extend.texi revision 259563
194742Sobrien@c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1996, 1998, 1999, 2000, 294742Sobrien@c 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc. 394742Sobrien 495253Sru@c This is part of the GCC manual. 594742Sobrien@c For copying conditions, see the file gcc.texi. 696991Srwatson 796991Srwatson@node C Extensions 896991Srwatson@chapter Extensions to the C Language Family 9102773Srwatson@cindex extensions, C language 10102773Srwatson@cindex C language extensions 1194854Ssos 1294917Simp@opindex pedantic 13126445SobrienGNU C provides several language features not found in ISO standard C@. 1494917Simp(The @option{-pedantic} option directs GCC to print a warning message if 1594917Simpany of these features is used.) To test for the availability of these 1694917Simpfeatures in conditional compilation, check for a predefined macro 17117751Smarkm@code{__GNUC__}, which is always defined under GCC@. 18117751Smarkm 19116149SmarkmThese extensions are available in C. Most of them are also available 20116149Smarkmin C++. @xref{C++ Extensions,,Extensions to the C++ Language}, for 21125244Snectarextensions that apply @emph{only} to C++. 22125244Snectar 2394847SjhbSome features that are in ISO C99 but not C89 or C++ are also, as 2494847Sjhbextensions, accepted by GCC in C89 mode and in C++. 2594847Sjhb 26126337Svkashyap@menu 27128023Svkashyap* Statement Exprs:: Putting statements and declarations inside expressions. 2894855Sscottl* Local Labels:: Labels local to a block. 29126054Sscottl* Labels as Values:: Getting pointers to labels, and computed gotos. 30126054Sscottl* Nested Functions:: As in Algol and Pascal, lexical scoping of functions. 31126054Sscottl* Constructing Calls:: Dispatching a call to another function. 32126054Sscottl* Typeof:: @code{typeof}: referring to the type of an expression. 33126054Sscottl* Conditionals:: Omitting the middle operand of a @samp{?:} expression. 34126054Sscottl* Long Long:: Double-word integers---@code{long long int}. 3594915Sken* Complex:: Data types for complex numbers. 3699607Smjacob* Decimal Float:: Decimal Floating Types. 3794915Sken* Hex Floats:: Hexadecimal floating-point constants. 3894915Sken* Zero Length:: Zero-length arrays. 3994915Sken* Variable Length:: Arrays whose length is computed at run time. 4094915Sken* Empty Structures:: Structures with no members. 4194915Sken* Variadic Macros:: Macros with a variable number of arguments. 42105411Snjl* Escaped Newlines:: Slightly looser rules for escaped newlines. 4394915Sken* Subscripting:: Any array can be subscripted, even if not an lvalue. 4494915Sken* Pointer Arith:: Arithmetic on @code{void}-pointers and function pointers. 4594915Sken* Initializers:: Non-constant initializers. 4699607Smjacob* Compound Literals:: Compound literals give structures, unions 47106734Smjacob or arrays as values. 48128435Stackerman* Designated Inits:: Labeling elements of initializers. 4997611Sbillf* Cast to Union:: Casting to union type from any member of the union. 5094918Sgshapiro* Case Ranges:: `case 1 ... 9' and such. 5194918Sgshapiro* Mixed Declarations:: Mixing declarations and code. 5294918Sgshapiro* Function Attributes:: Declaring that functions have no side effects, 5394918Sgshapiro or that they can never return. 5494918Sgshapiro* Attribute Syntax:: Formal syntax for attributes. 55118316Smbr* Function Prototypes:: Prototype declarations and old-style definitions. 5694955Smurray* C++ Comments:: C++ comments are recognized. 5795054Snectar* Dollar Signs:: Dollar sign is allowed in identifiers. 58125080Scperciva* Character Escapes:: @samp{\e} stands for the character @key{ESC}. 59106187Sdes* Variable Attributes:: Specifying attributes of variables. 60106187Sdes* Type Attributes:: Specifying attributes of types. 6195455Sdes* Alignment:: Inquiring about the alignment of a type or variable. 6298750Sdes* Inline:: Defining inline functions (as fast as macros). 6399606Sdes* Extended Asm:: Assembler instructions with C expressions as operands. 6499606Sdes (With them you can define ``built-in'' functions.) 6599606Sdes* Constraints:: Constraints for asm operands 6696268Sgad* Asm Labels:: Specifying the assembler name to use for a C symbol. 6796268Sgad* Explicit Reg Vars:: Defining variables residing in specified registers. 68116233Sgad* Alternate Keywords:: @code{__const__}, @code{__asm__}, etc., for header files. 6996268Sgad* Incomplete Enums:: @code{enum foo;}, with details to follow. 7096301Sgrog* Function Names:: Printable strings which are the name of the current 7196332Speter function. 7296332Speter* Return Address:: Getting the return or frame address of a function. 7396332Speter* Vector Extensions:: Using vector instructions through built-in functions. 7496332Speter* Offsetof:: Special syntax for implementing @code{offsetof}. 7596332Speter* Atomic Builtins:: Built-in functions for atomic memory access. 76100314Sru* Object Size Checking:: Built-in functions for limited buffer overflow 7796451Sru checking. 7897611Sbillf* Other Builtins:: Other built-in functions. 7998333Sanholt* Target Builtins:: Built-in functions specific to particular targets. 8098986Sjmallett* Target Format Checks:: Format checks specific to particular targets. 81111061Sjmallett* Pragmas:: Pragmas accepted by GCC. 8299732Sjoerg* Unnamed Fields:: Unnamed struct/union fields within structs/unions. 8399732Sjoerg* Thread-Local:: Per-thread variables. 84113692Snectar* Binary constants:: Binary constants using the @samp{0b} prefix. 85113692Snectar@end menu 86115825Sfanf 87126445Sobrien@node Statement Exprs 88117645Sdwmalone@section Statements and Declarations in Expressions 89118204Sbp@cindex statements inside expressions 90118204Sbp@cindex declarations inside expressions 91118204Sbp@cindex expressions containing statements 92118204Sbp@cindex macros, statements in expressions 93127337Smlaier 94126445Sobrien@c the above section title wrapped and causes an underfull hbox.. i 95129082Spjd@c changed it from "within" to "in". --mew 4feb93 96129082SpjdA compound statement enclosed in parentheses may appear as an expression 97115822Sdougbin GNU C@. This allows you to use loops, switches, and local variables 98126445Sobrienwithin an expression. 99115822Sdougb 100115822SdougbRecall that a compound statement is a sequence of statements surrounded 101115822Sdougbby braces; in this construct, parentheses go around the braces. For 102115822Sdougbexample: 103115822Sdougb 104115822Sdougb@smallexample 105115822Sdougb(@{ int y = foo (); int z; 106115822Sdougb if (y > 0) z = y; 107115822Sdougb else z = - y; 108115822Sdougb z; @}) 109115822Sdougb@end smallexample 110115822Sdougb 111115822Sdougb@noindent 112115822Sdougbis a valid (though slightly more complex than necessary) expression 113115822Sdougbfor the absolute value of @code{foo ()}. 114115822Sdougb 115115822SdougbThe last thing in the compound statement should be an expression 116115822Sdougbfollowed by a semicolon; the value of this subexpression serves as the 117115822Sdougbvalue of the entire construct. (If you use some other kind of statement 118115822Sdougblast within the braces, the construct has type @code{void}, and thus 119115822Sdougbeffectively no value.) 120115895Sguido 121115822SdougbThis feature is especially useful in making macro definitions ``safe'' (so 122115895Sguidothat they evaluate each operand exactly once). For example, the 123115895Sguido``maximum'' function is commonly defined as a macro in standard C as 124115895Sguidofollows: 125115822Sdougb 126115822Sdougb@smallexample 127115822Sdougb#define max(a,b) ((a) > (b) ? (a) : (b)) 128115822Sdougb@end smallexample 129115822Sdougb 130115822Sdougb@noindent 131115822Sdougb@cindex side effects, macro argument 132115822SdougbBut this definition computes either @var{a} or @var{b} twice, with bad 133115822Sdougbresults if the operand has side effects. In GNU C, if you know the 134115822Sdougbtype of the operands (here taken as @code{int}), you can define 135115822Sdougbthe macro safely as follows: 136115822Sdougb 137115822Sdougb@smallexample 138115822Sdougb#define maxint(a,b) \ 139115822Sdougb (@{int _a = (a), _b = (b); _a > _b ? _a : _b; @}) 140115822Sdougb@end smallexample 141115822Sdougb 142115822SdougbEmbedded statements are not allowed in constant expressions, such as 143115822Sdougbthe value of an enumeration constant, the width of a bit-field, or 144115822Sdougbthe initial value of a static variable. 145115822Sdougb 146115822SdougbIf you don't know the type of the operand, you can still do this, but you 147115822Sdougbmust use @code{typeof} (@pxref{Typeof}). 148115822Sdougb 149115822SdougbIn G++, the result value of a statement expression undergoes array and 150115822Sdougbfunction pointer decay, and is returned by value to the enclosing 151115822Sdougbexpression. For instance, if @code{A} is a class, then 152115822Sdougb 153115822Sdougb@smallexample 154115822Sdougb A a; 155115895Sguido 156115895Sguido (@{a;@}).Foo () 157115895Sguido@end smallexample 158115895Sguido 159115822Sdougb@noindent 160115822Sdougbwill construct a temporary @code{A} object to hold the result of the 161115822Sdougbstatement expression, and that will be used to invoke @code{Foo}. 162115822SdougbTherefore the @code{this} pointer observed by @code{Foo} will not be the 163address of @code{a}. 164 165Any temporaries created within a statement within a statement expression 166will be destroyed at the statement's end. This makes statement 167expressions inside macros slightly different from function calls. In 168the latter case temporaries introduced during argument evaluation will 169be destroyed at the end of the statement that includes the function 170call. In the statement expression case they will be destroyed during 171the statement expression. For instance, 172 173@smallexample 174#define macro(a) (@{__typeof__(a) b = (a); b + 3; @}) 175template<typename T> T function(T a) @{ T b = a; return b + 3; @} 176 177void foo () 178@{ 179 macro (X ()); 180 function (X ()); 181@} 182@end smallexample 183 184@noindent 185will have different places where temporaries are destroyed. For the 186@code{macro} case, the temporary @code{X} will be destroyed just after 187the initialization of @code{b}. In the @code{function} case that 188temporary will be destroyed when the function returns. 189 190These considerations mean that it is probably a bad idea to use 191statement-expressions of this form in header files that are designed to 192work with C++. (Note that some versions of the GNU C Library contained 193header files using statement-expression that lead to precisely this 194bug.) 195 196Jumping into a statement expression with @code{goto} or using a 197@code{switch} statement outside the statement expression with a 198@code{case} or @code{default} label inside the statement expression is 199not permitted. Jumping into a statement expression with a computed 200@code{goto} (@pxref{Labels as Values}) yields undefined behavior. 201Jumping out of a statement expression is permitted, but if the 202statement expression is part of a larger expression then it is 203unspecified which other subexpressions of that expression have been 204evaluated except where the language definition requires certain 205subexpressions to be evaluated before or after the statement 206expression. In any case, as with a function call the evaluation of a 207statement expression is not interleaved with the evaluation of other 208parts of the containing expression. For example, 209 210@smallexample 211 foo (), ((@{ bar1 (); goto a; 0; @}) + bar2 ()), baz(); 212@end smallexample 213 214@noindent 215will call @code{foo} and @code{bar1} and will not call @code{baz} but 216may or may not call @code{bar2}. If @code{bar2} is called, it will be 217called after @code{foo} and before @code{bar1} 218 219@node Local Labels 220@section Locally Declared Labels 221@cindex local labels 222@cindex macros, local labels 223 224GCC allows you to declare @dfn{local labels} in any nested block 225scope. A local label is just like an ordinary label, but you can 226only reference it (with a @code{goto} statement, or by taking its 227address) within the block in which it was declared. 228 229A local label declaration looks like this: 230 231@smallexample 232__label__ @var{label}; 233@end smallexample 234 235@noindent 236or 237 238@smallexample 239__label__ @var{label1}, @var{label2}, /* @r{@dots{}} */; 240@end smallexample 241 242Local label declarations must come at the beginning of the block, 243before any ordinary declarations or statements. 244 245The label declaration defines the label @emph{name}, but does not define 246the label itself. You must do this in the usual way, with 247@code{@var{label}:}, within the statements of the statement expression. 248 249The local label feature is useful for complex macros. If a macro 250contains nested loops, a @code{goto} can be useful for breaking out of 251them. However, an ordinary label whose scope is the whole function 252cannot be used: if the macro can be expanded several times in one 253function, the label will be multiply defined in that function. A 254local label avoids this problem. For example: 255 256@smallexample 257#define SEARCH(value, array, target) \ 258do @{ \ 259 __label__ found; \ 260 typeof (target) _SEARCH_target = (target); \ 261 typeof (*(array)) *_SEARCH_array = (array); \ 262 int i, j; \ 263 int value; \ 264 for (i = 0; i < max; i++) \ 265 for (j = 0; j < max; j++) \ 266 if (_SEARCH_array[i][j] == _SEARCH_target) \ 267 @{ (value) = i; goto found; @} \ 268 (value) = -1; \ 269 found:; \ 270@} while (0) 271@end smallexample 272 273This could also be written using a statement-expression: 274 275@smallexample 276#define SEARCH(array, target) \ 277(@{ \ 278 __label__ found; \ 279 typeof (target) _SEARCH_target = (target); \ 280 typeof (*(array)) *_SEARCH_array = (array); \ 281 int i, j; \ 282 int value; \ 283 for (i = 0; i < max; i++) \ 284 for (j = 0; j < max; j++) \ 285 if (_SEARCH_array[i][j] == _SEARCH_target) \ 286 @{ value = i; goto found; @} \ 287 value = -1; \ 288 found: \ 289 value; \ 290@}) 291@end smallexample 292 293Local label declarations also make the labels they declare visible to 294nested functions, if there are any. @xref{Nested Functions}, for details. 295 296@node Labels as Values 297@section Labels as Values 298@cindex labels as values 299@cindex computed gotos 300@cindex goto with computed label 301@cindex address of a label 302 303You can get the address of a label defined in the current function 304(or a containing function) with the unary operator @samp{&&}. The 305value has type @code{void *}. This value is a constant and can be used 306wherever a constant of that type is valid. For example: 307 308@smallexample 309void *ptr; 310/* @r{@dots{}} */ 311ptr = &&foo; 312@end smallexample 313 314To use these values, you need to be able to jump to one. This is done 315with the computed goto statement@footnote{The analogous feature in 316Fortran is called an assigned goto, but that name seems inappropriate in 317C, where one can do more than simply store label addresses in label 318variables.}, @code{goto *@var{exp};}. For example, 319 320@smallexample 321goto *ptr; 322@end smallexample 323 324@noindent 325Any expression of type @code{void *} is allowed. 326 327One way of using these constants is in initializing a static array that 328will serve as a jump table: 329 330@smallexample 331static void *array[] = @{ &&foo, &&bar, &&hack @}; 332@end smallexample 333 334Then you can select a label with indexing, like this: 335 336@smallexample 337goto *array[i]; 338@end smallexample 339 340@noindent 341Note that this does not check whether the subscript is in bounds---array 342indexing in C never does that. 343 344Such an array of label values serves a purpose much like that of the 345@code{switch} statement. The @code{switch} statement is cleaner, so 346use that rather than an array unless the problem does not fit a 347@code{switch} statement very well. 348 349Another use of label values is in an interpreter for threaded code. 350The labels within the interpreter function can be stored in the 351threaded code for super-fast dispatching. 352 353You may not use this mechanism to jump to code in a different function. 354If you do that, totally unpredictable things will happen. The best way to 355avoid this is to store the label address only in automatic variables and 356never pass it as an argument. 357 358An alternate way to write the above example is 359 360@smallexample 361static const int array[] = @{ &&foo - &&foo, &&bar - &&foo, 362 &&hack - &&foo @}; 363goto *(&&foo + array[i]); 364@end smallexample 365 366@noindent 367This is more friendly to code living in shared libraries, as it reduces 368the number of dynamic relocations that are needed, and by consequence, 369allows the data to be read-only. 370 371@node Nested Functions 372@section Nested Functions 373@cindex nested functions 374@cindex downward funargs 375@cindex thunks 376 377A @dfn{nested function} is a function defined inside another function. 378(Nested functions are not supported for GNU C++.) The nested function's 379name is local to the block where it is defined. For example, here we 380define a nested function named @code{square}, and call it twice: 381 382@smallexample 383@group 384foo (double a, double b) 385@{ 386 double square (double z) @{ return z * z; @} 387 388 return square (a) + square (b); 389@} 390@end group 391@end smallexample 392 393The nested function can access all the variables of the containing 394function that are visible at the point of its definition. This is 395called @dfn{lexical scoping}. For example, here we show a nested 396function which uses an inherited variable named @code{offset}: 397 398@smallexample 399@group 400bar (int *array, int offset, int size) 401@{ 402 int access (int *array, int index) 403 @{ return array[index + offset]; @} 404 int i; 405 /* @r{@dots{}} */ 406 for (i = 0; i < size; i++) 407 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */ 408@} 409@end group 410@end smallexample 411 412Nested function definitions are permitted within functions in the places 413where variable definitions are allowed; that is, in any block, mixed 414with the other declarations and statements in the block. 415 416It is possible to call the nested function from outside the scope of its 417name by storing its address or passing the address to another function: 418 419@smallexample 420hack (int *array, int size) 421@{ 422 void store (int index, int value) 423 @{ array[index] = value; @} 424 425 intermediate (store, size); 426@} 427@end smallexample 428 429Here, the function @code{intermediate} receives the address of 430@code{store} as an argument. If @code{intermediate} calls @code{store}, 431the arguments given to @code{store} are used to store into @code{array}. 432But this technique works only so long as the containing function 433(@code{hack}, in this example) does not exit. 434 435If you try to call the nested function through its address after the 436containing function has exited, all hell will break loose. If you try 437to call it after a containing scope level has exited, and if it refers 438to some of the variables that are no longer in scope, you may be lucky, 439but it's not wise to take the risk. If, however, the nested function 440does not refer to anything that has gone out of scope, you should be 441safe. 442 443GCC implements taking the address of a nested function using a technique 444called @dfn{trampolines}. A paper describing them is available as 445 446@noindent 447@uref{http://people.debian.org/~aaronl/Usenix88-lexic.pdf}. 448 449A nested function can jump to a label inherited from a containing 450function, provided the label was explicitly declared in the containing 451function (@pxref{Local Labels}). Such a jump returns instantly to the 452containing function, exiting the nested function which did the 453@code{goto} and any intermediate functions as well. Here is an example: 454 455@smallexample 456@group 457bar (int *array, int offset, int size) 458@{ 459 __label__ failure; 460 int access (int *array, int index) 461 @{ 462 if (index > size) 463 goto failure; 464 return array[index + offset]; 465 @} 466 int i; 467 /* @r{@dots{}} */ 468 for (i = 0; i < size; i++) 469 /* @r{@dots{}} */ access (array, i) /* @r{@dots{}} */ 470 /* @r{@dots{}} */ 471 return 0; 472 473 /* @r{Control comes here from @code{access} 474 if it detects an error.} */ 475 failure: 476 return -1; 477@} 478@end group 479@end smallexample 480 481A nested function always has no linkage. Declaring one with 482@code{extern} or @code{static} is erroneous. If you need to declare the nested function 483before its definition, use @code{auto} (which is otherwise meaningless 484for function declarations). 485 486@smallexample 487bar (int *array, int offset, int size) 488@{ 489 __label__ failure; 490 auto int access (int *, int); 491 /* @r{@dots{}} */ 492 int access (int *array, int index) 493 @{ 494 if (index > size) 495 goto failure; 496 return array[index + offset]; 497 @} 498 /* @r{@dots{}} */ 499@} 500@end smallexample 501 502@node Constructing Calls 503@section Constructing Function Calls 504@cindex constructing calls 505@cindex forwarding calls 506 507Using the built-in functions described below, you can record 508the arguments a function received, and call another function 509with the same arguments, without knowing the number or types 510of the arguments. 511 512You can also record the return value of that function call, 513and later return that value, without knowing what data type 514the function tried to return (as long as your caller expects 515that data type). 516 517However, these built-in functions may interact badly with some 518sophisticated features or other extensions of the language. It 519is, therefore, not recommended to use them outside very simple 520functions acting as mere forwarders for their arguments. 521 522@deftypefn {Built-in Function} {void *} __builtin_apply_args () 523This built-in function returns a pointer to data 524describing how to perform a call with the same arguments as were passed 525to the current function. 526 527The function saves the arg pointer register, structure value address, 528and all registers that might be used to pass arguments to a function 529into a block of memory allocated on the stack. Then it returns the 530address of that block. 531@end deftypefn 532 533@deftypefn {Built-in Function} {void *} __builtin_apply (void (*@var{function})(), void *@var{arguments}, size_t @var{size}) 534This built-in function invokes @var{function} 535with a copy of the parameters described by @var{arguments} 536and @var{size}. 537 538The value of @var{arguments} should be the value returned by 539@code{__builtin_apply_args}. The argument @var{size} specifies the size 540of the stack argument data, in bytes. 541 542This function returns a pointer to data describing 543how to return whatever value was returned by @var{function}. The data 544is saved in a block of memory allocated on the stack. 545 546It is not always simple to compute the proper value for @var{size}. The 547value is used by @code{__builtin_apply} to compute the amount of data 548that should be pushed on the stack and copied from the incoming argument 549area. 550@end deftypefn 551 552@deftypefn {Built-in Function} {void} __builtin_return (void *@var{result}) 553This built-in function returns the value described by @var{result} from 554the containing function. You should specify, for @var{result}, a value 555returned by @code{__builtin_apply}. 556@end deftypefn 557 558@node Typeof 559@section Referring to a Type with @code{typeof} 560@findex typeof 561@findex sizeof 562@cindex macros, types of arguments 563 564Another way to refer to the type of an expression is with @code{typeof}. 565The syntax of using of this keyword looks like @code{sizeof}, but the 566construct acts semantically like a type name defined with @code{typedef}. 567 568There are two ways of writing the argument to @code{typeof}: with an 569expression or with a type. Here is an example with an expression: 570 571@smallexample 572typeof (x[0](1)) 573@end smallexample 574 575@noindent 576This assumes that @code{x} is an array of pointers to functions; 577the type described is that of the values of the functions. 578 579Here is an example with a typename as the argument: 580 581@smallexample 582typeof (int *) 583@end smallexample 584 585@noindent 586Here the type described is that of pointers to @code{int}. 587 588If you are writing a header file that must work when included in ISO C 589programs, write @code{__typeof__} instead of @code{typeof}. 590@xref{Alternate Keywords}. 591 592A @code{typeof}-construct can be used anywhere a typedef name could be 593used. For example, you can use it in a declaration, in a cast, or inside 594of @code{sizeof} or @code{typeof}. 595 596@code{typeof} is often useful in conjunction with the 597statements-within-expressions feature. Here is how the two together can 598be used to define a safe ``maximum'' macro that operates on any 599arithmetic type and evaluates each of its arguments exactly once: 600 601@smallexample 602#define max(a,b) \ 603 (@{ typeof (a) _a = (a); \ 604 typeof (b) _b = (b); \ 605 _a > _b ? _a : _b; @}) 606@end smallexample 607 608@cindex underscores in variables in macros 609@cindex @samp{_} in variables in macros 610@cindex local variables in macros 611@cindex variables, local, in macros 612@cindex macros, local variables in 613 614The reason for using names that start with underscores for the local 615variables is to avoid conflicts with variable names that occur within the 616expressions that are substituted for @code{a} and @code{b}. Eventually we 617hope to design a new form of declaration syntax that allows you to declare 618variables whose scopes start only after their initializers; this will be a 619more reliable way to prevent such conflicts. 620 621@noindent 622Some more examples of the use of @code{typeof}: 623 624@itemize @bullet 625@item 626This declares @code{y} with the type of what @code{x} points to. 627 628@smallexample 629typeof (*x) y; 630@end smallexample 631 632@item 633This declares @code{y} as an array of such values. 634 635@smallexample 636typeof (*x) y[4]; 637@end smallexample 638 639@item 640This declares @code{y} as an array of pointers to characters: 641 642@smallexample 643typeof (typeof (char *)[4]) y; 644@end smallexample 645 646@noindent 647It is equivalent to the following traditional C declaration: 648 649@smallexample 650char *y[4]; 651@end smallexample 652 653To see the meaning of the declaration using @code{typeof}, and why it 654might be a useful way to write, rewrite it with these macros: 655 656@smallexample 657#define pointer(T) typeof(T *) 658#define array(T, N) typeof(T [N]) 659@end smallexample 660 661@noindent 662Now the declaration can be rewritten this way: 663 664@smallexample 665array (pointer (char), 4) y; 666@end smallexample 667 668@noindent 669Thus, @code{array (pointer (char), 4)} is the type of arrays of 4 670pointers to @code{char}. 671@end itemize 672 673@emph{Compatibility Note:} In addition to @code{typeof}, GCC 2 supported 674a more limited extension which permitted one to write 675 676@smallexample 677typedef @var{T} = @var{expr}; 678@end smallexample 679 680@noindent 681with the effect of declaring @var{T} to have the type of the expression 682@var{expr}. This extension does not work with GCC 3 (versions between 6833.0 and 3.2 will crash; 3.2.1 and later give an error). Code which 684relies on it should be rewritten to use @code{typeof}: 685 686@smallexample 687typedef typeof(@var{expr}) @var{T}; 688@end smallexample 689 690@noindent 691This will work with all versions of GCC@. 692 693@node Conditionals 694@section Conditionals with Omitted Operands 695@cindex conditional expressions, extensions 696@cindex omitted middle-operands 697@cindex middle-operands, omitted 698@cindex extensions, @code{?:} 699@cindex @code{?:} extensions 700 701The middle operand in a conditional expression may be omitted. Then 702if the first operand is nonzero, its value is the value of the conditional 703expression. 704 705Therefore, the expression 706 707@smallexample 708x ? : y 709@end smallexample 710 711@noindent 712has the value of @code{x} if that is nonzero; otherwise, the value of 713@code{y}. 714 715This example is perfectly equivalent to 716 717@smallexample 718x ? x : y 719@end smallexample 720 721@cindex side effect in ?: 722@cindex ?: side effect 723@noindent 724In this simple case, the ability to omit the middle operand is not 725especially useful. When it becomes useful is when the first operand does, 726or may (if it is a macro argument), contain a side effect. Then repeating 727the operand in the middle would perform the side effect twice. Omitting 728the middle operand uses the value already computed without the undesirable 729effects of recomputing it. 730 731@node Long Long 732@section Double-Word Integers 733@cindex @code{long long} data types 734@cindex double-word arithmetic 735@cindex multiprecision arithmetic 736@cindex @code{LL} integer suffix 737@cindex @code{ULL} integer suffix 738 739ISO C99 supports data types for integers that are at least 64 bits wide, 740and as an extension GCC supports them in C89 mode and in C++. 741Simply write @code{long long int} for a signed integer, or 742@code{unsigned long long int} for an unsigned integer. To make an 743integer constant of type @code{long long int}, add the suffix @samp{LL} 744to the integer. To make an integer constant of type @code{unsigned long 745long int}, add the suffix @samp{ULL} to the integer. 746 747You can use these types in arithmetic like any other integer types. 748Addition, subtraction, and bitwise boolean operations on these types 749are open-coded on all types of machines. Multiplication is open-coded 750if the machine supports fullword-to-doubleword a widening multiply 751instruction. Division and shifts are open-coded only on machines that 752provide special support. The operations that are not open-coded use 753special library routines that come with GCC@. 754 755There may be pitfalls when you use @code{long long} types for function 756arguments, unless you declare function prototypes. If a function 757expects type @code{int} for its argument, and you pass a value of type 758@code{long long int}, confusion will result because the caller and the 759subroutine will disagree about the number of bytes for the argument. 760Likewise, if the function expects @code{long long int} and you pass 761@code{int}. The best way to avoid such problems is to use prototypes. 762 763@node Complex 764@section Complex Numbers 765@cindex complex numbers 766@cindex @code{_Complex} keyword 767@cindex @code{__complex__} keyword 768 769ISO C99 supports complex floating data types, and as an extension GCC 770supports them in C89 mode and in C++, and supports complex integer data 771types which are not part of ISO C99. You can declare complex types 772using the keyword @code{_Complex}. As an extension, the older GNU 773keyword @code{__complex__} is also supported. 774 775For example, @samp{_Complex double x;} declares @code{x} as a 776variable whose real part and imaginary part are both of type 777@code{double}. @samp{_Complex short int y;} declares @code{y} to 778have real and imaginary parts of type @code{short int}; this is not 779likely to be useful, but it shows that the set of complex types is 780complete. 781 782To write a constant with a complex data type, use the suffix @samp{i} or 783@samp{j} (either one; they are equivalent). For example, @code{2.5fi} 784has type @code{_Complex float} and @code{3i} has type 785@code{_Complex int}. Such a constant always has a pure imaginary 786value, but you can form any complex value you like by adding one to a 787real constant. This is a GNU extension; if you have an ISO C99 788conforming C library (such as GNU libc), and want to construct complex 789constants of floating type, you should include @code{<complex.h>} and 790use the macros @code{I} or @code{_Complex_I} instead. 791 792@cindex @code{__real__} keyword 793@cindex @code{__imag__} keyword 794To extract the real part of a complex-valued expression @var{exp}, write 795@code{__real__ @var{exp}}. Likewise, use @code{__imag__} to 796extract the imaginary part. This is a GNU extension; for values of 797floating type, you should use the ISO C99 functions @code{crealf}, 798@code{creal}, @code{creall}, @code{cimagf}, @code{cimag} and 799@code{cimagl}, declared in @code{<complex.h>} and also provided as 800built-in functions by GCC@. 801 802@cindex complex conjugation 803The operator @samp{~} performs complex conjugation when used on a value 804with a complex type. This is a GNU extension; for values of 805floating type, you should use the ISO C99 functions @code{conjf}, 806@code{conj} and @code{conjl}, declared in @code{<complex.h>} and also 807provided as built-in functions by GCC@. 808 809GCC can allocate complex automatic variables in a noncontiguous 810fashion; it's even possible for the real part to be in a register while 811the imaginary part is on the stack (or vice-versa). Only the DWARF2 812debug info format can represent this, so use of DWARF2 is recommended. 813If you are using the stabs debug info format, GCC describes a noncontiguous 814complex variable as if it were two separate variables of noncomplex type. 815If the variable's actual name is @code{foo}, the two fictitious 816variables are named @code{foo$real} and @code{foo$imag}. You can 817examine and set these two fictitious variables with your debugger. 818 819@node Decimal Float 820@section Decimal Floating Types 821@cindex decimal floating types 822@cindex @code{_Decimal32} data type 823@cindex @code{_Decimal64} data type 824@cindex @code{_Decimal128} data type 825@cindex @code{df} integer suffix 826@cindex @code{dd} integer suffix 827@cindex @code{dl} integer suffix 828@cindex @code{DF} integer suffix 829@cindex @code{DD} integer suffix 830@cindex @code{DL} integer suffix 831 832As an extension, the GNU C compiler supports decimal floating types as 833defined in the N1176 draft of ISO/IEC WDTR24732. Support for decimal 834floating types in GCC will evolve as the draft technical report changes. 835Calling conventions for any target might also change. Not all targets 836support decimal floating types. 837 838The decimal floating types are @code{_Decimal32}, @code{_Decimal64}, and 839@code{_Decimal128}. They use a radix of ten, unlike the floating types 840@code{float}, @code{double}, and @code{long double} whose radix is not 841specified by the C standard but is usually two. 842 843Support for decimal floating types includes the arithmetic operators 844add, subtract, multiply, divide; unary arithmetic operators; 845relational operators; equality operators; and conversions to and from 846integer and other floating types. Use a suffix @samp{df} or 847@samp{DF} in a literal constant of type @code{_Decimal32}, @samp{dd} 848or @samp{DD} for @code{_Decimal64}, and @samp{dl} or @samp{DL} for 849@code{_Decimal128}. 850 851GCC support of decimal float as specified by the draft technical report 852is incomplete: 853 854@itemize @bullet 855@item 856Translation time data type (TTDT) is not supported. 857 858@item 859Characteristics of decimal floating types are defined in header file 860@file{decfloat.h} rather than @file{float.h}. 861 862@item 863When the value of a decimal floating type cannot be represented in the 864integer type to which it is being converted, the result is undefined 865rather than the result value specified by the draft technical report. 866@end itemize 867 868Types @code{_Decimal32}, @code{_Decimal64}, and @code{_Decimal128} 869are supported by the DWARF2 debug information format. 870 871@node Hex Floats 872@section Hex Floats 873@cindex hex floats 874 875ISO C99 supports floating-point numbers written not only in the usual 876decimal notation, such as @code{1.55e1}, but also numbers such as 877@code{0x1.fp3} written in hexadecimal format. As a GNU extension, GCC 878supports this in C89 mode (except in some cases when strictly 879conforming) and in C++. In that format the 880@samp{0x} hex introducer and the @samp{p} or @samp{P} exponent field are 881mandatory. The exponent is a decimal number that indicates the power of 8822 by which the significant part will be multiplied. Thus @samp{0x1.f} is 883@tex 884$1 {15\over16}$, 885@end tex 886@ifnottex 8871 15/16, 888@end ifnottex 889@samp{p3} multiplies it by 8, and the value of @code{0x1.fp3} 890is the same as @code{1.55e1}. 891 892Unlike for floating-point numbers in the decimal notation the exponent 893is always required in the hexadecimal notation. Otherwise the compiler 894would not be able to resolve the ambiguity of, e.g., @code{0x1.f}. This 895could mean @code{1.0f} or @code{1.9375} since @samp{f} is also the 896extension for floating-point constants of type @code{float}. 897 898@node Zero Length 899@section Arrays of Length Zero 900@cindex arrays of length zero 901@cindex zero-length arrays 902@cindex length-zero arrays 903@cindex flexible array members 904 905Zero-length arrays are allowed in GNU C@. They are very useful as the 906last element of a structure which is really a header for a variable-length 907object: 908 909@smallexample 910struct line @{ 911 int length; 912 char contents[0]; 913@}; 914 915struct line *thisline = (struct line *) 916 malloc (sizeof (struct line) + this_length); 917thisline->length = this_length; 918@end smallexample 919 920In ISO C90, you would have to give @code{contents} a length of 1, which 921means either you waste space or complicate the argument to @code{malloc}. 922 923In ISO C99, you would use a @dfn{flexible array member}, which is 924slightly different in syntax and semantics: 925 926@itemize @bullet 927@item 928Flexible array members are written as @code{contents[]} without 929the @code{0}. 930 931@item 932Flexible array members have incomplete type, and so the @code{sizeof} 933operator may not be applied. As a quirk of the original implementation 934of zero-length arrays, @code{sizeof} evaluates to zero. 935 936@item 937Flexible array members may only appear as the last member of a 938@code{struct} that is otherwise non-empty. 939 940@item 941A structure containing a flexible array member, or a union containing 942such a structure (possibly recursively), may not be a member of a 943structure or an element of an array. (However, these uses are 944permitted by GCC as extensions.) 945@end itemize 946 947GCC versions before 3.0 allowed zero-length arrays to be statically 948initialized, as if they were flexible arrays. In addition to those 949cases that were useful, it also allowed initializations in situations 950that would corrupt later data. Non-empty initialization of zero-length 951arrays is now treated like any case where there are more initializer 952elements than the array holds, in that a suitable warning about "excess 953elements in array" is given, and the excess elements (all of them, in 954this case) are ignored. 955 956Instead GCC allows static initialization of flexible array members. 957This is equivalent to defining a new structure containing the original 958structure followed by an array of sufficient size to contain the data. 959I.e.@: in the following, @code{f1} is constructed as if it were declared 960like @code{f2}. 961 962@smallexample 963struct f1 @{ 964 int x; int y[]; 965@} f1 = @{ 1, @{ 2, 3, 4 @} @}; 966 967struct f2 @{ 968 struct f1 f1; int data[3]; 969@} f2 = @{ @{ 1 @}, @{ 2, 3, 4 @} @}; 970@end smallexample 971 972@noindent 973The convenience of this extension is that @code{f1} has the desired 974type, eliminating the need to consistently refer to @code{f2.f1}. 975 976This has symmetry with normal static arrays, in that an array of 977unknown size is also written with @code{[]}. 978 979Of course, this extension only makes sense if the extra data comes at 980the end of a top-level object, as otherwise we would be overwriting 981data at subsequent offsets. To avoid undue complication and confusion 982with initialization of deeply nested arrays, we simply disallow any 983non-empty initialization except when the structure is the top-level 984object. For example: 985 986@smallexample 987struct foo @{ int x; int y[]; @}; 988struct bar @{ struct foo z; @}; 989 990struct foo a = @{ 1, @{ 2, 3, 4 @} @}; // @r{Valid.} 991struct bar b = @{ @{ 1, @{ 2, 3, 4 @} @} @}; // @r{Invalid.} 992struct bar c = @{ @{ 1, @{ @} @} @}; // @r{Valid.} 993struct foo d[1] = @{ @{ 1 @{ 2, 3, 4 @} @} @}; // @r{Invalid.} 994@end smallexample 995 996@node Empty Structures 997@section Structures With No Members 998@cindex empty structures 999@cindex zero-size structures 1000 1001GCC permits a C structure to have no members: 1002 1003@smallexample 1004struct empty @{ 1005@}; 1006@end smallexample 1007 1008The structure will have size zero. In C++, empty structures are part 1009of the language. G++ treats empty structures as if they had a single 1010member of type @code{char}. 1011 1012@node Variable Length 1013@section Arrays of Variable Length 1014@cindex variable-length arrays 1015@cindex arrays of variable length 1016@cindex VLAs 1017 1018Variable-length automatic arrays are allowed in ISO C99, and as an 1019extension GCC accepts them in C89 mode and in C++. (However, GCC's 1020implementation of variable-length arrays does not yet conform in detail 1021to the ISO C99 standard.) These arrays are 1022declared like any other automatic arrays, but with a length that is not 1023a constant expression. The storage is allocated at the point of 1024declaration and deallocated when the brace-level is exited. For 1025example: 1026 1027@smallexample 1028FILE * 1029concat_fopen (char *s1, char *s2, char *mode) 1030@{ 1031 char str[strlen (s1) + strlen (s2) + 1]; 1032 strcpy (str, s1); 1033 strcat (str, s2); 1034 return fopen (str, mode); 1035@} 1036@end smallexample 1037 1038@cindex scope of a variable length array 1039@cindex variable-length array scope 1040@cindex deallocating variable length arrays 1041Jumping or breaking out of the scope of the array name deallocates the 1042storage. Jumping into the scope is not allowed; you get an error 1043message for it. 1044 1045@cindex @code{alloca} vs variable-length arrays 1046You can use the function @code{alloca} to get an effect much like 1047variable-length arrays. The function @code{alloca} is available in 1048many other C implementations (but not in all). On the other hand, 1049variable-length arrays are more elegant. 1050 1051There are other differences between these two methods. Space allocated 1052with @code{alloca} exists until the containing @emph{function} returns. 1053The space for a variable-length array is deallocated as soon as the array 1054name's scope ends. (If you use both variable-length arrays and 1055@code{alloca} in the same function, deallocation of a variable-length array 1056will also deallocate anything more recently allocated with @code{alloca}.) 1057 1058You can also use variable-length arrays as arguments to functions: 1059 1060@smallexample 1061struct entry 1062tester (int len, char data[len][len]) 1063@{ 1064 /* @r{@dots{}} */ 1065@} 1066@end smallexample 1067 1068The length of an array is computed once when the storage is allocated 1069and is remembered for the scope of the array in case you access it with 1070@code{sizeof}. 1071 1072If you want to pass the array first and the length afterward, you can 1073use a forward declaration in the parameter list---another GNU extension. 1074 1075@smallexample 1076struct entry 1077tester (int len; char data[len][len], int len) 1078@{ 1079 /* @r{@dots{}} */ 1080@} 1081@end smallexample 1082 1083@cindex parameter forward declaration 1084The @samp{int len} before the semicolon is a @dfn{parameter forward 1085declaration}, and it serves the purpose of making the name @code{len} 1086known when the declaration of @code{data} is parsed. 1087 1088You can write any number of such parameter forward declarations in the 1089parameter list. They can be separated by commas or semicolons, but the 1090last one must end with a semicolon, which is followed by the ``real'' 1091parameter declarations. Each forward declaration must match a ``real'' 1092declaration in parameter name and data type. ISO C99 does not support 1093parameter forward declarations. 1094 1095@node Variadic Macros 1096@section Macros with a Variable Number of Arguments. 1097@cindex variable number of arguments 1098@cindex macro with variable arguments 1099@cindex rest argument (in macro) 1100@cindex variadic macros 1101 1102In the ISO C standard of 1999, a macro can be declared to accept a 1103variable number of arguments much as a function can. The syntax for 1104defining the macro is similar to that of a function. Here is an 1105example: 1106 1107@smallexample 1108#define debug(format, ...) fprintf (stderr, format, __VA_ARGS__) 1109@end smallexample 1110 1111Here @samp{@dots{}} is a @dfn{variable argument}. In the invocation of 1112such a macro, it represents the zero or more tokens until the closing 1113parenthesis that ends the invocation, including any commas. This set of 1114tokens replaces the identifier @code{__VA_ARGS__} in the macro body 1115wherever it appears. See the CPP manual for more information. 1116 1117GCC has long supported variadic macros, and used a different syntax that 1118allowed you to give a name to the variable arguments just like any other 1119argument. Here is an example: 1120 1121@smallexample 1122#define debug(format, args...) fprintf (stderr, format, args) 1123@end smallexample 1124 1125This is in all ways equivalent to the ISO C example above, but arguably 1126more readable and descriptive. 1127 1128GNU CPP has two further variadic macro extensions, and permits them to 1129be used with either of the above forms of macro definition. 1130 1131In standard C, you are not allowed to leave the variable argument out 1132entirely; but you are allowed to pass an empty argument. For example, 1133this invocation is invalid in ISO C, because there is no comma after 1134the string: 1135 1136@smallexample 1137debug ("A message") 1138@end smallexample 1139 1140GNU CPP permits you to completely omit the variable arguments in this 1141way. In the above examples, the compiler would complain, though since 1142the expansion of the macro still has the extra comma after the format 1143string. 1144 1145To help solve this problem, CPP behaves specially for variable arguments 1146used with the token paste operator, @samp{##}. If instead you write 1147 1148@smallexample 1149#define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__) 1150@end smallexample 1151 1152and if the variable arguments are omitted or empty, the @samp{##} 1153operator causes the preprocessor to remove the comma before it. If you 1154do provide some variable arguments in your macro invocation, GNU CPP 1155does not complain about the paste operation and instead places the 1156variable arguments after the comma. Just like any other pasted macro 1157argument, these arguments are not macro expanded. 1158 1159@node Escaped Newlines 1160@section Slightly Looser Rules for Escaped Newlines 1161@cindex escaped newlines 1162@cindex newlines (escaped) 1163 1164Recently, the preprocessor has relaxed its treatment of escaped 1165newlines. Previously, the newline had to immediately follow a 1166backslash. The current implementation allows whitespace in the form 1167of spaces, horizontal and vertical tabs, and form feeds between the 1168backslash and the subsequent newline. The preprocessor issues a 1169warning, but treats it as a valid escaped newline and combines the two 1170lines to form a single logical line. This works within comments and 1171tokens, as well as between tokens. Comments are @emph{not} treated as 1172whitespace for the purposes of this relaxation, since they have not 1173yet been replaced with spaces. 1174 1175@node Subscripting 1176@section Non-Lvalue Arrays May Have Subscripts 1177@cindex subscripting 1178@cindex arrays, non-lvalue 1179 1180@cindex subscripting and function values 1181In ISO C99, arrays that are not lvalues still decay to pointers, and 1182may be subscripted, although they may not be modified or used after 1183the next sequence point and the unary @samp{&} operator may not be 1184applied to them. As an extension, GCC allows such arrays to be 1185subscripted in C89 mode, though otherwise they do not decay to 1186pointers outside C99 mode. For example, 1187this is valid in GNU C though not valid in C89: 1188 1189@smallexample 1190@group 1191struct foo @{int a[4];@}; 1192 1193struct foo f(); 1194 1195bar (int index) 1196@{ 1197 return f().a[index]; 1198@} 1199@end group 1200@end smallexample 1201 1202@node Pointer Arith 1203@section Arithmetic on @code{void}- and Function-Pointers 1204@cindex void pointers, arithmetic 1205@cindex void, size of pointer to 1206@cindex function pointers, arithmetic 1207@cindex function, size of pointer to 1208 1209In GNU C, addition and subtraction operations are supported on pointers to 1210@code{void} and on pointers to functions. This is done by treating the 1211size of a @code{void} or of a function as 1. 1212 1213A consequence of this is that @code{sizeof} is also allowed on @code{void} 1214and on function types, and returns 1. 1215 1216@opindex Wpointer-arith 1217The option @option{-Wpointer-arith} requests a warning if these extensions 1218are used. 1219 1220@node Initializers 1221@section Non-Constant Initializers 1222@cindex initializers, non-constant 1223@cindex non-constant initializers 1224 1225As in standard C++ and ISO C99, the elements of an aggregate initializer for an 1226automatic variable are not required to be constant expressions in GNU C@. 1227Here is an example of an initializer with run-time varying elements: 1228 1229@smallexample 1230foo (float f, float g) 1231@{ 1232 float beat_freqs[2] = @{ f-g, f+g @}; 1233 /* @r{@dots{}} */ 1234@} 1235@end smallexample 1236 1237@node Compound Literals 1238@section Compound Literals 1239@cindex constructor expressions 1240@cindex initializations in expressions 1241@cindex structures, constructor expression 1242@cindex expressions, constructor 1243@cindex compound literals 1244@c The GNU C name for what C99 calls compound literals was "constructor expressions". 1245 1246ISO C99 supports compound literals. A compound literal looks like 1247a cast containing an initializer. Its value is an object of the 1248type specified in the cast, containing the elements specified in 1249the initializer; it is an lvalue. As an extension, GCC supports 1250compound literals in C89 mode and in C++. 1251 1252Usually, the specified type is a structure. Assume that 1253@code{struct foo} and @code{structure} are declared as shown: 1254 1255@smallexample 1256struct foo @{int a; char b[2];@} structure; 1257@end smallexample 1258 1259@noindent 1260Here is an example of constructing a @code{struct foo} with a compound literal: 1261 1262@smallexample 1263structure = ((struct foo) @{x + y, 'a', 0@}); 1264@end smallexample 1265 1266@noindent 1267This is equivalent to writing the following: 1268 1269@smallexample 1270@{ 1271 struct foo temp = @{x + y, 'a', 0@}; 1272 structure = temp; 1273@} 1274@end smallexample 1275 1276You can also construct an array. If all the elements of the compound literal 1277are (made up of) simple constant expressions, suitable for use in 1278initializers of objects of static storage duration, then the compound 1279literal can be coerced to a pointer to its first element and used in 1280such an initializer, as shown here: 1281 1282@smallexample 1283char **foo = (char *[]) @{ "x", "y", "z" @}; 1284@end smallexample 1285 1286Compound literals for scalar types and union types are is 1287also allowed, but then the compound literal is equivalent 1288to a cast. 1289 1290As a GNU extension, GCC allows initialization of objects with static storage 1291duration by compound literals (which is not possible in ISO C99, because 1292the initializer is not a constant). 1293It is handled as if the object was initialized only with the bracket 1294enclosed list if the types of the compound literal and the object match. 1295The initializer list of the compound literal must be constant. 1296If the object being initialized has array type of unknown size, the size is 1297determined by compound literal size. 1298 1299@smallexample 1300static struct foo x = (struct foo) @{1, 'a', 'b'@}; 1301static int y[] = (int []) @{1, 2, 3@}; 1302static int z[] = (int [3]) @{1@}; 1303@end smallexample 1304 1305@noindent 1306The above lines are equivalent to the following: 1307@smallexample 1308static struct foo x = @{1, 'a', 'b'@}; 1309static int y[] = @{1, 2, 3@}; 1310static int z[] = @{1, 0, 0@}; 1311@end smallexample 1312 1313@node Designated Inits 1314@section Designated Initializers 1315@cindex initializers with labeled elements 1316@cindex labeled elements in initializers 1317@cindex case labels in initializers 1318@cindex designated initializers 1319 1320Standard C89 requires the elements of an initializer to appear in a fixed 1321order, the same as the order of the elements in the array or structure 1322being initialized. 1323 1324In ISO C99 you can give the elements in any order, specifying the array 1325indices or structure field names they apply to, and GNU C allows this as 1326an extension in C89 mode as well. This extension is not 1327implemented in GNU C++. 1328 1329To specify an array index, write 1330@samp{[@var{index}] =} before the element value. For example, 1331 1332@smallexample 1333int a[6] = @{ [4] = 29, [2] = 15 @}; 1334@end smallexample 1335 1336@noindent 1337is equivalent to 1338 1339@smallexample 1340int a[6] = @{ 0, 0, 15, 0, 29, 0 @}; 1341@end smallexample 1342 1343@noindent 1344The index values must be constant expressions, even if the array being 1345initialized is automatic. 1346 1347An alternative syntax for this which has been obsolete since GCC 2.5 but 1348GCC still accepts is to write @samp{[@var{index}]} before the element 1349value, with no @samp{=}. 1350 1351To initialize a range of elements to the same value, write 1352@samp{[@var{first} ... @var{last}] = @var{value}}. This is a GNU 1353extension. For example, 1354 1355@smallexample 1356int widths[] = @{ [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 @}; 1357@end smallexample 1358 1359@noindent 1360If the value in it has side-effects, the side-effects will happen only once, 1361not for each initialized field by the range initializer. 1362 1363@noindent 1364Note that the length of the array is the highest value specified 1365plus one. 1366 1367In a structure initializer, specify the name of a field to initialize 1368with @samp{.@var{fieldname} =} before the element value. For example, 1369given the following structure, 1370 1371@smallexample 1372struct point @{ int x, y; @}; 1373@end smallexample 1374 1375@noindent 1376the following initialization 1377 1378@smallexample 1379struct point p = @{ .y = yvalue, .x = xvalue @}; 1380@end smallexample 1381 1382@noindent 1383is equivalent to 1384 1385@smallexample 1386struct point p = @{ xvalue, yvalue @}; 1387@end smallexample 1388 1389Another syntax which has the same meaning, obsolete since GCC 2.5, is 1390@samp{@var{fieldname}:}, as shown here: 1391 1392@smallexample 1393struct point p = @{ y: yvalue, x: xvalue @}; 1394@end smallexample 1395 1396@cindex designators 1397The @samp{[@var{index}]} or @samp{.@var{fieldname}} is known as a 1398@dfn{designator}. You can also use a designator (or the obsolete colon 1399syntax) when initializing a union, to specify which element of the union 1400should be used. For example, 1401 1402@smallexample 1403union foo @{ int i; double d; @}; 1404 1405union foo f = @{ .d = 4 @}; 1406@end smallexample 1407 1408@noindent 1409will convert 4 to a @code{double} to store it in the union using 1410the second element. By contrast, casting 4 to type @code{union foo} 1411would store it into the union as the integer @code{i}, since it is 1412an integer. (@xref{Cast to Union}.) 1413 1414You can combine this technique of naming elements with ordinary C 1415initialization of successive elements. Each initializer element that 1416does not have a designator applies to the next consecutive element of the 1417array or structure. For example, 1418 1419@smallexample 1420int a[6] = @{ [1] = v1, v2, [4] = v4 @}; 1421@end smallexample 1422 1423@noindent 1424is equivalent to 1425 1426@smallexample 1427int a[6] = @{ 0, v1, v2, 0, v4, 0 @}; 1428@end smallexample 1429 1430Labeling the elements of an array initializer is especially useful 1431when the indices are characters or belong to an @code{enum} type. 1432For example: 1433 1434@smallexample 1435int whitespace[256] 1436 = @{ [' '] = 1, ['\t'] = 1, ['\h'] = 1, 1437 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 @}; 1438@end smallexample 1439 1440@cindex designator lists 1441You can also write a series of @samp{.@var{fieldname}} and 1442@samp{[@var{index}]} designators before an @samp{=} to specify a 1443nested subobject to initialize; the list is taken relative to the 1444subobject corresponding to the closest surrounding brace pair. For 1445example, with the @samp{struct point} declaration above: 1446 1447@smallexample 1448struct point ptarray[10] = @{ [2].y = yv2, [2].x = xv2, [0].x = xv0 @}; 1449@end smallexample 1450 1451@noindent 1452If the same field is initialized multiple times, it will have value from 1453the last initialization. If any such overridden initialization has 1454side-effect, it is unspecified whether the side-effect happens or not. 1455Currently, GCC will discard them and issue a warning. 1456 1457@node Case Ranges 1458@section Case Ranges 1459@cindex case ranges 1460@cindex ranges in case statements 1461 1462You can specify a range of consecutive values in a single @code{case} label, 1463like this: 1464 1465@smallexample 1466case @var{low} ... @var{high}: 1467@end smallexample 1468 1469@noindent 1470This has the same effect as the proper number of individual @code{case} 1471labels, one for each integer value from @var{low} to @var{high}, inclusive. 1472 1473This feature is especially useful for ranges of ASCII character codes: 1474 1475@smallexample 1476case 'A' ... 'Z': 1477@end smallexample 1478 1479@strong{Be careful:} Write spaces around the @code{...}, for otherwise 1480it may be parsed wrong when you use it with integer values. For example, 1481write this: 1482 1483@smallexample 1484case 1 ... 5: 1485@end smallexample 1486 1487@noindent 1488rather than this: 1489 1490@smallexample 1491case 1...5: 1492@end smallexample 1493 1494@node Cast to Union 1495@section Cast to a Union Type 1496@cindex cast to a union 1497@cindex union, casting to a 1498 1499A cast to union type is similar to other casts, except that the type 1500specified is a union type. You can specify the type either with 1501@code{union @var{tag}} or with a typedef name. A cast to union is actually 1502a constructor though, not a cast, and hence does not yield an lvalue like 1503normal casts. (@xref{Compound Literals}.) 1504 1505The types that may be cast to the union type are those of the members 1506of the union. Thus, given the following union and variables: 1507 1508@smallexample 1509union foo @{ int i; double d; @}; 1510int x; 1511double y; 1512@end smallexample 1513 1514@noindent 1515both @code{x} and @code{y} can be cast to type @code{union foo}. 1516 1517Using the cast as the right-hand side of an assignment to a variable of 1518union type is equivalent to storing in a member of the union: 1519 1520@smallexample 1521union foo u; 1522/* @r{@dots{}} */ 1523u = (union foo) x @equiv{} u.i = x 1524u = (union foo) y @equiv{} u.d = y 1525@end smallexample 1526 1527You can also use the union cast as a function argument: 1528 1529@smallexample 1530void hack (union foo); 1531/* @r{@dots{}} */ 1532hack ((union foo) x); 1533@end smallexample 1534 1535@node Mixed Declarations 1536@section Mixed Declarations and Code 1537@cindex mixed declarations and code 1538@cindex declarations, mixed with code 1539@cindex code, mixed with declarations 1540 1541ISO C99 and ISO C++ allow declarations and code to be freely mixed 1542within compound statements. As an extension, GCC also allows this in 1543C89 mode. For example, you could do: 1544 1545@smallexample 1546int i; 1547/* @r{@dots{}} */ 1548i++; 1549int j = i + 2; 1550@end smallexample 1551 1552Each identifier is visible from where it is declared until the end of 1553the enclosing block. 1554 1555@node Function Attributes 1556@section Declaring Attributes of Functions 1557@cindex function attributes 1558@cindex declaring attributes of functions 1559@cindex functions that never return 1560@cindex functions that return more than once 1561@cindex functions that have no side effects 1562@cindex functions in arbitrary sections 1563@cindex functions that behave like malloc 1564@cindex @code{volatile} applied to function 1565@cindex @code{const} applied to function 1566@cindex functions with @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style arguments 1567@cindex functions with non-null pointer arguments 1568@cindex functions that are passed arguments in registers on the 386 1569@cindex functions that pop the argument stack on the 386 1570@cindex functions that do not pop the argument stack on the 386 1571 1572In GNU C, you declare certain things about functions called in your program 1573which help the compiler optimize function calls and check your code more 1574carefully. 1575 1576The keyword @code{__attribute__} allows you to specify special 1577attributes when making a declaration. This keyword is followed by an 1578attribute specification inside double parentheses. The following 1579attributes are currently defined for functions on all targets: 1580@code{noreturn}, @code{returns_twice}, @code{noinline}, @code{always_inline}, 1581@code{flatten}, @code{pure}, @code{const}, @code{nothrow}, @code{sentinel}, 1582@code{format}, @code{format_arg}, @code{no_instrument_function}, 1583@code{section}, @code{constructor}, @code{destructor}, @code{used}, 1584@code{unused}, @code{deprecated}, @code{weak}, @code{malloc}, 1585@code{alias}, @code{warn_unused_result}, @code{nonnull}, 1586@code{gnu_inline} and @code{externally_visible}. Several other 1587attributes are defined for functions on particular target systems. Other 1588attributes, including @code{section} are supported for variables declarations 1589(@pxref{Variable Attributes}) and for types (@pxref{Type Attributes}). 1590 1591You may also specify attributes with @samp{__} preceding and following 1592each keyword. This allows you to use them in header files without 1593being concerned about a possible macro of the same name. For example, 1594you may use @code{__noreturn__} instead of @code{noreturn}. 1595 1596@xref{Attribute Syntax}, for details of the exact syntax for using 1597attributes. 1598 1599@table @code 1600@c Keep this table alphabetized by attribute name. Treat _ as space. 1601 1602@item alias ("@var{target}") 1603@cindex @code{alias} attribute 1604The @code{alias} attribute causes the declaration to be emitted as an 1605alias for another symbol, which must be specified. For instance, 1606 1607@smallexample 1608void __f () @{ /* @r{Do something.} */; @} 1609void f () __attribute__ ((weak, alias ("__f"))); 1610@end smallexample 1611 1612defines @samp{f} to be a weak alias for @samp{__f}. In C++, the 1613mangled name for the target must be used. It is an error if @samp{__f} 1614is not defined in the same translation unit. 1615 1616Not all target machines support this attribute. 1617 1618@item always_inline 1619@cindex @code{always_inline} function attribute 1620Generally, functions are not inlined unless optimization is specified. 1621For functions declared inline, this attribute inlines the function even 1622if no optimization level was specified. 1623 1624@item gnu_inline 1625@cindex @code{gnu_inline} function attribute 1626This attribute should be used with a function which is also declared 1627with the @code{inline} keyword. It directs GCC to treat the function 1628as if it were defined in gnu89 mode even when compiling in C99 or 1629gnu99 mode. 1630 1631If the function is declared @code{extern}, then this definition of the 1632function is used only for inlining. In no case is the function 1633compiled as a standalone function, not even if you take its address 1634explicitly. Such an address becomes an external reference, as if you 1635had only declared the function, and had not defined it. This has 1636almost the effect of a macro. The way to use this is to put a 1637function definition in a header file with this attribute, and put 1638another copy of the function, without @code{extern}, in a library 1639file. The definition in the header file will cause most calls to the 1640function to be inlined. If any uses of the function remain, they will 1641refer to the single copy in the library. Note that the two 1642definitions of the functions need not be precisely the same, although 1643if they do not have the same effect your program may behave oddly. 1644 1645If the function is neither @code{extern} nor @code{static}, then the 1646function is compiled as a standalone function, as well as being 1647inlined where possible. 1648 1649This is how GCC traditionally handled functions declared 1650@code{inline}. Since ISO C99 specifies a different semantics for 1651@code{inline}, this function attribute is provided as a transition 1652measure and as a useful feature in its own right. This attribute is 1653available in GCC 4.1.3 and later. It is available if either of the 1654preprocessor macros @code{__GNUC_GNU_INLINE__} or 1655@code{__GNUC_STDC_INLINE__} are defined. @xref{Inline,,An Inline 1656Function is As Fast As a Macro}. 1657 1658Note that since the first version of GCC to support C99 inline semantics 1659is 4.3, earlier versions of GCC which accept this attribute effectively 1660assume that it is always present, whether or not it is given explicitly. 1661In versions prior to 4.3, the only effect of explicitly including it is 1662to disable warnings about using inline functions in C99 mode. 1663 1664@cindex @code{flatten} function attribute 1665@item flatten 1666Generally, inlining into a function is limited. For a function marked with 1667this attribute, every call inside this function will be inlined, if possible. 1668Whether the function itself is considered for inlining depends on its size and 1669the current inlining parameters. The @code{flatten} attribute only works 1670reliably in unit-at-a-time mode. 1671 1672@item cdecl 1673@cindex functions that do pop the argument stack on the 386 1674@opindex mrtd 1675On the Intel 386, the @code{cdecl} attribute causes the compiler to 1676assume that the calling function will pop off the stack space used to 1677pass arguments. This is 1678useful to override the effects of the @option{-mrtd} switch. 1679 1680@item const 1681@cindex @code{const} function attribute 1682Many functions do not examine any values except their arguments, and 1683have no effects except the return value. Basically this is just slightly 1684more strict class than the @code{pure} attribute below, since function is not 1685allowed to read global memory. 1686 1687@cindex pointer arguments 1688Note that a function that has pointer arguments and examines the data 1689pointed to must @emph{not} be declared @code{const}. Likewise, a 1690function that calls a non-@code{const} function usually must not be 1691@code{const}. It does not make sense for a @code{const} function to 1692return @code{void}. 1693 1694The attribute @code{const} is not implemented in GCC versions earlier 1695than 2.5. An alternative way to declare that a function has no side 1696effects, which works in the current version and in some older versions, 1697is as follows: 1698 1699@smallexample 1700typedef int intfn (); 1701 1702extern const intfn square; 1703@end smallexample 1704 1705This approach does not work in GNU C++ from 2.6.0 on, since the language 1706specifies that the @samp{const} must be attached to the return value. 1707 1708@item constructor 1709@itemx destructor 1710@cindex @code{constructor} function attribute 1711@cindex @code{destructor} function attribute 1712The @code{constructor} attribute causes the function to be called 1713automatically before execution enters @code{main ()}. Similarly, the 1714@code{destructor} attribute causes the function to be called 1715automatically after @code{main ()} has completed or @code{exit ()} has 1716been called. Functions with these attributes are useful for 1717initializing data that will be used implicitly during the execution of 1718the program. 1719 1720@item deprecated 1721@cindex @code{deprecated} attribute. 1722The @code{deprecated} attribute results in a warning if the function 1723is used anywhere in the source file. This is useful when identifying 1724functions that are expected to be removed in a future version of a 1725program. The warning also includes the location of the declaration 1726of the deprecated function, to enable users to easily find further 1727information about why the function is deprecated, or what they should 1728do instead. Note that the warnings only occurs for uses: 1729 1730@smallexample 1731int old_fn () __attribute__ ((deprecated)); 1732int old_fn (); 1733int (*fn_ptr)() = old_fn; 1734@end smallexample 1735 1736results in a warning on line 3 but not line 2. 1737 1738The @code{deprecated} attribute can also be used for variables and 1739types (@pxref{Variable Attributes}, @pxref{Type Attributes}.) 1740 1741@item dllexport 1742@cindex @code{__declspec(dllexport)} 1743On Microsoft Windows targets and Symbian OS targets the 1744@code{dllexport} attribute causes the compiler to provide a global 1745pointer to a pointer in a DLL, so that it can be referenced with the 1746@code{dllimport} attribute. On Microsoft Windows targets, the pointer 1747name is formed by combining @code{_imp__} and the function or variable 1748name. 1749 1750You can use @code{__declspec(dllexport)} as a synonym for 1751@code{__attribute__ ((dllexport))} for compatibility with other 1752compilers. 1753 1754On systems that support the @code{visibility} attribute, this 1755attribute also implies ``default'' visibility, unless a 1756@code{visibility} attribute is explicitly specified. You should avoid 1757the use of @code{dllexport} with ``hidden'' or ``internal'' 1758visibility; in the future GCC may issue an error for those cases. 1759 1760Currently, the @code{dllexport} attribute is ignored for inlined 1761functions, unless the @option{-fkeep-inline-functions} flag has been 1762used. The attribute is also ignored for undefined symbols. 1763 1764When applied to C++ classes, the attribute marks defined non-inlined 1765member functions and static data members as exports. Static consts 1766initialized in-class are not marked unless they are also defined 1767out-of-class. 1768 1769For Microsoft Windows targets there are alternative methods for 1770including the symbol in the DLL's export table such as using a 1771@file{.def} file with an @code{EXPORTS} section or, with GNU ld, using 1772the @option{--export-all} linker flag. 1773 1774@item dllimport 1775@cindex @code{__declspec(dllimport)} 1776On Microsoft Windows and Symbian OS targets, the @code{dllimport} 1777attribute causes the compiler to reference a function or variable via 1778a global pointer to a pointer that is set up by the DLL exporting the 1779symbol. The attribute implies @code{extern} storage. On Microsoft 1780Windows targets, the pointer name is formed by combining @code{_imp__} 1781and the function or variable name. 1782 1783You can use @code{__declspec(dllimport)} as a synonym for 1784@code{__attribute__ ((dllimport))} for compatibility with other 1785compilers. 1786 1787Currently, the attribute is ignored for inlined functions. If the 1788attribute is applied to a symbol @emph{definition}, an error is reported. 1789If a symbol previously declared @code{dllimport} is later defined, the 1790attribute is ignored in subsequent references, and a warning is emitted. 1791The attribute is also overridden by a subsequent declaration as 1792@code{dllexport}. 1793 1794When applied to C++ classes, the attribute marks non-inlined 1795member functions and static data members as imports. However, the 1796attribute is ignored for virtual methods to allow creation of vtables 1797using thunks. 1798 1799On the SH Symbian OS target the @code{dllimport} attribute also has 1800another affect---it can cause the vtable and run-time type information 1801for a class to be exported. This happens when the class has a 1802dllimport'ed constructor or a non-inline, non-pure virtual function 1803and, for either of those two conditions, the class also has a inline 1804constructor or destructor and has a key function that is defined in 1805the current translation unit. 1806 1807For Microsoft Windows based targets the use of the @code{dllimport} 1808attribute on functions is not necessary, but provides a small 1809performance benefit by eliminating a thunk in the DLL@. The use of the 1810@code{dllimport} attribute on imported variables was required on older 1811versions of the GNU linker, but can now be avoided by passing the 1812@option{--enable-auto-import} switch to the GNU linker. As with 1813functions, using the attribute for a variable eliminates a thunk in 1814the DLL@. 1815 1816One drawback to using this attribute is that a pointer to a function 1817or variable marked as @code{dllimport} cannot be used as a constant 1818address. On Microsoft Windows targets, the attribute can be disabled 1819for functions by setting the @option{-mnop-fun-dllimport} flag. 1820 1821@item eightbit_data 1822@cindex eight bit data on the H8/300, H8/300H, and H8S 1823Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified 1824variable should be placed into the eight bit data section. 1825The compiler will generate more efficient code for certain operations 1826on data in the eight bit data area. Note the eight bit data area is limited to 1827256 bytes of data. 1828 1829You must use GAS and GLD from GNU binutils version 2.7 or later for 1830this attribute to work correctly. 1831 1832@item exception_handler 1833@cindex exception handler functions on the Blackfin processor 1834Use this attribute on the Blackfin to indicate that the specified function 1835is an exception handler. The compiler will generate function entry and 1836exit sequences suitable for use in an exception handler when this 1837attribute is present. 1838 1839@item far 1840@cindex functions which handle memory bank switching 1841On 68HC11 and 68HC12 the @code{far} attribute causes the compiler to 1842use a calling convention that takes care of switching memory banks when 1843entering and leaving a function. This calling convention is also the 1844default when using the @option{-mlong-calls} option. 1845 1846On 68HC12 the compiler will use the @code{call} and @code{rtc} instructions 1847to call and return from a function. 1848 1849On 68HC11 the compiler will generate a sequence of instructions 1850to invoke a board-specific routine to switch the memory bank and call the 1851real function. The board-specific routine simulates a @code{call}. 1852At the end of a function, it will jump to a board-specific routine 1853instead of using @code{rts}. The board-specific return routine simulates 1854the @code{rtc}. 1855 1856@item fastcall 1857@cindex functions that pop the argument stack on the 386 1858On the Intel 386, the @code{fastcall} attribute causes the compiler to 1859pass the first argument (if of integral type) in the register ECX and 1860the second argument (if of integral type) in the register EDX@. Subsequent 1861and other typed arguments are passed on the stack. The called function will 1862pop the arguments off the stack. If the number of arguments is variable all 1863arguments are pushed on the stack. 1864 1865@item format (@var{archetype}, @var{string-index}, @var{first-to-check}) 1866@cindex @code{format} function attribute 1867@opindex Wformat 1868The @code{format} attribute specifies that a function takes @code{printf}, 1869@code{scanf}, @code{strftime} or @code{strfmon} style arguments which 1870should be type-checked against a format string. For example, the 1871declaration: 1872 1873@smallexample 1874extern int 1875my_printf (void *my_object, const char *my_format, ...) 1876 __attribute__ ((format (printf, 2, 3))); 1877@end smallexample 1878 1879@noindent 1880causes the compiler to check the arguments in calls to @code{my_printf} 1881for consistency with the @code{printf} style format string argument 1882@code{my_format}. 1883 1884The parameter @var{archetype} determines how the format string is 1885interpreted, and should be @code{printf}, @code{scanf}, @code{strftime} 1886or @code{strfmon}. (You can also use @code{__printf__}, 1887@code{__scanf__}, @code{__strftime__} or @code{__strfmon__}.) The 1888parameter @var{string-index} specifies which argument is the format 1889string argument (starting from 1), while @var{first-to-check} is the 1890number of the first argument to check against the format string. For 1891functions where the arguments are not available to be checked (such as 1892@code{vprintf}), specify the third parameter as zero. In this case the 1893compiler only checks the format string for consistency. For 1894@code{strftime} formats, the third parameter is required to be zero. 1895Since non-static C++ methods have an implicit @code{this} argument, the 1896arguments of such methods should be counted from two, not one, when 1897giving values for @var{string-index} and @var{first-to-check}. 1898 1899In the example above, the format string (@code{my_format}) is the second 1900argument of the function @code{my_print}, and the arguments to check 1901start with the third argument, so the correct parameters for the format 1902attribute are 2 and 3. 1903 1904@opindex ffreestanding 1905@opindex fno-builtin 1906The @code{format} attribute allows you to identify your own functions 1907which take format strings as arguments, so that GCC can check the 1908calls to these functions for errors. The compiler always (unless 1909@option{-ffreestanding} or @option{-fno-builtin} is used) checks formats 1910for the standard library functions @code{printf}, @code{fprintf}, 1911@code{sprintf}, @code{scanf}, @code{fscanf}, @code{sscanf}, @code{strftime}, 1912@code{vprintf}, @code{vfprintf} and @code{vsprintf} whenever such 1913warnings are requested (using @option{-Wformat}), so there is no need to 1914modify the header file @file{stdio.h}. In C99 mode, the functions 1915@code{snprintf}, @code{vsnprintf}, @code{vscanf}, @code{vfscanf} and 1916@code{vsscanf} are also checked. Except in strictly conforming C 1917standard modes, the X/Open function @code{strfmon} is also checked as 1918are @code{printf_unlocked} and @code{fprintf_unlocked}. 1919@xref{C Dialect Options,,Options Controlling C Dialect}. 1920 1921The target may provide additional types of format checks. 1922@xref{Target Format Checks,,Format Checks Specific to Particular 1923Target Machines}. 1924 1925@item format_arg (@var{string-index}) 1926@cindex @code{format_arg} function attribute 1927@opindex Wformat-nonliteral 1928The @code{format_arg} attribute specifies that a function takes a format 1929string for a @code{printf}, @code{scanf}, @code{strftime} or 1930@code{strfmon} style function and modifies it (for example, to translate 1931it into another language), so the result can be passed to a 1932@code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} style 1933function (with the remaining arguments to the format function the same 1934as they would have been for the unmodified string). For example, the 1935declaration: 1936 1937@smallexample 1938extern char * 1939my_dgettext (char *my_domain, const char *my_format) 1940 __attribute__ ((format_arg (2))); 1941@end smallexample 1942 1943@noindent 1944causes the compiler to check the arguments in calls to a @code{printf}, 1945@code{scanf}, @code{strftime} or @code{strfmon} type function, whose 1946format string argument is a call to the @code{my_dgettext} function, for 1947consistency with the format string argument @code{my_format}. If the 1948@code{format_arg} attribute had not been specified, all the compiler 1949could tell in such calls to format functions would be that the format 1950string argument is not constant; this would generate a warning when 1951@option{-Wformat-nonliteral} is used, but the calls could not be checked 1952without the attribute. 1953 1954The parameter @var{string-index} specifies which argument is the format 1955string argument (starting from one). Since non-static C++ methods have 1956an implicit @code{this} argument, the arguments of such methods should 1957be counted from two. 1958 1959The @code{format-arg} attribute allows you to identify your own 1960functions which modify format strings, so that GCC can check the 1961calls to @code{printf}, @code{scanf}, @code{strftime} or @code{strfmon} 1962type function whose operands are a call to one of your own function. 1963The compiler always treats @code{gettext}, @code{dgettext}, and 1964@code{dcgettext} in this manner except when strict ISO C support is 1965requested by @option{-ansi} or an appropriate @option{-std} option, or 1966@option{-ffreestanding} or @option{-fno-builtin} 1967is used. @xref{C Dialect Options,,Options 1968Controlling C Dialect}. 1969 1970@item function_vector 1971@cindex calling functions through the function vector on the H8/300 processors 1972Use this attribute on the H8/300, H8/300H, and H8S to indicate that the specified 1973function should be called through the function vector. Calling a 1974function through the function vector will reduce code size, however; 1975the function vector has a limited size (maximum 128 entries on the H8/300 1976and 64 entries on the H8/300H and H8S) and shares space with the interrupt vector. 1977 1978You must use GAS and GLD from GNU binutils version 2.7 or later for 1979this attribute to work correctly. 1980 1981@item interrupt 1982@cindex interrupt handler functions 1983Use this attribute on the ARM, AVR, C4x, CRX, M32C, M32R/D, MS1, and Xstormy16 1984ports to indicate that the specified function is an interrupt handler. 1985The compiler will generate function entry and exit sequences suitable 1986for use in an interrupt handler when this attribute is present. 1987 1988Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H, H8S, and 1989SH processors can be specified via the @code{interrupt_handler} attribute. 1990 1991Note, on the AVR, interrupts will be enabled inside the function. 1992 1993Note, for the ARM, you can specify the kind of interrupt to be handled by 1994adding an optional parameter to the interrupt attribute like this: 1995 1996@smallexample 1997void f () __attribute__ ((interrupt ("IRQ"))); 1998@end smallexample 1999 2000Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT and UNDEF@. 2001 2002@item interrupt_handler 2003@cindex interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors 2004Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S, and SH to 2005indicate that the specified function is an interrupt handler. The compiler 2006will generate function entry and exit sequences suitable for use in an 2007interrupt handler when this attribute is present. 2008 2009@item kspisusp 2010@cindex User stack pointer in interrupts on the Blackfin 2011When used together with @code{interrupt_handler}, @code{exception_handler} 2012or @code{nmi_handler}, code will be generated to load the stack pointer 2013from the USP register in the function prologue. 2014 2015@item long_call/short_call 2016@cindex indirect calls on ARM 2017This attribute specifies how a particular function is called on 2018ARM@. Both attributes override the @option{-mlong-calls} (@pxref{ARM Options}) 2019command line switch and @code{#pragma long_calls} settings. The 2020@code{long_call} attribute indicates that the function might be far 2021away from the call site and require a different (more expensive) 2022calling sequence. The @code{short_call} attribute always places 2023the offset to the function from the call site into the @samp{BL} 2024instruction directly. 2025 2026@item longcall/shortcall 2027@cindex functions called via pointer on the RS/6000 and PowerPC 2028On the Blackfin, RS/6000 and PowerPC, the @code{longcall} attribute 2029indicates that the function might be far away from the call site and 2030require a different (more expensive) calling sequence. The 2031@code{shortcall} attribute indicates that the function is always close 2032enough for the shorter calling sequence to be used. These attributes 2033override both the @option{-mlongcall} switch and, on the RS/6000 and 2034PowerPC, the @code{#pragma longcall} setting. 2035 2036@xref{RS/6000 and PowerPC Options}, for more information on whether long 2037calls are necessary. 2038 2039@item long_call 2040@cindex indirect calls on MIPS 2041This attribute specifies how a particular function is called on MIPS@. 2042The attribute overrides the @option{-mlong-calls} (@pxref{MIPS Options}) 2043command line switch. This attribute causes the compiler to always call 2044the function by first loading its address into a register, and then using 2045the contents of that register. 2046 2047@item malloc 2048@cindex @code{malloc} attribute 2049The @code{malloc} attribute is used to tell the compiler that a function 2050may be treated as if any non-@code{NULL} pointer it returns cannot 2051alias any other pointer valid when the function returns. 2052This will often improve optimization. 2053Standard functions with this property include @code{malloc} and 2054@code{calloc}. @code{realloc}-like functions have this property as 2055long as the old pointer is never referred to (including comparing it 2056to the new pointer) after the function returns a non-@code{NULL} 2057value. 2058 2059@item model (@var{model-name}) 2060@cindex function addressability on the M32R/D 2061@cindex variable addressability on the IA-64 2062 2063On the M32R/D, use this attribute to set the addressability of an 2064object, and of the code generated for a function. The identifier 2065@var{model-name} is one of @code{small}, @code{medium}, or 2066@code{large}, representing each of the code models. 2067 2068Small model objects live in the lower 16MB of memory (so that their 2069addresses can be loaded with the @code{ld24} instruction), and are 2070callable with the @code{bl} instruction. 2071 2072Medium model objects may live anywhere in the 32-bit address space (the 2073compiler will generate @code{seth/add3} instructions to load their addresses), 2074and are callable with the @code{bl} instruction. 2075 2076Large model objects may live anywhere in the 32-bit address space (the 2077compiler will generate @code{seth/add3} instructions to load their addresses), 2078and may not be reachable with the @code{bl} instruction (the compiler will 2079generate the much slower @code{seth/add3/jl} instruction sequence). 2080 2081On IA-64, use this attribute to set the addressability of an object. 2082At present, the only supported identifier for @var{model-name} is 2083@code{small}, indicating addressability via ``small'' (22-bit) 2084addresses (so that their addresses can be loaded with the @code{addl} 2085instruction). Caveat: such addressing is by definition not position 2086independent and hence this attribute must not be used for objects 2087defined by shared libraries. 2088 2089@item naked 2090@cindex function without a prologue/epilogue code 2091Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate that the 2092specified function does not need prologue/epilogue sequences generated by 2093the compiler. It is up to the programmer to provide these sequences. 2094 2095@item near 2096@cindex functions which do not handle memory bank switching on 68HC11/68HC12 2097On 68HC11 and 68HC12 the @code{near} attribute causes the compiler to 2098use the normal calling convention based on @code{jsr} and @code{rts}. 2099This attribute can be used to cancel the effect of the @option{-mlong-calls} 2100option. 2101 2102@item nesting 2103@cindex Allow nesting in an interrupt handler on the Blackfin processor. 2104Use this attribute together with @code{interrupt_handler}, 2105@code{exception_handler} or @code{nmi_handler} to indicate that the function 2106entry code should enable nested interrupts or exceptions. 2107 2108@item nmi_handler 2109@cindex NMI handler functions on the Blackfin processor 2110Use this attribute on the Blackfin to indicate that the specified function 2111is an NMI handler. The compiler will generate function entry and 2112exit sequences suitable for use in an NMI handler when this 2113attribute is present. 2114 2115@item no_instrument_function 2116@cindex @code{no_instrument_function} function attribute 2117@opindex finstrument-functions 2118If @option{-finstrument-functions} is given, profiling function calls will 2119be generated at entry and exit of most user-compiled functions. 2120Functions with this attribute will not be so instrumented. 2121 2122@item noinline 2123@cindex @code{noinline} function attribute 2124This function attribute prevents a function from being considered for 2125inlining. 2126 2127@item nonnull (@var{arg-index}, @dots{}) 2128@cindex @code{nonnull} function attribute 2129The @code{nonnull} attribute specifies that some function parameters should 2130be non-null pointers. For instance, the declaration: 2131 2132@smallexample 2133extern void * 2134my_memcpy (void *dest, const void *src, size_t len) 2135 __attribute__((nonnull (1, 2))); 2136@end smallexample 2137 2138@noindent 2139causes the compiler to check that, in calls to @code{my_memcpy}, 2140arguments @var{dest} and @var{src} are non-null. If the compiler 2141determines that a null pointer is passed in an argument slot marked 2142as non-null, and the @option{-Wnonnull} option is enabled, a warning 2143is issued. The compiler may also choose to make optimizations based 2144on the knowledge that certain function arguments will not be null. 2145 2146If no argument index list is given to the @code{nonnull} attribute, 2147all pointer arguments are marked as non-null. To illustrate, the 2148following declaration is equivalent to the previous example: 2149 2150@smallexample 2151extern void * 2152my_memcpy (void *dest, const void *src, size_t len) 2153 __attribute__((nonnull)); 2154@end smallexample 2155 2156@item noreturn 2157@cindex @code{noreturn} function attribute 2158A few standard library functions, such as @code{abort} and @code{exit}, 2159cannot return. GCC knows this automatically. Some programs define 2160their own functions that never return. You can declare them 2161@code{noreturn} to tell the compiler this fact. For example, 2162 2163@smallexample 2164@group 2165void fatal () __attribute__ ((noreturn)); 2166 2167void 2168fatal (/* @r{@dots{}} */) 2169@{ 2170 /* @r{@dots{}} */ /* @r{Print error message.} */ /* @r{@dots{}} */ 2171 exit (1); 2172@} 2173@end group 2174@end smallexample 2175 2176The @code{noreturn} keyword tells the compiler to assume that 2177@code{fatal} cannot return. It can then optimize without regard to what 2178would happen if @code{fatal} ever did return. This makes slightly 2179better code. More importantly, it helps avoid spurious warnings of 2180uninitialized variables. 2181 2182The @code{noreturn} keyword does not affect the exceptional path when that 2183applies: a @code{noreturn}-marked function may still return to the caller 2184by throwing an exception or calling @code{longjmp}. 2185 2186Do not assume that registers saved by the calling function are 2187restored before calling the @code{noreturn} function. 2188 2189It does not make sense for a @code{noreturn} function to have a return 2190type other than @code{void}. 2191 2192The attribute @code{noreturn} is not implemented in GCC versions 2193earlier than 2.5. An alternative way to declare that a function does 2194not return, which works in the current version and in some older 2195versions, is as follows: 2196 2197@smallexample 2198typedef void voidfn (); 2199 2200volatile voidfn fatal; 2201@end smallexample 2202 2203This approach does not work in GNU C++. 2204 2205@item nothrow 2206@cindex @code{nothrow} function attribute 2207The @code{nothrow} attribute is used to inform the compiler that a 2208function cannot throw an exception. For example, most functions in 2209the standard C library can be guaranteed not to throw an exception 2210with the notable exceptions of @code{qsort} and @code{bsearch} that 2211take function pointer arguments. The @code{nothrow} attribute is not 2212implemented in GCC versions earlier than 3.3. 2213 2214@item pure 2215@cindex @code{pure} function attribute 2216Many functions have no effects except the return value and their 2217return value depends only on the parameters and/or global variables. 2218Such a function can be subject 2219to common subexpression elimination and loop optimization just as an 2220arithmetic operator would be. These functions should be declared 2221with the attribute @code{pure}. For example, 2222 2223@smallexample 2224int square (int) __attribute__ ((pure)); 2225@end smallexample 2226 2227@noindent 2228says that the hypothetical function @code{square} is safe to call 2229fewer times than the program says. 2230 2231Some of common examples of pure functions are @code{strlen} or @code{memcmp}. 2232Interesting non-pure functions are functions with infinite loops or those 2233depending on volatile memory or other system resource, that may change between 2234two consecutive calls (such as @code{feof} in a multithreading environment). 2235 2236The attribute @code{pure} is not implemented in GCC versions earlier 2237than 2.96. 2238 2239@item regparm (@var{number}) 2240@cindex @code{regparm} attribute 2241@cindex functions that are passed arguments in registers on the 386 2242On the Intel 386, the @code{regparm} attribute causes the compiler to 2243pass arguments number one to @var{number} if they are of integral type 2244in registers EAX, EDX, and ECX instead of on the stack. Functions that 2245take a variable number of arguments will continue to be passed all of their 2246arguments on the stack. 2247 2248Beware that on some ELF systems this attribute is unsuitable for 2249global functions in shared libraries with lazy binding (which is the 2250default). Lazy binding will send the first call via resolving code in 2251the loader, which might assume EAX, EDX and ECX can be clobbered, as 2252per the standard calling conventions. Solaris 8 is affected by this. 2253GNU systems with GLIBC 2.1 or higher, and FreeBSD, are believed to be 2254safe since the loaders there save all registers. (Lazy binding can be 2255disabled with the linker or the loader if desired, to avoid the 2256problem.) 2257 2258@item sseregparm 2259@cindex @code{sseregparm} attribute 2260On the Intel 386 with SSE support, the @code{sseregparm} attribute 2261causes the compiler to pass up to 3 floating point arguments in 2262SSE registers instead of on the stack. Functions that take a 2263variable number of arguments will continue to pass all of their 2264floating point arguments on the stack. 2265 2266@item force_align_arg_pointer 2267@cindex @code{force_align_arg_pointer} attribute 2268On the Intel x86, the @code{force_align_arg_pointer} attribute may be 2269applied to individual function definitions, generating an alternate 2270prologue and epilogue that realigns the runtime stack. This supports 2271mixing legacy codes that run with a 4-byte aligned stack with modern 2272codes that keep a 16-byte stack for SSE compatibility. The alternate 2273prologue and epilogue are slower and bigger than the regular ones, and 2274the alternate prologue requires a scratch register; this lowers the 2275number of registers available if used in conjunction with the 2276@code{regparm} attribute. The @code{force_align_arg_pointer} 2277attribute is incompatible with nested functions; this is considered a 2278hard error. 2279 2280@item returns_twice 2281@cindex @code{returns_twice} attribute 2282The @code{returns_twice} attribute tells the compiler that a function may 2283return more than one time. The compiler will ensure that all registers 2284are dead before calling such a function and will emit a warning about 2285the variables that may be clobbered after the second return from the 2286function. Examples of such functions are @code{setjmp} and @code{vfork}. 2287The @code{longjmp}-like counterpart of such function, if any, might need 2288to be marked with the @code{noreturn} attribute. 2289 2290@item saveall 2291@cindex save all registers on the Blackfin, H8/300, H8/300H, and H8S 2292Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to indicate that 2293all registers except the stack pointer should be saved in the prologue 2294regardless of whether they are used or not. 2295 2296@item section ("@var{section-name}") 2297@cindex @code{section} function attribute 2298Normally, the compiler places the code it generates in the @code{text} section. 2299Sometimes, however, you need additional sections, or you need certain 2300particular functions to appear in special sections. The @code{section} 2301attribute specifies that a function lives in a particular section. 2302For example, the declaration: 2303 2304@smallexample 2305extern void foobar (void) __attribute__ ((section ("bar"))); 2306@end smallexample 2307 2308@noindent 2309puts the function @code{foobar} in the @code{bar} section. 2310 2311Some file formats do not support arbitrary sections so the @code{section} 2312attribute is not available on all platforms. 2313If you need to map the entire contents of a module to a particular 2314section, consider using the facilities of the linker instead. 2315 2316@item sentinel 2317@cindex @code{sentinel} function attribute 2318This function attribute ensures that a parameter in a function call is 2319an explicit @code{NULL}. The attribute is only valid on variadic 2320functions. By default, the sentinel is located at position zero, the 2321last parameter of the function call. If an optional integer position 2322argument P is supplied to the attribute, the sentinel must be located at 2323position P counting backwards from the end of the argument list. 2324 2325@smallexample 2326__attribute__ ((sentinel)) 2327is equivalent to 2328__attribute__ ((sentinel(0))) 2329@end smallexample 2330 2331The attribute is automatically set with a position of 0 for the built-in 2332functions @code{execl} and @code{execlp}. The built-in function 2333@code{execle} has the attribute set with a position of 1. 2334 2335A valid @code{NULL} in this context is defined as zero with any pointer 2336type. If your system defines the @code{NULL} macro with an integer type 2337then you need to add an explicit cast. GCC replaces @code{stddef.h} 2338with a copy that redefines NULL appropriately. 2339 2340The warnings for missing or incorrect sentinels are enabled with 2341@option{-Wformat}. 2342 2343@item short_call 2344See long_call/short_call. 2345 2346@item shortcall 2347See longcall/shortcall. 2348 2349@item signal 2350@cindex signal handler functions on the AVR processors 2351Use this attribute on the AVR to indicate that the specified 2352function is a signal handler. The compiler will generate function 2353entry and exit sequences suitable for use in a signal handler when this 2354attribute is present. Interrupts will be disabled inside the function. 2355 2356@item sp_switch 2357Use this attribute on the SH to indicate an @code{interrupt_handler} 2358function should switch to an alternate stack. It expects a string 2359argument that names a global variable holding the address of the 2360alternate stack. 2361 2362@smallexample 2363void *alt_stack; 2364void f () __attribute__ ((interrupt_handler, 2365 sp_switch ("alt_stack"))); 2366@end smallexample 2367 2368@item stdcall 2369@cindex functions that pop the argument stack on the 386 2370On the Intel 386, the @code{stdcall} attribute causes the compiler to 2371assume that the called function will pop off the stack space used to 2372pass arguments, unless it takes a variable number of arguments. 2373 2374@item tiny_data 2375@cindex tiny data section on the H8/300H and H8S 2376Use this attribute on the H8/300H and H8S to indicate that the specified 2377variable should be placed into the tiny data section. 2378The compiler will generate more efficient code for loads and stores 2379on data in the tiny data section. Note the tiny data area is limited to 2380slightly under 32kbytes of data. 2381 2382@item trap_exit 2383Use this attribute on the SH for an @code{interrupt_handler} to return using 2384@code{trapa} instead of @code{rte}. This attribute expects an integer 2385argument specifying the trap number to be used. 2386 2387@item unused 2388@cindex @code{unused} attribute. 2389This attribute, attached to a function, means that the function is meant 2390to be possibly unused. GCC will not produce a warning for this 2391function. 2392 2393@item used 2394@cindex @code{used} attribute. 2395This attribute, attached to a function, means that code must be emitted 2396for the function even if it appears that the function is not referenced. 2397This is useful, for example, when the function is referenced only in 2398inline assembly. 2399 2400@item visibility ("@var{visibility_type}") 2401@cindex @code{visibility} attribute 2402This attribute affects the linkage of the declaration to which it is attached. 2403There are four supported @var{visibility_type} values: default, 2404hidden, protected or internal visibility. 2405 2406@smallexample 2407void __attribute__ ((visibility ("protected"))) 2408f () @{ /* @r{Do something.} */; @} 2409int i __attribute__ ((visibility ("hidden"))); 2410@end smallexample 2411 2412The possible values of @var{visibility_type} correspond to the 2413visibility settings in the ELF gABI. 2414 2415@table @dfn 2416@c keep this list of visibilities in alphabetical order. 2417 2418@item default 2419Default visibility is the normal case for the object file format. 2420This value is available for the visibility attribute to override other 2421options that may change the assumed visibility of entities. 2422 2423On ELF, default visibility means that the declaration is visible to other 2424modules and, in shared libraries, means that the declared entity may be 2425overridden. 2426 2427On Darwin, default visibility means that the declaration is visible to 2428other modules. 2429 2430Default visibility corresponds to ``external linkage'' in the language. 2431 2432@item hidden 2433Hidden visibility indicates that the entity declared will have a new 2434form of linkage, which we'll call ``hidden linkage''. Two 2435declarations of an object with hidden linkage refer to the same object 2436if they are in the same shared object. 2437 2438@item internal 2439Internal visibility is like hidden visibility, but with additional 2440processor specific semantics. Unless otherwise specified by the 2441psABI, GCC defines internal visibility to mean that a function is 2442@emph{never} called from another module. Compare this with hidden 2443functions which, while they cannot be referenced directly by other 2444modules, can be referenced indirectly via function pointers. By 2445indicating that a function cannot be called from outside the module, 2446GCC may for instance omit the load of a PIC register since it is known 2447that the calling function loaded the correct value. 2448 2449@item protected 2450Protected visibility is like default visibility except that it 2451indicates that references within the defining module will bind to the 2452definition in that module. That is, the declared entity cannot be 2453overridden by another module. 2454 2455@end table 2456 2457All visibilities are supported on many, but not all, ELF targets 2458(supported when the assembler supports the @samp{.visibility} 2459pseudo-op). Default visibility is supported everywhere. Hidden 2460visibility is supported on Darwin targets. 2461 2462The visibility attribute should be applied only to declarations which 2463would otherwise have external linkage. The attribute should be applied 2464consistently, so that the same entity should not be declared with 2465different settings of the attribute. 2466 2467In C++, the visibility attribute applies to types as well as functions 2468and objects, because in C++ types have linkage. A class must not have 2469greater visibility than its non-static data member types and bases, 2470and class members default to the visibility of their class. Also, a 2471declaration without explicit visibility is limited to the visibility 2472of its type. 2473 2474In C++, you can mark member functions and static member variables of a 2475class with the visibility attribute. This is useful if if you know a 2476particular method or static member variable should only be used from 2477one shared object; then you can mark it hidden while the rest of the 2478class has default visibility. Care must be taken to avoid breaking 2479the One Definition Rule; for example, it is usually not useful to mark 2480an inline method as hidden without marking the whole class as hidden. 2481 2482A C++ namespace declaration can also have the visibility attribute. 2483This attribute applies only to the particular namespace body, not to 2484other definitions of the same namespace; it is equivalent to using 2485@samp{#pragma GCC visibility} before and after the namespace 2486definition (@pxref{Visibility Pragmas}). 2487 2488In C++, if a template argument has limited visibility, this 2489restriction is implicitly propagated to the template instantiation. 2490Otherwise, template instantiations and specializations default to the 2491visibility of their template. 2492 2493If both the template and enclosing class have explicit visibility, the 2494visibility from the template is used. 2495 2496@item warn_unused_result 2497@cindex @code{warn_unused_result} attribute 2498The @code{warn_unused_result} attribute causes a warning to be emitted 2499if a caller of the function with this attribute does not use its 2500return value. This is useful for functions where not checking 2501the result is either a security problem or always a bug, such as 2502@code{realloc}. 2503 2504@smallexample 2505int fn () __attribute__ ((warn_unused_result)); 2506int foo () 2507@{ 2508 if (fn () < 0) return -1; 2509 fn (); 2510 return 0; 2511@} 2512@end smallexample 2513 2514results in warning on line 5. 2515 2516@item weak 2517@cindex @code{weak} attribute 2518The @code{weak} attribute causes the declaration to be emitted as a weak 2519symbol rather than a global. This is primarily useful in defining 2520library functions which can be overridden in user code, though it can 2521also be used with non-function declarations. Weak symbols are supported 2522for ELF targets, and also for a.out targets when using the GNU assembler 2523and linker. 2524 2525@item weakref 2526@itemx weakref ("@var{target}") 2527@cindex @code{weakref} attribute 2528The @code{weakref} attribute marks a declaration as a weak reference. 2529Without arguments, it should be accompanied by an @code{alias} attribute 2530naming the target symbol. Optionally, the @var{target} may be given as 2531an argument to @code{weakref} itself. In either case, @code{weakref} 2532implicitly marks the declaration as @code{weak}. Without a 2533@var{target}, given as an argument to @code{weakref} or to @code{alias}, 2534@code{weakref} is equivalent to @code{weak}. 2535 2536@smallexample 2537static int x() __attribute__ ((weakref ("y"))); 2538/* is equivalent to... */ 2539static int x() __attribute__ ((weak, weakref, alias ("y"))); 2540/* and to... */ 2541static int x() __attribute__ ((weakref)); 2542static int x() __attribute__ ((alias ("y"))); 2543@end smallexample 2544 2545A weak reference is an alias that does not by itself require a 2546definition to be given for the target symbol. If the target symbol is 2547only referenced through weak references, then the becomes a @code{weak} 2548undefined symbol. If it is directly referenced, however, then such 2549strong references prevail, and a definition will be required for the 2550symbol, not necessarily in the same translation unit. 2551 2552The effect is equivalent to moving all references to the alias to a 2553separate translation unit, renaming the alias to the aliased symbol, 2554declaring it as weak, compiling the two separate translation units and 2555performing a reloadable link on them. 2556 2557At present, a declaration to which @code{weakref} is attached can 2558only be @code{static}. 2559 2560@item externally_visible 2561@cindex @code{externally_visible} attribute. 2562This attribute, attached to a global variable or function nullify 2563effect of @option{-fwhole-program} command line option, so the object 2564remain visible outside the current compilation unit 2565 2566@end table 2567 2568You can specify multiple attributes in a declaration by separating them 2569by commas within the double parentheses or by immediately following an 2570attribute declaration with another attribute declaration. 2571 2572@cindex @code{#pragma}, reason for not using 2573@cindex pragma, reason for not using 2574Some people object to the @code{__attribute__} feature, suggesting that 2575ISO C's @code{#pragma} should be used instead. At the time 2576@code{__attribute__} was designed, there were two reasons for not doing 2577this. 2578 2579@enumerate 2580@item 2581It is impossible to generate @code{#pragma} commands from a macro. 2582 2583@item 2584There is no telling what the same @code{#pragma} might mean in another 2585compiler. 2586@end enumerate 2587 2588These two reasons applied to almost any application that might have been 2589proposed for @code{#pragma}. It was basically a mistake to use 2590@code{#pragma} for @emph{anything}. 2591 2592The ISO C99 standard includes @code{_Pragma}, which now allows pragmas 2593to be generated from macros. In addition, a @code{#pragma GCC} 2594namespace is now in use for GCC-specific pragmas. However, it has been 2595found convenient to use @code{__attribute__} to achieve a natural 2596attachment of attributes to their corresponding declarations, whereas 2597@code{#pragma GCC} is of use for constructs that do not naturally form 2598part of the grammar. @xref{Other Directives,,Miscellaneous 2599Preprocessing Directives, cpp, The GNU C Preprocessor}. 2600 2601@node Attribute Syntax 2602@section Attribute Syntax 2603@cindex attribute syntax 2604 2605This section describes the syntax with which @code{__attribute__} may be 2606used, and the constructs to which attribute specifiers bind, for the C 2607language. Some details may vary for C++. Because of infelicities in 2608the grammar for attributes, some forms described here may not be 2609successfully parsed in all cases. 2610 2611There are some problems with the semantics of attributes in C++. For 2612example, there are no manglings for attributes, although they may affect 2613code generation, so problems may arise when attributed types are used in 2614conjunction with templates or overloading. Similarly, @code{typeid} 2615does not distinguish between types with different attributes. Support 2616for attributes in C++ may be restricted in future to attributes on 2617declarations only, but not on nested declarators. 2618 2619@xref{Function Attributes}, for details of the semantics of attributes 2620applying to functions. @xref{Variable Attributes}, for details of the 2621semantics of attributes applying to variables. @xref{Type Attributes}, 2622for details of the semantics of attributes applying to structure, union 2623and enumerated types. 2624 2625An @dfn{attribute specifier} is of the form 2626@code{__attribute__ ((@var{attribute-list}))}. An @dfn{attribute list} 2627is a possibly empty comma-separated sequence of @dfn{attributes}, where 2628each attribute is one of the following: 2629 2630@itemize @bullet 2631@item 2632Empty. Empty attributes are ignored. 2633 2634@item 2635A word (which may be an identifier such as @code{unused}, or a reserved 2636word such as @code{const}). 2637 2638@item 2639A word, followed by, in parentheses, parameters for the attribute. 2640These parameters take one of the following forms: 2641 2642@itemize @bullet 2643@item 2644An identifier. For example, @code{mode} attributes use this form. 2645 2646@item 2647An identifier followed by a comma and a non-empty comma-separated list 2648of expressions. For example, @code{format} attributes use this form. 2649 2650@item 2651A possibly empty comma-separated list of expressions. For example, 2652@code{format_arg} attributes use this form with the list being a single 2653integer constant expression, and @code{alias} attributes use this form 2654with the list being a single string constant. 2655@end itemize 2656@end itemize 2657 2658An @dfn{attribute specifier list} is a sequence of one or more attribute 2659specifiers, not separated by any other tokens. 2660 2661In GNU C, an attribute specifier list may appear after the colon following a 2662label, other than a @code{case} or @code{default} label. The only 2663attribute it makes sense to use after a label is @code{unused}. This 2664feature is intended for code generated by programs which contains labels 2665that may be unused but which is compiled with @option{-Wall}. It would 2666not normally be appropriate to use in it human-written code, though it 2667could be useful in cases where the code that jumps to the label is 2668contained within an @code{#ifdef} conditional. GNU C++ does not permit 2669such placement of attribute lists, as it is permissible for a 2670declaration, which could begin with an attribute list, to be labelled in 2671C++. Declarations cannot be labelled in C90 or C99, so the ambiguity 2672does not arise there. 2673 2674An attribute specifier list may appear as part of a @code{struct}, 2675@code{union} or @code{enum} specifier. It may go either immediately 2676after the @code{struct}, @code{union} or @code{enum} keyword, or after 2677the closing brace. The former syntax is preferred. 2678Where attribute specifiers follow the closing brace, they are considered 2679to relate to the structure, union or enumerated type defined, not to any 2680enclosing declaration the type specifier appears in, and the type 2681defined is not complete until after the attribute specifiers. 2682@c Otherwise, there would be the following problems: a shift/reduce 2683@c conflict between attributes binding the struct/union/enum and 2684@c binding to the list of specifiers/qualifiers; and "aligned" 2685@c attributes could use sizeof for the structure, but the size could be 2686@c changed later by "packed" attributes. 2687 2688Otherwise, an attribute specifier appears as part of a declaration, 2689counting declarations of unnamed parameters and type names, and relates 2690to that declaration (which may be nested in another declaration, for 2691example in the case of a parameter declaration), or to a particular declarator 2692within a declaration. Where an 2693attribute specifier is applied to a parameter declared as a function or 2694an array, it should apply to the function or array rather than the 2695pointer to which the parameter is implicitly converted, but this is not 2696yet correctly implemented. 2697 2698Any list of specifiers and qualifiers at the start of a declaration may 2699contain attribute specifiers, whether or not such a list may in that 2700context contain storage class specifiers. (Some attributes, however, 2701are essentially in the nature of storage class specifiers, and only make 2702sense where storage class specifiers may be used; for example, 2703@code{section}.) There is one necessary limitation to this syntax: the 2704first old-style parameter declaration in a function definition cannot 2705begin with an attribute specifier, because such an attribute applies to 2706the function instead by syntax described below (which, however, is not 2707yet implemented in this case). In some other cases, attribute 2708specifiers are permitted by this grammar but not yet supported by the 2709compiler. All attribute specifiers in this place relate to the 2710declaration as a whole. In the obsolescent usage where a type of 2711@code{int} is implied by the absence of type specifiers, such a list of 2712specifiers and qualifiers may be an attribute specifier list with no 2713other specifiers or qualifiers. 2714 2715At present, the first parameter in a function prototype must have some 2716type specifier which is not an attribute specifier; this resolves an 2717ambiguity in the interpretation of @code{void f(int 2718(__attribute__((foo)) x))}, but is subject to change. At present, if 2719the parentheses of a function declarator contain only attributes then 2720those attributes are ignored, rather than yielding an error or warning 2721or implying a single parameter of type int, but this is subject to 2722change. 2723 2724An attribute specifier list may appear immediately before a declarator 2725(other than the first) in a comma-separated list of declarators in a 2726declaration of more than one identifier using a single list of 2727specifiers and qualifiers. Such attribute specifiers apply 2728only to the identifier before whose declarator they appear. For 2729example, in 2730 2731@smallexample 2732__attribute__((noreturn)) void d0 (void), 2733 __attribute__((format(printf, 1, 2))) d1 (const char *, ...), 2734 d2 (void) 2735@end smallexample 2736 2737@noindent 2738the @code{noreturn} attribute applies to all the functions 2739declared; the @code{format} attribute only applies to @code{d1}. 2740 2741An attribute specifier list may appear immediately before the comma, 2742@code{=} or semicolon terminating the declaration of an identifier other 2743than a function definition. At present, such attribute specifiers apply 2744to the declared object or function, but in future they may attach to the 2745outermost adjacent declarator. In simple cases there is no difference, 2746but, for example, in 2747 2748@smallexample 2749void (****f)(void) __attribute__((noreturn)); 2750@end smallexample 2751 2752@noindent 2753at present the @code{noreturn} attribute applies to @code{f}, which 2754causes a warning since @code{f} is not a function, but in future it may 2755apply to the function @code{****f}. The precise semantics of what 2756attributes in such cases will apply to are not yet specified. Where an 2757assembler name for an object or function is specified (@pxref{Asm 2758Labels}), at present the attribute must follow the @code{asm} 2759specification; in future, attributes before the @code{asm} specification 2760may apply to the adjacent declarator, and those after it to the declared 2761object or function. 2762 2763An attribute specifier list may, in future, be permitted to appear after 2764the declarator in a function definition (before any old-style parameter 2765declarations or the function body). 2766 2767Attribute specifiers may be mixed with type qualifiers appearing inside 2768the @code{[]} of a parameter array declarator, in the C99 construct by 2769which such qualifiers are applied to the pointer to which the array is 2770implicitly converted. Such attribute specifiers apply to the pointer, 2771not to the array, but at present this is not implemented and they are 2772ignored. 2773 2774An attribute specifier list may appear at the start of a nested 2775declarator. At present, there are some limitations in this usage: the 2776attributes correctly apply to the declarator, but for most individual 2777attributes the semantics this implies are not implemented. 2778When attribute specifiers follow the @code{*} of a pointer 2779declarator, they may be mixed with any type qualifiers present. 2780The following describes the formal semantics of this syntax. It will make the 2781most sense if you are familiar with the formal specification of 2782declarators in the ISO C standard. 2783 2784Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration @code{T 2785D1}, where @code{T} contains declaration specifiers that specify a type 2786@var{Type} (such as @code{int}) and @code{D1} is a declarator that 2787contains an identifier @var{ident}. The type specified for @var{ident} 2788for derived declarators whose type does not include an attribute 2789specifier is as in the ISO C standard. 2790 2791If @code{D1} has the form @code{( @var{attribute-specifier-list} D )}, 2792and the declaration @code{T D} specifies the type 2793``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then 2794@code{T D1} specifies the type ``@var{derived-declarator-type-list} 2795@var{attribute-specifier-list} @var{Type}'' for @var{ident}. 2796 2797If @code{D1} has the form @code{* 2798@var{type-qualifier-and-attribute-specifier-list} D}, and the 2799declaration @code{T D} specifies the type 2800``@var{derived-declarator-type-list} @var{Type}'' for @var{ident}, then 2801@code{T D1} specifies the type ``@var{derived-declarator-type-list} 2802@var{type-qualifier-and-attribute-specifier-list} @var{Type}'' for 2803@var{ident}. 2804 2805For example, 2806 2807@smallexample 2808void (__attribute__((noreturn)) ****f) (void); 2809@end smallexample 2810 2811@noindent 2812specifies the type ``pointer to pointer to pointer to pointer to 2813non-returning function returning @code{void}''. As another example, 2814 2815@smallexample 2816char *__attribute__((aligned(8))) *f; 2817@end smallexample 2818 2819@noindent 2820specifies the type ``pointer to 8-byte-aligned pointer to @code{char}''. 2821Note again that this does not work with most attributes; for example, 2822the usage of @samp{aligned} and @samp{noreturn} attributes given above 2823is not yet supported. 2824 2825For compatibility with existing code written for compiler versions that 2826did not implement attributes on nested declarators, some laxity is 2827allowed in the placing of attributes. If an attribute that only applies 2828to types is applied to a declaration, it will be treated as applying to 2829the type of that declaration. If an attribute that only applies to 2830declarations is applied to the type of a declaration, it will be treated 2831as applying to that declaration; and, for compatibility with code 2832placing the attributes immediately before the identifier declared, such 2833an attribute applied to a function return type will be treated as 2834applying to the function type, and such an attribute applied to an array 2835element type will be treated as applying to the array type. If an 2836attribute that only applies to function types is applied to a 2837pointer-to-function type, it will be treated as applying to the pointer 2838target type; if such an attribute is applied to a function return type 2839that is not a pointer-to-function type, it will be treated as applying 2840to the function type. 2841 2842@node Function Prototypes 2843@section Prototypes and Old-Style Function Definitions 2844@cindex function prototype declarations 2845@cindex old-style function definitions 2846@cindex promotion of formal parameters 2847 2848GNU C extends ISO C to allow a function prototype to override a later 2849old-style non-prototype definition. Consider the following example: 2850 2851@smallexample 2852/* @r{Use prototypes unless the compiler is old-fashioned.} */ 2853#ifdef __STDC__ 2854#define P(x) x 2855#else 2856#define P(x) () 2857#endif 2858 2859/* @r{Prototype function declaration.} */ 2860int isroot P((uid_t)); 2861 2862/* @r{Old-style function definition.} */ 2863int 2864isroot (x) /* @r{??? lossage here ???} */ 2865 uid_t x; 2866@{ 2867 return x == 0; 2868@} 2869@end smallexample 2870 2871Suppose the type @code{uid_t} happens to be @code{short}. ISO C does 2872not allow this example, because subword arguments in old-style 2873non-prototype definitions are promoted. Therefore in this example the 2874function definition's argument is really an @code{int}, which does not 2875match the prototype argument type of @code{short}. 2876 2877This restriction of ISO C makes it hard to write code that is portable 2878to traditional C compilers, because the programmer does not know 2879whether the @code{uid_t} type is @code{short}, @code{int}, or 2880@code{long}. Therefore, in cases like these GNU C allows a prototype 2881to override a later old-style definition. More precisely, in GNU C, a 2882function prototype argument type overrides the argument type specified 2883by a later old-style definition if the former type is the same as the 2884latter type before promotion. Thus in GNU C the above example is 2885equivalent to the following: 2886 2887@smallexample 2888int isroot (uid_t); 2889 2890int 2891isroot (uid_t x) 2892@{ 2893 return x == 0; 2894@} 2895@end smallexample 2896 2897@noindent 2898GNU C++ does not support old-style function definitions, so this 2899extension is irrelevant. 2900 2901@node C++ Comments 2902@section C++ Style Comments 2903@cindex // 2904@cindex C++ comments 2905@cindex comments, C++ style 2906 2907In GNU C, you may use C++ style comments, which start with @samp{//} and 2908continue until the end of the line. Many other C implementations allow 2909such comments, and they are included in the 1999 C standard. However, 2910C++ style comments are not recognized if you specify an @option{-std} 2911option specifying a version of ISO C before C99, or @option{-ansi} 2912(equivalent to @option{-std=c89}). 2913 2914@node Dollar Signs 2915@section Dollar Signs in Identifier Names 2916@cindex $ 2917@cindex dollar signs in identifier names 2918@cindex identifier names, dollar signs in 2919 2920In GNU C, you may normally use dollar signs in identifier names. 2921This is because many traditional C implementations allow such identifiers. 2922However, dollar signs in identifiers are not supported on a few target 2923machines, typically because the target assembler does not allow them. 2924 2925@node Character Escapes 2926@section The Character @key{ESC} in Constants 2927 2928You can use the sequence @samp{\e} in a string or character constant to 2929stand for the ASCII character @key{ESC}. 2930 2931@node Alignment 2932@section Inquiring on Alignment of Types or Variables 2933@cindex alignment 2934@cindex type alignment 2935@cindex variable alignment 2936 2937The keyword @code{__alignof__} allows you to inquire about how an object 2938is aligned, or the minimum alignment usually required by a type. Its 2939syntax is just like @code{sizeof}. 2940 2941For example, if the target machine requires a @code{double} value to be 2942aligned on an 8-byte boundary, then @code{__alignof__ (double)} is 8. 2943This is true on many RISC machines. On more traditional machine 2944designs, @code{__alignof__ (double)} is 4 or even 2. 2945 2946Some machines never actually require alignment; they allow reference to any 2947data type even at an odd address. For these machines, @code{__alignof__} 2948reports the @emph{recommended} alignment of a type. 2949 2950If the operand of @code{__alignof__} is an lvalue rather than a type, 2951its value is the required alignment for its type, taking into account 2952any minimum alignment specified with GCC's @code{__attribute__} 2953extension (@pxref{Variable Attributes}). For example, after this 2954declaration: 2955 2956@smallexample 2957struct foo @{ int x; char y; @} foo1; 2958@end smallexample 2959 2960@noindent 2961the value of @code{__alignof__ (foo1.y)} is 1, even though its actual 2962alignment is probably 2 or 4, the same as @code{__alignof__ (int)}. 2963 2964It is an error to ask for the alignment of an incomplete type. 2965 2966@node Variable Attributes 2967@section Specifying Attributes of Variables 2968@cindex attribute of variables 2969@cindex variable attributes 2970 2971The keyword @code{__attribute__} allows you to specify special 2972attributes of variables or structure fields. This keyword is followed 2973by an attribute specification inside double parentheses. Some 2974attributes are currently defined generically for variables. 2975Other attributes are defined for variables on particular target 2976systems. Other attributes are available for functions 2977(@pxref{Function Attributes}) and for types (@pxref{Type Attributes}). 2978Other front ends might define more attributes 2979(@pxref{C++ Extensions,,Extensions to the C++ Language}). 2980 2981You may also specify attributes with @samp{__} preceding and following 2982each keyword. This allows you to use them in header files without 2983being concerned about a possible macro of the same name. For example, 2984you may use @code{__aligned__} instead of @code{aligned}. 2985 2986@xref{Attribute Syntax}, for details of the exact syntax for using 2987attributes. 2988 2989@table @code 2990@cindex @code{aligned} attribute 2991@item aligned (@var{alignment}) 2992This attribute specifies a minimum alignment for the variable or 2993structure field, measured in bytes. For example, the declaration: 2994 2995@smallexample 2996int x __attribute__ ((aligned (16))) = 0; 2997@end smallexample 2998 2999@noindent 3000causes the compiler to allocate the global variable @code{x} on a 300116-byte boundary. On a 68040, this could be used in conjunction with 3002an @code{asm} expression to access the @code{move16} instruction which 3003requires 16-byte aligned operands. 3004 3005You can also specify the alignment of structure fields. For example, to 3006create a double-word aligned @code{int} pair, you could write: 3007 3008@smallexample 3009struct foo @{ int x[2] __attribute__ ((aligned (8))); @}; 3010@end smallexample 3011 3012@noindent 3013This is an alternative to creating a union with a @code{double} member 3014that forces the union to be double-word aligned. 3015 3016As in the preceding examples, you can explicitly specify the alignment 3017(in bytes) that you wish the compiler to use for a given variable or 3018structure field. Alternatively, you can leave out the alignment factor 3019and just ask the compiler to align a variable or field to the maximum 3020useful alignment for the target machine you are compiling for. For 3021example, you could write: 3022 3023@smallexample 3024short array[3] __attribute__ ((aligned)); 3025@end smallexample 3026 3027Whenever you leave out the alignment factor in an @code{aligned} attribute 3028specification, the compiler automatically sets the alignment for the declared 3029variable or field to the largest alignment which is ever used for any data 3030type on the target machine you are compiling for. Doing this can often make 3031copy operations more efficient, because the compiler can use whatever 3032instructions copy the biggest chunks of memory when performing copies to 3033or from the variables or fields that you have aligned this way. 3034 3035The @code{aligned} attribute can only increase the alignment; but you 3036can decrease it by specifying @code{packed} as well. See below. 3037 3038Note that the effectiveness of @code{aligned} attributes may be limited 3039by inherent limitations in your linker. On many systems, the linker is 3040only able to arrange for variables to be aligned up to a certain maximum 3041alignment. (For some linkers, the maximum supported alignment may 3042be very very small.) If your linker is only able to align variables 3043up to a maximum of 8 byte alignment, then specifying @code{aligned(16)} 3044in an @code{__attribute__} will still only provide you with 8 byte 3045alignment. See your linker documentation for further information. 3046 3047@item cleanup (@var{cleanup_function}) 3048@cindex @code{cleanup} attribute 3049The @code{cleanup} attribute runs a function when the variable goes 3050out of scope. This attribute can only be applied to auto function 3051scope variables; it may not be applied to parameters or variables 3052with static storage duration. The function must take one parameter, 3053a pointer to a type compatible with the variable. The return value 3054of the function (if any) is ignored. 3055 3056If @option{-fexceptions} is enabled, then @var{cleanup_function} 3057will be run during the stack unwinding that happens during the 3058processing of the exception. Note that the @code{cleanup} attribute 3059does not allow the exception to be caught, only to perform an action. 3060It is undefined what happens if @var{cleanup_function} does not 3061return normally. 3062 3063@item common 3064@itemx nocommon 3065@cindex @code{common} attribute 3066@cindex @code{nocommon} attribute 3067@opindex fcommon 3068@opindex fno-common 3069The @code{common} attribute requests GCC to place a variable in 3070``common'' storage. The @code{nocommon} attribute requests the 3071opposite---to allocate space for it directly. 3072 3073These attributes override the default chosen by the 3074@option{-fno-common} and @option{-fcommon} flags respectively. 3075 3076@item deprecated 3077@cindex @code{deprecated} attribute 3078The @code{deprecated} attribute results in a warning if the variable 3079is used anywhere in the source file. This is useful when identifying 3080variables that are expected to be removed in a future version of a 3081program. The warning also includes the location of the declaration 3082of the deprecated variable, to enable users to easily find further 3083information about why the variable is deprecated, or what they should 3084do instead. Note that the warning only occurs for uses: 3085 3086@smallexample 3087extern int old_var __attribute__ ((deprecated)); 3088extern int old_var; 3089int new_fn () @{ return old_var; @} 3090@end smallexample 3091 3092results in a warning on line 3 but not line 2. 3093 3094The @code{deprecated} attribute can also be used for functions and 3095types (@pxref{Function Attributes}, @pxref{Type Attributes}.) 3096 3097@item mode (@var{mode}) 3098@cindex @code{mode} attribute 3099This attribute specifies the data type for the declaration---whichever 3100type corresponds to the mode @var{mode}. This in effect lets you 3101request an integer or floating point type according to its width. 3102 3103You may also specify a mode of @samp{byte} or @samp{__byte__} to 3104indicate the mode corresponding to a one-byte integer, @samp{word} or 3105@samp{__word__} for the mode of a one-word integer, and @samp{pointer} 3106or @samp{__pointer__} for the mode used to represent pointers. 3107 3108@item packed 3109@cindex @code{packed} attribute 3110The @code{packed} attribute specifies that a variable or structure field 3111should have the smallest possible alignment---one byte for a variable, 3112and one bit for a field, unless you specify a larger value with the 3113@code{aligned} attribute. 3114 3115Here is a structure in which the field @code{x} is packed, so that it 3116immediately follows @code{a}: 3117 3118@smallexample 3119struct foo 3120@{ 3121 char a; 3122 int x[2] __attribute__ ((packed)); 3123@}; 3124@end smallexample 3125 3126@item section ("@var{section-name}") 3127@cindex @code{section} variable attribute 3128Normally, the compiler places the objects it generates in sections like 3129@code{data} and @code{bss}. Sometimes, however, you need additional sections, 3130or you need certain particular variables to appear in special sections, 3131for example to map to special hardware. The @code{section} 3132attribute specifies that a variable (or function) lives in a particular 3133section. For example, this small program uses several specific section names: 3134 3135@smallexample 3136struct duart a __attribute__ ((section ("DUART_A"))) = @{ 0 @}; 3137struct duart b __attribute__ ((section ("DUART_B"))) = @{ 0 @}; 3138char stack[10000] __attribute__ ((section ("STACK"))) = @{ 0 @}; 3139int init_data __attribute__ ((section ("INITDATA"))) = 0; 3140 3141main() 3142@{ 3143 /* @r{Initialize stack pointer} */ 3144 init_sp (stack + sizeof (stack)); 3145 3146 /* @r{Initialize initialized data} */ 3147 memcpy (&init_data, &data, &edata - &data); 3148 3149 /* @r{Turn on the serial ports} */ 3150 init_duart (&a); 3151 init_duart (&b); 3152@} 3153@end smallexample 3154 3155@noindent 3156Use the @code{section} attribute with an @emph{initialized} definition 3157of a @emph{global} variable, as shown in the example. GCC issues 3158a warning and otherwise ignores the @code{section} attribute in 3159uninitialized variable declarations. 3160 3161You may only use the @code{section} attribute with a fully initialized 3162global definition because of the way linkers work. The linker requires 3163each object be defined once, with the exception that uninitialized 3164variables tentatively go in the @code{common} (or @code{bss}) section 3165and can be multiply ``defined''. You can force a variable to be 3166initialized with the @option{-fno-common} flag or the @code{nocommon} 3167attribute. 3168 3169Some file formats do not support arbitrary sections so the @code{section} 3170attribute is not available on all platforms. 3171If you need to map the entire contents of a module to a particular 3172section, consider using the facilities of the linker instead. 3173 3174@item shared 3175@cindex @code{shared} variable attribute 3176On Microsoft Windows, in addition to putting variable definitions in a named 3177section, the section can also be shared among all running copies of an 3178executable or DLL@. For example, this small program defines shared data 3179by putting it in a named section @code{shared} and marking the section 3180shareable: 3181 3182@smallexample 3183int foo __attribute__((section ("shared"), shared)) = 0; 3184 3185int 3186main() 3187@{ 3188 /* @r{Read and write foo. All running 3189 copies see the same value.} */ 3190 return 0; 3191@} 3192@end smallexample 3193 3194@noindent 3195You may only use the @code{shared} attribute along with @code{section} 3196attribute with a fully initialized global definition because of the way 3197linkers work. See @code{section} attribute for more information. 3198 3199The @code{shared} attribute is only available on Microsoft Windows@. 3200 3201@item tls_model ("@var{tls_model}") 3202@cindex @code{tls_model} attribute 3203The @code{tls_model} attribute sets thread-local storage model 3204(@pxref{Thread-Local}) of a particular @code{__thread} variable, 3205overriding @option{-ftls-model=} command line switch on a per-variable 3206basis. 3207The @var{tls_model} argument should be one of @code{global-dynamic}, 3208@code{local-dynamic}, @code{initial-exec} or @code{local-exec}. 3209 3210Not all targets support this attribute. 3211 3212@item unused 3213This attribute, attached to a variable, means that the variable is meant 3214to be possibly unused. GCC will not produce a warning for this 3215variable. 3216 3217@item used 3218This attribute, attached to a variable, means that the variable must be 3219emitted even if it appears that the variable is not referenced. 3220 3221@item vector_size (@var{bytes}) 3222This attribute specifies the vector size for the variable, measured in 3223bytes. For example, the declaration: 3224 3225@smallexample 3226int foo __attribute__ ((vector_size (16))); 3227@end smallexample 3228 3229@noindent 3230causes the compiler to set the mode for @code{foo}, to be 16 bytes, 3231divided into @code{int} sized units. Assuming a 32-bit int (a vector of 32324 units of 4 bytes), the corresponding mode of @code{foo} will be V4SI@. 3233 3234This attribute is only applicable to integral and float scalars, 3235although arrays, pointers, and function return values are allowed in 3236conjunction with this construct. 3237 3238Aggregates with this attribute are invalid, even if they are of the same 3239size as a corresponding scalar. For example, the declaration: 3240 3241@smallexample 3242struct S @{ int a; @}; 3243struct S __attribute__ ((vector_size (16))) foo; 3244@end smallexample 3245 3246@noindent 3247is invalid even if the size of the structure is the same as the size of 3248the @code{int}. 3249 3250@item selectany 3251The @code{selectany} attribute causes an initialized global variable to 3252have link-once semantics. When multiple definitions of the variable are 3253encountered by the linker, the first is selected and the remainder are 3254discarded. Following usage by the Microsoft compiler, the linker is told 3255@emph{not} to warn about size or content differences of the multiple 3256definitions. 3257 3258Although the primary usage of this attribute is for POD types, the 3259attribute can also be applied to global C++ objects that are initialized 3260by a constructor. In this case, the static initialization and destruction 3261code for the object is emitted in each translation defining the object, 3262but the calls to the constructor and destructor are protected by a 3263link-once guard variable. 3264 3265The @code{selectany} attribute is only available on Microsoft Windows 3266targets. You can use @code{__declspec (selectany)} as a synonym for 3267@code{__attribute__ ((selectany))} for compatibility with other 3268compilers. 3269 3270@item weak 3271The @code{weak} attribute is described in @xref{Function Attributes}. 3272 3273@item dllimport 3274The @code{dllimport} attribute is described in @xref{Function Attributes}. 3275 3276@item dllexport 3277The @code{dllexport} attribute is described in @xref{Function Attributes}. 3278 3279@end table 3280 3281@subsection M32R/D Variable Attributes 3282 3283One attribute is currently defined for the M32R/D@. 3284 3285@table @code 3286@item model (@var{model-name}) 3287@cindex variable addressability on the M32R/D 3288Use this attribute on the M32R/D to set the addressability of an object. 3289The identifier @var{model-name} is one of @code{small}, @code{medium}, 3290or @code{large}, representing each of the code models. 3291 3292Small model objects live in the lower 16MB of memory (so that their 3293addresses can be loaded with the @code{ld24} instruction). 3294 3295Medium and large model objects may live anywhere in the 32-bit address space 3296(the compiler will generate @code{seth/add3} instructions to load their 3297addresses). 3298@end table 3299 3300@anchor{i386 Variable Attributes} 3301@subsection i386 Variable Attributes 3302 3303Two attributes are currently defined for i386 configurations: 3304@code{ms_struct} and @code{gcc_struct} 3305 3306@table @code 3307@item ms_struct 3308@itemx gcc_struct 3309@cindex @code{ms_struct} attribute 3310@cindex @code{gcc_struct} attribute 3311 3312If @code{packed} is used on a structure, or if bit-fields are used 3313it may be that the Microsoft ABI packs them differently 3314than GCC would normally pack them. Particularly when moving packed 3315data between functions compiled with GCC and the native Microsoft compiler 3316(either via function call or as data in a file), it may be necessary to access 3317either format. 3318 3319Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86 3320compilers to match the native Microsoft compiler. 3321 3322The Microsoft structure layout algorithm is fairly simple with the exception 3323of the bitfield packing: 3324 3325The padding and alignment of members of structures and whether a bit field 3326can straddle a storage-unit boundary 3327 3328@enumerate 3329@item Structure members are stored sequentially in the order in which they are 3330declared: the first member has the lowest memory address and the last member 3331the highest. 3332 3333@item Every data object has an alignment-requirement. The alignment-requirement 3334for all data except structures, unions, and arrays is either the size of the 3335object or the current packing size (specified with either the aligned attribute 3336or the pack pragma), whichever is less. For structures, unions, and arrays, 3337the alignment-requirement is the largest alignment-requirement of its members. 3338Every object is allocated an offset so that: 3339 3340offset % alignment-requirement == 0 3341 3342@item Adjacent bit fields are packed into the same 1-, 2-, or 4-byte allocation 3343unit if the integral types are the same size and if the next bit field fits 3344into the current allocation unit without crossing the boundary imposed by the 3345common alignment requirements of the bit fields. 3346@end enumerate 3347 3348Handling of zero-length bitfields: 3349 3350MSVC interprets zero-length bitfields in the following ways: 3351 3352@enumerate 3353@item If a zero-length bitfield is inserted between two bitfields that would 3354normally be coalesced, the bitfields will not be coalesced. 3355 3356For example: 3357 3358@smallexample 3359struct 3360 @{ 3361 unsigned long bf_1 : 12; 3362 unsigned long : 0; 3363 unsigned long bf_2 : 12; 3364 @} t1; 3365@end smallexample 3366 3367The size of @code{t1} would be 8 bytes with the zero-length bitfield. If the 3368zero-length bitfield were removed, @code{t1}'s size would be 4 bytes. 3369 3370@item If a zero-length bitfield is inserted after a bitfield, @code{foo}, and the 3371alignment of the zero-length bitfield is greater than the member that follows it, 3372@code{bar}, @code{bar} will be aligned as the type of the zero-length bitfield. 3373 3374For example: 3375 3376@smallexample 3377struct 3378 @{ 3379 char foo : 4; 3380 short : 0; 3381 char bar; 3382 @} t2; 3383 3384struct 3385 @{ 3386 char foo : 4; 3387 short : 0; 3388 double bar; 3389 @} t3; 3390@end smallexample 3391 3392For @code{t2}, @code{bar} will be placed at offset 2, rather than offset 1. 3393Accordingly, the size of @code{t2} will be 4. For @code{t3}, the zero-length 3394bitfield will not affect the alignment of @code{bar} or, as a result, the size 3395of the structure. 3396 3397Taking this into account, it is important to note the following: 3398 3399@enumerate 3400@item If a zero-length bitfield follows a normal bitfield, the type of the 3401zero-length bitfield may affect the alignment of the structure as whole. For 3402example, @code{t2} has a size of 4 bytes, since the zero-length bitfield follows a 3403normal bitfield, and is of type short. 3404 3405@item Even if a zero-length bitfield is not followed by a normal bitfield, it may 3406still affect the alignment of the structure: 3407 3408@smallexample 3409struct 3410 @{ 3411 char foo : 6; 3412 long : 0; 3413 @} t4; 3414@end smallexample 3415 3416Here, @code{t4} will take up 4 bytes. 3417@end enumerate 3418 3419@item Zero-length bitfields following non-bitfield members are ignored: 3420 3421@smallexample 3422struct 3423 @{ 3424 char foo; 3425 long : 0; 3426 char bar; 3427 @} t5; 3428@end smallexample 3429 3430Here, @code{t5} will take up 2 bytes. 3431@end enumerate 3432@end table 3433 3434@subsection PowerPC Variable Attributes 3435 3436Three attributes currently are defined for PowerPC configurations: 3437@code{altivec}, @code{ms_struct} and @code{gcc_struct}. 3438 3439For full documentation of the struct attributes please see the 3440documentation in the @xref{i386 Variable Attributes}, section. 3441 3442For documentation of @code{altivec} attribute please see the 3443documentation in the @xref{PowerPC Type Attributes}, section. 3444 3445@subsection Xstormy16 Variable Attributes 3446 3447One attribute is currently defined for xstormy16 configurations: 3448@code{below100} 3449 3450@table @code 3451@item below100 3452@cindex @code{below100} attribute 3453 3454If a variable has the @code{below100} attribute (@code{BELOW100} is 3455allowed also), GCC will place the variable in the first 0x100 bytes of 3456memory and use special opcodes to access it. Such variables will be 3457placed in either the @code{.bss_below100} section or the 3458@code{.data_below100} section. 3459 3460@end table 3461 3462@node Type Attributes 3463@section Specifying Attributes of Types 3464@cindex attribute of types 3465@cindex type attributes 3466 3467The keyword @code{__attribute__} allows you to specify special 3468attributes of @code{struct} and @code{union} types when you define 3469such types. This keyword is followed by an attribute specification 3470inside double parentheses. Seven attributes are currently defined for 3471types: @code{aligned}, @code{packed}, @code{transparent_union}, 3472@code{unused}, @code{deprecated}, @code{visibility}, and 3473@code{may_alias}. Other attributes are defined for functions 3474(@pxref{Function Attributes}) and for variables (@pxref{Variable 3475Attributes}). 3476 3477You may also specify any one of these attributes with @samp{__} 3478preceding and following its keyword. This allows you to use these 3479attributes in header files without being concerned about a possible 3480macro of the same name. For example, you may use @code{__aligned__} 3481instead of @code{aligned}. 3482 3483You may specify type attributes either in a @code{typedef} declaration 3484or in an enum, struct or union type declaration or definition. 3485 3486For an enum, struct or union type, you may specify attributes either 3487between the enum, struct or union tag and the name of the type, or 3488just past the closing curly brace of the @emph{definition}. The 3489former syntax is preferred. 3490 3491@xref{Attribute Syntax}, for details of the exact syntax for using 3492attributes. 3493 3494@table @code 3495@cindex @code{aligned} attribute 3496@item aligned (@var{alignment}) 3497This attribute specifies a minimum alignment (in bytes) for variables 3498of the specified type. For example, the declarations: 3499 3500@smallexample 3501struct S @{ short f[3]; @} __attribute__ ((aligned (8))); 3502typedef int more_aligned_int __attribute__ ((aligned (8))); 3503@end smallexample 3504 3505@noindent 3506force the compiler to insure (as far as it can) that each variable whose 3507type is @code{struct S} or @code{more_aligned_int} will be allocated and 3508aligned @emph{at least} on a 8-byte boundary. On a SPARC, having all 3509variables of type @code{struct S} aligned to 8-byte boundaries allows 3510the compiler to use the @code{ldd} and @code{std} (doubleword load and 3511store) instructions when copying one variable of type @code{struct S} to 3512another, thus improving run-time efficiency. 3513 3514Note that the alignment of any given @code{struct} or @code{union} type 3515is required by the ISO C standard to be at least a perfect multiple of 3516the lowest common multiple of the alignments of all of the members of 3517the @code{struct} or @code{union} in question. This means that you @emph{can} 3518effectively adjust the alignment of a @code{struct} or @code{union} 3519type by attaching an @code{aligned} attribute to any one of the members 3520of such a type, but the notation illustrated in the example above is a 3521more obvious, intuitive, and readable way to request the compiler to 3522adjust the alignment of an entire @code{struct} or @code{union} type. 3523 3524As in the preceding example, you can explicitly specify the alignment 3525(in bytes) that you wish the compiler to use for a given @code{struct} 3526or @code{union} type. Alternatively, you can leave out the alignment factor 3527and just ask the compiler to align a type to the maximum 3528useful alignment for the target machine you are compiling for. For 3529example, you could write: 3530 3531@smallexample 3532struct S @{ short f[3]; @} __attribute__ ((aligned)); 3533@end smallexample 3534 3535Whenever you leave out the alignment factor in an @code{aligned} 3536attribute specification, the compiler automatically sets the alignment 3537for the type to the largest alignment which is ever used for any data 3538type on the target machine you are compiling for. Doing this can often 3539make copy operations more efficient, because the compiler can use 3540whatever instructions copy the biggest chunks of memory when performing 3541copies to or from the variables which have types that you have aligned 3542this way. 3543 3544In the example above, if the size of each @code{short} is 2 bytes, then 3545the size of the entire @code{struct S} type is 6 bytes. The smallest 3546power of two which is greater than or equal to that is 8, so the 3547compiler sets the alignment for the entire @code{struct S} type to 8 3548bytes. 3549 3550Note that although you can ask the compiler to select a time-efficient 3551alignment for a given type and then declare only individual stand-alone 3552objects of that type, the compiler's ability to select a time-efficient 3553alignment is primarily useful only when you plan to create arrays of 3554variables having the relevant (efficiently aligned) type. If you 3555declare or use arrays of variables of an efficiently-aligned type, then 3556it is likely that your program will also be doing pointer arithmetic (or 3557subscripting, which amounts to the same thing) on pointers to the 3558relevant type, and the code that the compiler generates for these 3559pointer arithmetic operations will often be more efficient for 3560efficiently-aligned types than for other types. 3561 3562The @code{aligned} attribute can only increase the alignment; but you 3563can decrease it by specifying @code{packed} as well. See below. 3564 3565Note that the effectiveness of @code{aligned} attributes may be limited 3566by inherent limitations in your linker. On many systems, the linker is 3567only able to arrange for variables to be aligned up to a certain maximum 3568alignment. (For some linkers, the maximum supported alignment may 3569be very very small.) If your linker is only able to align variables 3570up to a maximum of 8 byte alignment, then specifying @code{aligned(16)} 3571in an @code{__attribute__} will still only provide you with 8 byte 3572alignment. See your linker documentation for further information. 3573 3574@item packed 3575This attribute, attached to @code{struct} or @code{union} type 3576definition, specifies that each member (other than zero-width bitfields) 3577of the structure or union is placed to minimize the memory required. When 3578attached to an @code{enum} definition, it indicates that the smallest 3579integral type should be used. 3580 3581@opindex fshort-enums 3582Specifying this attribute for @code{struct} and @code{union} types is 3583equivalent to specifying the @code{packed} attribute on each of the 3584structure or union members. Specifying the @option{-fshort-enums} 3585flag on the line is equivalent to specifying the @code{packed} 3586attribute on all @code{enum} definitions. 3587 3588In the following example @code{struct my_packed_struct}'s members are 3589packed closely together, but the internal layout of its @code{s} member 3590is not packed---to do that, @code{struct my_unpacked_struct} would need to 3591be packed too. 3592 3593@smallexample 3594struct my_unpacked_struct 3595 @{ 3596 char c; 3597 int i; 3598 @}; 3599 3600struct __attribute__ ((__packed__)) my_packed_struct 3601 @{ 3602 char c; 3603 int i; 3604 struct my_unpacked_struct s; 3605 @}; 3606@end smallexample 3607 3608You may only specify this attribute on the definition of a @code{enum}, 3609@code{struct} or @code{union}, not on a @code{typedef} which does not 3610also define the enumerated type, structure or union. 3611 3612@item transparent_union 3613This attribute, attached to a @code{union} type definition, indicates 3614that any function parameter having that union type causes calls to that 3615function to be treated in a special way. 3616 3617First, the argument corresponding to a transparent union type can be of 3618any type in the union; no cast is required. Also, if the union contains 3619a pointer type, the corresponding argument can be a null pointer 3620constant or a void pointer expression; and if the union contains a void 3621pointer type, the corresponding argument can be any pointer expression. 3622If the union member type is a pointer, qualifiers like @code{const} on 3623the referenced type must be respected, just as with normal pointer 3624conversions. 3625 3626Second, the argument is passed to the function using the calling 3627conventions of the first member of the transparent union, not the calling 3628conventions of the union itself. All members of the union must have the 3629same machine representation; this is necessary for this argument passing 3630to work properly. 3631 3632Transparent unions are designed for library functions that have multiple 3633interfaces for compatibility reasons. For example, suppose the 3634@code{wait} function must accept either a value of type @code{int *} to 3635comply with Posix, or a value of type @code{union wait *} to comply with 3636the 4.1BSD interface. If @code{wait}'s parameter were @code{void *}, 3637@code{wait} would accept both kinds of arguments, but it would also 3638accept any other pointer type and this would make argument type checking 3639less useful. Instead, @code{<sys/wait.h>} might define the interface 3640as follows: 3641 3642@smallexample 3643typedef union 3644 @{ 3645 int *__ip; 3646 union wait *__up; 3647 @} wait_status_ptr_t __attribute__ ((__transparent_union__)); 3648 3649pid_t wait (wait_status_ptr_t); 3650@end smallexample 3651 3652This interface allows either @code{int *} or @code{union wait *} 3653arguments to be passed, using the @code{int *} calling convention. 3654The program can call @code{wait} with arguments of either type: 3655 3656@smallexample 3657int w1 () @{ int w; return wait (&w); @} 3658int w2 () @{ union wait w; return wait (&w); @} 3659@end smallexample 3660 3661With this interface, @code{wait}'s implementation might look like this: 3662 3663@smallexample 3664pid_t wait (wait_status_ptr_t p) 3665@{ 3666 return waitpid (-1, p.__ip, 0); 3667@} 3668@end smallexample 3669 3670@item unused 3671When attached to a type (including a @code{union} or a @code{struct}), 3672this attribute means that variables of that type are meant to appear 3673possibly unused. GCC will not produce a warning for any variables of 3674that type, even if the variable appears to do nothing. This is often 3675the case with lock or thread classes, which are usually defined and then 3676not referenced, but contain constructors and destructors that have 3677nontrivial bookkeeping functions. 3678 3679@item deprecated 3680The @code{deprecated} attribute results in a warning if the type 3681is used anywhere in the source file. This is useful when identifying 3682types that are expected to be removed in a future version of a program. 3683If possible, the warning also includes the location of the declaration 3684of the deprecated type, to enable users to easily find further 3685information about why the type is deprecated, or what they should do 3686instead. Note that the warnings only occur for uses and then only 3687if the type is being applied to an identifier that itself is not being 3688declared as deprecated. 3689 3690@smallexample 3691typedef int T1 __attribute__ ((deprecated)); 3692T1 x; 3693typedef T1 T2; 3694T2 y; 3695typedef T1 T3 __attribute__ ((deprecated)); 3696T3 z __attribute__ ((deprecated)); 3697@end smallexample 3698 3699results in a warning on line 2 and 3 but not lines 4, 5, or 6. No 3700warning is issued for line 4 because T2 is not explicitly 3701deprecated. Line 5 has no warning because T3 is explicitly 3702deprecated. Similarly for line 6. 3703 3704The @code{deprecated} attribute can also be used for functions and 3705variables (@pxref{Function Attributes}, @pxref{Variable Attributes}.) 3706 3707@item may_alias 3708Accesses to objects with types with this attribute are not subjected to 3709type-based alias analysis, but are instead assumed to be able to alias 3710any other type of objects, just like the @code{char} type. See 3711@option{-fstrict-aliasing} for more information on aliasing issues. 3712 3713Example of use: 3714 3715@smallexample 3716typedef short __attribute__((__may_alias__)) short_a; 3717 3718int 3719main (void) 3720@{ 3721 int a = 0x12345678; 3722 short_a *b = (short_a *) &a; 3723 3724 b[1] = 0; 3725 3726 if (a == 0x12345678) 3727 abort(); 3728 3729 exit(0); 3730@} 3731@end smallexample 3732 3733If you replaced @code{short_a} with @code{short} in the variable 3734declaration, the above program would abort when compiled with 3735@option{-fstrict-aliasing}, which is on by default at @option{-O2} or 3736above in recent GCC versions. 3737 3738@item visibility 3739In C++, attribute visibility (@pxref{Function Attributes}) can also be 3740applied to class, struct, union and enum types. Unlike other type 3741attributes, the attribute must appear between the initial keyword and 3742the name of the type; it cannot appear after the body of the type. 3743 3744Note that the type visibility is applied to vague linkage entities 3745associated with the class (vtable, typeinfo node, etc.). In 3746particular, if a class is thrown as an exception in one shared object 3747and caught in another, the class must have default visibility. 3748Otherwise the two shared objects will be unable to use the same 3749typeinfo node and exception handling will break. 3750 3751@subsection ARM Type Attributes 3752 3753On those ARM targets that support @code{dllimport} (such as Symbian 3754OS), you can use the @code{notshared} attribute to indicate that the 3755virtual table and other similar data for a class should not be 3756exported from a DLL@. For example: 3757 3758@smallexample 3759class __declspec(notshared) C @{ 3760public: 3761 __declspec(dllimport) C(); 3762 virtual void f(); 3763@} 3764 3765__declspec(dllexport) 3766C::C() @{@} 3767@end smallexample 3768 3769In this code, @code{C::C} is exported from the current DLL, but the 3770virtual table for @code{C} is not exported. (You can use 3771@code{__attribute__} instead of @code{__declspec} if you prefer, but 3772most Symbian OS code uses @code{__declspec}.) 3773 3774@anchor{i386 Type Attributes} 3775@subsection i386 Type Attributes 3776 3777Two attributes are currently defined for i386 configurations: 3778@code{ms_struct} and @code{gcc_struct} 3779 3780@item ms_struct 3781@itemx gcc_struct 3782@cindex @code{ms_struct} 3783@cindex @code{gcc_struct} 3784 3785If @code{packed} is used on a structure, or if bit-fields are used 3786it may be that the Microsoft ABI packs them differently 3787than GCC would normally pack them. Particularly when moving packed 3788data between functions compiled with GCC and the native Microsoft compiler 3789(either via function call or as data in a file), it may be necessary to access 3790either format. 3791 3792Currently @option{-m[no-]ms-bitfields} is provided for the Microsoft Windows X86 3793compilers to match the native Microsoft compiler. 3794@end table 3795 3796To specify multiple attributes, separate them by commas within the 3797double parentheses: for example, @samp{__attribute__ ((aligned (16), 3798packed))}. 3799 3800@anchor{PowerPC Type Attributes} 3801@subsection PowerPC Type Attributes 3802 3803Three attributes currently are defined for PowerPC configurations: 3804@code{altivec}, @code{ms_struct} and @code{gcc_struct}. 3805 3806For full documentation of the struct attributes please see the 3807documentation in the @xref{i386 Type Attributes}, section. 3808 3809The @code{altivec} attribute allows one to declare AltiVec vector data 3810types supported by the AltiVec Programming Interface Manual. The 3811attribute requires an argument to specify one of three vector types: 3812@code{vector__}, @code{pixel__} (always followed by unsigned short), 3813and @code{bool__} (always followed by unsigned). 3814 3815@smallexample 3816__attribute__((altivec(vector__))) 3817__attribute__((altivec(pixel__))) unsigned short 3818__attribute__((altivec(bool__))) unsigned 3819@end smallexample 3820 3821These attributes mainly are intended to support the @code{__vector}, 3822@code{__pixel}, and @code{__bool} AltiVec keywords. 3823 3824@node Inline 3825@section An Inline Function is As Fast As a Macro 3826@cindex inline functions 3827@cindex integrating function code 3828@cindex open coding 3829@cindex macros, inline alternative 3830 3831By declaring a function inline, you can direct GCC to make 3832calls to that function faster. One way GCC can achieve this is to 3833integrate that function's code into the code for its callers. This 3834makes execution faster by eliminating the function-call overhead; in 3835addition, if any of the actual argument values are constant, their 3836known values may permit simplifications at compile time so that not 3837all of the inline function's code needs to be included. The effect on 3838code size is less predictable; object code may be larger or smaller 3839with function inlining, depending on the particular case. You can 3840also direct GCC to try to integrate all ``simple enough'' functions 3841into their callers with the option @option{-finline-functions}. 3842 3843GCC implements three different semantics of declaring a function 3844inline. One is available with @option{-std=gnu89}, another when 3845@option{-std=c99} or @option{-std=gnu99}, and the third is used when 3846compiling C++. 3847 3848To declare a function inline, use the @code{inline} keyword in its 3849declaration, like this: 3850 3851@smallexample 3852static inline int 3853inc (int *a) 3854@{ 3855 (*a)++; 3856@} 3857@end smallexample 3858 3859If you are writing a header file to be included in ISO C89 programs, write 3860@code{__inline__} instead of @code{inline}. @xref{Alternate Keywords}. 3861 3862The three types of inlining behave similarly in two important cases: 3863when the @code{inline} keyword is used on a @code{static} function, 3864like the example above, and when a function is first declared without 3865using the @code{inline} keyword and then is defined with 3866@code{inline}, like this: 3867 3868@smallexample 3869extern int inc (int *a); 3870inline int 3871inc (int *a) 3872@{ 3873 (*a)++; 3874@} 3875@end smallexample 3876 3877In both of these common cases, the program behaves the same as if you 3878had not used the @code{inline} keyword, except for its speed. 3879 3880@cindex inline functions, omission of 3881@opindex fkeep-inline-functions 3882When a function is both inline and @code{static}, if all calls to the 3883function are integrated into the caller, and the function's address is 3884never used, then the function's own assembler code is never referenced. 3885In this case, GCC does not actually output assembler code for the 3886function, unless you specify the option @option{-fkeep-inline-functions}. 3887Some calls cannot be integrated for various reasons (in particular, 3888calls that precede the function's definition cannot be integrated, and 3889neither can recursive calls within the definition). If there is a 3890nonintegrated call, then the function is compiled to assembler code as 3891usual. The function must also be compiled as usual if the program 3892refers to its address, because that can't be inlined. 3893 3894@cindex automatic @code{inline} for C++ member fns 3895@cindex @code{inline} automatic for C++ member fns 3896@cindex member fns, automatically @code{inline} 3897@cindex C++ member fns, automatically @code{inline} 3898@opindex fno-default-inline 3899As required by ISO C++, GCC considers member functions defined within 3900the body of a class to be marked inline even if they are 3901not explicitly declared with the @code{inline} keyword. You can 3902override this with @option{-fno-default-inline}; @pxref{C++ Dialect 3903Options,,Options Controlling C++ Dialect}. 3904 3905GCC does not inline any functions when not optimizing unless you specify 3906the @samp{always_inline} attribute for the function, like this: 3907 3908@smallexample 3909/* @r{Prototype.} */ 3910inline void foo (const char) __attribute__((always_inline)); 3911@end smallexample 3912 3913The remainder of this section is specific to GNU C89 inlining. 3914 3915@cindex non-static inline function 3916When an inline function is not @code{static}, then the compiler must assume 3917that there may be calls from other source files; since a global symbol can 3918be defined only once in any program, the function must not be defined in 3919the other source files, so the calls therein cannot be integrated. 3920Therefore, a non-@code{static} inline function is always compiled on its 3921own in the usual fashion. 3922 3923If you specify both @code{inline} and @code{extern} in the function 3924definition, then the definition is used only for inlining. In no case 3925is the function compiled on its own, not even if you refer to its 3926address explicitly. Such an address becomes an external reference, as 3927if you had only declared the function, and had not defined it. 3928 3929This combination of @code{inline} and @code{extern} has almost the 3930effect of a macro. The way to use it is to put a function definition in 3931a header file with these keywords, and put another copy of the 3932definition (lacking @code{inline} and @code{extern}) in a library file. 3933The definition in the header file will cause most calls to the function 3934to be inlined. If any uses of the function remain, they will refer to 3935the single copy in the library. 3936 3937@node Extended Asm 3938@section Assembler Instructions with C Expression Operands 3939@cindex extended @code{asm} 3940@cindex @code{asm} expressions 3941@cindex assembler instructions 3942@cindex registers 3943 3944In an assembler instruction using @code{asm}, you can specify the 3945operands of the instruction using C expressions. This means you need not 3946guess which registers or memory locations will contain the data you want 3947to use. 3948 3949You must specify an assembler instruction template much like what 3950appears in a machine description, plus an operand constraint string for 3951each operand. 3952 3953For example, here is how to use the 68881's @code{fsinx} instruction: 3954 3955@smallexample 3956asm ("fsinx %1,%0" : "=f" (result) : "f" (angle)); 3957@end smallexample 3958 3959@noindent 3960Here @code{angle} is the C expression for the input operand while 3961@code{result} is that of the output operand. Each has @samp{"f"} as its 3962operand constraint, saying that a floating point register is required. 3963The @samp{=} in @samp{=f} indicates that the operand is an output; all 3964output operands' constraints must use @samp{=}. The constraints use the 3965same language used in the machine description (@pxref{Constraints}). 3966 3967Each operand is described by an operand-constraint string followed by 3968the C expression in parentheses. A colon separates the assembler 3969template from the first output operand and another separates the last 3970output operand from the first input, if any. Commas separate the 3971operands within each group. The total number of operands is currently 3972limited to 30; this limitation may be lifted in some future version of 3973GCC@. 3974 3975If there are no output operands but there are input operands, you must 3976place two consecutive colons surrounding the place where the output 3977operands would go. 3978 3979As of GCC version 3.1, it is also possible to specify input and output 3980operands using symbolic names which can be referenced within the 3981assembler code. These names are specified inside square brackets 3982preceding the constraint string, and can be referenced inside the 3983assembler code using @code{%[@var{name}]} instead of a percentage sign 3984followed by the operand number. Using named operands the above example 3985could look like: 3986 3987@smallexample 3988asm ("fsinx %[angle],%[output]" 3989 : [output] "=f" (result) 3990 : [angle] "f" (angle)); 3991@end smallexample 3992 3993@noindent 3994Note that the symbolic operand names have no relation whatsoever to 3995other C identifiers. You may use any name you like, even those of 3996existing C symbols, but you must ensure that no two operands within the same 3997assembler construct use the same symbolic name. 3998 3999Output operand expressions must be lvalues; the compiler can check this. 4000The input operands need not be lvalues. The compiler cannot check 4001whether the operands have data types that are reasonable for the 4002instruction being executed. It does not parse the assembler instruction 4003template and does not know what it means or even whether it is valid 4004assembler input. The extended @code{asm} feature is most often used for 4005machine instructions the compiler itself does not know exist. If 4006the output expression cannot be directly addressed (for example, it is a 4007bit-field), your constraint must allow a register. In that case, GCC 4008will use the register as the output of the @code{asm}, and then store 4009that register into the output. 4010 4011The ordinary output operands must be write-only; GCC will assume that 4012the values in these operands before the instruction are dead and need 4013not be generated. Extended asm supports input-output or read-write 4014operands. Use the constraint character @samp{+} to indicate such an 4015operand and list it with the output operands. You should only use 4016read-write operands when the constraints for the operand (or the 4017operand in which only some of the bits are to be changed) allow a 4018register. 4019 4020You may, as an alternative, logically split its function into two 4021separate operands, one input operand and one write-only output 4022operand. The connection between them is expressed by constraints 4023which say they need to be in the same location when the instruction 4024executes. You can use the same C expression for both operands, or 4025different expressions. For example, here we write the (fictitious) 4026@samp{combine} instruction with @code{bar} as its read-only source 4027operand and @code{foo} as its read-write destination: 4028 4029@smallexample 4030asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar)); 4031@end smallexample 4032 4033@noindent 4034The constraint @samp{"0"} for operand 1 says that it must occupy the 4035same location as operand 0. A number in constraint is allowed only in 4036an input operand and it must refer to an output operand. 4037 4038Only a number in the constraint can guarantee that one operand will be in 4039the same place as another. The mere fact that @code{foo} is the value 4040of both operands is not enough to guarantee that they will be in the 4041same place in the generated assembler code. The following would not 4042work reliably: 4043 4044@smallexample 4045asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar)); 4046@end smallexample 4047 4048Various optimizations or reloading could cause operands 0 and 1 to be in 4049different registers; GCC knows no reason not to do so. For example, the 4050compiler might find a copy of the value of @code{foo} in one register and 4051use it for operand 1, but generate the output operand 0 in a different 4052register (copying it afterward to @code{foo}'s own address). Of course, 4053since the register for operand 1 is not even mentioned in the assembler 4054code, the result will not work, but GCC can't tell that. 4055 4056As of GCC version 3.1, one may write @code{[@var{name}]} instead of 4057the operand number for a matching constraint. For example: 4058 4059@smallexample 4060asm ("cmoveq %1,%2,%[result]" 4061 : [result] "=r"(result) 4062 : "r" (test), "r"(new), "[result]"(old)); 4063@end smallexample 4064 4065Sometimes you need to make an @code{asm} operand be a specific register, 4066but there's no matching constraint letter for that register @emph{by 4067itself}. To force the operand into that register, use a local variable 4068for the operand and specify the register in the variable declaration. 4069@xref{Explicit Reg Vars}. Then for the @code{asm} operand, use any 4070register constraint letter that matches the register: 4071 4072@smallexample 4073register int *p1 asm ("r0") = @dots{}; 4074register int *p2 asm ("r1") = @dots{}; 4075register int *result asm ("r0"); 4076asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2)); 4077@end smallexample 4078 4079@anchor{Example of asm with clobbered asm reg} 4080In the above example, beware that a register that is call-clobbered by 4081the target ABI will be overwritten by any function call in the 4082assignment, including library calls for arithmetic operators. 4083Assuming it is a call-clobbered register, this may happen to @code{r0} 4084above by the assignment to @code{p2}. If you have to use such a 4085register, use temporary variables for expressions between the register 4086assignment and use: 4087 4088@smallexample 4089int t1 = @dots{}; 4090register int *p1 asm ("r0") = @dots{}; 4091register int *p2 asm ("r1") = t1; 4092register int *result asm ("r0"); 4093asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2)); 4094@end smallexample 4095 4096Some instructions clobber specific hard registers. To describe this, 4097write a third colon after the input operands, followed by the names of 4098the clobbered hard registers (given as strings). Here is a realistic 4099example for the VAX: 4100 4101@smallexample 4102asm volatile ("movc3 %0,%1,%2" 4103 : /* @r{no outputs} */ 4104 : "g" (from), "g" (to), "g" (count) 4105 : "r0", "r1", "r2", "r3", "r4", "r5"); 4106@end smallexample 4107 4108You may not write a clobber description in a way that overlaps with an 4109input or output operand. For example, you may not have an operand 4110describing a register class with one member if you mention that register 4111in the clobber list. Variables declared to live in specific registers 4112(@pxref{Explicit Reg Vars}), and used as asm input or output operands must 4113have no part mentioned in the clobber description. 4114There is no way for you to specify that an input 4115operand is modified without also specifying it as an output 4116operand. Note that if all the output operands you specify are for this 4117purpose (and hence unused), you will then also need to specify 4118@code{volatile} for the @code{asm} construct, as described below, to 4119prevent GCC from deleting the @code{asm} statement as unused. 4120 4121If you refer to a particular hardware register from the assembler code, 4122you will probably have to list the register after the third colon to 4123tell the compiler the register's value is modified. In some assemblers, 4124the register names begin with @samp{%}; to produce one @samp{%} in the 4125assembler code, you must write @samp{%%} in the input. 4126 4127If your assembler instruction can alter the condition code register, add 4128@samp{cc} to the list of clobbered registers. GCC on some machines 4129represents the condition codes as a specific hardware register; 4130@samp{cc} serves to name this register. On other machines, the 4131condition code is handled differently, and specifying @samp{cc} has no 4132effect. But it is valid no matter what the machine. 4133 4134If your assembler instructions access memory in an unpredictable 4135fashion, add @samp{memory} to the list of clobbered registers. This 4136will cause GCC to not keep memory values cached in registers across the 4137assembler instruction and not optimize stores or loads to that memory. 4138You will also want to add the @code{volatile} keyword if the memory 4139affected is not listed in the inputs or outputs of the @code{asm}, as 4140the @samp{memory} clobber does not count as a side-effect of the 4141@code{asm}. If you know how large the accessed memory is, you can add 4142it as input or output but if this is not known, you should add 4143@samp{memory}. As an example, if you access ten bytes of a string, you 4144can use a memory input like: 4145 4146@smallexample 4147@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}. 4148@end smallexample 4149 4150Note that in the following example the memory input is necessary, 4151otherwise GCC might optimize the store to @code{x} away: 4152@smallexample 4153int foo () 4154@{ 4155 int x = 42; 4156 int *y = &x; 4157 int result; 4158 asm ("magic stuff accessing an 'int' pointed to by '%1'" 4159 "=&d" (r) : "a" (y), "m" (*y)); 4160 return result; 4161@} 4162@end smallexample 4163 4164You can put multiple assembler instructions together in a single 4165@code{asm} template, separated by the characters normally used in assembly 4166code for the system. A combination that works in most places is a newline 4167to break the line, plus a tab character to move to the instruction field 4168(written as @samp{\n\t}). Sometimes semicolons can be used, if the 4169assembler allows semicolons as a line-breaking character. Note that some 4170assembler dialects use semicolons to start a comment. 4171The input operands are guaranteed not to use any of the clobbered 4172registers, and neither will the output operands' addresses, so you can 4173read and write the clobbered registers as many times as you like. Here 4174is an example of multiple instructions in a template; it assumes the 4175subroutine @code{_foo} accepts arguments in registers 9 and 10: 4176 4177@smallexample 4178asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo" 4179 : /* no outputs */ 4180 : "g" (from), "g" (to) 4181 : "r9", "r10"); 4182@end smallexample 4183 4184Unless an output operand has the @samp{&} constraint modifier, GCC 4185may allocate it in the same register as an unrelated input operand, on 4186the assumption the inputs are consumed before the outputs are produced. 4187This assumption may be false if the assembler code actually consists of 4188more than one instruction. In such a case, use @samp{&} for each output 4189operand that may not overlap an input. @xref{Modifiers}. 4190 4191If you want to test the condition code produced by an assembler 4192instruction, you must include a branch and a label in the @code{asm} 4193construct, as follows: 4194 4195@smallexample 4196asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:" 4197 : "g" (result) 4198 : "g" (input)); 4199@end smallexample 4200 4201@noindent 4202This assumes your assembler supports local labels, as the GNU assembler 4203and most Unix assemblers do. 4204 4205Speaking of labels, jumps from one @code{asm} to another are not 4206supported. The compiler's optimizers do not know about these jumps, and 4207therefore they cannot take account of them when deciding how to 4208optimize. 4209 4210@cindex macros containing @code{asm} 4211Usually the most convenient way to use these @code{asm} instructions is to 4212encapsulate them in macros that look like functions. For example, 4213 4214@smallexample 4215#define sin(x) \ 4216(@{ double __value, __arg = (x); \ 4217 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \ 4218 __value; @}) 4219@end smallexample 4220 4221@noindent 4222Here the variable @code{__arg} is used to make sure that the instruction 4223operates on a proper @code{double} value, and to accept only those 4224arguments @code{x} which can convert automatically to a @code{double}. 4225 4226Another way to make sure the instruction operates on the correct data 4227type is to use a cast in the @code{asm}. This is different from using a 4228variable @code{__arg} in that it converts more different types. For 4229example, if the desired type were @code{int}, casting the argument to 4230@code{int} would accept a pointer with no complaint, while assigning the 4231argument to an @code{int} variable named @code{__arg} would warn about 4232using a pointer unless the caller explicitly casts it. 4233 4234If an @code{asm} has output operands, GCC assumes for optimization 4235purposes the instruction has no side effects except to change the output 4236operands. This does not mean instructions with a side effect cannot be 4237used, but you must be careful, because the compiler may eliminate them 4238if the output operands aren't used, or move them out of loops, or 4239replace two with one if they constitute a common subexpression. Also, 4240if your instruction does have a side effect on a variable that otherwise 4241appears not to change, the old value of the variable may be reused later 4242if it happens to be found in a register. 4243 4244You can prevent an @code{asm} instruction from being deleted 4245by writing the keyword @code{volatile} after 4246the @code{asm}. For example: 4247 4248@smallexample 4249#define get_and_set_priority(new) \ 4250(@{ int __old; \ 4251 asm volatile ("get_and_set_priority %0, %1" \ 4252 : "=g" (__old) : "g" (new)); \ 4253 __old; @}) 4254@end smallexample 4255 4256@noindent 4257The @code{volatile} keyword indicates that the instruction has 4258important side-effects. GCC will not delete a volatile @code{asm} if 4259it is reachable. (The instruction can still be deleted if GCC can 4260prove that control-flow will never reach the location of the 4261instruction.) Note that even a volatile @code{asm} instruction 4262can be moved relative to other code, including across jump 4263instructions. For example, on many targets there is a system 4264register which can be set to control the rounding mode of 4265floating point operations. You might try 4266setting it with a volatile @code{asm}, like this PowerPC example: 4267 4268@smallexample 4269 asm volatile("mtfsf 255,%0" : : "f" (fpenv)); 4270 sum = x + y; 4271@end smallexample 4272 4273@noindent 4274This will not work reliably, as the compiler may move the addition back 4275before the volatile @code{asm}. To make it work you need to add an 4276artificial dependency to the @code{asm} referencing a variable in the code 4277you don't want moved, for example: 4278 4279@smallexample 4280 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv)); 4281 sum = x + y; 4282@end smallexample 4283 4284Similarly, you can't expect a 4285sequence of volatile @code{asm} instructions to remain perfectly 4286consecutive. If you want consecutive output, use a single @code{asm}. 4287Also, GCC will perform some optimizations across a volatile @code{asm} 4288instruction; GCC does not ``forget everything'' when it encounters 4289a volatile @code{asm} instruction the way some other compilers do. 4290 4291An @code{asm} instruction without any output operands will be treated 4292identically to a volatile @code{asm} instruction. 4293 4294It is a natural idea to look for a way to give access to the condition 4295code left by the assembler instruction. However, when we attempted to 4296implement this, we found no way to make it work reliably. The problem 4297is that output operands might need reloading, which would result in 4298additional following ``store'' instructions. On most machines, these 4299instructions would alter the condition code before there was time to 4300test it. This problem doesn't arise for ordinary ``test'' and 4301``compare'' instructions because they don't have any output operands. 4302 4303For reasons similar to those described above, it is not possible to give 4304an assembler instruction access to the condition code left by previous 4305instructions. 4306 4307If you are writing a header file that should be includable in ISO C 4308programs, write @code{__asm__} instead of @code{asm}. @xref{Alternate 4309Keywords}. 4310 4311@subsection Size of an @code{asm} 4312 4313Some targets require that GCC track the size of each instruction used in 4314order to generate correct code. Because the final length of an 4315@code{asm} is only known by the assembler, GCC must make an estimate as 4316to how big it will be. The estimate is formed by counting the number of 4317statements in the pattern of the @code{asm} and multiplying that by the 4318length of the longest instruction on that processor. Statements in the 4319@code{asm} are identified by newline characters and whatever statement 4320separator characters are supported by the assembler; on most processors 4321this is the `@code{;}' character. 4322 4323Normally, GCC's estimate is perfectly adequate to ensure that correct 4324code is generated, but it is possible to confuse the compiler if you use 4325pseudo instructions or assembler macros that expand into multiple real 4326instructions or if you use assembler directives that expand to more 4327space in the object file than would be needed for a single instruction. 4328If this happens then the assembler will produce a diagnostic saying that 4329a label is unreachable. 4330 4331@subsection i386 floating point asm operands 4332 4333There are several rules on the usage of stack-like regs in 4334asm_operands insns. These rules apply only to the operands that are 4335stack-like regs: 4336 4337@enumerate 4338@item 4339Given a set of input regs that die in an asm_operands, it is 4340necessary to know which are implicitly popped by the asm, and 4341which must be explicitly popped by gcc. 4342 4343An input reg that is implicitly popped by the asm must be 4344explicitly clobbered, unless it is constrained to match an 4345output operand. 4346 4347@item 4348For any input reg that is implicitly popped by an asm, it is 4349necessary to know how to adjust the stack to compensate for the pop. 4350If any non-popped input is closer to the top of the reg-stack than 4351the implicitly popped reg, it would not be possible to know what the 4352stack looked like---it's not clear how the rest of the stack ``slides 4353up''. 4354 4355All implicitly popped input regs must be closer to the top of 4356the reg-stack than any input that is not implicitly popped. 4357 4358It is possible that if an input dies in an insn, reload might 4359use the input reg for an output reload. Consider this example: 4360 4361@smallexample 4362asm ("foo" : "=t" (a) : "f" (b)); 4363@end smallexample 4364 4365This asm says that input B is not popped by the asm, and that 4366the asm pushes a result onto the reg-stack, i.e., the stack is one 4367deeper after the asm than it was before. But, it is possible that 4368reload will think that it can use the same reg for both the input and 4369the output, if input B dies in this insn. 4370 4371If any input operand uses the @code{f} constraint, all output reg 4372constraints must use the @code{&} earlyclobber. 4373 4374The asm above would be written as 4375 4376@smallexample 4377asm ("foo" : "=&t" (a) : "f" (b)); 4378@end smallexample 4379 4380@item 4381Some operands need to be in particular places on the stack. All 4382output operands fall in this category---there is no other way to 4383know which regs the outputs appear in unless the user indicates 4384this in the constraints. 4385 4386Output operands must specifically indicate which reg an output 4387appears in after an asm. @code{=f} is not allowed: the operand 4388constraints must select a class with a single reg. 4389 4390@item 4391Output operands may not be ``inserted'' between existing stack regs. 4392Since no 387 opcode uses a read/write operand, all output operands 4393are dead before the asm_operands, and are pushed by the asm_operands. 4394It makes no sense to push anywhere but the top of the reg-stack. 4395 4396Output operands must start at the top of the reg-stack: output 4397operands may not ``skip'' a reg. 4398 4399@item 4400Some asm statements may need extra stack space for internal 4401calculations. This can be guaranteed by clobbering stack registers 4402unrelated to the inputs and outputs. 4403 4404@end enumerate 4405 4406Here are a couple of reasonable asms to want to write. This asm 4407takes one input, which is internally popped, and produces two outputs. 4408 4409@smallexample 4410asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp)); 4411@end smallexample 4412 4413This asm takes two inputs, which are popped by the @code{fyl2xp1} opcode, 4414and replaces them with one output. The user must code the @code{st(1)} 4415clobber for reg-stack.c to know that @code{fyl2xp1} pops both inputs. 4416 4417@smallexample 4418asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)"); 4419@end smallexample 4420 4421@include md.texi 4422 4423@node Asm Labels 4424@section Controlling Names Used in Assembler Code 4425@cindex assembler names for identifiers 4426@cindex names used in assembler code 4427@cindex identifiers, names in assembler code 4428 4429You can specify the name to be used in the assembler code for a C 4430function or variable by writing the @code{asm} (or @code{__asm__}) 4431keyword after the declarator as follows: 4432 4433@smallexample 4434int foo asm ("myfoo") = 2; 4435@end smallexample 4436 4437@noindent 4438This specifies that the name to be used for the variable @code{foo} in 4439the assembler code should be @samp{myfoo} rather than the usual 4440@samp{_foo}. 4441 4442On systems where an underscore is normally prepended to the name of a C 4443function or variable, this feature allows you to define names for the 4444linker that do not start with an underscore. 4445 4446It does not make sense to use this feature with a non-static local 4447variable since such variables do not have assembler names. If you are 4448trying to put the variable in a particular register, see @ref{Explicit 4449Reg Vars}. GCC presently accepts such code with a warning, but will 4450probably be changed to issue an error, rather than a warning, in the 4451future. 4452 4453You cannot use @code{asm} in this way in a function @emph{definition}; but 4454you can get the same effect by writing a declaration for the function 4455before its definition and putting @code{asm} there, like this: 4456 4457@smallexample 4458extern func () asm ("FUNC"); 4459 4460func (x, y) 4461 int x, y; 4462/* @r{@dots{}} */ 4463@end smallexample 4464 4465It is up to you to make sure that the assembler names you choose do not 4466conflict with any other assembler symbols. Also, you must not use a 4467register name; that would produce completely invalid assembler code. GCC 4468does not as yet have the ability to store static variables in registers. 4469Perhaps that will be added. 4470 4471@node Explicit Reg Vars 4472@section Variables in Specified Registers 4473@cindex explicit register variables 4474@cindex variables in specified registers 4475@cindex specified registers 4476@cindex registers, global allocation 4477 4478GNU C allows you to put a few global variables into specified hardware 4479registers. You can also specify the register in which an ordinary 4480register variable should be allocated. 4481 4482@itemize @bullet 4483@item 4484Global register variables reserve registers throughout the program. 4485This may be useful in programs such as programming language 4486interpreters which have a couple of global variables that are accessed 4487very often. 4488 4489@item 4490Local register variables in specific registers do not reserve the 4491registers, except at the point where they are used as input or output 4492operands in an @code{asm} statement and the @code{asm} statement itself is 4493not deleted. The compiler's data flow analysis is capable of determining 4494where the specified registers contain live values, and where they are 4495available for other uses. Stores into local register variables may be deleted 4496when they appear to be dead according to dataflow analysis. References 4497to local register variables may be deleted or moved or simplified. 4498 4499These local variables are sometimes convenient for use with the extended 4500@code{asm} feature (@pxref{Extended Asm}), if you want to write one 4501output of the assembler instruction directly into a particular register. 4502(This will work provided the register you specify fits the constraints 4503specified for that operand in the @code{asm}.) 4504@end itemize 4505 4506@menu 4507* Global Reg Vars:: 4508* Local Reg Vars:: 4509@end menu 4510 4511@node Global Reg Vars 4512@subsection Defining Global Register Variables 4513@cindex global register variables 4514@cindex registers, global variables in 4515 4516You can define a global register variable in GNU C like this: 4517 4518@smallexample 4519register int *foo asm ("a5"); 4520@end smallexample 4521 4522@noindent 4523Here @code{a5} is the name of the register which should be used. Choose a 4524register which is normally saved and restored by function calls on your 4525machine, so that library routines will not clobber it. 4526 4527Naturally the register name is cpu-dependent, so you would need to 4528conditionalize your program according to cpu type. The register 4529@code{a5} would be a good choice on a 68000 for a variable of pointer 4530type. On machines with register windows, be sure to choose a ``global'' 4531register that is not affected magically by the function call mechanism. 4532 4533In addition, operating systems on one type of cpu may differ in how they 4534name the registers; then you would need additional conditionals. For 4535example, some 68000 operating systems call this register @code{%a5}. 4536 4537Eventually there may be a way of asking the compiler to choose a register 4538automatically, but first we need to figure out how it should choose and 4539how to enable you to guide the choice. No solution is evident. 4540 4541Defining a global register variable in a certain register reserves that 4542register entirely for this use, at least within the current compilation. 4543The register will not be allocated for any other purpose in the functions 4544in the current compilation. The register will not be saved and restored by 4545these functions. Stores into this register are never deleted even if they 4546would appear to be dead, but references may be deleted or moved or 4547simplified. 4548 4549It is not safe to access the global register variables from signal 4550handlers, or from more than one thread of control, because the system 4551library routines may temporarily use the register for other things (unless 4552you recompile them specially for the task at hand). 4553 4554@cindex @code{qsort}, and global register variables 4555It is not safe for one function that uses a global register variable to 4556call another such function @code{foo} by way of a third function 4557@code{lose} that was compiled without knowledge of this variable (i.e.@: in a 4558different source file in which the variable wasn't declared). This is 4559because @code{lose} might save the register and put some other value there. 4560For example, you can't expect a global register variable to be available in 4561the comparison-function that you pass to @code{qsort}, since @code{qsort} 4562might have put something else in that register. (If you are prepared to 4563recompile @code{qsort} with the same global register variable, you can 4564solve this problem.) 4565 4566If you want to recompile @code{qsort} or other source files which do not 4567actually use your global register variable, so that they will not use that 4568register for any other purpose, then it suffices to specify the compiler 4569option @option{-ffixed-@var{reg}}. You need not actually add a global 4570register declaration to their source code. 4571 4572A function which can alter the value of a global register variable cannot 4573safely be called from a function compiled without this variable, because it 4574could clobber the value the caller expects to find there on return. 4575Therefore, the function which is the entry point into the part of the 4576program that uses the global register variable must explicitly save and 4577restore the value which belongs to its caller. 4578 4579@cindex register variable after @code{longjmp} 4580@cindex global register after @code{longjmp} 4581@cindex value after @code{longjmp} 4582@findex longjmp 4583@findex setjmp 4584On most machines, @code{longjmp} will restore to each global register 4585variable the value it had at the time of the @code{setjmp}. On some 4586machines, however, @code{longjmp} will not change the value of global 4587register variables. To be portable, the function that called @code{setjmp} 4588should make other arrangements to save the values of the global register 4589variables, and to restore them in a @code{longjmp}. This way, the same 4590thing will happen regardless of what @code{longjmp} does. 4591 4592All global register variable declarations must precede all function 4593definitions. If such a declaration could appear after function 4594definitions, the declaration would be too late to prevent the register from 4595being used for other purposes in the preceding functions. 4596 4597Global register variables may not have initial values, because an 4598executable file has no means to supply initial contents for a register. 4599 4600On the SPARC, there are reports that g3 @dots{} g7 are suitable 4601registers, but certain library functions, such as @code{getwd}, as well 4602as the subroutines for division and remainder, modify g3 and g4. g1 and 4603g2 are local temporaries. 4604 4605On the 68000, a2 @dots{} a5 should be suitable, as should d2 @dots{} d7. 4606Of course, it will not do to use more than a few of those. 4607 4608@node Local Reg Vars 4609@subsection Specifying Registers for Local Variables 4610@cindex local variables, specifying registers 4611@cindex specifying registers for local variables 4612@cindex registers for local variables 4613 4614You can define a local register variable with a specified register 4615like this: 4616 4617@smallexample 4618register int *foo asm ("a5"); 4619@end smallexample 4620 4621@noindent 4622Here @code{a5} is the name of the register which should be used. Note 4623that this is the same syntax used for defining global register 4624variables, but for a local variable it would appear within a function. 4625 4626Naturally the register name is cpu-dependent, but this is not a 4627problem, since specific registers are most often useful with explicit 4628assembler instructions (@pxref{Extended Asm}). Both of these things 4629generally require that you conditionalize your program according to 4630cpu type. 4631 4632In addition, operating systems on one type of cpu may differ in how they 4633name the registers; then you would need additional conditionals. For 4634example, some 68000 operating systems call this register @code{%a5}. 4635 4636Defining such a register variable does not reserve the register; it 4637remains available for other uses in places where flow control determines 4638the variable's value is not live. 4639 4640This option does not guarantee that GCC will generate code that has 4641this variable in the register you specify at all times. You may not 4642code an explicit reference to this register in the @emph{assembler 4643instruction template} part of an @code{asm} statement and assume it will 4644always refer to this variable. However, using the variable as an 4645@code{asm} @emph{operand} guarantees that the specified register is used 4646for the operand. 4647 4648Stores into local register variables may be deleted when they appear to be dead 4649according to dataflow analysis. References to local register variables may 4650be deleted or moved or simplified. 4651 4652As for global register variables, it's recommended that you choose a 4653register which is normally saved and restored by function calls on 4654your machine, so that library routines will not clobber it. A common 4655pitfall is to initialize multiple call-clobbered registers with 4656arbitrary expressions, where a function call or library call for an 4657arithmetic operator will overwrite a register value from a previous 4658assignment, for example @code{r0} below: 4659@smallexample 4660register int *p1 asm ("r0") = @dots{}; 4661register int *p2 asm ("r1") = @dots{}; 4662@end smallexample 4663In those cases, a solution is to use a temporary variable for 4664each arbitrary expression. @xref{Example of asm with clobbered asm reg}. 4665 4666@node Alternate Keywords 4667@section Alternate Keywords 4668@cindex alternate keywords 4669@cindex keywords, alternate 4670 4671@option{-ansi} and the various @option{-std} options disable certain 4672keywords. This causes trouble when you want to use GNU C extensions, or 4673a general-purpose header file that should be usable by all programs, 4674including ISO C programs. The keywords @code{asm}, @code{typeof} and 4675@code{inline} are not available in programs compiled with 4676@option{-ansi} or @option{-std} (although @code{inline} can be used in a 4677program compiled with @option{-std=c99}). The ISO C99 keyword 4678@code{restrict} is only available when @option{-std=gnu99} (which will 4679eventually be the default) or @option{-std=c99} (or the equivalent 4680@option{-std=iso9899:1999}) is used. 4681 4682The way to solve these problems is to put @samp{__} at the beginning and 4683end of each problematical keyword. For example, use @code{__asm__} 4684instead of @code{asm}, and @code{__inline__} instead of @code{inline}. 4685 4686Other C compilers won't accept these alternative keywords; if you want to 4687compile with another compiler, you can define the alternate keywords as 4688macros to replace them with the customary keywords. It looks like this: 4689 4690@smallexample 4691#ifndef __GNUC__ 4692#define __asm__ asm 4693#endif 4694@end smallexample 4695 4696@findex __extension__ 4697@opindex pedantic 4698@option{-pedantic} and other options cause warnings for many GNU C extensions. 4699You can 4700prevent such warnings within one expression by writing 4701@code{__extension__} before the expression. @code{__extension__} has no 4702effect aside from this. 4703 4704@node Incomplete Enums 4705@section Incomplete @code{enum} Types 4706 4707You can define an @code{enum} tag without specifying its possible values. 4708This results in an incomplete type, much like what you get if you write 4709@code{struct foo} without describing the elements. A later declaration 4710which does specify the possible values completes the type. 4711 4712You can't allocate variables or storage using the type while it is 4713incomplete. However, you can work with pointers to that type. 4714 4715This extension may not be very useful, but it makes the handling of 4716@code{enum} more consistent with the way @code{struct} and @code{union} 4717are handled. 4718 4719This extension is not supported by GNU C++. 4720 4721@node Function Names 4722@section Function Names as Strings 4723@cindex @code{__func__} identifier 4724@cindex @code{__FUNCTION__} identifier 4725@cindex @code{__PRETTY_FUNCTION__} identifier 4726 4727GCC provides three magic variables which hold the name of the current 4728function, as a string. The first of these is @code{__func__}, which 4729is part of the C99 standard: 4730 4731@display 4732The identifier @code{__func__} is implicitly declared by the translator 4733as if, immediately following the opening brace of each function 4734definition, the declaration 4735 4736@smallexample 4737static const char __func__[] = "function-name"; 4738@end smallexample 4739 4740appeared, where function-name is the name of the lexically-enclosing 4741function. This name is the unadorned name of the function. 4742@end display 4743 4744@code{__FUNCTION__} is another name for @code{__func__}. Older 4745versions of GCC recognize only this name. However, it is not 4746standardized. For maximum portability, we recommend you use 4747@code{__func__}, but provide a fallback definition with the 4748preprocessor: 4749 4750@smallexample 4751#if __STDC_VERSION__ < 199901L 4752# if __GNUC__ >= 2 4753# define __func__ __FUNCTION__ 4754# else 4755# define __func__ "<unknown>" 4756# endif 4757#endif 4758@end smallexample 4759 4760In C, @code{__PRETTY_FUNCTION__} is yet another name for 4761@code{__func__}. However, in C++, @code{__PRETTY_FUNCTION__} contains 4762the type signature of the function as well as its bare name. For 4763example, this program: 4764 4765@smallexample 4766extern "C" @{ 4767extern int printf (char *, ...); 4768@} 4769 4770class a @{ 4771 public: 4772 void sub (int i) 4773 @{ 4774 printf ("__FUNCTION__ = %s\n", __FUNCTION__); 4775 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__); 4776 @} 4777@}; 4778 4779int 4780main (void) 4781@{ 4782 a ax; 4783 ax.sub (0); 4784 return 0; 4785@} 4786@end smallexample 4787 4788@noindent 4789gives this output: 4790 4791@smallexample 4792__FUNCTION__ = sub 4793__PRETTY_FUNCTION__ = void a::sub(int) 4794@end smallexample 4795 4796These identifiers are not preprocessor macros. In GCC 3.3 and 4797earlier, in C only, @code{__FUNCTION__} and @code{__PRETTY_FUNCTION__} 4798were treated as string literals; they could be used to initialize 4799@code{char} arrays, and they could be concatenated with other string 4800literals. GCC 3.4 and later treat them as variables, like 4801@code{__func__}. In C++, @code{__FUNCTION__} and 4802@code{__PRETTY_FUNCTION__} have always been variables. 4803 4804@node Return Address 4805@section Getting the Return or Frame Address of a Function 4806 4807These functions may be used to get information about the callers of a 4808function. 4809 4810@deftypefn {Built-in Function} {void *} __builtin_return_address (unsigned int @var{level}) 4811This function returns the return address of the current function, or of 4812one of its callers. The @var{level} argument is number of frames to 4813scan up the call stack. A value of @code{0} yields the return address 4814of the current function, a value of @code{1} yields the return address 4815of the caller of the current function, and so forth. When inlining 4816the expected behavior is that the function will return the address of 4817the function that will be returned to. To work around this behavior use 4818the @code{noinline} function attribute. 4819 4820The @var{level} argument must be a constant integer. 4821 4822On some machines it may be impossible to determine the return address of 4823any function other than the current one; in such cases, or when the top 4824of the stack has been reached, this function will return @code{0} or a 4825random value. In addition, @code{__builtin_frame_address} may be used 4826to determine if the top of the stack has been reached. 4827 4828This function should only be used with a nonzero argument for debugging 4829purposes. 4830@end deftypefn 4831 4832@deftypefn {Built-in Function} {void *} __builtin_frame_address (unsigned int @var{level}) 4833This function is similar to @code{__builtin_return_address}, but it 4834returns the address of the function frame rather than the return address 4835of the function. Calling @code{__builtin_frame_address} with a value of 4836@code{0} yields the frame address of the current function, a value of 4837@code{1} yields the frame address of the caller of the current function, 4838and so forth. 4839 4840The frame is the area on the stack which holds local variables and saved 4841registers. The frame address is normally the address of the first word 4842pushed on to the stack by the function. However, the exact definition 4843depends upon the processor and the calling convention. If the processor 4844has a dedicated frame pointer register, and the function has a frame, 4845then @code{__builtin_frame_address} will return the value of the frame 4846pointer register. 4847 4848On some machines it may be impossible to determine the frame address of 4849any function other than the current one; in such cases, or when the top 4850of the stack has been reached, this function will return @code{0} if 4851the first frame pointer is properly initialized by the startup code. 4852 4853This function should only be used with a nonzero argument for debugging 4854purposes. 4855@end deftypefn 4856 4857@node Vector Extensions 4858@section Using vector instructions through built-in functions 4859 4860On some targets, the instruction set contains SIMD vector instructions that 4861operate on multiple values contained in one large register at the same time. 4862For example, on the i386 the MMX, 3Dnow! and SSE extensions can be used 4863this way. 4864 4865The first step in using these extensions is to provide the necessary data 4866types. This should be done using an appropriate @code{typedef}: 4867 4868@smallexample 4869typedef int v4si __attribute__ ((vector_size (16))); 4870@end smallexample 4871 4872The @code{int} type specifies the base type, while the attribute specifies 4873the vector size for the variable, measured in bytes. For example, the 4874declaration above causes the compiler to set the mode for the @code{v4si} 4875type to be 16 bytes wide and divided into @code{int} sized units. For 4876a 32-bit @code{int} this means a vector of 4 units of 4 bytes, and the 4877corresponding mode of @code{foo} will be @acronym{V4SI}. 4878 4879The @code{vector_size} attribute is only applicable to integral and 4880float scalars, although arrays, pointers, and function return values 4881are allowed in conjunction with this construct. 4882 4883All the basic integer types can be used as base types, both as signed 4884and as unsigned: @code{char}, @code{short}, @code{int}, @code{long}, 4885@code{long long}. In addition, @code{float} and @code{double} can be 4886used to build floating-point vector types. 4887 4888Specifying a combination that is not valid for the current architecture 4889will cause GCC to synthesize the instructions using a narrower mode. 4890For example, if you specify a variable of type @code{V4SI} and your 4891architecture does not allow for this specific SIMD type, GCC will 4892produce code that uses 4 @code{SIs}. 4893 4894The types defined in this manner can be used with a subset of normal C 4895operations. Currently, GCC will allow using the following operators 4896on these types: @code{+, -, *, /, unary minus, ^, |, &, ~}@. 4897 4898The operations behave like C++ @code{valarrays}. Addition is defined as 4899the addition of the corresponding elements of the operands. For 4900example, in the code below, each of the 4 elements in @var{a} will be 4901added to the corresponding 4 elements in @var{b} and the resulting 4902vector will be stored in @var{c}. 4903 4904@smallexample 4905typedef int v4si __attribute__ ((vector_size (16))); 4906 4907v4si a, b, c; 4908 4909c = a + b; 4910@end smallexample 4911 4912Subtraction, multiplication, division, and the logical operations 4913operate in a similar manner. Likewise, the result of using the unary 4914minus or complement operators on a vector type is a vector whose 4915elements are the negative or complemented values of the corresponding 4916elements in the operand. 4917 4918You can declare variables and use them in function calls and returns, as 4919well as in assignments and some casts. You can specify a vector type as 4920a return type for a function. Vector types can also be used as function 4921arguments. It is possible to cast from one vector type to another, 4922provided they are of the same size (in fact, you can also cast vectors 4923to and from other datatypes of the same size). 4924 4925You cannot operate between vectors of different lengths or different 4926signedness without a cast. 4927 4928A port that supports hardware vector operations, usually provides a set 4929of built-in functions that can be used to operate on vectors. For 4930example, a function to add two vectors and multiply the result by a 4931third could look like this: 4932 4933@smallexample 4934v4si f (v4si a, v4si b, v4si c) 4935@{ 4936 v4si tmp = __builtin_addv4si (a, b); 4937 return __builtin_mulv4si (tmp, c); 4938@} 4939 4940@end smallexample 4941 4942@node Offsetof 4943@section Offsetof 4944@findex __builtin_offsetof 4945 4946GCC implements for both C and C++ a syntactic extension to implement 4947the @code{offsetof} macro. 4948 4949@smallexample 4950primary: 4951 "__builtin_offsetof" "(" @code{typename} "," offsetof_member_designator ")" 4952 4953offsetof_member_designator: 4954 @code{identifier} 4955 | offsetof_member_designator "." @code{identifier} 4956 | offsetof_member_designator "[" @code{expr} "]" 4957@end smallexample 4958 4959This extension is sufficient such that 4960 4961@smallexample 4962#define offsetof(@var{type}, @var{member}) __builtin_offsetof (@var{type}, @var{member}) 4963@end smallexample 4964 4965is a suitable definition of the @code{offsetof} macro. In C++, @var{type} 4966may be dependent. In either case, @var{member} may consist of a single 4967identifier, or a sequence of member accesses and array references. 4968 4969@node Atomic Builtins 4970@section Built-in functions for atomic memory access 4971 4972The following builtins are intended to be compatible with those described 4973in the @cite{Intel Itanium Processor-specific Application Binary Interface}, 4974section 7.4. As such, they depart from the normal GCC practice of using 4975the ``__builtin_'' prefix, and further that they are overloaded such that 4976they work on multiple types. 4977 4978The definition given in the Intel documentation allows only for the use of 4979the types @code{int}, @code{long}, @code{long long} as well as their unsigned 4980counterparts. GCC will allow any integral scalar or pointer type that is 49811, 2, 4 or 8 bytes in length. 4982 4983Not all operations are supported by all target processors. If a particular 4984operation cannot be implemented on the target processor, a warning will be 4985generated and a call an external function will be generated. The external 4986function will carry the same name as the builtin, with an additional suffix 4987@samp{_@var{n}} where @var{n} is the size of the data type. 4988 4989@c ??? Should we have a mechanism to suppress this warning? This is almost 4990@c useful for implementing the operation under the control of an external 4991@c mutex. 4992 4993In most cases, these builtins are considered a @dfn{full barrier}. That is, 4994no memory operand will be moved across the operation, either forward or 4995backward. Further, instructions will be issued as necessary to prevent the 4996processor from speculating loads across the operation and from queuing stores 4997after the operation. 4998 4999All of the routines are are described in the Intel documentation to take 5000``an optional list of variables protected by the memory barrier''. It's 5001not clear what is meant by that; it could mean that @emph{only} the 5002following variables are protected, or it could mean that these variables 5003should in addition be protected. At present GCC ignores this list and 5004protects all variables which are globally accessible. If in the future 5005we make some use of this list, an empty list will continue to mean all 5006globally accessible variables. 5007 5008@table @code 5009@item @var{type} __sync_fetch_and_add (@var{type} *ptr, @var{type} value, ...) 5010@itemx @var{type} __sync_fetch_and_sub (@var{type} *ptr, @var{type} value, ...) 5011@itemx @var{type} __sync_fetch_and_or (@var{type} *ptr, @var{type} value, ...) 5012@itemx @var{type} __sync_fetch_and_and (@var{type} *ptr, @var{type} value, ...) 5013@itemx @var{type} __sync_fetch_and_xor (@var{type} *ptr, @var{type} value, ...) 5014@itemx @var{type} __sync_fetch_and_nand (@var{type} *ptr, @var{type} value, ...) 5015@findex __sync_fetch_and_add 5016@findex __sync_fetch_and_sub 5017@findex __sync_fetch_and_or 5018@findex __sync_fetch_and_and 5019@findex __sync_fetch_and_xor 5020@findex __sync_fetch_and_nand 5021These builtins perform the operation suggested by the name, and 5022returns the value that had previously been in memory. That is, 5023 5024@smallexample 5025@{ tmp = *ptr; *ptr @var{op}= value; return tmp; @} 5026@{ tmp = *ptr; *ptr = ~tmp & value; return tmp; @} // nand 5027@end smallexample 5028 5029@item @var{type} __sync_add_and_fetch (@var{type} *ptr, @var{type} value, ...) 5030@itemx @var{type} __sync_sub_and_fetch (@var{type} *ptr, @var{type} value, ...) 5031@itemx @var{type} __sync_or_and_fetch (@var{type} *ptr, @var{type} value, ...) 5032@itemx @var{type} __sync_and_and_fetch (@var{type} *ptr, @var{type} value, ...) 5033@itemx @var{type} __sync_xor_and_fetch (@var{type} *ptr, @var{type} value, ...) 5034@itemx @var{type} __sync_nand_and_fetch (@var{type} *ptr, @var{type} value, ...) 5035@findex __sync_add_and_fetch 5036@findex __sync_sub_and_fetch 5037@findex __sync_or_and_fetch 5038@findex __sync_and_and_fetch 5039@findex __sync_xor_and_fetch 5040@findex __sync_nand_and_fetch 5041These builtins perform the operation suggested by the name, and 5042return the new value. That is, 5043 5044@smallexample 5045@{ *ptr @var{op}= value; return *ptr; @} 5046@{ *ptr = ~*ptr & value; return *ptr; @} // nand 5047@end smallexample 5048 5049@item bool __sync_bool_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...) 5050@itemx @var{type} __sync_val_compare_and_swap (@var{type} *ptr, @var{type} oldval @var{type} newval, ...) 5051@findex __sync_bool_compare_and_swap 5052@findex __sync_val_compare_and_swap 5053These builtins perform an atomic compare and swap. That is, if the current 5054value of @code{*@var{ptr}} is @var{oldval}, then write @var{newval} into 5055@code{*@var{ptr}}. 5056 5057The ``bool'' version returns true if the comparison is successful and 5058@var{newval} was written. The ``val'' version returns the contents 5059of @code{*@var{ptr}} before the operation. 5060 5061@item __sync_synchronize (...) 5062@findex __sync_synchronize 5063This builtin issues a full memory barrier. 5064 5065@item @var{type} __sync_lock_test_and_set (@var{type} *ptr, @var{type} value, ...) 5066@findex __sync_lock_test_and_set 5067This builtin, as described by Intel, is not a traditional test-and-set 5068operation, but rather an atomic exchange operation. It writes @var{value} 5069into @code{*@var{ptr}}, and returns the previous contents of 5070@code{*@var{ptr}}. 5071 5072Many targets have only minimal support for such locks, and do not support 5073a full exchange operation. In this case, a target may support reduced 5074functionality here by which the @emph{only} valid value to store is the 5075immediate constant 1. The exact value actually stored in @code{*@var{ptr}} 5076is implementation defined. 5077 5078This builtin is not a full barrier, but rather an @dfn{acquire barrier}. 5079This means that references after the builtin cannot move to (or be 5080speculated to) before the builtin, but previous memory stores may not 5081be globally visible yet, and previous memory loads may not yet be 5082satisfied. 5083 5084@item void __sync_lock_release (@var{type} *ptr, ...) 5085@findex __sync_lock_release 5086This builtin releases the lock acquired by @code{__sync_lock_test_and_set}. 5087Normally this means writing the constant 0 to @code{*@var{ptr}}. 5088 5089This builtin is not a full barrier, but rather a @dfn{release barrier}. 5090This means that all previous memory stores are globally visible, and all 5091previous memory loads have been satisfied, but following memory reads 5092are not prevented from being speculated to before the barrier. 5093@end table 5094 5095@node Object Size Checking 5096@section Object Size Checking Builtins 5097@findex __builtin_object_size 5098@findex __builtin___memcpy_chk 5099@findex __builtin___mempcpy_chk 5100@findex __builtin___memmove_chk 5101@findex __builtin___memset_chk 5102@findex __builtin___strcpy_chk 5103@findex __builtin___stpcpy_chk 5104@findex __builtin___strncpy_chk 5105@findex __builtin___strcat_chk 5106@findex __builtin___strncat_chk 5107@findex __builtin___sprintf_chk 5108@findex __builtin___snprintf_chk 5109@findex __builtin___vsprintf_chk 5110@findex __builtin___vsnprintf_chk 5111@findex __builtin___printf_chk 5112@findex __builtin___vprintf_chk 5113@findex __builtin___fprintf_chk 5114@findex __builtin___vfprintf_chk 5115 5116GCC implements a limited buffer overflow protection mechanism 5117that can prevent some buffer overflow attacks. 5118 5119@deftypefn {Built-in Function} {size_t} __builtin_object_size (void * @var{ptr}, int @var{type}) 5120is a built-in construct that returns a constant number of bytes from 5121@var{ptr} to the end of the object @var{ptr} pointer points to 5122(if known at compile time). @code{__builtin_object_size} never evaluates 5123its arguments for side-effects. If there are any side-effects in them, it 5124returns @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0} 5125for @var{type} 2 or 3. If there are multiple objects @var{ptr} can 5126point to and all of them are known at compile time, the returned number 5127is the maximum of remaining byte counts in those objects if @var{type} & 2 is 51280 and minimum if nonzero. If it is not possible to determine which objects 5129@var{ptr} points to at compile time, @code{__builtin_object_size} should 5130return @code{(size_t) -1} for @var{type} 0 or 1 and @code{(size_t) 0} 5131for @var{type} 2 or 3. 5132 5133@var{type} is an integer constant from 0 to 3. If the least significant 5134bit is clear, objects are whole variables, if it is set, a closest 5135surrounding subobject is considered the object a pointer points to. 5136The second bit determines if maximum or minimum of remaining bytes 5137is computed. 5138 5139@smallexample 5140struct V @{ char buf1[10]; int b; char buf2[10]; @} var; 5141char *p = &var.buf1[1], *q = &var.b; 5142 5143/* Here the object p points to is var. */ 5144assert (__builtin_object_size (p, 0) == sizeof (var) - 1); 5145/* The subobject p points to is var.buf1. */ 5146assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1); 5147/* The object q points to is var. */ 5148assert (__builtin_object_size (q, 0) 5149 == (char *) (&var + 1) - (char *) &var.b); 5150/* The subobject q points to is var.b. */ 5151assert (__builtin_object_size (q, 1) == sizeof (var.b)); 5152@end smallexample 5153@end deftypefn 5154 5155There are built-in functions added for many common string operation 5156functions, e.g. for @code{memcpy} @code{__builtin___memcpy_chk} 5157built-in is provided. This built-in has an additional last argument, 5158which is the number of bytes remaining in object the @var{dest} 5159argument points to or @code{(size_t) -1} if the size is not known. 5160 5161The built-in functions are optimized into the normal string functions 5162like @code{memcpy} if the last argument is @code{(size_t) -1} or if 5163it is known at compile time that the destination object will not 5164be overflown. If the compiler can determine at compile time the 5165object will be always overflown, it issues a warning. 5166 5167The intended use can be e.g. 5168 5169@smallexample 5170#undef memcpy 5171#define bos0(dest) __builtin_object_size (dest, 0) 5172#define memcpy(dest, src, n) \ 5173 __builtin___memcpy_chk (dest, src, n, bos0 (dest)) 5174 5175char *volatile p; 5176char buf[10]; 5177/* It is unknown what object p points to, so this is optimized 5178 into plain memcpy - no checking is possible. */ 5179memcpy (p, "abcde", n); 5180/* Destination is known and length too. It is known at compile 5181 time there will be no overflow. */ 5182memcpy (&buf[5], "abcde", 5); 5183/* Destination is known, but the length is not known at compile time. 5184 This will result in __memcpy_chk call that can check for overflow 5185 at runtime. */ 5186memcpy (&buf[5], "abcde", n); 5187/* Destination is known and it is known at compile time there will 5188 be overflow. There will be a warning and __memcpy_chk call that 5189 will abort the program at runtime. */ 5190memcpy (&buf[6], "abcde", 5); 5191@end smallexample 5192 5193Such built-in functions are provided for @code{memcpy}, @code{mempcpy}, 5194@code{memmove}, @code{memset}, @code{strcpy}, @code{stpcpy}, @code{strncpy}, 5195@code{strcat} and @code{strncat}. 5196 5197There are also checking built-in functions for formatted output functions. 5198@smallexample 5199int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...); 5200int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os, 5201 const char *fmt, ...); 5202int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt, 5203 va_list ap); 5204int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os, 5205 const char *fmt, va_list ap); 5206@end smallexample 5207 5208The added @var{flag} argument is passed unchanged to @code{__sprintf_chk} 5209etc. functions and can contain implementation specific flags on what 5210additional security measures the checking function might take, such as 5211handling @code{%n} differently. 5212 5213The @var{os} argument is the object size @var{s} points to, like in the 5214other built-in functions. There is a small difference in the behavior 5215though, if @var{os} is @code{(size_t) -1}, the built-in functions are 5216optimized into the non-checking functions only if @var{flag} is 0, otherwise 5217the checking function is called with @var{os} argument set to 5218@code{(size_t) -1}. 5219 5220In addition to this, there are checking built-in functions 5221@code{__builtin___printf_chk}, @code{__builtin___vprintf_chk}, 5222@code{__builtin___fprintf_chk} and @code{__builtin___vfprintf_chk}. 5223These have just one additional argument, @var{flag}, right before 5224format string @var{fmt}. If the compiler is able to optimize them to 5225@code{fputc} etc. functions, it will, otherwise the checking function 5226should be called and the @var{flag} argument passed to it. 5227 5228@node Other Builtins 5229@section Other built-in functions provided by GCC 5230@cindex built-in functions 5231@findex __builtin_isgreater 5232@findex __builtin_isgreaterequal 5233@findex __builtin_isless 5234@findex __builtin_islessequal 5235@findex __builtin_islessgreater 5236@findex __builtin_isunordered 5237@findex __builtin_powi 5238@findex __builtin_powif 5239@findex __builtin_powil 5240@findex _Exit 5241@findex _exit 5242@findex abort 5243@findex abs 5244@findex acos 5245@findex acosf 5246@findex acosh 5247@findex acoshf 5248@findex acoshl 5249@findex acosl 5250@findex alloca 5251@findex asin 5252@findex asinf 5253@findex asinh 5254@findex asinhf 5255@findex asinhl 5256@findex asinl 5257@findex atan 5258@findex atan2 5259@findex atan2f 5260@findex atan2l 5261@findex atanf 5262@findex atanh 5263@findex atanhf 5264@findex atanhl 5265@findex atanl 5266@findex bcmp 5267@findex bzero 5268@findex cabs 5269@findex cabsf 5270@findex cabsl 5271@findex cacos 5272@findex cacosf 5273@findex cacosh 5274@findex cacoshf 5275@findex cacoshl 5276@findex cacosl 5277@findex calloc 5278@findex carg 5279@findex cargf 5280@findex cargl 5281@findex casin 5282@findex casinf 5283@findex casinh 5284@findex casinhf 5285@findex casinhl 5286@findex casinl 5287@findex catan 5288@findex catanf 5289@findex catanh 5290@findex catanhf 5291@findex catanhl 5292@findex catanl 5293@findex cbrt 5294@findex cbrtf 5295@findex cbrtl 5296@findex ccos 5297@findex ccosf 5298@findex ccosh 5299@findex ccoshf 5300@findex ccoshl 5301@findex ccosl 5302@findex ceil 5303@findex ceilf 5304@findex ceill 5305@findex cexp 5306@findex cexpf 5307@findex cexpl 5308@findex cimag 5309@findex cimagf 5310@findex cimagl 5311@findex clog 5312@findex clogf 5313@findex clogl 5314@findex conj 5315@findex conjf 5316@findex conjl 5317@findex copysign 5318@findex copysignf 5319@findex copysignl 5320@findex cos 5321@findex cosf 5322@findex cosh 5323@findex coshf 5324@findex coshl 5325@findex cosl 5326@findex cpow 5327@findex cpowf 5328@findex cpowl 5329@findex cproj 5330@findex cprojf 5331@findex cprojl 5332@findex creal 5333@findex crealf 5334@findex creall 5335@findex csin 5336@findex csinf 5337@findex csinh 5338@findex csinhf 5339@findex csinhl 5340@findex csinl 5341@findex csqrt 5342@findex csqrtf 5343@findex csqrtl 5344@findex ctan 5345@findex ctanf 5346@findex ctanh 5347@findex ctanhf 5348@findex ctanhl 5349@findex ctanl 5350@findex dcgettext 5351@findex dgettext 5352@findex drem 5353@findex dremf 5354@findex dreml 5355@findex erf 5356@findex erfc 5357@findex erfcf 5358@findex erfcl 5359@findex erff 5360@findex erfl 5361@findex exit 5362@findex exp 5363@findex exp10 5364@findex exp10f 5365@findex exp10l 5366@findex exp2 5367@findex exp2f 5368@findex exp2l 5369@findex expf 5370@findex expl 5371@findex expm1 5372@findex expm1f 5373@findex expm1l 5374@findex fabs 5375@findex fabsf 5376@findex fabsl 5377@findex fdim 5378@findex fdimf 5379@findex fdiml 5380@findex ffs 5381@findex floor 5382@findex floorf 5383@findex floorl 5384@findex fma 5385@findex fmaf 5386@findex fmal 5387@findex fmax 5388@findex fmaxf 5389@findex fmaxl 5390@findex fmin 5391@findex fminf 5392@findex fminl 5393@findex fmod 5394@findex fmodf 5395@findex fmodl 5396@findex fprintf 5397@findex fprintf_unlocked 5398@findex fputs 5399@findex fputs_unlocked 5400@findex frexp 5401@findex frexpf 5402@findex frexpl 5403@findex fscanf 5404@findex gamma 5405@findex gammaf 5406@findex gammal 5407@findex gettext 5408@findex hypot 5409@findex hypotf 5410@findex hypotl 5411@findex ilogb 5412@findex ilogbf 5413@findex ilogbl 5414@findex imaxabs 5415@findex index 5416@findex isalnum 5417@findex isalpha 5418@findex isascii 5419@findex isblank 5420@findex iscntrl 5421@findex isdigit 5422@findex isgraph 5423@findex islower 5424@findex isprint 5425@findex ispunct 5426@findex isspace 5427@findex isupper 5428@findex iswalnum 5429@findex iswalpha 5430@findex iswblank 5431@findex iswcntrl 5432@findex iswdigit 5433@findex iswgraph 5434@findex iswlower 5435@findex iswprint 5436@findex iswpunct 5437@findex iswspace 5438@findex iswupper 5439@findex iswxdigit 5440@findex isxdigit 5441@findex j0 5442@findex j0f 5443@findex j0l 5444@findex j1 5445@findex j1f 5446@findex j1l 5447@findex jn 5448@findex jnf 5449@findex jnl 5450@findex labs 5451@findex ldexp 5452@findex ldexpf 5453@findex ldexpl 5454@findex lgamma 5455@findex lgammaf 5456@findex lgammal 5457@findex llabs 5458@findex llrint 5459@findex llrintf 5460@findex llrintl 5461@findex llround 5462@findex llroundf 5463@findex llroundl 5464@findex log 5465@findex log10 5466@findex log10f 5467@findex log10l 5468@findex log1p 5469@findex log1pf 5470@findex log1pl 5471@findex log2 5472@findex log2f 5473@findex log2l 5474@findex logb 5475@findex logbf 5476@findex logbl 5477@findex logf 5478@findex logl 5479@findex lrint 5480@findex lrintf 5481@findex lrintl 5482@findex lround 5483@findex lroundf 5484@findex lroundl 5485@findex malloc 5486@findex memcmp 5487@findex memcpy 5488@findex mempcpy 5489@findex memset 5490@findex modf 5491@findex modff 5492@findex modfl 5493@findex nearbyint 5494@findex nearbyintf 5495@findex nearbyintl 5496@findex nextafter 5497@findex nextafterf 5498@findex nextafterl 5499@findex nexttoward 5500@findex nexttowardf 5501@findex nexttowardl 5502@findex pow 5503@findex pow10 5504@findex pow10f 5505@findex pow10l 5506@findex powf 5507@findex powl 5508@findex printf 5509@findex printf_unlocked 5510@findex putchar 5511@findex puts 5512@findex remainder 5513@findex remainderf 5514@findex remainderl 5515@findex remquo 5516@findex remquof 5517@findex remquol 5518@findex rindex 5519@findex rint 5520@findex rintf 5521@findex rintl 5522@findex round 5523@findex roundf 5524@findex roundl 5525@findex scalb 5526@findex scalbf 5527@findex scalbl 5528@findex scalbln 5529@findex scalblnf 5530@findex scalblnf 5531@findex scalbn 5532@findex scalbnf 5533@findex scanfnl 5534@findex signbit 5535@findex signbitf 5536@findex signbitl 5537@findex significand 5538@findex significandf 5539@findex significandl 5540@findex sin 5541@findex sincos 5542@findex sincosf 5543@findex sincosl 5544@findex sinf 5545@findex sinh 5546@findex sinhf 5547@findex sinhl 5548@findex sinl 5549@findex snprintf 5550@findex sprintf 5551@findex sqrt 5552@findex sqrtf 5553@findex sqrtl 5554@findex sscanf 5555@findex stpcpy 5556@findex stpncpy 5557@findex strcasecmp 5558@findex strcat 5559@findex strchr 5560@findex strcmp 5561@findex strcpy 5562@findex strcspn 5563@findex strdup 5564@findex strfmon 5565@findex strftime 5566@findex strlen 5567@findex strncasecmp 5568@findex strncat 5569@findex strncmp 5570@findex strncpy 5571@findex strndup 5572@findex strpbrk 5573@findex strrchr 5574@findex strspn 5575@findex strstr 5576@findex tan 5577@findex tanf 5578@findex tanh 5579@findex tanhf 5580@findex tanhl 5581@findex tanl 5582@findex tgamma 5583@findex tgammaf 5584@findex tgammal 5585@findex toascii 5586@findex tolower 5587@findex toupper 5588@findex towlower 5589@findex towupper 5590@findex trunc 5591@findex truncf 5592@findex truncl 5593@findex vfprintf 5594@findex vfscanf 5595@findex vprintf 5596@findex vscanf 5597@findex vsnprintf 5598@findex vsprintf 5599@findex vsscanf 5600@findex y0 5601@findex y0f 5602@findex y0l 5603@findex y1 5604@findex y1f 5605@findex y1l 5606@findex yn 5607@findex ynf 5608@findex ynl 5609 5610GCC provides a large number of built-in functions other than the ones 5611mentioned above. Some of these are for internal use in the processing 5612of exceptions or variable-length argument lists and will not be 5613documented here because they may change from time to time; we do not 5614recommend general use of these functions. 5615 5616The remaining functions are provided for optimization purposes. 5617 5618@opindex fno-builtin 5619GCC includes built-in versions of many of the functions in the standard 5620C library. The versions prefixed with @code{__builtin_} will always be 5621treated as having the same meaning as the C library function even if you 5622specify the @option{-fno-builtin} option. (@pxref{C Dialect Options}) 5623Many of these functions are only optimized in certain cases; if they are 5624not optimized in a particular case, a call to the library function will 5625be emitted. 5626 5627@opindex ansi 5628@opindex std 5629Outside strict ISO C mode (@option{-ansi}, @option{-std=c89} or 5630@option{-std=c99}), the functions 5631@code{_exit}, @code{alloca}, @code{bcmp}, @code{bzero}, 5632@code{dcgettext}, @code{dgettext}, @code{dremf}, @code{dreml}, 5633@code{drem}, @code{exp10f}, @code{exp10l}, @code{exp10}, @code{ffsll}, 5634@code{ffsl}, @code{ffs}, @code{fprintf_unlocked}, @code{fputs_unlocked}, 5635@code{gammaf}, @code{gammal}, @code{gamma}, @code{gettext}, 5636@code{index}, @code{isascii}, @code{j0f}, @code{j0l}, @code{j0}, 5637@code{j1f}, @code{j1l}, @code{j1}, @code{jnf}, @code{jnl}, @code{jn}, 5638@code{mempcpy}, @code{pow10f}, @code{pow10l}, @code{pow10}, 5639@code{printf_unlocked}, @code{rindex}, @code{scalbf}, @code{scalbl}, 5640@code{scalb}, @code{signbit}, @code{signbitf}, @code{signbitl}, 5641@code{significandf}, @code{significandl}, @code{significand}, 5642@code{sincosf}, @code{sincosl}, @code{sincos}, @code{stpcpy}, 5643@code{stpncpy}, @code{strcasecmp}, @code{strdup}, @code{strfmon}, 5644@code{strncasecmp}, @code{strndup}, @code{toascii}, @code{y0f}, 5645@code{y0l}, @code{y0}, @code{y1f}, @code{y1l}, @code{y1}, @code{ynf}, 5646@code{ynl} and @code{yn} 5647may be handled as built-in functions. 5648All these functions have corresponding versions 5649prefixed with @code{__builtin_}, which may be used even in strict C89 5650mode. 5651 5652The ISO C99 functions 5653@code{_Exit}, @code{acoshf}, @code{acoshl}, @code{acosh}, @code{asinhf}, 5654@code{asinhl}, @code{asinh}, @code{atanhf}, @code{atanhl}, @code{atanh}, 5655@code{cabsf}, @code{cabsl}, @code{cabs}, @code{cacosf}, @code{cacoshf}, 5656@code{cacoshl}, @code{cacosh}, @code{cacosl}, @code{cacos}, 5657@code{cargf}, @code{cargl}, @code{carg}, @code{casinf}, @code{casinhf}, 5658@code{casinhl}, @code{casinh}, @code{casinl}, @code{casin}, 5659@code{catanf}, @code{catanhf}, @code{catanhl}, @code{catanh}, 5660@code{catanl}, @code{catan}, @code{cbrtf}, @code{cbrtl}, @code{cbrt}, 5661@code{ccosf}, @code{ccoshf}, @code{ccoshl}, @code{ccosh}, @code{ccosl}, 5662@code{ccos}, @code{cexpf}, @code{cexpl}, @code{cexp}, @code{cimagf}, 5663@code{cimagl}, @code{cimag}, @code{clogf}, @code{clogl}, @code{clog}, 5664@code{conjf}, @code{conjl}, @code{conj}, @code{copysignf}, @code{copysignl}, 5665@code{copysign}, @code{cpowf}, @code{cpowl}, @code{cpow}, @code{cprojf}, 5666@code{cprojl}, @code{cproj}, @code{crealf}, @code{creall}, @code{creal}, 5667@code{csinf}, @code{csinhf}, @code{csinhl}, @code{csinh}, @code{csinl}, 5668@code{csin}, @code{csqrtf}, @code{csqrtl}, @code{csqrt}, @code{ctanf}, 5669@code{ctanhf}, @code{ctanhl}, @code{ctanh}, @code{ctanl}, @code{ctan}, 5670@code{erfcf}, @code{erfcl}, @code{erfc}, @code{erff}, @code{erfl}, 5671@code{erf}, @code{exp2f}, @code{exp2l}, @code{exp2}, @code{expm1f}, 5672@code{expm1l}, @code{expm1}, @code{fdimf}, @code{fdiml}, @code{fdim}, 5673@code{fmaf}, @code{fmal}, @code{fmaxf}, @code{fmaxl}, @code{fmax}, 5674@code{fma}, @code{fminf}, @code{fminl}, @code{fmin}, @code{hypotf}, 5675@code{hypotl}, @code{hypot}, @code{ilogbf}, @code{ilogbl}, @code{ilogb}, 5676@code{imaxabs}, @code{isblank}, @code{iswblank}, @code{lgammaf}, 5677@code{lgammal}, @code{lgamma}, @code{llabs}, @code{llrintf}, @code{llrintl}, 5678@code{llrint}, @code{llroundf}, @code{llroundl}, @code{llround}, 5679@code{log1pf}, @code{log1pl}, @code{log1p}, @code{log2f}, @code{log2l}, 5680@code{log2}, @code{logbf}, @code{logbl}, @code{logb}, @code{lrintf}, 5681@code{lrintl}, @code{lrint}, @code{lroundf}, @code{lroundl}, 5682@code{lround}, @code{nearbyintf}, @code{nearbyintl}, @code{nearbyint}, 5683@code{nextafterf}, @code{nextafterl}, @code{nextafter}, 5684@code{nexttowardf}, @code{nexttowardl}, @code{nexttoward}, 5685@code{remainderf}, @code{remainderl}, @code{remainder}, @code{remquof}, 5686@code{remquol}, @code{remquo}, @code{rintf}, @code{rintl}, @code{rint}, 5687@code{roundf}, @code{roundl}, @code{round}, @code{scalblnf}, 5688@code{scalblnl}, @code{scalbln}, @code{scalbnf}, @code{scalbnl}, 5689@code{scalbn}, @code{snprintf}, @code{tgammaf}, @code{tgammal}, 5690@code{tgamma}, @code{truncf}, @code{truncl}, @code{trunc}, 5691@code{vfscanf}, @code{vscanf}, @code{vsnprintf} and @code{vsscanf} 5692are handled as built-in functions 5693except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}). 5694 5695There are also built-in versions of the ISO C99 functions 5696@code{acosf}, @code{acosl}, @code{asinf}, @code{asinl}, @code{atan2f}, 5697@code{atan2l}, @code{atanf}, @code{atanl}, @code{ceilf}, @code{ceill}, 5698@code{cosf}, @code{coshf}, @code{coshl}, @code{cosl}, @code{expf}, 5699@code{expl}, @code{fabsf}, @code{fabsl}, @code{floorf}, @code{floorl}, 5700@code{fmodf}, @code{fmodl}, @code{frexpf}, @code{frexpl}, @code{ldexpf}, 5701@code{ldexpl}, @code{log10f}, @code{log10l}, @code{logf}, @code{logl}, 5702@code{modfl}, @code{modf}, @code{powf}, @code{powl}, @code{sinf}, 5703@code{sinhf}, @code{sinhl}, @code{sinl}, @code{sqrtf}, @code{sqrtl}, 5704@code{tanf}, @code{tanhf}, @code{tanhl} and @code{tanl} 5705that are recognized in any mode since ISO C90 reserves these names for 5706the purpose to which ISO C99 puts them. All these functions have 5707corresponding versions prefixed with @code{__builtin_}. 5708 5709The ISO C94 functions 5710@code{iswalnum}, @code{iswalpha}, @code{iswcntrl}, @code{iswdigit}, 5711@code{iswgraph}, @code{iswlower}, @code{iswprint}, @code{iswpunct}, 5712@code{iswspace}, @code{iswupper}, @code{iswxdigit}, @code{towlower} and 5713@code{towupper} 5714are handled as built-in functions 5715except in strict ISO C90 mode (@option{-ansi} or @option{-std=c89}). 5716 5717The ISO C90 functions 5718@code{abort}, @code{abs}, @code{acos}, @code{asin}, @code{atan2}, 5719@code{atan}, @code{calloc}, @code{ceil}, @code{cosh}, @code{cos}, 5720@code{exit}, @code{exp}, @code{fabs}, @code{floor}, @code{fmod}, 5721@code{fprintf}, @code{fputs}, @code{frexp}, @code{fscanf}, 5722@code{isalnum}, @code{isalpha}, @code{iscntrl}, @code{isdigit}, 5723@code{isgraph}, @code{islower}, @code{isprint}, @code{ispunct}, 5724@code{isspace}, @code{isupper}, @code{isxdigit}, @code{tolower}, 5725@code{toupper}, @code{labs}, @code{ldexp}, @code{log10}, @code{log}, 5726@code{malloc}, @code{memcmp}, @code{memcpy}, @code{memset}, @code{modf}, 5727@code{pow}, @code{printf}, @code{putchar}, @code{puts}, @code{scanf}, 5728@code{sinh}, @code{sin}, @code{snprintf}, @code{sprintf}, @code{sqrt}, 5729@code{sscanf}, @code{strcat}, @code{strchr}, @code{strcmp}, 5730@code{strcpy}, @code{strcspn}, @code{strlen}, @code{strncat}, 5731@code{strncmp}, @code{strncpy}, @code{strpbrk}, @code{strrchr}, 5732@code{strspn}, @code{strstr}, @code{tanh}, @code{tan}, @code{vfprintf}, 5733@code{vprintf} and @code{vsprintf} 5734are all recognized as built-in functions unless 5735@option{-fno-builtin} is specified (or @option{-fno-builtin-@var{function}} 5736is specified for an individual function). All of these functions have 5737corresponding versions prefixed with @code{__builtin_}. 5738 5739GCC provides built-in versions of the ISO C99 floating point comparison 5740macros that avoid raising exceptions for unordered operands. They have 5741the same names as the standard macros ( @code{isgreater}, 5742@code{isgreaterequal}, @code{isless}, @code{islessequal}, 5743@code{islessgreater}, and @code{isunordered}) , with @code{__builtin_} 5744prefixed. We intend for a library implementor to be able to simply 5745@code{#define} each standard macro to its built-in equivalent. 5746 5747@deftypefn {Built-in Function} int __builtin_types_compatible_p (@var{type1}, @var{type2}) 5748 5749You can use the built-in function @code{__builtin_types_compatible_p} to 5750determine whether two types are the same. 5751 5752This built-in function returns 1 if the unqualified versions of the 5753types @var{type1} and @var{type2} (which are types, not expressions) are 5754compatible, 0 otherwise. The result of this built-in function can be 5755used in integer constant expressions. 5756 5757This built-in function ignores top level qualifiers (e.g., @code{const}, 5758@code{volatile}). For example, @code{int} is equivalent to @code{const 5759int}. 5760 5761The type @code{int[]} and @code{int[5]} are compatible. On the other 5762hand, @code{int} and @code{char *} are not compatible, even if the size 5763of their types, on the particular architecture are the same. Also, the 5764amount of pointer indirection is taken into account when determining 5765similarity. Consequently, @code{short *} is not similar to 5766@code{short **}. Furthermore, two types that are typedefed are 5767considered compatible if their underlying types are compatible. 5768 5769An @code{enum} type is not considered to be compatible with another 5770@code{enum} type even if both are compatible with the same integer 5771type; this is what the C standard specifies. 5772For example, @code{enum @{foo, bar@}} is not similar to 5773@code{enum @{hot, dog@}}. 5774 5775You would typically use this function in code whose execution varies 5776depending on the arguments' types. For example: 5777 5778@smallexample 5779#define foo(x) \ 5780 (@{ \ 5781 typeof (x) tmp = (x); \ 5782 if (__builtin_types_compatible_p (typeof (x), long double)) \ 5783 tmp = foo_long_double (tmp); \ 5784 else if (__builtin_types_compatible_p (typeof (x), double)) \ 5785 tmp = foo_double (tmp); \ 5786 else if (__builtin_types_compatible_p (typeof (x), float)) \ 5787 tmp = foo_float (tmp); \ 5788 else \ 5789 abort (); \ 5790 tmp; \ 5791 @}) 5792@end smallexample 5793 5794@emph{Note:} This construct is only available for C@. 5795 5796@end deftypefn 5797 5798@deftypefn {Built-in Function} @var{type} __builtin_choose_expr (@var{const_exp}, @var{exp1}, @var{exp2}) 5799 5800You can use the built-in function @code{__builtin_choose_expr} to 5801evaluate code depending on the value of a constant expression. This 5802built-in function returns @var{exp1} if @var{const_exp}, which is a 5803constant expression that must be able to be determined at compile time, 5804is nonzero. Otherwise it returns 0. 5805 5806This built-in function is analogous to the @samp{? :} operator in C, 5807except that the expression returned has its type unaltered by promotion 5808rules. Also, the built-in function does not evaluate the expression 5809that was not chosen. For example, if @var{const_exp} evaluates to true, 5810@var{exp2} is not evaluated even if it has side-effects. 5811 5812This built-in function can return an lvalue if the chosen argument is an 5813lvalue. 5814 5815If @var{exp1} is returned, the return type is the same as @var{exp1}'s 5816type. Similarly, if @var{exp2} is returned, its return type is the same 5817as @var{exp2}. 5818 5819Example: 5820 5821@smallexample 5822#define foo(x) \ 5823 __builtin_choose_expr ( \ 5824 __builtin_types_compatible_p (typeof (x), double), \ 5825 foo_double (x), \ 5826 __builtin_choose_expr ( \ 5827 __builtin_types_compatible_p (typeof (x), float), \ 5828 foo_float (x), \ 5829 /* @r{The void expression results in a compile-time error} \ 5830 @r{when assigning the result to something.} */ \ 5831 (void)0)) 5832@end smallexample 5833 5834@emph{Note:} This construct is only available for C@. Furthermore, the 5835unused expression (@var{exp1} or @var{exp2} depending on the value of 5836@var{const_exp}) may still generate syntax errors. This may change in 5837future revisions. 5838 5839@end deftypefn 5840 5841@deftypefn {Built-in Function} int __builtin_constant_p (@var{exp}) 5842You can use the built-in function @code{__builtin_constant_p} to 5843determine if a value is known to be constant at compile-time and hence 5844that GCC can perform constant-folding on expressions involving that 5845value. The argument of the function is the value to test. The function 5846returns the integer 1 if the argument is known to be a compile-time 5847constant and 0 if it is not known to be a compile-time constant. A 5848return of 0 does not indicate that the value is @emph{not} a constant, 5849but merely that GCC cannot prove it is a constant with the specified 5850value of the @option{-O} option. 5851 5852You would typically use this function in an embedded application where 5853memory was a critical resource. If you have some complex calculation, 5854you may want it to be folded if it involves constants, but need to call 5855a function if it does not. For example: 5856 5857@smallexample 5858#define Scale_Value(X) \ 5859 (__builtin_constant_p (X) \ 5860 ? ((X) * SCALE + OFFSET) : Scale (X)) 5861@end smallexample 5862 5863You may use this built-in function in either a macro or an inline 5864function. However, if you use it in an inlined function and pass an 5865argument of the function as the argument to the built-in, GCC will 5866never return 1 when you call the inline function with a string constant 5867or compound literal (@pxref{Compound Literals}) and will not return 1 5868when you pass a constant numeric value to the inline function unless you 5869specify the @option{-O} option. 5870 5871You may also use @code{__builtin_constant_p} in initializers for static 5872data. For instance, you can write 5873 5874@smallexample 5875static const int table[] = @{ 5876 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1, 5877 /* @r{@dots{}} */ 5878@}; 5879@end smallexample 5880 5881@noindent 5882This is an acceptable initializer even if @var{EXPRESSION} is not a 5883constant expression. GCC must be more conservative about evaluating the 5884built-in in this case, because it has no opportunity to perform 5885optimization. 5886 5887Previous versions of GCC did not accept this built-in in data 5888initializers. The earliest version where it is completely safe is 58893.0.1. 5890@end deftypefn 5891 5892@deftypefn {Built-in Function} long __builtin_expect (long @var{exp}, long @var{c}) 5893@opindex fprofile-arcs 5894You may use @code{__builtin_expect} to provide the compiler with 5895branch prediction information. In general, you should prefer to 5896use actual profile feedback for this (@option{-fprofile-arcs}), as 5897programmers are notoriously bad at predicting how their programs 5898actually perform. However, there are applications in which this 5899data is hard to collect. 5900 5901The return value is the value of @var{exp}, which should be an 5902integral expression. The value of @var{c} must be a compile-time 5903constant. The semantics of the built-in are that it is expected 5904that @var{exp} == @var{c}. For example: 5905 5906@smallexample 5907if (__builtin_expect (x, 0)) 5908 foo (); 5909@end smallexample 5910 5911@noindent 5912would indicate that we do not expect to call @code{foo}, since 5913we expect @code{x} to be zero. Since you are limited to integral 5914expressions for @var{exp}, you should use constructions such as 5915 5916@smallexample 5917if (__builtin_expect (ptr != NULL, 1)) 5918 error (); 5919@end smallexample 5920 5921@noindent 5922when testing pointer or floating-point values. 5923@end deftypefn 5924 5925@deftypefn {Built-in Function} void __builtin_prefetch (const void *@var{addr}, ...) 5926This function is used to minimize cache-miss latency by moving data into 5927a cache before it is accessed. 5928You can insert calls to @code{__builtin_prefetch} into code for which 5929you know addresses of data in memory that is likely to be accessed soon. 5930If the target supports them, data prefetch instructions will be generated. 5931If the prefetch is done early enough before the access then the data will 5932be in the cache by the time it is accessed. 5933 5934The value of @var{addr} is the address of the memory to prefetch. 5935There are two optional arguments, @var{rw} and @var{locality}. 5936The value of @var{rw} is a compile-time constant one or zero; one 5937means that the prefetch is preparing for a write to the memory address 5938and zero, the default, means that the prefetch is preparing for a read. 5939The value @var{locality} must be a compile-time constant integer between 5940zero and three. A value of zero means that the data has no temporal 5941locality, so it need not be left in the cache after the access. A value 5942of three means that the data has a high degree of temporal locality and 5943should be left in all levels of cache possible. Values of one and two 5944mean, respectively, a low or moderate degree of temporal locality. The 5945default is three. 5946 5947@smallexample 5948for (i = 0; i < n; i++) 5949 @{ 5950 a[i] = a[i] + b[i]; 5951 __builtin_prefetch (&a[i+j], 1, 1); 5952 __builtin_prefetch (&b[i+j], 0, 1); 5953 /* @r{@dots{}} */ 5954 @} 5955@end smallexample 5956 5957Data prefetch does not generate faults if @var{addr} is invalid, but 5958the address expression itself must be valid. For example, a prefetch 5959of @code{p->next} will not fault if @code{p->next} is not a valid 5960address, but evaluation will fault if @code{p} is not a valid address. 5961 5962If the target does not support data prefetch, the address expression 5963is evaluated if it includes side effects but no other code is generated 5964and GCC does not issue a warning. 5965@end deftypefn 5966 5967@deftypefn {Built-in Function} double __builtin_huge_val (void) 5968Returns a positive infinity, if supported by the floating-point format, 5969else @code{DBL_MAX}. This function is suitable for implementing the 5970ISO C macro @code{HUGE_VAL}. 5971@end deftypefn 5972 5973@deftypefn {Built-in Function} float __builtin_huge_valf (void) 5974Similar to @code{__builtin_huge_val}, except the return type is @code{float}. 5975@end deftypefn 5976 5977@deftypefn {Built-in Function} {long double} __builtin_huge_vall (void) 5978Similar to @code{__builtin_huge_val}, except the return 5979type is @code{long double}. 5980@end deftypefn 5981 5982@deftypefn {Built-in Function} double __builtin_inf (void) 5983Similar to @code{__builtin_huge_val}, except a warning is generated 5984if the target floating-point format does not support infinities. 5985@end deftypefn 5986 5987@deftypefn {Built-in Function} _Decimal32 __builtin_infd32 (void) 5988Similar to @code{__builtin_inf}, except the return type is @code{_Decimal32}. 5989@end deftypefn 5990 5991@deftypefn {Built-in Function} _Decimal64 __builtin_infd64 (void) 5992Similar to @code{__builtin_inf}, except the return type is @code{_Decimal64}. 5993@end deftypefn 5994 5995@deftypefn {Built-in Function} _Decimal128 __builtin_infd128 (void) 5996Similar to @code{__builtin_inf}, except the return type is @code{_Decimal128}. 5997@end deftypefn 5998 5999@deftypefn {Built-in Function} float __builtin_inff (void) 6000Similar to @code{__builtin_inf}, except the return type is @code{float}. 6001This function is suitable for implementing the ISO C99 macro @code{INFINITY}. 6002@end deftypefn 6003 6004@deftypefn {Built-in Function} {long double} __builtin_infl (void) 6005Similar to @code{__builtin_inf}, except the return 6006type is @code{long double}. 6007@end deftypefn 6008 6009@deftypefn {Built-in Function} double __builtin_nan (const char *str) 6010This is an implementation of the ISO C99 function @code{nan}. 6011 6012Since ISO C99 defines this function in terms of @code{strtod}, which we 6013do not implement, a description of the parsing is in order. The string 6014is parsed as by @code{strtol}; that is, the base is recognized by 6015leading @samp{0} or @samp{0x} prefixes. The number parsed is placed 6016in the significand such that the least significant bit of the number 6017is at the least significant bit of the significand. The number is 6018truncated to fit the significand field provided. The significand is 6019forced to be a quiet NaN@. 6020 6021This function, if given a string literal all of which would have been 6022consumed by strtol, is evaluated early enough that it is considered a 6023compile-time constant. 6024@end deftypefn 6025 6026@deftypefn {Built-in Function} _Decimal32 __builtin_nand32 (const char *str) 6027Similar to @code{__builtin_nan}, except the return type is @code{_Decimal32}. 6028@end deftypefn 6029 6030@deftypefn {Built-in Function} _Decimal64 __builtin_nand64 (const char *str) 6031Similar to @code{__builtin_nan}, except the return type is @code{_Decimal64}. 6032@end deftypefn 6033 6034@deftypefn {Built-in Function} _Decimal128 __builtin_nand128 (const char *str) 6035Similar to @code{__builtin_nan}, except the return type is @code{_Decimal128}. 6036@end deftypefn 6037 6038@deftypefn {Built-in Function} float __builtin_nanf (const char *str) 6039Similar to @code{__builtin_nan}, except the return type is @code{float}. 6040@end deftypefn 6041 6042@deftypefn {Built-in Function} {long double} __builtin_nanl (const char *str) 6043Similar to @code{__builtin_nan}, except the return type is @code{long double}. 6044@end deftypefn 6045 6046@deftypefn {Built-in Function} double __builtin_nans (const char *str) 6047Similar to @code{__builtin_nan}, except the significand is forced 6048to be a signaling NaN@. The @code{nans} function is proposed by 6049@uref{http://www.open-std.org/jtc1/sc22/wg14/www/docs/n965.htm,,WG14 N965}. 6050@end deftypefn 6051 6052@deftypefn {Built-in Function} float __builtin_nansf (const char *str) 6053Similar to @code{__builtin_nans}, except the return type is @code{float}. 6054@end deftypefn 6055 6056@deftypefn {Built-in Function} {long double} __builtin_nansl (const char *str) 6057Similar to @code{__builtin_nans}, except the return type is @code{long double}. 6058@end deftypefn 6059 6060@deftypefn {Built-in Function} int __builtin_ffs (unsigned int x) 6061Returns one plus the index of the least significant 1-bit of @var{x}, or 6062if @var{x} is zero, returns zero. 6063@end deftypefn 6064 6065@deftypefn {Built-in Function} int __builtin_clz (unsigned int x) 6066Returns the number of leading 0-bits in @var{x}, starting at the most 6067significant bit position. If @var{x} is 0, the result is undefined. 6068@end deftypefn 6069 6070@deftypefn {Built-in Function} int __builtin_ctz (unsigned int x) 6071Returns the number of trailing 0-bits in @var{x}, starting at the least 6072significant bit position. If @var{x} is 0, the result is undefined. 6073@end deftypefn 6074 6075@deftypefn {Built-in Function} int __builtin_popcount (unsigned int x) 6076Returns the number of 1-bits in @var{x}. 6077@end deftypefn 6078 6079@deftypefn {Built-in Function} int __builtin_parity (unsigned int x) 6080Returns the parity of @var{x}, i.e.@: the number of 1-bits in @var{x} 6081modulo 2. 6082@end deftypefn 6083 6084@deftypefn {Built-in Function} int __builtin_ffsl (unsigned long) 6085Similar to @code{__builtin_ffs}, except the argument type is 6086@code{unsigned long}. 6087@end deftypefn 6088 6089@deftypefn {Built-in Function} int __builtin_clzl (unsigned long) 6090Similar to @code{__builtin_clz}, except the argument type is 6091@code{unsigned long}. 6092@end deftypefn 6093 6094@deftypefn {Built-in Function} int __builtin_ctzl (unsigned long) 6095Similar to @code{__builtin_ctz}, except the argument type is 6096@code{unsigned long}. 6097@end deftypefn 6098 6099@deftypefn {Built-in Function} int __builtin_popcountl (unsigned long) 6100Similar to @code{__builtin_popcount}, except the argument type is 6101@code{unsigned long}. 6102@end deftypefn 6103 6104@deftypefn {Built-in Function} int __builtin_parityl (unsigned long) 6105Similar to @code{__builtin_parity}, except the argument type is 6106@code{unsigned long}. 6107@end deftypefn 6108 6109@deftypefn {Built-in Function} int __builtin_ffsll (unsigned long long) 6110Similar to @code{__builtin_ffs}, except the argument type is 6111@code{unsigned long long}. 6112@end deftypefn 6113 6114@deftypefn {Built-in Function} int __builtin_clzll (unsigned long long) 6115Similar to @code{__builtin_clz}, except the argument type is 6116@code{unsigned long long}. 6117@end deftypefn 6118 6119@deftypefn {Built-in Function} int __builtin_ctzll (unsigned long long) 6120Similar to @code{__builtin_ctz}, except the argument type is 6121@code{unsigned long long}. 6122@end deftypefn 6123 6124@deftypefn {Built-in Function} int __builtin_popcountll (unsigned long long) 6125Similar to @code{__builtin_popcount}, except the argument type is 6126@code{unsigned long long}. 6127@end deftypefn 6128 6129@deftypefn {Built-in Function} int __builtin_parityll (unsigned long long) 6130Similar to @code{__builtin_parity}, except the argument type is 6131@code{unsigned long long}. 6132@end deftypefn 6133 6134@deftypefn {Built-in Function} double __builtin_powi (double, int) 6135Returns the first argument raised to the power of the second. Unlike the 6136@code{pow} function no guarantees about precision and rounding are made. 6137@end deftypefn 6138 6139@deftypefn {Built-in Function} float __builtin_powif (float, int) 6140Similar to @code{__builtin_powi}, except the argument and return types 6141are @code{float}. 6142@end deftypefn 6143 6144@deftypefn {Built-in Function} {long double} __builtin_powil (long double, int) 6145Similar to @code{__builtin_powi}, except the argument and return types 6146are @code{long double}. 6147@end deftypefn 6148 6149@deftypefn {Built-in Function} int32_t __builtin_bswap32 (int32_t x) 6150Returns @var{x} with the order of the bytes reversed; for example, 6151@code{0xaabbccdd} becomes @code{0xddccbbaa}. Byte here always means 6152exactly 8 bits. 6153@end deftypefn 6154 6155@deftypefn {Built-in Function} int64_t __builtin_bswap64 (int64_t x) 6156Similar to @code{__builtin_bswap32}, except the argument and return types 6157are 64-bit. 6158@end deftypefn 6159 6160@node Target Builtins 6161@section Built-in Functions Specific to Particular Target Machines 6162 6163On some target machines, GCC supports many built-in functions specific 6164to those machines. Generally these generate calls to specific machine 6165instructions, but allow the compiler to schedule those calls. 6166 6167@menu 6168* Alpha Built-in Functions:: 6169* ARM Built-in Functions:: 6170* Blackfin Built-in Functions:: 6171* FR-V Built-in Functions:: 6172* X86 Built-in Functions:: 6173* MIPS DSP Built-in Functions:: 6174* MIPS Paired-Single Support:: 6175* PowerPC AltiVec Built-in Functions:: 6176* SPARC VIS Built-in Functions:: 6177@end menu 6178 6179@node Alpha Built-in Functions 6180@subsection Alpha Built-in Functions 6181 6182These built-in functions are available for the Alpha family of 6183processors, depending on the command-line switches used. 6184 6185The following built-in functions are always available. They 6186all generate the machine instruction that is part of the name. 6187 6188@smallexample 6189long __builtin_alpha_implver (void) 6190long __builtin_alpha_rpcc (void) 6191long __builtin_alpha_amask (long) 6192long __builtin_alpha_cmpbge (long, long) 6193long __builtin_alpha_extbl (long, long) 6194long __builtin_alpha_extwl (long, long) 6195long __builtin_alpha_extll (long, long) 6196long __builtin_alpha_extql (long, long) 6197long __builtin_alpha_extwh (long, long) 6198long __builtin_alpha_extlh (long, long) 6199long __builtin_alpha_extqh (long, long) 6200long __builtin_alpha_insbl (long, long) 6201long __builtin_alpha_inswl (long, long) 6202long __builtin_alpha_insll (long, long) 6203long __builtin_alpha_insql (long, long) 6204long __builtin_alpha_inswh (long, long) 6205long __builtin_alpha_inslh (long, long) 6206long __builtin_alpha_insqh (long, long) 6207long __builtin_alpha_mskbl (long, long) 6208long __builtin_alpha_mskwl (long, long) 6209long __builtin_alpha_mskll (long, long) 6210long __builtin_alpha_mskql (long, long) 6211long __builtin_alpha_mskwh (long, long) 6212long __builtin_alpha_msklh (long, long) 6213long __builtin_alpha_mskqh (long, long) 6214long __builtin_alpha_umulh (long, long) 6215long __builtin_alpha_zap (long, long) 6216long __builtin_alpha_zapnot (long, long) 6217@end smallexample 6218 6219The following built-in functions are always with @option{-mmax} 6220or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{pca56} or 6221later. They all generate the machine instruction that is part 6222of the name. 6223 6224@smallexample 6225long __builtin_alpha_pklb (long) 6226long __builtin_alpha_pkwb (long) 6227long __builtin_alpha_unpkbl (long) 6228long __builtin_alpha_unpkbw (long) 6229long __builtin_alpha_minub8 (long, long) 6230long __builtin_alpha_minsb8 (long, long) 6231long __builtin_alpha_minuw4 (long, long) 6232long __builtin_alpha_minsw4 (long, long) 6233long __builtin_alpha_maxub8 (long, long) 6234long __builtin_alpha_maxsb8 (long, long) 6235long __builtin_alpha_maxuw4 (long, long) 6236long __builtin_alpha_maxsw4 (long, long) 6237long __builtin_alpha_perr (long, long) 6238@end smallexample 6239 6240The following built-in functions are always with @option{-mcix} 6241or @option{-mcpu=@var{cpu}} where @var{cpu} is @code{ev67} or 6242later. They all generate the machine instruction that is part 6243of the name. 6244 6245@smallexample 6246long __builtin_alpha_cttz (long) 6247long __builtin_alpha_ctlz (long) 6248long __builtin_alpha_ctpop (long) 6249@end smallexample 6250 6251The following builtins are available on systems that use the OSF/1 6252PALcode. Normally they invoke the @code{rduniq} and @code{wruniq} 6253PAL calls, but when invoked with @option{-mtls-kernel}, they invoke 6254@code{rdval} and @code{wrval}. 6255 6256@smallexample 6257void *__builtin_thread_pointer (void) 6258void __builtin_set_thread_pointer (void *) 6259@end smallexample 6260 6261@node ARM Built-in Functions 6262@subsection ARM Built-in Functions 6263 6264These built-in functions are available for the ARM family of 6265processors, when the @option{-mcpu=iwmmxt} switch is used: 6266 6267@smallexample 6268typedef int v2si __attribute__ ((vector_size (8))); 6269typedef short v4hi __attribute__ ((vector_size (8))); 6270typedef char v8qi __attribute__ ((vector_size (8))); 6271 6272int __builtin_arm_getwcx (int) 6273void __builtin_arm_setwcx (int, int) 6274int __builtin_arm_textrmsb (v8qi, int) 6275int __builtin_arm_textrmsh (v4hi, int) 6276int __builtin_arm_textrmsw (v2si, int) 6277int __builtin_arm_textrmub (v8qi, int) 6278int __builtin_arm_textrmuh (v4hi, int) 6279int __builtin_arm_textrmuw (v2si, int) 6280v8qi __builtin_arm_tinsrb (v8qi, int) 6281v4hi __builtin_arm_tinsrh (v4hi, int) 6282v2si __builtin_arm_tinsrw (v2si, int) 6283long long __builtin_arm_tmia (long long, int, int) 6284long long __builtin_arm_tmiabb (long long, int, int) 6285long long __builtin_arm_tmiabt (long long, int, int) 6286long long __builtin_arm_tmiaph (long long, int, int) 6287long long __builtin_arm_tmiatb (long long, int, int) 6288long long __builtin_arm_tmiatt (long long, int, int) 6289int __builtin_arm_tmovmskb (v8qi) 6290int __builtin_arm_tmovmskh (v4hi) 6291int __builtin_arm_tmovmskw (v2si) 6292long long __builtin_arm_waccb (v8qi) 6293long long __builtin_arm_wacch (v4hi) 6294long long __builtin_arm_waccw (v2si) 6295v8qi __builtin_arm_waddb (v8qi, v8qi) 6296v8qi __builtin_arm_waddbss (v8qi, v8qi) 6297v8qi __builtin_arm_waddbus (v8qi, v8qi) 6298v4hi __builtin_arm_waddh (v4hi, v4hi) 6299v4hi __builtin_arm_waddhss (v4hi, v4hi) 6300v4hi __builtin_arm_waddhus (v4hi, v4hi) 6301v2si __builtin_arm_waddw (v2si, v2si) 6302v2si __builtin_arm_waddwss (v2si, v2si) 6303v2si __builtin_arm_waddwus (v2si, v2si) 6304v8qi __builtin_arm_walign (v8qi, v8qi, int) 6305long long __builtin_arm_wand(long long, long long) 6306long long __builtin_arm_wandn (long long, long long) 6307v8qi __builtin_arm_wavg2b (v8qi, v8qi) 6308v8qi __builtin_arm_wavg2br (v8qi, v8qi) 6309v4hi __builtin_arm_wavg2h (v4hi, v4hi) 6310v4hi __builtin_arm_wavg2hr (v4hi, v4hi) 6311v8qi __builtin_arm_wcmpeqb (v8qi, v8qi) 6312v4hi __builtin_arm_wcmpeqh (v4hi, v4hi) 6313v2si __builtin_arm_wcmpeqw (v2si, v2si) 6314v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi) 6315v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi) 6316v2si __builtin_arm_wcmpgtsw (v2si, v2si) 6317v8qi __builtin_arm_wcmpgtub (v8qi, v8qi) 6318v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi) 6319v2si __builtin_arm_wcmpgtuw (v2si, v2si) 6320long long __builtin_arm_wmacs (long long, v4hi, v4hi) 6321long long __builtin_arm_wmacsz (v4hi, v4hi) 6322long long __builtin_arm_wmacu (long long, v4hi, v4hi) 6323long long __builtin_arm_wmacuz (v4hi, v4hi) 6324v4hi __builtin_arm_wmadds (v4hi, v4hi) 6325v4hi __builtin_arm_wmaddu (v4hi, v4hi) 6326v8qi __builtin_arm_wmaxsb (v8qi, v8qi) 6327v4hi __builtin_arm_wmaxsh (v4hi, v4hi) 6328v2si __builtin_arm_wmaxsw (v2si, v2si) 6329v8qi __builtin_arm_wmaxub (v8qi, v8qi) 6330v4hi __builtin_arm_wmaxuh (v4hi, v4hi) 6331v2si __builtin_arm_wmaxuw (v2si, v2si) 6332v8qi __builtin_arm_wminsb (v8qi, v8qi) 6333v4hi __builtin_arm_wminsh (v4hi, v4hi) 6334v2si __builtin_arm_wminsw (v2si, v2si) 6335v8qi __builtin_arm_wminub (v8qi, v8qi) 6336v4hi __builtin_arm_wminuh (v4hi, v4hi) 6337v2si __builtin_arm_wminuw (v2si, v2si) 6338v4hi __builtin_arm_wmulsm (v4hi, v4hi) 6339v4hi __builtin_arm_wmulul (v4hi, v4hi) 6340v4hi __builtin_arm_wmulum (v4hi, v4hi) 6341long long __builtin_arm_wor (long long, long long) 6342v2si __builtin_arm_wpackdss (long long, long long) 6343v2si __builtin_arm_wpackdus (long long, long long) 6344v8qi __builtin_arm_wpackhss (v4hi, v4hi) 6345v8qi __builtin_arm_wpackhus (v4hi, v4hi) 6346v4hi __builtin_arm_wpackwss (v2si, v2si) 6347v4hi __builtin_arm_wpackwus (v2si, v2si) 6348long long __builtin_arm_wrord (long long, long long) 6349long long __builtin_arm_wrordi (long long, int) 6350v4hi __builtin_arm_wrorh (v4hi, long long) 6351v4hi __builtin_arm_wrorhi (v4hi, int) 6352v2si __builtin_arm_wrorw (v2si, long long) 6353v2si __builtin_arm_wrorwi (v2si, int) 6354v2si __builtin_arm_wsadb (v8qi, v8qi) 6355v2si __builtin_arm_wsadbz (v8qi, v8qi) 6356v2si __builtin_arm_wsadh (v4hi, v4hi) 6357v2si __builtin_arm_wsadhz (v4hi, v4hi) 6358v4hi __builtin_arm_wshufh (v4hi, int) 6359long long __builtin_arm_wslld (long long, long long) 6360long long __builtin_arm_wslldi (long long, int) 6361v4hi __builtin_arm_wsllh (v4hi, long long) 6362v4hi __builtin_arm_wsllhi (v4hi, int) 6363v2si __builtin_arm_wsllw (v2si, long long) 6364v2si __builtin_arm_wsllwi (v2si, int) 6365long long __builtin_arm_wsrad (long long, long long) 6366long long __builtin_arm_wsradi (long long, int) 6367v4hi __builtin_arm_wsrah (v4hi, long long) 6368v4hi __builtin_arm_wsrahi (v4hi, int) 6369v2si __builtin_arm_wsraw (v2si, long long) 6370v2si __builtin_arm_wsrawi (v2si, int) 6371long long __builtin_arm_wsrld (long long, long long) 6372long long __builtin_arm_wsrldi (long long, int) 6373v4hi __builtin_arm_wsrlh (v4hi, long long) 6374v4hi __builtin_arm_wsrlhi (v4hi, int) 6375v2si __builtin_arm_wsrlw (v2si, long long) 6376v2si __builtin_arm_wsrlwi (v2si, int) 6377v8qi __builtin_arm_wsubb (v8qi, v8qi) 6378v8qi __builtin_arm_wsubbss (v8qi, v8qi) 6379v8qi __builtin_arm_wsubbus (v8qi, v8qi) 6380v4hi __builtin_arm_wsubh (v4hi, v4hi) 6381v4hi __builtin_arm_wsubhss (v4hi, v4hi) 6382v4hi __builtin_arm_wsubhus (v4hi, v4hi) 6383v2si __builtin_arm_wsubw (v2si, v2si) 6384v2si __builtin_arm_wsubwss (v2si, v2si) 6385v2si __builtin_arm_wsubwus (v2si, v2si) 6386v4hi __builtin_arm_wunpckehsb (v8qi) 6387v2si __builtin_arm_wunpckehsh (v4hi) 6388long long __builtin_arm_wunpckehsw (v2si) 6389v4hi __builtin_arm_wunpckehub (v8qi) 6390v2si __builtin_arm_wunpckehuh (v4hi) 6391long long __builtin_arm_wunpckehuw (v2si) 6392v4hi __builtin_arm_wunpckelsb (v8qi) 6393v2si __builtin_arm_wunpckelsh (v4hi) 6394long long __builtin_arm_wunpckelsw (v2si) 6395v4hi __builtin_arm_wunpckelub (v8qi) 6396v2si __builtin_arm_wunpckeluh (v4hi) 6397long long __builtin_arm_wunpckeluw (v2si) 6398v8qi __builtin_arm_wunpckihb (v8qi, v8qi) 6399v4hi __builtin_arm_wunpckihh (v4hi, v4hi) 6400v2si __builtin_arm_wunpckihw (v2si, v2si) 6401v8qi __builtin_arm_wunpckilb (v8qi, v8qi) 6402v4hi __builtin_arm_wunpckilh (v4hi, v4hi) 6403v2si __builtin_arm_wunpckilw (v2si, v2si) 6404long long __builtin_arm_wxor (long long, long long) 6405long long __builtin_arm_wzero () 6406@end smallexample 6407 6408@node Blackfin Built-in Functions 6409@subsection Blackfin Built-in Functions 6410 6411Currently, there are two Blackfin-specific built-in functions. These are 6412used for generating @code{CSYNC} and @code{SSYNC} machine insns without 6413using inline assembly; by using these built-in functions the compiler can 6414automatically add workarounds for hardware errata involving these 6415instructions. These functions are named as follows: 6416 6417@smallexample 6418void __builtin_bfin_csync (void) 6419void __builtin_bfin_ssync (void) 6420@end smallexample 6421 6422@node FR-V Built-in Functions 6423@subsection FR-V Built-in Functions 6424 6425GCC provides many FR-V-specific built-in functions. In general, 6426these functions are intended to be compatible with those described 6427by @cite{FR-V Family, Softune C/C++ Compiler Manual (V6), Fujitsu 6428Semiconductor}. The two exceptions are @code{__MDUNPACKH} and 6429@code{__MBTOHE}, the gcc forms of which pass 128-bit values by 6430pointer rather than by value. 6431 6432Most of the functions are named after specific FR-V instructions. 6433Such functions are said to be ``directly mapped'' and are summarized 6434here in tabular form. 6435 6436@menu 6437* Argument Types:: 6438* Directly-mapped Integer Functions:: 6439* Directly-mapped Media Functions:: 6440* Raw read/write Functions:: 6441* Other Built-in Functions:: 6442@end menu 6443 6444@node Argument Types 6445@subsubsection Argument Types 6446 6447The arguments to the built-in functions can be divided into three groups: 6448register numbers, compile-time constants and run-time values. In order 6449to make this classification clear at a glance, the arguments and return 6450values are given the following pseudo types: 6451 6452@multitable @columnfractions .20 .30 .15 .35 6453@item Pseudo type @tab Real C type @tab Constant? @tab Description 6454@item @code{uh} @tab @code{unsigned short} @tab No @tab an unsigned halfword 6455@item @code{uw1} @tab @code{unsigned int} @tab No @tab an unsigned word 6456@item @code{sw1} @tab @code{int} @tab No @tab a signed word 6457@item @code{uw2} @tab @code{unsigned long long} @tab No 6458@tab an unsigned doubleword 6459@item @code{sw2} @tab @code{long long} @tab No @tab a signed doubleword 6460@item @code{const} @tab @code{int} @tab Yes @tab an integer constant 6461@item @code{acc} @tab @code{int} @tab Yes @tab an ACC register number 6462@item @code{iacc} @tab @code{int} @tab Yes @tab an IACC register number 6463@end multitable 6464 6465These pseudo types are not defined by GCC, they are simply a notational 6466convenience used in this manual. 6467 6468Arguments of type @code{uh}, @code{uw1}, @code{sw1}, @code{uw2} 6469and @code{sw2} are evaluated at run time. They correspond to 6470register operands in the underlying FR-V instructions. 6471 6472@code{const} arguments represent immediate operands in the underlying 6473FR-V instructions. They must be compile-time constants. 6474 6475@code{acc} arguments are evaluated at compile time and specify the number 6476of an accumulator register. For example, an @code{acc} argument of 2 6477will select the ACC2 register. 6478 6479@code{iacc} arguments are similar to @code{acc} arguments but specify the 6480number of an IACC register. See @pxref{Other Built-in Functions} 6481for more details. 6482 6483@node Directly-mapped Integer Functions 6484@subsubsection Directly-mapped Integer Functions 6485 6486The functions listed below map directly to FR-V I-type instructions. 6487 6488@multitable @columnfractions .45 .32 .23 6489@item Function prototype @tab Example usage @tab Assembly output 6490@item @code{sw1 __ADDSS (sw1, sw1)} 6491@tab @code{@var{c} = __ADDSS (@var{a}, @var{b})} 6492@tab @code{ADDSS @var{a},@var{b},@var{c}} 6493@item @code{sw1 __SCAN (sw1, sw1)} 6494@tab @code{@var{c} = __SCAN (@var{a}, @var{b})} 6495@tab @code{SCAN @var{a},@var{b},@var{c}} 6496@item @code{sw1 __SCUTSS (sw1)} 6497@tab @code{@var{b} = __SCUTSS (@var{a})} 6498@tab @code{SCUTSS @var{a},@var{b}} 6499@item @code{sw1 __SLASS (sw1, sw1)} 6500@tab @code{@var{c} = __SLASS (@var{a}, @var{b})} 6501@tab @code{SLASS @var{a},@var{b},@var{c}} 6502@item @code{void __SMASS (sw1, sw1)} 6503@tab @code{__SMASS (@var{a}, @var{b})} 6504@tab @code{SMASS @var{a},@var{b}} 6505@item @code{void __SMSSS (sw1, sw1)} 6506@tab @code{__SMSSS (@var{a}, @var{b})} 6507@tab @code{SMSSS @var{a},@var{b}} 6508@item @code{void __SMU (sw1, sw1)} 6509@tab @code{__SMU (@var{a}, @var{b})} 6510@tab @code{SMU @var{a},@var{b}} 6511@item @code{sw2 __SMUL (sw1, sw1)} 6512@tab @code{@var{c} = __SMUL (@var{a}, @var{b})} 6513@tab @code{SMUL @var{a},@var{b},@var{c}} 6514@item @code{sw1 __SUBSS (sw1, sw1)} 6515@tab @code{@var{c} = __SUBSS (@var{a}, @var{b})} 6516@tab @code{SUBSS @var{a},@var{b},@var{c}} 6517@item @code{uw2 __UMUL (uw1, uw1)} 6518@tab @code{@var{c} = __UMUL (@var{a}, @var{b})} 6519@tab @code{UMUL @var{a},@var{b},@var{c}} 6520@end multitable 6521 6522@node Directly-mapped Media Functions 6523@subsubsection Directly-mapped Media Functions 6524 6525The functions listed below map directly to FR-V M-type instructions. 6526 6527@multitable @columnfractions .45 .32 .23 6528@item Function prototype @tab Example usage @tab Assembly output 6529@item @code{uw1 __MABSHS (sw1)} 6530@tab @code{@var{b} = __MABSHS (@var{a})} 6531@tab @code{MABSHS @var{a},@var{b}} 6532@item @code{void __MADDACCS (acc, acc)} 6533@tab @code{__MADDACCS (@var{b}, @var{a})} 6534@tab @code{MADDACCS @var{a},@var{b}} 6535@item @code{sw1 __MADDHSS (sw1, sw1)} 6536@tab @code{@var{c} = __MADDHSS (@var{a}, @var{b})} 6537@tab @code{MADDHSS @var{a},@var{b},@var{c}} 6538@item @code{uw1 __MADDHUS (uw1, uw1)} 6539@tab @code{@var{c} = __MADDHUS (@var{a}, @var{b})} 6540@tab @code{MADDHUS @var{a},@var{b},@var{c}} 6541@item @code{uw1 __MAND (uw1, uw1)} 6542@tab @code{@var{c} = __MAND (@var{a}, @var{b})} 6543@tab @code{MAND @var{a},@var{b},@var{c}} 6544@item @code{void __MASACCS (acc, acc)} 6545@tab @code{__MASACCS (@var{b}, @var{a})} 6546@tab @code{MASACCS @var{a},@var{b}} 6547@item @code{uw1 __MAVEH (uw1, uw1)} 6548@tab @code{@var{c} = __MAVEH (@var{a}, @var{b})} 6549@tab @code{MAVEH @var{a},@var{b},@var{c}} 6550@item @code{uw2 __MBTOH (uw1)} 6551@tab @code{@var{b} = __MBTOH (@var{a})} 6552@tab @code{MBTOH @var{a},@var{b}} 6553@item @code{void __MBTOHE (uw1 *, uw1)} 6554@tab @code{__MBTOHE (&@var{b}, @var{a})} 6555@tab @code{MBTOHE @var{a},@var{b}} 6556@item @code{void __MCLRACC (acc)} 6557@tab @code{__MCLRACC (@var{a})} 6558@tab @code{MCLRACC @var{a}} 6559@item @code{void __MCLRACCA (void)} 6560@tab @code{__MCLRACCA ()} 6561@tab @code{MCLRACCA} 6562@item @code{uw1 __Mcop1 (uw1, uw1)} 6563@tab @code{@var{c} = __Mcop1 (@var{a}, @var{b})} 6564@tab @code{Mcop1 @var{a},@var{b},@var{c}} 6565@item @code{uw1 __Mcop2 (uw1, uw1)} 6566@tab @code{@var{c} = __Mcop2 (@var{a}, @var{b})} 6567@tab @code{Mcop2 @var{a},@var{b},@var{c}} 6568@item @code{uw1 __MCPLHI (uw2, const)} 6569@tab @code{@var{c} = __MCPLHI (@var{a}, @var{b})} 6570@tab @code{MCPLHI @var{a},#@var{b},@var{c}} 6571@item @code{uw1 __MCPLI (uw2, const)} 6572@tab @code{@var{c} = __MCPLI (@var{a}, @var{b})} 6573@tab @code{MCPLI @var{a},#@var{b},@var{c}} 6574@item @code{void __MCPXIS (acc, sw1, sw1)} 6575@tab @code{__MCPXIS (@var{c}, @var{a}, @var{b})} 6576@tab @code{MCPXIS @var{a},@var{b},@var{c}} 6577@item @code{void __MCPXIU (acc, uw1, uw1)} 6578@tab @code{__MCPXIU (@var{c}, @var{a}, @var{b})} 6579@tab @code{MCPXIU @var{a},@var{b},@var{c}} 6580@item @code{void __MCPXRS (acc, sw1, sw1)} 6581@tab @code{__MCPXRS (@var{c}, @var{a}, @var{b})} 6582@tab @code{MCPXRS @var{a},@var{b},@var{c}} 6583@item @code{void __MCPXRU (acc, uw1, uw1)} 6584@tab @code{__MCPXRU (@var{c}, @var{a}, @var{b})} 6585@tab @code{MCPXRU @var{a},@var{b},@var{c}} 6586@item @code{uw1 __MCUT (acc, uw1)} 6587@tab @code{@var{c} = __MCUT (@var{a}, @var{b})} 6588@tab @code{MCUT @var{a},@var{b},@var{c}} 6589@item @code{uw1 __MCUTSS (acc, sw1)} 6590@tab @code{@var{c} = __MCUTSS (@var{a}, @var{b})} 6591@tab @code{MCUTSS @var{a},@var{b},@var{c}} 6592@item @code{void __MDADDACCS (acc, acc)} 6593@tab @code{__MDADDACCS (@var{b}, @var{a})} 6594@tab @code{MDADDACCS @var{a},@var{b}} 6595@item @code{void __MDASACCS (acc, acc)} 6596@tab @code{__MDASACCS (@var{b}, @var{a})} 6597@tab @code{MDASACCS @var{a},@var{b}} 6598@item @code{uw2 __MDCUTSSI (acc, const)} 6599@tab @code{@var{c} = __MDCUTSSI (@var{a}, @var{b})} 6600@tab @code{MDCUTSSI @var{a},#@var{b},@var{c}} 6601@item @code{uw2 __MDPACKH (uw2, uw2)} 6602@tab @code{@var{c} = __MDPACKH (@var{a}, @var{b})} 6603@tab @code{MDPACKH @var{a},@var{b},@var{c}} 6604@item @code{uw2 __MDROTLI (uw2, const)} 6605@tab @code{@var{c} = __MDROTLI (@var{a}, @var{b})} 6606@tab @code{MDROTLI @var{a},#@var{b},@var{c}} 6607@item @code{void __MDSUBACCS (acc, acc)} 6608@tab @code{__MDSUBACCS (@var{b}, @var{a})} 6609@tab @code{MDSUBACCS @var{a},@var{b}} 6610@item @code{void __MDUNPACKH (uw1 *, uw2)} 6611@tab @code{__MDUNPACKH (&@var{b}, @var{a})} 6612@tab @code{MDUNPACKH @var{a},@var{b}} 6613@item @code{uw2 __MEXPDHD (uw1, const)} 6614@tab @code{@var{c} = __MEXPDHD (@var{a}, @var{b})} 6615@tab @code{MEXPDHD @var{a},#@var{b},@var{c}} 6616@item @code{uw1 __MEXPDHW (uw1, const)} 6617@tab @code{@var{c} = __MEXPDHW (@var{a}, @var{b})} 6618@tab @code{MEXPDHW @var{a},#@var{b},@var{c}} 6619@item @code{uw1 __MHDSETH (uw1, const)} 6620@tab @code{@var{c} = __MHDSETH (@var{a}, @var{b})} 6621@tab @code{MHDSETH @var{a},#@var{b},@var{c}} 6622@item @code{sw1 __MHDSETS (const)} 6623@tab @code{@var{b} = __MHDSETS (@var{a})} 6624@tab @code{MHDSETS #@var{a},@var{b}} 6625@item @code{uw1 __MHSETHIH (uw1, const)} 6626@tab @code{@var{b} = __MHSETHIH (@var{b}, @var{a})} 6627@tab @code{MHSETHIH #@var{a},@var{b}} 6628@item @code{sw1 __MHSETHIS (sw1, const)} 6629@tab @code{@var{b} = __MHSETHIS (@var{b}, @var{a})} 6630@tab @code{MHSETHIS #@var{a},@var{b}} 6631@item @code{uw1 __MHSETLOH (uw1, const)} 6632@tab @code{@var{b} = __MHSETLOH (@var{b}, @var{a})} 6633@tab @code{MHSETLOH #@var{a},@var{b}} 6634@item @code{sw1 __MHSETLOS (sw1, const)} 6635@tab @code{@var{b} = __MHSETLOS (@var{b}, @var{a})} 6636@tab @code{MHSETLOS #@var{a},@var{b}} 6637@item @code{uw1 __MHTOB (uw2)} 6638@tab @code{@var{b} = __MHTOB (@var{a})} 6639@tab @code{MHTOB @var{a},@var{b}} 6640@item @code{void __MMACHS (acc, sw1, sw1)} 6641@tab @code{__MMACHS (@var{c}, @var{a}, @var{b})} 6642@tab @code{MMACHS @var{a},@var{b},@var{c}} 6643@item @code{void __MMACHU (acc, uw1, uw1)} 6644@tab @code{__MMACHU (@var{c}, @var{a}, @var{b})} 6645@tab @code{MMACHU @var{a},@var{b},@var{c}} 6646@item @code{void __MMRDHS (acc, sw1, sw1)} 6647@tab @code{__MMRDHS (@var{c}, @var{a}, @var{b})} 6648@tab @code{MMRDHS @var{a},@var{b},@var{c}} 6649@item @code{void __MMRDHU (acc, uw1, uw1)} 6650@tab @code{__MMRDHU (@var{c}, @var{a}, @var{b})} 6651@tab @code{MMRDHU @var{a},@var{b},@var{c}} 6652@item @code{void __MMULHS (acc, sw1, sw1)} 6653@tab @code{__MMULHS (@var{c}, @var{a}, @var{b})} 6654@tab @code{MMULHS @var{a},@var{b},@var{c}} 6655@item @code{void __MMULHU (acc, uw1, uw1)} 6656@tab @code{__MMULHU (@var{c}, @var{a}, @var{b})} 6657@tab @code{MMULHU @var{a},@var{b},@var{c}} 6658@item @code{void __MMULXHS (acc, sw1, sw1)} 6659@tab @code{__MMULXHS (@var{c}, @var{a}, @var{b})} 6660@tab @code{MMULXHS @var{a},@var{b},@var{c}} 6661@item @code{void __MMULXHU (acc, uw1, uw1)} 6662@tab @code{__MMULXHU (@var{c}, @var{a}, @var{b})} 6663@tab @code{MMULXHU @var{a},@var{b},@var{c}} 6664@item @code{uw1 __MNOT (uw1)} 6665@tab @code{@var{b} = __MNOT (@var{a})} 6666@tab @code{MNOT @var{a},@var{b}} 6667@item @code{uw1 __MOR (uw1, uw1)} 6668@tab @code{@var{c} = __MOR (@var{a}, @var{b})} 6669@tab @code{MOR @var{a},@var{b},@var{c}} 6670@item @code{uw1 __MPACKH (uh, uh)} 6671@tab @code{@var{c} = __MPACKH (@var{a}, @var{b})} 6672@tab @code{MPACKH @var{a},@var{b},@var{c}} 6673@item @code{sw2 __MQADDHSS (sw2, sw2)} 6674@tab @code{@var{c} = __MQADDHSS (@var{a}, @var{b})} 6675@tab @code{MQADDHSS @var{a},@var{b},@var{c}} 6676@item @code{uw2 __MQADDHUS (uw2, uw2)} 6677@tab @code{@var{c} = __MQADDHUS (@var{a}, @var{b})} 6678@tab @code{MQADDHUS @var{a},@var{b},@var{c}} 6679@item @code{void __MQCPXIS (acc, sw2, sw2)} 6680@tab @code{__MQCPXIS (@var{c}, @var{a}, @var{b})} 6681@tab @code{MQCPXIS @var{a},@var{b},@var{c}} 6682@item @code{void __MQCPXIU (acc, uw2, uw2)} 6683@tab @code{__MQCPXIU (@var{c}, @var{a}, @var{b})} 6684@tab @code{MQCPXIU @var{a},@var{b},@var{c}} 6685@item @code{void __MQCPXRS (acc, sw2, sw2)} 6686@tab @code{__MQCPXRS (@var{c}, @var{a}, @var{b})} 6687@tab @code{MQCPXRS @var{a},@var{b},@var{c}} 6688@item @code{void __MQCPXRU (acc, uw2, uw2)} 6689@tab @code{__MQCPXRU (@var{c}, @var{a}, @var{b})} 6690@tab @code{MQCPXRU @var{a},@var{b},@var{c}} 6691@item @code{sw2 __MQLCLRHS (sw2, sw2)} 6692@tab @code{@var{c} = __MQLCLRHS (@var{a}, @var{b})} 6693@tab @code{MQLCLRHS @var{a},@var{b},@var{c}} 6694@item @code{sw2 __MQLMTHS (sw2, sw2)} 6695@tab @code{@var{c} = __MQLMTHS (@var{a}, @var{b})} 6696@tab @code{MQLMTHS @var{a},@var{b},@var{c}} 6697@item @code{void __MQMACHS (acc, sw2, sw2)} 6698@tab @code{__MQMACHS (@var{c}, @var{a}, @var{b})} 6699@tab @code{MQMACHS @var{a},@var{b},@var{c}} 6700@item @code{void __MQMACHU (acc, uw2, uw2)} 6701@tab @code{__MQMACHU (@var{c}, @var{a}, @var{b})} 6702@tab @code{MQMACHU @var{a},@var{b},@var{c}} 6703@item @code{void __MQMACXHS (acc, sw2, sw2)} 6704@tab @code{__MQMACXHS (@var{c}, @var{a}, @var{b})} 6705@tab @code{MQMACXHS @var{a},@var{b},@var{c}} 6706@item @code{void __MQMULHS (acc, sw2, sw2)} 6707@tab @code{__MQMULHS (@var{c}, @var{a}, @var{b})} 6708@tab @code{MQMULHS @var{a},@var{b},@var{c}} 6709@item @code{void __MQMULHU (acc, uw2, uw2)} 6710@tab @code{__MQMULHU (@var{c}, @var{a}, @var{b})} 6711@tab @code{MQMULHU @var{a},@var{b},@var{c}} 6712@item @code{void __MQMULXHS (acc, sw2, sw2)} 6713@tab @code{__MQMULXHS (@var{c}, @var{a}, @var{b})} 6714@tab @code{MQMULXHS @var{a},@var{b},@var{c}} 6715@item @code{void __MQMULXHU (acc, uw2, uw2)} 6716@tab @code{__MQMULXHU (@var{c}, @var{a}, @var{b})} 6717@tab @code{MQMULXHU @var{a},@var{b},@var{c}} 6718@item @code{sw2 __MQSATHS (sw2, sw2)} 6719@tab @code{@var{c} = __MQSATHS (@var{a}, @var{b})} 6720@tab @code{MQSATHS @var{a},@var{b},@var{c}} 6721@item @code{uw2 __MQSLLHI (uw2, int)} 6722@tab @code{@var{c} = __MQSLLHI (@var{a}, @var{b})} 6723@tab @code{MQSLLHI @var{a},@var{b},@var{c}} 6724@item @code{sw2 __MQSRAHI (sw2, int)} 6725@tab @code{@var{c} = __MQSRAHI (@var{a}, @var{b})} 6726@tab @code{MQSRAHI @var{a},@var{b},@var{c}} 6727@item @code{sw2 __MQSUBHSS (sw2, sw2)} 6728@tab @code{@var{c} = __MQSUBHSS (@var{a}, @var{b})} 6729@tab @code{MQSUBHSS @var{a},@var{b},@var{c}} 6730@item @code{uw2 __MQSUBHUS (uw2, uw2)} 6731@tab @code{@var{c} = __MQSUBHUS (@var{a}, @var{b})} 6732@tab @code{MQSUBHUS @var{a},@var{b},@var{c}} 6733@item @code{void __MQXMACHS (acc, sw2, sw2)} 6734@tab @code{__MQXMACHS (@var{c}, @var{a}, @var{b})} 6735@tab @code{MQXMACHS @var{a},@var{b},@var{c}} 6736@item @code{void __MQXMACXHS (acc, sw2, sw2)} 6737@tab @code{__MQXMACXHS (@var{c}, @var{a}, @var{b})} 6738@tab @code{MQXMACXHS @var{a},@var{b},@var{c}} 6739@item @code{uw1 __MRDACC (acc)} 6740@tab @code{@var{b} = __MRDACC (@var{a})} 6741@tab @code{MRDACC @var{a},@var{b}} 6742@item @code{uw1 __MRDACCG (acc)} 6743@tab @code{@var{b} = __MRDACCG (@var{a})} 6744@tab @code{MRDACCG @var{a},@var{b}} 6745@item @code{uw1 __MROTLI (uw1, const)} 6746@tab @code{@var{c} = __MROTLI (@var{a}, @var{b})} 6747@tab @code{MROTLI @var{a},#@var{b},@var{c}} 6748@item @code{uw1 __MROTRI (uw1, const)} 6749@tab @code{@var{c} = __MROTRI (@var{a}, @var{b})} 6750@tab @code{MROTRI @var{a},#@var{b},@var{c}} 6751@item @code{sw1 __MSATHS (sw1, sw1)} 6752@tab @code{@var{c} = __MSATHS (@var{a}, @var{b})} 6753@tab @code{MSATHS @var{a},@var{b},@var{c}} 6754@item @code{uw1 __MSATHU (uw1, uw1)} 6755@tab @code{@var{c} = __MSATHU (@var{a}, @var{b})} 6756@tab @code{MSATHU @var{a},@var{b},@var{c}} 6757@item @code{uw1 __MSLLHI (uw1, const)} 6758@tab @code{@var{c} = __MSLLHI (@var{a}, @var{b})} 6759@tab @code{MSLLHI @var{a},#@var{b},@var{c}} 6760@item @code{sw1 __MSRAHI (sw1, const)} 6761@tab @code{@var{c} = __MSRAHI (@var{a}, @var{b})} 6762@tab @code{MSRAHI @var{a},#@var{b},@var{c}} 6763@item @code{uw1 __MSRLHI (uw1, const)} 6764@tab @code{@var{c} = __MSRLHI (@var{a}, @var{b})} 6765@tab @code{MSRLHI @var{a},#@var{b},@var{c}} 6766@item @code{void __MSUBACCS (acc, acc)} 6767@tab @code{__MSUBACCS (@var{b}, @var{a})} 6768@tab @code{MSUBACCS @var{a},@var{b}} 6769@item @code{sw1 __MSUBHSS (sw1, sw1)} 6770@tab @code{@var{c} = __MSUBHSS (@var{a}, @var{b})} 6771@tab @code{MSUBHSS @var{a},@var{b},@var{c}} 6772@item @code{uw1 __MSUBHUS (uw1, uw1)} 6773@tab @code{@var{c} = __MSUBHUS (@var{a}, @var{b})} 6774@tab @code{MSUBHUS @var{a},@var{b},@var{c}} 6775@item @code{void __MTRAP (void)} 6776@tab @code{__MTRAP ()} 6777@tab @code{MTRAP} 6778@item @code{uw2 __MUNPACKH (uw1)} 6779@tab @code{@var{b} = __MUNPACKH (@var{a})} 6780@tab @code{MUNPACKH @var{a},@var{b}} 6781@item @code{uw1 __MWCUT (uw2, uw1)} 6782@tab @code{@var{c} = __MWCUT (@var{a}, @var{b})} 6783@tab @code{MWCUT @var{a},@var{b},@var{c}} 6784@item @code{void __MWTACC (acc, uw1)} 6785@tab @code{__MWTACC (@var{b}, @var{a})} 6786@tab @code{MWTACC @var{a},@var{b}} 6787@item @code{void __MWTACCG (acc, uw1)} 6788@tab @code{__MWTACCG (@var{b}, @var{a})} 6789@tab @code{MWTACCG @var{a},@var{b}} 6790@item @code{uw1 __MXOR (uw1, uw1)} 6791@tab @code{@var{c} = __MXOR (@var{a}, @var{b})} 6792@tab @code{MXOR @var{a},@var{b},@var{c}} 6793@end multitable 6794 6795@node Raw read/write Functions 6796@subsubsection Raw read/write Functions 6797 6798This sections describes built-in functions related to read and write 6799instructions to access memory. These functions generate 6800@code{membar} instructions to flush the I/O load and stores where 6801appropriate, as described in Fujitsu's manual described above. 6802 6803@table @code 6804 6805@item unsigned char __builtin_read8 (void *@var{data}) 6806@item unsigned short __builtin_read16 (void *@var{data}) 6807@item unsigned long __builtin_read32 (void *@var{data}) 6808@item unsigned long long __builtin_read64 (void *@var{data}) 6809 6810@item void __builtin_write8 (void *@var{data}, unsigned char @var{datum}) 6811@item void __builtin_write16 (void *@var{data}, unsigned short @var{datum}) 6812@item void __builtin_write32 (void *@var{data}, unsigned long @var{datum}) 6813@item void __builtin_write64 (void *@var{data}, unsigned long long @var{datum}) 6814@end table 6815 6816@node Other Built-in Functions 6817@subsubsection Other Built-in Functions 6818 6819This section describes built-in functions that are not named after 6820a specific FR-V instruction. 6821 6822@table @code 6823@item sw2 __IACCreadll (iacc @var{reg}) 6824Return the full 64-bit value of IACC0@. The @var{reg} argument is reserved 6825for future expansion and must be 0. 6826 6827@item sw1 __IACCreadl (iacc @var{reg}) 6828Return the value of IACC0H if @var{reg} is 0 and IACC0L if @var{reg} is 1. 6829Other values of @var{reg} are rejected as invalid. 6830 6831@item void __IACCsetll (iacc @var{reg}, sw2 @var{x}) 6832Set the full 64-bit value of IACC0 to @var{x}. The @var{reg} argument 6833is reserved for future expansion and must be 0. 6834 6835@item void __IACCsetl (iacc @var{reg}, sw1 @var{x}) 6836Set IACC0H to @var{x} if @var{reg} is 0 and IACC0L to @var{x} if @var{reg} 6837is 1. Other values of @var{reg} are rejected as invalid. 6838 6839@item void __data_prefetch0 (const void *@var{x}) 6840Use the @code{dcpl} instruction to load the contents of address @var{x} 6841into the data cache. 6842 6843@item void __data_prefetch (const void *@var{x}) 6844Use the @code{nldub} instruction to load the contents of address @var{x} 6845into the data cache. The instruction will be issued in slot I1@. 6846@end table 6847 6848@node X86 Built-in Functions 6849@subsection X86 Built-in Functions 6850 6851These built-in functions are available for the i386 and x86-64 family 6852of computers, depending on the command-line switches used. 6853 6854Note that, if you specify command-line switches such as @option{-msse}, 6855the compiler could use the extended instruction sets even if the built-ins 6856are not used explicitly in the program. For this reason, applications 6857which perform runtime CPU detection must compile separate files for each 6858supported architecture, using the appropriate flags. In particular, 6859the file containing the CPU detection code should be compiled without 6860these options. 6861 6862The following machine modes are available for use with MMX built-in functions 6863(@pxref{Vector Extensions}): @code{V2SI} for a vector of two 32-bit integers, 6864@code{V4HI} for a vector of four 16-bit integers, and @code{V8QI} for a 6865vector of eight 8-bit integers. Some of the built-in functions operate on 6866MMX registers as a whole 64-bit entity, these use @code{DI} as their mode. 6867 6868If 3Dnow extensions are enabled, @code{V2SF} is used as a mode for a vector 6869of two 32-bit floating point values. 6870 6871If SSE extensions are enabled, @code{V4SF} is used for a vector of four 32-bit 6872floating point values. Some instructions use a vector of four 32-bit 6873integers, these use @code{V4SI}. Finally, some instructions operate on an 6874entire vector register, interpreting it as a 128-bit integer, these use mode 6875@code{TI}. 6876 6877The following built-in functions are made available by @option{-mmmx}. 6878All of them generate the machine instruction that is part of the name. 6879 6880@smallexample 6881v8qi __builtin_ia32_paddb (v8qi, v8qi) 6882v4hi __builtin_ia32_paddw (v4hi, v4hi) 6883v2si __builtin_ia32_paddd (v2si, v2si) 6884v8qi __builtin_ia32_psubb (v8qi, v8qi) 6885v4hi __builtin_ia32_psubw (v4hi, v4hi) 6886v2si __builtin_ia32_psubd (v2si, v2si) 6887v8qi __builtin_ia32_paddsb (v8qi, v8qi) 6888v4hi __builtin_ia32_paddsw (v4hi, v4hi) 6889v8qi __builtin_ia32_psubsb (v8qi, v8qi) 6890v4hi __builtin_ia32_psubsw (v4hi, v4hi) 6891v8qi __builtin_ia32_paddusb (v8qi, v8qi) 6892v4hi __builtin_ia32_paddusw (v4hi, v4hi) 6893v8qi __builtin_ia32_psubusb (v8qi, v8qi) 6894v4hi __builtin_ia32_psubusw (v4hi, v4hi) 6895v4hi __builtin_ia32_pmullw (v4hi, v4hi) 6896v4hi __builtin_ia32_pmulhw (v4hi, v4hi) 6897di __builtin_ia32_pand (di, di) 6898di __builtin_ia32_pandn (di,di) 6899di __builtin_ia32_por (di, di) 6900di __builtin_ia32_pxor (di, di) 6901v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi) 6902v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi) 6903v2si __builtin_ia32_pcmpeqd (v2si, v2si) 6904v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi) 6905v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi) 6906v2si __builtin_ia32_pcmpgtd (v2si, v2si) 6907v8qi __builtin_ia32_punpckhbw (v8qi, v8qi) 6908v4hi __builtin_ia32_punpckhwd (v4hi, v4hi) 6909v2si __builtin_ia32_punpckhdq (v2si, v2si) 6910v8qi __builtin_ia32_punpcklbw (v8qi, v8qi) 6911v4hi __builtin_ia32_punpcklwd (v4hi, v4hi) 6912v2si __builtin_ia32_punpckldq (v2si, v2si) 6913v8qi __builtin_ia32_packsswb (v4hi, v4hi) 6914v4hi __builtin_ia32_packssdw (v2si, v2si) 6915v8qi __builtin_ia32_packuswb (v4hi, v4hi) 6916@end smallexample 6917 6918The following built-in functions are made available either with 6919@option{-msse}, or with a combination of @option{-m3dnow} and 6920@option{-march=athlon}. All of them generate the machine 6921instruction that is part of the name. 6922 6923@smallexample 6924v4hi __builtin_ia32_pmulhuw (v4hi, v4hi) 6925v8qi __builtin_ia32_pavgb (v8qi, v8qi) 6926v4hi __builtin_ia32_pavgw (v4hi, v4hi) 6927v4hi __builtin_ia32_psadbw (v8qi, v8qi) 6928v8qi __builtin_ia32_pmaxub (v8qi, v8qi) 6929v4hi __builtin_ia32_pmaxsw (v4hi, v4hi) 6930v8qi __builtin_ia32_pminub (v8qi, v8qi) 6931v4hi __builtin_ia32_pminsw (v4hi, v4hi) 6932int __builtin_ia32_pextrw (v4hi, int) 6933v4hi __builtin_ia32_pinsrw (v4hi, int, int) 6934int __builtin_ia32_pmovmskb (v8qi) 6935void __builtin_ia32_maskmovq (v8qi, v8qi, char *) 6936void __builtin_ia32_movntq (di *, di) 6937void __builtin_ia32_sfence (void) 6938@end smallexample 6939 6940The following built-in functions are available when @option{-msse} is used. 6941All of them generate the machine instruction that is part of the name. 6942 6943@smallexample 6944int __builtin_ia32_comieq (v4sf, v4sf) 6945int __builtin_ia32_comineq (v4sf, v4sf) 6946int __builtin_ia32_comilt (v4sf, v4sf) 6947int __builtin_ia32_comile (v4sf, v4sf) 6948int __builtin_ia32_comigt (v4sf, v4sf) 6949int __builtin_ia32_comige (v4sf, v4sf) 6950int __builtin_ia32_ucomieq (v4sf, v4sf) 6951int __builtin_ia32_ucomineq (v4sf, v4sf) 6952int __builtin_ia32_ucomilt (v4sf, v4sf) 6953int __builtin_ia32_ucomile (v4sf, v4sf) 6954int __builtin_ia32_ucomigt (v4sf, v4sf) 6955int __builtin_ia32_ucomige (v4sf, v4sf) 6956v4sf __builtin_ia32_addps (v4sf, v4sf) 6957v4sf __builtin_ia32_subps (v4sf, v4sf) 6958v4sf __builtin_ia32_mulps (v4sf, v4sf) 6959v4sf __builtin_ia32_divps (v4sf, v4sf) 6960v4sf __builtin_ia32_addss (v4sf, v4sf) 6961v4sf __builtin_ia32_subss (v4sf, v4sf) 6962v4sf __builtin_ia32_mulss (v4sf, v4sf) 6963v4sf __builtin_ia32_divss (v4sf, v4sf) 6964v4si __builtin_ia32_cmpeqps (v4sf, v4sf) 6965v4si __builtin_ia32_cmpltps (v4sf, v4sf) 6966v4si __builtin_ia32_cmpleps (v4sf, v4sf) 6967v4si __builtin_ia32_cmpgtps (v4sf, v4sf) 6968v4si __builtin_ia32_cmpgeps (v4sf, v4sf) 6969v4si __builtin_ia32_cmpunordps (v4sf, v4sf) 6970v4si __builtin_ia32_cmpneqps (v4sf, v4sf) 6971v4si __builtin_ia32_cmpnltps (v4sf, v4sf) 6972v4si __builtin_ia32_cmpnleps (v4sf, v4sf) 6973v4si __builtin_ia32_cmpngtps (v4sf, v4sf) 6974v4si __builtin_ia32_cmpngeps (v4sf, v4sf) 6975v4si __builtin_ia32_cmpordps (v4sf, v4sf) 6976v4si __builtin_ia32_cmpeqss (v4sf, v4sf) 6977v4si __builtin_ia32_cmpltss (v4sf, v4sf) 6978v4si __builtin_ia32_cmpless (v4sf, v4sf) 6979v4si __builtin_ia32_cmpunordss (v4sf, v4sf) 6980v4si __builtin_ia32_cmpneqss (v4sf, v4sf) 6981v4si __builtin_ia32_cmpnlts (v4sf, v4sf) 6982v4si __builtin_ia32_cmpnless (v4sf, v4sf) 6983v4si __builtin_ia32_cmpordss (v4sf, v4sf) 6984v4sf __builtin_ia32_maxps (v4sf, v4sf) 6985v4sf __builtin_ia32_maxss (v4sf, v4sf) 6986v4sf __builtin_ia32_minps (v4sf, v4sf) 6987v4sf __builtin_ia32_minss (v4sf, v4sf) 6988v4sf __builtin_ia32_andps (v4sf, v4sf) 6989v4sf __builtin_ia32_andnps (v4sf, v4sf) 6990v4sf __builtin_ia32_orps (v4sf, v4sf) 6991v4sf __builtin_ia32_xorps (v4sf, v4sf) 6992v4sf __builtin_ia32_movss (v4sf, v4sf) 6993v4sf __builtin_ia32_movhlps (v4sf, v4sf) 6994v4sf __builtin_ia32_movlhps (v4sf, v4sf) 6995v4sf __builtin_ia32_unpckhps (v4sf, v4sf) 6996v4sf __builtin_ia32_unpcklps (v4sf, v4sf) 6997v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si) 6998v4sf __builtin_ia32_cvtsi2ss (v4sf, int) 6999v2si __builtin_ia32_cvtps2pi (v4sf) 7000int __builtin_ia32_cvtss2si (v4sf) 7001v2si __builtin_ia32_cvttps2pi (v4sf) 7002int __builtin_ia32_cvttss2si (v4sf) 7003v4sf __builtin_ia32_rcpps (v4sf) 7004v4sf __builtin_ia32_rsqrtps (v4sf) 7005v4sf __builtin_ia32_sqrtps (v4sf) 7006v4sf __builtin_ia32_rcpss (v4sf) 7007v4sf __builtin_ia32_rsqrtss (v4sf) 7008v4sf __builtin_ia32_sqrtss (v4sf) 7009v4sf __builtin_ia32_shufps (v4sf, v4sf, int) 7010void __builtin_ia32_movntps (float *, v4sf) 7011int __builtin_ia32_movmskps (v4sf) 7012@end smallexample 7013 7014The following built-in functions are available when @option{-msse} is used. 7015 7016@table @code 7017@item v4sf __builtin_ia32_loadaps (float *) 7018Generates the @code{movaps} machine instruction as a load from memory. 7019@item void __builtin_ia32_storeaps (float *, v4sf) 7020Generates the @code{movaps} machine instruction as a store to memory. 7021@item v4sf __builtin_ia32_loadups (float *) 7022Generates the @code{movups} machine instruction as a load from memory. 7023@item void __builtin_ia32_storeups (float *, v4sf) 7024Generates the @code{movups} machine instruction as a store to memory. 7025@item v4sf __builtin_ia32_loadsss (float *) 7026Generates the @code{movss} machine instruction as a load from memory. 7027@item void __builtin_ia32_storess (float *, v4sf) 7028Generates the @code{movss} machine instruction as a store to memory. 7029@item v4sf __builtin_ia32_loadhps (v4sf, v2si *) 7030Generates the @code{movhps} machine instruction as a load from memory. 7031@item v4sf __builtin_ia32_loadlps (v4sf, v2si *) 7032Generates the @code{movlps} machine instruction as a load from memory 7033@item void __builtin_ia32_storehps (v4sf, v2si *) 7034Generates the @code{movhps} machine instruction as a store to memory. 7035@item void __builtin_ia32_storelps (v4sf, v2si *) 7036Generates the @code{movlps} machine instruction as a store to memory. 7037@end table 7038 7039The following built-in functions are available when @option{-msse2} is used. 7040All of them generate the machine instruction that is part of the name. 7041 7042@smallexample 7043int __builtin_ia32_comisdeq (v2df, v2df) 7044int __builtin_ia32_comisdlt (v2df, v2df) 7045int __builtin_ia32_comisdle (v2df, v2df) 7046int __builtin_ia32_comisdgt (v2df, v2df) 7047int __builtin_ia32_comisdge (v2df, v2df) 7048int __builtin_ia32_comisdneq (v2df, v2df) 7049int __builtin_ia32_ucomisdeq (v2df, v2df) 7050int __builtin_ia32_ucomisdlt (v2df, v2df) 7051int __builtin_ia32_ucomisdle (v2df, v2df) 7052int __builtin_ia32_ucomisdgt (v2df, v2df) 7053int __builtin_ia32_ucomisdge (v2df, v2df) 7054int __builtin_ia32_ucomisdneq (v2df, v2df) 7055v2df __builtin_ia32_cmpeqpd (v2df, v2df) 7056v2df __builtin_ia32_cmpltpd (v2df, v2df) 7057v2df __builtin_ia32_cmplepd (v2df, v2df) 7058v2df __builtin_ia32_cmpgtpd (v2df, v2df) 7059v2df __builtin_ia32_cmpgepd (v2df, v2df) 7060v2df __builtin_ia32_cmpunordpd (v2df, v2df) 7061v2df __builtin_ia32_cmpneqpd (v2df, v2df) 7062v2df __builtin_ia32_cmpnltpd (v2df, v2df) 7063v2df __builtin_ia32_cmpnlepd (v2df, v2df) 7064v2df __builtin_ia32_cmpngtpd (v2df, v2df) 7065v2df __builtin_ia32_cmpngepd (v2df, v2df) 7066v2df __builtin_ia32_cmpordpd (v2df, v2df) 7067v2df __builtin_ia32_cmpeqsd (v2df, v2df) 7068v2df __builtin_ia32_cmpltsd (v2df, v2df) 7069v2df __builtin_ia32_cmplesd (v2df, v2df) 7070v2df __builtin_ia32_cmpunordsd (v2df, v2df) 7071v2df __builtin_ia32_cmpneqsd (v2df, v2df) 7072v2df __builtin_ia32_cmpnltsd (v2df, v2df) 7073v2df __builtin_ia32_cmpnlesd (v2df, v2df) 7074v2df __builtin_ia32_cmpordsd (v2df, v2df) 7075v2di __builtin_ia32_paddq (v2di, v2di) 7076v2di __builtin_ia32_psubq (v2di, v2di) 7077v2df __builtin_ia32_addpd (v2df, v2df) 7078v2df __builtin_ia32_subpd (v2df, v2df) 7079v2df __builtin_ia32_mulpd (v2df, v2df) 7080v2df __builtin_ia32_divpd (v2df, v2df) 7081v2df __builtin_ia32_addsd (v2df, v2df) 7082v2df __builtin_ia32_subsd (v2df, v2df) 7083v2df __builtin_ia32_mulsd (v2df, v2df) 7084v2df __builtin_ia32_divsd (v2df, v2df) 7085v2df __builtin_ia32_minpd (v2df, v2df) 7086v2df __builtin_ia32_maxpd (v2df, v2df) 7087v2df __builtin_ia32_minsd (v2df, v2df) 7088v2df __builtin_ia32_maxsd (v2df, v2df) 7089v2df __builtin_ia32_andpd (v2df, v2df) 7090v2df __builtin_ia32_andnpd (v2df, v2df) 7091v2df __builtin_ia32_orpd (v2df, v2df) 7092v2df __builtin_ia32_xorpd (v2df, v2df) 7093v2df __builtin_ia32_movsd (v2df, v2df) 7094v2df __builtin_ia32_unpckhpd (v2df, v2df) 7095v2df __builtin_ia32_unpcklpd (v2df, v2df) 7096v16qi __builtin_ia32_paddb128 (v16qi, v16qi) 7097v8hi __builtin_ia32_paddw128 (v8hi, v8hi) 7098v4si __builtin_ia32_paddd128 (v4si, v4si) 7099v2di __builtin_ia32_paddq128 (v2di, v2di) 7100v16qi __builtin_ia32_psubb128 (v16qi, v16qi) 7101v8hi __builtin_ia32_psubw128 (v8hi, v8hi) 7102v4si __builtin_ia32_psubd128 (v4si, v4si) 7103v2di __builtin_ia32_psubq128 (v2di, v2di) 7104v8hi __builtin_ia32_pmullw128 (v8hi, v8hi) 7105v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi) 7106v2di __builtin_ia32_pand128 (v2di, v2di) 7107v2di __builtin_ia32_pandn128 (v2di, v2di) 7108v2di __builtin_ia32_por128 (v2di, v2di) 7109v2di __builtin_ia32_pxor128 (v2di, v2di) 7110v16qi __builtin_ia32_pavgb128 (v16qi, v16qi) 7111v8hi __builtin_ia32_pavgw128 (v8hi, v8hi) 7112v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi) 7113v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi) 7114v4si __builtin_ia32_pcmpeqd128 (v4si, v4si) 7115v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi) 7116v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi) 7117v4si __builtin_ia32_pcmpgtd128 (v4si, v4si) 7118v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi) 7119v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi) 7120v16qi __builtin_ia32_pminub128 (v16qi, v16qi) 7121v8hi __builtin_ia32_pminsw128 (v8hi, v8hi) 7122v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi) 7123v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi) 7124v4si __builtin_ia32_punpckhdq128 (v4si, v4si) 7125v2di __builtin_ia32_punpckhqdq128 (v2di, v2di) 7126v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi) 7127v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi) 7128v4si __builtin_ia32_punpckldq128 (v4si, v4si) 7129v2di __builtin_ia32_punpcklqdq128 (v2di, v2di) 7130v16qi __builtin_ia32_packsswb128 (v16qi, v16qi) 7131v8hi __builtin_ia32_packssdw128 (v8hi, v8hi) 7132v16qi __builtin_ia32_packuswb128 (v16qi, v16qi) 7133v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi) 7134void __builtin_ia32_maskmovdqu (v16qi, v16qi) 7135v2df __builtin_ia32_loadupd (double *) 7136void __builtin_ia32_storeupd (double *, v2df) 7137v2df __builtin_ia32_loadhpd (v2df, double *) 7138v2df __builtin_ia32_loadlpd (v2df, double *) 7139int __builtin_ia32_movmskpd (v2df) 7140int __builtin_ia32_pmovmskb128 (v16qi) 7141void __builtin_ia32_movnti (int *, int) 7142void __builtin_ia32_movntpd (double *, v2df) 7143void __builtin_ia32_movntdq (v2df *, v2df) 7144v4si __builtin_ia32_pshufd (v4si, int) 7145v8hi __builtin_ia32_pshuflw (v8hi, int) 7146v8hi __builtin_ia32_pshufhw (v8hi, int) 7147v2di __builtin_ia32_psadbw128 (v16qi, v16qi) 7148v2df __builtin_ia32_sqrtpd (v2df) 7149v2df __builtin_ia32_sqrtsd (v2df) 7150v2df __builtin_ia32_shufpd (v2df, v2df, int) 7151v2df __builtin_ia32_cvtdq2pd (v4si) 7152v4sf __builtin_ia32_cvtdq2ps (v4si) 7153v4si __builtin_ia32_cvtpd2dq (v2df) 7154v2si __builtin_ia32_cvtpd2pi (v2df) 7155v4sf __builtin_ia32_cvtpd2ps (v2df) 7156v4si __builtin_ia32_cvttpd2dq (v2df) 7157v2si __builtin_ia32_cvttpd2pi (v2df) 7158v2df __builtin_ia32_cvtpi2pd (v2si) 7159int __builtin_ia32_cvtsd2si (v2df) 7160int __builtin_ia32_cvttsd2si (v2df) 7161long long __builtin_ia32_cvtsd2si64 (v2df) 7162long long __builtin_ia32_cvttsd2si64 (v2df) 7163v4si __builtin_ia32_cvtps2dq (v4sf) 7164v2df __builtin_ia32_cvtps2pd (v4sf) 7165v4si __builtin_ia32_cvttps2dq (v4sf) 7166v2df __builtin_ia32_cvtsi2sd (v2df, int) 7167v2df __builtin_ia32_cvtsi642sd (v2df, long long) 7168v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df) 7169v2df __builtin_ia32_cvtss2sd (v2df, v4sf) 7170void __builtin_ia32_clflush (const void *) 7171void __builtin_ia32_lfence (void) 7172void __builtin_ia32_mfence (void) 7173v16qi __builtin_ia32_loaddqu (const char *) 7174void __builtin_ia32_storedqu (char *, v16qi) 7175unsigned long long __builtin_ia32_pmuludq (v2si, v2si) 7176v2di __builtin_ia32_pmuludq128 (v4si, v4si) 7177v8hi __builtin_ia32_psllw128 (v8hi, v2di) 7178v4si __builtin_ia32_pslld128 (v4si, v2di) 7179v2di __builtin_ia32_psllq128 (v4si, v2di) 7180v8hi __builtin_ia32_psrlw128 (v8hi, v2di) 7181v4si __builtin_ia32_psrld128 (v4si, v2di) 7182v2di __builtin_ia32_psrlq128 (v2di, v2di) 7183v8hi __builtin_ia32_psraw128 (v8hi, v2di) 7184v4si __builtin_ia32_psrad128 (v4si, v2di) 7185v2di __builtin_ia32_pslldqi128 (v2di, int) 7186v8hi __builtin_ia32_psllwi128 (v8hi, int) 7187v4si __builtin_ia32_pslldi128 (v4si, int) 7188v2di __builtin_ia32_psllqi128 (v2di, int) 7189v2di __builtin_ia32_psrldqi128 (v2di, int) 7190v8hi __builtin_ia32_psrlwi128 (v8hi, int) 7191v4si __builtin_ia32_psrldi128 (v4si, int) 7192v2di __builtin_ia32_psrlqi128 (v2di, int) 7193v8hi __builtin_ia32_psrawi128 (v8hi, int) 7194v4si __builtin_ia32_psradi128 (v4si, int) 7195v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi) 7196@end smallexample 7197 7198The following built-in functions are available when @option{-msse3} is used. 7199All of them generate the machine instruction that is part of the name. 7200 7201@smallexample 7202v2df __builtin_ia32_addsubpd (v2df, v2df) 7203v4sf __builtin_ia32_addsubps (v4sf, v4sf) 7204v2df __builtin_ia32_haddpd (v2df, v2df) 7205v4sf __builtin_ia32_haddps (v4sf, v4sf) 7206v2df __builtin_ia32_hsubpd (v2df, v2df) 7207v4sf __builtin_ia32_hsubps (v4sf, v4sf) 7208v16qi __builtin_ia32_lddqu (char const *) 7209void __builtin_ia32_monitor (void *, unsigned int, unsigned int) 7210v2df __builtin_ia32_movddup (v2df) 7211v4sf __builtin_ia32_movshdup (v4sf) 7212v4sf __builtin_ia32_movsldup (v4sf) 7213void __builtin_ia32_mwait (unsigned int, unsigned int) 7214@end smallexample 7215 7216The following built-in functions are available when @option{-msse3} is used. 7217 7218@table @code 7219@item v2df __builtin_ia32_loadddup (double const *) 7220Generates the @code{movddup} machine instruction as a load from memory. 7221@end table 7222 7223The following built-in functions are available when @option{-mssse3} is used. 7224All of them generate the machine instruction that is part of the name 7225with MMX registers. 7226 7227@smallexample 7228v2si __builtin_ia32_phaddd (v2si, v2si) 7229v4hi __builtin_ia32_phaddw (v4hi, v4hi) 7230v4hi __builtin_ia32_phaddsw (v4hi, v4hi) 7231v2si __builtin_ia32_phsubd (v2si, v2si) 7232v4hi __builtin_ia32_phsubw (v4hi, v4hi) 7233v4hi __builtin_ia32_phsubsw (v4hi, v4hi) 7234v8qi __builtin_ia32_pmaddubsw (v8qi, v8qi) 7235v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi) 7236v8qi __builtin_ia32_pshufb (v8qi, v8qi) 7237v8qi __builtin_ia32_psignb (v8qi, v8qi) 7238v2si __builtin_ia32_psignd (v2si, v2si) 7239v4hi __builtin_ia32_psignw (v4hi, v4hi) 7240long long __builtin_ia32_palignr (long long, long long, int) 7241v8qi __builtin_ia32_pabsb (v8qi) 7242v2si __builtin_ia32_pabsd (v2si) 7243v4hi __builtin_ia32_pabsw (v4hi) 7244@end smallexample 7245 7246The following built-in functions are available when @option{-mssse3} is used. 7247All of them generate the machine instruction that is part of the name 7248with SSE registers. 7249 7250@smallexample 7251v4si __builtin_ia32_phaddd128 (v4si, v4si) 7252v8hi __builtin_ia32_phaddw128 (v8hi, v8hi) 7253v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi) 7254v4si __builtin_ia32_phsubd128 (v4si, v4si) 7255v8hi __builtin_ia32_phsubw128 (v8hi, v8hi) 7256v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi) 7257v16qi __builtin_ia32_pmaddubsw128 (v16qi, v16qi) 7258v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi) 7259v16qi __builtin_ia32_pshufb128 (v16qi, v16qi) 7260v16qi __builtin_ia32_psignb128 (v16qi, v16qi) 7261v4si __builtin_ia32_psignd128 (v4si, v4si) 7262v8hi __builtin_ia32_psignw128 (v8hi, v8hi) 7263v2di __builtin_ia32_palignr (v2di, v2di, int) 7264v16qi __builtin_ia32_pabsb128 (v16qi) 7265v4si __builtin_ia32_pabsd128 (v4si) 7266v8hi __builtin_ia32_pabsw128 (v8hi) 7267@end smallexample 7268 7269The following built-in functions are available when @option{-msse4a} is used. 7270 7271@smallexample 7272void _mm_stream_sd (double*,__m128d); 7273Generates the @code{movntsd} machine instruction. 7274void _mm_stream_ss (float*,__m128); 7275Generates the @code{movntss} machine instruction. 7276__m128i _mm_extract_si64 (__m128i, __m128i); 7277Generates the @code{extrq} machine instruction with only SSE register operands. 7278__m128i _mm_extracti_si64 (__m128i, int, int); 7279Generates the @code{extrq} machine instruction with SSE register and immediate operands. 7280__m128i _mm_insert_si64 (__m128i, __m128i); 7281Generates the @code{insertq} machine instruction with only SSE register operands. 7282__m128i _mm_inserti_si64 (__m128i, __m128i, int, int); 7283Generates the @code{insertq} machine instruction with SSE register and immediate operands. 7284@end smallexample 7285 7286The following built-in functions are available when @option{-m3dnow} is used. 7287All of them generate the machine instruction that is part of the name. 7288 7289@smallexample 7290void __builtin_ia32_femms (void) 7291v8qi __builtin_ia32_pavgusb (v8qi, v8qi) 7292v2si __builtin_ia32_pf2id (v2sf) 7293v2sf __builtin_ia32_pfacc (v2sf, v2sf) 7294v2sf __builtin_ia32_pfadd (v2sf, v2sf) 7295v2si __builtin_ia32_pfcmpeq (v2sf, v2sf) 7296v2si __builtin_ia32_pfcmpge (v2sf, v2sf) 7297v2si __builtin_ia32_pfcmpgt (v2sf, v2sf) 7298v2sf __builtin_ia32_pfmax (v2sf, v2sf) 7299v2sf __builtin_ia32_pfmin (v2sf, v2sf) 7300v2sf __builtin_ia32_pfmul (v2sf, v2sf) 7301v2sf __builtin_ia32_pfrcp (v2sf) 7302v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf) 7303v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf) 7304v2sf __builtin_ia32_pfrsqrt (v2sf) 7305v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf) 7306v2sf __builtin_ia32_pfsub (v2sf, v2sf) 7307v2sf __builtin_ia32_pfsubr (v2sf, v2sf) 7308v2sf __builtin_ia32_pi2fd (v2si) 7309v4hi __builtin_ia32_pmulhrw (v4hi, v4hi) 7310@end smallexample 7311 7312The following built-in functions are available when both @option{-m3dnow} 7313and @option{-march=athlon} are used. All of them generate the machine 7314instruction that is part of the name. 7315 7316@smallexample 7317v2si __builtin_ia32_pf2iw (v2sf) 7318v2sf __builtin_ia32_pfnacc (v2sf, v2sf) 7319v2sf __builtin_ia32_pfpnacc (v2sf, v2sf) 7320v2sf __builtin_ia32_pi2fw (v2si) 7321v2sf __builtin_ia32_pswapdsf (v2sf) 7322v2si __builtin_ia32_pswapdsi (v2si) 7323@end smallexample 7324 7325@node MIPS DSP Built-in Functions 7326@subsection MIPS DSP Built-in Functions 7327 7328The MIPS DSP Application-Specific Extension (ASE) includes new 7329instructions that are designed to improve the performance of DSP and 7330media applications. It provides instructions that operate on packed 73318-bit integer data, Q15 fractional data and Q31 fractional data. 7332 7333GCC supports MIPS DSP operations using both the generic 7334vector extensions (@pxref{Vector Extensions}) and a collection of 7335MIPS-specific built-in functions. Both kinds of support are 7336enabled by the @option{-mdsp} command-line option. 7337 7338At present, GCC only provides support for operations on 32-bit 7339vectors. The vector type associated with 8-bit integer data is 7340usually called @code{v4i8} and the vector type associated with Q15 is 7341usually called @code{v2q15}. They can be defined in C as follows: 7342 7343@smallexample 7344typedef char v4i8 __attribute__ ((vector_size(4))); 7345typedef short v2q15 __attribute__ ((vector_size(4))); 7346@end smallexample 7347 7348@code{v4i8} and @code{v2q15} values are initialized in the same way as 7349aggregates. For example: 7350 7351@smallexample 7352v4i8 a = @{1, 2, 3, 4@}; 7353v4i8 b; 7354b = (v4i8) @{5, 6, 7, 8@}; 7355 7356v2q15 c = @{0x0fcb, 0x3a75@}; 7357v2q15 d; 7358d = (v2q15) @{0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15@}; 7359@end smallexample 7360 7361@emph{Note:} The CPU's endianness determines the order in which values 7362are packed. On little-endian targets, the first value is the least 7363significant and the last value is the most significant. The opposite 7364order applies to big-endian targets. For example, the code above will 7365set the lowest byte of @code{a} to @code{1} on little-endian targets 7366and @code{4} on big-endian targets. 7367 7368@emph{Note:} Q15 and Q31 values must be initialized with their integer 7369representation. As shown in this example, the integer representation 7370of a Q15 value can be obtained by multiplying the fractional value by 7371@code{0x1.0p15}. The equivalent for Q31 values is to multiply by 7372@code{0x1.0p31}. 7373 7374The table below lists the @code{v4i8} and @code{v2q15} operations for which 7375hardware support exists. @code{a} and @code{b} are @code{v4i8} values, 7376and @code{c} and @code{d} are @code{v2q15} values. 7377 7378@multitable @columnfractions .50 .50 7379@item C code @tab MIPS instruction 7380@item @code{a + b} @tab @code{addu.qb} 7381@item @code{c + d} @tab @code{addq.ph} 7382@item @code{a - b} @tab @code{subu.qb} 7383@item @code{c - d} @tab @code{subq.ph} 7384@end multitable 7385 7386It is easier to describe the DSP built-in functions if we first define 7387the following types: 7388 7389@smallexample 7390typedef int q31; 7391typedef int i32; 7392typedef long long a64; 7393@end smallexample 7394 7395@code{q31} and @code{i32} are actually the same as @code{int}, but we 7396use @code{q31} to indicate a Q31 fractional value and @code{i32} to 7397indicate a 32-bit integer value. Similarly, @code{a64} is the same as 7398@code{long long}, but we use @code{a64} to indicate values that will 7399be placed in one of the four DSP accumulators (@code{$ac0}, 7400@code{$ac1}, @code{$ac2} or @code{$ac3}). 7401 7402Also, some built-in functions prefer or require immediate numbers as 7403parameters, because the corresponding DSP instructions accept both immediate 7404numbers and register operands, or accept immediate numbers only. The 7405immediate parameters are listed as follows. 7406 7407@smallexample 7408imm0_7: 0 to 7. 7409imm0_15: 0 to 15. 7410imm0_31: 0 to 31. 7411imm0_63: 0 to 63. 7412imm0_255: 0 to 255. 7413imm_n32_31: -32 to 31. 7414imm_n512_511: -512 to 511. 7415@end smallexample 7416 7417The following built-in functions map directly to a particular MIPS DSP 7418instruction. Please refer to the architecture specification 7419for details on what each instruction does. 7420 7421@smallexample 7422v2q15 __builtin_mips_addq_ph (v2q15, v2q15) 7423v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15) 7424q31 __builtin_mips_addq_s_w (q31, q31) 7425v4i8 __builtin_mips_addu_qb (v4i8, v4i8) 7426v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8) 7427v2q15 __builtin_mips_subq_ph (v2q15, v2q15) 7428v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15) 7429q31 __builtin_mips_subq_s_w (q31, q31) 7430v4i8 __builtin_mips_subu_qb (v4i8, v4i8) 7431v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8) 7432i32 __builtin_mips_addsc (i32, i32) 7433i32 __builtin_mips_addwc (i32, i32) 7434i32 __builtin_mips_modsub (i32, i32) 7435i32 __builtin_mips_raddu_w_qb (v4i8) 7436v2q15 __builtin_mips_absq_s_ph (v2q15) 7437q31 __builtin_mips_absq_s_w (q31) 7438v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15) 7439v2q15 __builtin_mips_precrq_ph_w (q31, q31) 7440v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31) 7441v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15) 7442q31 __builtin_mips_preceq_w_phl (v2q15) 7443q31 __builtin_mips_preceq_w_phr (v2q15) 7444v2q15 __builtin_mips_precequ_ph_qbl (v4i8) 7445v2q15 __builtin_mips_precequ_ph_qbr (v4i8) 7446v2q15 __builtin_mips_precequ_ph_qbla (v4i8) 7447v2q15 __builtin_mips_precequ_ph_qbra (v4i8) 7448v2q15 __builtin_mips_preceu_ph_qbl (v4i8) 7449v2q15 __builtin_mips_preceu_ph_qbr (v4i8) 7450v2q15 __builtin_mips_preceu_ph_qbla (v4i8) 7451v2q15 __builtin_mips_preceu_ph_qbra (v4i8) 7452v4i8 __builtin_mips_shll_qb (v4i8, imm0_7) 7453v4i8 __builtin_mips_shll_qb (v4i8, i32) 7454v2q15 __builtin_mips_shll_ph (v2q15, imm0_15) 7455v2q15 __builtin_mips_shll_ph (v2q15, i32) 7456v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15) 7457v2q15 __builtin_mips_shll_s_ph (v2q15, i32) 7458q31 __builtin_mips_shll_s_w (q31, imm0_31) 7459q31 __builtin_mips_shll_s_w (q31, i32) 7460v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7) 7461v4i8 __builtin_mips_shrl_qb (v4i8, i32) 7462v2q15 __builtin_mips_shra_ph (v2q15, imm0_15) 7463v2q15 __builtin_mips_shra_ph (v2q15, i32) 7464v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15) 7465v2q15 __builtin_mips_shra_r_ph (v2q15, i32) 7466q31 __builtin_mips_shra_r_w (q31, imm0_31) 7467q31 __builtin_mips_shra_r_w (q31, i32) 7468v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15) 7469v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15) 7470v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15) 7471q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15) 7472q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15) 7473a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8) 7474a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8) 7475a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8) 7476a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8) 7477a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15) 7478a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31) 7479a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15) 7480a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31) 7481a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15) 7482a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15) 7483a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15) 7484a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15) 7485a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15) 7486i32 __builtin_mips_bitrev (i32) 7487i32 __builtin_mips_insv (i32, i32) 7488v4i8 __builtin_mips_repl_qb (imm0_255) 7489v4i8 __builtin_mips_repl_qb (i32) 7490v2q15 __builtin_mips_repl_ph (imm_n512_511) 7491v2q15 __builtin_mips_repl_ph (i32) 7492void __builtin_mips_cmpu_eq_qb (v4i8, v4i8) 7493void __builtin_mips_cmpu_lt_qb (v4i8, v4i8) 7494void __builtin_mips_cmpu_le_qb (v4i8, v4i8) 7495i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8) 7496i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8) 7497i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8) 7498void __builtin_mips_cmp_eq_ph (v2q15, v2q15) 7499void __builtin_mips_cmp_lt_ph (v2q15, v2q15) 7500void __builtin_mips_cmp_le_ph (v2q15, v2q15) 7501v4i8 __builtin_mips_pick_qb (v4i8, v4i8) 7502v2q15 __builtin_mips_pick_ph (v2q15, v2q15) 7503v2q15 __builtin_mips_packrl_ph (v2q15, v2q15) 7504i32 __builtin_mips_extr_w (a64, imm0_31) 7505i32 __builtin_mips_extr_w (a64, i32) 7506i32 __builtin_mips_extr_r_w (a64, imm0_31) 7507i32 __builtin_mips_extr_s_h (a64, i32) 7508i32 __builtin_mips_extr_rs_w (a64, imm0_31) 7509i32 __builtin_mips_extr_rs_w (a64, i32) 7510i32 __builtin_mips_extr_s_h (a64, imm0_31) 7511i32 __builtin_mips_extr_r_w (a64, i32) 7512i32 __builtin_mips_extp (a64, imm0_31) 7513i32 __builtin_mips_extp (a64, i32) 7514i32 __builtin_mips_extpdp (a64, imm0_31) 7515i32 __builtin_mips_extpdp (a64, i32) 7516a64 __builtin_mips_shilo (a64, imm_n32_31) 7517a64 __builtin_mips_shilo (a64, i32) 7518a64 __builtin_mips_mthlip (a64, i32) 7519void __builtin_mips_wrdsp (i32, imm0_63) 7520i32 __builtin_mips_rddsp (imm0_63) 7521i32 __builtin_mips_lbux (void *, i32) 7522i32 __builtin_mips_lhx (void *, i32) 7523i32 __builtin_mips_lwx (void *, i32) 7524i32 __builtin_mips_bposge32 (void) 7525@end smallexample 7526 7527@node MIPS Paired-Single Support 7528@subsection MIPS Paired-Single Support 7529 7530The MIPS64 architecture includes a number of instructions that 7531operate on pairs of single-precision floating-point values. 7532Each pair is packed into a 64-bit floating-point register, 7533with one element being designated the ``upper half'' and 7534the other being designated the ``lower half''. 7535 7536GCC supports paired-single operations using both the generic 7537vector extensions (@pxref{Vector Extensions}) and a collection of 7538MIPS-specific built-in functions. Both kinds of support are 7539enabled by the @option{-mpaired-single} command-line option. 7540 7541The vector type associated with paired-single values is usually 7542called @code{v2sf}. It can be defined in C as follows: 7543 7544@smallexample 7545typedef float v2sf __attribute__ ((vector_size (8))); 7546@end smallexample 7547 7548@code{v2sf} values are initialized in the same way as aggregates. 7549For example: 7550 7551@smallexample 7552v2sf a = @{1.5, 9.1@}; 7553v2sf b; 7554float e, f; 7555b = (v2sf) @{e, f@}; 7556@end smallexample 7557 7558@emph{Note:} The CPU's endianness determines which value is stored in 7559the upper half of a register and which value is stored in the lower half. 7560On little-endian targets, the first value is the lower one and the second 7561value is the upper one. The opposite order applies to big-endian targets. 7562For example, the code above will set the lower half of @code{a} to 7563@code{1.5} on little-endian targets and @code{9.1} on big-endian targets. 7564 7565@menu 7566* Paired-Single Arithmetic:: 7567* Paired-Single Built-in Functions:: 7568* MIPS-3D Built-in Functions:: 7569@end menu 7570 7571@node Paired-Single Arithmetic 7572@subsubsection Paired-Single Arithmetic 7573 7574The table below lists the @code{v2sf} operations for which hardware 7575support exists. @code{a}, @code{b} and @code{c} are @code{v2sf} 7576values and @code{x} is an integral value. 7577 7578@multitable @columnfractions .50 .50 7579@item C code @tab MIPS instruction 7580@item @code{a + b} @tab @code{add.ps} 7581@item @code{a - b} @tab @code{sub.ps} 7582@item @code{-a} @tab @code{neg.ps} 7583@item @code{a * b} @tab @code{mul.ps} 7584@item @code{a * b + c} @tab @code{madd.ps} 7585@item @code{a * b - c} @tab @code{msub.ps} 7586@item @code{-(a * b + c)} @tab @code{nmadd.ps} 7587@item @code{-(a * b - c)} @tab @code{nmsub.ps} 7588@item @code{x ? a : b} @tab @code{movn.ps}/@code{movz.ps} 7589@end multitable 7590 7591Note that the multiply-accumulate instructions can be disabled 7592using the command-line option @code{-mno-fused-madd}. 7593 7594@node Paired-Single Built-in Functions 7595@subsubsection Paired-Single Built-in Functions 7596 7597The following paired-single functions map directly to a particular 7598MIPS instruction. Please refer to the architecture specification 7599for details on what each instruction does. 7600 7601@table @code 7602@item v2sf __builtin_mips_pll_ps (v2sf, v2sf) 7603Pair lower lower (@code{pll.ps}). 7604 7605@item v2sf __builtin_mips_pul_ps (v2sf, v2sf) 7606Pair upper lower (@code{pul.ps}). 7607 7608@item v2sf __builtin_mips_plu_ps (v2sf, v2sf) 7609Pair lower upper (@code{plu.ps}). 7610 7611@item v2sf __builtin_mips_puu_ps (v2sf, v2sf) 7612Pair upper upper (@code{puu.ps}). 7613 7614@item v2sf __builtin_mips_cvt_ps_s (float, float) 7615Convert pair to paired single (@code{cvt.ps.s}). 7616 7617@item float __builtin_mips_cvt_s_pl (v2sf) 7618Convert pair lower to single (@code{cvt.s.pl}). 7619 7620@item float __builtin_mips_cvt_s_pu (v2sf) 7621Convert pair upper to single (@code{cvt.s.pu}). 7622 7623@item v2sf __builtin_mips_abs_ps (v2sf) 7624Absolute value (@code{abs.ps}). 7625 7626@item v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int) 7627Align variable (@code{alnv.ps}). 7628 7629@emph{Note:} The value of the third parameter must be 0 or 4 7630modulo 8, otherwise the result will be unpredictable. Please read the 7631instruction description for details. 7632@end table 7633 7634The following multi-instruction functions are also available. 7635In each case, @var{cond} can be any of the 16 floating-point conditions: 7636@code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult}, 7637@code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, @code{ngl}, 7638@code{lt}, @code{nge}, @code{le} or @code{ngt}. 7639 7640@table @code 7641@item v2sf __builtin_mips_movt_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7642@itemx v2sf __builtin_mips_movf_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7643Conditional move based on floating point comparison (@code{c.@var{cond}.ps}, 7644@code{movt.ps}/@code{movf.ps}). 7645 7646The @code{movt} functions return the value @var{x} computed by: 7647 7648@smallexample 7649c.@var{cond}.ps @var{cc},@var{a},@var{b} 7650mov.ps @var{x},@var{c} 7651movt.ps @var{x},@var{d},@var{cc} 7652@end smallexample 7653 7654The @code{movf} functions are similar but use @code{movf.ps} instead 7655of @code{movt.ps}. 7656 7657@item int __builtin_mips_upper_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7658@itemx int __builtin_mips_lower_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7659Comparison of two paired-single values (@code{c.@var{cond}.ps}, 7660@code{bc1t}/@code{bc1f}). 7661 7662These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps} 7663and return either the upper or lower half of the result. For example: 7664 7665@smallexample 7666v2sf a, b; 7667if (__builtin_mips_upper_c_eq_ps (a, b)) 7668 upper_halves_are_equal (); 7669else 7670 upper_halves_are_unequal (); 7671 7672if (__builtin_mips_lower_c_eq_ps (a, b)) 7673 lower_halves_are_equal (); 7674else 7675 lower_halves_are_unequal (); 7676@end smallexample 7677@end table 7678 7679@node MIPS-3D Built-in Functions 7680@subsubsection MIPS-3D Built-in Functions 7681 7682The MIPS-3D Application-Specific Extension (ASE) includes additional 7683paired-single instructions that are designed to improve the performance 7684of 3D graphics operations. Support for these instructions is controlled 7685by the @option{-mips3d} command-line option. 7686 7687The functions listed below map directly to a particular MIPS-3D 7688instruction. Please refer to the architecture specification for 7689more details on what each instruction does. 7690 7691@table @code 7692@item v2sf __builtin_mips_addr_ps (v2sf, v2sf) 7693Reduction add (@code{addr.ps}). 7694 7695@item v2sf __builtin_mips_mulr_ps (v2sf, v2sf) 7696Reduction multiply (@code{mulr.ps}). 7697 7698@item v2sf __builtin_mips_cvt_pw_ps (v2sf) 7699Convert paired single to paired word (@code{cvt.pw.ps}). 7700 7701@item v2sf __builtin_mips_cvt_ps_pw (v2sf) 7702Convert paired word to paired single (@code{cvt.ps.pw}). 7703 7704@item float __builtin_mips_recip1_s (float) 7705@itemx double __builtin_mips_recip1_d (double) 7706@itemx v2sf __builtin_mips_recip1_ps (v2sf) 7707Reduced precision reciprocal (sequence step 1) (@code{recip1.@var{fmt}}). 7708 7709@item float __builtin_mips_recip2_s (float, float) 7710@itemx double __builtin_mips_recip2_d (double, double) 7711@itemx v2sf __builtin_mips_recip2_ps (v2sf, v2sf) 7712Reduced precision reciprocal (sequence step 2) (@code{recip2.@var{fmt}}). 7713 7714@item float __builtin_mips_rsqrt1_s (float) 7715@itemx double __builtin_mips_rsqrt1_d (double) 7716@itemx v2sf __builtin_mips_rsqrt1_ps (v2sf) 7717Reduced precision reciprocal square root (sequence step 1) 7718(@code{rsqrt1.@var{fmt}}). 7719 7720@item float __builtin_mips_rsqrt2_s (float, float) 7721@itemx double __builtin_mips_rsqrt2_d (double, double) 7722@itemx v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf) 7723Reduced precision reciprocal square root (sequence step 2) 7724(@code{rsqrt2.@var{fmt}}). 7725@end table 7726 7727The following multi-instruction functions are also available. 7728In each case, @var{cond} can be any of the 16 floating-point conditions: 7729@code{f}, @code{un}, @code{eq}, @code{ueq}, @code{olt}, @code{ult}, 7730@code{ole}, @code{ule}, @code{sf}, @code{ngle}, @code{seq}, 7731@code{ngl}, @code{lt}, @code{nge}, @code{le} or @code{ngt}. 7732 7733@table @code 7734@item int __builtin_mips_cabs_@var{cond}_s (float @var{a}, float @var{b}) 7735@itemx int __builtin_mips_cabs_@var{cond}_d (double @var{a}, double @var{b}) 7736Absolute comparison of two scalar values (@code{cabs.@var{cond}.@var{fmt}}, 7737@code{bc1t}/@code{bc1f}). 7738 7739These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.s} 7740or @code{cabs.@var{cond}.d} and return the result as a boolean value. 7741For example: 7742 7743@smallexample 7744float a, b; 7745if (__builtin_mips_cabs_eq_s (a, b)) 7746 true (); 7747else 7748 false (); 7749@end smallexample 7750 7751@item int __builtin_mips_upper_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7752@itemx int __builtin_mips_lower_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7753Absolute comparison of two paired-single values (@code{cabs.@var{cond}.ps}, 7754@code{bc1t}/@code{bc1f}). 7755 7756These functions compare @var{a} and @var{b} using @code{cabs.@var{cond}.ps} 7757and return either the upper or lower half of the result. For example: 7758 7759@smallexample 7760v2sf a, b; 7761if (__builtin_mips_upper_cabs_eq_ps (a, b)) 7762 upper_halves_are_equal (); 7763else 7764 upper_halves_are_unequal (); 7765 7766if (__builtin_mips_lower_cabs_eq_ps (a, b)) 7767 lower_halves_are_equal (); 7768else 7769 lower_halves_are_unequal (); 7770@end smallexample 7771 7772@item v2sf __builtin_mips_movt_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7773@itemx v2sf __builtin_mips_movf_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7774Conditional move based on absolute comparison (@code{cabs.@var{cond}.ps}, 7775@code{movt.ps}/@code{movf.ps}). 7776 7777The @code{movt} functions return the value @var{x} computed by: 7778 7779@smallexample 7780cabs.@var{cond}.ps @var{cc},@var{a},@var{b} 7781mov.ps @var{x},@var{c} 7782movt.ps @var{x},@var{d},@var{cc} 7783@end smallexample 7784 7785The @code{movf} functions are similar but use @code{movf.ps} instead 7786of @code{movt.ps}. 7787 7788@item int __builtin_mips_any_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7789@itemx int __builtin_mips_all_c_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7790@itemx int __builtin_mips_any_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7791@itemx int __builtin_mips_all_cabs_@var{cond}_ps (v2sf @var{a}, v2sf @var{b}) 7792Comparison of two paired-single values 7793(@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps}, 7794@code{bc1any2t}/@code{bc1any2f}). 7795 7796These functions compare @var{a} and @var{b} using @code{c.@var{cond}.ps} 7797or @code{cabs.@var{cond}.ps}. The @code{any} forms return true if either 7798result is true and the @code{all} forms return true if both results are true. 7799For example: 7800 7801@smallexample 7802v2sf a, b; 7803if (__builtin_mips_any_c_eq_ps (a, b)) 7804 one_is_true (); 7805else 7806 both_are_false (); 7807 7808if (__builtin_mips_all_c_eq_ps (a, b)) 7809 both_are_true (); 7810else 7811 one_is_false (); 7812@end smallexample 7813 7814@item int __builtin_mips_any_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7815@itemx int __builtin_mips_all_c_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7816@itemx int __builtin_mips_any_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7817@itemx int __builtin_mips_all_cabs_@var{cond}_4s (v2sf @var{a}, v2sf @var{b}, v2sf @var{c}, v2sf @var{d}) 7818Comparison of four paired-single values 7819(@code{c.@var{cond}.ps}/@code{cabs.@var{cond}.ps}, 7820@code{bc1any4t}/@code{bc1any4f}). 7821 7822These functions use @code{c.@var{cond}.ps} or @code{cabs.@var{cond}.ps} 7823to compare @var{a} with @var{b} and to compare @var{c} with @var{d}. 7824The @code{any} forms return true if any of the four results are true 7825and the @code{all} forms return true if all four results are true. 7826For example: 7827 7828@smallexample 7829v2sf a, b, c, d; 7830if (__builtin_mips_any_c_eq_4s (a, b, c, d)) 7831 some_are_true (); 7832else 7833 all_are_false (); 7834 7835if (__builtin_mips_all_c_eq_4s (a, b, c, d)) 7836 all_are_true (); 7837else 7838 some_are_false (); 7839@end smallexample 7840@end table 7841 7842@node PowerPC AltiVec Built-in Functions 7843@subsection PowerPC AltiVec Built-in Functions 7844 7845GCC provides an interface for the PowerPC family of processors to access 7846the AltiVec operations described in Motorola's AltiVec Programming 7847Interface Manual. The interface is made available by including 7848@code{<altivec.h>} and using @option{-maltivec} and 7849@option{-mabi=altivec}. The interface supports the following vector 7850types. 7851 7852@smallexample 7853vector unsigned char 7854vector signed char 7855vector bool char 7856 7857vector unsigned short 7858vector signed short 7859vector bool short 7860vector pixel 7861 7862vector unsigned int 7863vector signed int 7864vector bool int 7865vector float 7866@end smallexample 7867 7868GCC's implementation of the high-level language interface available from 7869C and C++ code differs from Motorola's documentation in several ways. 7870 7871@itemize @bullet 7872 7873@item 7874A vector constant is a list of constant expressions within curly braces. 7875 7876@item 7877A vector initializer requires no cast if the vector constant is of the 7878same type as the variable it is initializing. 7879 7880@item 7881If @code{signed} or @code{unsigned} is omitted, the signedness of the 7882vector type is the default signedness of the base type. The default 7883varies depending on the operating system, so a portable program should 7884always specify the signedness. 7885 7886@item 7887Compiling with @option{-maltivec} adds keywords @code{__vector}, 7888@code{__pixel}, and @code{__bool}. Macros @option{vector}, 7889@code{pixel}, and @code{bool} are defined in @code{<altivec.h>} and can 7890be undefined. 7891 7892@item 7893GCC allows using a @code{typedef} name as the type specifier for a 7894vector type. 7895 7896@item 7897For C, overloaded functions are implemented with macros so the following 7898does not work: 7899 7900@smallexample 7901 vec_add ((vector signed int)@{1, 2, 3, 4@}, foo); 7902@end smallexample 7903 7904Since @code{vec_add} is a macro, the vector constant in the example 7905is treated as four separate arguments. Wrap the entire argument in 7906parentheses for this to work. 7907@end itemize 7908 7909@emph{Note:} Only the @code{<altivec.h>} interface is supported. 7910Internally, GCC uses built-in functions to achieve the functionality in 7911the aforementioned header file, but they are not supported and are 7912subject to change without notice. 7913 7914The following interfaces are supported for the generic and specific 7915AltiVec operations and the AltiVec predicates. In cases where there 7916is a direct mapping between generic and specific operations, only the 7917generic names are shown here, although the specific operations can also 7918be used. 7919 7920Arguments that are documented as @code{const int} require literal 7921integral values within the range required for that operation. 7922 7923@smallexample 7924vector signed char vec_abs (vector signed char); 7925vector signed short vec_abs (vector signed short); 7926vector signed int vec_abs (vector signed int); 7927vector float vec_abs (vector float); 7928 7929vector signed char vec_abss (vector signed char); 7930vector signed short vec_abss (vector signed short); 7931vector signed int vec_abss (vector signed int); 7932 7933vector signed char vec_add (vector bool char, vector signed char); 7934vector signed char vec_add (vector signed char, vector bool char); 7935vector signed char vec_add (vector signed char, vector signed char); 7936vector unsigned char vec_add (vector bool char, vector unsigned char); 7937vector unsigned char vec_add (vector unsigned char, vector bool char); 7938vector unsigned char vec_add (vector unsigned char, 7939 vector unsigned char); 7940vector signed short vec_add (vector bool short, vector signed short); 7941vector signed short vec_add (vector signed short, vector bool short); 7942vector signed short vec_add (vector signed short, vector signed short); 7943vector unsigned short vec_add (vector bool short, 7944 vector unsigned short); 7945vector unsigned short vec_add (vector unsigned short, 7946 vector bool short); 7947vector unsigned short vec_add (vector unsigned short, 7948 vector unsigned short); 7949vector signed int vec_add (vector bool int, vector signed int); 7950vector signed int vec_add (vector signed int, vector bool int); 7951vector signed int vec_add (vector signed int, vector signed int); 7952vector unsigned int vec_add (vector bool int, vector unsigned int); 7953vector unsigned int vec_add (vector unsigned int, vector bool int); 7954vector unsigned int vec_add (vector unsigned int, vector unsigned int); 7955vector float vec_add (vector float, vector float); 7956 7957vector float vec_vaddfp (vector float, vector float); 7958 7959vector signed int vec_vadduwm (vector bool int, vector signed int); 7960vector signed int vec_vadduwm (vector signed int, vector bool int); 7961vector signed int vec_vadduwm (vector signed int, vector signed int); 7962vector unsigned int vec_vadduwm (vector bool int, vector unsigned int); 7963vector unsigned int vec_vadduwm (vector unsigned int, vector bool int); 7964vector unsigned int vec_vadduwm (vector unsigned int, 7965 vector unsigned int); 7966 7967vector signed short vec_vadduhm (vector bool short, 7968 vector signed short); 7969vector signed short vec_vadduhm (vector signed short, 7970 vector bool short); 7971vector signed short vec_vadduhm (vector signed short, 7972 vector signed short); 7973vector unsigned short vec_vadduhm (vector bool short, 7974 vector unsigned short); 7975vector unsigned short vec_vadduhm (vector unsigned short, 7976 vector bool short); 7977vector unsigned short vec_vadduhm (vector unsigned short, 7978 vector unsigned short); 7979 7980vector signed char vec_vaddubm (vector bool char, vector signed char); 7981vector signed char vec_vaddubm (vector signed char, vector bool char); 7982vector signed char vec_vaddubm (vector signed char, vector signed char); 7983vector unsigned char vec_vaddubm (vector bool char, 7984 vector unsigned char); 7985vector unsigned char vec_vaddubm (vector unsigned char, 7986 vector bool char); 7987vector unsigned char vec_vaddubm (vector unsigned char, 7988 vector unsigned char); 7989 7990vector unsigned int vec_addc (vector unsigned int, vector unsigned int); 7991 7992vector unsigned char vec_adds (vector bool char, vector unsigned char); 7993vector unsigned char vec_adds (vector unsigned char, vector bool char); 7994vector unsigned char vec_adds (vector unsigned char, 7995 vector unsigned char); 7996vector signed char vec_adds (vector bool char, vector signed char); 7997vector signed char vec_adds (vector signed char, vector bool char); 7998vector signed char vec_adds (vector signed char, vector signed char); 7999vector unsigned short vec_adds (vector bool short, 8000 vector unsigned short); 8001vector unsigned short vec_adds (vector unsigned short, 8002 vector bool short); 8003vector unsigned short vec_adds (vector unsigned short, 8004 vector unsigned short); 8005vector signed short vec_adds (vector bool short, vector signed short); 8006vector signed short vec_adds (vector signed short, vector bool short); 8007vector signed short vec_adds (vector signed short, vector signed short); 8008vector unsigned int vec_adds (vector bool int, vector unsigned int); 8009vector unsigned int vec_adds (vector unsigned int, vector bool int); 8010vector unsigned int vec_adds (vector unsigned int, vector unsigned int); 8011vector signed int vec_adds (vector bool int, vector signed int); 8012vector signed int vec_adds (vector signed int, vector bool int); 8013vector signed int vec_adds (vector signed int, vector signed int); 8014 8015vector signed int vec_vaddsws (vector bool int, vector signed int); 8016vector signed int vec_vaddsws (vector signed int, vector bool int); 8017vector signed int vec_vaddsws (vector signed int, vector signed int); 8018 8019vector unsigned int vec_vadduws (vector bool int, vector unsigned int); 8020vector unsigned int vec_vadduws (vector unsigned int, vector bool int); 8021vector unsigned int vec_vadduws (vector unsigned int, 8022 vector unsigned int); 8023 8024vector signed short vec_vaddshs (vector bool short, 8025 vector signed short); 8026vector signed short vec_vaddshs (vector signed short, 8027 vector bool short); 8028vector signed short vec_vaddshs (vector signed short, 8029 vector signed short); 8030 8031vector unsigned short vec_vadduhs (vector bool short, 8032 vector unsigned short); 8033vector unsigned short vec_vadduhs (vector unsigned short, 8034 vector bool short); 8035vector unsigned short vec_vadduhs (vector unsigned short, 8036 vector unsigned short); 8037 8038vector signed char vec_vaddsbs (vector bool char, vector signed char); 8039vector signed char vec_vaddsbs (vector signed char, vector bool char); 8040vector signed char vec_vaddsbs (vector signed char, vector signed char); 8041 8042vector unsigned char vec_vaddubs (vector bool char, 8043 vector unsigned char); 8044vector unsigned char vec_vaddubs (vector unsigned char, 8045 vector bool char); 8046vector unsigned char vec_vaddubs (vector unsigned char, 8047 vector unsigned char); 8048 8049vector float vec_and (vector float, vector float); 8050vector float vec_and (vector float, vector bool int); 8051vector float vec_and (vector bool int, vector float); 8052vector bool int vec_and (vector bool int, vector bool int); 8053vector signed int vec_and (vector bool int, vector signed int); 8054vector signed int vec_and (vector signed int, vector bool int); 8055vector signed int vec_and (vector signed int, vector signed int); 8056vector unsigned int vec_and (vector bool int, vector unsigned int); 8057vector unsigned int vec_and (vector unsigned int, vector bool int); 8058vector unsigned int vec_and (vector unsigned int, vector unsigned int); 8059vector bool short vec_and (vector bool short, vector bool short); 8060vector signed short vec_and (vector bool short, vector signed short); 8061vector signed short vec_and (vector signed short, vector bool short); 8062vector signed short vec_and (vector signed short, vector signed short); 8063vector unsigned short vec_and (vector bool short, 8064 vector unsigned short); 8065vector unsigned short vec_and (vector unsigned short, 8066 vector bool short); 8067vector unsigned short vec_and (vector unsigned short, 8068 vector unsigned short); 8069vector signed char vec_and (vector bool char, vector signed char); 8070vector bool char vec_and (vector bool char, vector bool char); 8071vector signed char vec_and (vector signed char, vector bool char); 8072vector signed char vec_and (vector signed char, vector signed char); 8073vector unsigned char vec_and (vector bool char, vector unsigned char); 8074vector unsigned char vec_and (vector unsigned char, vector bool char); 8075vector unsigned char vec_and (vector unsigned char, 8076 vector unsigned char); 8077 8078vector float vec_andc (vector float, vector float); 8079vector float vec_andc (vector float, vector bool int); 8080vector float vec_andc (vector bool int, vector float); 8081vector bool int vec_andc (vector bool int, vector bool int); 8082vector signed int vec_andc (vector bool int, vector signed int); 8083vector signed int vec_andc (vector signed int, vector bool int); 8084vector signed int vec_andc (vector signed int, vector signed int); 8085vector unsigned int vec_andc (vector bool int, vector unsigned int); 8086vector unsigned int vec_andc (vector unsigned int, vector bool int); 8087vector unsigned int vec_andc (vector unsigned int, vector unsigned int); 8088vector bool short vec_andc (vector bool short, vector bool short); 8089vector signed short vec_andc (vector bool short, vector signed short); 8090vector signed short vec_andc (vector signed short, vector bool short); 8091vector signed short vec_andc (vector signed short, vector signed short); 8092vector unsigned short vec_andc (vector bool short, 8093 vector unsigned short); 8094vector unsigned short vec_andc (vector unsigned short, 8095 vector bool short); 8096vector unsigned short vec_andc (vector unsigned short, 8097 vector unsigned short); 8098vector signed char vec_andc (vector bool char, vector signed char); 8099vector bool char vec_andc (vector bool char, vector bool char); 8100vector signed char vec_andc (vector signed char, vector bool char); 8101vector signed char vec_andc (vector signed char, vector signed char); 8102vector unsigned char vec_andc (vector bool char, vector unsigned char); 8103vector unsigned char vec_andc (vector unsigned char, vector bool char); 8104vector unsigned char vec_andc (vector unsigned char, 8105 vector unsigned char); 8106 8107vector unsigned char vec_avg (vector unsigned char, 8108 vector unsigned char); 8109vector signed char vec_avg (vector signed char, vector signed char); 8110vector unsigned short vec_avg (vector unsigned short, 8111 vector unsigned short); 8112vector signed short vec_avg (vector signed short, vector signed short); 8113vector unsigned int vec_avg (vector unsigned int, vector unsigned int); 8114vector signed int vec_avg (vector signed int, vector signed int); 8115 8116vector signed int vec_vavgsw (vector signed int, vector signed int); 8117 8118vector unsigned int vec_vavguw (vector unsigned int, 8119 vector unsigned int); 8120 8121vector signed short vec_vavgsh (vector signed short, 8122 vector signed short); 8123 8124vector unsigned short vec_vavguh (vector unsigned short, 8125 vector unsigned short); 8126 8127vector signed char vec_vavgsb (vector signed char, vector signed char); 8128 8129vector unsigned char vec_vavgub (vector unsigned char, 8130 vector unsigned char); 8131 8132vector float vec_ceil (vector float); 8133 8134vector signed int vec_cmpb (vector float, vector float); 8135 8136vector bool char vec_cmpeq (vector signed char, vector signed char); 8137vector bool char vec_cmpeq (vector unsigned char, vector unsigned char); 8138vector bool short vec_cmpeq (vector signed short, vector signed short); 8139vector bool short vec_cmpeq (vector unsigned short, 8140 vector unsigned short); 8141vector bool int vec_cmpeq (vector signed int, vector signed int); 8142vector bool int vec_cmpeq (vector unsigned int, vector unsigned int); 8143vector bool int vec_cmpeq (vector float, vector float); 8144 8145vector bool int vec_vcmpeqfp (vector float, vector float); 8146 8147vector bool int vec_vcmpequw (vector signed int, vector signed int); 8148vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int); 8149 8150vector bool short vec_vcmpequh (vector signed short, 8151 vector signed short); 8152vector bool short vec_vcmpequh (vector unsigned short, 8153 vector unsigned short); 8154 8155vector bool char vec_vcmpequb (vector signed char, vector signed char); 8156vector bool char vec_vcmpequb (vector unsigned char, 8157 vector unsigned char); 8158 8159vector bool int vec_cmpge (vector float, vector float); 8160 8161vector bool char vec_cmpgt (vector unsigned char, vector unsigned char); 8162vector bool char vec_cmpgt (vector signed char, vector signed char); 8163vector bool short vec_cmpgt (vector unsigned short, 8164 vector unsigned short); 8165vector bool short vec_cmpgt (vector signed short, vector signed short); 8166vector bool int vec_cmpgt (vector unsigned int, vector unsigned int); 8167vector bool int vec_cmpgt (vector signed int, vector signed int); 8168vector bool int vec_cmpgt (vector float, vector float); 8169 8170vector bool int vec_vcmpgtfp (vector float, vector float); 8171 8172vector bool int vec_vcmpgtsw (vector signed int, vector signed int); 8173 8174vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int); 8175 8176vector bool short vec_vcmpgtsh (vector signed short, 8177 vector signed short); 8178 8179vector bool short vec_vcmpgtuh (vector unsigned short, 8180 vector unsigned short); 8181 8182vector bool char vec_vcmpgtsb (vector signed char, vector signed char); 8183 8184vector bool char vec_vcmpgtub (vector unsigned char, 8185 vector unsigned char); 8186 8187vector bool int vec_cmple (vector float, vector float); 8188 8189vector bool char vec_cmplt (vector unsigned char, vector unsigned char); 8190vector bool char vec_cmplt (vector signed char, vector signed char); 8191vector bool short vec_cmplt (vector unsigned short, 8192 vector unsigned short); 8193vector bool short vec_cmplt (vector signed short, vector signed short); 8194vector bool int vec_cmplt (vector unsigned int, vector unsigned int); 8195vector bool int vec_cmplt (vector signed int, vector signed int); 8196vector bool int vec_cmplt (vector float, vector float); 8197 8198vector float vec_ctf (vector unsigned int, const int); 8199vector float vec_ctf (vector signed int, const int); 8200 8201vector float vec_vcfsx (vector signed int, const int); 8202 8203vector float vec_vcfux (vector unsigned int, const int); 8204 8205vector signed int vec_cts (vector float, const int); 8206 8207vector unsigned int vec_ctu (vector float, const int); 8208 8209void vec_dss (const int); 8210 8211void vec_dssall (void); 8212 8213void vec_dst (const vector unsigned char *, int, const int); 8214void vec_dst (const vector signed char *, int, const int); 8215void vec_dst (const vector bool char *, int, const int); 8216void vec_dst (const vector unsigned short *, int, const int); 8217void vec_dst (const vector signed short *, int, const int); 8218void vec_dst (const vector bool short *, int, const int); 8219void vec_dst (const vector pixel *, int, const int); 8220void vec_dst (const vector unsigned int *, int, const int); 8221void vec_dst (const vector signed int *, int, const int); 8222void vec_dst (const vector bool int *, int, const int); 8223void vec_dst (const vector float *, int, const int); 8224void vec_dst (const unsigned char *, int, const int); 8225void vec_dst (const signed char *, int, const int); 8226void vec_dst (const unsigned short *, int, const int); 8227void vec_dst (const short *, int, const int); 8228void vec_dst (const unsigned int *, int, const int); 8229void vec_dst (const int *, int, const int); 8230void vec_dst (const unsigned long *, int, const int); 8231void vec_dst (const long *, int, const int); 8232void vec_dst (const float *, int, const int); 8233 8234void vec_dstst (const vector unsigned char *, int, const int); 8235void vec_dstst (const vector signed char *, int, const int); 8236void vec_dstst (const vector bool char *, int, const int); 8237void vec_dstst (const vector unsigned short *, int, const int); 8238void vec_dstst (const vector signed short *, int, const int); 8239void vec_dstst (const vector bool short *, int, const int); 8240void vec_dstst (const vector pixel *, int, const int); 8241void vec_dstst (const vector unsigned int *, int, const int); 8242void vec_dstst (const vector signed int *, int, const int); 8243void vec_dstst (const vector bool int *, int, const int); 8244void vec_dstst (const vector float *, int, const int); 8245void vec_dstst (const unsigned char *, int, const int); 8246void vec_dstst (const signed char *, int, const int); 8247void vec_dstst (const unsigned short *, int, const int); 8248void vec_dstst (const short *, int, const int); 8249void vec_dstst (const unsigned int *, int, const int); 8250void vec_dstst (const int *, int, const int); 8251void vec_dstst (const unsigned long *, int, const int); 8252void vec_dstst (const long *, int, const int); 8253void vec_dstst (const float *, int, const int); 8254 8255void vec_dststt (const vector unsigned char *, int, const int); 8256void vec_dststt (const vector signed char *, int, const int); 8257void vec_dststt (const vector bool char *, int, const int); 8258void vec_dststt (const vector unsigned short *, int, const int); 8259void vec_dststt (const vector signed short *, int, const int); 8260void vec_dststt (const vector bool short *, int, const int); 8261void vec_dststt (const vector pixel *, int, const int); 8262void vec_dststt (const vector unsigned int *, int, const int); 8263void vec_dststt (const vector signed int *, int, const int); 8264void vec_dststt (const vector bool int *, int, const int); 8265void vec_dststt (const vector float *, int, const int); 8266void vec_dststt (const unsigned char *, int, const int); 8267void vec_dststt (const signed char *, int, const int); 8268void vec_dststt (const unsigned short *, int, const int); 8269void vec_dststt (const short *, int, const int); 8270void vec_dststt (const unsigned int *, int, const int); 8271void vec_dststt (const int *, int, const int); 8272void vec_dststt (const unsigned long *, int, const int); 8273void vec_dststt (const long *, int, const int); 8274void vec_dststt (const float *, int, const int); 8275 8276void vec_dstt (const vector unsigned char *, int, const int); 8277void vec_dstt (const vector signed char *, int, const int); 8278void vec_dstt (const vector bool char *, int, const int); 8279void vec_dstt (const vector unsigned short *, int, const int); 8280void vec_dstt (const vector signed short *, int, const int); 8281void vec_dstt (const vector bool short *, int, const int); 8282void vec_dstt (const vector pixel *, int, const int); 8283void vec_dstt (const vector unsigned int *, int, const int); 8284void vec_dstt (const vector signed int *, int, const int); 8285void vec_dstt (const vector bool int *, int, const int); 8286void vec_dstt (const vector float *, int, const int); 8287void vec_dstt (const unsigned char *, int, const int); 8288void vec_dstt (const signed char *, int, const int); 8289void vec_dstt (const unsigned short *, int, const int); 8290void vec_dstt (const short *, int, const int); 8291void vec_dstt (const unsigned int *, int, const int); 8292void vec_dstt (const int *, int, const int); 8293void vec_dstt (const unsigned long *, int, const int); 8294void vec_dstt (const long *, int, const int); 8295void vec_dstt (const float *, int, const int); 8296 8297vector float vec_expte (vector float); 8298 8299vector float vec_floor (vector float); 8300 8301vector float vec_ld (int, const vector float *); 8302vector float vec_ld (int, const float *); 8303vector bool int vec_ld (int, const vector bool int *); 8304vector signed int vec_ld (int, const vector signed int *); 8305vector signed int vec_ld (int, const int *); 8306vector signed int vec_ld (int, const long *); 8307vector unsigned int vec_ld (int, const vector unsigned int *); 8308vector unsigned int vec_ld (int, const unsigned int *); 8309vector unsigned int vec_ld (int, const unsigned long *); 8310vector bool short vec_ld (int, const vector bool short *); 8311vector pixel vec_ld (int, const vector pixel *); 8312vector signed short vec_ld (int, const vector signed short *); 8313vector signed short vec_ld (int, const short *); 8314vector unsigned short vec_ld (int, const vector unsigned short *); 8315vector unsigned short vec_ld (int, const unsigned short *); 8316vector bool char vec_ld (int, const vector bool char *); 8317vector signed char vec_ld (int, const vector signed char *); 8318vector signed char vec_ld (int, const signed char *); 8319vector unsigned char vec_ld (int, const vector unsigned char *); 8320vector unsigned char vec_ld (int, const unsigned char *); 8321 8322vector signed char vec_lde (int, const signed char *); 8323vector unsigned char vec_lde (int, const unsigned char *); 8324vector signed short vec_lde (int, const short *); 8325vector unsigned short vec_lde (int, const unsigned short *); 8326vector float vec_lde (int, const float *); 8327vector signed int vec_lde (int, const int *); 8328vector unsigned int vec_lde (int, const unsigned int *); 8329vector signed int vec_lde (int, const long *); 8330vector unsigned int vec_lde (int, const unsigned long *); 8331 8332vector float vec_lvewx (int, float *); 8333vector signed int vec_lvewx (int, int *); 8334vector unsigned int vec_lvewx (int, unsigned int *); 8335vector signed int vec_lvewx (int, long *); 8336vector unsigned int vec_lvewx (int, unsigned long *); 8337 8338vector signed short vec_lvehx (int, short *); 8339vector unsigned short vec_lvehx (int, unsigned short *); 8340 8341vector signed char vec_lvebx (int, char *); 8342vector unsigned char vec_lvebx (int, unsigned char *); 8343 8344vector float vec_ldl (int, const vector float *); 8345vector float vec_ldl (int, const float *); 8346vector bool int vec_ldl (int, const vector bool int *); 8347vector signed int vec_ldl (int, const vector signed int *); 8348vector signed int vec_ldl (int, const int *); 8349vector signed int vec_ldl (int, const long *); 8350vector unsigned int vec_ldl (int, const vector unsigned int *); 8351vector unsigned int vec_ldl (int, const unsigned int *); 8352vector unsigned int vec_ldl (int, const unsigned long *); 8353vector bool short vec_ldl (int, const vector bool short *); 8354vector pixel vec_ldl (int, const vector pixel *); 8355vector signed short vec_ldl (int, const vector signed short *); 8356vector signed short vec_ldl (int, const short *); 8357vector unsigned short vec_ldl (int, const vector unsigned short *); 8358vector unsigned short vec_ldl (int, const unsigned short *); 8359vector bool char vec_ldl (int, const vector bool char *); 8360vector signed char vec_ldl (int, const vector signed char *); 8361vector signed char vec_ldl (int, const signed char *); 8362vector unsigned char vec_ldl (int, const vector unsigned char *); 8363vector unsigned char vec_ldl (int, const unsigned char *); 8364 8365vector float vec_loge (vector float); 8366 8367vector unsigned char vec_lvsl (int, const volatile unsigned char *); 8368vector unsigned char vec_lvsl (int, const volatile signed char *); 8369vector unsigned char vec_lvsl (int, const volatile unsigned short *); 8370vector unsigned char vec_lvsl (int, const volatile short *); 8371vector unsigned char vec_lvsl (int, const volatile unsigned int *); 8372vector unsigned char vec_lvsl (int, const volatile int *); 8373vector unsigned char vec_lvsl (int, const volatile unsigned long *); 8374vector unsigned char vec_lvsl (int, const volatile long *); 8375vector unsigned char vec_lvsl (int, const volatile float *); 8376 8377vector unsigned char vec_lvsr (int, const volatile unsigned char *); 8378vector unsigned char vec_lvsr (int, const volatile signed char *); 8379vector unsigned char vec_lvsr (int, const volatile unsigned short *); 8380vector unsigned char vec_lvsr (int, const volatile short *); 8381vector unsigned char vec_lvsr (int, const volatile unsigned int *); 8382vector unsigned char vec_lvsr (int, const volatile int *); 8383vector unsigned char vec_lvsr (int, const volatile unsigned long *); 8384vector unsigned char vec_lvsr (int, const volatile long *); 8385vector unsigned char vec_lvsr (int, const volatile float *); 8386 8387vector float vec_madd (vector float, vector float, vector float); 8388 8389vector signed short vec_madds (vector signed short, 8390 vector signed short, 8391 vector signed short); 8392 8393vector unsigned char vec_max (vector bool char, vector unsigned char); 8394vector unsigned char vec_max (vector unsigned char, vector bool char); 8395vector unsigned char vec_max (vector unsigned char, 8396 vector unsigned char); 8397vector signed char vec_max (vector bool char, vector signed char); 8398vector signed char vec_max (vector signed char, vector bool char); 8399vector signed char vec_max (vector signed char, vector signed char); 8400vector unsigned short vec_max (vector bool short, 8401 vector unsigned short); 8402vector unsigned short vec_max (vector unsigned short, 8403 vector bool short); 8404vector unsigned short vec_max (vector unsigned short, 8405 vector unsigned short); 8406vector signed short vec_max (vector bool short, vector signed short); 8407vector signed short vec_max (vector signed short, vector bool short); 8408vector signed short vec_max (vector signed short, vector signed short); 8409vector unsigned int vec_max (vector bool int, vector unsigned int); 8410vector unsigned int vec_max (vector unsigned int, vector bool int); 8411vector unsigned int vec_max (vector unsigned int, vector unsigned int); 8412vector signed int vec_max (vector bool int, vector signed int); 8413vector signed int vec_max (vector signed int, vector bool int); 8414vector signed int vec_max (vector signed int, vector signed int); 8415vector float vec_max (vector float, vector float); 8416 8417vector float vec_vmaxfp (vector float, vector float); 8418 8419vector signed int vec_vmaxsw (vector bool int, vector signed int); 8420vector signed int vec_vmaxsw (vector signed int, vector bool int); 8421vector signed int vec_vmaxsw (vector signed int, vector signed int); 8422 8423vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int); 8424vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int); 8425vector unsigned int vec_vmaxuw (vector unsigned int, 8426 vector unsigned int); 8427 8428vector signed short vec_vmaxsh (vector bool short, vector signed short); 8429vector signed short vec_vmaxsh (vector signed short, vector bool short); 8430vector signed short vec_vmaxsh (vector signed short, 8431 vector signed short); 8432 8433vector unsigned short vec_vmaxuh (vector bool short, 8434 vector unsigned short); 8435vector unsigned short vec_vmaxuh (vector unsigned short, 8436 vector bool short); 8437vector unsigned short vec_vmaxuh (vector unsigned short, 8438 vector unsigned short); 8439 8440vector signed char vec_vmaxsb (vector bool char, vector signed char); 8441vector signed char vec_vmaxsb (vector signed char, vector bool char); 8442vector signed char vec_vmaxsb (vector signed char, vector signed char); 8443 8444vector unsigned char vec_vmaxub (vector bool char, 8445 vector unsigned char); 8446vector unsigned char vec_vmaxub (vector unsigned char, 8447 vector bool char); 8448vector unsigned char vec_vmaxub (vector unsigned char, 8449 vector unsigned char); 8450 8451vector bool char vec_mergeh (vector bool char, vector bool char); 8452vector signed char vec_mergeh (vector signed char, vector signed char); 8453vector unsigned char vec_mergeh (vector unsigned char, 8454 vector unsigned char); 8455vector bool short vec_mergeh (vector bool short, vector bool short); 8456vector pixel vec_mergeh (vector pixel, vector pixel); 8457vector signed short vec_mergeh (vector signed short, 8458 vector signed short); 8459vector unsigned short vec_mergeh (vector unsigned short, 8460 vector unsigned short); 8461vector float vec_mergeh (vector float, vector float); 8462vector bool int vec_mergeh (vector bool int, vector bool int); 8463vector signed int vec_mergeh (vector signed int, vector signed int); 8464vector unsigned int vec_mergeh (vector unsigned int, 8465 vector unsigned int); 8466 8467vector float vec_vmrghw (vector float, vector float); 8468vector bool int vec_vmrghw (vector bool int, vector bool int); 8469vector signed int vec_vmrghw (vector signed int, vector signed int); 8470vector unsigned int vec_vmrghw (vector unsigned int, 8471 vector unsigned int); 8472 8473vector bool short vec_vmrghh (vector bool short, vector bool short); 8474vector signed short vec_vmrghh (vector signed short, 8475 vector signed short); 8476vector unsigned short vec_vmrghh (vector unsigned short, 8477 vector unsigned short); 8478vector pixel vec_vmrghh (vector pixel, vector pixel); 8479 8480vector bool char vec_vmrghb (vector bool char, vector bool char); 8481vector signed char vec_vmrghb (vector signed char, vector signed char); 8482vector unsigned char vec_vmrghb (vector unsigned char, 8483 vector unsigned char); 8484 8485vector bool char vec_mergel (vector bool char, vector bool char); 8486vector signed char vec_mergel (vector signed char, vector signed char); 8487vector unsigned char vec_mergel (vector unsigned char, 8488 vector unsigned char); 8489vector bool short vec_mergel (vector bool short, vector bool short); 8490vector pixel vec_mergel (vector pixel, vector pixel); 8491vector signed short vec_mergel (vector signed short, 8492 vector signed short); 8493vector unsigned short vec_mergel (vector unsigned short, 8494 vector unsigned short); 8495vector float vec_mergel (vector float, vector float); 8496vector bool int vec_mergel (vector bool int, vector bool int); 8497vector signed int vec_mergel (vector signed int, vector signed int); 8498vector unsigned int vec_mergel (vector unsigned int, 8499 vector unsigned int); 8500 8501vector float vec_vmrglw (vector float, vector float); 8502vector signed int vec_vmrglw (vector signed int, vector signed int); 8503vector unsigned int vec_vmrglw (vector unsigned int, 8504 vector unsigned int); 8505vector bool int vec_vmrglw (vector bool int, vector bool int); 8506 8507vector bool short vec_vmrglh (vector bool short, vector bool short); 8508vector signed short vec_vmrglh (vector signed short, 8509 vector signed short); 8510vector unsigned short vec_vmrglh (vector unsigned short, 8511 vector unsigned short); 8512vector pixel vec_vmrglh (vector pixel, vector pixel); 8513 8514vector bool char vec_vmrglb (vector bool char, vector bool char); 8515vector signed char vec_vmrglb (vector signed char, vector signed char); 8516vector unsigned char vec_vmrglb (vector unsigned char, 8517 vector unsigned char); 8518 8519vector unsigned short vec_mfvscr (void); 8520 8521vector unsigned char vec_min (vector bool char, vector unsigned char); 8522vector unsigned char vec_min (vector unsigned char, vector bool char); 8523vector unsigned char vec_min (vector unsigned char, 8524 vector unsigned char); 8525vector signed char vec_min (vector bool char, vector signed char); 8526vector signed char vec_min (vector signed char, vector bool char); 8527vector signed char vec_min (vector signed char, vector signed char); 8528vector unsigned short vec_min (vector bool short, 8529 vector unsigned short); 8530vector unsigned short vec_min (vector unsigned short, 8531 vector bool short); 8532vector unsigned short vec_min (vector unsigned short, 8533 vector unsigned short); 8534vector signed short vec_min (vector bool short, vector signed short); 8535vector signed short vec_min (vector signed short, vector bool short); 8536vector signed short vec_min (vector signed short, vector signed short); 8537vector unsigned int vec_min (vector bool int, vector unsigned int); 8538vector unsigned int vec_min (vector unsigned int, vector bool int); 8539vector unsigned int vec_min (vector unsigned int, vector unsigned int); 8540vector signed int vec_min (vector bool int, vector signed int); 8541vector signed int vec_min (vector signed int, vector bool int); 8542vector signed int vec_min (vector signed int, vector signed int); 8543vector float vec_min (vector float, vector float); 8544 8545vector float vec_vminfp (vector float, vector float); 8546 8547vector signed int vec_vminsw (vector bool int, vector signed int); 8548vector signed int vec_vminsw (vector signed int, vector bool int); 8549vector signed int vec_vminsw (vector signed int, vector signed int); 8550 8551vector unsigned int vec_vminuw (vector bool int, vector unsigned int); 8552vector unsigned int vec_vminuw (vector unsigned int, vector bool int); 8553vector unsigned int vec_vminuw (vector unsigned int, 8554 vector unsigned int); 8555 8556vector signed short vec_vminsh (vector bool short, vector signed short); 8557vector signed short vec_vminsh (vector signed short, vector bool short); 8558vector signed short vec_vminsh (vector signed short, 8559 vector signed short); 8560 8561vector unsigned short vec_vminuh (vector bool short, 8562 vector unsigned short); 8563vector unsigned short vec_vminuh (vector unsigned short, 8564 vector bool short); 8565vector unsigned short vec_vminuh (vector unsigned short, 8566 vector unsigned short); 8567 8568vector signed char vec_vminsb (vector bool char, vector signed char); 8569vector signed char vec_vminsb (vector signed char, vector bool char); 8570vector signed char vec_vminsb (vector signed char, vector signed char); 8571 8572vector unsigned char vec_vminub (vector bool char, 8573 vector unsigned char); 8574vector unsigned char vec_vminub (vector unsigned char, 8575 vector bool char); 8576vector unsigned char vec_vminub (vector unsigned char, 8577 vector unsigned char); 8578 8579vector signed short vec_mladd (vector signed short, 8580 vector signed short, 8581 vector signed short); 8582vector signed short vec_mladd (vector signed short, 8583 vector unsigned short, 8584 vector unsigned short); 8585vector signed short vec_mladd (vector unsigned short, 8586 vector signed short, 8587 vector signed short); 8588vector unsigned short vec_mladd (vector unsigned short, 8589 vector unsigned short, 8590 vector unsigned short); 8591 8592vector signed short vec_mradds (vector signed short, 8593 vector signed short, 8594 vector signed short); 8595 8596vector unsigned int vec_msum (vector unsigned char, 8597 vector unsigned char, 8598 vector unsigned int); 8599vector signed int vec_msum (vector signed char, 8600 vector unsigned char, 8601 vector signed int); 8602vector unsigned int vec_msum (vector unsigned short, 8603 vector unsigned short, 8604 vector unsigned int); 8605vector signed int vec_msum (vector signed short, 8606 vector signed short, 8607 vector signed int); 8608 8609vector signed int vec_vmsumshm (vector signed short, 8610 vector signed short, 8611 vector signed int); 8612 8613vector unsigned int vec_vmsumuhm (vector unsigned short, 8614 vector unsigned short, 8615 vector unsigned int); 8616 8617vector signed int vec_vmsummbm (vector signed char, 8618 vector unsigned char, 8619 vector signed int); 8620 8621vector unsigned int vec_vmsumubm (vector unsigned char, 8622 vector unsigned char, 8623 vector unsigned int); 8624 8625vector unsigned int vec_msums (vector unsigned short, 8626 vector unsigned short, 8627 vector unsigned int); 8628vector signed int vec_msums (vector signed short, 8629 vector signed short, 8630 vector signed int); 8631 8632vector signed int vec_vmsumshs (vector signed short, 8633 vector signed short, 8634 vector signed int); 8635 8636vector unsigned int vec_vmsumuhs (vector unsigned short, 8637 vector unsigned short, 8638 vector unsigned int); 8639 8640void vec_mtvscr (vector signed int); 8641void vec_mtvscr (vector unsigned int); 8642void vec_mtvscr (vector bool int); 8643void vec_mtvscr (vector signed short); 8644void vec_mtvscr (vector unsigned short); 8645void vec_mtvscr (vector bool short); 8646void vec_mtvscr (vector pixel); 8647void vec_mtvscr (vector signed char); 8648void vec_mtvscr (vector unsigned char); 8649void vec_mtvscr (vector bool char); 8650 8651vector unsigned short vec_mule (vector unsigned char, 8652 vector unsigned char); 8653vector signed short vec_mule (vector signed char, 8654 vector signed char); 8655vector unsigned int vec_mule (vector unsigned short, 8656 vector unsigned short); 8657vector signed int vec_mule (vector signed short, vector signed short); 8658 8659vector signed int vec_vmulesh (vector signed short, 8660 vector signed short); 8661 8662vector unsigned int vec_vmuleuh (vector unsigned short, 8663 vector unsigned short); 8664 8665vector signed short vec_vmulesb (vector signed char, 8666 vector signed char); 8667 8668vector unsigned short vec_vmuleub (vector unsigned char, 8669 vector unsigned char); 8670 8671vector unsigned short vec_mulo (vector unsigned char, 8672 vector unsigned char); 8673vector signed short vec_mulo (vector signed char, vector signed char); 8674vector unsigned int vec_mulo (vector unsigned short, 8675 vector unsigned short); 8676vector signed int vec_mulo (vector signed short, vector signed short); 8677 8678vector signed int vec_vmulosh (vector signed short, 8679 vector signed short); 8680 8681vector unsigned int vec_vmulouh (vector unsigned short, 8682 vector unsigned short); 8683 8684vector signed short vec_vmulosb (vector signed char, 8685 vector signed char); 8686 8687vector unsigned short vec_vmuloub (vector unsigned char, 8688 vector unsigned char); 8689 8690vector float vec_nmsub (vector float, vector float, vector float); 8691 8692vector float vec_nor (vector float, vector float); 8693vector signed int vec_nor (vector signed int, vector signed int); 8694vector unsigned int vec_nor (vector unsigned int, vector unsigned int); 8695vector bool int vec_nor (vector bool int, vector bool int); 8696vector signed short vec_nor (vector signed short, vector signed short); 8697vector unsigned short vec_nor (vector unsigned short, 8698 vector unsigned short); 8699vector bool short vec_nor (vector bool short, vector bool short); 8700vector signed char vec_nor (vector signed char, vector signed char); 8701vector unsigned char vec_nor (vector unsigned char, 8702 vector unsigned char); 8703vector bool char vec_nor (vector bool char, vector bool char); 8704 8705vector float vec_or (vector float, vector float); 8706vector float vec_or (vector float, vector bool int); 8707vector float vec_or (vector bool int, vector float); 8708vector bool int vec_or (vector bool int, vector bool int); 8709vector signed int vec_or (vector bool int, vector signed int); 8710vector signed int vec_or (vector signed int, vector bool int); 8711vector signed int vec_or (vector signed int, vector signed int); 8712vector unsigned int vec_or (vector bool int, vector unsigned int); 8713vector unsigned int vec_or (vector unsigned int, vector bool int); 8714vector unsigned int vec_or (vector unsigned int, vector unsigned int); 8715vector bool short vec_or (vector bool short, vector bool short); 8716vector signed short vec_or (vector bool short, vector signed short); 8717vector signed short vec_or (vector signed short, vector bool short); 8718vector signed short vec_or (vector signed short, vector signed short); 8719vector unsigned short vec_or (vector bool short, vector unsigned short); 8720vector unsigned short vec_or (vector unsigned short, vector bool short); 8721vector unsigned short vec_or (vector unsigned short, 8722 vector unsigned short); 8723vector signed char vec_or (vector bool char, vector signed char); 8724vector bool char vec_or (vector bool char, vector bool char); 8725vector signed char vec_or (vector signed char, vector bool char); 8726vector signed char vec_or (vector signed char, vector signed char); 8727vector unsigned char vec_or (vector bool char, vector unsigned char); 8728vector unsigned char vec_or (vector unsigned char, vector bool char); 8729vector unsigned char vec_or (vector unsigned char, 8730 vector unsigned char); 8731 8732vector signed char vec_pack (vector signed short, vector signed short); 8733vector unsigned char vec_pack (vector unsigned short, 8734 vector unsigned short); 8735vector bool char vec_pack (vector bool short, vector bool short); 8736vector signed short vec_pack (vector signed int, vector signed int); 8737vector unsigned short vec_pack (vector unsigned int, 8738 vector unsigned int); 8739vector bool short vec_pack (vector bool int, vector bool int); 8740 8741vector bool short vec_vpkuwum (vector bool int, vector bool int); 8742vector signed short vec_vpkuwum (vector signed int, vector signed int); 8743vector unsigned short vec_vpkuwum (vector unsigned int, 8744 vector unsigned int); 8745 8746vector bool char vec_vpkuhum (vector bool short, vector bool short); 8747vector signed char vec_vpkuhum (vector signed short, 8748 vector signed short); 8749vector unsigned char vec_vpkuhum (vector unsigned short, 8750 vector unsigned short); 8751 8752vector pixel vec_packpx (vector unsigned int, vector unsigned int); 8753 8754vector unsigned char vec_packs (vector unsigned short, 8755 vector unsigned short); 8756vector signed char vec_packs (vector signed short, vector signed short); 8757vector unsigned short vec_packs (vector unsigned int, 8758 vector unsigned int); 8759vector signed short vec_packs (vector signed int, vector signed int); 8760 8761vector signed short vec_vpkswss (vector signed int, vector signed int); 8762 8763vector unsigned short vec_vpkuwus (vector unsigned int, 8764 vector unsigned int); 8765 8766vector signed char vec_vpkshss (vector signed short, 8767 vector signed short); 8768 8769vector unsigned char vec_vpkuhus (vector unsigned short, 8770 vector unsigned short); 8771 8772vector unsigned char vec_packsu (vector unsigned short, 8773 vector unsigned short); 8774vector unsigned char vec_packsu (vector signed short, 8775 vector signed short); 8776vector unsigned short vec_packsu (vector unsigned int, 8777 vector unsigned int); 8778vector unsigned short vec_packsu (vector signed int, vector signed int); 8779 8780vector unsigned short vec_vpkswus (vector signed int, 8781 vector signed int); 8782 8783vector unsigned char vec_vpkshus (vector signed short, 8784 vector signed short); 8785 8786vector float vec_perm (vector float, 8787 vector float, 8788 vector unsigned char); 8789vector signed int vec_perm (vector signed int, 8790 vector signed int, 8791 vector unsigned char); 8792vector unsigned int vec_perm (vector unsigned int, 8793 vector unsigned int, 8794 vector unsigned char); 8795vector bool int vec_perm (vector bool int, 8796 vector bool int, 8797 vector unsigned char); 8798vector signed short vec_perm (vector signed short, 8799 vector signed short, 8800 vector unsigned char); 8801vector unsigned short vec_perm (vector unsigned short, 8802 vector unsigned short, 8803 vector unsigned char); 8804vector bool short vec_perm (vector bool short, 8805 vector bool short, 8806 vector unsigned char); 8807vector pixel vec_perm (vector pixel, 8808 vector pixel, 8809 vector unsigned char); 8810vector signed char vec_perm (vector signed char, 8811 vector signed char, 8812 vector unsigned char); 8813vector unsigned char vec_perm (vector unsigned char, 8814 vector unsigned char, 8815 vector unsigned char); 8816vector bool char vec_perm (vector bool char, 8817 vector bool char, 8818 vector unsigned char); 8819 8820vector float vec_re (vector float); 8821 8822vector signed char vec_rl (vector signed char, 8823 vector unsigned char); 8824vector unsigned char vec_rl (vector unsigned char, 8825 vector unsigned char); 8826vector signed short vec_rl (vector signed short, vector unsigned short); 8827vector unsigned short vec_rl (vector unsigned short, 8828 vector unsigned short); 8829vector signed int vec_rl (vector signed int, vector unsigned int); 8830vector unsigned int vec_rl (vector unsigned int, vector unsigned int); 8831 8832vector signed int vec_vrlw (vector signed int, vector unsigned int); 8833vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int); 8834 8835vector signed short vec_vrlh (vector signed short, 8836 vector unsigned short); 8837vector unsigned short vec_vrlh (vector unsigned short, 8838 vector unsigned short); 8839 8840vector signed char vec_vrlb (vector signed char, vector unsigned char); 8841vector unsigned char vec_vrlb (vector unsigned char, 8842 vector unsigned char); 8843 8844vector float vec_round (vector float); 8845 8846vector float vec_rsqrte (vector float); 8847 8848vector float vec_sel (vector float, vector float, vector bool int); 8849vector float vec_sel (vector float, vector float, vector unsigned int); 8850vector signed int vec_sel (vector signed int, 8851 vector signed int, 8852 vector bool int); 8853vector signed int vec_sel (vector signed int, 8854 vector signed int, 8855 vector unsigned int); 8856vector unsigned int vec_sel (vector unsigned int, 8857 vector unsigned int, 8858 vector bool int); 8859vector unsigned int vec_sel (vector unsigned int, 8860 vector unsigned int, 8861 vector unsigned int); 8862vector bool int vec_sel (vector bool int, 8863 vector bool int, 8864 vector bool int); 8865vector bool int vec_sel (vector bool int, 8866 vector bool int, 8867 vector unsigned int); 8868vector signed short vec_sel (vector signed short, 8869 vector signed short, 8870 vector bool short); 8871vector signed short vec_sel (vector signed short, 8872 vector signed short, 8873 vector unsigned short); 8874vector unsigned short vec_sel (vector unsigned short, 8875 vector unsigned short, 8876 vector bool short); 8877vector unsigned short vec_sel (vector unsigned short, 8878 vector unsigned short, 8879 vector unsigned short); 8880vector bool short vec_sel (vector bool short, 8881 vector bool short, 8882 vector bool short); 8883vector bool short vec_sel (vector bool short, 8884 vector bool short, 8885 vector unsigned short); 8886vector signed char vec_sel (vector signed char, 8887 vector signed char, 8888 vector bool char); 8889vector signed char vec_sel (vector signed char, 8890 vector signed char, 8891 vector unsigned char); 8892vector unsigned char vec_sel (vector unsigned char, 8893 vector unsigned char, 8894 vector bool char); 8895vector unsigned char vec_sel (vector unsigned char, 8896 vector unsigned char, 8897 vector unsigned char); 8898vector bool char vec_sel (vector bool char, 8899 vector bool char, 8900 vector bool char); 8901vector bool char vec_sel (vector bool char, 8902 vector bool char, 8903 vector unsigned char); 8904 8905vector signed char vec_sl (vector signed char, 8906 vector unsigned char); 8907vector unsigned char vec_sl (vector unsigned char, 8908 vector unsigned char); 8909vector signed short vec_sl (vector signed short, vector unsigned short); 8910vector unsigned short vec_sl (vector unsigned short, 8911 vector unsigned short); 8912vector signed int vec_sl (vector signed int, vector unsigned int); 8913vector unsigned int vec_sl (vector unsigned int, vector unsigned int); 8914 8915vector signed int vec_vslw (vector signed int, vector unsigned int); 8916vector unsigned int vec_vslw (vector unsigned int, vector unsigned int); 8917 8918vector signed short vec_vslh (vector signed short, 8919 vector unsigned short); 8920vector unsigned short vec_vslh (vector unsigned short, 8921 vector unsigned short); 8922 8923vector signed char vec_vslb (vector signed char, vector unsigned char); 8924vector unsigned char vec_vslb (vector unsigned char, 8925 vector unsigned char); 8926 8927vector float vec_sld (vector float, vector float, const int); 8928vector signed int vec_sld (vector signed int, 8929 vector signed int, 8930 const int); 8931vector unsigned int vec_sld (vector unsigned int, 8932 vector unsigned int, 8933 const int); 8934vector bool int vec_sld (vector bool int, 8935 vector bool int, 8936 const int); 8937vector signed short vec_sld (vector signed short, 8938 vector signed short, 8939 const int); 8940vector unsigned short vec_sld (vector unsigned short, 8941 vector unsigned short, 8942 const int); 8943vector bool short vec_sld (vector bool short, 8944 vector bool short, 8945 const int); 8946vector pixel vec_sld (vector pixel, 8947 vector pixel, 8948 const int); 8949vector signed char vec_sld (vector signed char, 8950 vector signed char, 8951 const int); 8952vector unsigned char vec_sld (vector unsigned char, 8953 vector unsigned char, 8954 const int); 8955vector bool char vec_sld (vector bool char, 8956 vector bool char, 8957 const int); 8958 8959vector signed int vec_sll (vector signed int, 8960 vector unsigned int); 8961vector signed int vec_sll (vector signed int, 8962 vector unsigned short); 8963vector signed int vec_sll (vector signed int, 8964 vector unsigned char); 8965vector unsigned int vec_sll (vector unsigned int, 8966 vector unsigned int); 8967vector unsigned int vec_sll (vector unsigned int, 8968 vector unsigned short); 8969vector unsigned int vec_sll (vector unsigned int, 8970 vector unsigned char); 8971vector bool int vec_sll (vector bool int, 8972 vector unsigned int); 8973vector bool int vec_sll (vector bool int, 8974 vector unsigned short); 8975vector bool int vec_sll (vector bool int, 8976 vector unsigned char); 8977vector signed short vec_sll (vector signed short, 8978 vector unsigned int); 8979vector signed short vec_sll (vector signed short, 8980 vector unsigned short); 8981vector signed short vec_sll (vector signed short, 8982 vector unsigned char); 8983vector unsigned short vec_sll (vector unsigned short, 8984 vector unsigned int); 8985vector unsigned short vec_sll (vector unsigned short, 8986 vector unsigned short); 8987vector unsigned short vec_sll (vector unsigned short, 8988 vector unsigned char); 8989vector bool short vec_sll (vector bool short, vector unsigned int); 8990vector bool short vec_sll (vector bool short, vector unsigned short); 8991vector bool short vec_sll (vector bool short, vector unsigned char); 8992vector pixel vec_sll (vector pixel, vector unsigned int); 8993vector pixel vec_sll (vector pixel, vector unsigned short); 8994vector pixel vec_sll (vector pixel, vector unsigned char); 8995vector signed char vec_sll (vector signed char, vector unsigned int); 8996vector signed char vec_sll (vector signed char, vector unsigned short); 8997vector signed char vec_sll (vector signed char, vector unsigned char); 8998vector unsigned char vec_sll (vector unsigned char, 8999 vector unsigned int); 9000vector unsigned char vec_sll (vector unsigned char, 9001 vector unsigned short); 9002vector unsigned char vec_sll (vector unsigned char, 9003 vector unsigned char); 9004vector bool char vec_sll (vector bool char, vector unsigned int); 9005vector bool char vec_sll (vector bool char, vector unsigned short); 9006vector bool char vec_sll (vector bool char, vector unsigned char); 9007 9008vector float vec_slo (vector float, vector signed char); 9009vector float vec_slo (vector float, vector unsigned char); 9010vector signed int vec_slo (vector signed int, vector signed char); 9011vector signed int vec_slo (vector signed int, vector unsigned char); 9012vector unsigned int vec_slo (vector unsigned int, vector signed char); 9013vector unsigned int vec_slo (vector unsigned int, vector unsigned char); 9014vector signed short vec_slo (vector signed short, vector signed char); 9015vector signed short vec_slo (vector signed short, vector unsigned char); 9016vector unsigned short vec_slo (vector unsigned short, 9017 vector signed char); 9018vector unsigned short vec_slo (vector unsigned short, 9019 vector unsigned char); 9020vector pixel vec_slo (vector pixel, vector signed char); 9021vector pixel vec_slo (vector pixel, vector unsigned char); 9022vector signed char vec_slo (vector signed char, vector signed char); 9023vector signed char vec_slo (vector signed char, vector unsigned char); 9024vector unsigned char vec_slo (vector unsigned char, vector signed char); 9025vector unsigned char vec_slo (vector unsigned char, 9026 vector unsigned char); 9027 9028vector signed char vec_splat (vector signed char, const int); 9029vector unsigned char vec_splat (vector unsigned char, const int); 9030vector bool char vec_splat (vector bool char, const int); 9031vector signed short vec_splat (vector signed short, const int); 9032vector unsigned short vec_splat (vector unsigned short, const int); 9033vector bool short vec_splat (vector bool short, const int); 9034vector pixel vec_splat (vector pixel, const int); 9035vector float vec_splat (vector float, const int); 9036vector signed int vec_splat (vector signed int, const int); 9037vector unsigned int vec_splat (vector unsigned int, const int); 9038vector bool int vec_splat (vector bool int, const int); 9039 9040vector float vec_vspltw (vector float, const int); 9041vector signed int vec_vspltw (vector signed int, const int); 9042vector unsigned int vec_vspltw (vector unsigned int, const int); 9043vector bool int vec_vspltw (vector bool int, const int); 9044 9045vector bool short vec_vsplth (vector bool short, const int); 9046vector signed short vec_vsplth (vector signed short, const int); 9047vector unsigned short vec_vsplth (vector unsigned short, const int); 9048vector pixel vec_vsplth (vector pixel, const int); 9049 9050vector signed char vec_vspltb (vector signed char, const int); 9051vector unsigned char vec_vspltb (vector unsigned char, const int); 9052vector bool char vec_vspltb (vector bool char, const int); 9053 9054vector signed char vec_splat_s8 (const int); 9055 9056vector signed short vec_splat_s16 (const int); 9057 9058vector signed int vec_splat_s32 (const int); 9059 9060vector unsigned char vec_splat_u8 (const int); 9061 9062vector unsigned short vec_splat_u16 (const int); 9063 9064vector unsigned int vec_splat_u32 (const int); 9065 9066vector signed char vec_sr (vector signed char, vector unsigned char); 9067vector unsigned char vec_sr (vector unsigned char, 9068 vector unsigned char); 9069vector signed short vec_sr (vector signed short, 9070 vector unsigned short); 9071vector unsigned short vec_sr (vector unsigned short, 9072 vector unsigned short); 9073vector signed int vec_sr (vector signed int, vector unsigned int); 9074vector unsigned int vec_sr (vector unsigned int, vector unsigned int); 9075 9076vector signed int vec_vsrw (vector signed int, vector unsigned int); 9077vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int); 9078 9079vector signed short vec_vsrh (vector signed short, 9080 vector unsigned short); 9081vector unsigned short vec_vsrh (vector unsigned short, 9082 vector unsigned short); 9083 9084vector signed char vec_vsrb (vector signed char, vector unsigned char); 9085vector unsigned char vec_vsrb (vector unsigned char, 9086 vector unsigned char); 9087 9088vector signed char vec_sra (vector signed char, vector unsigned char); 9089vector unsigned char vec_sra (vector unsigned char, 9090 vector unsigned char); 9091vector signed short vec_sra (vector signed short, 9092 vector unsigned short); 9093vector unsigned short vec_sra (vector unsigned short, 9094 vector unsigned short); 9095vector signed int vec_sra (vector signed int, vector unsigned int); 9096vector unsigned int vec_sra (vector unsigned int, vector unsigned int); 9097 9098vector signed int vec_vsraw (vector signed int, vector unsigned int); 9099vector unsigned int vec_vsraw (vector unsigned int, 9100 vector unsigned int); 9101 9102vector signed short vec_vsrah (vector signed short, 9103 vector unsigned short); 9104vector unsigned short vec_vsrah (vector unsigned short, 9105 vector unsigned short); 9106 9107vector signed char vec_vsrab (vector signed char, vector unsigned char); 9108vector unsigned char vec_vsrab (vector unsigned char, 9109 vector unsigned char); 9110 9111vector signed int vec_srl (vector signed int, vector unsigned int); 9112vector signed int vec_srl (vector signed int, vector unsigned short); 9113vector signed int vec_srl (vector signed int, vector unsigned char); 9114vector unsigned int vec_srl (vector unsigned int, vector unsigned int); 9115vector unsigned int vec_srl (vector unsigned int, 9116 vector unsigned short); 9117vector unsigned int vec_srl (vector unsigned int, vector unsigned char); 9118vector bool int vec_srl (vector bool int, vector unsigned int); 9119vector bool int vec_srl (vector bool int, vector unsigned short); 9120vector bool int vec_srl (vector bool int, vector unsigned char); 9121vector signed short vec_srl (vector signed short, vector unsigned int); 9122vector signed short vec_srl (vector signed short, 9123 vector unsigned short); 9124vector signed short vec_srl (vector signed short, vector unsigned char); 9125vector unsigned short vec_srl (vector unsigned short, 9126 vector unsigned int); 9127vector unsigned short vec_srl (vector unsigned short, 9128 vector unsigned short); 9129vector unsigned short vec_srl (vector unsigned short, 9130 vector unsigned char); 9131vector bool short vec_srl (vector bool short, vector unsigned int); 9132vector bool short vec_srl (vector bool short, vector unsigned short); 9133vector bool short vec_srl (vector bool short, vector unsigned char); 9134vector pixel vec_srl (vector pixel, vector unsigned int); 9135vector pixel vec_srl (vector pixel, vector unsigned short); 9136vector pixel vec_srl (vector pixel, vector unsigned char); 9137vector signed char vec_srl (vector signed char, vector unsigned int); 9138vector signed char vec_srl (vector signed char, vector unsigned short); 9139vector signed char vec_srl (vector signed char, vector unsigned char); 9140vector unsigned char vec_srl (vector unsigned char, 9141 vector unsigned int); 9142vector unsigned char vec_srl (vector unsigned char, 9143 vector unsigned short); 9144vector unsigned char vec_srl (vector unsigned char, 9145 vector unsigned char); 9146vector bool char vec_srl (vector bool char, vector unsigned int); 9147vector bool char vec_srl (vector bool char, vector unsigned short); 9148vector bool char vec_srl (vector bool char, vector unsigned char); 9149 9150vector float vec_sro (vector float, vector signed char); 9151vector float vec_sro (vector float, vector unsigned char); 9152vector signed int vec_sro (vector signed int, vector signed char); 9153vector signed int vec_sro (vector signed int, vector unsigned char); 9154vector unsigned int vec_sro (vector unsigned int, vector signed char); 9155vector unsigned int vec_sro (vector unsigned int, vector unsigned char); 9156vector signed short vec_sro (vector signed short, vector signed char); 9157vector signed short vec_sro (vector signed short, vector unsigned char); 9158vector unsigned short vec_sro (vector unsigned short, 9159 vector signed char); 9160vector unsigned short vec_sro (vector unsigned short, 9161 vector unsigned char); 9162vector pixel vec_sro (vector pixel, vector signed char); 9163vector pixel vec_sro (vector pixel, vector unsigned char); 9164vector signed char vec_sro (vector signed char, vector signed char); 9165vector signed char vec_sro (vector signed char, vector unsigned char); 9166vector unsigned char vec_sro (vector unsigned char, vector signed char); 9167vector unsigned char vec_sro (vector unsigned char, 9168 vector unsigned char); 9169 9170void vec_st (vector float, int, vector float *); 9171void vec_st (vector float, int, float *); 9172void vec_st (vector signed int, int, vector signed int *); 9173void vec_st (vector signed int, int, int *); 9174void vec_st (vector unsigned int, int, vector unsigned int *); 9175void vec_st (vector unsigned int, int, unsigned int *); 9176void vec_st (vector bool int, int, vector bool int *); 9177void vec_st (vector bool int, int, unsigned int *); 9178void vec_st (vector bool int, int, int *); 9179void vec_st (vector signed short, int, vector signed short *); 9180void vec_st (vector signed short, int, short *); 9181void vec_st (vector unsigned short, int, vector unsigned short *); 9182void vec_st (vector unsigned short, int, unsigned short *); 9183void vec_st (vector bool short, int, vector bool short *); 9184void vec_st (vector bool short, int, unsigned short *); 9185void vec_st (vector pixel, int, vector pixel *); 9186void vec_st (vector pixel, int, unsigned short *); 9187void vec_st (vector pixel, int, short *); 9188void vec_st (vector bool short, int, short *); 9189void vec_st (vector signed char, int, vector signed char *); 9190void vec_st (vector signed char, int, signed char *); 9191void vec_st (vector unsigned char, int, vector unsigned char *); 9192void vec_st (vector unsigned char, int, unsigned char *); 9193void vec_st (vector bool char, int, vector bool char *); 9194void vec_st (vector bool char, int, unsigned char *); 9195void vec_st (vector bool char, int, signed char *); 9196 9197void vec_ste (vector signed char, int, signed char *); 9198void vec_ste (vector unsigned char, int, unsigned char *); 9199void vec_ste (vector bool char, int, signed char *); 9200void vec_ste (vector bool char, int, unsigned char *); 9201void vec_ste (vector signed short, int, short *); 9202void vec_ste (vector unsigned short, int, unsigned short *); 9203void vec_ste (vector bool short, int, short *); 9204void vec_ste (vector bool short, int, unsigned short *); 9205void vec_ste (vector pixel, int, short *); 9206void vec_ste (vector pixel, int, unsigned short *); 9207void vec_ste (vector float, int, float *); 9208void vec_ste (vector signed int, int, int *); 9209void vec_ste (vector unsigned int, int, unsigned int *); 9210void vec_ste (vector bool int, int, int *); 9211void vec_ste (vector bool int, int, unsigned int *); 9212 9213void vec_stvewx (vector float, int, float *); 9214void vec_stvewx (vector signed int, int, int *); 9215void vec_stvewx (vector unsigned int, int, unsigned int *); 9216void vec_stvewx (vector bool int, int, int *); 9217void vec_stvewx (vector bool int, int, unsigned int *); 9218 9219void vec_stvehx (vector signed short, int, short *); 9220void vec_stvehx (vector unsigned short, int, unsigned short *); 9221void vec_stvehx (vector bool short, int, short *); 9222void vec_stvehx (vector bool short, int, unsigned short *); 9223void vec_stvehx (vector pixel, int, short *); 9224void vec_stvehx (vector pixel, int, unsigned short *); 9225 9226void vec_stvebx (vector signed char, int, signed char *); 9227void vec_stvebx (vector unsigned char, int, unsigned char *); 9228void vec_stvebx (vector bool char, int, signed char *); 9229void vec_stvebx (vector bool char, int, unsigned char *); 9230 9231void vec_stl (vector float, int, vector float *); 9232void vec_stl (vector float, int, float *); 9233void vec_stl (vector signed int, int, vector signed int *); 9234void vec_stl (vector signed int, int, int *); 9235void vec_stl (vector unsigned int, int, vector unsigned int *); 9236void vec_stl (vector unsigned int, int, unsigned int *); 9237void vec_stl (vector bool int, int, vector bool int *); 9238void vec_stl (vector bool int, int, unsigned int *); 9239void vec_stl (vector bool int, int, int *); 9240void vec_stl (vector signed short, int, vector signed short *); 9241void vec_stl (vector signed short, int, short *); 9242void vec_stl (vector unsigned short, int, vector unsigned short *); 9243void vec_stl (vector unsigned short, int, unsigned short *); 9244void vec_stl (vector bool short, int, vector bool short *); 9245void vec_stl (vector bool short, int, unsigned short *); 9246void vec_stl (vector bool short, int, short *); 9247void vec_stl (vector pixel, int, vector pixel *); 9248void vec_stl (vector pixel, int, unsigned short *); 9249void vec_stl (vector pixel, int, short *); 9250void vec_stl (vector signed char, int, vector signed char *); 9251void vec_stl (vector signed char, int, signed char *); 9252void vec_stl (vector unsigned char, int, vector unsigned char *); 9253void vec_stl (vector unsigned char, int, unsigned char *); 9254void vec_stl (vector bool char, int, vector bool char *); 9255void vec_stl (vector bool char, int, unsigned char *); 9256void vec_stl (vector bool char, int, signed char *); 9257 9258vector signed char vec_sub (vector bool char, vector signed char); 9259vector signed char vec_sub (vector signed char, vector bool char); 9260vector signed char vec_sub (vector signed char, vector signed char); 9261vector unsigned char vec_sub (vector bool char, vector unsigned char); 9262vector unsigned char vec_sub (vector unsigned char, vector bool char); 9263vector unsigned char vec_sub (vector unsigned char, 9264 vector unsigned char); 9265vector signed short vec_sub (vector bool short, vector signed short); 9266vector signed short vec_sub (vector signed short, vector bool short); 9267vector signed short vec_sub (vector signed short, vector signed short); 9268vector unsigned short vec_sub (vector bool short, 9269 vector unsigned short); 9270vector unsigned short vec_sub (vector unsigned short, 9271 vector bool short); 9272vector unsigned short vec_sub (vector unsigned short, 9273 vector unsigned short); 9274vector signed int vec_sub (vector bool int, vector signed int); 9275vector signed int vec_sub (vector signed int, vector bool int); 9276vector signed int vec_sub (vector signed int, vector signed int); 9277vector unsigned int vec_sub (vector bool int, vector unsigned int); 9278vector unsigned int vec_sub (vector unsigned int, vector bool int); 9279vector unsigned int vec_sub (vector unsigned int, vector unsigned int); 9280vector float vec_sub (vector float, vector float); 9281 9282vector float vec_vsubfp (vector float, vector float); 9283 9284vector signed int vec_vsubuwm (vector bool int, vector signed int); 9285vector signed int vec_vsubuwm (vector signed int, vector bool int); 9286vector signed int vec_vsubuwm (vector signed int, vector signed int); 9287vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int); 9288vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int); 9289vector unsigned int vec_vsubuwm (vector unsigned int, 9290 vector unsigned int); 9291 9292vector signed short vec_vsubuhm (vector bool short, 9293 vector signed short); 9294vector signed short vec_vsubuhm (vector signed short, 9295 vector bool short); 9296vector signed short vec_vsubuhm (vector signed short, 9297 vector signed short); 9298vector unsigned short vec_vsubuhm (vector bool short, 9299 vector unsigned short); 9300vector unsigned short vec_vsubuhm (vector unsigned short, 9301 vector bool short); 9302vector unsigned short vec_vsubuhm (vector unsigned short, 9303 vector unsigned short); 9304 9305vector signed char vec_vsububm (vector bool char, vector signed char); 9306vector signed char vec_vsububm (vector signed char, vector bool char); 9307vector signed char vec_vsububm (vector signed char, vector signed char); 9308vector unsigned char vec_vsububm (vector bool char, 9309 vector unsigned char); 9310vector unsigned char vec_vsububm (vector unsigned char, 9311 vector bool char); 9312vector unsigned char vec_vsububm (vector unsigned char, 9313 vector unsigned char); 9314 9315vector unsigned int vec_subc (vector unsigned int, vector unsigned int); 9316 9317vector unsigned char vec_subs (vector bool char, vector unsigned char); 9318vector unsigned char vec_subs (vector unsigned char, vector bool char); 9319vector unsigned char vec_subs (vector unsigned char, 9320 vector unsigned char); 9321vector signed char vec_subs (vector bool char, vector signed char); 9322vector signed char vec_subs (vector signed char, vector bool char); 9323vector signed char vec_subs (vector signed char, vector signed char); 9324vector unsigned short vec_subs (vector bool short, 9325 vector unsigned short); 9326vector unsigned short vec_subs (vector unsigned short, 9327 vector bool short); 9328vector unsigned short vec_subs (vector unsigned short, 9329 vector unsigned short); 9330vector signed short vec_subs (vector bool short, vector signed short); 9331vector signed short vec_subs (vector signed short, vector bool short); 9332vector signed short vec_subs (vector signed short, vector signed short); 9333vector unsigned int vec_subs (vector bool int, vector unsigned int); 9334vector unsigned int vec_subs (vector unsigned int, vector bool int); 9335vector unsigned int vec_subs (vector unsigned int, vector unsigned int); 9336vector signed int vec_subs (vector bool int, vector signed int); 9337vector signed int vec_subs (vector signed int, vector bool int); 9338vector signed int vec_subs (vector signed int, vector signed int); 9339 9340vector signed int vec_vsubsws (vector bool int, vector signed int); 9341vector signed int vec_vsubsws (vector signed int, vector bool int); 9342vector signed int vec_vsubsws (vector signed int, vector signed int); 9343 9344vector unsigned int vec_vsubuws (vector bool int, vector unsigned int); 9345vector unsigned int vec_vsubuws (vector unsigned int, vector bool int); 9346vector unsigned int vec_vsubuws (vector unsigned int, 9347 vector unsigned int); 9348 9349vector signed short vec_vsubshs (vector bool short, 9350 vector signed short); 9351vector signed short vec_vsubshs (vector signed short, 9352 vector bool short); 9353vector signed short vec_vsubshs (vector signed short, 9354 vector signed short); 9355 9356vector unsigned short vec_vsubuhs (vector bool short, 9357 vector unsigned short); 9358vector unsigned short vec_vsubuhs (vector unsigned short, 9359 vector bool short); 9360vector unsigned short vec_vsubuhs (vector unsigned short, 9361 vector unsigned short); 9362 9363vector signed char vec_vsubsbs (vector bool char, vector signed char); 9364vector signed char vec_vsubsbs (vector signed char, vector bool char); 9365vector signed char vec_vsubsbs (vector signed char, vector signed char); 9366 9367vector unsigned char vec_vsububs (vector bool char, 9368 vector unsigned char); 9369vector unsigned char vec_vsububs (vector unsigned char, 9370 vector bool char); 9371vector unsigned char vec_vsububs (vector unsigned char, 9372 vector unsigned char); 9373 9374vector unsigned int vec_sum4s (vector unsigned char, 9375 vector unsigned int); 9376vector signed int vec_sum4s (vector signed char, vector signed int); 9377vector signed int vec_sum4s (vector signed short, vector signed int); 9378 9379vector signed int vec_vsum4shs (vector signed short, vector signed int); 9380 9381vector signed int vec_vsum4sbs (vector signed char, vector signed int); 9382 9383vector unsigned int vec_vsum4ubs (vector unsigned char, 9384 vector unsigned int); 9385 9386vector signed int vec_sum2s (vector signed int, vector signed int); 9387 9388vector signed int vec_sums (vector signed int, vector signed int); 9389 9390vector float vec_trunc (vector float); 9391 9392vector signed short vec_unpackh (vector signed char); 9393vector bool short vec_unpackh (vector bool char); 9394vector signed int vec_unpackh (vector signed short); 9395vector bool int vec_unpackh (vector bool short); 9396vector unsigned int vec_unpackh (vector pixel); 9397 9398vector bool int vec_vupkhsh (vector bool short); 9399vector signed int vec_vupkhsh (vector signed short); 9400 9401vector unsigned int vec_vupkhpx (vector pixel); 9402 9403vector bool short vec_vupkhsb (vector bool char); 9404vector signed short vec_vupkhsb (vector signed char); 9405 9406vector signed short vec_unpackl (vector signed char); 9407vector bool short vec_unpackl (vector bool char); 9408vector unsigned int vec_unpackl (vector pixel); 9409vector signed int vec_unpackl (vector signed short); 9410vector bool int vec_unpackl (vector bool short); 9411 9412vector unsigned int vec_vupklpx (vector pixel); 9413 9414vector bool int vec_vupklsh (vector bool short); 9415vector signed int vec_vupklsh (vector signed short); 9416 9417vector bool short vec_vupklsb (vector bool char); 9418vector signed short vec_vupklsb (vector signed char); 9419 9420vector float vec_xor (vector float, vector float); 9421vector float vec_xor (vector float, vector bool int); 9422vector float vec_xor (vector bool int, vector float); 9423vector bool int vec_xor (vector bool int, vector bool int); 9424vector signed int vec_xor (vector bool int, vector signed int); 9425vector signed int vec_xor (vector signed int, vector bool int); 9426vector signed int vec_xor (vector signed int, vector signed int); 9427vector unsigned int vec_xor (vector bool int, vector unsigned int); 9428vector unsigned int vec_xor (vector unsigned int, vector bool int); 9429vector unsigned int vec_xor (vector unsigned int, vector unsigned int); 9430vector bool short vec_xor (vector bool short, vector bool short); 9431vector signed short vec_xor (vector bool short, vector signed short); 9432vector signed short vec_xor (vector signed short, vector bool short); 9433vector signed short vec_xor (vector signed short, vector signed short); 9434vector unsigned short vec_xor (vector bool short, 9435 vector unsigned short); 9436vector unsigned short vec_xor (vector unsigned short, 9437 vector bool short); 9438vector unsigned short vec_xor (vector unsigned short, 9439 vector unsigned short); 9440vector signed char vec_xor (vector bool char, vector signed char); 9441vector bool char vec_xor (vector bool char, vector bool char); 9442vector signed char vec_xor (vector signed char, vector bool char); 9443vector signed char vec_xor (vector signed char, vector signed char); 9444vector unsigned char vec_xor (vector bool char, vector unsigned char); 9445vector unsigned char vec_xor (vector unsigned char, vector bool char); 9446vector unsigned char vec_xor (vector unsigned char, 9447 vector unsigned char); 9448 9449int vec_all_eq (vector signed char, vector bool char); 9450int vec_all_eq (vector signed char, vector signed char); 9451int vec_all_eq (vector unsigned char, vector bool char); 9452int vec_all_eq (vector unsigned char, vector unsigned char); 9453int vec_all_eq (vector bool char, vector bool char); 9454int vec_all_eq (vector bool char, vector unsigned char); 9455int vec_all_eq (vector bool char, vector signed char); 9456int vec_all_eq (vector signed short, vector bool short); 9457int vec_all_eq (vector signed short, vector signed short); 9458int vec_all_eq (vector unsigned short, vector bool short); 9459int vec_all_eq (vector unsigned short, vector unsigned short); 9460int vec_all_eq (vector bool short, vector bool short); 9461int vec_all_eq (vector bool short, vector unsigned short); 9462int vec_all_eq (vector bool short, vector signed short); 9463int vec_all_eq (vector pixel, vector pixel); 9464int vec_all_eq (vector signed int, vector bool int); 9465int vec_all_eq (vector signed int, vector signed int); 9466int vec_all_eq (vector unsigned int, vector bool int); 9467int vec_all_eq (vector unsigned int, vector unsigned int); 9468int vec_all_eq (vector bool int, vector bool int); 9469int vec_all_eq (vector bool int, vector unsigned int); 9470int vec_all_eq (vector bool int, vector signed int); 9471int vec_all_eq (vector float, vector float); 9472 9473int vec_all_ge (vector bool char, vector unsigned char); 9474int vec_all_ge (vector unsigned char, vector bool char); 9475int vec_all_ge (vector unsigned char, vector unsigned char); 9476int vec_all_ge (vector bool char, vector signed char); 9477int vec_all_ge (vector signed char, vector bool char); 9478int vec_all_ge (vector signed char, vector signed char); 9479int vec_all_ge (vector bool short, vector unsigned short); 9480int vec_all_ge (vector unsigned short, vector bool short); 9481int vec_all_ge (vector unsigned short, vector unsigned short); 9482int vec_all_ge (vector signed short, vector signed short); 9483int vec_all_ge (vector bool short, vector signed short); 9484int vec_all_ge (vector signed short, vector bool short); 9485int vec_all_ge (vector bool int, vector unsigned int); 9486int vec_all_ge (vector unsigned int, vector bool int); 9487int vec_all_ge (vector unsigned int, vector unsigned int); 9488int vec_all_ge (vector bool int, vector signed int); 9489int vec_all_ge (vector signed int, vector bool int); 9490int vec_all_ge (vector signed int, vector signed int); 9491int vec_all_ge (vector float, vector float); 9492 9493int vec_all_gt (vector bool char, vector unsigned char); 9494int vec_all_gt (vector unsigned char, vector bool char); 9495int vec_all_gt (vector unsigned char, vector unsigned char); 9496int vec_all_gt (vector bool char, vector signed char); 9497int vec_all_gt (vector signed char, vector bool char); 9498int vec_all_gt (vector signed char, vector signed char); 9499int vec_all_gt (vector bool short, vector unsigned short); 9500int vec_all_gt (vector unsigned short, vector bool short); 9501int vec_all_gt (vector unsigned short, vector unsigned short); 9502int vec_all_gt (vector bool short, vector signed short); 9503int vec_all_gt (vector signed short, vector bool short); 9504int vec_all_gt (vector signed short, vector signed short); 9505int vec_all_gt (vector bool int, vector unsigned int); 9506int vec_all_gt (vector unsigned int, vector bool int); 9507int vec_all_gt (vector unsigned int, vector unsigned int); 9508int vec_all_gt (vector bool int, vector signed int); 9509int vec_all_gt (vector signed int, vector bool int); 9510int vec_all_gt (vector signed int, vector signed int); 9511int vec_all_gt (vector float, vector float); 9512 9513int vec_all_in (vector float, vector float); 9514 9515int vec_all_le (vector bool char, vector unsigned char); 9516int vec_all_le (vector unsigned char, vector bool char); 9517int vec_all_le (vector unsigned char, vector unsigned char); 9518int vec_all_le (vector bool char, vector signed char); 9519int vec_all_le (vector signed char, vector bool char); 9520int vec_all_le (vector signed char, vector signed char); 9521int vec_all_le (vector bool short, vector unsigned short); 9522int vec_all_le (vector unsigned short, vector bool short); 9523int vec_all_le (vector unsigned short, vector unsigned short); 9524int vec_all_le (vector bool short, vector signed short); 9525int vec_all_le (vector signed short, vector bool short); 9526int vec_all_le (vector signed short, vector signed short); 9527int vec_all_le (vector bool int, vector unsigned int); 9528int vec_all_le (vector unsigned int, vector bool int); 9529int vec_all_le (vector unsigned int, vector unsigned int); 9530int vec_all_le (vector bool int, vector signed int); 9531int vec_all_le (vector signed int, vector bool int); 9532int vec_all_le (vector signed int, vector signed int); 9533int vec_all_le (vector float, vector float); 9534 9535int vec_all_lt (vector bool char, vector unsigned char); 9536int vec_all_lt (vector unsigned char, vector bool char); 9537int vec_all_lt (vector unsigned char, vector unsigned char); 9538int vec_all_lt (vector bool char, vector signed char); 9539int vec_all_lt (vector signed char, vector bool char); 9540int vec_all_lt (vector signed char, vector signed char); 9541int vec_all_lt (vector bool short, vector unsigned short); 9542int vec_all_lt (vector unsigned short, vector bool short); 9543int vec_all_lt (vector unsigned short, vector unsigned short); 9544int vec_all_lt (vector bool short, vector signed short); 9545int vec_all_lt (vector signed short, vector bool short); 9546int vec_all_lt (vector signed short, vector signed short); 9547int vec_all_lt (vector bool int, vector unsigned int); 9548int vec_all_lt (vector unsigned int, vector bool int); 9549int vec_all_lt (vector unsigned int, vector unsigned int); 9550int vec_all_lt (vector bool int, vector signed int); 9551int vec_all_lt (vector signed int, vector bool int); 9552int vec_all_lt (vector signed int, vector signed int); 9553int vec_all_lt (vector float, vector float); 9554 9555int vec_all_nan (vector float); 9556 9557int vec_all_ne (vector signed char, vector bool char); 9558int vec_all_ne (vector signed char, vector signed char); 9559int vec_all_ne (vector unsigned char, vector bool char); 9560int vec_all_ne (vector unsigned char, vector unsigned char); 9561int vec_all_ne (vector bool char, vector bool char); 9562int vec_all_ne (vector bool char, vector unsigned char); 9563int vec_all_ne (vector bool char, vector signed char); 9564int vec_all_ne (vector signed short, vector bool short); 9565int vec_all_ne (vector signed short, vector signed short); 9566int vec_all_ne (vector unsigned short, vector bool short); 9567int vec_all_ne (vector unsigned short, vector unsigned short); 9568int vec_all_ne (vector bool short, vector bool short); 9569int vec_all_ne (vector bool short, vector unsigned short); 9570int vec_all_ne (vector bool short, vector signed short); 9571int vec_all_ne (vector pixel, vector pixel); 9572int vec_all_ne (vector signed int, vector bool int); 9573int vec_all_ne (vector signed int, vector signed int); 9574int vec_all_ne (vector unsigned int, vector bool int); 9575int vec_all_ne (vector unsigned int, vector unsigned int); 9576int vec_all_ne (vector bool int, vector bool int); 9577int vec_all_ne (vector bool int, vector unsigned int); 9578int vec_all_ne (vector bool int, vector signed int); 9579int vec_all_ne (vector float, vector float); 9580 9581int vec_all_nge (vector float, vector float); 9582 9583int vec_all_ngt (vector float, vector float); 9584 9585int vec_all_nle (vector float, vector float); 9586 9587int vec_all_nlt (vector float, vector float); 9588 9589int vec_all_numeric (vector float); 9590 9591int vec_any_eq (vector signed char, vector bool char); 9592int vec_any_eq (vector signed char, vector signed char); 9593int vec_any_eq (vector unsigned char, vector bool char); 9594int vec_any_eq (vector unsigned char, vector unsigned char); 9595int vec_any_eq (vector bool char, vector bool char); 9596int vec_any_eq (vector bool char, vector unsigned char); 9597int vec_any_eq (vector bool char, vector signed char); 9598int vec_any_eq (vector signed short, vector bool short); 9599int vec_any_eq (vector signed short, vector signed short); 9600int vec_any_eq (vector unsigned short, vector bool short); 9601int vec_any_eq (vector unsigned short, vector unsigned short); 9602int vec_any_eq (vector bool short, vector bool short); 9603int vec_any_eq (vector bool short, vector unsigned short); 9604int vec_any_eq (vector bool short, vector signed short); 9605int vec_any_eq (vector pixel, vector pixel); 9606int vec_any_eq (vector signed int, vector bool int); 9607int vec_any_eq (vector signed int, vector signed int); 9608int vec_any_eq (vector unsigned int, vector bool int); 9609int vec_any_eq (vector unsigned int, vector unsigned int); 9610int vec_any_eq (vector bool int, vector bool int); 9611int vec_any_eq (vector bool int, vector unsigned int); 9612int vec_any_eq (vector bool int, vector signed int); 9613int vec_any_eq (vector float, vector float); 9614 9615int vec_any_ge (vector signed char, vector bool char); 9616int vec_any_ge (vector unsigned char, vector bool char); 9617int vec_any_ge (vector unsigned char, vector unsigned char); 9618int vec_any_ge (vector signed char, vector signed char); 9619int vec_any_ge (vector bool char, vector unsigned char); 9620int vec_any_ge (vector bool char, vector signed char); 9621int vec_any_ge (vector unsigned short, vector bool short); 9622int vec_any_ge (vector unsigned short, vector unsigned short); 9623int vec_any_ge (vector signed short, vector signed short); 9624int vec_any_ge (vector signed short, vector bool short); 9625int vec_any_ge (vector bool short, vector unsigned short); 9626int vec_any_ge (vector bool short, vector signed short); 9627int vec_any_ge (vector signed int, vector bool int); 9628int vec_any_ge (vector unsigned int, vector bool int); 9629int vec_any_ge (vector unsigned int, vector unsigned int); 9630int vec_any_ge (vector signed int, vector signed int); 9631int vec_any_ge (vector bool int, vector unsigned int); 9632int vec_any_ge (vector bool int, vector signed int); 9633int vec_any_ge (vector float, vector float); 9634 9635int vec_any_gt (vector bool char, vector unsigned char); 9636int vec_any_gt (vector unsigned char, vector bool char); 9637int vec_any_gt (vector unsigned char, vector unsigned char); 9638int vec_any_gt (vector bool char, vector signed char); 9639int vec_any_gt (vector signed char, vector bool char); 9640int vec_any_gt (vector signed char, vector signed char); 9641int vec_any_gt (vector bool short, vector unsigned short); 9642int vec_any_gt (vector unsigned short, vector bool short); 9643int vec_any_gt (vector unsigned short, vector unsigned short); 9644int vec_any_gt (vector bool short, vector signed short); 9645int vec_any_gt (vector signed short, vector bool short); 9646int vec_any_gt (vector signed short, vector signed short); 9647int vec_any_gt (vector bool int, vector unsigned int); 9648int vec_any_gt (vector unsigned int, vector bool int); 9649int vec_any_gt (vector unsigned int, vector unsigned int); 9650int vec_any_gt (vector bool int, vector signed int); 9651int vec_any_gt (vector signed int, vector bool int); 9652int vec_any_gt (vector signed int, vector signed int); 9653int vec_any_gt (vector float, vector float); 9654 9655int vec_any_le (vector bool char, vector unsigned char); 9656int vec_any_le (vector unsigned char, vector bool char); 9657int vec_any_le (vector unsigned char, vector unsigned char); 9658int vec_any_le (vector bool char, vector signed char); 9659int vec_any_le (vector signed char, vector bool char); 9660int vec_any_le (vector signed char, vector signed char); 9661int vec_any_le (vector bool short, vector unsigned short); 9662int vec_any_le (vector unsigned short, vector bool short); 9663int vec_any_le (vector unsigned short, vector unsigned short); 9664int vec_any_le (vector bool short, vector signed short); 9665int vec_any_le (vector signed short, vector bool short); 9666int vec_any_le (vector signed short, vector signed short); 9667int vec_any_le (vector bool int, vector unsigned int); 9668int vec_any_le (vector unsigned int, vector bool int); 9669int vec_any_le (vector unsigned int, vector unsigned int); 9670int vec_any_le (vector bool int, vector signed int); 9671int vec_any_le (vector signed int, vector bool int); 9672int vec_any_le (vector signed int, vector signed int); 9673int vec_any_le (vector float, vector float); 9674 9675int vec_any_lt (vector bool char, vector unsigned char); 9676int vec_any_lt (vector unsigned char, vector bool char); 9677int vec_any_lt (vector unsigned char, vector unsigned char); 9678int vec_any_lt (vector bool char, vector signed char); 9679int vec_any_lt (vector signed char, vector bool char); 9680int vec_any_lt (vector signed char, vector signed char); 9681int vec_any_lt (vector bool short, vector unsigned short); 9682int vec_any_lt (vector unsigned short, vector bool short); 9683int vec_any_lt (vector unsigned short, vector unsigned short); 9684int vec_any_lt (vector bool short, vector signed short); 9685int vec_any_lt (vector signed short, vector bool short); 9686int vec_any_lt (vector signed short, vector signed short); 9687int vec_any_lt (vector bool int, vector unsigned int); 9688int vec_any_lt (vector unsigned int, vector bool int); 9689int vec_any_lt (vector unsigned int, vector unsigned int); 9690int vec_any_lt (vector bool int, vector signed int); 9691int vec_any_lt (vector signed int, vector bool int); 9692int vec_any_lt (vector signed int, vector signed int); 9693int vec_any_lt (vector float, vector float); 9694 9695int vec_any_nan (vector float); 9696 9697int vec_any_ne (vector signed char, vector bool char); 9698int vec_any_ne (vector signed char, vector signed char); 9699int vec_any_ne (vector unsigned char, vector bool char); 9700int vec_any_ne (vector unsigned char, vector unsigned char); 9701int vec_any_ne (vector bool char, vector bool char); 9702int vec_any_ne (vector bool char, vector unsigned char); 9703int vec_any_ne (vector bool char, vector signed char); 9704int vec_any_ne (vector signed short, vector bool short); 9705int vec_any_ne (vector signed short, vector signed short); 9706int vec_any_ne (vector unsigned short, vector bool short); 9707int vec_any_ne (vector unsigned short, vector unsigned short); 9708int vec_any_ne (vector bool short, vector bool short); 9709int vec_any_ne (vector bool short, vector unsigned short); 9710int vec_any_ne (vector bool short, vector signed short); 9711int vec_any_ne (vector pixel, vector pixel); 9712int vec_any_ne (vector signed int, vector bool int); 9713int vec_any_ne (vector signed int, vector signed int); 9714int vec_any_ne (vector unsigned int, vector bool int); 9715int vec_any_ne (vector unsigned int, vector unsigned int); 9716int vec_any_ne (vector bool int, vector bool int); 9717int vec_any_ne (vector bool int, vector unsigned int); 9718int vec_any_ne (vector bool int, vector signed int); 9719int vec_any_ne (vector float, vector float); 9720 9721int vec_any_nge (vector float, vector float); 9722 9723int vec_any_ngt (vector float, vector float); 9724 9725int vec_any_nle (vector float, vector float); 9726 9727int vec_any_nlt (vector float, vector float); 9728 9729int vec_any_numeric (vector float); 9730 9731int vec_any_out (vector float, vector float); 9732@end smallexample 9733 9734@node SPARC VIS Built-in Functions 9735@subsection SPARC VIS Built-in Functions 9736 9737GCC supports SIMD operations on the SPARC using both the generic vector 9738extensions (@pxref{Vector Extensions}) as well as built-in functions for 9739the SPARC Visual Instruction Set (VIS). When you use the @option{-mvis} 9740switch, the VIS extension is exposed as the following built-in functions: 9741 9742@smallexample 9743typedef int v2si __attribute__ ((vector_size (8))); 9744typedef short v4hi __attribute__ ((vector_size (8))); 9745typedef short v2hi __attribute__ ((vector_size (4))); 9746typedef char v8qi __attribute__ ((vector_size (8))); 9747typedef char v4qi __attribute__ ((vector_size (4))); 9748 9749void * __builtin_vis_alignaddr (void *, long); 9750int64_t __builtin_vis_faligndatadi (int64_t, int64_t); 9751v2si __builtin_vis_faligndatav2si (v2si, v2si); 9752v4hi __builtin_vis_faligndatav4hi (v4si, v4si); 9753v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi); 9754 9755v4hi __builtin_vis_fexpand (v4qi); 9756 9757v4hi __builtin_vis_fmul8x16 (v4qi, v4hi); 9758v4hi __builtin_vis_fmul8x16au (v4qi, v4hi); 9759v4hi __builtin_vis_fmul8x16al (v4qi, v4hi); 9760v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi); 9761v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi); 9762v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi); 9763v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi); 9764 9765v4qi __builtin_vis_fpack16 (v4hi); 9766v8qi __builtin_vis_fpack32 (v2si, v2si); 9767v2hi __builtin_vis_fpackfix (v2si); 9768v8qi __builtin_vis_fpmerge (v4qi, v4qi); 9769 9770int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t); 9771@end smallexample 9772 9773@node Target Format Checks 9774@section Format Checks Specific to Particular Target Machines 9775 9776For some target machines, GCC supports additional options to the 9777format attribute 9778(@pxref{Function Attributes,,Declaring Attributes of Functions}). 9779 9780@menu 9781* Solaris Format Checks:: 9782@end menu 9783 9784@node Solaris Format Checks 9785@subsection Solaris Format Checks 9786 9787Solaris targets support the @code{cmn_err} (or @code{__cmn_err__}) format 9788check. @code{cmn_err} accepts a subset of the standard @code{printf} 9789conversions, and the two-argument @code{%b} conversion for displaying 9790bit-fields. See the Solaris man page for @code{cmn_err} for more information. 9791 9792@node Pragmas 9793@section Pragmas Accepted by GCC 9794@cindex pragmas 9795@cindex #pragma 9796 9797GCC supports several types of pragmas, primarily in order to compile 9798code originally written for other compilers. Note that in general 9799we do not recommend the use of pragmas; @xref{Function Attributes}, 9800for further explanation. 9801 9802@menu 9803* ARM Pragmas:: 9804* M32C Pragmas:: 9805* RS/6000 and PowerPC Pragmas:: 9806* Darwin Pragmas:: 9807* Solaris Pragmas:: 9808* Symbol-Renaming Pragmas:: 9809* Structure-Packing Pragmas:: 9810* Weak Pragmas:: 9811* Diagnostic Pragmas:: 9812* Visibility Pragmas:: 9813@end menu 9814 9815@node ARM Pragmas 9816@subsection ARM Pragmas 9817 9818The ARM target defines pragmas for controlling the default addition of 9819@code{long_call} and @code{short_call} attributes to functions. 9820@xref{Function Attributes}, for information about the effects of these 9821attributes. 9822 9823@table @code 9824@item long_calls 9825@cindex pragma, long_calls 9826Set all subsequent functions to have the @code{long_call} attribute. 9827 9828@item no_long_calls 9829@cindex pragma, no_long_calls 9830Set all subsequent functions to have the @code{short_call} attribute. 9831 9832@item long_calls_off 9833@cindex pragma, long_calls_off 9834Do not affect the @code{long_call} or @code{short_call} attributes of 9835subsequent functions. 9836@end table 9837 9838@node M32C Pragmas 9839@subsection M32C Pragmas 9840 9841@table @code 9842@item memregs @var{number} 9843@cindex pragma, memregs 9844Overrides the command line option @code{-memregs=} for the current 9845file. Use with care! This pragma must be before any function in the 9846file, and mixing different memregs values in different objects may 9847make them incompatible. This pragma is useful when a 9848performance-critical function uses a memreg for temporary values, 9849as it may allow you to reduce the number of memregs used. 9850 9851@end table 9852 9853@node RS/6000 and PowerPC Pragmas 9854@subsection RS/6000 and PowerPC Pragmas 9855 9856The RS/6000 and PowerPC targets define one pragma for controlling 9857whether or not the @code{longcall} attribute is added to function 9858declarations by default. This pragma overrides the @option{-mlongcall} 9859option, but not the @code{longcall} and @code{shortcall} attributes. 9860@xref{RS/6000 and PowerPC Options}, for more information about when long 9861calls are and are not necessary. 9862 9863@table @code 9864@item longcall (1) 9865@cindex pragma, longcall 9866Apply the @code{longcall} attribute to all subsequent function 9867declarations. 9868 9869@item longcall (0) 9870Do not apply the @code{longcall} attribute to subsequent function 9871declarations. 9872@end table 9873 9874@c Describe c4x pragmas here. 9875@c Describe h8300 pragmas here. 9876@c Describe sh pragmas here. 9877@c Describe v850 pragmas here. 9878 9879@node Darwin Pragmas 9880@subsection Darwin Pragmas 9881 9882The following pragmas are available for all architectures running the 9883Darwin operating system. These are useful for compatibility with other 9884Mac OS compilers. 9885 9886@table @code 9887@item mark @var{tokens}@dots{} 9888@cindex pragma, mark 9889This pragma is accepted, but has no effect. 9890 9891@item options align=@var{alignment} 9892@cindex pragma, options align 9893This pragma sets the alignment of fields in structures. The values of 9894@var{alignment} may be @code{mac68k}, to emulate m68k alignment, or 9895@code{power}, to emulate PowerPC alignment. Uses of this pragma nest 9896properly; to restore the previous setting, use @code{reset} for the 9897@var{alignment}. 9898 9899@item segment @var{tokens}@dots{} 9900@cindex pragma, segment 9901This pragma is accepted, but has no effect. 9902 9903@item unused (@var{var} [, @var{var}]@dots{}) 9904@cindex pragma, unused 9905This pragma declares variables to be possibly unused. GCC will not 9906produce warnings for the listed variables. The effect is similar to 9907that of the @code{unused} attribute, except that this pragma may appear 9908anywhere within the variables' scopes. 9909@end table 9910 9911@node Solaris Pragmas 9912@subsection Solaris Pragmas 9913 9914The Solaris target supports @code{#pragma redefine_extname} 9915(@pxref{Symbol-Renaming Pragmas}). It also supports additional 9916@code{#pragma} directives for compatibility with the system compiler. 9917 9918@table @code 9919@item align @var{alignment} (@var{variable} [, @var{variable}]...) 9920@cindex pragma, align 9921 9922Increase the minimum alignment of each @var{variable} to @var{alignment}. 9923This is the same as GCC's @code{aligned} attribute @pxref{Variable 9924Attributes}). Macro expansion occurs on the arguments to this pragma 9925when compiling C. It does not currently occur when compiling C++, but 9926this is a bug which may be fixed in a future release. 9927 9928@item fini (@var{function} [, @var{function}]...) 9929@cindex pragma, fini 9930 9931This pragma causes each listed @var{function} to be called after 9932main, or during shared module unloading, by adding a call to the 9933@code{.fini} section. 9934 9935@item init (@var{function} [, @var{function}]...) 9936@cindex pragma, init 9937 9938This pragma causes each listed @var{function} to be called during 9939initialization (before @code{main}) or during shared module loading, by 9940adding a call to the @code{.init} section. 9941 9942@end table 9943 9944@node Symbol-Renaming Pragmas 9945@subsection Symbol-Renaming Pragmas 9946 9947For compatibility with the Solaris and Tru64 UNIX system headers, GCC 9948supports two @code{#pragma} directives which change the name used in 9949assembly for a given declaration. These pragmas are only available on 9950platforms whose system headers need them. To get this effect on all 9951platforms supported by GCC, use the asm labels extension (@pxref{Asm 9952Labels}). 9953 9954@table @code 9955@item redefine_extname @var{oldname} @var{newname} 9956@cindex pragma, redefine_extname 9957 9958This pragma gives the C function @var{oldname} the assembly symbol 9959@var{newname}. The preprocessor macro @code{__PRAGMA_REDEFINE_EXTNAME} 9960will be defined if this pragma is available (currently only on 9961Solaris). 9962 9963@item extern_prefix @var{string} 9964@cindex pragma, extern_prefix 9965 9966This pragma causes all subsequent external function and variable 9967declarations to have @var{string} prepended to their assembly symbols. 9968This effect may be terminated with another @code{extern_prefix} pragma 9969whose argument is an empty string. The preprocessor macro 9970@code{__PRAGMA_EXTERN_PREFIX} will be defined if this pragma is 9971available (currently only on Tru64 UNIX)@. 9972@end table 9973 9974These pragmas and the asm labels extension interact in a complicated 9975manner. Here are some corner cases you may want to be aware of. 9976 9977@enumerate 9978@item Both pragmas silently apply only to declarations with external 9979linkage. Asm labels do not have this restriction. 9980 9981@item In C++, both pragmas silently apply only to declarations with 9982``C'' linkage. Again, asm labels do not have this restriction. 9983 9984@item If any of the three ways of changing the assembly name of a 9985declaration is applied to a declaration whose assembly name has 9986already been determined (either by a previous use of one of these 9987features, or because the compiler needed the assembly name in order to 9988generate code), and the new name is different, a warning issues and 9989the name does not change. 9990 9991@item The @var{oldname} used by @code{#pragma redefine_extname} is 9992always the C-language name. 9993 9994@item If @code{#pragma extern_prefix} is in effect, and a declaration 9995occurs with an asm label attached, the prefix is silently ignored for 9996that declaration. 9997 9998@item If @code{#pragma extern_prefix} and @code{#pragma redefine_extname} 9999apply to the same declaration, whichever triggered first wins, and a 10000warning issues if they contradict each other. (We would like to have 10001@code{#pragma redefine_extname} always win, for consistency with asm 10002labels, but if @code{#pragma extern_prefix} triggers first we have no 10003way of knowing that that happened.) 10004@end enumerate 10005 10006@node Structure-Packing Pragmas 10007@subsection Structure-Packing Pragmas 10008 10009For compatibility with Win32, GCC supports a set of @code{#pragma} 10010directives which change the maximum alignment of members of structures 10011(other than zero-width bitfields), unions, and classes subsequently 10012defined. The @var{n} value below always is required to be a small power 10013of two and specifies the new alignment in bytes. 10014 10015@enumerate 10016@item @code{#pragma pack(@var{n})} simply sets the new alignment. 10017@item @code{#pragma pack()} sets the alignment to the one that was in 10018effect when compilation started (see also command line option 10019@option{-fpack-struct[=<n>]} @pxref{Code Gen Options}). 10020@item @code{#pragma pack(push[,@var{n}])} pushes the current alignment 10021setting on an internal stack and then optionally sets the new alignment. 10022@item @code{#pragma pack(pop)} restores the alignment setting to the one 10023saved at the top of the internal stack (and removes that stack entry). 10024Note that @code{#pragma pack([@var{n}])} does not influence this internal 10025stack; thus it is possible to have @code{#pragma pack(push)} followed by 10026multiple @code{#pragma pack(@var{n})} instances and finalized by a single 10027@code{#pragma pack(pop)}. 10028@end enumerate 10029 10030Some targets, e.g. i386 and powerpc, support the @code{ms_struct} 10031@code{#pragma} which lays out a structure as the documented 10032@code{__attribute__ ((ms_struct))}. 10033@enumerate 10034@item @code{#pragma ms_struct on} turns on the layout for structures 10035declared. 10036@item @code{#pragma ms_struct off} turns off the layout for structures 10037declared. 10038@item @code{#pragma ms_struct reset} goes back to the default layout. 10039@end enumerate 10040 10041@node Weak Pragmas 10042@subsection Weak Pragmas 10043 10044For compatibility with SVR4, GCC supports a set of @code{#pragma} 10045directives for declaring symbols to be weak, and defining weak 10046aliases. 10047 10048@table @code 10049@item #pragma weak @var{symbol} 10050@cindex pragma, weak 10051This pragma declares @var{symbol} to be weak, as if the declaration 10052had the attribute of the same name. The pragma may appear before 10053or after the declaration of @var{symbol}, but must appear before 10054either its first use or its definition. It is not an error for 10055@var{symbol} to never be defined at all. 10056 10057@item #pragma weak @var{symbol1} = @var{symbol2} 10058This pragma declares @var{symbol1} to be a weak alias of @var{symbol2}. 10059It is an error if @var{symbol2} is not defined in the current 10060translation unit. 10061@end table 10062 10063@node Diagnostic Pragmas 10064@subsection Diagnostic Pragmas 10065 10066GCC allows the user to selectively enable or disable certain types of 10067diagnostics, and change the kind of the diagnostic. For example, a 10068project's policy might require that all sources compile with 10069@option{-Werror} but certain files might have exceptions allowing 10070specific types of warnings. Or, a project might selectively enable 10071diagnostics and treat them as errors depending on which preprocessor 10072macros are defined. 10073 10074@table @code 10075@item #pragma GCC diagnostic @var{kind} @var{option} 10076@cindex pragma, diagnostic 10077 10078Modifies the disposition of a diagnostic. Note that not all 10079diagnostics are modifiable; at the moment only warnings (normally 10080controlled by @samp{-W...}) can be controlled, and not all of them. 10081Use @option{-fdiagnostics-show-option} to determine which diagnostics 10082are controllable and which option controls them. 10083 10084@var{kind} is @samp{error} to treat this diagnostic as an error, 10085@samp{warning} to treat it like a warning (even if @option{-Werror} is 10086in effect), or @samp{ignored} if the diagnostic is to be ignored. 10087@var{option} is a double quoted string which matches the command line 10088option. 10089 10090@example 10091#pragma GCC diagnostic warning "-Wformat" 10092#pragma GCC diagnostic error "-Wformat" 10093#pragma GCC diagnostic ignored "-Wformat" 10094@end example 10095 10096Note that these pragmas override any command line options. Also, 10097while it is syntactically valid to put these pragmas anywhere in your 10098sources, the only supported location for them is before any data or 10099functions are defined. Doing otherwise may result in unpredictable 10100results depending on how the optimizer manages your sources. If the 10101same option is listed multiple times, the last one specified is the 10102one that is in effect. This pragma is not intended to be a general 10103purpose replacement for command line options, but for implementing 10104strict control over project policies. 10105 10106@end table 10107 10108@node Visibility Pragmas 10109@subsection Visibility Pragmas 10110 10111@table @code 10112@item #pragma GCC visibility push(@var{visibility}) 10113@itemx #pragma GCC visibility pop 10114@cindex pragma, visibility 10115 10116This pragma allows the user to set the visibility for multiple 10117declarations without having to give each a visibility attribute 10118@xref{Function Attributes}, for more information about visibility and 10119the attribute syntax. 10120 10121In C++, @samp{#pragma GCC visibility} affects only namespace-scope 10122declarations. Class members and template specializations are not 10123affected; if you want to override the visibility for a particular 10124member or instantiation, you must use an attribute. 10125 10126@end table 10127 10128@node Unnamed Fields 10129@section Unnamed struct/union fields within structs/unions 10130@cindex struct 10131@cindex union 10132 10133For compatibility with other compilers, GCC allows you to define 10134a structure or union that contains, as fields, structures and unions 10135without names. For example: 10136 10137@smallexample 10138struct @{ 10139 int a; 10140 union @{ 10141 int b; 10142 float c; 10143 @}; 10144 int d; 10145@} foo; 10146@end smallexample 10147 10148In this example, the user would be able to access members of the unnamed 10149union with code like @samp{foo.b}. Note that only unnamed structs and 10150unions are allowed, you may not have, for example, an unnamed 10151@code{int}. 10152 10153You must never create such structures that cause ambiguous field definitions. 10154For example, this structure: 10155 10156@smallexample 10157struct @{ 10158 int a; 10159 struct @{ 10160 int a; 10161 @}; 10162@} foo; 10163@end smallexample 10164 10165It is ambiguous which @code{a} is being referred to with @samp{foo.a}. 10166Such constructs are not supported and must be avoided. In the future, 10167such constructs may be detected and treated as compilation errors. 10168 10169@opindex fms-extensions 10170Unless @option{-fms-extensions} is used, the unnamed field must be a 10171structure or union definition without a tag (for example, @samp{struct 10172@{ int a; @};}). If @option{-fms-extensions} is used, the field may 10173also be a definition with a tag such as @samp{struct foo @{ int a; 10174@};}, a reference to a previously defined structure or union such as 10175@samp{struct foo;}, or a reference to a @code{typedef} name for a 10176previously defined structure or union type. 10177 10178@node Thread-Local 10179@section Thread-Local Storage 10180@cindex Thread-Local Storage 10181@cindex @acronym{TLS} 10182@cindex __thread 10183 10184Thread-local storage (@acronym{TLS}) is a mechanism by which variables 10185are allocated such that there is one instance of the variable per extant 10186thread. The run-time model GCC uses to implement this originates 10187in the IA-64 processor-specific ABI, but has since been migrated 10188to other processors as well. It requires significant support from 10189the linker (@command{ld}), dynamic linker (@command{ld.so}), and 10190system libraries (@file{libc.so} and @file{libpthread.so}), so it 10191is not available everywhere. 10192 10193At the user level, the extension is visible with a new storage 10194class keyword: @code{__thread}. For example: 10195 10196@smallexample 10197__thread int i; 10198extern __thread struct state s; 10199static __thread char *p; 10200@end smallexample 10201 10202The @code{__thread} specifier may be used alone, with the @code{extern} 10203or @code{static} specifiers, but with no other storage class specifier. 10204When used with @code{extern} or @code{static}, @code{__thread} must appear 10205immediately after the other storage class specifier. 10206 10207The @code{__thread} specifier may be applied to any global, file-scoped 10208static, function-scoped static, or static data member of a class. It may 10209not be applied to block-scoped automatic or non-static data member. 10210 10211When the address-of operator is applied to a thread-local variable, it is 10212evaluated at run-time and returns the address of the current thread's 10213instance of that variable. An address so obtained may be used by any 10214thread. When a thread terminates, any pointers to thread-local variables 10215in that thread become invalid. 10216 10217No static initialization may refer to the address of a thread-local variable. 10218 10219In C++, if an initializer is present for a thread-local variable, it must 10220be a @var{constant-expression}, as defined in 5.19.2 of the ANSI/ISO C++ 10221standard. 10222 10223See @uref{http://people.redhat.com/drepper/tls.pdf, 10224ELF Handling For Thread-Local Storage} for a detailed explanation of 10225the four thread-local storage addressing models, and how the run-time 10226is expected to function. 10227 10228@menu 10229* C99 Thread-Local Edits:: 10230* C++98 Thread-Local Edits:: 10231@end menu 10232 10233@node C99 Thread-Local Edits 10234@subsection ISO/IEC 9899:1999 Edits for Thread-Local Storage 10235 10236The following are a set of changes to ISO/IEC 9899:1999 (aka C99) 10237that document the exact semantics of the language extension. 10238 10239@itemize @bullet 10240@item 10241@cite{5.1.2 Execution environments} 10242 10243Add new text after paragraph 1 10244 10245@quotation 10246Within either execution environment, a @dfn{thread} is a flow of 10247control within a program. It is implementation defined whether 10248or not there may be more than one thread associated with a program. 10249It is implementation defined how threads beyond the first are 10250created, the name and type of the function called at thread 10251startup, and how threads may be terminated. However, objects 10252with thread storage duration shall be initialized before thread 10253startup. 10254@end quotation 10255 10256@item 10257@cite{6.2.4 Storage durations of objects} 10258 10259Add new text before paragraph 3 10260 10261@quotation 10262An object whose identifier is declared with the storage-class 10263specifier @w{@code{__thread}} has @dfn{thread storage duration}. 10264Its lifetime is the entire execution of the thread, and its 10265stored value is initialized only once, prior to thread startup. 10266@end quotation 10267 10268@item 10269@cite{6.4.1 Keywords} 10270 10271Add @code{__thread}. 10272 10273@item 10274@cite{6.7.1 Storage-class specifiers} 10275 10276Add @code{__thread} to the list of storage class specifiers in 10277paragraph 1. 10278 10279Change paragraph 2 to 10280 10281@quotation 10282With the exception of @code{__thread}, at most one storage-class 10283specifier may be given [@dots{}]. The @code{__thread} specifier may 10284be used alone, or immediately following @code{extern} or 10285@code{static}. 10286@end quotation 10287 10288Add new text after paragraph 6 10289 10290@quotation 10291The declaration of an identifier for a variable that has 10292block scope that specifies @code{__thread} shall also 10293specify either @code{extern} or @code{static}. 10294 10295The @code{__thread} specifier shall be used only with 10296variables. 10297@end quotation 10298@end itemize 10299 10300@node C++98 Thread-Local Edits 10301@subsection ISO/IEC 14882:1998 Edits for Thread-Local Storage 10302 10303The following are a set of changes to ISO/IEC 14882:1998 (aka C++98) 10304that document the exact semantics of the language extension. 10305 10306@itemize @bullet 10307@item 10308@b{[intro.execution]} 10309 10310New text after paragraph 4 10311 10312@quotation 10313A @dfn{thread} is a flow of control within the abstract machine. 10314It is implementation defined whether or not there may be more than 10315one thread. 10316@end quotation 10317 10318New text after paragraph 7 10319 10320@quotation 10321It is unspecified whether additional action must be taken to 10322ensure when and whether side effects are visible to other threads. 10323@end quotation 10324 10325@item 10326@b{[lex.key]} 10327 10328Add @code{__thread}. 10329 10330@item 10331@b{[basic.start.main]} 10332 10333Add after paragraph 5 10334 10335@quotation 10336The thread that begins execution at the @code{main} function is called 10337the @dfn{main thread}. It is implementation defined how functions 10338beginning threads other than the main thread are designated or typed. 10339A function so designated, as well as the @code{main} function, is called 10340a @dfn{thread startup function}. It is implementation defined what 10341happens if a thread startup function returns. It is implementation 10342defined what happens to other threads when any thread calls @code{exit}. 10343@end quotation 10344 10345@item 10346@b{[basic.start.init]} 10347 10348Add after paragraph 4 10349 10350@quotation 10351The storage for an object of thread storage duration shall be 10352statically initialized before the first statement of the thread startup 10353function. An object of thread storage duration shall not require 10354dynamic initialization. 10355@end quotation 10356 10357@item 10358@b{[basic.start.term]} 10359 10360Add after paragraph 3 10361 10362@quotation 10363The type of an object with thread storage duration shall not have a 10364non-trivial destructor, nor shall it be an array type whose elements 10365(directly or indirectly) have non-trivial destructors. 10366@end quotation 10367 10368@item 10369@b{[basic.stc]} 10370 10371Add ``thread storage duration'' to the list in paragraph 1. 10372 10373Change paragraph 2 10374 10375@quotation 10376Thread, static, and automatic storage durations are associated with 10377objects introduced by declarations [@dots{}]. 10378@end quotation 10379 10380Add @code{__thread} to the list of specifiers in paragraph 3. 10381 10382@item 10383@b{[basic.stc.thread]} 10384 10385New section before @b{[basic.stc.static]} 10386 10387@quotation 10388The keyword @code{__thread} applied to a non-local object gives the 10389object thread storage duration. 10390 10391A local variable or class data member declared both @code{static} 10392and @code{__thread} gives the variable or member thread storage 10393duration. 10394@end quotation 10395 10396@item 10397@b{[basic.stc.static]} 10398 10399Change paragraph 1 10400 10401@quotation 10402All objects which have neither thread storage duration, dynamic 10403storage duration nor are local [@dots{}]. 10404@end quotation 10405 10406@item 10407@b{[dcl.stc]} 10408 10409Add @code{__thread} to the list in paragraph 1. 10410 10411Change paragraph 1 10412 10413@quotation 10414With the exception of @code{__thread}, at most one 10415@var{storage-class-specifier} shall appear in a given 10416@var{decl-specifier-seq}. The @code{__thread} specifier may 10417be used alone, or immediately following the @code{extern} or 10418@code{static} specifiers. [@dots{}] 10419@end quotation 10420 10421Add after paragraph 5 10422 10423@quotation 10424The @code{__thread} specifier can be applied only to the names of objects 10425and to anonymous unions. 10426@end quotation 10427 10428@item 10429@b{[class.mem]} 10430 10431Add after paragraph 6 10432 10433@quotation 10434Non-@code{static} members shall not be @code{__thread}. 10435@end quotation 10436@end itemize 10437 10438@node Binary constants 10439@section Binary constants using the @samp{0b} prefix 10440@cindex Binary constants using the @samp{0b} prefix 10441 10442Integer constants can be written as binary constants, consisting of a 10443sequence of @samp{0} and @samp{1} digits, prefixed by @samp{0b} or 10444@samp{0B}. This is particularly useful in environments that operate a 10445lot on the bit-level (like microcontrollers). 10446 10447The following statements are identical: 10448 10449@smallexample 10450i = 42; 10451i = 0x2a; 10452i = 052; 10453i = 0b101010; 10454@end smallexample 10455 10456The type of these constants follows the same rules as for octal or 10457hexadecimal integer constants, so suffixes like @samp{L} or @samp{UL} 10458can be applied. 10459 10460@node C++ Extensions 10461@chapter Extensions to the C++ Language 10462@cindex extensions, C++ language 10463@cindex C++ language extensions 10464 10465The GNU compiler provides these extensions to the C++ language (and you 10466can also use most of the C language extensions in your C++ programs). If you 10467want to write code that checks whether these features are available, you can 10468test for the GNU compiler the same way as for C programs: check for a 10469predefined macro @code{__GNUC__}. You can also use @code{__GNUG__} to 10470test specifically for GNU C++ (@pxref{Common Predefined Macros,, 10471Predefined Macros,cpp,The GNU C Preprocessor}). 10472 10473@menu 10474* Volatiles:: What constitutes an access to a volatile object. 10475* Restricted Pointers:: C99 restricted pointers and references. 10476* Vague Linkage:: Where G++ puts inlines, vtables and such. 10477* C++ Interface:: You can use a single C++ header file for both 10478 declarations and definitions. 10479* Template Instantiation:: Methods for ensuring that exactly one copy of 10480 each needed template instantiation is emitted. 10481* Bound member functions:: You can extract a function pointer to the 10482 method denoted by a @samp{->*} or @samp{.*} expression. 10483* C++ Attributes:: Variable, function, and type attributes for C++ only. 10484* Namespace Association:: Strong using-directives for namespace association. 10485* Java Exceptions:: Tweaking exception handling to work with Java. 10486* Deprecated Features:: Things will disappear from g++. 10487* Backwards Compatibility:: Compatibilities with earlier definitions of C++. 10488@end menu 10489 10490@node Volatiles 10491@section When is a Volatile Object Accessed? 10492@cindex accessing volatiles 10493@cindex volatile read 10494@cindex volatile write 10495@cindex volatile access 10496 10497Both the C and C++ standard have the concept of volatile objects. These 10498are normally accessed by pointers and used for accessing hardware. The 10499standards encourage compilers to refrain from optimizations concerning 10500accesses to volatile objects. The C standard leaves it implementation 10501defined as to what constitutes a volatile access. The C++ standard omits 10502to specify this, except to say that C++ should behave in a similar manner 10503to C with respect to volatiles, where possible. The minimum either 10504standard specifies is that at a sequence point all previous accesses to 10505volatile objects have stabilized and no subsequent accesses have 10506occurred. Thus an implementation is free to reorder and combine 10507volatile accesses which occur between sequence points, but cannot do so 10508for accesses across a sequence point. The use of volatiles does not 10509allow you to violate the restriction on updating objects multiple times 10510within a sequence point. 10511 10512@xref{Qualifiers implementation, , Volatile qualifier and the C compiler}. 10513 10514The behavior differs slightly between C and C++ in the non-obvious cases: 10515 10516@smallexample 10517volatile int *src = @var{somevalue}; 10518*src; 10519@end smallexample 10520 10521With C, such expressions are rvalues, and GCC interprets this either as a 10522read of the volatile object being pointed to or only as request to evaluate 10523the side-effects. The C++ standard specifies that such expressions do not 10524undergo lvalue to rvalue conversion, and that the type of the dereferenced 10525object may be incomplete. The C++ standard does not specify explicitly 10526that it is this lvalue to rvalue conversion which may be responsible for 10527causing an access. However, there is reason to believe that it is, 10528because otherwise certain simple expressions become undefined. However, 10529because it would surprise most programmers, G++ treats dereferencing a 10530pointer to volatile object of complete type when the value is unused as 10531GCC would do for an equivalent type in C. When the object has incomplete 10532type, G++ issues a warning; if you wish to force an error, you must 10533force a conversion to rvalue with, for instance, a static cast. 10534 10535When using a reference to volatile, G++ does not treat equivalent 10536expressions as accesses to volatiles, but instead issues a warning that 10537no volatile is accessed. The rationale for this is that otherwise it 10538becomes difficult to determine where volatile access occur, and not 10539possible to ignore the return value from functions returning volatile 10540references. Again, if you wish to force a read, cast the reference to 10541an rvalue. 10542 10543@node Restricted Pointers 10544@section Restricting Pointer Aliasing 10545@cindex restricted pointers 10546@cindex restricted references 10547@cindex restricted this pointer 10548 10549As with the C front end, G++ understands the C99 feature of restricted pointers, 10550specified with the @code{__restrict__}, or @code{__restrict} type 10551qualifier. Because you cannot compile C++ by specifying the @option{-std=c99} 10552language flag, @code{restrict} is not a keyword in C++. 10553 10554In addition to allowing restricted pointers, you can specify restricted 10555references, which indicate that the reference is not aliased in the local 10556context. 10557 10558@smallexample 10559void fn (int *__restrict__ rptr, int &__restrict__ rref) 10560@{ 10561 /* @r{@dots{}} */ 10562@} 10563@end smallexample 10564 10565@noindent 10566In the body of @code{fn}, @var{rptr} points to an unaliased integer and 10567@var{rref} refers to a (different) unaliased integer. 10568 10569You may also specify whether a member function's @var{this} pointer is 10570unaliased by using @code{__restrict__} as a member function qualifier. 10571 10572@smallexample 10573void T::fn () __restrict__ 10574@{ 10575 /* @r{@dots{}} */ 10576@} 10577@end smallexample 10578 10579@noindent 10580Within the body of @code{T::fn}, @var{this} will have the effective 10581definition @code{T *__restrict__ const this}. Notice that the 10582interpretation of a @code{__restrict__} member function qualifier is 10583different to that of @code{const} or @code{volatile} qualifier, in that it 10584is applied to the pointer rather than the object. This is consistent with 10585other compilers which implement restricted pointers. 10586 10587As with all outermost parameter qualifiers, @code{__restrict__} is 10588ignored in function definition matching. This means you only need to 10589specify @code{__restrict__} in a function definition, rather than 10590in a function prototype as well. 10591 10592@node Vague Linkage 10593@section Vague Linkage 10594@cindex vague linkage 10595 10596There are several constructs in C++ which require space in the object 10597file but are not clearly tied to a single translation unit. We say that 10598these constructs have ``vague linkage''. Typically such constructs are 10599emitted wherever they are needed, though sometimes we can be more 10600clever. 10601 10602@table @asis 10603@item Inline Functions 10604Inline functions are typically defined in a header file which can be 10605included in many different compilations. Hopefully they can usually be 10606inlined, but sometimes an out-of-line copy is necessary, if the address 10607of the function is taken or if inlining fails. In general, we emit an 10608out-of-line copy in all translation units where one is needed. As an 10609exception, we only emit inline virtual functions with the vtable, since 10610it will always require a copy. 10611 10612Local static variables and string constants used in an inline function 10613are also considered to have vague linkage, since they must be shared 10614between all inlined and out-of-line instances of the function. 10615 10616@item VTables 10617@cindex vtable 10618C++ virtual functions are implemented in most compilers using a lookup 10619table, known as a vtable. The vtable contains pointers to the virtual 10620functions provided by a class, and each object of the class contains a 10621pointer to its vtable (or vtables, in some multiple-inheritance 10622situations). If the class declares any non-inline, non-pure virtual 10623functions, the first one is chosen as the ``key method'' for the class, 10624and the vtable is only emitted in the translation unit where the key 10625method is defined. 10626 10627@emph{Note:} If the chosen key method is later defined as inline, the 10628vtable will still be emitted in every translation unit which defines it. 10629Make sure that any inline virtuals are declared inline in the class 10630body, even if they are not defined there. 10631 10632@item type_info objects 10633@cindex type_info 10634@cindex RTTI 10635C++ requires information about types to be written out in order to 10636implement @samp{dynamic_cast}, @samp{typeid} and exception handling. 10637For polymorphic classes (classes with virtual functions), the type_info 10638object is written out along with the vtable so that @samp{dynamic_cast} 10639can determine the dynamic type of a class object at runtime. For all 10640other types, we write out the type_info object when it is used: when 10641applying @samp{typeid} to an expression, throwing an object, or 10642referring to a type in a catch clause or exception specification. 10643 10644@item Template Instantiations 10645Most everything in this section also applies to template instantiations, 10646but there are other options as well. 10647@xref{Template Instantiation,,Where's the Template?}. 10648 10649@end table 10650 10651When used with GNU ld version 2.8 or later on an ELF system such as 10652GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of 10653these constructs will be discarded at link time. This is known as 10654COMDAT support. 10655 10656On targets that don't support COMDAT, but do support weak symbols, GCC 10657will use them. This way one copy will override all the others, but 10658the unused copies will still take up space in the executable. 10659 10660For targets which do not support either COMDAT or weak symbols, 10661most entities with vague linkage will be emitted as local symbols to 10662avoid duplicate definition errors from the linker. This will not happen 10663for local statics in inlines, however, as having multiple copies will 10664almost certainly break things. 10665 10666@xref{C++ Interface,,Declarations and Definitions in One Header}, for 10667another way to control placement of these constructs. 10668 10669@node C++ Interface 10670@section #pragma interface and implementation 10671 10672@cindex interface and implementation headers, C++ 10673@cindex C++ interface and implementation headers 10674@cindex pragmas, interface and implementation 10675 10676@code{#pragma interface} and @code{#pragma implementation} provide the 10677user with a way of explicitly directing the compiler to emit entities 10678with vague linkage (and debugging information) in a particular 10679translation unit. 10680 10681@emph{Note:} As of GCC 2.7.2, these @code{#pragma}s are not useful in 10682most cases, because of COMDAT support and the ``key method'' heuristic 10683mentioned in @ref{Vague Linkage}. Using them can actually cause your 10684program to grow due to unnecessary out-of-line copies of inline 10685functions. Currently (3.4) the only benefit of these 10686@code{#pragma}s is reduced duplication of debugging information, and 10687that should be addressed soon on DWARF 2 targets with the use of 10688COMDAT groups. 10689 10690@table @code 10691@item #pragma interface 10692@itemx #pragma interface "@var{subdir}/@var{objects}.h" 10693@kindex #pragma interface 10694Use this directive in @emph{header files} that define object classes, to save 10695space in most of the object files that use those classes. Normally, 10696local copies of certain information (backup copies of inline member 10697functions, debugging information, and the internal tables that implement 10698virtual functions) must be kept in each object file that includes class 10699definitions. You can use this pragma to avoid such duplication. When a 10700header file containing @samp{#pragma interface} is included in a 10701compilation, this auxiliary information will not be generated (unless 10702the main input source file itself uses @samp{#pragma implementation}). 10703Instead, the object files will contain references to be resolved at link 10704time. 10705 10706The second form of this directive is useful for the case where you have 10707multiple headers with the same name in different directories. If you 10708use this form, you must specify the same string to @samp{#pragma 10709implementation}. 10710 10711@item #pragma implementation 10712@itemx #pragma implementation "@var{objects}.h" 10713@kindex #pragma implementation 10714Use this pragma in a @emph{main input file}, when you want full output from 10715included header files to be generated (and made globally visible). The 10716included header file, in turn, should use @samp{#pragma interface}. 10717Backup copies of inline member functions, debugging information, and the 10718internal tables used to implement virtual functions are all generated in 10719implementation files. 10720 10721@cindex implied @code{#pragma implementation} 10722@cindex @code{#pragma implementation}, implied 10723@cindex naming convention, implementation headers 10724If you use @samp{#pragma implementation} with no argument, it applies to 10725an include file with the same basename@footnote{A file's @dfn{basename} 10726was the name stripped of all leading path information and of trailing 10727suffixes, such as @samp{.h} or @samp{.C} or @samp{.cc}.} as your source 10728file. For example, in @file{allclass.cc}, giving just 10729@samp{#pragma implementation} 10730by itself is equivalent to @samp{#pragma implementation "allclass.h"}. 10731 10732In versions of GNU C++ prior to 2.6.0 @file{allclass.h} was treated as 10733an implementation file whenever you would include it from 10734@file{allclass.cc} even if you never specified @samp{#pragma 10735implementation}. This was deemed to be more trouble than it was worth, 10736however, and disabled. 10737 10738Use the string argument if you want a single implementation file to 10739include code from multiple header files. (You must also use 10740@samp{#include} to include the header file; @samp{#pragma 10741implementation} only specifies how to use the file---it doesn't actually 10742include it.) 10743 10744There is no way to split up the contents of a single header file into 10745multiple implementation files. 10746@end table 10747 10748@cindex inlining and C++ pragmas 10749@cindex C++ pragmas, effect on inlining 10750@cindex pragmas in C++, effect on inlining 10751@samp{#pragma implementation} and @samp{#pragma interface} also have an 10752effect on function inlining. 10753 10754If you define a class in a header file marked with @samp{#pragma 10755interface}, the effect on an inline function defined in that class is 10756similar to an explicit @code{extern} declaration---the compiler emits 10757no code at all to define an independent version of the function. Its 10758definition is used only for inlining with its callers. 10759 10760@opindex fno-implement-inlines 10761Conversely, when you include the same header file in a main source file 10762that declares it as @samp{#pragma implementation}, the compiler emits 10763code for the function itself; this defines a version of the function 10764that can be found via pointers (or by callers compiled without 10765inlining). If all calls to the function can be inlined, you can avoid 10766emitting the function by compiling with @option{-fno-implement-inlines}. 10767If any calls were not inlined, you will get linker errors. 10768 10769@node Template Instantiation 10770@section Where's the Template? 10771@cindex template instantiation 10772 10773C++ templates are the first language feature to require more 10774intelligence from the environment than one usually finds on a UNIX 10775system. Somehow the compiler and linker have to make sure that each 10776template instance occurs exactly once in the executable if it is needed, 10777and not at all otherwise. There are two basic approaches to this 10778problem, which are referred to as the Borland model and the Cfront model. 10779 10780@table @asis 10781@item Borland model 10782Borland C++ solved the template instantiation problem by adding the code 10783equivalent of common blocks to their linker; the compiler emits template 10784instances in each translation unit that uses them, and the linker 10785collapses them together. The advantage of this model is that the linker 10786only has to consider the object files themselves; there is no external 10787complexity to worry about. This disadvantage is that compilation time 10788is increased because the template code is being compiled repeatedly. 10789Code written for this model tends to include definitions of all 10790templates in the header file, since they must be seen to be 10791instantiated. 10792 10793@item Cfront model 10794The AT&T C++ translator, Cfront, solved the template instantiation 10795problem by creating the notion of a template repository, an 10796automatically maintained place where template instances are stored. A 10797more modern version of the repository works as follows: As individual 10798object files are built, the compiler places any template definitions and 10799instantiations encountered in the repository. At link time, the link 10800wrapper adds in the objects in the repository and compiles any needed 10801instances that were not previously emitted. The advantages of this 10802model are more optimal compilation speed and the ability to use the 10803system linker; to implement the Borland model a compiler vendor also 10804needs to replace the linker. The disadvantages are vastly increased 10805complexity, and thus potential for error; for some code this can be 10806just as transparent, but in practice it can been very difficult to build 10807multiple programs in one directory and one program in multiple 10808directories. Code written for this model tends to separate definitions 10809of non-inline member templates into a separate file, which should be 10810compiled separately. 10811@end table 10812 10813When used with GNU ld version 2.8 or later on an ELF system such as 10814GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the 10815Borland model. On other systems, G++ implements neither automatic 10816model. 10817 10818A future version of G++ will support a hybrid model whereby the compiler 10819will emit any instantiations for which the template definition is 10820included in the compile, and store template definitions and 10821instantiation context information into the object file for the rest. 10822The link wrapper will extract that information as necessary and invoke 10823the compiler to produce the remaining instantiations. The linker will 10824then combine duplicate instantiations. 10825 10826In the mean time, you have the following options for dealing with 10827template instantiations: 10828 10829@enumerate 10830@item 10831@opindex frepo 10832Compile your template-using code with @option{-frepo}. The compiler will 10833generate files with the extension @samp{.rpo} listing all of the 10834template instantiations used in the corresponding object files which 10835could be instantiated there; the link wrapper, @samp{collect2}, will 10836then update the @samp{.rpo} files to tell the compiler where to place 10837those instantiations and rebuild any affected object files. The 10838link-time overhead is negligible after the first pass, as the compiler 10839will continue to place the instantiations in the same files. 10840 10841This is your best option for application code written for the Borland 10842model, as it will just work. Code written for the Cfront model will 10843need to be modified so that the template definitions are available at 10844one or more points of instantiation; usually this is as simple as adding 10845@code{#include <tmethods.cc>} to the end of each template header. 10846 10847For library code, if you want the library to provide all of the template 10848instantiations it needs, just try to link all of its object files 10849together; the link will fail, but cause the instantiations to be 10850generated as a side effect. Be warned, however, that this may cause 10851conflicts if multiple libraries try to provide the same instantiations. 10852For greater control, use explicit instantiation as described in the next 10853option. 10854 10855@item 10856@opindex fno-implicit-templates 10857Compile your code with @option{-fno-implicit-templates} to disable the 10858implicit generation of template instances, and explicitly instantiate 10859all the ones you use. This approach requires more knowledge of exactly 10860which instances you need than do the others, but it's less 10861mysterious and allows greater control. You can scatter the explicit 10862instantiations throughout your program, perhaps putting them in the 10863translation units where the instances are used or the translation units 10864that define the templates themselves; you can put all of the explicit 10865instantiations you need into one big file; or you can create small files 10866like 10867 10868@smallexample 10869#include "Foo.h" 10870#include "Foo.cc" 10871 10872template class Foo<int>; 10873template ostream& operator << 10874 (ostream&, const Foo<int>&); 10875@end smallexample 10876 10877for each of the instances you need, and create a template instantiation 10878library from those. 10879 10880If you are using Cfront-model code, you can probably get away with not 10881using @option{-fno-implicit-templates} when compiling files that don't 10882@samp{#include} the member template definitions. 10883 10884If you use one big file to do the instantiations, you may want to 10885compile it without @option{-fno-implicit-templates} so you get all of the 10886instances required by your explicit instantiations (but not by any 10887other files) without having to specify them as well. 10888 10889G++ has extended the template instantiation syntax given in the ISO 10890standard to allow forward declaration of explicit instantiations 10891(with @code{extern}), instantiation of the compiler support data for a 10892template class (i.e.@: the vtable) without instantiating any of its 10893members (with @code{inline}), and instantiation of only the static data 10894members of a template class, without the support data or member 10895functions (with (@code{static}): 10896 10897@smallexample 10898extern template int max (int, int); 10899inline template class Foo<int>; 10900static template class Foo<int>; 10901@end smallexample 10902 10903@item 10904Do nothing. Pretend G++ does implement automatic instantiation 10905management. Code written for the Borland model will work fine, but 10906each translation unit will contain instances of each of the templates it 10907uses. In a large program, this can lead to an unacceptable amount of code 10908duplication. 10909@end enumerate 10910 10911@node Bound member functions 10912@section Extracting the function pointer from a bound pointer to member function 10913@cindex pmf 10914@cindex pointer to member function 10915@cindex bound pointer to member function 10916 10917In C++, pointer to member functions (PMFs) are implemented using a wide 10918pointer of sorts to handle all the possible call mechanisms; the PMF 10919needs to store information about how to adjust the @samp{this} pointer, 10920and if the function pointed to is virtual, where to find the vtable, and 10921where in the vtable to look for the member function. If you are using 10922PMFs in an inner loop, you should really reconsider that decision. If 10923that is not an option, you can extract the pointer to the function that 10924would be called for a given object/PMF pair and call it directly inside 10925the inner loop, to save a bit of time. 10926 10927Note that you will still be paying the penalty for the call through a 10928function pointer; on most modern architectures, such a call defeats the 10929branch prediction features of the CPU@. This is also true of normal 10930virtual function calls. 10931 10932The syntax for this extension is 10933 10934@smallexample 10935extern A a; 10936extern int (A::*fp)(); 10937typedef int (*fptr)(A *); 10938 10939fptr p = (fptr)(a.*fp); 10940@end smallexample 10941 10942For PMF constants (i.e.@: expressions of the form @samp{&Klasse::Member}), 10943no object is needed to obtain the address of the function. They can be 10944converted to function pointers directly: 10945 10946@smallexample 10947fptr p1 = (fptr)(&A::foo); 10948@end smallexample 10949 10950@opindex Wno-pmf-conversions 10951You must specify @option{-Wno-pmf-conversions} to use this extension. 10952 10953@node C++ Attributes 10954@section C++-Specific Variable, Function, and Type Attributes 10955 10956Some attributes only make sense for C++ programs. 10957 10958@table @code 10959@item init_priority (@var{priority}) 10960@cindex init_priority attribute 10961 10962 10963In Standard C++, objects defined at namespace scope are guaranteed to be 10964initialized in an order in strict accordance with that of their definitions 10965@emph{in a given translation unit}. No guarantee is made for initializations 10966across translation units. However, GNU C++ allows users to control the 10967order of initialization of objects defined at namespace scope with the 10968@code{init_priority} attribute by specifying a relative @var{priority}, 10969a constant integral expression currently bounded between 101 and 65535 10970inclusive. Lower numbers indicate a higher priority. 10971 10972In the following example, @code{A} would normally be created before 10973@code{B}, but the @code{init_priority} attribute has reversed that order: 10974 10975@smallexample 10976Some_Class A __attribute__ ((init_priority (2000))); 10977Some_Class B __attribute__ ((init_priority (543))); 10978@end smallexample 10979 10980@noindent 10981Note that the particular values of @var{priority} do not matter; only their 10982relative ordering. 10983 10984@item java_interface 10985@cindex java_interface attribute 10986 10987This type attribute informs C++ that the class is a Java interface. It may 10988only be applied to classes declared within an @code{extern "Java"} block. 10989Calls to methods declared in this interface will be dispatched using GCJ's 10990interface table mechanism, instead of regular virtual table dispatch. 10991 10992@end table 10993 10994See also @xref{Namespace Association}. 10995 10996@node Namespace Association 10997@section Namespace Association 10998 10999@strong{Caution:} The semantics of this extension are not fully 11000defined. Users should refrain from using this extension as its 11001semantics may change subtly over time. It is possible that this 11002extension will be removed in future versions of G++. 11003 11004A using-directive with @code{__attribute ((strong))} is stronger 11005than a normal using-directive in two ways: 11006 11007@itemize @bullet 11008@item 11009Templates from the used namespace can be specialized and explicitly 11010instantiated as though they were members of the using namespace. 11011 11012@item 11013The using namespace is considered an associated namespace of all 11014templates in the used namespace for purposes of argument-dependent 11015name lookup. 11016@end itemize 11017 11018The used namespace must be nested within the using namespace so that 11019normal unqualified lookup works properly. 11020 11021This is useful for composing a namespace transparently from 11022implementation namespaces. For example: 11023 11024@smallexample 11025namespace std @{ 11026 namespace debug @{ 11027 template <class T> struct A @{ @}; 11028 @} 11029 using namespace debug __attribute ((__strong__)); 11030 template <> struct A<int> @{ @}; // @r{ok to specialize} 11031 11032 template <class T> void f (A<T>); 11033@} 11034 11035int main() 11036@{ 11037 f (std::A<float>()); // @r{lookup finds} std::f 11038 f (std::A<int>()); 11039@} 11040@end smallexample 11041 11042@node Java Exceptions 11043@section Java Exceptions 11044 11045The Java language uses a slightly different exception handling model 11046from C++. Normally, GNU C++ will automatically detect when you are 11047writing C++ code that uses Java exceptions, and handle them 11048appropriately. However, if C++ code only needs to execute destructors 11049when Java exceptions are thrown through it, GCC will guess incorrectly. 11050Sample problematic code is: 11051 11052@smallexample 11053 struct S @{ ~S(); @}; 11054 extern void bar(); // @r{is written in Java, and may throw exceptions} 11055 void foo() 11056 @{ 11057 S s; 11058 bar(); 11059 @} 11060@end smallexample 11061 11062@noindent 11063The usual effect of an incorrect guess is a link failure, complaining of 11064a missing routine called @samp{__gxx_personality_v0}. 11065 11066You can inform the compiler that Java exceptions are to be used in a 11067translation unit, irrespective of what it might think, by writing 11068@samp{@w{#pragma GCC java_exceptions}} at the head of the file. This 11069@samp{#pragma} must appear before any functions that throw or catch 11070exceptions, or run destructors when exceptions are thrown through them. 11071 11072You cannot mix Java and C++ exceptions in the same translation unit. It 11073is believed to be safe to throw a C++ exception from one file through 11074another file compiled for the Java exception model, or vice versa, but 11075there may be bugs in this area. 11076 11077@node Deprecated Features 11078@section Deprecated Features 11079 11080In the past, the GNU C++ compiler was extended to experiment with new 11081features, at a time when the C++ language was still evolving. Now that 11082the C++ standard is complete, some of those features are superseded by 11083superior alternatives. Using the old features might cause a warning in 11084some cases that the feature will be dropped in the future. In other 11085cases, the feature might be gone already. 11086 11087While the list below is not exhaustive, it documents some of the options 11088that are now deprecated: 11089 11090@table @code 11091@item -fexternal-templates 11092@itemx -falt-external-templates 11093These are two of the many ways for G++ to implement template 11094instantiation. @xref{Template Instantiation}. The C++ standard clearly 11095defines how template definitions have to be organized across 11096implementation units. G++ has an implicit instantiation mechanism that 11097should work just fine for standard-conforming code. 11098 11099@item -fstrict-prototype 11100@itemx -fno-strict-prototype 11101Previously it was possible to use an empty prototype parameter list to 11102indicate an unspecified number of parameters (like C), rather than no 11103parameters, as C++ demands. This feature has been removed, except where 11104it is required for backwards compatibility @xref{Backwards Compatibility}. 11105@end table 11106 11107G++ allows a virtual function returning @samp{void *} to be overridden 11108by one returning a different pointer type. This extension to the 11109covariant return type rules is now deprecated and will be removed from a 11110future version. 11111 11112The G++ minimum and maximum operators (@samp{<?} and @samp{>?}) and 11113their compound forms (@samp{<?=}) and @samp{>?=}) have been deprecated 11114and will be removed in a future version. Code using these operators 11115should be modified to use @code{std::min} and @code{std::max} instead. 11116 11117The named return value extension has been deprecated, and is now 11118removed from G++. 11119 11120The use of initializer lists with new expressions has been deprecated, 11121and is now removed from G++. 11122 11123Floating and complex non-type template parameters have been deprecated, 11124and are now removed from G++. 11125 11126The implicit typename extension has been deprecated and is now 11127removed from G++. 11128 11129The use of default arguments in function pointers, function typedefs 11130and other places where they are not permitted by the standard is 11131deprecated and will be removed from a future version of G++. 11132 11133G++ allows floating-point literals to appear in integral constant expressions, 11134e.g. @samp{ enum E @{ e = int(2.2 * 3.7) @} } 11135This extension is deprecated and will be removed from a future version. 11136 11137G++ allows static data members of const floating-point type to be declared 11138with an initializer in a class definition. The standard only allows 11139initializers for static members of const integral types and const 11140enumeration types so this extension has been deprecated and will be removed 11141from a future version. 11142 11143@node Backwards Compatibility 11144@section Backwards Compatibility 11145@cindex Backwards Compatibility 11146@cindex ARM [Annotated C++ Reference Manual] 11147 11148Now that there is a definitive ISO standard C++, G++ has a specification 11149to adhere to. The C++ language evolved over time, and features that 11150used to be acceptable in previous drafts of the standard, such as the ARM 11151[Annotated C++ Reference Manual], are no longer accepted. In order to allow 11152compilation of C++ written to such drafts, G++ contains some backwards 11153compatibilities. @emph{All such backwards compatibility features are 11154liable to disappear in future versions of G++.} They should be considered 11155deprecated @xref{Deprecated Features}. 11156 11157@table @code 11158@item For scope 11159If a variable is declared at for scope, it used to remain in scope until 11160the end of the scope which contained the for statement (rather than just 11161within the for scope). G++ retains this, but issues a warning, if such a 11162variable is accessed outside the for scope. 11163 11164@item Implicit C language 11165Old C system header files did not contain an @code{extern "C" @{@dots{}@}} 11166scope to set the language. On such systems, all header files are 11167implicitly scoped inside a C language scope. Also, an empty prototype 11168@code{()} will be treated as an unspecified number of arguments, rather 11169than no arguments, as C++ demands. 11170@end table 11171