smallexample
9104
9105@node Offsetof
9106@section Offsetof
9107
9108G++ uses a syntactic extension to implement the @code{offsetof} macro.
9109
9110In particular:
9111
9112@smallexample
9113  __offsetof__ (expression)
9114@end smallexample
9115
9116is equivalent to the parenthesized expression, except that the
9117expression is considered an integral constant expression even if it
9118contains certain operators that are not normally permitted in an
9119integral constant expression.  Users should never use
9120@code{__offsetof__} directly; the only valid use of
9121@code{__offsetof__} is to implement the @code{offsetof} macro in
9122@code{<stddef.h>}.
9123
9124@node Java Exceptions
9125@section Java Exceptions
9126
9127The Java language uses a slightly different exception handling model
9128from C++.  Normally, GNU C++ will automatically detect when you are
9129writing C++ code that uses Java exceptions, and handle them
9130appropriately.  However, if C++ code only needs to execute destructors
9131when Java exceptions are thrown through it, GCC will guess incorrectly.
9132Sample problematic code is:
9133
9134@smallexample
9135  struct S @{ ~S(); @};
9136  extern void bar();    // is written in Java, and may throw exceptions
9137  void foo()
9138  @{
9139    S s;
9140    bar();
9141  @}
9142@end smallexample
9143
9144@noindent
9145The usual effect of an incorrect guess is a link failure, complaining of
9146a missing routine called @samp{__gxx_personality_v0}.
9147
9148You can inform the compiler that Java exceptions are to be used in a
9149translation unit, irrespective of what it might think, by writing
9150@samp{@w{#pragma GCC java_exceptions}} at the head of the file.  This
9151@samp{#pragma} must appear before any functions that throw or catch
9152exceptions, or run destructors when exceptions are thrown through them.
9153
9154You cannot mix Java and C++ exceptions in the same translation unit.  It
9155is believed to be safe to throw a C++ exception from one file through
9156another file compiled for the Java exception model, or vice versa, but
9157there may be bugs in this area.
9158
9159@node Deprecated Features
9160@section Deprecated Features
9161
9162In the past, the GNU C++ compiler was extended to experiment with new
9163features, at a time when the C++ language was still evolving.  Now that
9164the C++ standard is complete, some of those features are superseded by
9165superior alternatives.  Using the old features might cause a warning in
9166some cases that the feature will be dropped in the future.  In other
9167cases, the feature might be gone already.
9168
9169While the list below is not exhaustive, it documents some of the options
9170that are now deprecated:
9171
9172@table @code
9173@item -fexternal-templates
9174@itemx -falt-external-templates
9175These are two of the many ways for G++ to implement template
9176instantiation.  @xref{Template Instantiation}.  The C++ standard clearly
9177defines how template definitions have to be organized across
9178implementation units.  G++ has an implicit instantiation mechanism that
9179should work just fine for standard-conforming code.
9180
9181@item -fstrict-prototype
9182@itemx -fno-strict-prototype
9183Previously it was possible to use an empty prototype parameter list to
9184indicate an unspecified number of parameters (like C), rather than no
9185parameters, as C++ demands.  This feature has been removed, except where
9186it is required for backwards compatibility @xref{Backwards Compatibility}.
9187@end table
9188
9189The named return value extension has been deprecated, and is now
9190removed from G++.
9191
9192The use of initializer lists with new expressions has been deprecated,
9193and is now removed from G++.
9194
9195Floating and complex non-type template parameters have been deprecated,
9196and are now removed from G++.
9197
9198The implicit typename extension has been deprecated and is now
9199removed from G++.
9200
9201The use of default arguments in function pointers, function typedefs and
9202and other places where they are not permitted by the standard is
9203deprecated and will be removed from a future version of G++.
9204
9205@node Backwards Compatibility
9206@section Backwards Compatibility
9207@cindex Backwards Compatibility
9208@cindex ARM [Annotated C++ Reference Manual]
9209
9210Now that there is a definitive ISO standard C++, G++ has a specification
9211to adhere to.  The C++ language evolved over time, and features that
9212used to be acceptable in previous drafts of the standard, such as the ARM
9213[Annotated C++ Reference Manual], are no longer accepted.  In order to allow
9214compilation of C++ written to such drafts, G++ contains some backwards
9215compatibilities.  @emph{All such backwards compatibility features are
9216liable to disappear in future versions of G++.} They should be considered
9217deprecated @xref{Deprecated Features}.
9218
9219@table @code
9220@item For scope
9221If a variable is declared at for scope, it used to remain in scope until
9222the end of the scope which contained the for statement (rather than just
9223within the for scope).  G++ retains this, but issues a warning, if such a
9224variable is accessed outside the for scope.
9225
9226@item Implicit C language
9227Old C system header files did not contain an @code{extern "C" @{@dots{}@}}
9228scope to set the language.  On such systems, all header files are
9229implicitly scoped inside a C language scope.  Also, an empty prototype
9230@code{()} will be treated as an unspecified number of arguments, rather
9231than no arguments, as C++ demands.
9232@end table
9233