1@c Copyright (C) 1988-2015 Free Software Foundation, Inc. 2@c This is part of the GCC manual. 3@c For copying conditions, see the file gcc.texi. 4 5@node Target Macros 6@chapter Target Description Macros and Functions 7@cindex machine description macros 8@cindex target description macros 9@cindex macros, target description 10@cindex @file{tm.h} macros 11 12In addition to the file @file{@var{machine}.md}, a machine description 13includes a C header file conventionally given the name 14@file{@var{machine}.h} and a C source file named @file{@var{machine}.c}. 15The header file defines numerous macros that convey the information 16about the target machine that does not fit into the scheme of the 17@file{.md} file. The file @file{tm.h} should be a link to 18@file{@var{machine}.h}. The header file @file{config.h} includes 19@file{tm.h} and most compiler source files include @file{config.h}. The 20source file defines a variable @code{targetm}, which is a structure 21containing pointers to functions and data relating to the target 22machine. @file{@var{machine}.c} should also contain their definitions, 23if they are not defined elsewhere in GCC, and other functions called 24through the macros defined in the @file{.h} file. 25 26@menu 27* Target Structure:: The @code{targetm} variable. 28* Driver:: Controlling how the driver runs the compilation passes. 29* Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}. 30* Per-Function Data:: Defining data structures for per-function information. 31* Storage Layout:: Defining sizes and alignments of data. 32* Type Layout:: Defining sizes and properties of basic user data types. 33* Registers:: Naming and describing the hardware registers. 34* Register Classes:: Defining the classes of hardware registers. 35* Stack and Calling:: Defining which way the stack grows and by how much. 36* Varargs:: Defining the varargs macros. 37* Trampolines:: Code set up at run time to enter a nested function. 38* Library Calls:: Controlling how library routines are implicitly called. 39* Addressing Modes:: Defining addressing modes valid for memory operands. 40* Anchored Addresses:: Defining how @option{-fsection-anchors} should work. 41* Condition Code:: Defining how insns update the condition code. 42* Costs:: Defining relative costs of different operations. 43* Scheduling:: Adjusting the behavior of the instruction scheduler. 44* Sections:: Dividing storage into text, data, and other sections. 45* PIC:: Macros for position independent code. 46* Assembler Format:: Defining how to write insns and pseudo-ops to output. 47* Debugging Info:: Defining the format of debugging output. 48* Floating Point:: Handling floating point for cross-compilers. 49* Mode Switching:: Insertion of mode-switching instructions. 50* Target Attributes:: Defining target-specific uses of @code{__attribute__}. 51* Emulated TLS:: Emulated TLS support. 52* MIPS Coprocessors:: MIPS coprocessor support and how to customize it. 53* PCH Target:: Validity checking for precompiled headers. 54* C++ ABI:: Controlling C++ ABI changes. 55* Named Address Spaces:: Adding support for named address spaces 56* Misc:: Everything else. 57@end menu 58 59@node Target Structure 60@section The Global @code{targetm} Variable 61@cindex target hooks 62@cindex target functions 63 64@deftypevar {struct gcc_target} targetm 65The target @file{.c} file must define the global @code{targetm} variable 66which contains pointers to functions and data relating to the target 67machine. The variable is declared in @file{target.h}; 68@file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is 69used to initialize the variable, and macros for the default initializers 70for elements of the structure. The @file{.c} file should override those 71macros for which the default definition is inappropriate. For example: 72@smallexample 73#include "target.h" 74#include "target-def.h" 75 76/* @r{Initialize the GCC target structure.} */ 77 78#undef TARGET_COMP_TYPE_ATTRIBUTES 79#define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes 80 81struct gcc_target targetm = TARGET_INITIALIZER; 82@end smallexample 83@end deftypevar 84 85Where a macro should be defined in the @file{.c} file in this manner to 86form part of the @code{targetm} structure, it is documented below as a 87``Target Hook'' with a prototype. Many macros will change in future 88from being defined in the @file{.h} file to being part of the 89@code{targetm} structure. 90 91Similarly, there is a @code{targetcm} variable for hooks that are 92specific to front ends for C-family languages, documented as ``C 93Target Hook''. This is declared in @file{c-family/c-target.h}, the 94initializer @code{TARGETCM_INITIALIZER} in 95@file{c-family/c-target-def.h}. If targets initialize @code{targetcm} 96themselves, they should set @code{target_has_targetcm=yes} in 97@file{config.gcc}; otherwise a default definition is used. 98 99Similarly, there is a @code{targetm_common} variable for hooks that 100are shared between the compiler driver and the compilers proper, 101documented as ``Common Target Hook''. This is declared in 102@file{common/common-target.h}, the initializer 103@code{TARGETM_COMMON_INITIALIZER} in 104@file{common/common-target-def.h}. If targets initialize 105@code{targetm_common} themselves, they should set 106@code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a 107default definition is used. 108 109@node Driver 110@section Controlling the Compilation Driver, @file{gcc} 111@cindex driver 112@cindex controlling the compilation driver 113 114@c prevent bad page break with this line 115You can control the compilation driver. 116 117@defmac DRIVER_SELF_SPECS 118A list of specs for the driver itself. It should be a suitable 119initializer for an array of strings, with no surrounding braces. 120 121The driver applies these specs to its own command line between loading 122default @file{specs} files (but not command-line specified ones) and 123choosing the multilib directory or running any subcommands. It 124applies them in the order given, so each spec can depend on the 125options added by earlier ones. It is also possible to remove options 126using @samp{%<@var{option}} in the usual way. 127 128This macro can be useful when a port has several interdependent target 129options. It provides a way of standardizing the command line so 130that the other specs are easier to write. 131 132Do not define this macro if it does not need to do anything. 133@end defmac 134 135@defmac OPTION_DEFAULT_SPECS 136A list of specs used to support configure-time default options (i.e.@: 137@option{--with} options) in the driver. It should be a suitable initializer 138for an array of structures, each containing two strings, without the 139outermost pair of surrounding braces. 140 141The first item in the pair is the name of the default. This must match 142the code in @file{config.gcc} for the target. The second item is a spec 143to apply if a default with this name was specified. The string 144@samp{%(VALUE)} in the spec will be replaced by the value of the default 145everywhere it occurs. 146 147The driver will apply these specs to its own command line between loading 148default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using 149the same mechanism as @code{DRIVER_SELF_SPECS}. 150 151Do not define this macro if it does not need to do anything. 152@end defmac 153 154@defmac CPP_SPEC 155A C string constant that tells the GCC driver program options to 156pass to CPP@. It can also specify how to translate options you 157give to GCC into options for GCC to pass to the CPP@. 158 159Do not define this macro if it does not need to do anything. 160@end defmac 161 162@defmac CPLUSPLUS_CPP_SPEC 163This macro is just like @code{CPP_SPEC}, but is used for C++, rather 164than C@. If you do not define this macro, then the value of 165@code{CPP_SPEC} (if any) will be used instead. 166@end defmac 167 168@defmac CC1_SPEC 169A C string constant that tells the GCC driver program options to 170pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language 171front ends. 172It can also specify how to translate options you give to GCC into options 173for GCC to pass to front ends. 174 175Do not define this macro if it does not need to do anything. 176@end defmac 177 178@defmac CC1PLUS_SPEC 179A C string constant that tells the GCC driver program options to 180pass to @code{cc1plus}. It can also specify how to translate options you 181give to GCC into options for GCC to pass to the @code{cc1plus}. 182 183Do not define this macro if it does not need to do anything. 184Note that everything defined in CC1_SPEC is already passed to 185@code{cc1plus} so there is no need to duplicate the contents of 186CC1_SPEC in CC1PLUS_SPEC@. 187@end defmac 188 189@defmac ASM_SPEC 190A C string constant that tells the GCC driver program options to 191pass to the assembler. It can also specify how to translate options 192you give to GCC into options for GCC to pass to the assembler. 193See the file @file{sun3.h} for an example of this. 194 195Do not define this macro if it does not need to do anything. 196@end defmac 197 198@defmac ASM_FINAL_SPEC 199A C string constant that tells the GCC driver program how to 200run any programs which cleanup after the normal assembler. 201Normally, this is not needed. See the file @file{mips.h} for 202an example of this. 203 204Do not define this macro if it does not need to do anything. 205@end defmac 206 207@defmac AS_NEEDS_DASH_FOR_PIPED_INPUT 208Define this macro, with no value, if the driver should give the assembler 209an argument consisting of a single dash, @option{-}, to instruct it to 210read from its standard input (which will be a pipe connected to the 211output of the compiler proper). This argument is given after any 212@option{-o} option specifying the name of the output file. 213 214If you do not define this macro, the assembler is assumed to read its 215standard input if given no non-option arguments. If your assembler 216cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct; 217see @file{mips.h} for instance. 218@end defmac 219 220@defmac LINK_SPEC 221A C string constant that tells the GCC driver program options to 222pass to the linker. It can also specify how to translate options you 223give to GCC into options for GCC to pass to the linker. 224 225Do not define this macro if it does not need to do anything. 226@end defmac 227 228@defmac LIB_SPEC 229Another C string constant used much like @code{LINK_SPEC}. The difference 230between the two is that @code{LIB_SPEC} is used at the end of the 231command given to the linker. 232 233If this macro is not defined, a default is provided that 234loads the standard C library from the usual place. See @file{gcc.c}. 235@end defmac 236 237@defmac LIBGCC_SPEC 238Another C string constant that tells the GCC driver program 239how and when to place a reference to @file{libgcc.a} into the 240linker command line. This constant is placed both before and after 241the value of @code{LIB_SPEC}. 242 243If this macro is not defined, the GCC driver provides a default that 244passes the string @option{-lgcc} to the linker. 245@end defmac 246 247@defmac REAL_LIBGCC_SPEC 248By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the 249@code{LIBGCC_SPEC} is not directly used by the driver program but is 250instead modified to refer to different versions of @file{libgcc.a} 251depending on the values of the command line flags @option{-static}, 252@option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On 253targets where these modifications are inappropriate, define 254@code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the 255driver how to place a reference to @file{libgcc} on the link command 256line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified. 257@end defmac 258 259@defmac USE_LD_AS_NEEDED 260A macro that controls the modifications to @code{LIBGCC_SPEC} 261mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be 262generated that uses @option{--as-needed} or equivalent options and the 263shared @file{libgcc} in place of the 264static exception handler library, when linking without any of 265@code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}. 266@end defmac 267 268@defmac LINK_EH_SPEC 269If defined, this C string constant is added to @code{LINK_SPEC}. 270When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects 271the modifications to @code{LIBGCC_SPEC} mentioned in 272@code{REAL_LIBGCC_SPEC}. 273@end defmac 274 275@defmac STARTFILE_SPEC 276Another C string constant used much like @code{LINK_SPEC}. The 277difference between the two is that @code{STARTFILE_SPEC} is used at 278the very beginning of the command given to the linker. 279 280If this macro is not defined, a default is provided that loads the 281standard C startup file from the usual place. See @file{gcc.c}. 282@end defmac 283 284@defmac ENDFILE_SPEC 285Another C string constant used much like @code{LINK_SPEC}. The 286difference between the two is that @code{ENDFILE_SPEC} is used at 287the very end of the command given to the linker. 288 289Do not define this macro if it does not need to do anything. 290@end defmac 291 292@defmac THREAD_MODEL_SPEC 293GCC @code{-v} will print the thread model GCC was configured to use. 294However, this doesn't work on platforms that are multilibbed on thread 295models, such as AIX 4.3. On such platforms, define 296@code{THREAD_MODEL_SPEC} such that it evaluates to a string without 297blanks that names one of the recognized thread models. @code{%*}, the 298default value of this macro, will expand to the value of 299@code{thread_file} set in @file{config.gcc}. 300@end defmac 301 302@defmac SYSROOT_SUFFIX_SPEC 303Define this macro to add a suffix to the target sysroot when GCC is 304configured with a sysroot. This will cause GCC to search for usr/lib, 305et al, within sysroot+suffix. 306@end defmac 307 308@defmac SYSROOT_HEADERS_SUFFIX_SPEC 309Define this macro to add a headers_suffix to the target sysroot when 310GCC is configured with a sysroot. This will cause GCC to pass the 311updated sysroot+headers_suffix to CPP, causing it to search for 312usr/include, et al, within sysroot+headers_suffix. 313@end defmac 314 315@defmac EXTRA_SPECS 316Define this macro to provide additional specifications to put in the 317@file{specs} file that can be used in various specifications like 318@code{CC1_SPEC}. 319 320The definition should be an initializer for an array of structures, 321containing a string constant, that defines the specification name, and a 322string constant that provides the specification. 323 324Do not define this macro if it does not need to do anything. 325 326@code{EXTRA_SPECS} is useful when an architecture contains several 327related targets, which have various @code{@dots{}_SPECS} which are similar 328to each other, and the maintainer would like one central place to keep 329these definitions. 330 331For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to 332define either @code{_CALL_SYSV} when the System V calling sequence is 333used or @code{_CALL_AIX} when the older AIX-based calling sequence is 334used. 335 336The @file{config/rs6000/rs6000.h} target file defines: 337 338@smallexample 339#define EXTRA_SPECS \ 340 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @}, 341 342#define CPP_SYS_DEFAULT "" 343@end smallexample 344 345The @file{config/rs6000/sysv.h} target file defines: 346@smallexample 347#undef CPP_SPEC 348#define CPP_SPEC \ 349"%@{posix: -D_POSIX_SOURCE @} \ 350%@{mcall-sysv: -D_CALL_SYSV @} \ 351%@{!mcall-sysv: %(cpp_sysv_default) @} \ 352%@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}" 353 354#undef CPP_SYSV_DEFAULT 355#define CPP_SYSV_DEFAULT "-D_CALL_SYSV" 356@end smallexample 357 358while the @file{config/rs6000/eabiaix.h} target file defines 359@code{CPP_SYSV_DEFAULT} as: 360 361@smallexample 362#undef CPP_SYSV_DEFAULT 363#define CPP_SYSV_DEFAULT "-D_CALL_AIX" 364@end smallexample 365@end defmac 366 367@defmac LINK_LIBGCC_SPECIAL_1 368Define this macro if the driver program should find the library 369@file{libgcc.a}. If you do not define this macro, the driver program will pass 370the argument @option{-lgcc} to tell the linker to do the search. 371@end defmac 372 373@defmac LINK_GCC_C_SEQUENCE_SPEC 374The sequence in which libgcc and libc are specified to the linker. 375By default this is @code{%G %L %G}. 376@end defmac 377 378@defmac LINK_COMMAND_SPEC 379A C string constant giving the complete command line need to execute the 380linker. When you do this, you will need to update your port each time a 381change is made to the link command line within @file{gcc.c}. Therefore, 382define this macro only if you need to completely redefine the command 383line for invoking the linker and there is no other way to accomplish 384the effect you need. Overriding this macro may be avoidable by overriding 385@code{LINK_GCC_C_SEQUENCE_SPEC} instead. 386@end defmac 387 388@deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT 389True if @file{..} components should always be removed from directory names computed relative to GCC's internal directories, false (default) if such components should be preserved and directory names containing them passed to other tools such as the linker. 390@end deftypevr 391 392@defmac MULTILIB_DEFAULTS 393Define this macro as a C expression for the initializer of an array of 394string to tell the driver program which options are defaults for this 395target and thus do not need to be handled specially when using 396@code{MULTILIB_OPTIONS}. 397 398Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in 399the target makefile fragment or if none of the options listed in 400@code{MULTILIB_OPTIONS} are set by default. 401@xref{Target Fragment}. 402@end defmac 403 404@defmac RELATIVE_PREFIX_NOT_LINKDIR 405Define this macro to tell @command{gcc} that it should only translate 406a @option{-B} prefix into a @option{-L} linker option if the prefix 407indicates an absolute file name. 408@end defmac 409 410@defmac MD_EXEC_PREFIX 411If defined, this macro is an additional prefix to try after 412@code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched 413when the compiler is built as a cross 414compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it 415to the list of directories used to find the assembler in @file{configure.in}. 416@end defmac 417 418@defmac STANDARD_STARTFILE_PREFIX 419Define this macro as a C string constant if you wish to override the 420standard choice of @code{libdir} as the default prefix to 421try when searching for startup files such as @file{crt0.o}. 422@code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler 423is built as a cross compiler. 424@end defmac 425 426@defmac STANDARD_STARTFILE_PREFIX_1 427Define this macro as a C string constant if you wish to override the 428standard choice of @code{/lib} as a prefix to try after the default prefix 429when searching for startup files such as @file{crt0.o}. 430@code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler 431is built as a cross compiler. 432@end defmac 433 434@defmac STANDARD_STARTFILE_PREFIX_2 435Define this macro as a C string constant if you wish to override the 436standard choice of @code{/lib} as yet another prefix to try after the 437default prefix when searching for startup files such as @file{crt0.o}. 438@code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler 439is built as a cross compiler. 440@end defmac 441 442@defmac MD_STARTFILE_PREFIX 443If defined, this macro supplies an additional prefix to try after the 444standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the 445compiler is built as a cross compiler. 446@end defmac 447 448@defmac MD_STARTFILE_PREFIX_1 449If defined, this macro supplies yet another prefix to try after the 450standard prefixes. It is not searched when the compiler is built as a 451cross compiler. 452@end defmac 453 454@defmac INIT_ENVIRONMENT 455Define this macro as a C string constant if you wish to set environment 456variables for programs called by the driver, such as the assembler and 457loader. The driver passes the value of this macro to @code{putenv} to 458initialize the necessary environment variables. 459@end defmac 460 461@defmac LOCAL_INCLUDE_DIR 462Define this macro as a C string constant if you wish to override the 463standard choice of @file{/usr/local/include} as the default prefix to 464try when searching for local header files. @code{LOCAL_INCLUDE_DIR} 465comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in 466@file{config.gcc}, normally @file{/usr/include}) in the search order. 467 468Cross compilers do not search either @file{/usr/local/include} or its 469replacement. 470@end defmac 471 472@defmac NATIVE_SYSTEM_HEADER_COMPONENT 473The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}. 474See @code{INCLUDE_DEFAULTS}, below, for the description of components. 475If you do not define this macro, no component is used. 476@end defmac 477 478@defmac INCLUDE_DEFAULTS 479Define this macro if you wish to override the entire default search path 480for include files. For a native compiler, the default search path 481usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR}, 482@code{GPLUSPLUS_INCLUDE_DIR}, and 483@code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR} 484and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile}, 485and specify private search areas for GCC@. The directory 486@code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs. 487 488The definition should be an initializer for an array of structures. 489Each array element should have four elements: the directory name (a 490string constant), the component name (also a string constant), a flag 491for C++-only directories, 492and a flag showing that the includes in the directory don't need to be 493wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of 494the array with a null element. 495 496The component name denotes what GNU package the include file is part of, 497if any, in all uppercase letters. For example, it might be @samp{GCC} 498or @samp{BINUTILS}. If the package is part of a vendor-supplied 499operating system, code the component name as @samp{0}. 500 501For example, here is the definition used for VAX/VMS: 502 503@smallexample 504#define INCLUDE_DEFAULTS \ 505@{ \ 506 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \ 507 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \ 508 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \ 509 @{ ".", 0, 0, 0@}, \ 510 @{ 0, 0, 0, 0@} \ 511@} 512@end smallexample 513@end defmac 514 515Here is the order of prefixes tried for exec files: 516 517@enumerate 518@item 519Any prefixes specified by the user with @option{-B}. 520 521@item 522The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX} 523is not set and the compiler has not been installed in the configure-time 524@var{prefix}, the location in which the compiler has actually been installed. 525 526@item 527The directories specified by the environment variable @code{COMPILER_PATH}. 528 529@item 530The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed 531in the configured-time @var{prefix}. 532 533@item 534The location @file{/usr/libexec/gcc/}, but only if this is a native compiler. 535 536@item 537The location @file{/usr/lib/gcc/}, but only if this is a native compiler. 538 539@item 540The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native 541compiler. 542@end enumerate 543 544Here is the order of prefixes tried for startfiles: 545 546@enumerate 547@item 548Any prefixes specified by the user with @option{-B}. 549 550@item 551The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined 552value based on the installed toolchain location. 553 554@item 555The directories specified by the environment variable @code{LIBRARY_PATH} 556(or port-specific name; native only, cross compilers do not use this). 557 558@item 559The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed 560in the configured @var{prefix} or this is a native compiler. 561 562@item 563The location @file{/usr/lib/gcc/}, but only if this is a native compiler. 564 565@item 566The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native 567compiler. 568 569@item 570The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a 571native compiler, or we have a target system root. 572 573@item 574The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a 575native compiler, or we have a target system root. 576 577@item 578The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications. 579If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and 580the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix. 581 582@item 583The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native 584compiler, or we have a target system root. The default for this macro is 585@file{/lib/}. 586 587@item 588The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native 589compiler, or we have a target system root. The default for this macro is 590@file{/usr/lib/}. 591@end enumerate 592 593@node Run-time Target 594@section Run-time Target Specification 595@cindex run-time target specification 596@cindex predefined macros 597@cindex target specifications 598 599@c prevent bad page break with this line 600Here are run-time target specifications. 601 602@defmac TARGET_CPU_CPP_BUILTINS () 603This function-like macro expands to a block of code that defines 604built-in preprocessor macros and assertions for the target CPU, using 605the functions @code{builtin_define}, @code{builtin_define_std} and 606@code{builtin_assert}. When the front end 607calls this macro it provides a trailing semicolon, and since it has 608finished command line option processing your code can use those 609results freely. 610 611@code{builtin_assert} takes a string in the form you pass to the 612command-line option @option{-A}, such as @code{cpu=mips}, and creates 613the assertion. @code{builtin_define} takes a string in the form 614accepted by option @option{-D} and unconditionally defines the macro. 615 616@code{builtin_define_std} takes a string representing the name of an 617object-like macro. If it doesn't lie in the user's namespace, 618@code{builtin_define_std} defines it unconditionally. Otherwise, it 619defines a version with two leading underscores, and another version 620with two leading and trailing underscores, and defines the original 621only if an ISO standard was not requested on the command line. For 622example, passing @code{unix} defines @code{__unix}, @code{__unix__} 623and possibly @code{unix}; passing @code{_mips} defines @code{__mips}, 624@code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64} 625defines only @code{_ABI64}. 626 627You can also test for the C dialect being compiled. The variable 628@code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus} 629or @code{clk_objective_c}. Note that if we are preprocessing 630assembler, this variable will be @code{clk_c} but the function-like 631macro @code{preprocessing_asm_p()} will return true, so you might want 632to check for that first. If you need to check for strict ANSI, the 633variable @code{flag_iso} can be used. The function-like macro 634@code{preprocessing_trad_p()} can be used to check for traditional 635preprocessing. 636@end defmac 637 638@defmac TARGET_OS_CPP_BUILTINS () 639Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional 640and is used for the target operating system instead. 641@end defmac 642 643@defmac TARGET_OBJFMT_CPP_BUILTINS () 644Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional 645and is used for the target object format. @file{elfos.h} uses this 646macro to define @code{__ELF__}, so you probably do not need to define 647it yourself. 648@end defmac 649 650@deftypevar {extern int} target_flags 651This variable is declared in @file{options.h}, which is included before 652any target-specific headers. 653@end deftypevar 654 655@deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS 656This variable specifies the initial value of @code{target_flags}. 657Its default setting is 0. 658@end deftypevr 659 660@cindex optional hardware or system features 661@cindex features, optional, in system conventions 662 663@deftypefn {Common Target Hook} bool TARGET_HANDLE_OPTION (struct gcc_options *@var{opts}, struct gcc_options *@var{opts_set}, const struct cl_decoded_option *@var{decoded}, location_t @var{loc}) 664This hook is called whenever the user specifies one of the 665target-specific options described by the @file{.opt} definition files 666(@pxref{Options}). It has the opportunity to do some option-specific 667processing and should return true if the option is valid. The default 668definition does nothing but return true. 669 670@var{decoded} specifies the option and its arguments. @var{opts} and 671@var{opts_set} are the @code{gcc_options} structures to be used for 672storing option state, and @var{loc} is the location at which the 673option was passed (@code{UNKNOWN_LOCATION} except for options passed 674via attributes). 675@end deftypefn 676 677@deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value}) 678This target hook is called whenever the user specifies one of the 679target-specific C language family options described by the @file{.opt} 680definition files(@pxref{Options}). It has the opportunity to do some 681option-specific processing and should return true if the option is 682valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The 683default definition does nothing but return false. 684 685In general, you should use @code{TARGET_HANDLE_OPTION} to handle 686options. However, if processing an option requires routines that are 687only available in the C (and related language) front ends, then you 688should use @code{TARGET_HANDLE_C_OPTION} instead. 689@end deftypefn 690 691@deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string}) 692Targets may provide a string object type that can be used within and between C, C++ and their respective Objective-C dialects. A string object might, for example, embed encoding and length information. These objects are considered opaque to the compiler and handled as references. An ideal implementation makes the composition of the string object match that of the Objective-C @code{NSString} (@code{NXString} for GNUStep), allowing efficient interworking between C-only and Objective-C code. If a target implements string objects then this hook should return a reference to such an object constructed from the normal `C' string representation provided in @var{string}. At present, the hook is used by Objective-C only, to obtain a common-format string object when the target provides one. 693@end deftypefn 694 695@deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE (const char *@var{classname}) 696Declare that Objective C class @var{classname} is referenced by the current TU. 697@end deftypefn 698 699@deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_CLASS_DEFINITION (const char *@var{classname}) 700Declare that Objective C class @var{classname} is defined by the current TU. 701@end deftypefn 702 703@deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref}) 704If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object. 705@end deftypefn 706 707@deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list}) 708If a target implements string objects then this hook should should provide a facility to check the function arguments in @var{args_list} against the format specifiers in @var{format_arg} where the type of @var{format_arg} is one recognized as a valid string reference type. 709@end deftypefn 710 711@deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void) 712This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE} 713but is called when the optimize level is changed via an attribute or 714pragma or when it is reset at the end of the code affected by the 715attribute or pragma. It is not called at the beginning of compilation 716when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these 717actions then, you should have @code{TARGET_OPTION_OVERRIDE} call 718@code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}. 719@end deftypefn 720 721@defmac C_COMMON_OVERRIDE_OPTIONS 722This is similar to the @code{TARGET_OPTION_OVERRIDE} hook 723but is only used in the C 724language frontends (C, Objective-C, C++, Objective-C++) and so can be 725used to alter option flag variables which only exist in those 726frontends. 727@end defmac 728 729@deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE 730Some machines may desire to change what optimizations are performed for 731various optimization levels. This variable, if defined, describes 732options to enable at particular sets of optimization levels. These 733options are processed once 734just after the optimization level is determined and before the remainder 735of the command options have been parsed, so may be overridden by other 736options passed explicitly. 737 738This processing is run once at program startup and when the optimization 739options are changed via @code{#pragma GCC optimize} or by using the 740@code{optimize} attribute. 741@end deftypevr 742 743@deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts}) 744Set target-dependent initial values of fields in @var{opts}. 745@end deftypefn 746 747@deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void) 748Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}. 749@end deftypefn 750 751@defmac SWITCHABLE_TARGET 752Some targets need to switch between substantially different subtargets 753during compilation. For example, the MIPS target has one subtarget for 754the traditional MIPS architecture and another for MIPS16. Source code 755can switch between these two subarchitectures using the @code{mips16} 756and @code{nomips16} attributes. 757 758Such subtargets can differ in things like the set of available 759registers, the set of available instructions, the costs of various 760operations, and so on. GCC caches a lot of this type of information 761in global variables, and recomputing them for each subtarget takes a 762significant amount of time. The compiler therefore provides a facility 763for maintaining several versions of the global variables and quickly 764switching between them; see @file{target-globals.h} for details. 765 766Define this macro to 1 if your target needs this facility. The default 767is 0. 768@end defmac 769 770@deftypefn {Target Hook} bool TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P (void) 771Returns true if the target supports IEEE 754 floating-point exceptions and rounding modes, false otherwise. This is intended to relate to the @code{float} and @code{double} types, but not necessarily @code{long double}. By default, returns true if the @code{adddf3} instruction pattern is available and false otherwise, on the assumption that hardware floating point supports exceptions and rounding modes but software floating point does not. 772@end deftypefn 773 774@node Per-Function Data 775@section Defining data structures for per-function information. 776@cindex per-function data 777@cindex data structures 778 779If the target needs to store information on a per-function basis, GCC 780provides a macro and a couple of variables to allow this. Note, just 781using statics to store the information is a bad idea, since GCC supports 782nested functions, so you can be halfway through encoding one function 783when another one comes along. 784 785GCC defines a data structure called @code{struct function} which 786contains all of the data specific to an individual function. This 787structure contains a field called @code{machine} whose type is 788@code{struct machine_function *}, which can be used by targets to point 789to their own specific data. 790 791If a target needs per-function specific data it should define the type 792@code{struct machine_function} and also the macro @code{INIT_EXPANDERS}. 793This macro should be used to initialize the function pointer 794@code{init_machine_status}. This pointer is explained below. 795 796One typical use of per-function, target specific data is to create an 797RTX to hold the register containing the function's return address. This 798RTX can then be used to implement the @code{__builtin_return_address} 799function, for level 0. 800 801Note---earlier implementations of GCC used a single data area to hold 802all of the per-function information. Thus when processing of a nested 803function began the old per-function data had to be pushed onto a 804stack, and when the processing was finished, it had to be popped off the 805stack. GCC used to provide function pointers called 806@code{save_machine_status} and @code{restore_machine_status} to handle 807the saving and restoring of the target specific information. Since the 808single data area approach is no longer used, these pointers are no 809longer supported. 810 811@defmac INIT_EXPANDERS 812Macro called to initialize any target specific information. This macro 813is called once per function, before generation of any RTL has begun. 814The intention of this macro is to allow the initialization of the 815function pointer @code{init_machine_status}. 816@end defmac 817 818@deftypevar {void (*)(struct function *)} init_machine_status 819If this function pointer is non-@code{NULL} it will be called once per 820function, before function compilation starts, in order to allow the 821target to perform any target specific initialization of the 822@code{struct function} structure. It is intended that this would be 823used to initialize the @code{machine} of that structure. 824 825@code{struct machine_function} structures are expected to be freed by GC@. 826Generally, any memory that they reference must be allocated by using 827GC allocation, including the structure itself. 828@end deftypevar 829 830@node Storage Layout 831@section Storage Layout 832@cindex storage layout 833 834Note that the definitions of the macros in this table which are sizes or 835alignments measured in bits do not need to be constant. They can be C 836expressions that refer to static variables, such as the @code{target_flags}. 837@xref{Run-time Target}. 838 839@defmac BITS_BIG_ENDIAN 840Define this macro to have the value 1 if the most significant bit in a 841byte has the lowest number; otherwise define it to have the value zero. 842This means that bit-field instructions count from the most significant 843bit. If the machine has no bit-field instructions, then this must still 844be defined, but it doesn't matter which value it is defined to. This 845macro need not be a constant. 846 847This macro does not affect the way structure fields are packed into 848bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}. 849@end defmac 850 851@defmac BYTES_BIG_ENDIAN 852Define this macro to have the value 1 if the most significant byte in a 853word has the lowest number. This macro need not be a constant. 854@end defmac 855 856@defmac WORDS_BIG_ENDIAN 857Define this macro to have the value 1 if, in a multiword object, the 858most significant word has the lowest number. This applies to both 859memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the 860order of words in memory is not the same as the order in registers. This 861macro need not be a constant. 862@end defmac 863 864@defmac REG_WORDS_BIG_ENDIAN 865On some machines, the order of words in a multiword object differs between 866registers in memory. In such a situation, define this macro to describe 867the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls 868the order of words in memory. 869@end defmac 870 871@defmac FLOAT_WORDS_BIG_ENDIAN 872Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or 873@code{TFmode} floating point numbers are stored in memory with the word 874containing the sign bit at the lowest address; otherwise define it to 875have the value 0. This macro need not be a constant. 876 877You need not define this macro if the ordering is the same as for 878multi-word integers. 879@end defmac 880 881@defmac BITS_PER_WORD 882Number of bits in a word. If you do not define this macro, the default 883is @code{BITS_PER_UNIT * UNITS_PER_WORD}. 884@end defmac 885 886@defmac MAX_BITS_PER_WORD 887Maximum number of bits in a word. If this is undefined, the default is 888@code{BITS_PER_WORD}. Otherwise, it is the constant value that is the 889largest value that @code{BITS_PER_WORD} can have at run-time. 890@end defmac 891 892@defmac UNITS_PER_WORD 893Number of storage units in a word; normally the size of a general-purpose 894register, a power of two from 1 or 8. 895@end defmac 896 897@defmac MIN_UNITS_PER_WORD 898Minimum number of units in a word. If this is undefined, the default is 899@code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the 900smallest value that @code{UNITS_PER_WORD} can have at run-time. 901@end defmac 902 903@defmac POINTER_SIZE 904Width of a pointer, in bits. You must specify a value no wider than the 905width of @code{Pmode}. If it is not equal to the width of @code{Pmode}, 906you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify 907a value the default is @code{BITS_PER_WORD}. 908@end defmac 909 910@defmac POINTERS_EXTEND_UNSIGNED 911A C expression that determines how pointers should be extended from 912@code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is 913greater than zero if pointers should be zero-extended, zero if they 914should be sign-extended, and negative if some other sort of conversion 915is needed. In the last case, the extension is done by the target's 916@code{ptr_extend} instruction. 917 918You need not define this macro if the @code{ptr_mode}, @code{Pmode} 919and @code{word_mode} are all the same width. 920@end defmac 921 922@defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type}) 923A macro to update @var{m} and @var{unsignedp} when an object whose type 924is @var{type} and which has the specified mode and signedness is to be 925stored in a register. This macro is only called when @var{type} is a 926scalar type. 927 928On most RISC machines, which only have operations that operate on a full 929register, define this macro to set @var{m} to @code{word_mode} if 930@var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most 931cases, only integer modes should be widened because wider-precision 932floating-point operations are usually more expensive than their narrower 933counterparts. 934 935For most machines, the macro definition does not change @var{unsignedp}. 936However, some machines, have instructions that preferentially handle 937either signed or unsigned quantities of certain modes. For example, on 938the DEC Alpha, 32-bit loads from memory and 32-bit add instructions 939sign-extend the result to 64 bits. On such machines, set 940@var{unsignedp} according to which kind of extension is more efficient. 941 942Do not define this macro if it would never modify @var{m}. 943@end defmac 944 945@deftypefn {Target Hook} machine_mode TARGET_PROMOTE_FUNCTION_MODE (const_tree @var{type}, machine_mode @var{mode}, int *@var{punsignedp}, const_tree @var{funtype}, int @var{for_return}) 946Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or 947function return values. The target hook should return the new mode 948and possibly change @code{*@var{punsignedp}} if the promotion should 949change signedness. This function is called only for scalar @emph{or 950pointer} types. 951 952@var{for_return} allows to distinguish the promotion of arguments and 953return values. If it is @code{1}, a return value is being promoted and 954@code{TARGET_FUNCTION_VALUE} must perform the same promotions done here. 955If it is @code{2}, the returned mode should be that of the register in 956which an incoming parameter is copied, or the outgoing result is computed; 957then the hook should return the same mode as @code{promote_mode}, though 958the signedness may be different. 959 960@var{type} can be NULL when promoting function arguments of libcalls. 961 962The default is to not promote arguments and return values. You can 963also define the hook to @code{default_promote_function_mode_always_promote} 964if you would like to apply the same rules given by @code{PROMOTE_MODE}. 965@end deftypefn 966 967@defmac PARM_BOUNDARY 968Normal alignment required for function parameters on the stack, in 969bits. All stack parameters receive at least this much alignment 970regardless of data type. On most machines, this is the same as the 971size of an integer. 972@end defmac 973 974@defmac STACK_BOUNDARY 975Define this macro to the minimum alignment enforced by hardware for the 976stack pointer on this machine. The definition is a C expression for the 977desired alignment (measured in bits). This value is used as a default 978if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines, 979this should be the same as @code{PARM_BOUNDARY}. 980@end defmac 981 982@defmac PREFERRED_STACK_BOUNDARY 983Define this macro if you wish to preserve a certain alignment for the 984stack pointer, greater than what the hardware enforces. The definition 985is a C expression for the desired alignment (measured in bits). This 986macro must evaluate to a value equal to or larger than 987@code{STACK_BOUNDARY}. 988@end defmac 989 990@defmac INCOMING_STACK_BOUNDARY 991Define this macro if the incoming stack boundary may be different 992from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate 993to a value equal to or larger than @code{STACK_BOUNDARY}. 994@end defmac 995 996@defmac FUNCTION_BOUNDARY 997Alignment required for a function entry point, in bits. 998@end defmac 999 1000@defmac BIGGEST_ALIGNMENT 1001Biggest alignment that any data type can require on this machine, in 1002bits. Note that this is not the biggest alignment that is supported, 1003just the biggest alignment that, when violated, may cause a fault. 1004@end defmac 1005 1006@deftypevr {Target Hook} HOST_WIDE_INT TARGET_ABSOLUTE_BIGGEST_ALIGNMENT 1007If defined, this target hook specifies the absolute biggest alignment 1008that a type or variable can have on this machine, otherwise, 1009@code{BIGGEST_ALIGNMENT} is used. 1010@end deftypevr 1011 1012@defmac MALLOC_ABI_ALIGNMENT 1013Alignment, in bits, a C conformant malloc implementation has to 1014provide. If not defined, the default value is @code{BITS_PER_WORD}. 1015@end defmac 1016 1017@defmac ATTRIBUTE_ALIGNED_VALUE 1018Alignment used by the @code{__attribute__ ((aligned))} construct. If 1019not defined, the default value is @code{BIGGEST_ALIGNMENT}. 1020@end defmac 1021 1022@defmac MINIMUM_ATOMIC_ALIGNMENT 1023If defined, the smallest alignment, in bits, that can be given to an 1024object that can be referenced in one operation, without disturbing any 1025nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger 1026on machines that don't have byte or half-word store operations. 1027@end defmac 1028 1029@defmac BIGGEST_FIELD_ALIGNMENT 1030Biggest alignment that any structure or union field can require on this 1031machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for 1032structure and union fields only, unless the field alignment has been set 1033by the @code{__attribute__ ((aligned (@var{n})))} construct. 1034@end defmac 1035 1036@defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed}) 1037An expression for the alignment of a structure field @var{field} if the 1038alignment computed in the usual way (including applying of 1039@code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the 1040alignment) is @var{computed}. It overrides alignment only if the 1041field alignment has not been set by the 1042@code{__attribute__ ((aligned (@var{n})))} construct. 1043@end defmac 1044 1045@defmac MAX_STACK_ALIGNMENT 1046Biggest stack alignment guaranteed by the backend. Use this macro 1047to specify the maximum alignment of a variable on stack. 1048 1049If not defined, the default value is @code{STACK_BOUNDARY}. 1050 1051@c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}. 1052@c But the fix for PR 32893 indicates that we can only guarantee 1053@c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not 1054@c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported. 1055@end defmac 1056 1057@defmac MAX_OFILE_ALIGNMENT 1058Biggest alignment supported by the object file format of this machine. 1059Use this macro to limit the alignment which can be specified using the 1060@code{__attribute__ ((aligned (@var{n})))} construct. If not defined, 1061the default value is @code{BIGGEST_ALIGNMENT}. 1062 1063On systems that use ELF, the default (in @file{config/elfos.h}) is 1064the largest supported 32-bit ELF section alignment representable on 1065a 32-bit host e.g. @samp{(((uint64_t) 1 << 28) * 8)}. 1066On 32-bit ELF the largest supported section alignment in bits is 1067@samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts. 1068@end defmac 1069 1070@defmac DATA_ALIGNMENT (@var{type}, @var{basic-align}) 1071If defined, a C expression to compute the alignment for a variable in 1072the static store. @var{type} is the data type, and @var{basic-align} is 1073the alignment that the object would ordinarily have. The value of this 1074macro is used instead of that alignment to align the object. 1075 1076If this macro is not defined, then @var{basic-align} is used. 1077 1078@findex strcpy 1079One use of this macro is to increase alignment of medium-size data to 1080make it all fit in fewer cache lines. Another is to cause character 1081arrays to be word-aligned so that @code{strcpy} calls that copy 1082constants to character arrays can be done inline. 1083@end defmac 1084 1085@defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align}) 1086Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates 1087some alignment increase, instead of optimization only purposes. E.g.@ 1088AMD x86-64 psABI says that variables with array type larger than 15 bytes 1089must be aligned to 16 byte boundaries. 1090 1091If this macro is not defined, then @var{basic-align} is used. 1092@end defmac 1093 1094@defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align}) 1095If defined, a C expression to compute the alignment given to a constant 1096that is being placed in memory. @var{constant} is the constant and 1097@var{basic-align} is the alignment that the object would ordinarily 1098have. The value of this macro is used instead of that alignment to 1099align the object. 1100 1101If this macro is not defined, then @var{basic-align} is used. 1102 1103The typical use of this macro is to increase alignment for string 1104constants to be word aligned so that @code{strcpy} calls that copy 1105constants can be done inline. 1106@end defmac 1107 1108@defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align}) 1109If defined, a C expression to compute the alignment for a variable in 1110the local store. @var{type} is the data type, and @var{basic-align} is 1111the alignment that the object would ordinarily have. The value of this 1112macro is used instead of that alignment to align the object. 1113 1114If this macro is not defined, then @var{basic-align} is used. 1115 1116One use of this macro is to increase alignment of medium-size data to 1117make it all fit in fewer cache lines. 1118 1119If the value of this macro has a type, it should be an unsigned type. 1120@end defmac 1121 1122@deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type}) 1123This hook can be used to define the alignment for a vector of type 1124@var{type}, in order to comply with a platform ABI. The default is to 1125require natural alignment for vector types. The alignment returned by 1126this hook must be a power-of-two multiple of the default alignment of 1127the vector element type. 1128@end deftypefn 1129 1130@defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align}) 1131If defined, a C expression to compute the alignment for stack slot. 1132@var{type} is the data type, @var{mode} is the widest mode available, 1133and @var{basic-align} is the alignment that the slot would ordinarily 1134have. The value of this macro is used instead of that alignment to 1135align the slot. 1136 1137If this macro is not defined, then @var{basic-align} is used when 1138@var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will 1139be used. 1140 1141This macro is to set alignment of stack slot to the maximum alignment 1142of all possible modes which the slot may have. 1143 1144If the value of this macro has a type, it should be an unsigned type. 1145@end defmac 1146 1147@defmac LOCAL_DECL_ALIGNMENT (@var{decl}) 1148If defined, a C expression to compute the alignment for a local 1149variable @var{decl}. 1150 1151If this macro is not defined, then 1152@code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))} 1153is used. 1154 1155One use of this macro is to increase alignment of medium-size data to 1156make it all fit in fewer cache lines. 1157 1158If the value of this macro has a type, it should be an unsigned type. 1159@end defmac 1160 1161@defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align}) 1162If defined, a C expression to compute the minimum required alignment 1163for dynamic stack realignment purposes for @var{exp} (a type or decl), 1164@var{mode}, assuming normal alignment @var{align}. 1165 1166If this macro is not defined, then @var{align} will be used. 1167@end defmac 1168 1169@defmac EMPTY_FIELD_BOUNDARY 1170Alignment in bits to be given to a structure bit-field that follows an 1171empty field such as @code{int : 0;}. 1172 1173If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro. 1174@end defmac 1175 1176@defmac STRUCTURE_SIZE_BOUNDARY 1177Number of bits which any structure or union's size must be a multiple of. 1178Each structure or union's size is rounded up to a multiple of this. 1179 1180If you do not define this macro, the default is the same as 1181@code{BITS_PER_UNIT}. 1182@end defmac 1183 1184@defmac STRICT_ALIGNMENT 1185Define this macro to be the value 1 if instructions will fail to work 1186if given data not on the nominal alignment. If instructions will merely 1187go slower in that case, define this macro as 0. 1188@end defmac 1189 1190@defmac PCC_BITFIELD_TYPE_MATTERS 1191Define this if you wish to imitate the way many other C compilers handle 1192alignment of bit-fields and the structures that contain them. 1193 1194The behavior is that the type written for a named bit-field (@code{int}, 1195@code{short}, or other integer type) imposes an alignment for the entire 1196structure, as if the structure really did contain an ordinary field of 1197that type. In addition, the bit-field is placed within the structure so 1198that it would fit within such a field, not crossing a boundary for it. 1199 1200Thus, on most machines, a named bit-field whose type is written as 1201@code{int} would not cross a four-byte boundary, and would force 1202four-byte alignment for the whole structure. (The alignment used may 1203not be four bytes; it is controlled by the other alignment parameters.) 1204 1205An unnamed bit-field will not affect the alignment of the containing 1206structure. 1207 1208If the macro is defined, its definition should be a C expression; 1209a nonzero value for the expression enables this behavior. 1210 1211Note that if this macro is not defined, or its value is zero, some 1212bit-fields may cross more than one alignment boundary. The compiler can 1213support such references if there are @samp{insv}, @samp{extv}, and 1214@samp{extzv} insns that can directly reference memory. 1215 1216The other known way of making bit-fields work is to define 1217@code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}. 1218Then every structure can be accessed with fullwords. 1219 1220Unless the machine has bit-field instructions or you define 1221@code{STRUCTURE_SIZE_BOUNDARY} that way, you must define 1222@code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value. 1223 1224If your aim is to make GCC use the same conventions for laying out 1225bit-fields as are used by another compiler, here is how to investigate 1226what the other compiler does. Compile and run this program: 1227 1228@smallexample 1229struct foo1 1230@{ 1231 char x; 1232 char :0; 1233 char y; 1234@}; 1235 1236struct foo2 1237@{ 1238 char x; 1239 int :0; 1240 char y; 1241@}; 1242 1243main () 1244@{ 1245 printf ("Size of foo1 is %d\n", 1246 sizeof (struct foo1)); 1247 printf ("Size of foo2 is %d\n", 1248 sizeof (struct foo2)); 1249 exit (0); 1250@} 1251@end smallexample 1252 1253If this prints 2 and 5, then the compiler's behavior is what you would 1254get from @code{PCC_BITFIELD_TYPE_MATTERS}. 1255@end defmac 1256 1257@defmac BITFIELD_NBYTES_LIMITED 1258Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited 1259to aligning a bit-field within the structure. 1260@end defmac 1261 1262@deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void) 1263When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine 1264whether unnamed bitfields affect the alignment of the containing 1265structure. The hook should return true if the structure should inherit 1266the alignment requirements of an unnamed bitfield's type. 1267@end deftypefn 1268 1269@deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void) 1270This target hook should return @code{true} if accesses to volatile bitfields 1271should use the narrowest mode possible. It should return @code{false} if 1272these accesses should use the bitfield container type. 1273 1274The default is @code{false}. 1275@end deftypefn 1276 1277@deftypefn {Target Hook} bool TARGET_MEMBER_TYPE_FORCES_BLK (const_tree @var{field}, machine_mode @var{mode}) 1278Return true if a structure, union or array containing @var{field} should 1279be accessed using @code{BLKMODE}. 1280 1281If @var{field} is the only field in the structure, @var{mode} is its 1282mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the 1283case where structures of one field would require the structure's mode to 1284retain the field's mode. 1285 1286Normally, this is not needed. 1287@end deftypefn 1288 1289@defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified}) 1290Define this macro as an expression for the alignment of a type (given 1291by @var{type} as a tree node) if the alignment computed in the usual 1292way is @var{computed} and the alignment explicitly specified was 1293@var{specified}. 1294 1295The default is to use @var{specified} if it is larger; otherwise, use 1296the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT} 1297@end defmac 1298 1299@defmac MAX_FIXED_MODE_SIZE 1300An integer expression for the size in bits of the largest integer 1301machine mode that should actually be used. All integer machine modes of 1302this size or smaller can be used for structures and unions with the 1303appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE 1304(DImode)} is assumed. 1305@end defmac 1306 1307@defmac STACK_SAVEAREA_MODE (@var{save_level}) 1308If defined, an expression of type @code{machine_mode} that 1309specifies the mode of the save area operand of a 1310@code{save_stack_@var{level}} named pattern (@pxref{Standard Names}). 1311@var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or 1312@code{SAVE_NONLOCAL} and selects which of the three named patterns is 1313having its mode specified. 1314 1315You need not define this macro if it always returns @code{Pmode}. You 1316would most commonly define this macro if the 1317@code{save_stack_@var{level}} patterns need to support both a 32- and a 131864-bit mode. 1319@end defmac 1320 1321@defmac STACK_SIZE_MODE 1322If defined, an expression of type @code{machine_mode} that 1323specifies the mode of the size increment operand of an 1324@code{allocate_stack} named pattern (@pxref{Standard Names}). 1325 1326You need not define this macro if it always returns @code{word_mode}. 1327You would most commonly define this macro if the @code{allocate_stack} 1328pattern needs to support both a 32- and a 64-bit mode. 1329@end defmac 1330 1331@deftypefn {Target Hook} machine_mode TARGET_LIBGCC_CMP_RETURN_MODE (void) 1332This target hook should return the mode to be used for the return value 1333of compare instructions expanded to libgcc calls. If not defined 1334@code{word_mode} is returned which is the right choice for a majority of 1335targets. 1336@end deftypefn 1337 1338@deftypefn {Target Hook} machine_mode TARGET_LIBGCC_SHIFT_COUNT_MODE (void) 1339This target hook should return the mode to be used for the shift count operand 1340of shift instructions expanded to libgcc calls. If not defined 1341@code{word_mode} is returned which is the right choice for a majority of 1342targets. 1343@end deftypefn 1344 1345@deftypefn {Target Hook} machine_mode TARGET_UNWIND_WORD_MODE (void) 1346Return machine mode to be used for @code{_Unwind_Word} type. 1347The default is to use @code{word_mode}. 1348@end deftypefn 1349 1350@deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type}) 1351This target hook returns @code{true} if bit-fields in the given 1352@var{record_type} are to be laid out following the rules of Microsoft 1353Visual C/C++, namely: (i) a bit-field won't share the same storage 1354unit with the previous bit-field if their underlying types have 1355different sizes, and the bit-field will be aligned to the highest 1356alignment of the underlying types of itself and of the previous 1357bit-field; (ii) a zero-sized bit-field will affect the alignment of 1358the whole enclosing structure, even if it is unnamed; except that 1359(iii) a zero-sized bit-field will be disregarded unless it follows 1360another bit-field of nonzero size. If this hook returns @code{true}, 1361other macros that control bit-field layout are ignored. 1362 1363When a bit-field is inserted into a packed record, the whole size 1364of the underlying type is used by one or more same-size adjacent 1365bit-fields (that is, if its long:3, 32 bits is used in the record, 1366and any additional adjacent long bit-fields are packed into the same 1367chunk of 32 bits. However, if the size changes, a new field of that 1368size is allocated). In an unpacked record, this is the same as using 1369alignment, but not equivalent when packing. 1370 1371If both MS bit-fields and @samp{__attribute__((packed))} are used, 1372the latter will take precedence. If @samp{__attribute__((packed))} is 1373used on a single field when MS bit-fields are in use, it will take 1374precedence for that field, but the alignment of the rest of the structure 1375may affect its placement. 1376@end deftypefn 1377 1378@deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void) 1379Returns true if the target supports decimal floating point. 1380@end deftypefn 1381 1382@deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void) 1383Returns true if the target supports fixed-point arithmetic. 1384@end deftypefn 1385 1386@deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void) 1387This hook is called just before expansion into rtl, allowing the target 1388to perform additional initializations or analysis before the expansion. 1389For example, the rs6000 port uses it to allocate a scratch stack slot 1390for use in copying SDmode values between memory and floating point 1391registers whenever the function being expanded has any SDmode 1392usage. 1393@end deftypefn 1394 1395@deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void) 1396This hook allows the backend to perform additional instantiations on rtl 1397that are not actually in any insns yet, but will be later. 1398@end deftypefn 1399 1400@deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type}) 1401If your target defines any fundamental types, or any types your target 1402uses should be mangled differently from the default, define this hook 1403to return the appropriate encoding for these types as part of a C++ 1404mangled name. The @var{type} argument is the tree structure representing 1405the type to be mangled. The hook may be applied to trees which are 1406not target-specific fundamental types; it should return @code{NULL} 1407for all such types, as well as arguments it does not recognize. If the 1408return value is not @code{NULL}, it must point to a statically-allocated 1409string constant. 1410 1411Target-specific fundamental types might be new fundamental types or 1412qualified versions of ordinary fundamental types. Encode new 1413fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name} 1414is the name used for the type in source code, and @var{n} is the 1415length of @var{name} in decimal. Encode qualified versions of 1416ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where 1417@var{name} is the name used for the type qualifier in source code, 1418@var{n} is the length of @var{name} as above, and @var{code} is the 1419code used to represent the unqualified version of this type. (See 1420@code{write_builtin_type} in @file{cp/mangle.c} for the list of 1421codes.) In both cases the spaces are for clarity; do not include any 1422spaces in your string. 1423 1424This hook is applied to types prior to typedef resolution. If the mangled 1425name for a particular type depends only on that type's main variant, you 1426can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT} 1427before mangling. 1428 1429The default version of this hook always returns @code{NULL}, which is 1430appropriate for a target that does not define any new fundamental 1431types. 1432@end deftypefn 1433 1434@node Type Layout 1435@section Layout of Source Language Data Types 1436 1437These macros define the sizes and other characteristics of the standard 1438basic data types used in programs being compiled. Unlike the macros in 1439the previous section, these apply to specific features of C and related 1440languages, rather than to fundamental aspects of storage layout. 1441 1442@defmac INT_TYPE_SIZE 1443A C expression for the size in bits of the type @code{int} on the 1444target machine. If you don't define this, the default is one word. 1445@end defmac 1446 1447@defmac SHORT_TYPE_SIZE 1448A C expression for the size in bits of the type @code{short} on the 1449target machine. If you don't define this, the default is half a word. 1450(If this would be less than one storage unit, it is rounded up to one 1451unit.) 1452@end defmac 1453 1454@defmac LONG_TYPE_SIZE 1455A C expression for the size in bits of the type @code{long} on the 1456target machine. If you don't define this, the default is one word. 1457@end defmac 1458 1459@defmac ADA_LONG_TYPE_SIZE 1460On some machines, the size used for the Ada equivalent of the type 1461@code{long} by a native Ada compiler differs from that used by C@. In 1462that situation, define this macro to be a C expression to be used for 1463the size of that type. If you don't define this, the default is the 1464value of @code{LONG_TYPE_SIZE}. 1465@end defmac 1466 1467@defmac LONG_LONG_TYPE_SIZE 1468A C expression for the size in bits of the type @code{long long} on the 1469target machine. If you don't define this, the default is two 1470words. If you want to support GNU Ada on your machine, the value of this 1471macro must be at least 64. 1472@end defmac 1473 1474@defmac CHAR_TYPE_SIZE 1475A C expression for the size in bits of the type @code{char} on the 1476target machine. If you don't define this, the default is 1477@code{BITS_PER_UNIT}. 1478@end defmac 1479 1480@defmac BOOL_TYPE_SIZE 1481A C expression for the size in bits of the C++ type @code{bool} and 1482C99 type @code{_Bool} on the target machine. If you don't define 1483this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}. 1484@end defmac 1485 1486@defmac FLOAT_TYPE_SIZE 1487A C expression for the size in bits of the type @code{float} on the 1488target machine. If you don't define this, the default is one word. 1489@end defmac 1490 1491@defmac DOUBLE_TYPE_SIZE 1492A C expression for the size in bits of the type @code{double} on the 1493target machine. If you don't define this, the default is two 1494words. 1495@end defmac 1496 1497@defmac LONG_DOUBLE_TYPE_SIZE 1498A C expression for the size in bits of the type @code{long double} on 1499the target machine. If you don't define this, the default is two 1500words. 1501@end defmac 1502 1503@defmac SHORT_FRACT_TYPE_SIZE 1504A C expression for the size in bits of the type @code{short _Fract} on 1505the target machine. If you don't define this, the default is 1506@code{BITS_PER_UNIT}. 1507@end defmac 1508 1509@defmac FRACT_TYPE_SIZE 1510A C expression for the size in bits of the type @code{_Fract} on 1511the target machine. If you don't define this, the default is 1512@code{BITS_PER_UNIT * 2}. 1513@end defmac 1514 1515@defmac LONG_FRACT_TYPE_SIZE 1516A C expression for the size in bits of the type @code{long _Fract} on 1517the target machine. If you don't define this, the default is 1518@code{BITS_PER_UNIT * 4}. 1519@end defmac 1520 1521@defmac LONG_LONG_FRACT_TYPE_SIZE 1522A C expression for the size in bits of the type @code{long long _Fract} on 1523the target machine. If you don't define this, the default is 1524@code{BITS_PER_UNIT * 8}. 1525@end defmac 1526 1527@defmac SHORT_ACCUM_TYPE_SIZE 1528A C expression for the size in bits of the type @code{short _Accum} on 1529the target machine. If you don't define this, the default is 1530@code{BITS_PER_UNIT * 2}. 1531@end defmac 1532 1533@defmac ACCUM_TYPE_SIZE 1534A C expression for the size in bits of the type @code{_Accum} on 1535the target machine. If you don't define this, the default is 1536@code{BITS_PER_UNIT * 4}. 1537@end defmac 1538 1539@defmac LONG_ACCUM_TYPE_SIZE 1540A C expression for the size in bits of the type @code{long _Accum} on 1541the target machine. If you don't define this, the default is 1542@code{BITS_PER_UNIT * 8}. 1543@end defmac 1544 1545@defmac LONG_LONG_ACCUM_TYPE_SIZE 1546A C expression for the size in bits of the type @code{long long _Accum} on 1547the target machine. If you don't define this, the default is 1548@code{BITS_PER_UNIT * 16}. 1549@end defmac 1550 1551@defmac LIBGCC2_GNU_PREFIX 1552This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target 1553hook and should be defined if that hook is overriden to be true. It 1554causes function names in libgcc to be changed to use a @code{__gnu_} 1555prefix for their name rather than the default @code{__}. A port which 1556uses this macro should also arrange to use @file{t-gnu-prefix} in 1557the libgcc @file{config.host}. 1558@end defmac 1559 1560@defmac TARGET_FLT_EVAL_METHOD 1561A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h}, 1562assuming, if applicable, that the floating-point control word is in its 1563default state. If you do not define this macro the value of 1564@code{FLT_EVAL_METHOD} will be zero. 1565@end defmac 1566 1567@defmac WIDEST_HARDWARE_FP_SIZE 1568A C expression for the size in bits of the widest floating-point format 1569supported by the hardware. If you define this macro, you must specify a 1570value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}. 1571If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE} 1572is the default. 1573@end defmac 1574 1575@defmac DEFAULT_SIGNED_CHAR 1576An expression whose value is 1 or 0, according to whether the type 1577@code{char} should be signed or unsigned by default. The user can 1578always override this default with the options @option{-fsigned-char} 1579and @option{-funsigned-char}. 1580@end defmac 1581 1582@deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void) 1583This target hook should return true if the compiler should give an 1584@code{enum} type only as many bytes as it takes to represent the range 1585of possible values of that type. It should return false if all 1586@code{enum} types should be allocated like @code{int}. 1587 1588The default is to return false. 1589@end deftypefn 1590 1591@defmac SIZE_TYPE 1592A C expression for a string describing the name of the data type to use 1593for size values. The typedef name @code{size_t} is defined using the 1594contents of the string. 1595 1596The string can contain more than one keyword. If so, separate them with 1597spaces, and write first any length keyword, then @code{unsigned} if 1598appropriate, and finally @code{int}. The string must exactly match one 1599of the data type names defined in the function 1600@code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}. 1601You may not omit @code{int} or change the order---that would cause the 1602compiler to crash on startup. 1603 1604If you don't define this macro, the default is @code{"long unsigned 1605int"}. 1606@end defmac 1607 1608@defmac SIZETYPE 1609GCC defines internal types (@code{sizetype}, @code{ssizetype}, 1610@code{bitsizetype} and @code{sbitsizetype}) for expressions 1611dealing with size. This macro is a C expression for a string describing 1612the name of the data type from which the precision of @code{sizetype} 1613is extracted. 1614 1615The string has the same restrictions as @code{SIZE_TYPE} string. 1616 1617If you don't define this macro, the default is @code{SIZE_TYPE}. 1618@end defmac 1619 1620@defmac PTRDIFF_TYPE 1621A C expression for a string describing the name of the data type to use 1622for the result of subtracting two pointers. The typedef name 1623@code{ptrdiff_t} is defined using the contents of the string. See 1624@code{SIZE_TYPE} above for more information. 1625 1626If you don't define this macro, the default is @code{"long int"}. 1627@end defmac 1628 1629@defmac WCHAR_TYPE 1630A C expression for a string describing the name of the data type to use 1631for wide characters. The typedef name @code{wchar_t} is defined using 1632the contents of the string. See @code{SIZE_TYPE} above for more 1633information. 1634 1635If you don't define this macro, the default is @code{"int"}. 1636@end defmac 1637 1638@defmac WCHAR_TYPE_SIZE 1639A C expression for the size in bits of the data type for wide 1640characters. This is used in @code{cpp}, which cannot make use of 1641@code{WCHAR_TYPE}. 1642@end defmac 1643 1644@defmac WINT_TYPE 1645A C expression for a string describing the name of the data type to 1646use for wide characters passed to @code{printf} and returned from 1647@code{getwc}. The typedef name @code{wint_t} is defined using the 1648contents of the string. See @code{SIZE_TYPE} above for more 1649information. 1650 1651If you don't define this macro, the default is @code{"unsigned int"}. 1652@end defmac 1653 1654@defmac INTMAX_TYPE 1655A C expression for a string describing the name of the data type that 1656can represent any value of any standard or extended signed integer type. 1657The typedef name @code{intmax_t} is defined using the contents of the 1658string. See @code{SIZE_TYPE} above for more information. 1659 1660If you don't define this macro, the default is the first of 1661@code{"int"}, @code{"long int"}, or @code{"long long int"} that has as 1662much precision as @code{long long int}. 1663@end defmac 1664 1665@defmac UINTMAX_TYPE 1666A C expression for a string describing the name of the data type that 1667can represent any value of any standard or extended unsigned integer 1668type. The typedef name @code{uintmax_t} is defined using the contents 1669of the string. See @code{SIZE_TYPE} above for more information. 1670 1671If you don't define this macro, the default is the first of 1672@code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long 1673unsigned int"} that has as much precision as @code{long long unsigned 1674int}. 1675@end defmac 1676 1677@defmac SIG_ATOMIC_TYPE 1678@defmacx INT8_TYPE 1679@defmacx INT16_TYPE 1680@defmacx INT32_TYPE 1681@defmacx INT64_TYPE 1682@defmacx UINT8_TYPE 1683@defmacx UINT16_TYPE 1684@defmacx UINT32_TYPE 1685@defmacx UINT64_TYPE 1686@defmacx INT_LEAST8_TYPE 1687@defmacx INT_LEAST16_TYPE 1688@defmacx INT_LEAST32_TYPE 1689@defmacx INT_LEAST64_TYPE 1690@defmacx UINT_LEAST8_TYPE 1691@defmacx UINT_LEAST16_TYPE 1692@defmacx UINT_LEAST32_TYPE 1693@defmacx UINT_LEAST64_TYPE 1694@defmacx INT_FAST8_TYPE 1695@defmacx INT_FAST16_TYPE 1696@defmacx INT_FAST32_TYPE 1697@defmacx INT_FAST64_TYPE 1698@defmacx UINT_FAST8_TYPE 1699@defmacx UINT_FAST16_TYPE 1700@defmacx UINT_FAST32_TYPE 1701@defmacx UINT_FAST64_TYPE 1702@defmacx INTPTR_TYPE 1703@defmacx UINTPTR_TYPE 1704C expressions for the standard types @code{sig_atomic_t}, 1705@code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t}, 1706@code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t}, 1707@code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t}, 1708@code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t}, 1709@code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t}, 1710@code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t}, 1711@code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t}, 1712@code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See 1713@code{SIZE_TYPE} above for more information. 1714 1715If any of these macros evaluates to a null pointer, the corresponding 1716type is not supported; if GCC is configured to provide 1717@code{<stdint.h>} in such a case, the header provided may not conform 1718to C99, depending on the type in question. The defaults for all of 1719these macros are null pointers. 1720@end defmac 1721 1722@defmac TARGET_PTRMEMFUNC_VBIT_LOCATION 1723The C++ compiler represents a pointer-to-member-function with a struct 1724that looks like: 1725 1726@smallexample 1727 struct @{ 1728 union @{ 1729 void (*fn)(); 1730 ptrdiff_t vtable_index; 1731 @}; 1732 ptrdiff_t delta; 1733 @}; 1734@end smallexample 1735 1736@noindent 1737The C++ compiler must use one bit to indicate whether the function that 1738will be called through a pointer-to-member-function is virtual. 1739Normally, we assume that the low-order bit of a function pointer must 1740always be zero. Then, by ensuring that the vtable_index is odd, we can 1741distinguish which variant of the union is in use. But, on some 1742platforms function pointers can be odd, and so this doesn't work. In 1743that case, we use the low-order bit of the @code{delta} field, and shift 1744the remainder of the @code{delta} field to the left. 1745 1746GCC will automatically make the right selection about where to store 1747this bit using the @code{FUNCTION_BOUNDARY} setting for your platform. 1748However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY} 1749set such that functions always start at even addresses, but the lowest 1750bit of pointers to functions indicate whether the function at that 1751address is in ARM or Thumb mode. If this is the case of your 1752architecture, you should define this macro to 1753@code{ptrmemfunc_vbit_in_delta}. 1754 1755In general, you should not have to define this macro. On architectures 1756in which function addresses are always even, according to 1757@code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to 1758@code{ptrmemfunc_vbit_in_pfn}. 1759@end defmac 1760 1761@defmac TARGET_VTABLE_USES_DESCRIPTORS 1762Normally, the C++ compiler uses function pointers in vtables. This 1763macro allows the target to change to use ``function descriptors'' 1764instead. Function descriptors are found on targets for whom a 1765function pointer is actually a small data structure. Normally the 1766data structure consists of the actual code address plus a data 1767pointer to which the function's data is relative. 1768 1769If vtables are used, the value of this macro should be the number 1770of words that the function descriptor occupies. 1771@end defmac 1772 1773@defmac TARGET_VTABLE_ENTRY_ALIGN 1774By default, the vtable entries are void pointers, the so the alignment 1775is the same as pointer alignment. The value of this macro specifies 1776the alignment of the vtable entry in bits. It should be defined only 1777when special alignment is necessary. */ 1778@end defmac 1779 1780@defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE 1781There are a few non-descriptor entries in the vtable at offsets below 1782zero. If these entries must be padded (say, to preserve the alignment 1783specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number 1784of words in each data entry. 1785@end defmac 1786 1787@node Registers 1788@section Register Usage 1789@cindex register usage 1790 1791This section explains how to describe what registers the target machine 1792has, and how (in general) they can be used. 1793 1794The description of which registers a specific instruction can use is 1795done with register classes; see @ref{Register Classes}. For information 1796on using registers to access a stack frame, see @ref{Frame Registers}. 1797For passing values in registers, see @ref{Register Arguments}. 1798For returning values in registers, see @ref{Scalar Return}. 1799 1800@menu 1801* Register Basics:: Number and kinds of registers. 1802* Allocation Order:: Order in which registers are allocated. 1803* Values in Registers:: What kinds of values each reg can hold. 1804* Leaf Functions:: Renumbering registers for leaf functions. 1805* Stack Registers:: Handling a register stack such as 80387. 1806@end menu 1807 1808@node Register Basics 1809@subsection Basic Characteristics of Registers 1810 1811@c prevent bad page break with this line 1812Registers have various characteristics. 1813 1814@defmac FIRST_PSEUDO_REGISTER 1815Number of hardware registers known to the compiler. They receive 1816numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first 1817pseudo register's number really is assigned the number 1818@code{FIRST_PSEUDO_REGISTER}. 1819@end defmac 1820 1821@defmac FIXED_REGISTERS 1822@cindex fixed register 1823An initializer that says which registers are used for fixed purposes 1824all throughout the compiled code and are therefore not available for 1825general allocation. These would include the stack pointer, the frame 1826pointer (except on machines where that can be used as a general 1827register when no frame pointer is needed), the program counter on 1828machines where that is considered one of the addressable registers, 1829and any other numbered register with a standard use. 1830 1831This information is expressed as a sequence of numbers, separated by 1832commas and surrounded by braces. The @var{n}th number is 1 if 1833register @var{n} is fixed, 0 otherwise. 1834 1835The table initialized from this macro, and the table initialized by 1836the following one, may be overridden at run time either automatically, 1837by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by 1838the user with the command options @option{-ffixed-@var{reg}}, 1839@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}. 1840@end defmac 1841 1842@defmac CALL_USED_REGISTERS 1843@cindex call-used register 1844@cindex call-clobbered register 1845@cindex call-saved register 1846Like @code{FIXED_REGISTERS} but has 1 for each register that is 1847clobbered (in general) by function calls as well as for fixed 1848registers. This macro therefore identifies the registers that are not 1849available for general allocation of values that must live across 1850function calls. 1851 1852If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler 1853automatically saves it on function entry and restores it on function 1854exit, if the register is used within the function. 1855@end defmac 1856 1857@defmac CALL_REALLY_USED_REGISTERS 1858@cindex call-used register 1859@cindex call-clobbered register 1860@cindex call-saved register 1861Like @code{CALL_USED_REGISTERS} except this macro doesn't require 1862that the entire set of @code{FIXED_REGISTERS} be included. 1863(@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}). 1864This macro is optional. If not specified, it defaults to the value 1865of @code{CALL_USED_REGISTERS}. 1866@end defmac 1867 1868@defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode}) 1869@cindex call-used register 1870@cindex call-clobbered register 1871@cindex call-saved register 1872A C expression that is nonzero if it is not permissible to store a 1873value of mode @var{mode} in hard register number @var{regno} across a 1874call without some part of it being clobbered. For most machines this 1875macro need not be defined. It is only required for machines that do not 1876preserve the entire contents of a register across a call. 1877@end defmac 1878 1879@findex fixed_regs 1880@findex call_used_regs 1881@findex global_regs 1882@findex reg_names 1883@findex reg_class_contents 1884@deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void) 1885This hook may conditionally modify five variables 1886@code{fixed_regs}, @code{call_used_regs}, @code{global_regs}, 1887@code{reg_names}, and @code{reg_class_contents}, to take into account 1888any dependence of these register sets on target flags. The first three 1889of these are of type @code{char []} (interpreted as Boolean vectors). 1890@code{global_regs} is a @code{const char *[]}, and 1891@code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is 1892called, @code{fixed_regs}, @code{call_used_regs}, 1893@code{reg_class_contents}, and @code{reg_names} have been initialized 1894from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS}, 1895@code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively. 1896@code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}}, 1897@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}} 1898command options have been applied. 1899 1900@cindex disabling certain registers 1901@cindex controlling register usage 1902If the usage of an entire class of registers depends on the target 1903flags, you may indicate this to GCC by using this macro to modify 1904@code{fixed_regs} and @code{call_used_regs} to 1 for each of the 1905registers in the classes which should not be used by GCC@. Also make 1906@code{define_register_constraint}s return @code{NO_REGS} for constraints 1907that shouldn't be used. 1908 1909(However, if this class is not included in @code{GENERAL_REGS} and all 1910of the insn patterns whose constraints permit this class are 1911controlled by target switches, then GCC will automatically avoid using 1912these registers when the target switches are opposed to them.) 1913@end deftypefn 1914 1915@defmac INCOMING_REGNO (@var{out}) 1916Define this macro if the target machine has register windows. This C 1917expression returns the register number as seen by the called function 1918corresponding to the register number @var{out} as seen by the calling 1919function. Return @var{out} if register number @var{out} is not an 1920outbound register. 1921@end defmac 1922 1923@defmac OUTGOING_REGNO (@var{in}) 1924Define this macro if the target machine has register windows. This C 1925expression returns the register number as seen by the calling function 1926corresponding to the register number @var{in} as seen by the called 1927function. Return @var{in} if register number @var{in} is not an inbound 1928register. 1929@end defmac 1930 1931@defmac LOCAL_REGNO (@var{regno}) 1932Define this macro if the target machine has register windows. This C 1933expression returns true if the register is call-saved but is in the 1934register window. Unlike most call-saved registers, such registers 1935need not be explicitly restored on function exit or during non-local 1936gotos. 1937@end defmac 1938 1939@defmac PC_REGNUM 1940If the program counter has a register number, define this as that 1941register number. Otherwise, do not define it. 1942@end defmac 1943 1944@node Allocation Order 1945@subsection Order of Allocation of Registers 1946@cindex order of register allocation 1947@cindex register allocation order 1948 1949@c prevent bad page break with this line 1950Registers are allocated in order. 1951 1952@defmac REG_ALLOC_ORDER 1953If defined, an initializer for a vector of integers, containing the 1954numbers of hard registers in the order in which GCC should prefer 1955to use them (from most preferred to least). 1956 1957If this macro is not defined, registers are used lowest numbered first 1958(all else being equal). 1959 1960One use of this macro is on machines where the highest numbered 1961registers must always be saved and the save-multiple-registers 1962instruction supports only sequences of consecutive registers. On such 1963machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists 1964the highest numbered allocable register first. 1965@end defmac 1966 1967@defmac ADJUST_REG_ALLOC_ORDER 1968A C statement (sans semicolon) to choose the order in which to allocate 1969hard registers for pseudo-registers local to a basic block. 1970 1971Store the desired register order in the array @code{reg_alloc_order}. 1972Element 0 should be the register to allocate first; element 1, the next 1973register; and so on. 1974 1975The macro body should not assume anything about the contents of 1976@code{reg_alloc_order} before execution of the macro. 1977 1978On most machines, it is not necessary to define this macro. 1979@end defmac 1980 1981@defmac HONOR_REG_ALLOC_ORDER 1982Normally, IRA tries to estimate the costs for saving a register in the 1983prologue and restoring it in the epilogue. This discourages it from 1984using call-saved registers. If a machine wants to ensure that IRA 1985allocates registers in the order given by REG_ALLOC_ORDER even if some 1986call-saved registers appear earlier than call-used ones, then define this 1987macro as a C expression to nonzero. Default is 0. 1988@end defmac 1989 1990@defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno}) 1991In some case register allocation order is not enough for the 1992Integrated Register Allocator (@acronym{IRA}) to generate a good code. 1993If this macro is defined, it should return a floating point value 1994based on @var{regno}. The cost of using @var{regno} for a pseudo will 1995be increased by approximately the pseudo's usage frequency times the 1996value returned by this macro. Not defining this macro is equivalent 1997to having it always return @code{0.0}. 1998 1999On most machines, it is not necessary to define this macro. 2000@end defmac 2001 2002@node Values in Registers 2003@subsection How Values Fit in Registers 2004 2005This section discusses the macros that describe which kinds of values 2006(specifically, which machine modes) each register can hold, and how many 2007consecutive registers are needed for a given mode. 2008 2009@defmac HARD_REGNO_NREGS (@var{regno}, @var{mode}) 2010A C expression for the number of consecutive hard registers, starting 2011at register number @var{regno}, required to hold a value of mode 2012@var{mode}. This macro must never return zero, even if a register 2013cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK 2014and/or CANNOT_CHANGE_MODE_CLASS instead. 2015 2016On a machine where all registers are exactly one word, a suitable 2017definition of this macro is 2018 2019@smallexample 2020#define HARD_REGNO_NREGS(REGNO, MODE) \ 2021 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ 2022 / UNITS_PER_WORD) 2023@end smallexample 2024@end defmac 2025 2026@defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode}) 2027A C expression that is nonzero if a value of mode @var{mode}, stored 2028in memory, ends with padding that causes it to take up more space than 2029in registers starting at register number @var{regno} (as determined by 2030multiplying GCC's notion of the size of the register when containing 2031this mode by the number of registers returned by 2032@code{HARD_REGNO_NREGS}). By default this is zero. 2033 2034For example, if a floating-point value is stored in three 32-bit 2035registers but takes up 128 bits in memory, then this would be 2036nonzero. 2037 2038This macros only needs to be defined if there are cases where 2039@code{subreg_get_info} 2040would otherwise wrongly determine that a @code{subreg} can be 2041represented by an offset to the register number, when in fact such a 2042@code{subreg} would contain some of the padding not stored in 2043registers and so not be representable. 2044@end defmac 2045 2046@defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode}) 2047For values of @var{regno} and @var{mode} for which 2048@code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression 2049returning the greater number of registers required to hold the value 2050including any padding. In the example above, the value would be four. 2051@end defmac 2052 2053@defmac REGMODE_NATURAL_SIZE (@var{mode}) 2054Define this macro if the natural size of registers that hold values 2055of mode @var{mode} is not the word size. It is a C expression that 2056should give the natural size in bytes for the specified mode. It is 2057used by the register allocator to try to optimize its results. This 2058happens for example on SPARC 64-bit where the natural size of 2059floating-point registers is still 32-bit. 2060@end defmac 2061 2062@defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode}) 2063A C expression that is nonzero if it is permissible to store a value 2064of mode @var{mode} in hard register number @var{regno} (or in several 2065registers starting with that one). For a machine where all registers 2066are equivalent, a suitable definition is 2067 2068@smallexample 2069#define HARD_REGNO_MODE_OK(REGNO, MODE) 1 2070@end smallexample 2071 2072You need not include code to check for the numbers of fixed registers, 2073because the allocation mechanism considers them to be always occupied. 2074 2075@cindex register pairs 2076On some machines, double-precision values must be kept in even/odd 2077register pairs. You can implement that by defining this macro to reject 2078odd register numbers for such modes. 2079 2080The minimum requirement for a mode to be OK in a register is that the 2081@samp{mov@var{mode}} instruction pattern support moves between the 2082register and other hard register in the same class and that moving a 2083value into the register and back out not alter it. 2084 2085Since the same instruction used to move @code{word_mode} will work for 2086all narrower integer modes, it is not necessary on any machine for 2087@code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided 2088you define patterns @samp{movhi}, etc., to take advantage of this. This 2089is useful because of the interaction between @code{HARD_REGNO_MODE_OK} 2090and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes 2091to be tieable. 2092 2093Many machines have special registers for floating point arithmetic. 2094Often people assume that floating point machine modes are allowed only 2095in floating point registers. This is not true. Any registers that 2096can hold integers can safely @emph{hold} a floating point machine 2097mode, whether or not floating arithmetic can be done on it in those 2098registers. Integer move instructions can be used to move the values. 2099 2100On some machines, though, the converse is true: fixed-point machine 2101modes may not go in floating registers. This is true if the floating 2102registers normalize any value stored in them, because storing a 2103non-floating value there would garble it. In this case, 2104@code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in 2105floating registers. But if the floating registers do not automatically 2106normalize, if you can store any bit pattern in one and retrieve it 2107unchanged without a trap, then any machine mode may go in a floating 2108register, so you can define this macro to say so. 2109 2110The primary significance of special floating registers is rather that 2111they are the registers acceptable in floating point arithmetic 2112instructions. However, this is of no concern to 2113@code{HARD_REGNO_MODE_OK}. You handle it by writing the proper 2114constraints for those instructions. 2115 2116On some machines, the floating registers are especially slow to access, 2117so that it is better to store a value in a stack frame than in such a 2118register if floating point arithmetic is not being done. As long as the 2119floating registers are not in class @code{GENERAL_REGS}, they will not 2120be used unless some pattern's constraint asks for one. 2121@end defmac 2122 2123@defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to}) 2124A C expression that is nonzero if it is OK to rename a hard register 2125@var{from} to another hard register @var{to}. 2126 2127One common use of this macro is to prevent renaming of a register to 2128another register that is not saved by a prologue in an interrupt 2129handler. 2130 2131The default is always nonzero. 2132@end defmac 2133 2134@defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2}) 2135A C expression that is nonzero if a value of mode 2136@var{mode1} is accessible in mode @var{mode2} without copying. 2137 2138If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and 2139@code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for 2140any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})} 2141should be nonzero. If they differ for any @var{r}, you should define 2142this macro to return zero unless some other mechanism ensures the 2143accessibility of the value in a narrower mode. 2144 2145You should define this macro to return nonzero in as many cases as 2146possible since doing so will allow GCC to perform better register 2147allocation. 2148@end defmac 2149 2150@deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno}) 2151This target hook should return @code{true} if it is OK to use a hard register 2152@var{regno} as scratch reg in peephole2. 2153 2154One common use of this macro is to prevent using of a register that 2155is not saved by a prologue in an interrupt handler. 2156 2157The default version of this hook always returns @code{true}. 2158@end deftypefn 2159 2160@defmac AVOID_CCMODE_COPIES 2161Define this macro if the compiler should avoid copies to/from @code{CCmode} 2162registers. You should only define this macro if support for copying to/from 2163@code{CCmode} is incomplete. 2164@end defmac 2165 2166@node Leaf Functions 2167@subsection Handling Leaf Functions 2168 2169@cindex leaf functions 2170@cindex functions, leaf 2171On some machines, a leaf function (i.e., one which makes no calls) can run 2172more efficiently if it does not make its own register window. Often this 2173means it is required to receive its arguments in the registers where they 2174are passed by the caller, instead of the registers where they would 2175normally arrive. 2176 2177The special treatment for leaf functions generally applies only when 2178other conditions are met; for example, often they may use only those 2179registers for its own variables and temporaries. We use the term ``leaf 2180function'' to mean a function that is suitable for this special 2181handling, so that functions with no calls are not necessarily ``leaf 2182functions''. 2183 2184GCC assigns register numbers before it knows whether the function is 2185suitable for leaf function treatment. So it needs to renumber the 2186registers in order to output a leaf function. The following macros 2187accomplish this. 2188 2189@defmac LEAF_REGISTERS 2190Name of a char vector, indexed by hard register number, which 2191contains 1 for a register that is allowable in a candidate for leaf 2192function treatment. 2193 2194If leaf function treatment involves renumbering the registers, then the 2195registers marked here should be the ones before renumbering---those that 2196GCC would ordinarily allocate. The registers which will actually be 2197used in the assembler code, after renumbering, should not be marked with 1 2198in this vector. 2199 2200Define this macro only if the target machine offers a way to optimize 2201the treatment of leaf functions. 2202@end defmac 2203 2204@defmac LEAF_REG_REMAP (@var{regno}) 2205A C expression whose value is the register number to which @var{regno} 2206should be renumbered, when a function is treated as a leaf function. 2207 2208If @var{regno} is a register number which should not appear in a leaf 2209function before renumbering, then the expression should yield @minus{}1, which 2210will cause the compiler to abort. 2211 2212Define this macro only if the target machine offers a way to optimize the 2213treatment of leaf functions, and registers need to be renumbered to do 2214this. 2215@end defmac 2216 2217@findex current_function_is_leaf 2218@findex current_function_uses_only_leaf_regs 2219@code{TARGET_ASM_FUNCTION_PROLOGUE} and 2220@code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions 2221specially. They can test the C variable @code{current_function_is_leaf} 2222which is nonzero for leaf functions. @code{current_function_is_leaf} is 2223set prior to local register allocation and is valid for the remaining 2224compiler passes. They can also test the C variable 2225@code{current_function_uses_only_leaf_regs} which is nonzero for leaf 2226functions which only use leaf registers. 2227@code{current_function_uses_only_leaf_regs} is valid after all passes 2228that modify the instructions have been run and is only useful if 2229@code{LEAF_REGISTERS} is defined. 2230@c changed this to fix overfull. ALSO: why the "it" at the beginning 2231@c of the next paragraph?! --mew 2feb93 2232 2233@node Stack Registers 2234@subsection Registers That Form a Stack 2235 2236There are special features to handle computers where some of the 2237``registers'' form a stack. Stack registers are normally written by 2238pushing onto the stack, and are numbered relative to the top of the 2239stack. 2240 2241Currently, GCC can only handle one group of stack-like registers, and 2242they must be consecutively numbered. Furthermore, the existing 2243support for stack-like registers is specific to the 80387 floating 2244point coprocessor. If you have a new architecture that uses 2245stack-like registers, you will need to do substantial work on 2246@file{reg-stack.c} and write your machine description to cooperate 2247with it, as well as defining these macros. 2248 2249@defmac STACK_REGS 2250Define this if the machine has any stack-like registers. 2251@end defmac 2252 2253@defmac STACK_REG_COVER_CLASS 2254This is a cover class containing the stack registers. Define this if 2255the machine has any stack-like registers. 2256@end defmac 2257 2258@defmac FIRST_STACK_REG 2259The number of the first stack-like register. This one is the top 2260of the stack. 2261@end defmac 2262 2263@defmac LAST_STACK_REG 2264The number of the last stack-like register. This one is the bottom of 2265the stack. 2266@end defmac 2267 2268@node Register Classes 2269@section Register Classes 2270@cindex register class definitions 2271@cindex class definitions, register 2272 2273On many machines, the numbered registers are not all equivalent. 2274For example, certain registers may not be allowed for indexed addressing; 2275certain registers may not be allowed in some instructions. These machine 2276restrictions are described to the compiler using @dfn{register classes}. 2277 2278You define a number of register classes, giving each one a name and saying 2279which of the registers belong to it. Then you can specify register classes 2280that are allowed as operands to particular instruction patterns. 2281 2282@findex ALL_REGS 2283@findex NO_REGS 2284In general, each register will belong to several classes. In fact, one 2285class must be named @code{ALL_REGS} and contain all the registers. Another 2286class must be named @code{NO_REGS} and contain no registers. Often the 2287union of two classes will be another class; however, this is not required. 2288 2289@findex GENERAL_REGS 2290One of the classes must be named @code{GENERAL_REGS}. There is nothing 2291terribly special about the name, but the operand constraint letters 2292@samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is 2293the same as @code{ALL_REGS}, just define it as a macro which expands 2294to @code{ALL_REGS}. 2295 2296Order the classes so that if class @var{x} is contained in class @var{y} 2297then @var{x} has a lower class number than @var{y}. 2298 2299The way classes other than @code{GENERAL_REGS} are specified in operand 2300constraints is through machine-dependent operand constraint letters. 2301You can define such letters to correspond to various classes, then use 2302them in operand constraints. 2303 2304You must define the narrowest register classes for allocatable 2305registers, so that each class either has no subclasses, or that for 2306some mode, the move cost between registers within the class is 2307cheaper than moving a register in the class to or from memory 2308(@pxref{Costs}). 2309 2310You should define a class for the union of two classes whenever some 2311instruction allows both classes. For example, if an instruction allows 2312either a floating point (coprocessor) register or a general register for a 2313certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS} 2314which includes both of them. Otherwise you will get suboptimal code, 2315or even internal compiler errors when reload cannot find a register in the 2316class computed via @code{reg_class_subunion}. 2317 2318You must also specify certain redundant information about the register 2319classes: for each class, which classes contain it and which ones are 2320contained in it; for each pair of classes, the largest class contained 2321in their union. 2322 2323When a value occupying several consecutive registers is expected in a 2324certain class, all the registers used must belong to that class. 2325Therefore, register classes cannot be used to enforce a requirement for 2326a register pair to start with an even-numbered register. The way to 2327specify this requirement is with @code{HARD_REGNO_MODE_OK}. 2328 2329Register classes used for input-operands of bitwise-and or shift 2330instructions have a special requirement: each such class must have, for 2331each fixed-point machine mode, a subclass whose registers can transfer that 2332mode to or from memory. For example, on some machines, the operations for 2333single-byte values (@code{QImode}) are limited to certain registers. When 2334this is so, each register class that is used in a bitwise-and or shift 2335instruction must have a subclass consisting of registers from which 2336single-byte values can be loaded or stored. This is so that 2337@code{PREFERRED_RELOAD_CLASS} can always have a possible value to return. 2338 2339@deftp {Data type} {enum reg_class} 2340An enumerated type that must be defined with all the register class names 2341as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS} 2342must be the last register class, followed by one more enumerated value, 2343@code{LIM_REG_CLASSES}, which is not a register class but rather 2344tells how many classes there are. 2345 2346Each register class has a number, which is the value of casting 2347the class name to type @code{int}. The number serves as an index 2348in many of the tables described below. 2349@end deftp 2350 2351@defmac N_REG_CLASSES 2352The number of distinct register classes, defined as follows: 2353 2354@smallexample 2355#define N_REG_CLASSES (int) LIM_REG_CLASSES 2356@end smallexample 2357@end defmac 2358 2359@defmac REG_CLASS_NAMES 2360An initializer containing the names of the register classes as C string 2361constants. These names are used in writing some of the debugging dumps. 2362@end defmac 2363 2364@defmac REG_CLASS_CONTENTS 2365An initializer containing the contents of the register classes, as integers 2366which are bit masks. The @var{n}th integer specifies the contents of class 2367@var{n}. The way the integer @var{mask} is interpreted is that 2368register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1. 2369 2370When the machine has more than 32 registers, an integer does not suffice. 2371Then the integers are replaced by sub-initializers, braced groupings containing 2372several integers. Each sub-initializer must be suitable as an initializer 2373for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}. 2374In this situation, the first integer in each sub-initializer corresponds to 2375registers 0 through 31, the second integer to registers 32 through 63, and 2376so on. 2377@end defmac 2378 2379@defmac REGNO_REG_CLASS (@var{regno}) 2380A C expression whose value is a register class containing hard register 2381@var{regno}. In general there is more than one such class; choose a class 2382which is @dfn{minimal}, meaning that no smaller class also contains the 2383register. 2384@end defmac 2385 2386@defmac BASE_REG_CLASS 2387A macro whose definition is the name of the class to which a valid 2388base register must belong. A base register is one used in an address 2389which is the register value plus a displacement. 2390@end defmac 2391 2392@defmac MODE_BASE_REG_CLASS (@var{mode}) 2393This is a variation of the @code{BASE_REG_CLASS} macro which allows 2394the selection of a base register in a mode dependent manner. If 2395@var{mode} is VOIDmode then it should return the same value as 2396@code{BASE_REG_CLASS}. 2397@end defmac 2398 2399@defmac MODE_BASE_REG_REG_CLASS (@var{mode}) 2400A C expression whose value is the register class to which a valid 2401base register must belong in order to be used in a base plus index 2402register address. You should define this macro if base plus index 2403addresses have different requirements than other base register uses. 2404@end defmac 2405 2406@defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code}) 2407A C expression whose value is the register class to which a valid 2408base register for a memory reference in mode @var{mode} to address 2409space @var{address_space} must belong. @var{outer_code} and @var{index_code} 2410define the context in which the base register occurs. @var{outer_code} is 2411the code of the immediately enclosing expression (@code{MEM} for the top level 2412of an address, @code{ADDRESS} for something that occurs in an 2413@code{address_operand}). @var{index_code} is the code of the corresponding 2414index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise. 2415@end defmac 2416 2417@defmac INDEX_REG_CLASS 2418A macro whose definition is the name of the class to which a valid 2419index register must belong. An index register is one used in an 2420address where its value is either multiplied by a scale factor or 2421added to another register (as well as added to a displacement). 2422@end defmac 2423 2424@defmac REGNO_OK_FOR_BASE_P (@var{num}) 2425A C expression which is nonzero if register number @var{num} is 2426suitable for use as a base register in operand addresses. 2427@end defmac 2428 2429@defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode}) 2430A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that 2431that expression may examine the mode of the memory reference in 2432@var{mode}. You should define this macro if the mode of the memory 2433reference affects whether a register may be used as a base register. If 2434you define this macro, the compiler will use it instead of 2435@code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for 2436addresses that appear outside a @code{MEM}, i.e., as an 2437@code{address_operand}. 2438@end defmac 2439 2440@defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode}) 2441A C expression which is nonzero if register number @var{num} is suitable for 2442use as a base register in base plus index operand addresses, accessing 2443memory in mode @var{mode}. It may be either a suitable hard register or a 2444pseudo register that has been allocated such a hard register. You should 2445define this macro if base plus index addresses have different requirements 2446than other base register uses. 2447 2448Use of this macro is deprecated; please use the more general 2449@code{REGNO_MODE_CODE_OK_FOR_BASE_P}. 2450@end defmac 2451 2452@defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code}) 2453A C expression which is nonzero if register number @var{num} is 2454suitable for use as a base register in operand addresses, accessing 2455memory in mode @var{mode} in address space @var{address_space}. 2456This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except 2457that that expression may examine the context in which the register 2458appears in the memory reference. @var{outer_code} is the code of the 2459immediately enclosing expression (@code{MEM} if at the top level of the 2460address, @code{ADDRESS} for something that occurs in an 2461@code{address_operand}). @var{index_code} is the code of the 2462corresponding index expression if @var{outer_code} is @code{PLUS}; 2463@code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses 2464that appear outside a @code{MEM}, i.e., as an @code{address_operand}. 2465@end defmac 2466 2467@defmac REGNO_OK_FOR_INDEX_P (@var{num}) 2468A C expression which is nonzero if register number @var{num} is 2469suitable for use as an index register in operand addresses. It may be 2470either a suitable hard register or a pseudo register that has been 2471allocated such a hard register. 2472 2473The difference between an index register and a base register is that 2474the index register may be scaled. If an address involves the sum of 2475two registers, neither one of them scaled, then either one may be 2476labeled the ``base'' and the other the ``index''; but whichever 2477labeling is used must fit the machine's constraints of which registers 2478may serve in each capacity. The compiler will try both labelings, 2479looking for one that is valid, and will reload one or both registers 2480only if neither labeling works. 2481@end defmac 2482 2483@deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass}) 2484A target hook that places additional preference on the register class to use when it is necessary to rename a register in class @var{rclass} to another class, or perhaps @var{NO_REGS}, if no preferred register class is found or hook @code{preferred_rename_class} is not implemented. Sometimes returning a more restrictive class makes better code. For example, on ARM, thumb-2 instructions using @code{LO_REGS} may be smaller than instructions using @code{GENERIC_REGS}. By returning @code{LO_REGS} from @code{preferred_rename_class}, code size can be reduced. 2485@end deftypefn 2486 2487@deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass}) 2488A target hook that places additional restrictions on the register class 2489to use when it is necessary to copy value @var{x} into a register in class 2490@var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps 2491another, smaller class. 2492 2493The default version of this hook always returns value of @code{rclass} argument. 2494 2495Sometimes returning a more restrictive class makes better code. For 2496example, on the 68000, when @var{x} is an integer constant that is in range 2497for a @samp{moveq} instruction, the value of this macro is always 2498@code{DATA_REGS} as long as @var{rclass} includes the data registers. 2499Requiring a data register guarantees that a @samp{moveq} will be used. 2500 2501One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return 2502@var{rclass} is if @var{x} is a legitimate constant which cannot be 2503loaded into some register class. By returning @code{NO_REGS} you can 2504force @var{x} into a memory location. For example, rs6000 can load 2505immediate values into general-purpose registers, but does not have an 2506instruction for loading an immediate value into a floating-point 2507register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when 2508@var{x} is a floating-point constant. If the constant can't be loaded 2509into any kind of register, code generation will be better if 2510@code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead 2511of using @code{TARGET_PREFERRED_RELOAD_CLASS}. 2512 2513If an insn has pseudos in it after register allocation, reload will go 2514through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS} 2515to find the best one. Returning @code{NO_REGS}, in this case, makes 2516reload add a @code{!} in front of the constraint: the x86 back-end uses 2517this feature to discourage usage of 387 registers when math is done in 2518the SSE registers (and vice versa). 2519@end deftypefn 2520 2521@defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class}) 2522A C expression that places additional restrictions on the register class 2523to use when it is necessary to copy value @var{x} into a register in class 2524@var{class}. The value is a register class; perhaps @var{class}, or perhaps 2525another, smaller class. On many machines, the following definition is 2526safe: 2527 2528@smallexample 2529#define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS 2530@end smallexample 2531 2532Sometimes returning a more restrictive class makes better code. For 2533example, on the 68000, when @var{x} is an integer constant that is in range 2534for a @samp{moveq} instruction, the value of this macro is always 2535@code{DATA_REGS} as long as @var{class} includes the data registers. 2536Requiring a data register guarantees that a @samp{moveq} will be used. 2537 2538One case where @code{PREFERRED_RELOAD_CLASS} must not return 2539@var{class} is if @var{x} is a legitimate constant which cannot be 2540loaded into some register class. By returning @code{NO_REGS} you can 2541force @var{x} into a memory location. For example, rs6000 can load 2542immediate values into general-purpose registers, but does not have an 2543instruction for loading an immediate value into a floating-point 2544register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when 2545@var{x} is a floating-point constant. If the constant can't be loaded 2546into any kind of register, code generation will be better if 2547@code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead 2548of using @code{TARGET_PREFERRED_RELOAD_CLASS}. 2549 2550If an insn has pseudos in it after register allocation, reload will go 2551through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS} 2552to find the best one. Returning @code{NO_REGS}, in this case, makes 2553reload add a @code{!} in front of the constraint: the x86 back-end uses 2554this feature to discourage usage of 387 registers when math is done in 2555the SSE registers (and vice versa). 2556@end defmac 2557 2558@deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass}) 2559Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of 2560input reloads. 2561 2562The default version of this hook always returns value of @code{rclass} 2563argument. 2564 2565You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage 2566reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}. 2567@end deftypefn 2568 2569@defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class}) 2570A C expression that places additional restrictions on the register class 2571to use when it is necessary to be able to hold a value of mode 2572@var{mode} in a reload register for which class @var{class} would 2573ordinarily be used. 2574 2575Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when 2576there are certain modes that simply can't go in certain reload classes. 2577 2578The value is a register class; perhaps @var{class}, or perhaps another, 2579smaller class. 2580 2581Don't define this macro unless the target machine has limitations which 2582require the macro to do something nontrivial. 2583@end defmac 2584 2585@deftypefn {Target Hook} reg_class_t TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, reg_class_t @var{reload_class}, machine_mode @var{reload_mode}, secondary_reload_info *@var{sri}) 2586Many machines have some registers that cannot be copied directly to or 2587from memory or even from other types of registers. An example is the 2588@samp{MQ} register, which on most machines, can only be copied to or 2589from general registers, but not memory. Below, we shall be using the 2590term 'intermediate register' when a move operation cannot be performed 2591directly, but has to be done by copying the source into the intermediate 2592register first, and then copying the intermediate register to the 2593destination. An intermediate register always has the same mode as 2594source and destination. Since it holds the actual value being copied, 2595reload might apply optimizations to re-use an intermediate register 2596and eliding the copy from the source when it can determine that the 2597intermediate register still holds the required value. 2598 2599Another kind of secondary reload is required on some machines which 2600allow copying all registers to and from memory, but require a scratch 2601register for stores to some memory locations (e.g., those with symbolic 2602address on the RT, and those with certain symbolic address on the SPARC 2603when compiling PIC)@. Scratch registers need not have the same mode 2604as the value being copied, and usually hold a different value than 2605that being copied. Special patterns in the md file are needed to 2606describe how the copy is performed with the help of the scratch register; 2607these patterns also describe the number, register class(es) and mode(s) 2608of the scratch register(s). 2609 2610In some cases, both an intermediate and a scratch register are required. 2611 2612For input reloads, this target hook is called with nonzero @var{in_p}, 2613and @var{x} is an rtx that needs to be copied to a register of class 2614@var{reload_class} in @var{reload_mode}. For output reloads, this target 2615hook is called with zero @var{in_p}, and a register of class @var{reload_class} 2616needs to be copied to rtx @var{x} in @var{reload_mode}. 2617 2618If copying a register of @var{reload_class} from/to @var{x} requires 2619an intermediate register, the hook @code{secondary_reload} should 2620return the register class required for this intermediate register. 2621If no intermediate register is required, it should return NO_REGS. 2622If more than one intermediate register is required, describe the one 2623that is closest in the copy chain to the reload register. 2624 2625If scratch registers are needed, you also have to describe how to 2626perform the copy from/to the reload register to/from this 2627closest intermediate register. Or if no intermediate register is 2628required, but still a scratch register is needed, describe the 2629copy from/to the reload register to/from the reload operand @var{x}. 2630 2631You do this by setting @code{sri->icode} to the instruction code of a pattern 2632in the md file which performs the move. Operands 0 and 1 are the output 2633and input of this copy, respectively. Operands from operand 2 onward are 2634for scratch operands. These scratch operands must have a mode, and a 2635single-register-class 2636@c [later: or memory] 2637output constraint. 2638 2639When an intermediate register is used, the @code{secondary_reload} 2640hook will be called again to determine how to copy the intermediate 2641register to/from the reload operand @var{x}, so your hook must also 2642have code to handle the register class of the intermediate operand. 2643 2644@c [For later: maybe we'll allow multi-alternative reload patterns - 2645@c the port maintainer could name a mov<mode> pattern that has clobbers - 2646@c and match the constraints of input and output to determine the required 2647@c alternative. A restriction would be that constraints used to match 2648@c against reloads registers would have to be written as register class 2649@c constraints, or we need a new target macro / hook that tells us if an 2650@c arbitrary constraint can match an unknown register of a given class. 2651@c Such a macro / hook would also be useful in other places.] 2652 2653 2654@var{x} might be a pseudo-register or a @code{subreg} of a 2655pseudo-register, which could either be in a hard register or in memory. 2656Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is 2657in memory and the hard register number if it is in a register. 2658 2659Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are 2660currently not supported. For the time being, you will have to continue 2661to use @code{SECONDARY_MEMORY_NEEDED} for that purpose. 2662 2663@code{copy_cost} also uses this target hook to find out how values are 2664copied. If you want it to include some extra cost for the need to allocate 2665(a) scratch register(s), set @code{sri->extra_cost} to the additional cost. 2666Or if two dependent moves are supposed to have a lower cost than the sum 2667of the individual moves due to expected fortuitous scheduling and/or special 2668forwarding logic, you can set @code{sri->extra_cost} to a negative amount. 2669@end deftypefn 2670 2671@defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2672@defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2673@defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2674These macros are obsolete, new ports should use the target hook 2675@code{TARGET_SECONDARY_RELOAD} instead. 2676 2677These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD} 2678target hook. Older ports still define these macros to indicate to the 2679reload phase that it may 2680need to allocate at least one register for a reload in addition to the 2681register to contain the data. Specifically, if copying @var{x} to a 2682register @var{class} in @var{mode} requires an intermediate register, 2683you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the 2684largest register class all of whose registers can be used as 2685intermediate registers or scratch registers. 2686 2687If copying a register @var{class} in @var{mode} to @var{x} requires an 2688intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS} 2689was supposed to be defined be defined to return the largest register 2690class required. If the 2691requirements for input and output reloads were the same, the macro 2692@code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both 2693macros identically. 2694 2695The values returned by these macros are often @code{GENERAL_REGS}. 2696Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x} 2697can be directly copied to or from a register of @var{class} in 2698@var{mode} without requiring a scratch register. Do not define this 2699macro if it would always return @code{NO_REGS}. 2700 2701If a scratch register is required (either with or without an 2702intermediate register), you were supposed to define patterns for 2703@samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required 2704(@pxref{Standard Names}. These patterns, which were normally 2705implemented with a @code{define_expand}, should be similar to the 2706@samp{mov@var{m}} patterns, except that operand 2 is the scratch 2707register. 2708 2709These patterns need constraints for the reload register and scratch 2710register that 2711contain a single register class. If the original reload register (whose 2712class is @var{class}) can meet the constraint given in the pattern, the 2713value returned by these macros is used for the class of the scratch 2714register. Otherwise, two additional reload registers are required. 2715Their classes are obtained from the constraints in the insn pattern. 2716 2717@var{x} might be a pseudo-register or a @code{subreg} of a 2718pseudo-register, which could either be in a hard register or in memory. 2719Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is 2720in memory and the hard register number if it is in a register. 2721 2722These macros should not be used in the case where a particular class of 2723registers can only be copied to memory and not to another class of 2724registers. In that case, secondary reload registers are not needed and 2725would not be helpful. Instead, a stack location must be used to perform 2726the copy and the @code{mov@var{m}} pattern should use memory as an 2727intermediate storage. This case often occurs between floating-point and 2728general registers. 2729@end defmac 2730 2731@defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m}) 2732Certain machines have the property that some registers cannot be copied 2733to some other registers without using memory. Define this macro on 2734those machines to be a C expression that is nonzero if objects of mode 2735@var{m} in registers of @var{class1} can only be copied to registers of 2736class @var{class2} by storing a register of @var{class1} into memory 2737and loading that memory location into a register of @var{class2}. 2738 2739Do not define this macro if its value would always be zero. 2740@end defmac 2741 2742@defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode}) 2743Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler 2744allocates a stack slot for a memory location needed for register copies. 2745If this macro is defined, the compiler instead uses the memory location 2746defined by this macro. 2747 2748Do not define this macro if you do not define 2749@code{SECONDARY_MEMORY_NEEDED}. 2750@end defmac 2751 2752@defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode}) 2753When the compiler needs a secondary memory location to copy between two 2754registers of mode @var{mode}, it normally allocates sufficient memory to 2755hold a quantity of @code{BITS_PER_WORD} bits and performs the store and 2756load operations in a mode that many bits wide and whose class is the 2757same as that of @var{mode}. 2758 2759This is right thing to do on most machines because it ensures that all 2760bits of the register are copied and prevents accesses to the registers 2761in a narrower mode, which some machines prohibit for floating-point 2762registers. 2763 2764However, this default behavior is not correct on some machines, such as 2765the DEC Alpha, that store short integers in floating-point registers 2766differently than in integer registers. On those machines, the default 2767widening will not work correctly and you must define this macro to 2768suppress that widening in some cases. See the file @file{alpha.h} for 2769details. 2770 2771Do not define this macro if you do not define 2772@code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that 2773is @code{BITS_PER_WORD} bits wide is correct for your machine. 2774@end defmac 2775 2776@deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass}) 2777A target hook which returns @code{true} if pseudos that have been assigned 2778to registers of class @var{rclass} would likely be spilled because 2779registers of @var{rclass} are needed for spill registers. 2780 2781The default version of this target hook returns @code{true} if @var{rclass} 2782has exactly one register and @code{false} otherwise. On most machines, this 2783default should be used. For generally register-starved machines, such as 2784i386, or machines with right register constraints, such as SH, this hook 2785can be used to avoid excessive spilling. 2786 2787This hook is also used by some of the global intra-procedural code 2788transformations to throtle code motion, to avoid increasing register 2789pressure. 2790@end deftypefn 2791 2792@deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, machine_mode @var{mode}) 2793A target hook returns the maximum number of consecutive registers 2794of class @var{rclass} needed to hold a value of mode @var{mode}. 2795 2796This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact, 2797the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass}, 2798@var{mode})} target hook should be the maximum value of 2799@code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno} 2800values in the class @var{rclass}. 2801 2802This target hook helps control the handling of multiple-word values 2803in the reload pass. 2804 2805The default version of this target hook returns the size of @var{mode} 2806in words. 2807@end deftypefn 2808 2809@defmac CLASS_MAX_NREGS (@var{class}, @var{mode}) 2810A C expression for the maximum number of consecutive registers 2811of class @var{class} needed to hold a value of mode @var{mode}. 2812 2813This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact, 2814the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})} 2815should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, 2816@var{mode})} for all @var{regno} values in the class @var{class}. 2817 2818This macro helps control the handling of multiple-word values 2819in the reload pass. 2820@end defmac 2821 2822@defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class}) 2823If defined, a C expression that returns nonzero for a @var{class} for which 2824a change from mode @var{from} to mode @var{to} is invalid. 2825 2826For example, loading 32-bit integer or floating-point objects into 2827floating-point registers on Alpha extends them to 64 bits. 2828Therefore loading a 64-bit object and then storing it as a 32-bit object 2829does not store the low-order 32 bits, as would be the case for a normal 2830register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS} 2831as below: 2832 2833@smallexample 2834#define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \ 2835 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \ 2836 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0) 2837@end smallexample 2838 2839Even if storing from a register in mode @var{to} would be valid, 2840if both @var{from} and @code{raw_reg_mode} for @var{class} are wider 2841than @code{word_mode}, then we must prevent @var{to} narrowing the 2842mode. This happens when the middle-end assumes that it can load 2843or store pieces of an @var{N}-word pseudo, and that the pseudo will 2844eventually be allocated to @var{N} @code{word_mode} hard registers. 2845Failure to prevent this kind of mode change will result in the 2846entire @code{raw_reg_mode} being modified instead of the partial 2847value that the middle-end intended. 2848 2849@end defmac 2850 2851@deftypefn {Target Hook} bool TARGET_LRA_P (void) 2852A target hook which returns true if we use LRA instead of reload pass. It means that LRA was ported to the target. The default version of this target hook returns always false. 2853@end deftypefn 2854 2855@deftypefn {Target Hook} int TARGET_REGISTER_PRIORITY (int) 2856A target hook which returns the register priority number to which the register @var{hard_regno} belongs to. The bigger the number, the more preferable the hard register usage (when all other conditions are the same). This hook can be used to prefer some hard register over others in LRA. For example, some x86-64 register usage needs additional prefix which makes instructions longer. The hook can return lower priority number for such registers make them less favorable and as result making the generated code smaller. The default version of this target hook returns always zero. 2857@end deftypefn 2858 2859@deftypefn {Target Hook} bool TARGET_REGISTER_USAGE_LEVELING_P (void) 2860A target hook which returns true if we need register usage leveling. That means if a few hard registers are equally good for the assignment, we choose the least used hard register. The register usage leveling may be profitable for some targets. Don't use the usage leveling for targets with conditional execution or targets with big register files as it hurts if-conversion and cross-jumping optimizations. The default version of this target hook returns always false. 2861@end deftypefn 2862 2863@deftypefn {Target Hook} bool TARGET_DIFFERENT_ADDR_DISPLACEMENT_P (void) 2864A target hook which returns true if an address with the same structure can have different maximal legitimate displacement. For example, the displacement can depend on memory mode or on operand combinations in the insn. The default version of this target hook returns always false. 2865@end deftypefn 2866 2867@deftypefn {Target Hook} bool TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P (rtx @var{subst}) 2868A target hook which returns @code{true} if @var{subst} can't 2869substitute safely pseudos with equivalent memory values during 2870register allocation. 2871The default version of this target hook returns @code{false}. 2872On most machines, this default should be used. For generally 2873machines with non orthogonal register usage for addressing, such 2874as SH, this hook can be used to avoid excessive spilling. 2875@end deftypefn 2876 2877@deftypefn {Target Hook} bool TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT (rtx *@var{disp}, rtx *@var{offset}, machine_mode @var{mode}) 2878A target hook which returns @code{true} if *@var{disp} is 2879legitimezed to valid address displacement with subtracting *@var{offset} 2880at memory mode @var{mode}. 2881The default version of this target hook returns @code{false}. 2882This hook will benefit machines with limited base plus displacement 2883addressing. 2884@end deftypefn 2885 2886@deftypefn {Target Hook} reg_class_t TARGET_SPILL_CLASS (reg_class_t, @var{machine_mode}) 2887This hook defines a class of registers which could be used for spilling pseudos of the given mode and class, or @code{NO_REGS} if only memory should be used. Not defining this hook is equivalent to returning @code{NO_REGS} for all inputs. 2888@end deftypefn 2889 2890@deftypefn {Target Hook} machine_mode TARGET_CSTORE_MODE (enum insn_code @var{icode}) 2891This hook defines the machine mode to use for the boolean result of conditional store patterns. The ICODE argument is the instruction code for the cstore being performed. Not definiting this hook is the same as accepting the mode encoded into operand 0 of the cstore expander patterns. 2892@end deftypefn 2893 2894@node Stack and Calling 2895@section Stack Layout and Calling Conventions 2896@cindex calling conventions 2897 2898@c prevent bad page break with this line 2899This describes the stack layout and calling conventions. 2900 2901@menu 2902* Frame Layout:: 2903* Exception Handling:: 2904* Stack Checking:: 2905* Frame Registers:: 2906* Elimination:: 2907* Stack Arguments:: 2908* Register Arguments:: 2909* Scalar Return:: 2910* Aggregate Return:: 2911* Caller Saves:: 2912* Function Entry:: 2913* Profiling:: 2914* Tail Calls:: 2915* Stack Smashing Protection:: 2916* Miscellaneous Register Hooks:: 2917@end menu 2918 2919@node Frame Layout 2920@subsection Basic Stack Layout 2921@cindex stack frame layout 2922@cindex frame layout 2923 2924@c prevent bad page break with this line 2925Here is the basic stack layout. 2926 2927@defmac STACK_GROWS_DOWNWARD 2928Define this macro if pushing a word onto the stack moves the stack 2929pointer to a smaller address. 2930 2931When we say, ``define this macro if @dots{}'', it means that the 2932compiler checks this macro only with @code{#ifdef} so the precise 2933definition used does not matter. 2934@end defmac 2935 2936@defmac STACK_PUSH_CODE 2937This macro defines the operation used when something is pushed 2938on the stack. In RTL, a push operation will be 2939@code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})} 2940 2941The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC}, 2942and @code{POST_INC}. Which of these is correct depends on 2943the stack direction and on whether the stack pointer points 2944to the last item on the stack or whether it points to the 2945space for the next item on the stack. 2946 2947The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is 2948defined, which is almost always right, and @code{PRE_INC} otherwise, 2949which is often wrong. 2950@end defmac 2951 2952@defmac FRAME_GROWS_DOWNWARD 2953Define this macro to nonzero value if the addresses of local variable slots 2954are at negative offsets from the frame pointer. 2955@end defmac 2956 2957@defmac ARGS_GROW_DOWNWARD 2958Define this macro if successive arguments to a function occupy decreasing 2959addresses on the stack. 2960@end defmac 2961 2962@defmac STARTING_FRAME_OFFSET 2963Offset from the frame pointer to the first local variable slot to be allocated. 2964 2965If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by 2966subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}. 2967Otherwise, it is found by adding the length of the first slot to the 2968value @code{STARTING_FRAME_OFFSET}. 2969@c i'm not sure if the above is still correct.. had to change it to get 2970@c rid of an overfull. --mew 2feb93 2971@end defmac 2972 2973@defmac STACK_ALIGNMENT_NEEDED 2974Define to zero to disable final alignment of the stack during reload. 2975The nonzero default for this macro is suitable for most ports. 2976 2977On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there 2978is a register save block following the local block that doesn't require 2979alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable 2980stack alignment and do it in the backend. 2981@end defmac 2982 2983@defmac STACK_POINTER_OFFSET 2984Offset from the stack pointer register to the first location at which 2985outgoing arguments are placed. If not specified, the default value of 2986zero is used. This is the proper value for most machines. 2987 2988If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above 2989the first location at which outgoing arguments are placed. 2990@end defmac 2991 2992@defmac FIRST_PARM_OFFSET (@var{fundecl}) 2993Offset from the argument pointer register to the first argument's 2994address. On some machines it may depend on the data type of the 2995function. 2996 2997If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above 2998the first argument's address. 2999@end defmac 3000 3001@defmac STACK_DYNAMIC_OFFSET (@var{fundecl}) 3002Offset from the stack pointer register to an item dynamically allocated 3003on the stack, e.g., by @code{alloca}. 3004 3005The default value for this macro is @code{STACK_POINTER_OFFSET} plus the 3006length of the outgoing arguments. The default is correct for most 3007machines. See @file{function.c} for details. 3008@end defmac 3009 3010@defmac INITIAL_FRAME_ADDRESS_RTX 3011A C expression whose value is RTL representing the address of the initial 3012stack frame. This address is passed to @code{RETURN_ADDR_RTX} and 3013@code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable 3014default value will be used. Define this macro in order to make frame pointer 3015elimination work in the presence of @code{__builtin_frame_address (count)} and 3016@code{__builtin_return_address (count)} for @code{count} not equal to zero. 3017@end defmac 3018 3019@defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr}) 3020A C expression whose value is RTL representing the address in a stack 3021frame where the pointer to the caller's frame is stored. Assume that 3022@var{frameaddr} is an RTL expression for the address of the stack frame 3023itself. 3024 3025If you don't define this macro, the default is to return the value 3026of @var{frameaddr}---that is, the stack frame address is also the 3027address of the stack word that points to the previous frame. 3028@end defmac 3029 3030@defmac SETUP_FRAME_ADDRESSES 3031If defined, a C expression that produces the machine-specific code to 3032setup the stack so that arbitrary frames can be accessed. For example, 3033on the SPARC, we must flush all of the register windows to the stack 3034before we can access arbitrary stack frames. You will seldom need to 3035define this macro. 3036@end defmac 3037 3038@deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void) 3039This target hook should return an rtx that is used to store 3040the address of the current frame into the built in @code{setjmp} buffer. 3041The default value, @code{virtual_stack_vars_rtx}, is correct for most 3042machines. One reason you may need to define this target hook is if 3043@code{hard_frame_pointer_rtx} is the appropriate value on your machine. 3044@end deftypefn 3045 3046@defmac FRAME_ADDR_RTX (@var{frameaddr}) 3047A C expression whose value is RTL representing the value of the frame 3048address for the current frame. @var{frameaddr} is the frame pointer 3049of the current frame. This is used for __builtin_frame_address. 3050You need only define this macro if the frame address is not the same 3051as the frame pointer. Most machines do not need to define it. 3052@end defmac 3053 3054@defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr}) 3055A C expression whose value is RTL representing the value of the return 3056address for the frame @var{count} steps up from the current frame, after 3057the prologue. @var{frameaddr} is the frame pointer of the @var{count} 3058frame, or the frame pointer of the @var{count} @minus{} 1 frame if 3059@code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero. 3060 3061The value of the expression must always be the correct address when 3062@var{count} is zero, but may be @code{NULL_RTX} if there is no way to 3063determine the return address of other frames. 3064@end defmac 3065 3066@defmac RETURN_ADDR_IN_PREVIOUS_FRAME 3067Define this macro to nonzero value if the return address of a particular 3068stack frame is accessed from the frame pointer of the previous stack 3069frame. The zero default for this macro is suitable for most ports. 3070@end defmac 3071 3072@defmac INCOMING_RETURN_ADDR_RTX 3073A C expression whose value is RTL representing the location of the 3074incoming return address at the beginning of any function, before the 3075prologue. This RTL is either a @code{REG}, indicating that the return 3076value is saved in @samp{REG}, or a @code{MEM} representing a location in 3077the stack. 3078 3079You only need to define this macro if you want to support call frame 3080debugging information like that provided by DWARF 2. 3081 3082If this RTL is a @code{REG}, you should also define 3083@code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}. 3084@end defmac 3085 3086@defmac DWARF_ALT_FRAME_RETURN_COLUMN 3087A C expression whose value is an integer giving a DWARF 2 column 3088number that may be used as an alternative return column. The column 3089must not correspond to any gcc hard register (that is, it must not 3090be in the range of @code{DWARF_FRAME_REGNUM}). 3091 3092This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a 3093general register, but an alternative column needs to be used for signal 3094frames. Some targets have also used different frame return columns 3095over time. 3096@end defmac 3097 3098@defmac DWARF_ZERO_REG 3099A C expression whose value is an integer giving a DWARF 2 register 3100number that is considered to always have the value zero. This should 3101only be defined if the target has an architected zero register, and 3102someone decided it was a good idea to use that register number to 3103terminate the stack backtrace. New ports should avoid this. 3104@end defmac 3105 3106@deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index}) 3107This target hook allows the backend to emit frame-related insns that 3108contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging 3109info engine will invoke it on insns of the form 3110@smallexample 3111(set (reg) (unspec [@dots{}] UNSPEC_INDEX)) 3112@end smallexample 3113and 3114@smallexample 3115(set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)). 3116@end smallexample 3117to let the backend emit the call frame instructions. @var{label} is 3118the CFI label attached to the insn, @var{pattern} is the pattern of 3119the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}. 3120@end deftypefn 3121 3122@defmac INCOMING_FRAME_SP_OFFSET 3123A C expression whose value is an integer giving the offset, in bytes, 3124from the value of the stack pointer register to the top of the stack 3125frame at the beginning of any function, before the prologue. The top of 3126the frame is defined to be the value of the stack pointer in the 3127previous frame, just before the call instruction. 3128 3129You only need to define this macro if you want to support call frame 3130debugging information like that provided by DWARF 2. 3131@end defmac 3132 3133@defmac ARG_POINTER_CFA_OFFSET (@var{fundecl}) 3134A C expression whose value is an integer giving the offset, in bytes, 3135from the argument pointer to the canonical frame address (cfa). The 3136final value should coincide with that calculated by 3137@code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable 3138during virtual register instantiation. 3139 3140The default value for this macro is 3141@code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size}, 3142which is correct for most machines; in general, the arguments are found 3143immediately before the stack frame. Note that this is not the case on 3144some targets that save registers into the caller's frame, such as SPARC 3145and rs6000, and so such targets need to define this macro. 3146 3147You only need to define this macro if the default is incorrect, and you 3148want to support call frame debugging information like that provided by 3149DWARF 2. 3150@end defmac 3151 3152@defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl}) 3153If defined, a C expression whose value is an integer giving the offset 3154in bytes from the frame pointer to the canonical frame address (cfa). 3155The final value should coincide with that calculated by 3156@code{INCOMING_FRAME_SP_OFFSET}. 3157 3158Normally the CFA is calculated as an offset from the argument pointer, 3159via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is 3160variable due to the ABI, this may not be possible. If this macro is 3161defined, it implies that the virtual register instantiation should be 3162based on the frame pointer instead of the argument pointer. Only one 3163of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET} 3164should be defined. 3165@end defmac 3166 3167@defmac CFA_FRAME_BASE_OFFSET (@var{fundecl}) 3168If defined, a C expression whose value is an integer giving the offset 3169in bytes from the canonical frame address (cfa) to the frame base used 3170in DWARF 2 debug information. The default is zero. A different value 3171may reduce the size of debug information on some ports. 3172@end defmac 3173 3174@node Exception Handling 3175@subsection Exception Handling Support 3176@cindex exception handling 3177 3178@defmac EH_RETURN_DATA_REGNO (@var{N}) 3179A C expression whose value is the @var{N}th register number used for 3180data by exception handlers, or @code{INVALID_REGNUM} if fewer than 3181@var{N} registers are usable. 3182 3183The exception handling library routines communicate with the exception 3184handlers via a set of agreed upon registers. Ideally these registers 3185should be call-clobbered; it is possible to use call-saved registers, 3186but may negatively impact code size. The target must support at least 31872 data registers, but should define 4 if there are enough free registers. 3188 3189You must define this macro if you want to support call frame exception 3190handling like that provided by DWARF 2. 3191@end defmac 3192 3193@defmac EH_RETURN_STACKADJ_RTX 3194A C expression whose value is RTL representing a location in which 3195to store a stack adjustment to be applied before function return. 3196This is used to unwind the stack to an exception handler's call frame. 3197It will be assigned zero on code paths that return normally. 3198 3199Typically this is a call-clobbered hard register that is otherwise 3200untouched by the epilogue, but could also be a stack slot. 3201 3202Do not define this macro if the stack pointer is saved and restored 3203by the regular prolog and epilog code in the call frame itself; in 3204this case, the exception handling library routines will update the 3205stack location to be restored in place. Otherwise, you must define 3206this macro if you want to support call frame exception handling like 3207that provided by DWARF 2. 3208@end defmac 3209 3210@defmac EH_RETURN_HANDLER_RTX 3211A C expression whose value is RTL representing a location in which 3212to store the address of an exception handler to which we should 3213return. It will not be assigned on code paths that return normally. 3214 3215Typically this is the location in the call frame at which the normal 3216return address is stored. For targets that return by popping an 3217address off the stack, this might be a memory address just below 3218the @emph{target} call frame rather than inside the current call 3219frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already 3220been assigned, so it may be used to calculate the location of the 3221target call frame. 3222 3223Some targets have more complex requirements than storing to an 3224address calculable during initial code generation. In that case 3225the @code{eh_return} instruction pattern should be used instead. 3226 3227If you want to support call frame exception handling, you must 3228define either this macro or the @code{eh_return} instruction pattern. 3229@end defmac 3230 3231@defmac RETURN_ADDR_OFFSET 3232If defined, an integer-valued C expression for which rtl will be generated 3233to add it to the exception handler address before it is searched in the 3234exception handling tables, and to subtract it again from the address before 3235using it to return to the exception handler. 3236@end defmac 3237 3238@defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global}) 3239This macro chooses the encoding of pointers embedded in the exception 3240handling sections. If at all possible, this should be defined such 3241that the exception handling section will not require dynamic relocations, 3242and so may be read-only. 3243 3244@var{code} is 0 for data, 1 for code labels, 2 for function pointers. 3245@var{global} is true if the symbol may be affected by dynamic relocations. 3246The macro should return a combination of the @code{DW_EH_PE_*} defines 3247as found in @file{dwarf2.h}. 3248 3249If this macro is not defined, pointers will not be encoded but 3250represented directly. 3251@end defmac 3252 3253@defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done}) 3254This macro allows the target to emit whatever special magic is required 3255to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}. 3256Generic code takes care of pc-relative and indirect encodings; this must 3257be defined if the target uses text-relative or data-relative encodings. 3258 3259This is a C statement that branches to @var{done} if the format was 3260handled. @var{encoding} is the format chosen, @var{size} is the number 3261of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF} 3262to be emitted. 3263@end defmac 3264 3265@defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}) 3266This macro allows the target to add CPU and operating system specific 3267code to the call-frame unwinder for use when there is no unwind data 3268available. The most common reason to implement this macro is to unwind 3269through signal frames. 3270 3271This macro is called from @code{uw_frame_state_for} in 3272@file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and 3273@file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; 3274@var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra} 3275for the address of the code being executed and @code{context->cfa} for 3276the stack pointer value. If the frame can be decoded, the register 3277save addresses should be updated in @var{fs} and the macro should 3278evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded, 3279the macro should evaluate to @code{_URC_END_OF_STACK}. 3280 3281For proper signal handling in Java this macro is accompanied by 3282@code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers. 3283@end defmac 3284 3285@defmac MD_HANDLE_UNWABI (@var{context}, @var{fs}) 3286This macro allows the target to add operating system specific code to the 3287call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive, 3288usually used for signal or interrupt frames. 3289 3290This macro is called from @code{uw_update_context} in libgcc's 3291@file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; 3292@var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi} 3293for the abi and context in the @code{.unwabi} directive. If the 3294@code{.unwabi} directive can be handled, the register save addresses should 3295be updated in @var{fs}. 3296@end defmac 3297 3298@defmac TARGET_USES_WEAK_UNWIND_INFO 3299A C expression that evaluates to true if the target requires unwind 3300info to be given comdat linkage. Define it to be @code{1} if comdat 3301linkage is necessary. The default is @code{0}. 3302@end defmac 3303 3304@node Stack Checking 3305@subsection Specifying How Stack Checking is Done 3306 3307GCC will check that stack references are within the boundaries of the 3308stack, if the option @option{-fstack-check} is specified, in one of 3309three ways: 3310 3311@enumerate 3312@item 3313If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC 3314will assume that you have arranged for full stack checking to be done 3315at appropriate places in the configuration files. GCC will not do 3316other special processing. 3317 3318@item 3319If @code{STACK_CHECK_BUILTIN} is zero and the value of the 3320@code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume 3321that you have arranged for static stack checking (checking of the 3322static stack frame of functions) to be done at appropriate places 3323in the configuration files. GCC will only emit code to do dynamic 3324stack checking (checking on dynamic stack allocations) using the third 3325approach below. 3326 3327@item 3328If neither of the above are true, GCC will generate code to periodically 3329``probe'' the stack pointer using the values of the macros defined below. 3330@end enumerate 3331 3332If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined, 3333GCC will change its allocation strategy for large objects if the option 3334@option{-fstack-check} is specified: they will always be allocated 3335dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes. 3336 3337@defmac STACK_CHECK_BUILTIN 3338A nonzero value if stack checking is done by the configuration files in a 3339machine-dependent manner. You should define this macro if stack checking 3340is required by the ABI of your machine or if you would like to do stack 3341checking in some more efficient way than the generic approach. The default 3342value of this macro is zero. 3343@end defmac 3344 3345@defmac STACK_CHECK_STATIC_BUILTIN 3346A nonzero value if static stack checking is done by the configuration files 3347in a machine-dependent manner. You should define this macro if you would 3348like to do static stack checking in some more efficient way than the generic 3349approach. The default value of this macro is zero. 3350@end defmac 3351 3352@defmac STACK_CHECK_PROBE_INTERVAL_EXP 3353An integer specifying the interval at which GCC must generate stack probe 3354instructions, defined as 2 raised to this integer. You will normally 3355define this macro so that the interval be no larger than the size of 3356the ``guard pages'' at the end of a stack area. The default value 3357of 12 (4096-byte interval) is suitable for most systems. 3358@end defmac 3359 3360@defmac STACK_CHECK_MOVING_SP 3361An integer which is nonzero if GCC should move the stack pointer page by page 3362when doing probes. This can be necessary on systems where the stack pointer 3363contains the bottom address of the memory area accessible to the executing 3364thread at any point in time. In this situation an alternate signal stack 3365is required in order to be able to recover from a stack overflow. The 3366default value of this macro is zero. 3367@end defmac 3368 3369@defmac STACK_CHECK_PROTECT 3370The number of bytes of stack needed to recover from a stack overflow, for 3371languages where such a recovery is supported. The default value of 75 words 3372with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and 33738192 bytes with other exception handling mechanisms should be adequate for 3374most machines. 3375@end defmac 3376 3377The following macros are relevant only if neither STACK_CHECK_BUILTIN 3378nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether 3379in the opposite case. 3380 3381@defmac STACK_CHECK_MAX_FRAME_SIZE 3382The maximum size of a stack frame, in bytes. GCC will generate probe 3383instructions in non-leaf functions to ensure at least this many bytes of 3384stack are available. If a stack frame is larger than this size, stack 3385checking will not be reliable and GCC will issue a warning. The 3386default is chosen so that GCC only generates one instruction on most 3387systems. You should normally not change the default value of this macro. 3388@end defmac 3389 3390@defmac STACK_CHECK_FIXED_FRAME_SIZE 3391GCC uses this value to generate the above warning message. It 3392represents the amount of fixed frame used by a function, not including 3393space for any callee-saved registers, temporaries and user variables. 3394You need only specify an upper bound for this amount and will normally 3395use the default of four words. 3396@end defmac 3397 3398@defmac STACK_CHECK_MAX_VAR_SIZE 3399The maximum size, in bytes, of an object that GCC will place in the 3400fixed area of the stack frame when the user specifies 3401@option{-fstack-check}. 3402GCC computed the default from the values of the above macros and you will 3403normally not need to override that default. 3404@end defmac 3405 3406@need 2000 3407@node Frame Registers 3408@subsection Registers That Address the Stack Frame 3409 3410@c prevent bad page break with this line 3411This discusses registers that address the stack frame. 3412 3413@defmac STACK_POINTER_REGNUM 3414The register number of the stack pointer register, which must also be a 3415fixed register according to @code{FIXED_REGISTERS}. On most machines, 3416the hardware determines which register this is. 3417@end defmac 3418 3419@defmac FRAME_POINTER_REGNUM 3420The register number of the frame pointer register, which is used to 3421access automatic variables in the stack frame. On some machines, the 3422hardware determines which register this is. On other machines, you can 3423choose any register you wish for this purpose. 3424@end defmac 3425 3426@defmac HARD_FRAME_POINTER_REGNUM 3427On some machines the offset between the frame pointer and starting 3428offset of the automatic variables is not known until after register 3429allocation has been done (for example, because the saved registers are 3430between these two locations). On those machines, define 3431@code{FRAME_POINTER_REGNUM} the number of a special, fixed register to 3432be used internally until the offset is known, and define 3433@code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number 3434used for the frame pointer. 3435 3436You should define this macro only in the very rare circumstances when it 3437is not possible to calculate the offset between the frame pointer and 3438the automatic variables until after register allocation has been 3439completed. When this macro is defined, you must also indicate in your 3440definition of @code{ELIMINABLE_REGS} how to eliminate 3441@code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM} 3442or @code{STACK_POINTER_REGNUM}. 3443 3444Do not define this macro if it would be the same as 3445@code{FRAME_POINTER_REGNUM}. 3446@end defmac 3447 3448@defmac ARG_POINTER_REGNUM 3449The register number of the arg pointer register, which is used to access 3450the function's argument list. On some machines, this is the same as the 3451frame pointer register. On some machines, the hardware determines which 3452register this is. On other machines, you can choose any register you 3453wish for this purpose. If this is not the same register as the frame 3454pointer register, then you must mark it as a fixed register according to 3455@code{FIXED_REGISTERS}, or arrange to be able to eliminate it 3456(@pxref{Elimination}). 3457@end defmac 3458 3459@defmac HARD_FRAME_POINTER_IS_FRAME_POINTER 3460Define this to a preprocessor constant that is nonzero if 3461@code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be 3462the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM 3463== FRAME_POINTER_REGNUM)}; you only need to define this macro if that 3464definition is not suitable for use in preprocessor conditionals. 3465@end defmac 3466 3467@defmac HARD_FRAME_POINTER_IS_ARG_POINTER 3468Define this to a preprocessor constant that is nonzero if 3469@code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the 3470same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM == 3471ARG_POINTER_REGNUM)}; you only need to define this macro if that 3472definition is not suitable for use in preprocessor conditionals. 3473@end defmac 3474 3475@defmac RETURN_ADDRESS_POINTER_REGNUM 3476The register number of the return address pointer register, which is used to 3477access the current function's return address from the stack. On some 3478machines, the return address is not at a fixed offset from the frame 3479pointer or stack pointer or argument pointer. This register can be defined 3480to point to the return address on the stack, and then be converted by 3481@code{ELIMINABLE_REGS} into either the frame pointer or stack pointer. 3482 3483Do not define this macro unless there is no other way to get the return 3484address from the stack. 3485@end defmac 3486 3487@defmac STATIC_CHAIN_REGNUM 3488@defmacx STATIC_CHAIN_INCOMING_REGNUM 3489Register numbers used for passing a function's static chain pointer. If 3490register windows are used, the register number as seen by the called 3491function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register 3492number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If 3493these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need 3494not be defined. 3495 3496The static chain register need not be a fixed register. 3497 3498If the static chain is passed in memory, these macros should not be 3499defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used. 3500@end defmac 3501 3502@deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl_or_type}, bool @var{incoming_p}) 3503This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for 3504targets that may use different static chain locations for different 3505nested functions. This may be required if the target has function 3506attributes that affect the calling conventions of the function and 3507those calling conventions use different static chain locations. 3508 3509The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al. 3510 3511If the static chain is passed in memory, this hook should be used to 3512provide rtx giving @code{mem} expressions that denote where they are stored. 3513Often the @code{mem} expression as seen by the caller will be at an offset 3514from the stack pointer and the @code{mem} expression as seen by the callee 3515will be at an offset from the frame pointer. 3516@findex stack_pointer_rtx 3517@findex frame_pointer_rtx 3518@findex arg_pointer_rtx 3519The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and 3520@code{arg_pointer_rtx} will have been initialized and should be used 3521to refer to those items. 3522@end deftypefn 3523 3524@defmac DWARF_FRAME_REGISTERS 3525This macro specifies the maximum number of hard registers that can be 3526saved in a call frame. This is used to size data structures used in 3527DWARF2 exception handling. 3528 3529Prior to GCC 3.0, this macro was needed in order to establish a stable 3530exception handling ABI in the face of adding new hard registers for ISA 3531extensions. In GCC 3.0 and later, the EH ABI is insulated from changes 3532in the number of hard registers. Nevertheless, this macro can still be 3533used to reduce the runtime memory requirements of the exception handling 3534routines, which can be substantial if the ISA contains a lot of 3535registers that are not call-saved. 3536 3537If this macro is not defined, it defaults to 3538@code{FIRST_PSEUDO_REGISTER}. 3539@end defmac 3540 3541@defmac PRE_GCC3_DWARF_FRAME_REGISTERS 3542 3543This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided 3544for backward compatibility in pre GCC 3.0 compiled code. 3545 3546If this macro is not defined, it defaults to 3547@code{DWARF_FRAME_REGISTERS}. 3548@end defmac 3549 3550@defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno}) 3551 3552Define this macro if the target's representation for dwarf registers 3553is different than the internal representation for unwind column. 3554Given a dwarf register, this macro should return the internal unwind 3555column number to use instead. 3556 3557See the PowerPC's SPE target for an example. 3558@end defmac 3559 3560@defmac DWARF_FRAME_REGNUM (@var{regno}) 3561 3562Define this macro if the target's representation for dwarf registers 3563used in .eh_frame or .debug_frame is different from that used in other 3564debug info sections. Given a GCC hard register number, this macro 3565should return the .eh_frame register number. The default is 3566@code{DBX_REGISTER_NUMBER (@var{regno})}. 3567 3568@end defmac 3569 3570@defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh}) 3571 3572Define this macro to map register numbers held in the call frame info 3573that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that 3574should be output in .debug_frame (@code{@var{for_eh}} is zero) and 3575.eh_frame (@code{@var{for_eh}} is nonzero). The default is to 3576return @code{@var{regno}}. 3577 3578@end defmac 3579 3580@defmac REG_VALUE_IN_UNWIND_CONTEXT 3581 3582Define this macro if the target stores register values as 3583@code{_Unwind_Word} type in unwind context. It should be defined if 3584target register size is larger than the size of @code{void *}. The 3585default is to store register values as @code{void *} type. 3586 3587@end defmac 3588 3589@defmac ASSUME_EXTENDED_UNWIND_CONTEXT 3590 3591Define this macro to be 1 if the target always uses extended unwind 3592context with version, args_size and by_value fields. If it is undefined, 3593it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is 3594defined and 0 otherwise. 3595 3596@end defmac 3597 3598@node Elimination 3599@subsection Eliminating Frame Pointer and Arg Pointer 3600 3601@c prevent bad page break with this line 3602This is about eliminating the frame pointer and arg pointer. 3603 3604@deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void) 3605This target hook should return @code{true} if a function must have and use 3606a frame pointer. This target hook is called in the reload pass. If its return 3607value is @code{true} the function will have a frame pointer. 3608 3609This target hook can in principle examine the current function and decide 3610according to the facts, but on most machines the constant @code{false} or the 3611constant @code{true} suffices. Use @code{false} when the machine allows code 3612to be generated with no frame pointer, and doing so saves some time or space. 3613Use @code{true} when there is no possible advantage to avoiding a frame 3614pointer. 3615 3616In certain cases, the compiler does not know how to produce valid code 3617without a frame pointer. The compiler recognizes those cases and 3618automatically gives the function a frame pointer regardless of what 3619@code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about 3620them. 3621 3622In a function that does not require a frame pointer, the frame pointer 3623register can be allocated for ordinary usage, unless you mark it as a 3624fixed register. See @code{FIXED_REGISTERS} for more information. 3625 3626Default return value is @code{false}. 3627@end deftypefn 3628 3629@findex get_frame_size 3630@defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var}) 3631A C statement to store in the variable @var{depth-var} the difference 3632between the frame pointer and the stack pointer values immediately after 3633the function prologue. The value would be computed from information 3634such as the result of @code{get_frame_size ()} and the tables of 3635registers @code{regs_ever_live} and @code{call_used_regs}. 3636 3637If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and 3638need not be defined. Otherwise, it must be defined even if 3639@code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that 3640case, you may set @var{depth-var} to anything. 3641@end defmac 3642 3643@defmac ELIMINABLE_REGS 3644If defined, this macro specifies a table of register pairs used to 3645eliminate unneeded registers that point into the stack frame. If it is not 3646defined, the only elimination attempted by the compiler is to replace 3647references to the frame pointer with references to the stack pointer. 3648 3649The definition of this macro is a list of structure initializations, each 3650of which specifies an original and replacement register. 3651 3652On some machines, the position of the argument pointer is not known until 3653the compilation is completed. In such a case, a separate hard register 3654must be used for the argument pointer. This register can be eliminated by 3655replacing it with either the frame pointer or the argument pointer, 3656depending on whether or not the frame pointer has been eliminated. 3657 3658In this case, you might specify: 3659@smallexample 3660#define ELIMINABLE_REGS \ 3661@{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \ 3662 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \ 3663 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@} 3664@end smallexample 3665 3666Note that the elimination of the argument pointer with the stack pointer is 3667specified first since that is the preferred elimination. 3668@end defmac 3669 3670@deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg}) 3671This target hook should returns @code{true} if the compiler is allowed to 3672try to replace register number @var{from_reg} with register number 3673@var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS} 3674is defined, and will usually be @code{true}, since most of the cases 3675preventing register elimination are things that the compiler already 3676knows about. 3677 3678Default return value is @code{true}. 3679@end deftypefn 3680 3681@defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var}) 3682This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It 3683specifies the initial difference between the specified pair of 3684registers. This macro must be defined if @code{ELIMINABLE_REGS} is 3685defined. 3686@end defmac 3687 3688@node Stack Arguments 3689@subsection Passing Function Arguments on the Stack 3690@cindex arguments on stack 3691@cindex stack arguments 3692 3693The macros in this section control how arguments are passed 3694on the stack. See the following section for other macros that 3695control passing certain arguments in registers. 3696 3697@deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype}) 3698This target hook returns @code{true} if an argument declared in a 3699prototype as an integral type smaller than @code{int} should actually be 3700passed as an @code{int}. In addition to avoiding errors in certain 3701cases of mismatch, it also makes for better code on certain machines. 3702The default is to not promote prototypes. 3703@end deftypefn 3704 3705@defmac PUSH_ARGS 3706A C expression. If nonzero, push insns will be used to pass 3707outgoing arguments. 3708If the target machine does not have a push instruction, set it to zero. 3709That directs GCC to use an alternate strategy: to 3710allocate the entire argument block and then store the arguments into 3711it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too. 3712@end defmac 3713 3714@defmac PUSH_ARGS_REVERSED 3715A C expression. If nonzero, function arguments will be evaluated from 3716last to first, rather than from first to last. If this macro is not 3717defined, it defaults to @code{PUSH_ARGS} on targets where the stack 3718and args grow in opposite directions, and 0 otherwise. 3719@end defmac 3720 3721@defmac PUSH_ROUNDING (@var{npushed}) 3722A C expression that is the number of bytes actually pushed onto the 3723stack when an instruction attempts to push @var{npushed} bytes. 3724 3725On some machines, the definition 3726 3727@smallexample 3728#define PUSH_ROUNDING(BYTES) (BYTES) 3729@end smallexample 3730 3731@noindent 3732will suffice. But on other machines, instructions that appear 3733to push one byte actually push two bytes in an attempt to maintain 3734alignment. Then the definition should be 3735 3736@smallexample 3737#define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) 3738@end smallexample 3739 3740If the value of this macro has a type, it should be an unsigned type. 3741@end defmac 3742 3743@findex outgoing_args_size 3744@findex crtl->outgoing_args_size 3745@defmac ACCUMULATE_OUTGOING_ARGS 3746A C expression. If nonzero, the maximum amount of space required for outgoing arguments 3747will be computed and placed into 3748@code{crtl->outgoing_args_size}. No space will be pushed 3749onto the stack for each call; instead, the function prologue should 3750increase the stack frame size by this amount. 3751 3752Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS} 3753is not proper. 3754@end defmac 3755 3756@defmac REG_PARM_STACK_SPACE (@var{fndecl}) 3757Define this macro if functions should assume that stack space has been 3758allocated for arguments even when their values are passed in 3759registers. 3760 3761The value of this macro is the size, in bytes, of the area reserved for 3762arguments passed in registers for the function represented by @var{fndecl}, 3763which can be zero if GCC is calling a library function. 3764The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself 3765of the function. 3766 3767This space can be allocated by the caller, or be a part of the 3768machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says 3769which. 3770@end defmac 3771@c above is overfull. not sure what to do. --mew 5feb93 did 3772@c something, not sure if it looks good. --mew 10feb93 3773 3774@defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl}) 3775Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments. 3776Define this macro if space guaranteed when compiling a function body 3777is different to space required when making a call, a situation that 3778can arise with K&R style function definitions. 3779@end defmac 3780 3781@defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype}) 3782Define this to a nonzero value if it is the responsibility of the 3783caller to allocate the area reserved for arguments passed in registers 3784when calling a function of @var{fntype}. @var{fntype} may be NULL 3785if the function called is a library function. 3786 3787If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls 3788whether the space for these arguments counts in the value of 3789@code{crtl->outgoing_args_size}. 3790@end defmac 3791 3792@defmac STACK_PARMS_IN_REG_PARM_AREA 3793Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the 3794stack parameters don't skip the area specified by it. 3795@c i changed this, makes more sens and it should have taken care of the 3796@c overfull.. not as specific, tho. --mew 5feb93 3797 3798Normally, when a parameter is not passed in registers, it is placed on the 3799stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro 3800suppresses this behavior and causes the parameter to be passed on the 3801stack in its natural location. 3802@end defmac 3803 3804@deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size}) 3805This target hook returns the number of bytes of its own arguments that 3806a function pops on returning, or 0 if the function pops no arguments 3807and the caller must therefore pop them all after the function returns. 3808 3809@var{fundecl} is a C variable whose value is a tree node that describes 3810the function in question. Normally it is a node of type 3811@code{FUNCTION_DECL} that describes the declaration of the function. 3812From this you can obtain the @code{DECL_ATTRIBUTES} of the function. 3813 3814@var{funtype} is a C variable whose value is a tree node that 3815describes the function in question. Normally it is a node of type 3816@code{FUNCTION_TYPE} that describes the data type of the function. 3817From this it is possible to obtain the data types of the value and 3818arguments (if known). 3819 3820When a call to a library function is being considered, @var{fundecl} 3821will contain an identifier node for the library function. Thus, if 3822you need to distinguish among various library functions, you can do so 3823by their names. Note that ``library function'' in this context means 3824a function used to perform arithmetic, whose name is known specially 3825in the compiler and was not mentioned in the C code being compiled. 3826 3827@var{size} is the number of bytes of arguments passed on the 3828stack. If a variable number of bytes is passed, it is zero, and 3829argument popping will always be the responsibility of the calling function. 3830 3831On the VAX, all functions always pop their arguments, so the definition 3832of this macro is @var{size}. On the 68000, using the standard 3833calling convention, no functions pop their arguments, so the value of 3834the macro is always 0 in this case. But an alternative calling 3835convention is available in which functions that take a fixed number of 3836arguments pop them but other functions (such as @code{printf}) pop 3837nothing (the caller pops all). When this convention is in use, 3838@var{funtype} is examined to determine whether a function takes a fixed 3839number of arguments. 3840@end deftypefn 3841 3842@defmac CALL_POPS_ARGS (@var{cum}) 3843A C expression that should indicate the number of bytes a call sequence 3844pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS} 3845when compiling a function call. 3846 3847@var{cum} is the variable in which all arguments to the called function 3848have been accumulated. 3849 3850On certain architectures, such as the SH5, a call trampoline is used 3851that pops certain registers off the stack, depending on the arguments 3852that have been passed to the function. Since this is a property of the 3853call site, not of the called function, @code{RETURN_POPS_ARGS} is not 3854appropriate. 3855@end defmac 3856 3857@node Register Arguments 3858@subsection Passing Arguments in Registers 3859@cindex arguments in registers 3860@cindex registers arguments 3861 3862This section describes the macros which let you control how various 3863types of arguments are passed in registers or how they are arranged in 3864the stack. 3865 3866@deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named}) 3867Return an RTX indicating whether a function argument is passed in a 3868register and if so, which register. 3869 3870The arguments are @var{ca}, which summarizes all the previous 3871arguments; @var{mode}, the machine mode of the argument; @var{type}, 3872the data type of the argument as a tree node or 0 if that is not known 3873(which happens for C support library functions); and @var{named}, 3874which is @code{true} for an ordinary argument and @code{false} for 3875nameless arguments that correspond to @samp{@dots{}} in the called 3876function's prototype. @var{type} can be an incomplete type if a 3877syntax error has previously occurred. 3878 3879The return value is usually either a @code{reg} RTX for the hard 3880register in which to pass the argument, or zero to pass the argument 3881on the stack. 3882 3883The return value can be a @code{const_int} which means argument is 3884passed in a target specific slot with specified number. Target hooks 3885should be used to store or load argument in such case. See 3886@code{TARGET_STORE_BOUNDS_FOR_ARG} and @code{TARGET_LOAD_BOUNDS_FOR_ARG} 3887for more information. 3888 3889The value of the expression can also be a @code{parallel} RTX@. This is 3890used when an argument is passed in multiple locations. The mode of the 3891@code{parallel} should be the mode of the entire argument. The 3892@code{parallel} holds any number of @code{expr_list} pairs; each one 3893describes where part of the argument is passed. In each 3894@code{expr_list} the first operand must be a @code{reg} RTX for the hard 3895register in which to pass this part of the argument, and the mode of the 3896register RTX indicates how large this part of the argument is. The 3897second operand of the @code{expr_list} is a @code{const_int} which gives 3898the offset in bytes into the entire argument of where this part starts. 3899As a special exception the first @code{expr_list} in the @code{parallel} 3900RTX may have a first operand of zero. This indicates that the entire 3901argument is also stored on the stack. 3902 3903The last time this hook is called, it is called with @code{MODE == 3904VOIDmode}, and its result is passed to the @code{call} or @code{call_value} 3905pattern as operands 2 and 3 respectively. 3906 3907@cindex @file{stdarg.h} and register arguments 3908The usual way to make the ISO library @file{stdarg.h} work on a 3909machine where some arguments are usually passed in registers, is to 3910cause nameless arguments to be passed on the stack instead. This is 3911done by making @code{TARGET_FUNCTION_ARG} return 0 whenever 3912@var{named} is @code{false}. 3913 3914@cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG} 3915@cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG} 3916You may use the hook @code{targetm.calls.must_pass_in_stack} 3917in the definition of this macro to determine if this argument is of a 3918type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE} 3919is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an 3920argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is 3921defined, the argument will be computed in the stack and then loaded into 3922a register. 3923@end deftypefn 3924 3925@deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (machine_mode @var{mode}, const_tree @var{type}) 3926This target hook should return @code{true} if we should not pass @var{type} 3927solely in registers. The file @file{expr.h} defines a 3928definition that is usually appropriate, refer to @file{expr.h} for additional 3929documentation. 3930@end deftypefn 3931 3932@deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named}) 3933Define this hook if the target machine has ``register windows'', so 3934that the register in which a function sees an arguments is not 3935necessarily the same as the one in which the caller passed the 3936argument. 3937 3938For such machines, @code{TARGET_FUNCTION_ARG} computes the register in 3939which the caller passes the value, and 3940@code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar 3941fashion to tell the function being called where the arguments will 3942arrive. 3943 3944If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined, 3945@code{TARGET_FUNCTION_ARG} serves both purposes. 3946@end deftypefn 3947 3948@deftypefn {Target Hook} bool TARGET_USE_PSEUDO_PIC_REG (void) 3949This hook should return 1 in case pseudo register should be created 3950for pic_offset_table_rtx during function expand. 3951@end deftypefn 3952 3953@deftypefn {Target Hook} void TARGET_INIT_PIC_REG (void) 3954Perform a target dependent initialization of pic_offset_table_rtx. 3955This hook is called at the start of register allocation. 3956@end deftypefn 3957 3958@deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, machine_mode @var{mode}, tree @var{type}, bool @var{named}) 3959This target hook returns the number of bytes at the beginning of an 3960argument that must be put in registers. The value must be zero for 3961arguments that are passed entirely in registers or that are entirely 3962pushed on the stack. 3963 3964On some machines, certain arguments must be passed partially in 3965registers and partially in memory. On these machines, typically the 3966first few words of arguments are passed in registers, and the rest 3967on the stack. If a multi-word argument (a @code{double} or a 3968structure) crosses that boundary, its first few words must be passed 3969in registers and the rest must be pushed. This macro tells the 3970compiler when this occurs, and how many bytes should go in registers. 3971 3972@code{TARGET_FUNCTION_ARG} for these arguments should return the first 3973register to be used by the caller for this argument; likewise 3974@code{TARGET_FUNCTION_INCOMING_ARG}, for the called function. 3975@end deftypefn 3976 3977@deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named}) 3978This target hook should return @code{true} if an argument at the 3979position indicated by @var{cum} should be passed by reference. This 3980predicate is queried after target independent reasons for being 3981passed by reference, such as @code{TREE_ADDRESSABLE (type)}. 3982 3983If the hook returns true, a copy of that argument is made in memory and a 3984pointer to the argument is passed instead of the argument itself. 3985The pointer is passed in whatever way is appropriate for passing a pointer 3986to that type. 3987@end deftypefn 3988 3989@deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named}) 3990The function argument described by the parameters to this hook is 3991known to be passed by reference. The hook should return true if the 3992function argument should be copied by the callee instead of copied 3993by the caller. 3994 3995For any argument for which the hook returns true, if it can be 3996determined that the argument is not modified, then a copy need 3997not be generated. 3998 3999The default version of this hook always returns false. 4000@end deftypefn 4001 4002@defmac CUMULATIVE_ARGS 4003A C type for declaring a variable that is used as the first argument 4004of @code{TARGET_FUNCTION_ARG} and other related values. For some 4005target machines, the type @code{int} suffices and can hold the number 4006of bytes of argument so far. 4007 4008There is no need to record in @code{CUMULATIVE_ARGS} anything about the 4009arguments that have been passed on the stack. The compiler has other 4010variables to keep track of that. For target machines on which all 4011arguments are passed on the stack, there is no need to store anything in 4012@code{CUMULATIVE_ARGS}; however, the data structure must exist and 4013should not be empty, so use @code{int}. 4014@end defmac 4015 4016@defmac OVERRIDE_ABI_FORMAT (@var{fndecl}) 4017If defined, this macro is called before generating any code for a 4018function, but after the @var{cfun} descriptor for the function has been 4019created. The back end may use this macro to update @var{cfun} to 4020reflect an ABI other than that which would normally be used by default. 4021If the compiler is generating code for a compiler-generated function, 4022@var{fndecl} may be @code{NULL}. 4023@end defmac 4024 4025@defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args}) 4026A C statement (sans semicolon) for initializing the variable 4027@var{cum} for the state at the beginning of the argument list. The 4028variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype} 4029is the tree node for the data type of the function which will receive 4030the args, or 0 if the args are to a compiler support library function. 4031For direct calls that are not libcalls, @var{fndecl} contain the 4032declaration node of the function. @var{fndecl} is also set when 4033@code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function 4034being compiled. @var{n_named_args} is set to the number of named 4035arguments, including a structure return address if it is passed as a 4036parameter, when making a call. When processing incoming arguments, 4037@var{n_named_args} is set to @minus{}1. 4038 4039When processing a call to a compiler support library function, 4040@var{libname} identifies which one. It is a @code{symbol_ref} rtx which 4041contains the name of the function, as a string. @var{libname} is 0 when 4042an ordinary C function call is being processed. Thus, each time this 4043macro is called, either @var{libname} or @var{fntype} is nonzero, but 4044never both of them at once. 4045@end defmac 4046 4047@defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname}) 4048Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls, 4049it gets a @code{MODE} argument instead of @var{fntype}, that would be 4050@code{NULL}. @var{indirect} would always be zero, too. If this macro 4051is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, 40520)} is used instead. 4053@end defmac 4054 4055@defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname}) 4056Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of 4057finding the arguments for the function being compiled. If this macro is 4058undefined, @code{INIT_CUMULATIVE_ARGS} is used instead. 4059 4060The value passed for @var{libname} is always 0, since library routines 4061with special calling conventions are never compiled with GCC@. The 4062argument @var{libname} exists for symmetry with 4063@code{INIT_CUMULATIVE_ARGS}. 4064@c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe. 4065@c --mew 5feb93 i switched the order of the sentences. --mew 10feb93 4066@end defmac 4067 4068@deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named}) 4069This hook updates the summarizer variable pointed to by @var{ca} to 4070advance past an argument in the argument list. The values @var{mode}, 4071@var{type} and @var{named} describe that argument. Once this is done, 4072the variable @var{cum} is suitable for analyzing the @emph{following} 4073argument with @code{TARGET_FUNCTION_ARG}, etc. 4074 4075This hook need not do anything if the argument in question was passed 4076on the stack. The compiler knows how to track the amount of stack space 4077used for arguments without any special help. 4078@end deftypefn 4079 4080@defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type}) 4081If defined, a C expression that is the number of bytes to add to the 4082offset of the argument passed in memory. This is needed for the SPU, 4083which passes @code{char} and @code{short} arguments in the preferred 4084slot that is in the middle of the quad word instead of starting at the 4085top. 4086@end defmac 4087 4088@defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type}) 4089If defined, a C expression which determines whether, and in which direction, 4090to pad out an argument with extra space. The value should be of type 4091@code{enum direction}: either @code{upward} to pad above the argument, 4092@code{downward} to pad below, or @code{none} to inhibit padding. 4093 4094The @emph{amount} of padding is not controlled by this macro, but by the 4095target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is 4096always just enough to reach the next multiple of that boundary. 4097 4098This macro has a default definition which is right for most systems. 4099For little-endian machines, the default is to pad upward. For 4100big-endian machines, the default is to pad downward for an argument of 4101constant size shorter than an @code{int}, and upward otherwise. 4102@end defmac 4103 4104@defmac PAD_VARARGS_DOWN 4105If defined, a C expression which determines whether the default 4106implementation of va_arg will attempt to pad down before reading the 4107next argument, if that argument is smaller than its aligned space as 4108controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such 4109arguments are padded down if @code{BYTES_BIG_ENDIAN} is true. 4110@end defmac 4111 4112@defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first}) 4113Specify padding for the last element of a block move between registers and 4114memory. @var{first} is nonzero if this is the only element. Defining this 4115macro allows better control of register function parameters on big-endian 4116machines, without using @code{PARALLEL} rtl. In particular, 4117@code{MUST_PASS_IN_STACK} need not test padding and mode of types in 4118registers, as there is no longer a "wrong" part of a register; For example, 4119a three byte aggregate may be passed in the high part of a register if so 4120required. 4121@end defmac 4122 4123@deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (machine_mode @var{mode}, const_tree @var{type}) 4124This hook returns the alignment boundary, in bits, of an argument 4125with the specified mode and type. The default hook returns 4126@code{PARM_BOUNDARY} for all arguments. 4127@end deftypefn 4128 4129@deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (machine_mode @var{mode}, const_tree @var{type}) 4130Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY}, 4131which is the default value for this hook. You can define this hook to 4132return a different value if an argument size must be rounded to a larger 4133value. 4134@end deftypefn 4135 4136@defmac FUNCTION_ARG_REGNO_P (@var{regno}) 4137A C expression that is nonzero if @var{regno} is the number of a hard 4138register in which function arguments are sometimes passed. This does 4139@emph{not} include implicit arguments such as the static chain and 4140the structure-value address. On many machines, no registers can be 4141used for this purpose since all function arguments are pushed on the 4142stack. 4143@end defmac 4144 4145@deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type}) 4146This hook should return true if parameter of type @var{type} are passed 4147as two scalar parameters. By default, GCC will attempt to pack complex 4148arguments into the target's word size. Some ABIs require complex arguments 4149to be split and treated as their individual components. For example, on 4150AIX64, complex floats should be passed in a pair of floating point 4151registers, even though a complex float would fit in one 64-bit floating 4152point register. 4153 4154The default value of this hook is @code{NULL}, which is treated as always 4155false. 4156@end deftypefn 4157 4158@deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void) 4159This hook returns a type node for @code{va_list} for the target. 4160The default version of the hook returns @code{void*}. 4161@end deftypefn 4162 4163@deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree}) 4164This target hook is used in function @code{c_common_nodes_and_builtins} 4165to iterate through the target specific builtin types for va_list. The 4166variable @var{idx} is used as iterator. @var{pname} has to be a pointer 4167to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed 4168variable. 4169The arguments @var{pname} and @var{ptree} are used to store the result of 4170this macro and are set to the name of the va_list builtin type and its 4171internal type. 4172If the return value of this macro is zero, then there is no more element. 4173Otherwise the @var{IDX} should be increased for the next call of this 4174macro to iterate through all types. 4175@end deftypefn 4176 4177@deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl}) 4178This hook returns the va_list type of the calling convention specified by 4179@var{fndecl}. 4180The default version of this hook returns @code{va_list_type_node}. 4181@end deftypefn 4182 4183@deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type}) 4184This hook returns the va_list type of the calling convention specified by the 4185type of @var{type}. If @var{type} is not a valid va_list type, it returns 4186@code{NULL_TREE}. 4187@end deftypefn 4188 4189@deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, gimple_seq *@var{pre_p}, gimple_seq *@var{post_p}) 4190This hook performs target-specific gimplification of 4191@code{VA_ARG_EXPR}. The first two parameters correspond to the 4192arguments to @code{va_arg}; the latter two are as in 4193@code{gimplify.c:gimplify_expr}. 4194@end deftypefn 4195 4196@deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (machine_mode @var{mode}) 4197Define this to return nonzero if the port can handle pointers 4198with machine mode @var{mode}. The default version of this 4199hook returns true for both @code{ptr_mode} and @code{Pmode}. 4200@end deftypefn 4201 4202@deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref *@var{ref}) 4203Define this to return nonzero if the memory reference @var{ref} may alias with the system C library errno location. The default version of this hook assumes the system C library errno location is either a declaration of type int or accessed by dereferencing a pointer to int. 4204@end deftypefn 4205 4206@deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (machine_mode @var{mode}) 4207Define this to return nonzero if the port is prepared to handle 4208insns involving scalar mode @var{mode}. For a scalar mode to be 4209considered supported, all the basic arithmetic and comparisons 4210must work. 4211 4212The default version of this hook returns true for any mode 4213required to handle the basic C types (as defined by the port). 4214Included here are the double-word arithmetic supported by the 4215code in @file{optabs.c}. 4216@end deftypefn 4217 4218@deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (machine_mode @var{mode}) 4219Define this to return nonzero if the port is prepared to handle 4220insns involving vector mode @var{mode}. At the very least, it 4221must have move patterns for this mode. 4222@end deftypefn 4223 4224@deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems}) 4225Return true if GCC should try to use a scalar mode to store an array 4226of @var{nelems} elements, given that each element has mode @var{mode}. 4227Returning true here overrides the usual @code{MAX_FIXED_MODE} limit 4228and allows GCC to use any defined integer mode. 4229 4230One use of this hook is to support vector load and store operations 4231that operate on several homogeneous vectors. For example, ARM NEON 4232has operations like: 4233 4234@smallexample 4235int8x8x3_t vld3_s8 (const int8_t *) 4236@end smallexample 4237 4238where the return type is defined as: 4239 4240@smallexample 4241typedef struct int8x8x3_t 4242@{ 4243 int8x8_t val[3]; 4244@} int8x8x3_t; 4245@end smallexample 4246 4247If this hook allows @code{val} to have a scalar mode, then 4248@code{int8x8x3_t} can have the same mode. GCC can then store 4249@code{int8x8x3_t}s in registers rather than forcing them onto the stack. 4250@end deftypefn 4251 4252@deftypefn {Target Hook} bool TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P (machine_mode @var{mode}) 4253Define this to return nonzero if libgcc provides support for the 4254floating-point mode @var{mode}, which is known to pass 4255@code{TARGET_SCALAR_MODE_SUPPORTED_P}. The default version of this 4256hook returns true for all of @code{SFmode}, @code{DFmode}, 4257@code{XFmode} and @code{TFmode}, if such modes exist. 4258@end deftypefn 4259 4260@deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (machine_mode @var{mode}) 4261Define this to return nonzero for machine modes for which the port has 4262small register classes. If this target hook returns nonzero for a given 4263@var{mode}, the compiler will try to minimize the lifetime of registers 4264in @var{mode}. The hook may be called with @code{VOIDmode} as argument. 4265In this case, the hook is expected to return nonzero if it returns nonzero 4266for any mode. 4267 4268On some machines, it is risky to let hard registers live across arbitrary 4269insns. Typically, these machines have instructions that require values 4270to be in specific registers (like an accumulator), and reload will fail 4271if the required hard register is used for another purpose across such an 4272insn. 4273 4274Passes before reload do not know which hard registers will be used 4275in an instruction, but the machine modes of the registers set or used in 4276the instruction are already known. And for some machines, register 4277classes are small for, say, integer registers but not for floating point 4278registers. For example, the AMD x86-64 architecture requires specific 4279registers for the legacy x86 integer instructions, but there are many 4280SSE registers for floating point operations. On such targets, a good 4281strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P} 4282machine modes but zero for the SSE register classes. 4283 4284The default version of this hook returns false for any mode. It is always 4285safe to redefine this hook to return with a nonzero value. But if you 4286unnecessarily define it, you will reduce the amount of optimizations 4287that can be performed in some cases. If you do not define this hook 4288to return a nonzero value when it is required, the compiler will run out 4289of spill registers and print a fatal error message. 4290@end deftypefn 4291 4292@node Scalar Return 4293@subsection How Scalar Function Values Are Returned 4294@cindex return values in registers 4295@cindex values, returned by functions 4296@cindex scalars, returned as values 4297 4298This section discusses the macros that control returning scalars as 4299values---values that can fit in registers. 4300 4301@deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing}) 4302 4303Define this to return an RTX representing the place where a function 4304returns or receives a value of data type @var{ret_type}, a tree node 4305representing a data type. @var{fn_decl_or_type} is a tree node 4306representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a 4307function being called. If @var{outgoing} is false, the hook should 4308compute the register in which the caller will see the return value. 4309Otherwise, the hook should return an RTX representing the place where 4310a function returns a value. 4311 4312On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant. 4313(Actually, on most machines, scalar values are returned in the same 4314place regardless of mode.) The value of the expression is usually a 4315@code{reg} RTX for the hard register where the return value is stored. 4316The value can also be a @code{parallel} RTX, if the return value is in 4317multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the 4318@code{parallel} form. Note that the callee will populate every 4319location specified in the @code{parallel}, but if the first element of 4320the @code{parallel} contains the whole return value, callers will use 4321that element as the canonical location and ignore the others. The m68k 4322port uses this type of @code{parallel} to return pointers in both 4323@samp{%a0} (the canonical location) and @samp{%d0}. 4324 4325If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply 4326the same promotion rules specified in @code{PROMOTE_MODE} if 4327@var{valtype} is a scalar type. 4328 4329If the precise function being called is known, @var{func} is a tree 4330node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null 4331pointer. This makes it possible to use a different value-returning 4332convention for specific functions when all their calls are 4333known. 4334 4335Some target machines have ``register windows'' so that the register in 4336which a function returns its value is not the same as the one in which 4337the caller sees the value. For such machines, you should return 4338different RTX depending on @var{outgoing}. 4339 4340@code{TARGET_FUNCTION_VALUE} is not used for return values with 4341aggregate data types, because these are returned in another way. See 4342@code{TARGET_STRUCT_VALUE_RTX} and related macros, below. 4343@end deftypefn 4344 4345@defmac FUNCTION_VALUE (@var{valtype}, @var{func}) 4346This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for 4347a new target instead. 4348@end defmac 4349 4350@defmac LIBCALL_VALUE (@var{mode}) 4351A C expression to create an RTX representing the place where a library 4352function returns a value of mode @var{mode}. 4353 4354Note that ``library function'' in this context means a compiler 4355support routine, used to perform arithmetic, whose name is known 4356specially by the compiler and was not mentioned in the C code being 4357compiled. 4358@end defmac 4359 4360@deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (machine_mode @var{mode}, const_rtx @var{fun}) 4361Define this hook if the back-end needs to know the name of the libcall 4362function in order to determine where the result should be returned. 4363 4364The mode of the result is given by @var{mode} and the name of the called 4365library function is given by @var{fun}. The hook should return an RTX 4366representing the place where the library function result will be returned. 4367 4368If this hook is not defined, then LIBCALL_VALUE will be used. 4369@end deftypefn 4370 4371@defmac FUNCTION_VALUE_REGNO_P (@var{regno}) 4372A C expression that is nonzero if @var{regno} is the number of a hard 4373register in which the values of called function may come back. 4374 4375A register whose use for returning values is limited to serving as the 4376second of a pair (for a value of type @code{double}, say) need not be 4377recognized by this macro. So for most machines, this definition 4378suffices: 4379 4380@smallexample 4381#define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) 4382@end smallexample 4383 4384If the machine has register windows, so that the caller and the called 4385function use different registers for the return value, this macro 4386should recognize only the caller's register numbers. 4387 4388This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P} 4389for a new target instead. 4390@end defmac 4391 4392@deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno}) 4393A target hook that return @code{true} if @var{regno} is the number of a hard 4394register in which the values of called function may come back. 4395 4396A register whose use for returning values is limited to serving as the 4397second of a pair (for a value of type @code{double}, say) need not be 4398recognized by this target hook. 4399 4400If the machine has register windows, so that the caller and the called 4401function use different registers for the return value, this target hook 4402should recognize only the caller's register numbers. 4403 4404If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used. 4405@end deftypefn 4406 4407@defmac APPLY_RESULT_SIZE 4408Define this macro if @samp{untyped_call} and @samp{untyped_return} 4409need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for 4410saving and restoring an arbitrary return value. 4411@end defmac 4412 4413@deftypevr {Target Hook} bool TARGET_OMIT_STRUCT_RETURN_REG 4414Normally, when a function returns a structure by memory, the address 4415is passed as an invisible pointer argument, but the compiler also 4416arranges to return the address from the function like it would a normal 4417pointer return value. Define this to true if that behaviour is 4418undesirable on your target. 4419@end deftypevr 4420 4421@deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type}) 4422This hook should return true if values of type @var{type} are returned 4423at the most significant end of a register (in other words, if they are 4424padded at the least significant end). You can assume that @var{type} 4425is returned in a register; the caller is required to check this. 4426 4427Note that the register provided by @code{TARGET_FUNCTION_VALUE} must 4428be able to hold the complete return value. For example, if a 1-, 2- 4429or 3-byte structure is returned at the most significant end of a 44304-byte register, @code{TARGET_FUNCTION_VALUE} should provide an 4431@code{SImode} rtx. 4432@end deftypefn 4433 4434@node Aggregate Return 4435@subsection How Large Values Are Returned 4436@cindex aggregates as return values 4437@cindex large return values 4438@cindex returning aggregate values 4439@cindex structure value address 4440 4441When a function value's mode is @code{BLKmode} (and in some other 4442cases), the value is not returned according to 4443@code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the 4444caller passes the address of a block of memory in which the value 4445should be stored. This address is called the @dfn{structure value 4446address}. 4447 4448This section describes how to control returning structure values in 4449memory. 4450 4451@deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype}) 4452This target hook should return a nonzero value to say to return the 4453function value in memory, just as large structures are always returned. 4454Here @var{type} will be the data type of the value, and @var{fntype} 4455will be the type of the function doing the returning, or @code{NULL} for 4456libcalls. 4457 4458Note that values of mode @code{BLKmode} must be explicitly handled 4459by this function. Also, the option @option{-fpcc-struct-return} 4460takes effect regardless of this macro. On most systems, it is 4461possible to leave the hook undefined; this causes a default 4462definition to be used, whose value is the constant 1 for @code{BLKmode} 4463values, and 0 otherwise. 4464 4465Do not use this hook to indicate that structures and unions should always 4466be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN} 4467to indicate this. 4468@end deftypefn 4469 4470@defmac DEFAULT_PCC_STRUCT_RETURN 4471Define this macro to be 1 if all structure and union return values must be 4472in memory. Since this results in slower code, this should be defined 4473only if needed for compatibility with other compilers or with an ABI@. 4474If you define this macro to be 0, then the conventions used for structure 4475and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY} 4476target hook. 4477 4478If not defined, this defaults to the value 1. 4479@end defmac 4480 4481@deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming}) 4482This target hook should return the location of the structure value 4483address (normally a @code{mem} or @code{reg}), or 0 if the address is 4484passed as an ``invisible'' first argument. Note that @var{fndecl} may 4485be @code{NULL}, for libcalls. You do not need to define this target 4486hook if the address is always passed as an ``invisible'' first 4487argument. 4488 4489On some architectures the place where the structure value address 4490is found by the called function is not the same place that the 4491caller put it. This can be due to register windows, or it could 4492be because the function prologue moves it to a different place. 4493@var{incoming} is @code{1} or @code{2} when the location is needed in 4494the context of the called function, and @code{0} in the context of 4495the caller. 4496 4497If @var{incoming} is nonzero and the address is to be found on the 4498stack, return a @code{mem} which refers to the frame pointer. If 4499@var{incoming} is @code{2}, the result is being used to fetch the 4500structure value address at the beginning of a function. If you need 4501to emit adjusting code, you should do it at this point. 4502@end deftypefn 4503 4504@defmac PCC_STATIC_STRUCT_RETURN 4505Define this macro if the usual system convention on the target machine 4506for returning structures and unions is for the called function to return 4507the address of a static variable containing the value. 4508 4509Do not define this if the usual system convention is for the caller to 4510pass an address to the subroutine. 4511 4512This macro has effect in @option{-fpcc-struct-return} mode, but it does 4513nothing when you use @option{-freg-struct-return} mode. 4514@end defmac 4515 4516@deftypefn {Target Hook} machine_mode TARGET_GET_RAW_RESULT_MODE (int @var{regno}) 4517This target hook returns the mode to be used when accessing raw return registers in @code{__builtin_return}. Define this macro if the value in @var{reg_raw_mode} is not correct. 4518@end deftypefn 4519 4520@deftypefn {Target Hook} machine_mode TARGET_GET_RAW_ARG_MODE (int @var{regno}) 4521This target hook returns the mode to be used when accessing raw argument registers in @code{__builtin_apply_args}. Define this macro if the value in @var{reg_raw_mode} is not correct. 4522@end deftypefn 4523 4524@node Caller Saves 4525@subsection Caller-Saves Register Allocation 4526 4527If you enable it, GCC can save registers around function calls. This 4528makes it possible to use call-clobbered registers to hold variables that 4529must live across calls. 4530 4531@defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs}) 4532A C expression specifying which mode is required for saving @var{nregs} 4533of a pseudo-register in call-clobbered hard register @var{regno}. If 4534@var{regno} is unsuitable for caller save, @code{VOIDmode} should be 4535returned. For most machines this macro need not be defined since GCC 4536will select the smallest suitable mode. 4537@end defmac 4538 4539@node Function Entry 4540@subsection Function Entry and Exit 4541@cindex function entry and exit 4542@cindex prologue 4543@cindex epilogue 4544 4545This section describes the macros that output function entry 4546(@dfn{prologue}) and exit (@dfn{epilogue}) code. 4547 4548@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) 4549If defined, a function that outputs the assembler code for entry to a 4550function. The prologue is responsible for setting up the stack frame, 4551initializing the frame pointer register, saving registers that must be 4552saved, and allocating @var{size} additional bytes of storage for the 4553local variables. @var{size} is an integer. @var{file} is a stdio 4554stream to which the assembler code should be output. 4555 4556The label for the beginning of the function need not be output by this 4557macro. That has already been done when the macro is run. 4558 4559@findex regs_ever_live 4560To determine which registers to save, the macro can refer to the array 4561@code{regs_ever_live}: element @var{r} is nonzero if hard register 4562@var{r} is used anywhere within the function. This implies the function 4563prologue should save register @var{r}, provided it is not one of the 4564call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use 4565@code{regs_ever_live}.) 4566 4567On machines that have ``register windows'', the function entry code does 4568not save on the stack the registers that are in the windows, even if 4569they are supposed to be preserved by function calls; instead it takes 4570appropriate steps to ``push'' the register stack, if any non-call-used 4571registers are used in the function. 4572 4573@findex frame_pointer_needed 4574On machines where functions may or may not have frame-pointers, the 4575function entry code must vary accordingly; it must set up the frame 4576pointer if one is wanted, and not otherwise. To determine whether a 4577frame pointer is in wanted, the macro can refer to the variable 4578@code{frame_pointer_needed}. The variable's value will be 1 at run 4579time in a function that needs a frame pointer. @xref{Elimination}. 4580 4581The function entry code is responsible for allocating any stack space 4582required for the function. This stack space consists of the regions 4583listed below. In most cases, these regions are allocated in the 4584order listed, with the last listed region closest to the top of the 4585stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and 4586the highest address if it is not defined). You can use a different order 4587for a machine if doing so is more convenient or required for 4588compatibility reasons. Except in cases where required by standard 4589or by a debugger, there is no reason why the stack layout used by GCC 4590need agree with that used by other compilers for a machine. 4591@end deftypefn 4592 4593@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file}) 4594If defined, a function that outputs assembler code at the end of a 4595prologue. This should be used when the function prologue is being 4596emitted as RTL, and you have some extra assembler that needs to be 4597emitted. @xref{prologue instruction pattern}. 4598@end deftypefn 4599 4600@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file}) 4601If defined, a function that outputs assembler code at the start of an 4602epilogue. This should be used when the function epilogue is being 4603emitted as RTL, and you have some extra assembler that needs to be 4604emitted. @xref{epilogue instruction pattern}. 4605@end deftypefn 4606 4607@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) 4608If defined, a function that outputs the assembler code for exit from a 4609function. The epilogue is responsible for restoring the saved 4610registers and stack pointer to their values when the function was 4611called, and returning control to the caller. This macro takes the 4612same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the 4613registers to restore are determined from @code{regs_ever_live} and 4614@code{CALL_USED_REGISTERS} in the same way. 4615 4616On some machines, there is a single instruction that does all the work 4617of returning from the function. On these machines, give that 4618instruction the name @samp{return} and do not define the macro 4619@code{TARGET_ASM_FUNCTION_EPILOGUE} at all. 4620 4621Do not define a pattern named @samp{return} if you want the 4622@code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target 4623switches to control whether return instructions or epilogues are used, 4624define a @samp{return} pattern with a validity condition that tests the 4625target switches appropriately. If the @samp{return} pattern's validity 4626condition is false, epilogues will be used. 4627 4628On machines where functions may or may not have frame-pointers, the 4629function exit code must vary accordingly. Sometimes the code for these 4630two cases is completely different. To determine whether a frame pointer 4631is wanted, the macro can refer to the variable 4632@code{frame_pointer_needed}. The variable's value will be 1 when compiling 4633a function that needs a frame pointer. 4634 4635Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and 4636@code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially. 4637The C variable @code{current_function_is_leaf} is nonzero for such a 4638function. @xref{Leaf Functions}. 4639 4640On some machines, some functions pop their arguments on exit while 4641others leave that for the caller to do. For example, the 68020 when 4642given @option{-mrtd} pops arguments in functions that take a fixed 4643number of arguments. 4644 4645@findex pops_args 4646@findex crtl->args.pops_args 4647Your definition of the macro @code{RETURN_POPS_ARGS} decides which 4648functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE} 4649needs to know what was decided. The number of bytes of the current 4650function's arguments that this function should pop is available in 4651@code{crtl->args.pops_args}. @xref{Scalar Return}. 4652@end deftypefn 4653 4654@itemize @bullet 4655@item 4656@findex pretend_args_size 4657@findex crtl->args.pretend_args_size 4658A region of @code{crtl->args.pretend_args_size} bytes of 4659uninitialized space just underneath the first argument arriving on the 4660stack. (This may not be at the very start of the allocated stack region 4661if the calling sequence has pushed anything else since pushing the stack 4662arguments. But usually, on such machines, nothing else has been pushed 4663yet, because the function prologue itself does all the pushing.) This 4664region is used on machines where an argument may be passed partly in 4665registers and partly in memory, and, in some cases to support the 4666features in @code{<stdarg.h>}. 4667 4668@item 4669An area of memory used to save certain registers used by the function. 4670The size of this area, which may also include space for such things as 4671the return address and pointers to previous stack frames, is 4672machine-specific and usually depends on which registers have been used 4673in the function. Machines with register windows often do not require 4674a save area. 4675 4676@item 4677A region of at least @var{size} bytes, possibly rounded up to an allocation 4678boundary, to contain the local variables of the function. On some machines, 4679this region and the save area may occur in the opposite order, with the 4680save area closer to the top of the stack. 4681 4682@item 4683@cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames 4684Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of 4685@code{crtl->outgoing_args_size} bytes to be used for outgoing 4686argument lists of the function. @xref{Stack Arguments}. 4687@end itemize 4688 4689@defmac EXIT_IGNORE_STACK 4690Define this macro as a C expression that is nonzero if the return 4691instruction or the function epilogue ignores the value of the stack 4692pointer; in other words, if it is safe to delete an instruction to 4693adjust the stack pointer before a return from the function. The 4694default is 0. 4695 4696Note that this macro's value is relevant only for functions for which 4697frame pointers are maintained. It is never safe to delete a final 4698stack adjustment in a function that has no frame pointer, and the 4699compiler knows this regardless of @code{EXIT_IGNORE_STACK}. 4700@end defmac 4701 4702@defmac EPILOGUE_USES (@var{regno}) 4703Define this macro as a C expression that is nonzero for registers that are 4704used by the epilogue or the @samp{return} pattern. The stack and frame 4705pointer registers are already assumed to be used as needed. 4706@end defmac 4707 4708@defmac EH_USES (@var{regno}) 4709Define this macro as a C expression that is nonzero for registers that are 4710used by the exception handling mechanism, and so should be considered live 4711on entry to an exception edge. 4712@end defmac 4713 4714@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function}) 4715A function that outputs the assembler code for a thunk 4716function, used to implement C++ virtual function calls with multiple 4717inheritance. The thunk acts as a wrapper around a virtual function, 4718adjusting the implicit object parameter before handing control off to 4719the real function. 4720 4721First, emit code to add the integer @var{delta} to the location that 4722contains the incoming first argument. Assume that this argument 4723contains a pointer, and is the one used to pass the @code{this} pointer 4724in C++. This is the incoming argument @emph{before} the function prologue, 4725e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of 4726all other incoming arguments. 4727 4728Then, if @var{vcall_offset} is nonzero, an additional adjustment should be 4729made after adding @code{delta}. In particular, if @var{p} is the 4730adjusted pointer, the following adjustment should be made: 4731 4732@smallexample 4733p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)] 4734@end smallexample 4735 4736After the additions, emit code to jump to @var{function}, which is a 4737@code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does 4738not touch the return address. Hence returning from @var{FUNCTION} will 4739return to whoever called the current @samp{thunk}. 4740 4741The effect must be as if @var{function} had been called directly with 4742the adjusted first argument. This macro is responsible for emitting all 4743of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE} 4744and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked. 4745 4746The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function} 4747have already been extracted from it.) It might possibly be useful on 4748some targets, but probably not. 4749 4750If you do not define this macro, the target-independent code in the C++ 4751front end will generate a less efficient heavyweight thunk that calls 4752@var{function} instead of jumping to it. The generic approach does 4753not support varargs. 4754@end deftypefn 4755 4756@deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (const_tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, const_tree @var{function}) 4757A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able 4758to output the assembler code for the thunk function specified by the 4759arguments it is passed, and false otherwise. In the latter case, the 4760generic approach will be used by the C++ front end, with the limitations 4761previously exposed. 4762@end deftypefn 4763 4764@node Profiling 4765@subsection Generating Code for Profiling 4766@cindex profiling, code generation 4767 4768These macros will help you generate code for profiling. 4769 4770@defmac FUNCTION_PROFILER (@var{file}, @var{labelno}) 4771A C statement or compound statement to output to @var{file} some 4772assembler code to call the profiling subroutine @code{mcount}. 4773 4774@findex mcount 4775The details of how @code{mcount} expects to be called are determined by 4776your operating system environment, not by GCC@. To figure them out, 4777compile a small program for profiling using the system's installed C 4778compiler and look at the assembler code that results. 4779 4780Older implementations of @code{mcount} expect the address of a counter 4781variable to be loaded into some register. The name of this variable is 4782@samp{LP} followed by the number @var{labelno}, so you would generate 4783the name using @samp{LP%d} in a @code{fprintf}. 4784@end defmac 4785 4786@defmac PROFILE_HOOK 4787A C statement or compound statement to output to @var{file} some assembly 4788code to call the profiling subroutine @code{mcount} even the target does 4789not support profiling. 4790@end defmac 4791 4792@defmac NO_PROFILE_COUNTERS 4793Define this macro to be an expression with a nonzero value if the 4794@code{mcount} subroutine on your system does not need a counter variable 4795allocated for each function. This is true for almost all modern 4796implementations. If you define this macro, you must not use the 4797@var{labelno} argument to @code{FUNCTION_PROFILER}. 4798@end defmac 4799 4800@defmac PROFILE_BEFORE_PROLOGUE 4801Define this macro if the code for function profiling should come before 4802the function prologue. Normally, the profiling code comes after. 4803@end defmac 4804 4805@deftypefn {Target Hook} bool TARGET_KEEP_LEAF_WHEN_PROFILED (void) 4806This target hook returns true if the target wants the leaf flag for the current function to stay true even if it calls mcount. This might make sense for targets using the leaf flag only to determine whether a stack frame needs to be generated or not and for which the call to mcount is generated before the function prologue. 4807@end deftypefn 4808 4809@node Tail Calls 4810@subsection Permitting tail calls 4811@cindex tail calls 4812 4813@deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp}) 4814True if it is OK to do sibling call optimization for the specified 4815call expression @var{exp}. @var{decl} will be the called function, 4816or @code{NULL} if this is an indirect call. 4817 4818It is not uncommon for limitations of calling conventions to prevent 4819tail calls to functions outside the current unit of translation, or 4820during PIC compilation. The hook is used to enforce these restrictions, 4821as the @code{sibcall} md pattern can not fail, or fall over to a 4822``normal'' call. The criteria for successful sibling call optimization 4823may vary greatly between different architectures. 4824@end deftypefn 4825 4826@deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs}) 4827Add any hard registers to @var{regs} that are live on entry to the 4828function. This hook only needs to be defined to provide registers that 4829cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved 4830registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM, 4831TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES, 4832FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM. 4833@end deftypefn 4834 4835@deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{}) 4836This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes. 4837@end deftypefn 4838 4839@deftypefn {Target Hook} bool TARGET_WARN_FUNC_RETURN (tree) 4840True if a function's return statements should be checked for matching the function's return type. This includes checking for falling off the end of a non-void function. Return false if no such check should be made. 4841@end deftypefn 4842 4843@node Stack Smashing Protection 4844@subsection Stack smashing protection 4845@cindex stack smashing protection 4846 4847@deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void) 4848This hook returns a @code{DECL} node for the external variable to use 4849for the stack protection guard. This variable is initialized by the 4850runtime to some random value and is used to initialize the guard value 4851that is placed at the top of the local stack frame. The type of this 4852variable must be @code{ptr_type_node}. 4853 4854The default version of this hook creates a variable called 4855@samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}. 4856@end deftypefn 4857 4858@deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void) 4859This hook returns a @code{CALL_EXPR} that alerts the runtime that the 4860stack protect guard variable has been modified. This expression should 4861involve a call to a @code{noreturn} function. 4862 4863The default version of this hook invokes a function called 4864@samp{__stack_chk_fail}, taking no arguments. This function is 4865normally defined in @file{libgcc2.c}. 4866@end deftypefn 4867 4868@deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts}) 4869Whether this target supports splitting the stack when the options described in @var{opts} have been passed. This is called after options have been parsed, so the target may reject splitting the stack in some configurations. The default version of this hook returns false. If @var{report} is true, this function may issue a warning or error; if @var{report} is false, it must simply return a value 4870@end deftypefn 4871 4872@node Miscellaneous Register Hooks 4873@subsection Miscellaneous register hooks 4874@cindex miscellaneous register hooks 4875 4876@deftypevr {Target Hook} bool TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS 4877Set to true if each call that binds to a local definition explicitly 4878clobbers or sets all non-fixed registers modified by performing the call. 4879That is, by the call pattern itself, or by code that might be inserted by the 4880linker (e.g. stubs, veneers, branch islands), but not including those 4881modifiable by the callee. The affected registers may be mentioned explicitly 4882in the call pattern, or included as clobbers in CALL_INSN_FUNCTION_USAGE. 4883The default version of this hook is set to false. The purpose of this hook 4884is to enable the fipa-ra optimization. 4885@end deftypevr 4886 4887@node Varargs 4888@section Implementing the Varargs Macros 4889@cindex varargs implementation 4890 4891GCC comes with an implementation of @code{<varargs.h>} and 4892@code{<stdarg.h>} that work without change on machines that pass arguments 4893on the stack. Other machines require their own implementations of 4894varargs, and the two machine independent header files must have 4895conditionals to include it. 4896 4897ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in 4898the calling convention for @code{va_start}. The traditional 4899implementation takes just one argument, which is the variable in which 4900to store the argument pointer. The ISO implementation of 4901@code{va_start} takes an additional second argument. The user is 4902supposed to write the last named argument of the function here. 4903 4904However, @code{va_start} should not use this argument. The way to find 4905the end of the named arguments is with the built-in functions described 4906below. 4907 4908@defmac __builtin_saveregs () 4909Use this built-in function to save the argument registers in memory so 4910that the varargs mechanism can access them. Both ISO and traditional 4911versions of @code{va_start} must use @code{__builtin_saveregs}, unless 4912you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead. 4913 4914On some machines, @code{__builtin_saveregs} is open-coded under the 4915control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On 4916other machines, it calls a routine written in assembler language, 4917found in @file{libgcc2.c}. 4918 4919Code generated for the call to @code{__builtin_saveregs} appears at the 4920beginning of the function, as opposed to where the call to 4921@code{__builtin_saveregs} is written, regardless of what the code is. 4922This is because the registers must be saved before the function starts 4923to use them for its own purposes. 4924@c i rewrote the first sentence above to fix an overfull hbox. --mew 4925@c 10feb93 4926@end defmac 4927 4928@defmac __builtin_next_arg (@var{lastarg}) 4929This builtin returns the address of the first anonymous stack 4930argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it 4931returns the address of the location above the first anonymous stack 4932argument. Use it in @code{va_start} to initialize the pointer for 4933fetching arguments from the stack. Also use it in @code{va_start} to 4934verify that the second parameter @var{lastarg} is the last named argument 4935of the current function. 4936@end defmac 4937 4938@defmac __builtin_classify_type (@var{object}) 4939Since each machine has its own conventions for which data types are 4940passed in which kind of register, your implementation of @code{va_arg} 4941has to embody these conventions. The easiest way to categorize the 4942specified data type is to use @code{__builtin_classify_type} together 4943with @code{sizeof} and @code{__alignof__}. 4944 4945@code{__builtin_classify_type} ignores the value of @var{object}, 4946considering only its data type. It returns an integer describing what 4947kind of type that is---integer, floating, pointer, structure, and so on. 4948 4949The file @file{typeclass.h} defines an enumeration that you can use to 4950interpret the values of @code{__builtin_classify_type}. 4951@end defmac 4952 4953These machine description macros help implement varargs: 4954 4955@deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void) 4956If defined, this hook produces the machine-specific code for a call to 4957@code{__builtin_saveregs}. This code will be moved to the very 4958beginning of the function, before any parameter access are made. The 4959return value of this function should be an RTX that contains the value 4960to use as the return of @code{__builtin_saveregs}. 4961@end deftypefn 4962 4963@deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (cumulative_args_t @var{args_so_far}, machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time}) 4964This target hook offers an alternative to using 4965@code{__builtin_saveregs} and defining the hook 4966@code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous 4967register arguments into the stack so that all the arguments appear to 4968have been passed consecutively on the stack. Once this is done, you can 4969use the standard implementation of varargs that works for machines that 4970pass all their arguments on the stack. 4971 4972The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data 4973structure, containing the values that are obtained after processing the 4974named arguments. The arguments @var{mode} and @var{type} describe the 4975last named argument---its machine mode and its data type as a tree node. 4976 4977The target hook should do two things: first, push onto the stack all the 4978argument registers @emph{not} used for the named arguments, and second, 4979store the size of the data thus pushed into the @code{int}-valued 4980variable pointed to by @var{pretend_args_size}. The value that you 4981store here will serve as additional offset for setting up the stack 4982frame. 4983 4984Because you must generate code to push the anonymous arguments at 4985compile time without knowing their data types, 4986@code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that 4987have just a single category of argument register and use it uniformly 4988for all data types. 4989 4990If the argument @var{second_time} is nonzero, it means that the 4991arguments of the function are being analyzed for the second time. This 4992happens for an inline function, which is not actually compiled until the 4993end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should 4994not generate any instructions in this case. 4995@end deftypefn 4996 4997@deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca}) 4998Define this hook to return @code{true} if the location where a function 4999argument is passed depends on whether or not it is a named argument. 5000 5001This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG} 5002is set for varargs and stdarg functions. If this hook returns 5003@code{true}, the @var{named} argument is always true for named 5004arguments, and false for unnamed arguments. If it returns @code{false}, 5005but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true}, 5006then all arguments are treated as named. Otherwise, all named arguments 5007except the last are treated as named. 5008 5009You need not define this hook if it always returns @code{false}. 5010@end deftypefn 5011 5012@deftypefn {Target Hook} void TARGET_CALL_ARGS (rtx, @var{tree}) 5013While generating RTL for a function call, this target hook is invoked once 5014for each argument passed to the function, either a register returned by 5015@code{TARGET_FUNCTION_ARG} or a memory location. It is called just 5016before the point where argument registers are stored. The type of the 5017function to be called is also passed as the second argument; it is 5018@code{NULL_TREE} for libcalls. The @code{TARGET_END_CALL_ARGS} hook is 5019invoked just after the code to copy the return reg has been emitted. 5020This functionality can be used to perform special setup of call argument 5021registers if a target needs it. 5022For functions without arguments, the hook is called once with @code{pc_rtx} 5023passed instead of an argument register. 5024Most ports do not need to implement anything for this hook. 5025@end deftypefn 5026 5027@deftypefn {Target Hook} void TARGET_END_CALL_ARGS (void) 5028This target hook is invoked while generating RTL for a function call, 5029just after the point where the return reg is copied into a pseudo. It 5030signals that all the call argument and return registers for the just 5031emitted call are now no longer in use. 5032Most ports do not need to implement anything for this hook. 5033@end deftypefn 5034 5035@deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca}) 5036If you need to conditionally change ABIs so that one works with 5037@code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither 5038@code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was 5039defined, then define this hook to return @code{true} if 5040@code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise. 5041Otherwise, you should not define this hook. 5042@end deftypefn 5043 5044@deftypefn {Target Hook} rtx TARGET_LOAD_BOUNDS_FOR_ARG (rtx @var{slot}, rtx @var{arg}, rtx @var{slot_no}) 5045This hook is used by expand pass to emit insn to load bounds of 5046@var{arg} passed in @var{slot}. Expand pass uses this hook in case 5047bounds of @var{arg} are not passed in register. If @var{slot} is a 5048memory, then bounds are loaded as for regular pointer loaded from 5049memory. If @var{slot} is not a memory then @var{slot_no} is an integer 5050constant holding number of the target dependent special slot which 5051should be used to obtain bounds. Hook returns RTX holding loaded bounds. 5052@end deftypefn 5053 5054@deftypefn {Target Hook} void TARGET_STORE_BOUNDS_FOR_ARG (rtx @var{arg}, rtx @var{slot}, rtx @var{bounds}, rtx @var{slot_no}) 5055This hook is used by expand pass to emit insns to store @var{bounds} of 5056@var{arg} passed in @var{slot}. Expand pass uses this hook in case 5057@var{bounds} of @var{arg} are not passed in register. If @var{slot} is a 5058memory, then @var{bounds} are stored as for regular pointer stored in 5059memory. If @var{slot} is not a memory then @var{slot_no} is an integer 5060constant holding number of the target dependent special slot which 5061should be used to store @var{bounds}. 5062@end deftypefn 5063 5064@deftypefn {Target Hook} rtx TARGET_LOAD_RETURNED_BOUNDS (rtx @var{slot}) 5065This hook is used by expand pass to emit insn to load bounds 5066returned by function call in @var{slot}. Hook returns RTX holding 5067loaded bounds. 5068@end deftypefn 5069 5070@deftypefn {Target Hook} void TARGET_STORE_RETURNED_BOUNDS (rtx @var{slot}, rtx @var{bounds}) 5071This hook is used by expand pass to emit insn to store @var{bounds} 5072returned by function call into @var{slot}. 5073@end deftypefn 5074 5075@deftypefn {Target Hook} rtx TARGET_CHKP_FUNCTION_VALUE_BOUNDS (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing}) 5076Define this to return an RTX representing the place where a function 5077returns bounds for returned pointers. Arguments meaning is similar to 5078@code{TARGET_FUNCTION_VALUE}. 5079@end deftypefn 5080 5081@deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARG_BOUNDS (cumulative_args_t @var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time}) 5082Use it to store bounds for anonymous register arguments stored 5083into the stack. Arguments meaning is similar to 5084@code{TARGET_SETUP_INCOMING_VARARGS}. 5085@end deftypefn 5086 5087@node Trampolines 5088@section Trampolines for Nested Functions 5089@cindex trampolines for nested functions 5090@cindex nested functions, trampolines for 5091 5092A @dfn{trampoline} is a small piece of code that is created at run time 5093when the address of a nested function is taken. It normally resides on 5094the stack, in the stack frame of the containing function. These macros 5095tell GCC how to generate code to allocate and initialize a 5096trampoline. 5097 5098The instructions in the trampoline must do two things: load a constant 5099address into the static chain register, and jump to the real address of 5100the nested function. On CISC machines such as the m68k, this requires 5101two instructions, a move immediate and a jump. Then the two addresses 5102exist in the trampoline as word-long immediate operands. On RISC 5103machines, it is often necessary to load each address into a register in 5104two parts. Then pieces of each address form separate immediate 5105operands. 5106 5107The code generated to initialize the trampoline must store the variable 5108parts---the static chain value and the function address---into the 5109immediate operands of the instructions. On a CISC machine, this is 5110simply a matter of copying each address to a memory reference at the 5111proper offset from the start of the trampoline. On a RISC machine, it 5112may be necessary to take out pieces of the address and store them 5113separately. 5114 5115@deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f}) 5116This hook is called by @code{assemble_trampoline_template} to output, 5117on the stream @var{f}, assembler code for a block of data that contains 5118the constant parts of a trampoline. This code should not include a 5119label---the label is taken care of automatically. 5120 5121If you do not define this hook, it means no template is needed 5122for the target. Do not define this hook on systems where the block move 5123code to copy the trampoline into place would be larger than the code 5124to generate it on the spot. 5125@end deftypefn 5126 5127@defmac TRAMPOLINE_SECTION 5128Return the section into which the trampoline template is to be placed 5129(@pxref{Sections}). The default value is @code{readonly_data_section}. 5130@end defmac 5131 5132@defmac TRAMPOLINE_SIZE 5133A C expression for the size in bytes of the trampoline, as an integer. 5134@end defmac 5135 5136@defmac TRAMPOLINE_ALIGNMENT 5137Alignment required for trampolines, in bits. 5138 5139If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT} 5140is used for aligning trampolines. 5141@end defmac 5142 5143@deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain}) 5144This hook is called to initialize a trampoline. 5145@var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl} 5146is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an 5147RTX for the static chain value that should be passed to the function 5148when it is called. 5149 5150If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the 5151first thing this hook should do is emit a block move into @var{m_tramp} 5152from the memory block returned by @code{assemble_trampoline_template}. 5153Note that the block move need only cover the constant parts of the 5154trampoline. If the target isolates the variable parts of the trampoline 5155to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied. 5156 5157If the target requires any other actions, such as flushing caches or 5158enabling stack execution, these actions should be performed after 5159initializing the trampoline proper. 5160@end deftypefn 5161 5162@deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr}) 5163This hook should perform any machine-specific adjustment in 5164the address of the trampoline. Its argument contains the address of the 5165memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case 5166the address to be used for a function call should be different from the 5167address at which the template was stored, the different address should 5168be returned; otherwise @var{addr} should be returned unchanged. 5169If this hook is not defined, @var{addr} will be used for function calls. 5170@end deftypefn 5171 5172Implementing trampolines is difficult on many machines because they have 5173separate instruction and data caches. Writing into a stack location 5174fails to clear the memory in the instruction cache, so when the program 5175jumps to that location, it executes the old contents. 5176 5177Here are two possible solutions. One is to clear the relevant parts of 5178the instruction cache whenever a trampoline is set up. The other is to 5179make all trampolines identical, by having them jump to a standard 5180subroutine. The former technique makes trampoline execution faster; the 5181latter makes initialization faster. 5182 5183To clear the instruction cache when a trampoline is initialized, define 5184the following macro. 5185 5186@defmac CLEAR_INSN_CACHE (@var{beg}, @var{end}) 5187If defined, expands to a C expression clearing the @emph{instruction 5188cache} in the specified interval. The definition of this macro would 5189typically be a series of @code{asm} statements. Both @var{beg} and 5190@var{end} are both pointer expressions. 5191@end defmac 5192 5193To use a standard subroutine, define the following macro. In addition, 5194you must make sure that the instructions in a trampoline fill an entire 5195cache line with identical instructions, or else ensure that the 5196beginning of the trampoline code is always aligned at the same point in 5197its cache line. Look in @file{m68k.h} as a guide. 5198 5199@defmac TRANSFER_FROM_TRAMPOLINE 5200Define this macro if trampolines need a special subroutine to do their 5201work. The macro should expand to a series of @code{asm} statements 5202which will be compiled with GCC@. They go in a library function named 5203@code{__transfer_from_trampoline}. 5204 5205If you need to avoid executing the ordinary prologue code of a compiled 5206C function when you jump to the subroutine, you can do so by placing a 5207special label of your own in the assembler code. Use one @code{asm} 5208statement to generate an assembler label, and another to make the label 5209global. Then trampolines can use that label to jump directly to your 5210special assembler code. 5211@end defmac 5212 5213@node Library Calls 5214@section Implicit Calls to Library Routines 5215@cindex library subroutine names 5216@cindex @file{libgcc.a} 5217 5218@c prevent bad page break with this line 5219Here is an explanation of implicit calls to library routines. 5220 5221@defmac DECLARE_LIBRARY_RENAMES 5222This macro, if defined, should expand to a piece of C code that will get 5223expanded when compiling functions for libgcc.a. It can be used to 5224provide alternate names for GCC's internal library functions if there 5225are ABI-mandated names that the compiler should provide. 5226@end defmac 5227 5228@findex set_optab_libfunc 5229@findex init_one_libfunc 5230@deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void) 5231This hook should declare additional library routines or rename 5232existing ones, using the functions @code{set_optab_libfunc} and 5233@code{init_one_libfunc} defined in @file{optabs.c}. 5234@code{init_optabs} calls this macro after initializing all the normal 5235library routines. 5236 5237The default is to do nothing. Most ports don't need to define this hook. 5238@end deftypefn 5239 5240@deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX 5241If false (the default), internal library routines start with two 5242underscores. If set to true, these routines start with @code{__gnu_} 5243instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This 5244currently only affects functions defined in @file{libgcc2.c}. If this 5245is set to true, the @file{tm.h} file must also 5246@code{#define LIBGCC2_GNU_PREFIX}. 5247@end deftypevr 5248 5249@defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison}) 5250This macro should return @code{true} if the library routine that 5251implements the floating point comparison operator @var{comparison} in 5252mode @var{mode} will return a boolean, and @var{false} if it will 5253return a tristate. 5254 5255GCC's own floating point libraries return tristates from the 5256comparison operators, so the default returns false always. Most ports 5257don't need to define this macro. 5258@end defmac 5259 5260@defmac TARGET_LIB_INT_CMP_BIASED 5261This macro should evaluate to @code{true} if the integer comparison 5262functions (like @code{__cmpdi2}) return 0 to indicate that the first 5263operand is smaller than the second, 1 to indicate that they are equal, 5264and 2 to indicate that the first operand is greater than the second. 5265If this macro evaluates to @code{false} the comparison functions return 5266@minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines 5267in @file{libgcc.a}, you do not need to define this macro. 5268@end defmac 5269 5270@defmac TARGET_HAS_NO_HW_DIVIDE 5271This macro should be defined if the target has no hardware divide 5272instructions. If this macro is defined, GCC will use an algorithm which 5273make use of simple logical and arithmetic operations for 64-bit 5274division. If the macro is not defined, GCC will use an algorithm which 5275make use of a 64-bit by 32-bit divide primitive. 5276@end defmac 5277 5278@cindex @code{EDOM}, implicit usage 5279@findex matherr 5280@defmac TARGET_EDOM 5281The value of @code{EDOM} on the target machine, as a C integer constant 5282expression. If you don't define this macro, GCC does not attempt to 5283deposit the value of @code{EDOM} into @code{errno} directly. Look in 5284@file{/usr/include/errno.h} to find the value of @code{EDOM} on your 5285system. 5286 5287If you do not define @code{TARGET_EDOM}, then compiled code reports 5288domain errors by calling the library function and letting it report the 5289error. If mathematical functions on your system use @code{matherr} when 5290there is an error, then you should leave @code{TARGET_EDOM} undefined so 5291that @code{matherr} is used normally. 5292@end defmac 5293 5294@cindex @code{errno}, implicit usage 5295@defmac GEN_ERRNO_RTX 5296Define this macro as a C expression to create an rtl expression that 5297refers to the global ``variable'' @code{errno}. (On certain systems, 5298@code{errno} may not actually be a variable.) If you don't define this 5299macro, a reasonable default is used. 5300@end defmac 5301 5302@deftypefn {Target Hook} bool TARGET_LIBC_HAS_FUNCTION (enum function_class @var{fn_class}) 5303This hook determines whether a function from a class of functions 5304@var{fn_class} is present at the runtime. 5305@end deftypefn 5306 5307@defmac NEXT_OBJC_RUNTIME 5308Set this macro to 1 to use the "NeXT" Objective-C message sending conventions 5309by default. This calling convention involves passing the object, the selector 5310and the method arguments all at once to the method-lookup library function. 5311This is the usual setting when targeting Darwin/Mac OS X systems, which have 5312the NeXT runtime installed. 5313 5314If the macro is set to 0, the "GNU" Objective-C message sending convention 5315will be used by default. This convention passes just the object and the 5316selector to the method-lookup function, which returns a pointer to the method. 5317 5318In either case, it remains possible to select code-generation for the alternate 5319scheme, by means of compiler command line switches. 5320@end defmac 5321 5322@node Addressing Modes 5323@section Addressing Modes 5324@cindex addressing modes 5325 5326@c prevent bad page break with this line 5327This is about addressing modes. 5328 5329@defmac HAVE_PRE_INCREMENT 5330@defmacx HAVE_PRE_DECREMENT 5331@defmacx HAVE_POST_INCREMENT 5332@defmacx HAVE_POST_DECREMENT 5333A C expression that is nonzero if the machine supports pre-increment, 5334pre-decrement, post-increment, or post-decrement addressing respectively. 5335@end defmac 5336 5337@defmac HAVE_PRE_MODIFY_DISP 5338@defmacx HAVE_POST_MODIFY_DISP 5339A C expression that is nonzero if the machine supports pre- or 5340post-address side-effect generation involving constants other than 5341the size of the memory operand. 5342@end defmac 5343 5344@defmac HAVE_PRE_MODIFY_REG 5345@defmacx HAVE_POST_MODIFY_REG 5346A C expression that is nonzero if the machine supports pre- or 5347post-address side-effect generation involving a register displacement. 5348@end defmac 5349 5350@defmac CONSTANT_ADDRESS_P (@var{x}) 5351A C expression that is 1 if the RTX @var{x} is a constant which 5352is a valid address. On most machines the default definition of 5353@code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)} 5354is acceptable, but a few machines are more restrictive as to which 5355constant addresses are supported. 5356@end defmac 5357 5358@defmac CONSTANT_P (@var{x}) 5359@code{CONSTANT_P}, which is defined by target-independent code, 5360accepts integer-values expressions whose values are not explicitly 5361known, such as @code{symbol_ref}, @code{label_ref}, and @code{high} 5362expressions and @code{const} arithmetic expressions, in addition to 5363@code{const_int} and @code{const_double} expressions. 5364@end defmac 5365 5366@defmac MAX_REGS_PER_ADDRESS 5367A number, the maximum number of registers that can appear in a valid 5368memory address. Note that it is up to you to specify a value equal to 5369the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever 5370accept. 5371@end defmac 5372 5373@deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{x}, bool @var{strict}) 5374A function that returns whether @var{x} (an RTX) is a legitimate memory 5375address on the target machine for a memory operand of mode @var{mode}. 5376 5377Legitimate addresses are defined in two variants: a strict variant and a 5378non-strict one. The @var{strict} parameter chooses which variant is 5379desired by the caller. 5380 5381The strict variant is used in the reload pass. It must be defined so 5382that any pseudo-register that has not been allocated a hard register is 5383considered a memory reference. This is because in contexts where some 5384kind of register is required, a pseudo-register with no hard register 5385must be rejected. For non-hard registers, the strict variant should look 5386up the @code{reg_renumber} array; it should then proceed using the hard 5387register number in the array, or treat the pseudo as a memory reference 5388if the array holds @code{-1}. 5389 5390The non-strict variant is used in other passes. It must be defined to 5391accept all pseudo-registers in every context where some kind of 5392register is required. 5393 5394Normally, constant addresses which are the sum of a @code{symbol_ref} 5395and an integer are stored inside a @code{const} RTX to mark them as 5396constant. Therefore, there is no need to recognize such sums 5397specifically as legitimate addresses. Normally you would simply 5398recognize any @code{const} as legitimate. 5399 5400Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant 5401sums that are not marked with @code{const}. It assumes that a naked 5402@code{plus} indicates indexing. If so, then you @emph{must} reject such 5403naked constant sums as illegitimate addresses, so that none of them will 5404be given to @code{PRINT_OPERAND_ADDRESS}. 5405 5406@cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation 5407On some machines, whether a symbolic address is legitimate depends on 5408the section that the address refers to. On these machines, define the 5409target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information 5410into the @code{symbol_ref}, and then check for it here. When you see a 5411@code{const}, you will have to look inside it to find the 5412@code{symbol_ref} in order to determine the section. @xref{Assembler 5413Format}. 5414 5415@cindex @code{GO_IF_LEGITIMATE_ADDRESS} 5416Some ports are still using a deprecated legacy substitute for 5417this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro 5418has this syntax: 5419 5420@example 5421#define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label}) 5422@end example 5423 5424@noindent 5425and should @code{goto @var{label}} if the address @var{x} is a valid 5426address on the target machine for a memory operand of mode @var{mode}. 5427 5428@findex REG_OK_STRICT 5429Compiler source files that want to use the strict variant of this 5430macro define the macro @code{REG_OK_STRICT}. You should use an 5431@code{#ifdef REG_OK_STRICT} conditional to define the strict variant in 5432that case and the non-strict variant otherwise. 5433 5434Using the hook is usually simpler because it limits the number of 5435files that are recompiled when changes are made. 5436@end deftypefn 5437 5438@defmac TARGET_MEM_CONSTRAINT 5439A single character to be used instead of the default @code{'m'} 5440character for general memory addresses. This defines the constraint 5441letter which matches the memory addresses accepted by 5442@code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to 5443support new address formats in your back end without changing the 5444semantics of the @code{'m'} constraint. This is necessary in order to 5445preserve functionality of inline assembly constructs using the 5446@code{'m'} constraint. 5447@end defmac 5448 5449@defmac FIND_BASE_TERM (@var{x}) 5450A C expression to determine the base term of address @var{x}, 5451or to provide a simplified version of @var{x} from which @file{alias.c} 5452can easily find the base term. This macro is used in only two places: 5453@code{find_base_value} and @code{find_base_term} in @file{alias.c}. 5454 5455It is always safe for this macro to not be defined. It exists so 5456that alias analysis can understand machine-dependent addresses. 5457 5458The typical use of this macro is to handle addresses containing 5459a label_ref or symbol_ref within an UNSPEC@. 5460@end defmac 5461 5462@deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode}) 5463This hook is given an invalid memory address @var{x} for an 5464operand of mode @var{mode} and should try to return a valid memory 5465address. 5466 5467@findex break_out_memory_refs 5468@var{x} will always be the result of a call to @code{break_out_memory_refs}, 5469and @var{oldx} will be the operand that was given to that function to produce 5470@var{x}. 5471 5472The code of the hook should not alter the substructure of 5473@var{x}. If it transforms @var{x} into a more legitimate form, it 5474should return the new @var{x}. 5475 5476It is not necessary for this hook to come up with a legitimate address, 5477with the exception of native TLS addresses (@pxref{Emulated TLS}). 5478The compiler has standard ways of doing so in all cases. In fact, if 5479the target supports only emulated TLS, it 5480is safe to omit this hook or make it return @var{x} if it cannot find 5481a valid way to legitimize the address. But often a machine-dependent 5482strategy can generate better code. 5483@end deftypefn 5484 5485@defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win}) 5486A C compound statement that attempts to replace @var{x}, which is an address 5487that needs reloading, with a valid memory address for an operand of mode 5488@var{mode}. @var{win} will be a C statement label elsewhere in the code. 5489It is not necessary to define this macro, but it might be useful for 5490performance reasons. 5491 5492For example, on the i386, it is sometimes possible to use a single 5493reload register instead of two by reloading a sum of two pseudo 5494registers into a register. On the other hand, for number of RISC 5495processors offsets are limited so that often an intermediate address 5496needs to be generated in order to address a stack slot. By defining 5497@code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses 5498generated for adjacent some stack slots can be made identical, and thus 5499be shared. 5500 5501@emph{Note}: This macro should be used with caution. It is necessary 5502to know something of how reload works in order to effectively use this, 5503and it is quite easy to produce macros that build in too much knowledge 5504of reload internals. 5505 5506@emph{Note}: This macro must be able to reload an address created by a 5507previous invocation of this macro. If it fails to handle such addresses 5508then the compiler may generate incorrect code or abort. 5509 5510@findex push_reload 5511The macro definition should use @code{push_reload} to indicate parts that 5512need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually 5513suitable to be passed unaltered to @code{push_reload}. 5514 5515The code generated by this macro must not alter the substructure of 5516@var{x}. If it transforms @var{x} into a more legitimate form, it 5517should assign @var{x} (which will always be a C variable) a new value. 5518This also applies to parts that you change indirectly by calling 5519@code{push_reload}. 5520 5521@findex strict_memory_address_p 5522The macro definition may use @code{strict_memory_address_p} to test if 5523the address has become legitimate. 5524 5525@findex copy_rtx 5526If you want to change only a part of @var{x}, one standard way of doing 5527this is to use @code{copy_rtx}. Note, however, that it unshares only a 5528single level of rtl. Thus, if the part to be changed is not at the 5529top level, you'll need to replace first the top level. 5530It is not necessary for this macro to come up with a legitimate 5531address; but often a machine-dependent strategy can generate better code. 5532@end defmac 5533 5534@deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr}, addr_space_t @var{addrspace}) 5535This hook returns @code{true} if memory address @var{addr} in address 5536space @var{addrspace} can have 5537different meanings depending on the machine mode of the memory 5538reference it is used for or if the address is valid for some modes 5539but not others. 5540 5541Autoincrement and autodecrement addresses typically have mode-dependent 5542effects because the amount of the increment or decrement is the size 5543of the operand being addressed. Some machines have other mode-dependent 5544addresses. Many RISC machines have no mode-dependent addresses. 5545 5546You may assume that @var{addr} is a valid address for the machine. 5547 5548The default version of this hook returns @code{false}. 5549@end deftypefn 5550 5551@deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (machine_mode @var{mode}, rtx @var{x}) 5552This hook returns true if @var{x} is a legitimate constant for a 5553@var{mode}-mode immediate operand on the target machine. You can assume that 5554@var{x} satisfies @code{CONSTANT_P}, so you need not check this. 5555 5556The default definition returns true. 5557@end deftypefn 5558 5559@deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x}) 5560This hook is used to undo the possibly obfuscating effects of the 5561@code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target 5562macros. Some backend implementations of these macros wrap symbol 5563references inside an @code{UNSPEC} rtx to represent PIC or similar 5564addressing modes. This target hook allows GCC's optimizers to understand 5565the semantics of these opaque @code{UNSPEC}s by converting them back 5566into their original form. 5567@end deftypefn 5568 5569@deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x}) 5570This hook should return true if @var{x} should not be emitted into 5571debug sections. 5572@end deftypefn 5573 5574@deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (machine_mode @var{mode}, rtx @var{x}) 5575This hook should return true if @var{x} is of a form that cannot (or 5576should not) be spilled to the constant pool. @var{mode} is the mode 5577of @var{x}. 5578 5579The default version of this hook returns false. 5580 5581The primary reason to define this hook is to prevent reload from 5582deciding that a non-legitimate constant would be better reloaded 5583from the constant pool instead of spilling and reloading a register 5584holding the constant. This restriction is often true of addresses 5585of TLS symbols for various targets. 5586@end deftypefn 5587 5588@deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (machine_mode @var{mode}, const_rtx @var{x}) 5589This hook should return true if pool entries for constant @var{x} can 5590be placed in an @code{object_block} structure. @var{mode} is the mode 5591of @var{x}. 5592 5593The default version returns false for all constants. 5594@end deftypefn 5595 5596@deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_DECL_P (const_tree @var{decl}) 5597This hook should return true if pool entries for @var{decl} should 5598be placed in an @code{object_block} structure. 5599 5600The default version returns true for all decls. 5601@end deftypefn 5602 5603@deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (unsigned @var{fn}, bool @var{md_fn}, bool @var{sqrt}) 5604This hook should return the DECL of a function that implements reciprocal of 5605the builtin function with builtin function code @var{fn}, or 5606@code{NULL_TREE} if such a function is not available. @var{md_fn} is true 5607when @var{fn} is a code of a machine-dependent builtin function. When 5608@var{sqrt} is true, additional optimizations that apply only to the reciprocal 5609of a square root function are performed, and only reciprocals of @code{sqrt} 5610function are valid. 5611@end deftypefn 5612 5613@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void) 5614This hook should return the DECL of a function @var{f} that given an 5615address @var{addr} as an argument returns a mask @var{m} that can be 5616used to extract from two vectors the relevant data that resides in 5617@var{addr} in case @var{addr} is not properly aligned. 5618 5619The autovectorizer, when vectorizing a load operation from an address 5620@var{addr} that may be unaligned, will generate two vector loads from 5621the two aligned addresses around @var{addr}. It then generates a 5622@code{REALIGN_LOAD} operation to extract the relevant data from the 5623two loaded vectors. The first two arguments to @code{REALIGN_LOAD}, 5624@var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and 5625the third argument, @var{OFF}, defines how the data will be extracted 5626from these two vectors: if @var{OFF} is 0, then the returned vector is 5627@var{v2}; otherwise, the returned vector is composed from the last 5628@var{VS}-@var{OFF} elements of @var{v1} concatenated to the first 5629@var{OFF} elements of @var{v2}. 5630 5631If this hook is defined, the autovectorizer will generate a call 5632to @var{f} (using the DECL tree that this hook returns) and will 5633use the return value of @var{f} as the argument @var{OFF} to 5634@code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f} 5635should comply with the semantics expected by @code{REALIGN_LOAD} 5636described above. 5637If this hook is not defined, then @var{addr} will be used as 5638the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low 5639log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered. 5640@end deftypefn 5641 5642@deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign}) 5643Returns cost of different scalar or vector statements for vectorization cost model. 5644For vector memory operations the cost may depend on type (@var{vectype}) and 5645misalignment value (@var{misalign}). 5646@end deftypefn 5647 5648@deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed}) 5649Return true if vector alignment is reachable (by peeling N iterations) for the given type. 5650@end deftypefn 5651 5652@deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST_OK (machine_mode, const unsigned char *@var{sel}) 5653Return true if a vector created for @code{vec_perm_const} is valid. 5654@end deftypefn 5655 5656@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type}) 5657This hook should return the DECL of a function that implements conversion of the 5658input vector of type @var{src_type} to type @var{dest_type}. 5659The value of @var{code} is one of the enumerators in @code{enum tree_code} and 5660specifies how the conversion is to be applied 5661(truncation, rounding, etc.). 5662 5663If this hook is defined, the autovectorizer will use the 5664@code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing 5665conversion. Otherwise, it will return @code{NULL_TREE}. 5666@end deftypefn 5667 5668@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in}) 5669This hook should return the decl of a function that implements the 5670vectorized variant of the builtin function with builtin function code 5671@var{code} or @code{NULL_TREE} if such a function is not available. 5672The value of @var{fndecl} is the builtin function declaration. The 5673return type of the vectorized function shall be of vector type 5674@var{vec_type_out} and the argument types should be @var{vec_type_in}. 5675@end deftypefn 5676 5677@deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed}) 5678This hook should return true if the target supports misaligned vector 5679store/load of a specific factor denoted in the @var{misalignment} 5680parameter. The vector store/load should be of machine mode @var{mode} and 5681the elements in the vectors should be of type @var{type}. @var{is_packed} 5682parameter is true if the memory access is defined in a packed struct. 5683@end deftypefn 5684 5685@deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_PREFERRED_SIMD_MODE (machine_mode @var{mode}) 5686This hook should return the preferred mode for vectorizing scalar 5687mode @var{mode}. The default is 5688equal to @code{word_mode}, because the vectorizer can do some 5689transformations even in absence of specialized @acronym{SIMD} hardware. 5690@end deftypefn 5691 5692@deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void) 5693This hook should return a mask of sizes that should be iterated over 5694after trying to autovectorize using the vector size derived from the 5695mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}. 5696The default is zero which means to not iterate over other vector sizes. 5697@end deftypefn 5698 5699@deftypefn {Target Hook} {void *} TARGET_VECTORIZE_INIT_COST (struct loop *@var{loop_info}) 5700This hook should initialize target-specific data structures in preparation for modeling the costs of vectorizing a loop or basic block. The default allocates three unsigned integers for accumulating costs for the prologue, body, and epilogue of the loop or basic block. If @var{loop_info} is non-NULL, it identifies the loop being vectorized; otherwise a single block is being vectorized. 5701@end deftypefn 5702 5703@deftypefn {Target Hook} unsigned TARGET_VECTORIZE_ADD_STMT_COST (void *@var{data}, int @var{count}, enum vect_cost_for_stmt @var{kind}, struct _stmt_vec_info *@var{stmt_info}, int @var{misalign}, enum vect_cost_model_location @var{where}) 5704This hook should update the target-specific @var{data} in response to adding @var{count} copies of the given @var{kind} of statement to a loop or basic block. The default adds the builtin vectorizer cost for the copies of the statement to the accumulator specified by @var{where}, (the prologue, body, or epilogue) and returns the amount added. The return value should be viewed as a tentative cost that may later be revised. 5705@end deftypefn 5706 5707@deftypefn {Target Hook} void TARGET_VECTORIZE_FINISH_COST (void *@var{data}, unsigned *@var{prologue_cost}, unsigned *@var{body_cost}, unsigned *@var{epilogue_cost}) 5708This hook should complete calculations of the cost of vectorizing a loop or basic block based on @var{data}, and return the prologue, body, and epilogue costs as unsigned integers. The default returns the value of the three accumulators. 5709@end deftypefn 5710 5711@deftypefn {Target Hook} void TARGET_VECTORIZE_DESTROY_COST_DATA (void *@var{data}) 5712This hook should release @var{data} and any related data structures allocated by TARGET_VECTORIZE_INIT_COST. The default releases the accumulator. 5713@end deftypefn 5714 5715@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_LOAD (tree) 5716This hook should return the built-in decl needed to load a vector of the given type within a transaction. 5717@end deftypefn 5718 5719@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_TM_STORE (tree) 5720This hook should return the built-in decl needed to store a vector of the given type within a transaction. 5721@end deftypefn 5722 5723@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale}) 5724Target builtin that implements vector gather operation. @var{mem_vectype} 5725is the vector type of the load and @var{index_type} is scalar type of 5726the index, scaled by @var{scale}. 5727The default is @code{NULL_TREE} which means to not vectorize gather 5728loads. 5729@end deftypefn 5730 5731@deftypefn {Target Hook} int TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN (struct cgraph_node *@var{}, struct cgraph_simd_clone *@var{}, @var{tree}, @var{int}) 5732This hook should set @var{vecsize_mangle}, @var{vecsize_int}, @var{vecsize_float} 5733fields in @var{simd_clone} structure pointed by @var{clone_info} argument and also 5734@var{simdlen} field if it was previously 0. 5735The hook should return 0 if SIMD clones shouldn't be emitted, 5736or number of @var{vecsize_mangle} variants that should be emitted. 5737@end deftypefn 5738 5739@deftypefn {Target Hook} void TARGET_SIMD_CLONE_ADJUST (struct cgraph_node *@var{}) 5740This hook should add implicit @code{attribute(target("..."))} attribute 5741to SIMD clone @var{node} if needed. 5742@end deftypefn 5743 5744@deftypefn {Target Hook} int TARGET_SIMD_CLONE_USABLE (struct cgraph_node *@var{}) 5745This hook should return -1 if SIMD clone @var{node} shouldn't be used 5746in vectorized loops in current function, or non-negative number if it is 5747usable. In that case, the smaller the number is, the more desirable it is 5748to use it. 5749@end deftypefn 5750 5751@node Anchored Addresses 5752@section Anchored Addresses 5753@cindex anchored addresses 5754@cindex @option{-fsection-anchors} 5755 5756GCC usually addresses every static object as a separate entity. 5757For example, if we have: 5758 5759@smallexample 5760static int a, b, c; 5761int foo (void) @{ return a + b + c; @} 5762@end smallexample 5763 5764the code for @code{foo} will usually calculate three separate symbolic 5765addresses: those of @code{a}, @code{b} and @code{c}. On some targets, 5766it would be better to calculate just one symbolic address and access 5767the three variables relative to it. The equivalent pseudocode would 5768be something like: 5769 5770@smallexample 5771int foo (void) 5772@{ 5773 register int *xr = &x; 5774 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 5775@} 5776@end smallexample 5777 5778(which isn't valid C). We refer to shared addresses like @code{x} as 5779``section anchors''. Their use is controlled by @option{-fsection-anchors}. 5780 5781The hooks below describe the target properties that GCC needs to know 5782in order to make effective use of section anchors. It won't use 5783section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET} 5784or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value. 5785 5786@deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET 5787The minimum offset that should be applied to a section anchor. 5788On most targets, it should be the smallest offset that can be 5789applied to a base register while still giving a legitimate address 5790for every mode. The default value is 0. 5791@end deftypevr 5792 5793@deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET 5794Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive) 5795offset that should be applied to section anchors. The default 5796value is 0. 5797@end deftypevr 5798 5799@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x}) 5800Write the assembly code to define section anchor @var{x}, which is a 5801@code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true. 5802The hook is called with the assembly output position set to the beginning 5803of @code{SYMBOL_REF_BLOCK (@var{x})}. 5804 5805If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses 5806it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}. 5807If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition 5808is @code{NULL}, which disables the use of section anchors altogether. 5809@end deftypefn 5810 5811@deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x}) 5812Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF} 5813@var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and 5814@samp{!SYMBOL_REF_ANCHOR_P (@var{x})}. 5815 5816The default version is correct for most targets, but you might need to 5817intercept this hook to handle things like target-specific attributes 5818or target-specific sections. 5819@end deftypefn 5820 5821@node Condition Code 5822@section Condition Code Status 5823@cindex condition code status 5824 5825The macros in this section can be split in two families, according to the 5826two ways of representing condition codes in GCC. 5827 5828The first representation is the so called @code{(cc0)} representation 5829(@pxref{Jump Patterns}), where all instructions can have an implicit 5830clobber of the condition codes. The second is the condition code 5831register representation, which provides better schedulability for 5832architectures that do have a condition code register, but on which 5833most instructions do not affect it. The latter category includes 5834most RISC machines. 5835 5836The implicit clobbering poses a strong restriction on the placement of 5837the definition and use of the condition code. In the past the definition 5838and use were always adjacent. However, recent changes to support trapping 5839arithmatic may result in the definition and user being in different blocks. 5840Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally, 5841the definition may be the source of exception handling edges. 5842 5843These restrictions can prevent important 5844optimizations on some machines. For example, on the IBM RS/6000, there 5845is a delay for taken branches unless the condition code register is set 5846three instructions earlier than the conditional branch. The instruction 5847scheduler cannot perform this optimization if it is not permitted to 5848separate the definition and use of the condition code register. 5849 5850For this reason, it is possible and suggested to use a register to 5851represent the condition code for new ports. If there is a specific 5852condition code register in the machine, use a hard register. If the 5853condition code or comparison result can be placed in any general register, 5854or if there are multiple condition registers, use a pseudo register. 5855Registers used to store the condition code value will usually have a mode 5856that is in class @code{MODE_CC}. 5857 5858Alternatively, you can use @code{BImode} if the comparison operator is 5859specified already in the compare instruction. In this case, you are not 5860interested in most macros in this section. 5861 5862@menu 5863* CC0 Condition Codes:: Old style representation of condition codes. 5864* MODE_CC Condition Codes:: Modern representation of condition codes. 5865@end menu 5866 5867@node CC0 Condition Codes 5868@subsection Representation of condition codes using @code{(cc0)} 5869@findex cc0 5870 5871@findex cc_status 5872The file @file{conditions.h} defines a variable @code{cc_status} to 5873describe how the condition code was computed (in case the interpretation of 5874the condition code depends on the instruction that it was set by). This 5875variable contains the RTL expressions on which the condition code is 5876currently based, and several standard flags. 5877 5878Sometimes additional machine-specific flags must be defined in the machine 5879description header file. It can also add additional machine-specific 5880information by defining @code{CC_STATUS_MDEP}. 5881 5882@defmac CC_STATUS_MDEP 5883C code for a data type which is used for declaring the @code{mdep} 5884component of @code{cc_status}. It defaults to @code{int}. 5885 5886This macro is not used on machines that do not use @code{cc0}. 5887@end defmac 5888 5889@defmac CC_STATUS_MDEP_INIT 5890A C expression to initialize the @code{mdep} field to ``empty''. 5891The default definition does nothing, since most machines don't use 5892the field anyway. If you want to use the field, you should probably 5893define this macro to initialize it. 5894 5895This macro is not used on machines that do not use @code{cc0}. 5896@end defmac 5897 5898@defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn}) 5899A C compound statement to set the components of @code{cc_status} 5900appropriately for an insn @var{insn} whose body is @var{exp}. It is 5901this macro's responsibility to recognize insns that set the condition 5902code as a byproduct of other activity as well as those that explicitly 5903set @code{(cc0)}. 5904 5905This macro is not used on machines that do not use @code{cc0}. 5906 5907If there are insns that do not set the condition code but do alter 5908other machine registers, this macro must check to see whether they 5909invalidate the expressions that the condition code is recorded as 5910reflecting. For example, on the 68000, insns that store in address 5911registers do not set the condition code, which means that usually 5912@code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such 5913insns. But suppose that the previous insn set the condition code 5914based on location @samp{a4@@(102)} and the current insn stores a new 5915value in @samp{a4}. Although the condition code is not changed by 5916this, it will no longer be true that it reflects the contents of 5917@samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter 5918@code{cc_status} in this case to say that nothing is known about the 5919condition code value. 5920 5921The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal 5922with the results of peephole optimization: insns whose patterns are 5923@code{parallel} RTXs containing various @code{reg}, @code{mem} or 5924constants which are just the operands. The RTL structure of these 5925insns is not sufficient to indicate what the insns actually do. What 5926@code{NOTICE_UPDATE_CC} should do when it sees one is just to run 5927@code{CC_STATUS_INIT}. 5928 5929A possible definition of @code{NOTICE_UPDATE_CC} is to call a function 5930that looks at an attribute (@pxref{Insn Attributes}) named, for example, 5931@samp{cc}. This avoids having detailed information about patterns in 5932two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}. 5933@end defmac 5934 5935@node MODE_CC Condition Codes 5936@subsection Representation of condition codes using registers 5937@findex CCmode 5938@findex MODE_CC 5939 5940@defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y}) 5941On many machines, the condition code may be produced by other instructions 5942than compares, for example the branch can use directly the condition 5943code set by a subtract instruction. However, on some machines 5944when the condition code is set this way some bits (such as the overflow 5945bit) are not set in the same way as a test instruction, so that a different 5946branch instruction must be used for some conditional branches. When 5947this happens, use the machine mode of the condition code register to 5948record different formats of the condition code register. Modes can 5949also be used to record which compare instruction (e.g. a signed or an 5950unsigned comparison) produced the condition codes. 5951 5952If other modes than @code{CCmode} are required, add them to 5953@file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose 5954a mode given an operand of a compare. This is needed because the modes 5955have to be chosen not only during RTL generation but also, for example, 5956by instruction combination. The result of @code{SELECT_CC_MODE} should 5957be consistent with the mode used in the patterns; for example to support 5958the case of the add on the SPARC discussed above, we have the pattern 5959 5960@smallexample 5961(define_insn "" 5962 [(set (reg:CC_NOOV 0) 5963 (compare:CC_NOOV 5964 (plus:SI (match_operand:SI 0 "register_operand" "%r") 5965 (match_operand:SI 1 "arith_operand" "rI")) 5966 (const_int 0)))] 5967 "" 5968 "@dots{}") 5969@end smallexample 5970 5971@noindent 5972together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode} 5973for comparisons whose argument is a @code{plus}: 5974 5975@smallexample 5976#define SELECT_CC_MODE(OP,X,Y) \ 5977 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ 5978 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \ 5979 ? CCFPEmode : CCFPmode) \ 5980 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ 5981 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \ 5982 ? CC_NOOVmode : CCmode)) 5983@end smallexample 5984 5985Another reason to use modes is to retain information on which operands 5986were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in 5987this section. 5988 5989You should define this macro if and only if you define extra CC modes 5990in @file{@var{machine}-modes.def}. 5991@end defmac 5992 5993@deftypefn {Target Hook} void TARGET_CANONICALIZE_COMPARISON (int *@var{code}, rtx *@var{op0}, rtx *@var{op1}, bool @var{op0_preserve_value}) 5994On some machines not all possible comparisons are defined, but you can 5995convert an invalid comparison into a valid one. For example, the Alpha 5996does not have a @code{GT} comparison, but you can use an @code{LT} 5997comparison instead and swap the order of the operands. 5998 5999On such machines, implement this hook to do any required conversions. 6000@var{code} is the initial comparison code and @var{op0} and @var{op1} 6001are the left and right operands of the comparison, respectively. If 6002@var{op0_preserve_value} is @code{true} the implementation is not 6003allowed to change the value of @var{op0} since the value might be used 6004in RTXs which aren't comparisons. E.g. the implementation is not 6005allowed to swap operands in that case. 6006 6007GCC will not assume that the comparison resulting from this macro is 6008valid but will see if the resulting insn matches a pattern in the 6009@file{md} file. 6010 6011You need not to implement this hook if it would never change the 6012comparison code or operands. 6013@end deftypefn 6014 6015@defmac REVERSIBLE_CC_MODE (@var{mode}) 6016A C expression whose value is one if it is always safe to reverse a 6017comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE} 6018can ever return @var{mode} for a floating-point inequality comparison, 6019then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero. 6020 6021You need not define this macro if it would always returns zero or if the 6022floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}. 6023For example, here is the definition used on the SPARC, where floating-point 6024inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}: 6025 6026@smallexample 6027#define REVERSIBLE_CC_MODE(MODE) \ 6028 ((MODE) != CCFPEmode && (MODE) != CCFPmode) 6029@end smallexample 6030@end defmac 6031 6032@defmac REVERSE_CONDITION (@var{code}, @var{mode}) 6033A C expression whose value is reversed condition code of the @var{code} for 6034comparison done in CC_MODE @var{mode}. The macro is used only in case 6035@code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case 6036machine has some non-standard way how to reverse certain conditionals. For 6037instance in case all floating point conditions are non-trapping, compiler may 6038freely convert unordered compares to ordered ones. Then definition may look 6039like: 6040 6041@smallexample 6042#define REVERSE_CONDITION(CODE, MODE) \ 6043 ((MODE) != CCFPmode ? reverse_condition (CODE) \ 6044 : reverse_condition_maybe_unordered (CODE)) 6045@end smallexample 6046@end defmac 6047 6048@deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2}) 6049On targets which do not use @code{(cc0)}, and which use a hard 6050register rather than a pseudo-register to hold condition codes, the 6051regular CSE passes are often not able to identify cases in which the 6052hard register is set to a common value. Use this hook to enable a 6053small pass which optimizes such cases. This hook should return true 6054to enable this pass, and it should set the integers to which its 6055arguments point to the hard register numbers used for condition codes. 6056When there is only one such register, as is true on most systems, the 6057integer pointed to by @var{p2} should be set to 6058@code{INVALID_REGNUM}. 6059 6060The default version of this hook returns false. 6061@end deftypefn 6062 6063@deftypefn {Target Hook} machine_mode TARGET_CC_MODES_COMPATIBLE (machine_mode @var{m1}, machine_mode @var{m2}) 6064On targets which use multiple condition code modes in class 6065@code{MODE_CC}, it is sometimes the case that a comparison can be 6066validly done in more than one mode. On such a system, define this 6067target hook to take two mode arguments and to return a mode in which 6068both comparisons may be validly done. If there is no such mode, 6069return @code{VOIDmode}. 6070 6071The default version of this hook checks whether the modes are the 6072same. If they are, it returns that mode. If they are different, it 6073returns @code{VOIDmode}. 6074@end deftypefn 6075 6076@deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM 6077If the target has a dedicated flags register, and it needs to use the post-reload comparison elimination pass, then this value should be set appropriately. 6078@end deftypevr 6079 6080@node Costs 6081@section Describing Relative Costs of Operations 6082@cindex costs of instructions 6083@cindex relative costs 6084@cindex speed of instructions 6085 6086These macros let you describe the relative speed of various operations 6087on the target machine. 6088 6089@defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to}) 6090A C expression for the cost of moving data of mode @var{mode} from a 6091register in class @var{from} to one in class @var{to}. The classes are 6092expressed using the enumeration values such as @code{GENERAL_REGS}. A 6093value of 2 is the default; other values are interpreted relative to 6094that. 6095 6096It is not required that the cost always equal 2 when @var{from} is the 6097same as @var{to}; on some machines it is expensive to move between 6098registers if they are not general registers. 6099 6100If reload sees an insn consisting of a single @code{set} between two 6101hard registers, and if @code{REGISTER_MOVE_COST} applied to their 6102classes returns a value of 2, reload does not check to ensure that the 6103constraints of the insn are met. Setting a cost of other than 2 will 6104allow reload to verify that the constraints are met. You should do this 6105if the @samp{mov@var{m}} pattern's constraints do not allow such copying. 6106 6107These macros are obsolete, new ports should use the target hook 6108@code{TARGET_REGISTER_MOVE_COST} instead. 6109@end defmac 6110 6111@deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to}) 6112This target hook should return the cost of moving data of mode @var{mode} 6113from a register in class @var{from} to one in class @var{to}. The classes 6114are expressed using the enumeration values such as @code{GENERAL_REGS}. 6115A value of 2 is the default; other values are interpreted relative to 6116that. 6117 6118It is not required that the cost always equal 2 when @var{from} is the 6119same as @var{to}; on some machines it is expensive to move between 6120registers if they are not general registers. 6121 6122If reload sees an insn consisting of a single @code{set} between two 6123hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their 6124classes returns a value of 2, reload does not check to ensure that the 6125constraints of the insn are met. Setting a cost of other than 2 will 6126allow reload to verify that the constraints are met. You should do this 6127if the @samp{mov@var{m}} pattern's constraints do not allow such copying. 6128 6129The default version of this function returns 2. 6130@end deftypefn 6131 6132@defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in}) 6133A C expression for the cost of moving data of mode @var{mode} between a 6134register of class @var{class} and memory; @var{in} is zero if the value 6135is to be written to memory, nonzero if it is to be read in. This cost 6136is relative to those in @code{REGISTER_MOVE_COST}. If moving between 6137registers and memory is more expensive than between two registers, you 6138should define this macro to express the relative cost. 6139 6140If you do not define this macro, GCC uses a default cost of 4 plus 6141the cost of copying via a secondary reload register, if one is 6142needed. If your machine requires a secondary reload register to copy 6143between memory and a register of @var{class} but the reload mechanism is 6144more complex than copying via an intermediate, define this macro to 6145reflect the actual cost of the move. 6146 6147GCC defines the function @code{memory_move_secondary_cost} if 6148secondary reloads are needed. It computes the costs due to copying via 6149a secondary register. If your machine copies from memory using a 6150secondary register in the conventional way but the default base value of 61514 is not correct for your machine, define this macro to add some other 6152value to the result of that function. The arguments to that function 6153are the same as to this macro. 6154 6155These macros are obsolete, new ports should use the target hook 6156@code{TARGET_MEMORY_MOVE_COST} instead. 6157@end defmac 6158 6159@deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in}) 6160This target hook should return the cost of moving data of mode @var{mode} 6161between a register of class @var{rclass} and memory; @var{in} is @code{false} 6162if the value is to be written to memory, @code{true} if it is to be read in. 6163This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}. 6164If moving between registers and memory is more expensive than between two 6165registers, you should add this target hook to express the relative cost. 6166 6167If you do not add this target hook, GCC uses a default cost of 4 plus 6168the cost of copying via a secondary reload register, if one is 6169needed. If your machine requires a secondary reload register to copy 6170between memory and a register of @var{rclass} but the reload mechanism is 6171more complex than copying via an intermediate, use this target hook to 6172reflect the actual cost of the move. 6173 6174GCC defines the function @code{memory_move_secondary_cost} if 6175secondary reloads are needed. It computes the costs due to copying via 6176a secondary register. If your machine copies from memory using a 6177secondary register in the conventional way but the default base value of 61784 is not correct for your machine, use this target hook to add some other 6179value to the result of that function. The arguments to that function 6180are the same as to this target hook. 6181@end deftypefn 6182 6183@defmac BRANCH_COST (@var{speed_p}, @var{predictable_p}) 6184A C expression for the cost of a branch instruction. A value of 1 is 6185the default; other values are interpreted relative to that. Parameter 6186@var{speed_p} is true when the branch in question should be optimized 6187for speed. When it is false, @code{BRANCH_COST} should return a value 6188optimal for code size rather than performance. @var{predictable_p} is 6189true for well-predicted branches. On many architectures the 6190@code{BRANCH_COST} can be reduced then. 6191@end defmac 6192 6193Here are additional macros which do not specify precise relative costs, 6194but only that certain actions are more expensive than GCC would 6195ordinarily expect. 6196 6197@defmac SLOW_BYTE_ACCESS 6198Define this macro as a C expression which is nonzero if accessing less 6199than a word of memory (i.e.@: a @code{char} or a @code{short}) is no 6200faster than accessing a word of memory, i.e., if such access 6201require more than one instruction or if there is no difference in cost 6202between byte and (aligned) word loads. 6203 6204When this macro is not defined, the compiler will access a field by 6205finding the smallest containing object; when it is defined, a fullword 6206load will be used if alignment permits. Unless bytes accesses are 6207faster than word accesses, using word accesses is preferable since it 6208may eliminate subsequent memory access if subsequent accesses occur to 6209other fields in the same word of the structure, but to different bytes. 6210@end defmac 6211 6212@defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment}) 6213Define this macro to be the value 1 if memory accesses described by the 6214@var{mode} and @var{alignment} parameters have a cost many times greater 6215than aligned accesses, for example if they are emulated in a trap 6216handler. 6217 6218When this macro is nonzero, the compiler will act as if 6219@code{STRICT_ALIGNMENT} were nonzero when generating code for block 6220moves. This can cause significantly more instructions to be produced. 6221Therefore, do not set this macro nonzero if unaligned accesses only add a 6222cycle or two to the time for a memory access. 6223 6224If the value of this macro is always zero, it need not be defined. If 6225this macro is defined, it should produce a nonzero value when 6226@code{STRICT_ALIGNMENT} is nonzero. 6227@end defmac 6228 6229@defmac MOVE_RATIO (@var{speed}) 6230The threshold of number of scalar memory-to-memory move insns, @emph{below} 6231which a sequence of insns should be generated instead of a 6232string move insn or a library call. Increasing the value will always 6233make code faster, but eventually incurs high cost in increased code size. 6234 6235Note that on machines where the corresponding move insn is a 6236@code{define_expand} that emits a sequence of insns, this macro counts 6237the number of such sequences. 6238 6239The parameter @var{speed} is true if the code is currently being 6240optimized for speed rather than size. 6241 6242If you don't define this, a reasonable default is used. 6243@end defmac 6244 6245@deftypefn {Target Hook} bool TARGET_USE_BY_PIECES_INFRASTRUCTURE_P (unsigned HOST_WIDE_INT @var{size}, unsigned int @var{alignment}, enum by_pieces_operation @var{op}, bool @var{speed_p}) 6246GCC will attempt several strategies when asked to copy between 6247two areas of memory, or to set, clear or store to memory, for example 6248when copying a @code{struct}. The @code{by_pieces} infrastructure 6249implements such memory operations as a sequence of load, store or move 6250insns. Alternate strategies are to expand the 6251@code{movmem} or @code{setmem} optabs, to emit a library call, or to emit 6252unit-by-unit, loop-based operations. 6253 6254This target hook should return true if, for a memory operation with a 6255given @var{size} and @var{alignment}, using the @code{by_pieces} 6256infrastructure is expected to result in better code generation. 6257Both @var{size} and @var{alignment} are measured in terms of storage 6258units. 6259 6260The parameter @var{op} is one of: @code{CLEAR_BY_PIECES}, 6261@code{MOVE_BY_PIECES}, @code{SET_BY_PIECES}, @code{STORE_BY_PIECES}. 6262These describe the type of memory operation under consideration. 6263 6264The parameter @var{speed_p} is true if the code is currently being 6265optimized for speed rather than size. 6266 6267Returning true for higher values of @var{size} can improve code generation 6268for speed if the target does not provide an implementation of the 6269@code{movmem} or @code{setmem} standard names, if the @code{movmem} or 6270@code{setmem} implementation would be more expensive than a sequence of 6271insns, or if the overhead of a library call would dominate that of 6272the body of the memory operation. 6273 6274Returning true for higher values of @code{size} may also cause an increase 6275in code size, for example where the number of insns emitted to perform a 6276move would be greater than that of a library call. 6277@end deftypefn 6278 6279@defmac MOVE_MAX_PIECES 6280A C expression used by @code{move_by_pieces} to determine the largest unit 6281a load or store used to copy memory is. Defaults to @code{MOVE_MAX}. 6282@end defmac 6283 6284@defmac CLEAR_RATIO (@var{speed}) 6285The threshold of number of scalar move insns, @emph{below} which a sequence 6286of insns should be generated to clear memory instead of a string clear insn 6287or a library call. Increasing the value will always make code faster, but 6288eventually incurs high cost in increased code size. 6289 6290The parameter @var{speed} is true if the code is currently being 6291optimized for speed rather than size. 6292 6293If you don't define this, a reasonable default is used. 6294@end defmac 6295 6296@defmac SET_RATIO (@var{speed}) 6297The threshold of number of scalar move insns, @emph{below} which a sequence 6298of insns should be generated to set memory to a constant value, instead of 6299a block set insn or a library call. 6300Increasing the value will always make code faster, but 6301eventually incurs high cost in increased code size. 6302 6303The parameter @var{speed} is true if the code is currently being 6304optimized for speed rather than size. 6305 6306If you don't define this, it defaults to the value of @code{MOVE_RATIO}. 6307@end defmac 6308 6309@defmac USE_LOAD_POST_INCREMENT (@var{mode}) 6310A C expression used to determine whether a load postincrement is a good 6311thing to use for a given mode. Defaults to the value of 6312@code{HAVE_POST_INCREMENT}. 6313@end defmac 6314 6315@defmac USE_LOAD_POST_DECREMENT (@var{mode}) 6316A C expression used to determine whether a load postdecrement is a good 6317thing to use for a given mode. Defaults to the value of 6318@code{HAVE_POST_DECREMENT}. 6319@end defmac 6320 6321@defmac USE_LOAD_PRE_INCREMENT (@var{mode}) 6322A C expression used to determine whether a load preincrement is a good 6323thing to use for a given mode. Defaults to the value of 6324@code{HAVE_PRE_INCREMENT}. 6325@end defmac 6326 6327@defmac USE_LOAD_PRE_DECREMENT (@var{mode}) 6328A C expression used to determine whether a load predecrement is a good 6329thing to use for a given mode. Defaults to the value of 6330@code{HAVE_PRE_DECREMENT}. 6331@end defmac 6332 6333@defmac USE_STORE_POST_INCREMENT (@var{mode}) 6334A C expression used to determine whether a store postincrement is a good 6335thing to use for a given mode. Defaults to the value of 6336@code{HAVE_POST_INCREMENT}. 6337@end defmac 6338 6339@defmac USE_STORE_POST_DECREMENT (@var{mode}) 6340A C expression used to determine whether a store postdecrement is a good 6341thing to use for a given mode. Defaults to the value of 6342@code{HAVE_POST_DECREMENT}. 6343@end defmac 6344 6345@defmac USE_STORE_PRE_INCREMENT (@var{mode}) 6346This macro is used to determine whether a store preincrement is a good 6347thing to use for a given mode. Defaults to the value of 6348@code{HAVE_PRE_INCREMENT}. 6349@end defmac 6350 6351@defmac USE_STORE_PRE_DECREMENT (@var{mode}) 6352This macro is used to determine whether a store predecrement is a good 6353thing to use for a given mode. Defaults to the value of 6354@code{HAVE_PRE_DECREMENT}. 6355@end defmac 6356 6357@defmac NO_FUNCTION_CSE 6358Define this macro if it is as good or better to call a constant 6359function address than to call an address kept in a register. 6360@end defmac 6361 6362@defmac LOGICAL_OP_NON_SHORT_CIRCUIT 6363Define this macro if a non-short-circuit operation produced by 6364@samp{fold_range_test ()} is optimal. This macro defaults to true if 6365@code{BRANCH_COST} is greater than or equal to the value 2. 6366@end defmac 6367 6368@deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int @var{opno}, int *@var{total}, bool @var{speed}) 6369This target hook describes the relative costs of RTL expressions. 6370 6371The cost may depend on the precise form of the expression, which is 6372available for examination in @var{x}, and the fact that @var{x} appears 6373as operand @var{opno} of an expression with rtx code @var{outer_code}. 6374That is, the hook can assume that there is some rtx @var{y} such 6375that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that 6376either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or 6377(b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}. 6378 6379@var{code} is @var{x}'s expression code---redundant, since it can be 6380obtained with @code{GET_CODE (@var{x})}. 6381 6382In implementing this hook, you can use the construct 6383@code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast 6384instructions. 6385 6386On entry to the hook, @code{*@var{total}} contains a default estimate 6387for the cost of the expression. The hook should modify this value as 6388necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)} 6389for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus 6390operations, and @code{COSTS_N_INSNS (1)} for all other operations. 6391 6392When optimizing for code size, i.e.@: when @code{speed} is 6393false, this target hook should be used to estimate the relative 6394size cost of an expression, again relative to @code{COSTS_N_INSNS}. 6395 6396The hook returns true when all subexpressions of @var{x} have been 6397processed, and false when @code{rtx_cost} should recurse. 6398@end deftypefn 6399 6400@deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, machine_mode @var{mode}, addr_space_t @var{as}, bool @var{speed}) 6401This hook computes the cost of an addressing mode that contains 6402@var{address}. If not defined, the cost is computed from 6403the @var{address} expression and the @code{TARGET_RTX_COST} hook. 6404 6405For most CISC machines, the default cost is a good approximation of the 6406true cost of the addressing mode. However, on RISC machines, all 6407instructions normally have the same length and execution time. Hence 6408all addresses will have equal costs. 6409 6410In cases where more than one form of an address is known, the form with 6411the lowest cost will be used. If multiple forms have the same, lowest, 6412cost, the one that is the most complex will be used. 6413 6414For example, suppose an address that is equal to the sum of a register 6415and a constant is used twice in the same basic block. When this macro 6416is not defined, the address will be computed in a register and memory 6417references will be indirect through that register. On machines where 6418the cost of the addressing mode containing the sum is no higher than 6419that of a simple indirect reference, this will produce an additional 6420instruction and possibly require an additional register. Proper 6421specification of this macro eliminates this overhead for such machines. 6422 6423This hook is never called with an invalid address. 6424 6425On machines where an address involving more than one register is as 6426cheap as an address computation involving only one register, defining 6427@code{TARGET_ADDRESS_COST} to reflect this can cause two registers to 6428be live over a region of code where only one would have been if 6429@code{TARGET_ADDRESS_COST} were not defined in that manner. This effect 6430should be considered in the definition of this macro. Equivalent costs 6431should probably only be given to addresses with different numbers of 6432registers on machines with lots of registers. 6433@end deftypefn 6434 6435@node Scheduling 6436@section Adjusting the Instruction Scheduler 6437 6438The instruction scheduler may need a fair amount of machine-specific 6439adjustment in order to produce good code. GCC provides several target 6440hooks for this purpose. It is usually enough to define just a few of 6441them: try the first ones in this list first. 6442 6443@deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void) 6444This hook returns the maximum number of instructions that can ever 6445issue at the same time on the target machine. The default is one. 6446Although the insn scheduler can define itself the possibility of issue 6447an insn on the same cycle, the value can serve as an additional 6448constraint to issue insns on the same simulated processor cycle (see 6449hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}). 6450This value must be constant over the entire compilation. If you need 6451it to vary depending on what the instructions are, you must use 6452@samp{TARGET_SCHED_VARIABLE_ISSUE}. 6453@end deftypefn 6454 6455@deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx_insn *@var{insn}, int @var{more}) 6456This hook is executed by the scheduler after it has scheduled an insn 6457from the ready list. It should return the number of insns which can 6458still be issued in the current cycle. The default is 6459@samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and 6460@code{USE}, which normally are not counted against the issue rate. 6461You should define this hook if some insns take more machine resources 6462than others, so that fewer insns can follow them in the same cycle. 6463@var{file} is either a null pointer, or a stdio stream to write any 6464debug output to. @var{verbose} is the verbose level provided by 6465@option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that 6466was scheduled. 6467@end deftypefn 6468 6469@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx_insn *@var{insn}, rtx @var{link}, rtx_insn *@var{dep_insn}, int @var{cost}) 6470This function corrects the value of @var{cost} based on the 6471relationship between @var{insn} and @var{dep_insn} through the 6472dependence @var{link}. It should return the new value. The default 6473is to make no adjustment to @var{cost}. This can be used for example 6474to specify to the scheduler using the traditional pipeline description 6475that an output- or anti-dependence does not incur the same cost as a 6476data-dependence. If the scheduler using the automaton based pipeline 6477description, the cost of anti-dependence is zero and the cost of 6478output-dependence is maximum of one and the difference of latency 6479times of the first and the second insns. If these values are not 6480acceptable, you could use the hook to modify them too. See also 6481@pxref{Processor pipeline description}. 6482@end deftypefn 6483 6484@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx_insn *@var{insn}, int @var{priority}) 6485This hook adjusts the integer scheduling priority @var{priority} of 6486@var{insn}. It should return the new priority. Increase the priority to 6487execute @var{insn} earlier, reduce the priority to execute @var{insn} 6488later. Do not define this hook if you do not need to adjust the 6489scheduling priorities of insns. 6490@end deftypefn 6491 6492@deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock}) 6493This hook is executed by the scheduler after it has scheduled the ready 6494list, to allow the machine description to reorder it (for example to 6495combine two small instructions together on @samp{VLIW} machines). 6496@var{file} is either a null pointer, or a stdio stream to write any 6497debug output to. @var{verbose} is the verbose level provided by 6498@option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready 6499list of instructions that are ready to be scheduled. @var{n_readyp} is 6500a pointer to the number of elements in the ready list. The scheduler 6501reads the ready list in reverse order, starting with 6502@var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock} 6503is the timer tick of the scheduler. You may modify the ready list and 6504the number of ready insns. The return value is the number of insns that 6505can issue this cycle; normally this is just @code{issue_rate}. See also 6506@samp{TARGET_SCHED_REORDER2}. 6507@end deftypefn 6508 6509@deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock}) 6510Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That 6511function is called whenever the scheduler starts a new cycle. This one 6512is called once per iteration over a cycle, immediately after 6513@samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and 6514return the number of insns to be scheduled in the same cycle. Defining 6515this hook can be useful if there are frequent situations where 6516scheduling one insn causes other insns to become ready in the same 6517cycle. These other insns can then be taken into account properly. 6518@end deftypefn 6519 6520@deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_P (void) 6521This hook is used to check whether target platform supports macro fusion. 6522@end deftypefn 6523 6524@deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_PAIR_P (rtx_insn *@var{prev}, rtx_insn *@var{curr}) 6525This hook is used to check whether two insns should be macro fused for 6526a target microarchitecture. If this hook returns true for the given insn pair 6527(@var{prev} and @var{curr}), the scheduler will put them into a sched 6528group, and they will not be scheduled apart. The two insns will be either 6529two SET insns or a compare and a conditional jump and this hook should 6530validate any dependencies needed to fuse the two insns together. 6531@end deftypefn 6532 6533@deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx_insn *@var{head}, rtx_insn *@var{tail}) 6534This hook is called after evaluation forward dependencies of insns in 6535chain given by two parameter values (@var{head} and @var{tail} 6536correspondingly) but before insns scheduling of the insn chain. For 6537example, it can be used for better insn classification if it requires 6538analysis of dependencies. This hook can use backward and forward 6539dependencies of the insn scheduler because they are already 6540calculated. 6541@end deftypefn 6542 6543@deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready}) 6544This hook is executed by the scheduler at the beginning of each block of 6545instructions that are to be scheduled. @var{file} is either a null 6546pointer, or a stdio stream to write any debug output to. @var{verbose} 6547is the verbose level provided by @option{-fsched-verbose-@var{n}}. 6548@var{max_ready} is the maximum number of insns in the current scheduling 6549region that can be live at the same time. This can be used to allocate 6550scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}. 6551@end deftypefn 6552 6553@deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose}) 6554This hook is executed by the scheduler at the end of each block of 6555instructions that are to be scheduled. It can be used to perform 6556cleanup of any actions done by the other scheduling hooks. @var{file} 6557is either a null pointer, or a stdio stream to write any debug output 6558to. @var{verbose} is the verbose level provided by 6559@option{-fsched-verbose-@var{n}}. 6560@end deftypefn 6561 6562@deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid}) 6563This hook is executed by the scheduler after function level initializations. 6564@var{file} is either a null pointer, or a stdio stream to write any debug output to. 6565@var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}. 6566@var{old_max_uid} is the maximum insn uid when scheduling begins. 6567@end deftypefn 6568 6569@deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose}) 6570This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}. 6571@var{file} is either a null pointer, or a stdio stream to write any debug output to. 6572@var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}. 6573@end deftypefn 6574 6575@deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void) 6576The hook returns an RTL insn. The automaton state used in the 6577pipeline hazard recognizer is changed as if the insn were scheduled 6578when the new simulated processor cycle starts. Usage of the hook may 6579simplify the automaton pipeline description for some @acronym{VLIW} 6580processors. If the hook is defined, it is used only for the automaton 6581based pipeline description. The default is not to change the state 6582when the new simulated processor cycle starts. 6583@end deftypefn 6584 6585@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void) 6586The hook can be used to initialize data used by the previous hook. 6587@end deftypefn 6588 6589@deftypefn {Target Hook} {rtx_insn *} TARGET_SCHED_DFA_POST_CYCLE_INSN (void) 6590The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used 6591to changed the state as if the insn were scheduled when the new 6592simulated processor cycle finishes. 6593@end deftypefn 6594 6595@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void) 6596The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but 6597used to initialize data used by the previous hook. 6598@end deftypefn 6599 6600@deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void) 6601The hook to notify target that the current simulated cycle is about to finish. 6602The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used 6603to change the state in more complicated situations - e.g., when advancing 6604state on a single insn is not enough. 6605@end deftypefn 6606 6607@deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void) 6608The hook to notify target that new simulated cycle has just started. 6609The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used 6610to change the state in more complicated situations - e.g., when advancing 6611state on a single insn is not enough. 6612@end deftypefn 6613 6614@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void) 6615This hook controls better choosing an insn from the ready insn queue 6616for the @acronym{DFA}-based insn scheduler. Usually the scheduler 6617chooses the first insn from the queue. If the hook returns a positive 6618value, an additional scheduler code tries all permutations of 6619@samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()} 6620subsequent ready insns to choose an insn whose issue will result in 6621maximal number of issued insns on the same cycle. For the 6622@acronym{VLIW} processor, the code could actually solve the problem of 6623packing simple insns into the @acronym{VLIW} insn. Of course, if the 6624rules of @acronym{VLIW} packing are described in the automaton. 6625 6626This code also could be used for superscalar @acronym{RISC} 6627processors. Let us consider a superscalar @acronym{RISC} processor 6628with 3 pipelines. Some insns can be executed in pipelines @var{A} or 6629@var{B}, some insns can be executed only in pipelines @var{B} or 6630@var{C}, and one insn can be executed in pipeline @var{B}. The 6631processor may issue the 1st insn into @var{A} and the 2nd one into 6632@var{B}. In this case, the 3rd insn will wait for freeing @var{B} 6633until the next cycle. If the scheduler issues the 3rd insn the first, 6634the processor could issue all 3 insns per cycle. 6635 6636Actually this code demonstrates advantages of the automaton based 6637pipeline hazard recognizer. We try quickly and easy many insn 6638schedules to choose the best one. 6639 6640The default is no multipass scheduling. 6641@end deftypefn 6642 6643@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx_insn *@var{insn}, int @var{ready_index}) 6644 6645This hook controls what insns from the ready insn queue will be 6646considered for the multipass insn scheduling. If the hook returns 6647zero for @var{insn}, the insn will be considered in multipass scheduling. 6648Positive return values will remove @var{insn} from consideration on 6649the current round of multipass scheduling. 6650Negative return values will remove @var{insn} from consideration for given 6651number of cycles. 6652Backends should be careful about returning non-zero for highest priority 6653instruction at position 0 in the ready list. @var{ready_index} is passed 6654to allow backends make correct judgements. 6655 6656The default is that any ready insns can be chosen to be issued. 6657@end deftypefn 6658 6659@deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN (void *@var{data}, signed char *@var{ready_try}, int @var{n_ready}, bool @var{first_cycle_insn_p}) 6660This hook prepares the target backend for a new round of multipass 6661scheduling. 6662@end deftypefn 6663 6664@deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE (void *@var{data}, signed char *@var{ready_try}, int @var{n_ready}, rtx_insn *@var{insn}, const void *@var{prev_data}) 6665This hook is called when multipass scheduling evaluates instruction INSN. 6666@end deftypefn 6667 6668@deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, signed char *@var{ready_try}, int @var{n_ready}) 6669This is called when multipass scheduling backtracks from evaluation of 6670an instruction. 6671@end deftypefn 6672 6673@deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data}) 6674This hook notifies the target about the result of the concluded current 6675round of multipass scheduling. 6676@end deftypefn 6677 6678@deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data}) 6679This hook initializes target-specific data used in multipass scheduling. 6680@end deftypefn 6681 6682@deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data}) 6683This hook finalizes target-specific data used in multipass scheduling. 6684@end deftypefn 6685 6686@deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *@var{dump}, int @var{verbose}, rtx_insn *@var{insn}, int @var{last_clock}, int @var{clock}, int *@var{sort_p}) 6687This hook is called by the insn scheduler before issuing @var{insn} 6688on cycle @var{clock}. If the hook returns nonzero, 6689@var{insn} is not issued on this processor cycle. Instead, 6690the processor cycle is advanced. If *@var{sort_p} 6691is zero, the insn ready queue is not sorted on the new cycle 6692start as usually. @var{dump} and @var{verbose} specify the file and 6693verbosity level to use for debugging output. 6694@var{last_clock} and @var{clock} are, respectively, the 6695processor cycle on which the previous insn has been issued, 6696and the current processor cycle. 6697@end deftypefn 6698 6699@deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance}) 6700This hook is used to define which dependences are considered costly by 6701the target, so costly that it is not advisable to schedule the insns that 6702are involved in the dependence too close to one another. The parameters 6703to this hook are as follows: The first parameter @var{_dep} is the dependence 6704being evaluated. The second parameter @var{cost} is the cost of the 6705dependence as estimated by the scheduler, and the third 6706parameter @var{distance} is the distance in cycles between the two insns. 6707The hook returns @code{true} if considering the distance between the two 6708insns the dependence between them is considered costly by the target, 6709and @code{false} otherwise. 6710 6711Defining this hook can be useful in multiple-issue out-of-order machines, 6712where (a) it's practically hopeless to predict the actual data/resource 6713delays, however: (b) there's a better chance to predict the actual grouping 6714that will be formed, and (c) correctly emulating the grouping can be very 6715important. In such targets one may want to allow issuing dependent insns 6716closer to one another---i.e., closer than the dependence distance; however, 6717not in cases of ``costly dependences'', which this hooks allows to define. 6718@end deftypefn 6719 6720@deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void) 6721This hook is called by the insn scheduler after emitting a new instruction to 6722the instruction stream. The hook notifies a target backend to extend its 6723per instruction data structures. 6724@end deftypefn 6725 6726@deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void) 6727Return a pointer to a store large enough to hold target scheduling context. 6728@end deftypefn 6729 6730@deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p}) 6731Initialize store pointed to by @var{tc} to hold target scheduling context. 6732It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the 6733beginning of the block. Otherwise, copy the current context into @var{tc}. 6734@end deftypefn 6735 6736@deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc}) 6737Copy target scheduling context pointed to by @var{tc} to the current context. 6738@end deftypefn 6739 6740@deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc}) 6741Deallocate internal data in target scheduling context pointed to by @var{tc}. 6742@end deftypefn 6743 6744@deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc}) 6745Deallocate a store for target scheduling context pointed to by @var{tc}. 6746@end deftypefn 6747 6748@deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx_insn *@var{insn}, unsigned int @var{dep_status}, rtx *@var{new_pat}) 6749This hook is called by the insn scheduler when @var{insn} has only 6750speculative dependencies and therefore can be scheduled speculatively. 6751The hook is used to check if the pattern of @var{insn} has a speculative 6752version and, in case of successful check, to generate that speculative 6753pattern. The hook should return 1, if the instruction has a speculative form, 6754or @minus{}1, if it doesn't. @var{request} describes the type of requested 6755speculation. If the return value equals 1 then @var{new_pat} is assigned 6756the generated speculative pattern. 6757@end deftypefn 6758 6759@deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (unsigned int @var{dep_status}) 6760This hook is called by the insn scheduler during generation of recovery code 6761for @var{insn}. It should return @code{true}, if the corresponding check 6762instruction should branch to recovery code, or @code{false} otherwise. 6763@end deftypefn 6764 6765@deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx_insn *@var{insn}, rtx_insn *@var{label}, unsigned int @var{ds}) 6766This hook is called by the insn scheduler to generate a pattern for recovery 6767check instruction. If @var{mutate_p} is zero, then @var{insn} is a 6768speculative instruction for which the check should be generated. 6769@var{label} is either a label of a basic block, where recovery code should 6770be emitted, or a null pointer, when requested check doesn't branch to 6771recovery code (a simple check). If @var{mutate_p} is nonzero, then 6772a pattern for a branchy check corresponding to a simple check denoted by 6773@var{insn} should be generated. In this case @var{label} can't be null. 6774@end deftypefn 6775 6776@deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info}) 6777This hook is used by the insn scheduler to find out what features should be 6778enabled/used. 6779The structure *@var{spec_info} should be filled in by the target. 6780The structure describes speculation types that can be used in the scheduler. 6781@end deftypefn 6782 6783@deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g}) 6784This hook is called by the swing modulo scheduler to calculate a 6785resource-based lower bound which is based on the resources available in 6786the machine and the resources required by each instruction. The target 6787backend can use @var{g} to calculate such bound. A very simple lower 6788bound will be used in case this hook is not implemented: the total number 6789of instructions divided by the issue rate. 6790@end deftypefn 6791 6792@deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx_insn *@var{insn}, int @var{x}) 6793This hook is called by Haifa Scheduler. It returns true if dispatch scheduling 6794is supported in hardware and the condition specified in the parameter is true. 6795@end deftypefn 6796 6797@deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx_insn *@var{insn}, int @var{x}) 6798This hook is called by Haifa Scheduler. It performs the operation specified 6799in its second parameter. 6800@end deftypefn 6801 6802@deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE 6803True if the processor has an exposed pipeline, which means that not just 6804the order of instructions is important for correctness when scheduling, but 6805also the latencies of operations. 6806@end deftypevr 6807 6808@deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, machine_mode @var{mode}) 6809This hook is called by tree reassociator to determine a level of 6810parallelism required in output calculations chain. 6811@end deftypefn 6812 6813@deftypefn {Target Hook} void TARGET_SCHED_FUSION_PRIORITY (rtx_insn *@var{insn}, int @var{max_pri}, int *@var{fusion_pri}, int *@var{pri}) 6814This hook is called by scheduling fusion pass. It calculates fusion 6815priorities for each instruction passed in by parameter. The priorities 6816are returned via pointer parameters. 6817 6818@var{insn} is the instruction whose priorities need to be calculated. 6819@var{max_pri} is the maximum priority can be returned in any cases. 6820@var{fusion_pri} is the pointer parameter through which @var{insn}'s 6821fusion priority should be calculated and returned. 6822@var{pri} is the pointer parameter through which @var{insn}'s priority 6823should be calculated and returned. 6824 6825Same @var{fusion_pri} should be returned for instructions which should 6826be scheduled together. Different @var{pri} should be returned for 6827instructions with same @var{fusion_pri}. @var{fusion_pri} is the major 6828sort key, @var{pri} is the minor sort key. All instructions will be 6829scheduled according to the two priorities. All priorities calculated 6830should be between 0 (exclusive) and @var{max_pri} (inclusive). To avoid 6831false dependencies, @var{fusion_pri} of instructions which need to be 6832scheduled together should be smaller than @var{fusion_pri} of irrelevant 6833instructions. 6834 6835Given below example: 6836 6837@smallexample 6838 ldr r10, [r1, 4] 6839 add r4, r4, r10 6840 ldr r15, [r2, 8] 6841 sub r5, r5, r15 6842 ldr r11, [r1, 0] 6843 add r4, r4, r11 6844 ldr r16, [r2, 12] 6845 sub r5, r5, r16 6846@end smallexample 6847 6848On targets like ARM/AArch64, the two pairs of consecutive loads should be 6849merged. Since peephole2 pass can't help in this case unless consecutive 6850loads are actually next to each other in instruction flow. That's where 6851this scheduling fusion pass works. This hook calculates priority for each 6852instruction based on its fustion type, like: 6853 6854@smallexample 6855 ldr r10, [r1, 4] ; fusion_pri=99, pri=96 6856 add r4, r4, r10 ; fusion_pri=100, pri=100 6857 ldr r15, [r2, 8] ; fusion_pri=98, pri=92 6858 sub r5, r5, r15 ; fusion_pri=100, pri=100 6859 ldr r11, [r1, 0] ; fusion_pri=99, pri=100 6860 add r4, r4, r11 ; fusion_pri=100, pri=100 6861 ldr r16, [r2, 12] ; fusion_pri=98, pri=88 6862 sub r5, r5, r16 ; fusion_pri=100, pri=100 6863@end smallexample 6864 6865Scheduling fusion pass then sorts all ready to issue instructions according 6866to the priorities. As a result, instructions of same fusion type will be 6867pushed together in instruction flow, like: 6868 6869@smallexample 6870 ldr r11, [r1, 0] 6871 ldr r10, [r1, 4] 6872 ldr r15, [r2, 8] 6873 ldr r16, [r2, 12] 6874 add r4, r4, r10 6875 sub r5, r5, r15 6876 add r4, r4, r11 6877 sub r5, r5, r16 6878@end smallexample 6879 6880Now peephole2 pass can simply merge the two pairs of loads. 6881 6882Since scheduling fusion pass relies on peephole2 to do real fusion 6883work, it is only enabled by default when peephole2 is in effect. 6884 6885This is firstly introduced on ARM/AArch64 targets, please refer to 6886the hook implementation for how different fusion types are supported. 6887@end deftypefn 6888 6889@node Sections 6890@section Dividing the Output into Sections (Texts, Data, @dots{}) 6891@c the above section title is WAY too long. maybe cut the part between 6892@c the (...)? --mew 10feb93 6893 6894An object file is divided into sections containing different types of 6895data. In the most common case, there are three sections: the @dfn{text 6896section}, which holds instructions and read-only data; the @dfn{data 6897section}, which holds initialized writable data; and the @dfn{bss 6898section}, which holds uninitialized data. Some systems have other kinds 6899of sections. 6900 6901@file{varasm.c} provides several well-known sections, such as 6902@code{text_section}, @code{data_section} and @code{bss_section}. 6903The normal way of controlling a @code{@var{foo}_section} variable 6904is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro, 6905as described below. The macros are only read once, when @file{varasm.c} 6906initializes itself, so their values must be run-time constants. 6907They may however depend on command-line flags. 6908 6909@emph{Note:} Some run-time files, such @file{crtstuff.c}, also make 6910use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them 6911to be string literals. 6912 6913Some assemblers require a different string to be written every time a 6914section is selected. If your assembler falls into this category, you 6915should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use 6916@code{get_unnamed_section} to set up the sections. 6917 6918You must always create a @code{text_section}, either by defining 6919@code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section} 6920in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of 6921@code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not 6922create a distinct @code{readonly_data_section}, the default is to 6923reuse @code{text_section}. 6924 6925All the other @file{varasm.c} sections are optional, and are null 6926if the target does not provide them. 6927 6928@defmac TEXT_SECTION_ASM_OP 6929A C expression whose value is a string, including spacing, containing the 6930assembler operation that should precede instructions and read-only data. 6931Normally @code{"\t.text"} is right. 6932@end defmac 6933 6934@defmac HOT_TEXT_SECTION_NAME 6935If defined, a C string constant for the name of the section containing most 6936frequently executed functions of the program. If not defined, GCC will provide 6937a default definition if the target supports named sections. 6938@end defmac 6939 6940@defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME 6941If defined, a C string constant for the name of the section containing unlikely 6942executed functions in the program. 6943@end defmac 6944 6945@defmac DATA_SECTION_ASM_OP 6946A C expression whose value is a string, including spacing, containing the 6947assembler operation to identify the following data as writable initialized 6948data. Normally @code{"\t.data"} is right. 6949@end defmac 6950 6951@defmac SDATA_SECTION_ASM_OP 6952If defined, a C expression whose value is a string, including spacing, 6953containing the assembler operation to identify the following data as 6954initialized, writable small data. 6955@end defmac 6956 6957@defmac READONLY_DATA_SECTION_ASM_OP 6958A C expression whose value is a string, including spacing, containing the 6959assembler operation to identify the following data as read-only initialized 6960data. 6961@end defmac 6962 6963@defmac BSS_SECTION_ASM_OP 6964If defined, a C expression whose value is a string, including spacing, 6965containing the assembler operation to identify the following data as 6966uninitialized global data. If not defined, and 6967@code{ASM_OUTPUT_ALIGNED_BSS} not defined, 6968uninitialized global data will be output in the data section if 6969@option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be 6970used. 6971@end defmac 6972 6973@defmac SBSS_SECTION_ASM_OP 6974If defined, a C expression whose value is a string, including spacing, 6975containing the assembler operation to identify the following data as 6976uninitialized, writable small data. 6977@end defmac 6978 6979@defmac TLS_COMMON_ASM_OP 6980If defined, a C expression whose value is a string containing the 6981assembler operation to identify the following data as thread-local 6982common data. The default is @code{".tls_common"}. 6983@end defmac 6984 6985@defmac TLS_SECTION_ASM_FLAG 6986If defined, a C expression whose value is a character constant 6987containing the flag used to mark a section as a TLS section. The 6988default is @code{'T'}. 6989@end defmac 6990 6991@defmac INIT_SECTION_ASM_OP 6992If defined, a C expression whose value is a string, including spacing, 6993containing the assembler operation to identify the following data as 6994initialization code. If not defined, GCC will assume such a section does 6995not exist. This section has no corresponding @code{init_section} 6996variable; it is used entirely in runtime code. 6997@end defmac 6998 6999@defmac FINI_SECTION_ASM_OP 7000If defined, a C expression whose value is a string, including spacing, 7001containing the assembler operation to identify the following data as 7002finalization code. If not defined, GCC will assume such a section does 7003not exist. This section has no corresponding @code{fini_section} 7004variable; it is used entirely in runtime code. 7005@end defmac 7006 7007@defmac INIT_ARRAY_SECTION_ASM_OP 7008If defined, a C expression whose value is a string, including spacing, 7009containing the assembler operation to identify the following data as 7010part of the @code{.init_array} (or equivalent) section. If not 7011defined, GCC will assume such a section does not exist. Do not define 7012both this macro and @code{INIT_SECTION_ASM_OP}. 7013@end defmac 7014 7015@defmac FINI_ARRAY_SECTION_ASM_OP 7016If defined, a C expression whose value is a string, including spacing, 7017containing the assembler operation to identify the following data as 7018part of the @code{.fini_array} (or equivalent) section. If not 7019defined, GCC will assume such a section does not exist. Do not define 7020both this macro and @code{FINI_SECTION_ASM_OP}. 7021@end defmac 7022 7023@defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function}) 7024If defined, an ASM statement that switches to a different section 7025via @var{section_op}, calls @var{function}, and switches back to 7026the text section. This is used in @file{crtstuff.c} if 7027@code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls 7028to initialization and finalization functions from the init and fini 7029sections. By default, this macro uses a simple function call. Some 7030ports need hand-crafted assembly code to avoid dependencies on 7031registers initialized in the function prologue or to ensure that 7032constant pools don't end up too far way in the text section. 7033@end defmac 7034 7035@defmac TARGET_LIBGCC_SDATA_SECTION 7036If defined, a string which names the section into which small 7037variables defined in crtstuff and libgcc should go. This is useful 7038when the target has options for optimizing access to small data, and 7039you want the crtstuff and libgcc routines to be conservative in what 7040they expect of your application yet liberal in what your application 7041expects. For example, for targets with a @code{.sdata} section (like 7042MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't 7043require small data support from your application, but use this macro 7044to put small data into @code{.sdata} so that your application can 7045access these variables whether it uses small data or not. 7046@end defmac 7047 7048@defmac FORCE_CODE_SECTION_ALIGN 7049If defined, an ASM statement that aligns a code section to some 7050arbitrary boundary. This is used to force all fragments of the 7051@code{.init} and @code{.fini} sections to have to same alignment 7052and thus prevent the linker from having to add any padding. 7053@end defmac 7054 7055@defmac JUMP_TABLES_IN_TEXT_SECTION 7056Define this macro to be an expression with a nonzero value if jump 7057tables (for @code{tablejump} insns) should be output in the text 7058section, along with the assembler instructions. Otherwise, the 7059readonly data section is used. 7060 7061This macro is irrelevant if there is no separate readonly data section. 7062@end defmac 7063 7064@deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void) 7065Define this hook if you need to do something special to set up the 7066@file{varasm.c} sections, or if your target has some special sections 7067of its own that you need to create. 7068 7069GCC calls this hook after processing the command line, but before writing 7070any assembly code, and before calling any of the section-returning hooks 7071described below. 7072@end deftypefn 7073 7074@deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void) 7075Return a mask describing how relocations should be treated when 7076selecting sections. Bit 1 should be set if global relocations 7077should be placed in a read-write section; bit 0 should be set if 7078local relocations should be placed in a read-write section. 7079 7080The default version of this function returns 3 when @option{-fpic} 7081is in effect, and 0 otherwise. The hook is typically redefined 7082when the target cannot support (some kinds of) dynamic relocations 7083in read-only sections even in executables. 7084@end deftypefn 7085 7086@deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align}) 7087Return the section into which @var{exp} should be placed. You can 7088assume that @var{exp} is either a @code{VAR_DECL} node or a constant of 7089some sort. @var{reloc} indicates whether the initial value of @var{exp} 7090requires link-time relocations. Bit 0 is set when variable contains 7091local relocations only, while bit 1 is set for global relocations. 7092@var{align} is the constant alignment in bits. 7093 7094The default version of this function takes care of putting read-only 7095variables in @code{readonly_data_section}. 7096 7097See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}. 7098@end deftypefn 7099 7100@defmac USE_SELECT_SECTION_FOR_FUNCTIONS 7101Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called 7102for @code{FUNCTION_DECL}s as well as for variables and constants. 7103 7104In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the 7105function has been determined to be likely to be called, and nonzero if 7106it is unlikely to be called. 7107@end defmac 7108 7109@deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc}) 7110Build up a unique section name, expressed as a @code{STRING_CST} node, 7111and assign it to @samp{DECL_SECTION_NAME (@var{decl})}. 7112As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether 7113the initial value of @var{exp} requires link-time relocations. 7114 7115The default version of this function appends the symbol name to the 7116ELF section name that would normally be used for the symbol. For 7117example, the function @code{foo} would be placed in @code{.text.foo}. 7118Whatever the actual target object format, this is often good enough. 7119@end deftypefn 7120 7121@deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl}) 7122Return the readonly data section associated with 7123@samp{DECL_SECTION_NAME (@var{decl})}. 7124The default version of this function selects @code{.gnu.linkonce.r.name} if 7125the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name} 7126if function is in @code{.text.name}, and the normal readonly-data section 7127otherwise. 7128@end deftypefn 7129 7130@deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX 7131Usually, the compiler uses the prefix @code{".rodata"} to construct 7132section names for mergeable constant data. Define this macro to override 7133the string if a different section name should be used. 7134@end deftypevr 7135 7136@deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void) 7137Return the section that should be used for transactional memory clone tables. 7138@end deftypefn 7139 7140@deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align}) 7141Return the section into which a constant @var{x}, of mode @var{mode}, 7142should be placed. You can assume that @var{x} is some kind of 7143constant in RTL@. The argument @var{mode} is redundant except in the 7144case of a @code{const_int} rtx. @var{align} is the constant alignment 7145in bits. 7146 7147The default version of this function takes care of putting symbolic 7148constants in @code{flag_pic} mode in @code{data_section} and everything 7149else in @code{readonly_data_section}. 7150@end deftypefn 7151 7152@deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id}) 7153Define this hook if you need to postprocess the assembler name generated 7154by target-independent code. The @var{id} provided to this hook will be 7155the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C, 7156or the mangled name of the @var{decl} in C++). The return value of the 7157hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on 7158your target system. The default implementation of this hook just 7159returns the @var{id} provided. 7160@end deftypefn 7161 7162@deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p}) 7163Define this hook if references to a symbol or a constant must be 7164treated differently depending on something about the variable or 7165function named by the symbol (such as what section it is in). 7166 7167The hook is executed immediately after rtl has been created for 7168@var{decl}, which may be a variable or function declaration or 7169an entry in the constant pool. In either case, @var{rtl} is the 7170rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})} 7171in this hook; that field may not have been initialized yet. 7172 7173In the case of a constant, it is safe to assume that the rtl is 7174a @code{mem} whose address is a @code{symbol_ref}. Most decls 7175will also have this form, but that is not guaranteed. Global 7176register variables, for instance, will have a @code{reg} for their 7177rtl. (Normally the right thing to do with such unusual rtl is 7178leave it alone.) 7179 7180The @var{new_decl_p} argument will be true if this is the first time 7181that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will 7182be false for subsequent invocations, which will happen for duplicate 7183declarations. Whether or not anything must be done for the duplicate 7184declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}. 7185@var{new_decl_p} is always true when the hook is called for a constant. 7186 7187@cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO} 7188The usual thing for this hook to do is to record flags in the 7189@code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}. 7190Historically, the name string was modified if it was necessary to 7191encode more than one bit of information, but this practice is now 7192discouraged; use @code{SYMBOL_REF_FLAGS}. 7193 7194The default definition of this hook, @code{default_encode_section_info} 7195in @file{varasm.c}, sets a number of commonly-useful bits in 7196@code{SYMBOL_REF_FLAGS}. Check whether the default does what you need 7197before overriding it. 7198@end deftypefn 7199 7200@deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name}) 7201Decode @var{name} and return the real name part, sans 7202the characters that @code{TARGET_ENCODE_SECTION_INFO} 7203may have added. 7204@end deftypefn 7205 7206@deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp}) 7207Returns true if @var{exp} should be placed into a ``small data'' section. 7208The default version of this hook always returns false. 7209@end deftypefn 7210 7211@deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION 7212Contains the value true if the target places read-only 7213``small data'' into a separate section. The default value is false. 7214@end deftypevr 7215 7216@deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void) 7217It returns true if target wants profile code emitted before prologue. 7218 7219The default version of this hook use the target macro 7220@code{PROFILE_BEFORE_PROLOGUE}. 7221@end deftypefn 7222 7223@deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp}) 7224Returns true if @var{exp} names an object for which name resolution 7225rules must resolve to the current ``module'' (dynamic shared library 7226or executable image). 7227 7228The default version of this hook implements the name resolution rules 7229for ELF, which has a looser model of global name binding than other 7230currently supported object file formats. 7231@end deftypefn 7232 7233@deftypevr {Target Hook} bool TARGET_HAVE_TLS 7234Contains the value true if the target supports thread-local storage. 7235The default value is false. 7236@end deftypevr 7237 7238 7239@node PIC 7240@section Position Independent Code 7241@cindex position independent code 7242@cindex PIC 7243 7244This section describes macros that help implement generation of position 7245independent code. Simply defining these macros is not enough to 7246generate valid PIC; you must also add support to the hook 7247@code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro 7248@code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You 7249must modify the definition of @samp{movsi} to do something appropriate 7250when the source operand contains a symbolic address. You may also 7251need to alter the handling of switch statements so that they use 7252relative addresses. 7253@c i rearranged the order of the macros above to try to force one of 7254@c them to the next line, to eliminate an overfull hbox. --mew 10feb93 7255 7256@defmac PIC_OFFSET_TABLE_REGNUM 7257The register number of the register used to address a table of static 7258data addresses in memory. In some cases this register is defined by a 7259processor's ``application binary interface'' (ABI)@. When this macro 7260is defined, RTL is generated for this register once, as with the stack 7261pointer and frame pointer registers. If this macro is not defined, it 7262is up to the machine-dependent files to allocate such a register (if 7263necessary). Note that this register must be fixed when in use (e.g.@: 7264when @code{flag_pic} is true). 7265@end defmac 7266 7267@defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED 7268A C expression that is nonzero if the register defined by 7269@code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined, 7270the default is zero. Do not define 7271this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined. 7272@end defmac 7273 7274@defmac LEGITIMATE_PIC_OPERAND_P (@var{x}) 7275A C expression that is nonzero if @var{x} is a legitimate immediate 7276operand on the target machine when generating position independent code. 7277You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not 7278check this. You can also assume @var{flag_pic} is true, so you need not 7279check it either. You need not define this macro if all constants 7280(including @code{SYMBOL_REF}) can be immediate operands when generating 7281position independent code. 7282@end defmac 7283 7284@node Assembler Format 7285@section Defining the Output Assembler Language 7286 7287This section describes macros whose principal purpose is to describe how 7288to write instructions in assembler language---rather than what the 7289instructions do. 7290 7291@menu 7292* File Framework:: Structural information for the assembler file. 7293* Data Output:: Output of constants (numbers, strings, addresses). 7294* Uninitialized Data:: Output of uninitialized variables. 7295* Label Output:: Output and generation of labels. 7296* Initialization:: General principles of initialization 7297 and termination routines. 7298* Macros for Initialization:: 7299 Specific macros that control the handling of 7300 initialization and termination routines. 7301* Instruction Output:: Output of actual instructions. 7302* Dispatch Tables:: Output of jump tables. 7303* Exception Region Output:: Output of exception region code. 7304* Alignment Output:: Pseudo ops for alignment and skipping data. 7305@end menu 7306 7307@node File Framework 7308@subsection The Overall Framework of an Assembler File 7309@cindex assembler format 7310@cindex output of assembler code 7311 7312@c prevent bad page break with this line 7313This describes the overall framework of an assembly file. 7314 7315@findex default_file_start 7316@deftypefn {Target Hook} void TARGET_ASM_FILE_START (void) 7317Output to @code{asm_out_file} any text which the assembler expects to 7318find at the beginning of a file. The default behavior is controlled 7319by two flags, documented below. Unless your target's assembler is 7320quite unusual, if you override the default, you should call 7321@code{default_file_start} at some point in your target hook. This 7322lets other target files rely on these variables. 7323@end deftypefn 7324 7325@deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF 7326If this flag is true, the text of the macro @code{ASM_APP_OFF} will be 7327printed as the very first line in the assembly file, unless 7328@option{-fverbose-asm} is in effect. (If that macro has been defined 7329to the empty string, this variable has no effect.) With the normal 7330definition of @code{ASM_APP_OFF}, the effect is to notify the GNU 7331assembler that it need not bother stripping comments or extra 7332whitespace from its input. This allows it to work a bit faster. 7333 7334The default is false. You should not set it to true unless you have 7335verified that your port does not generate any extra whitespace or 7336comments that will cause GAS to issue errors in NO_APP mode. 7337@end deftypevr 7338 7339@deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE 7340If this flag is true, @code{output_file_directive} will be called 7341for the primary source file, immediately after printing 7342@code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect 7343this to be done. The default is false. 7344@end deftypevr 7345 7346@deftypefn {Target Hook} void TARGET_ASM_FILE_END (void) 7347Output to @code{asm_out_file} any text which the assembler expects 7348to find at the end of a file. The default is to output nothing. 7349@end deftypefn 7350 7351@deftypefun void file_end_indicate_exec_stack () 7352Some systems use a common convention, the @samp{.note.GNU-stack} 7353special section, to indicate whether or not an object file relies on 7354the stack being executable. If your system uses this convention, you 7355should define @code{TARGET_ASM_FILE_END} to this function. If you 7356need to do other things in that hook, have your hook function call 7357this function. 7358@end deftypefun 7359 7360@deftypefn {Target Hook} void TARGET_ASM_LTO_START (void) 7361Output to @code{asm_out_file} any text which the assembler expects 7362to find at the start of an LTO section. The default is to output 7363nothing. 7364@end deftypefn 7365 7366@deftypefn {Target Hook} void TARGET_ASM_LTO_END (void) 7367Output to @code{asm_out_file} any text which the assembler expects 7368to find at the end of an LTO section. The default is to output 7369nothing. 7370@end deftypefn 7371 7372@deftypefn {Target Hook} void TARGET_ASM_CODE_END (void) 7373Output to @code{asm_out_file} any text which is needed before emitting 7374unwind info and debug info at the end of a file. Some targets emit 7375here PIC setup thunks that cannot be emitted at the end of file, 7376because they couldn't have unwind info then. The default is to output 7377nothing. 7378@end deftypefn 7379 7380@defmac ASM_COMMENT_START 7381A C string constant describing how to begin a comment in the target 7382assembler language. The compiler assumes that the comment will end at 7383the end of the line. 7384@end defmac 7385 7386@defmac ASM_APP_ON 7387A C string constant for text to be output before each @code{asm} 7388statement or group of consecutive ones. Normally this is 7389@code{"#APP"}, which is a comment that has no effect on most 7390assemblers but tells the GNU assembler that it must check the lines 7391that follow for all valid assembler constructs. 7392@end defmac 7393 7394@defmac ASM_APP_OFF 7395A C string constant for text to be output after each @code{asm} 7396statement or group of consecutive ones. Normally this is 7397@code{"#NO_APP"}, which tells the GNU assembler to resume making the 7398time-saving assumptions that are valid for ordinary compiler output. 7399@end defmac 7400 7401@defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) 7402A C statement to output COFF information or DWARF debugging information 7403which indicates that filename @var{name} is the current source file to 7404the stdio stream @var{stream}. 7405 7406This macro need not be defined if the standard form of output 7407for the file format in use is appropriate. 7408@end defmac 7409 7410@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name}) 7411Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}. 7412 7413 This target hook need not be defined if the standard form of output for the file format in use is appropriate. 7414@end deftypefn 7415 7416@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_IDENT (const char *@var{name}) 7417Output a string based on @var{name}, suitable for the @samp{#ident} directive, or the equivalent directive or pragma in non-C-family languages. If this hook is not defined, nothing is output for the @samp{#ident} directive. 7418@end deftypefn 7419 7420@defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string}) 7421A C statement to output the string @var{string} to the stdio stream 7422@var{stream}. If you do not call the function @code{output_quoted_string} 7423in your config files, GCC will only call it to output filenames to 7424the assembler source. So you can use it to canonicalize the format 7425of the filename using this macro. 7426@end defmac 7427 7428@deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl}) 7429Output assembly directives to switch to section @var{name}. The section 7430should have attributes as specified by @var{flags}, which is a bit mask 7431of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl} 7432is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which 7433this section is associated. 7434@end deftypefn 7435 7436@deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit}) 7437Return preferred text (sub)section for function @var{decl}. 7438Main purpose of this function is to separate cold, normal and hot 7439functions. @var{startup} is true when function is known to be used only 7440at startup (from static constructors or it is @code{main()}). 7441@var{exit} is true when function is known to be used only at exit 7442(from static destructors). 7443Return NULL if function should go to default text section. 7444@end deftypefn 7445 7446@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold}) 7447Used by the target to emit any assembler directives or additional labels needed when a function is partitioned between different sections. Output should be written to @var{file}. The function decl is available as @var{decl} and the new section is `cold' if @var{new_is_cold} is @code{true}. 7448@end deftypefn 7449 7450@deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS 7451This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}. 7452It must not be modified by command-line option processing. 7453@end deftypevr 7454 7455@anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS} 7456@deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS 7457This flag is true if we can create zeroed data by switching to a BSS 7458section and then using @code{ASM_OUTPUT_SKIP} to allocate the space. 7459This is true on most ELF targets. 7460@end deftypevr 7461 7462@deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc}) 7463Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION} 7464based on a variable or function decl, a section name, and whether or not the 7465declaration's initializer may contain runtime relocations. @var{decl} may be 7466null, in which case read-write data should be assumed. 7467 7468The default version of this function handles choosing code vs data, 7469read-only vs read-write data, and @code{flag_pic}. You should only 7470need to override this if your target has special flags that might be 7471set via @code{__attribute__}. 7472@end deftypefn 7473 7474@deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text}) 7475Provides the target with the ability to record the gcc command line 7476switches that have been passed to the compiler, and options that are 7477enabled. The @var{type} argument specifies what is being recorded. 7478It can take the following values: 7479 7480@table @gcctabopt 7481@item SWITCH_TYPE_PASSED 7482@var{text} is a command line switch that has been set by the user. 7483 7484@item SWITCH_TYPE_ENABLED 7485@var{text} is an option which has been enabled. This might be as a 7486direct result of a command line switch, or because it is enabled by 7487default or because it has been enabled as a side effect of a different 7488command line switch. For example, the @option{-O2} switch enables 7489various different individual optimization passes. 7490 7491@item SWITCH_TYPE_DESCRIPTIVE 7492@var{text} is either NULL or some descriptive text which should be 7493ignored. If @var{text} is NULL then it is being used to warn the 7494target hook that either recording is starting or ending. The first 7495time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the 7496warning is for start up and the second time the warning is for 7497wind down. This feature is to allow the target hook to make any 7498necessary preparations before it starts to record switches and to 7499perform any necessary tidying up after it has finished recording 7500switches. 7501 7502@item SWITCH_TYPE_LINE_START 7503This option can be ignored by this target hook. 7504 7505@item SWITCH_TYPE_LINE_END 7506This option can be ignored by this target hook. 7507@end table 7508 7509The hook's return value must be zero. Other return values may be 7510supported in the future. 7511 7512By default this hook is set to NULL, but an example implementation is 7513provided for ELF based targets. Called @var{elf_record_gcc_switches}, 7514it records the switches as ASCII text inside a new, string mergeable 7515section in the assembler output file. The name of the new section is 7516provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target 7517hook. 7518@end deftypefn 7519 7520@deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION 7521This is the name of the section that will be created by the example 7522ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target 7523hook. 7524@end deftypevr 7525 7526@need 2000 7527@node Data Output 7528@subsection Output of Data 7529 7530 7531@deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP 7532@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP 7533@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP 7534@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP 7535@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP 7536@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP 7537@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP 7538@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP 7539@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP 7540These hooks specify assembly directives for creating certain kinds 7541of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a 7542byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an 7543aligned two-byte object, and so on. Any of the hooks may be 7544@code{NULL}, indicating that no suitable directive is available. 7545 7546The compiler will print these strings at the start of a new line, 7547followed immediately by the object's initial value. In most cases, 7548the string should contain a tab, a pseudo-op, and then another tab. 7549@end deftypevr 7550 7551@deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p}) 7552The @code{assemble_integer} function uses this hook to output an 7553integer object. @var{x} is the object's value, @var{size} is its size 7554in bytes and @var{aligned_p} indicates whether it is aligned. The 7555function should return @code{true} if it was able to output the 7556object. If it returns false, @code{assemble_integer} will try to 7557split the object into smaller parts. 7558 7559The default implementation of this hook will use the 7560@code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false} 7561when the relevant string is @code{NULL}. 7562@end deftypefn 7563 7564@deftypefn {Target Hook} void TARGET_ASM_DECL_END (void) 7565Define this hook if the target assembler requires a special marker to 7566terminate an initialized variable declaration. 7567@end deftypefn 7568 7569@deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x}) 7570A target hook to recognize @var{rtx} patterns that @code{output_addr_const} 7571can't deal with, and output assembly code to @var{file} corresponding to 7572the pattern @var{x}. This may be used to allow machine-dependent 7573@code{UNSPEC}s to appear within constants. 7574 7575If target hook fails to recognize a pattern, it must return @code{false}, 7576so that a standard error message is printed. If it prints an error message 7577itself, by calling, for example, @code{output_operand_lossage}, it may just 7578return @code{true}. 7579@end deftypefn 7580 7581@defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len}) 7582A C statement to output to the stdio stream @var{stream} an assembler 7583instruction to assemble a string constant containing the @var{len} 7584bytes at @var{ptr}. @var{ptr} will be a C expression of type 7585@code{char *} and @var{len} a C expression of type @code{int}. 7586 7587If the assembler has a @code{.ascii} pseudo-op as found in the 7588Berkeley Unix assembler, do not define the macro 7589@code{ASM_OUTPUT_ASCII}. 7590@end defmac 7591 7592@defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n}) 7593A C statement to output word @var{n} of a function descriptor for 7594@var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS} 7595is defined, and is otherwise unused. 7596@end defmac 7597 7598@defmac CONSTANT_POOL_BEFORE_FUNCTION 7599You may define this macro as a C expression. You should define the 7600expression to have a nonzero value if GCC should output the constant 7601pool for a function before the code for the function, or a zero value if 7602GCC should output the constant pool after the function. If you do 7603not define this macro, the usual case, GCC will output the constant 7604pool before the function. 7605@end defmac 7606 7607@defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size}) 7608A C statement to output assembler commands to define the start of the 7609constant pool for a function. @var{funname} is a string giving 7610the name of the function. Should the return type of the function 7611be required, it can be obtained via @var{fundecl}. @var{size} 7612is the size, in bytes, of the constant pool that will be written 7613immediately after this call. 7614 7615If no constant-pool prefix is required, the usual case, this macro need 7616not be defined. 7617@end defmac 7618 7619@defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto}) 7620A C statement (with or without semicolon) to output a constant in the 7621constant pool, if it needs special treatment. (This macro need not do 7622anything for RTL expressions that can be output normally.) 7623 7624The argument @var{file} is the standard I/O stream to output the 7625assembler code on. @var{x} is the RTL expression for the constant to 7626output, and @var{mode} is the machine mode (in case @var{x} is a 7627@samp{const_int}). @var{align} is the required alignment for the value 7628@var{x}; you should output an assembler directive to force this much 7629alignment. 7630 7631The argument @var{labelno} is a number to use in an internal label for 7632the address of this pool entry. The definition of this macro is 7633responsible for outputting the label definition at the proper place. 7634Here is how to do this: 7635 7636@smallexample 7637@code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno}); 7638@end smallexample 7639 7640When you output a pool entry specially, you should end with a 7641@code{goto} to the label @var{jumpto}. This will prevent the same pool 7642entry from being output a second time in the usual manner. 7643 7644You need not define this macro if it would do nothing. 7645@end defmac 7646 7647@defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size}) 7648A C statement to output assembler commands to at the end of the constant 7649pool for a function. @var{funname} is a string giving the name of the 7650function. Should the return type of the function be required, you can 7651obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the 7652constant pool that GCC wrote immediately before this call. 7653 7654If no constant-pool epilogue is required, the usual case, you need not 7655define this macro. 7656@end defmac 7657 7658@defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR}) 7659Define this macro as a C expression which is nonzero if @var{C} is 7660used as a logical line separator by the assembler. @var{STR} points 7661to the position in the string where @var{C} was found; this can be used if 7662a line separator uses multiple characters. 7663 7664If you do not define this macro, the default is that only 7665the character @samp{;} is treated as a logical line separator. 7666@end defmac 7667 7668@deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN 7669@deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN 7670These target hooks are C string constants, describing the syntax in the 7671assembler for grouping arithmetic expressions. If not overridden, they 7672default to normal parentheses, which is correct for most assemblers. 7673@end deftypevr 7674 7675These macros are provided by @file{real.h} for writing the definitions 7676of @code{ASM_OUTPUT_DOUBLE} and the like: 7677 7678@defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l}) 7679@defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l}) 7680@defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l}) 7681@defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l}) 7682@defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l}) 7683@defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l}) 7684These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the 7685target's floating point representation, and store its bit pattern in 7686the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and 7687@code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a 7688simple @code{long int}. For the others, it should be an array of 7689@code{long int}. The number of elements in this array is determined 7690by the size of the desired target floating point data type: 32 bits of 7691it go in each @code{long int} array element. Each array element holds 769232 bits of the result, even if @code{long int} is wider than 32 bits 7693on the host machine. 7694 7695The array element values are designed so that you can print them out 7696using @code{fprintf} in the order they should appear in the target 7697machine's memory. 7698@end defmac 7699 7700@node Uninitialized Data 7701@subsection Output of Uninitialized Variables 7702 7703Each of the macros in this section is used to do the whole job of 7704outputting a single uninitialized variable. 7705 7706@defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) 7707A C statement (sans semicolon) to output to the stdio stream 7708@var{stream} the assembler definition of a common-label named 7709@var{name} whose size is @var{size} bytes. The variable @var{rounded} 7710is the size rounded up to whatever alignment the caller wants. It is 7711possible that @var{size} may be zero, for instance if a struct with no 7712other member than a zero-length array is defined. In this case, the 7713backend must output a symbol definition that allocates at least one 7714byte, both so that the address of the resulting object does not compare 7715equal to any other, and because some object formats cannot even express 7716the concept of a zero-sized common symbol, as that is how they represent 7717an ordinary undefined external. 7718 7719Use the expression @code{assemble_name (@var{stream}, @var{name})} to 7720output the name itself; before and after that, output the additional 7721assembler syntax for defining the name, and a newline. 7722 7723This macro controls how the assembler definitions of uninitialized 7724common global variables are output. 7725@end defmac 7726 7727@defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment}) 7728Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a 7729separate, explicit argument. If you define this macro, it is used in 7730place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in 7731handling the required alignment of the variable. The alignment is specified 7732as the number of bits. 7733@end defmac 7734 7735@defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 7736Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the 7737variable to be output, if there is one, or @code{NULL_TREE} if there 7738is no corresponding variable. If you define this macro, GCC will use it 7739in place of both @code{ASM_OUTPUT_COMMON} and 7740@code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see 7741the variable's decl in order to chose what to output. 7742@end defmac 7743 7744@defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 7745A C statement (sans semicolon) to output to the stdio stream 7746@var{stream} the assembler definition of uninitialized global @var{decl} named 7747@var{name} whose size is @var{size} bytes. The variable @var{alignment} 7748is the alignment specified as the number of bits. 7749 7750Try to use function @code{asm_output_aligned_bss} defined in file 7751@file{varasm.c} when defining this macro. If unable, use the expression 7752@code{assemble_name (@var{stream}, @var{name})} to output the name itself; 7753before and after that, output the additional assembler syntax for defining 7754the name, and a newline. 7755 7756There are two ways of handling global BSS@. One is to define this macro. 7757The other is to have @code{TARGET_ASM_SELECT_SECTION} return a 7758switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}). 7759You do not need to do both. 7760 7761Some languages do not have @code{common} data, and require a 7762non-common form of global BSS in order to handle uninitialized globals 7763efficiently. C++ is one example of this. However, if the target does 7764not support global BSS, the front end may choose to make globals 7765common in order to save space in the object file. 7766@end defmac 7767 7768@defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) 7769A C statement (sans semicolon) to output to the stdio stream 7770@var{stream} the assembler definition of a local-common-label named 7771@var{name} whose size is @var{size} bytes. The variable @var{rounded} 7772is the size rounded up to whatever alignment the caller wants. 7773 7774Use the expression @code{assemble_name (@var{stream}, @var{name})} to 7775output the name itself; before and after that, output the additional 7776assembler syntax for defining the name, and a newline. 7777 7778This macro controls how the assembler definitions of uninitialized 7779static variables are output. 7780@end defmac 7781 7782@defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment}) 7783Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a 7784separate, explicit argument. If you define this macro, it is used in 7785place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in 7786handling the required alignment of the variable. The alignment is specified 7787as the number of bits. 7788@end defmac 7789 7790@defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 7791Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the 7792variable to be output, if there is one, or @code{NULL_TREE} if there 7793is no corresponding variable. If you define this macro, GCC will use it 7794in place of both @code{ASM_OUTPUT_DECL} and 7795@code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see 7796the variable's decl in order to chose what to output. 7797@end defmac 7798 7799@node Label Output 7800@subsection Output and Generation of Labels 7801 7802@c prevent bad page break with this line 7803This is about outputting labels. 7804 7805@findex assemble_name 7806@defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name}) 7807A C statement (sans semicolon) to output to the stdio stream 7808@var{stream} the assembler definition of a label named @var{name}. 7809Use the expression @code{assemble_name (@var{stream}, @var{name})} to 7810output the name itself; before and after that, output the additional 7811assembler syntax for defining the name, and a newline. A default 7812definition of this macro is provided which is correct for most systems. 7813@end defmac 7814 7815@defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl}) 7816A C statement (sans semicolon) to output to the stdio stream 7817@var{stream} the assembler definition of a label named @var{name} of 7818a function. 7819Use the expression @code{assemble_name (@var{stream}, @var{name})} to 7820output the name itself; before and after that, output the additional 7821assembler syntax for defining the name, and a newline. A default 7822definition of this macro is provided which is correct for most systems. 7823 7824If this macro is not defined, then the function name is defined in the 7825usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 7826@end defmac 7827 7828@findex assemble_name_raw 7829@defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name}) 7830Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known 7831to refer to a compiler-generated label. The default definition uses 7832@code{assemble_name_raw}, which is like @code{assemble_name} except 7833that it is more efficient. 7834@end defmac 7835 7836@defmac SIZE_ASM_OP 7837A C string containing the appropriate assembler directive to specify the 7838size of a symbol, without any arguments. On systems that use ELF, the 7839default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other 7840systems, the default is not to define this macro. 7841 7842Define this macro only if it is correct to use the default definitions 7843of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE} 7844for your system. If you need your own custom definitions of those 7845macros, or if you do not need explicit symbol sizes at all, do not 7846define this macro. 7847@end defmac 7848 7849@defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size}) 7850A C statement (sans semicolon) to output to the stdio stream 7851@var{stream} a directive telling the assembler that the size of the 7852symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}. 7853If you define @code{SIZE_ASM_OP}, a default definition of this macro is 7854provided. 7855@end defmac 7856 7857@defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name}) 7858A C statement (sans semicolon) to output to the stdio stream 7859@var{stream} a directive telling the assembler to calculate the size of 7860the symbol @var{name} by subtracting its address from the current 7861address. 7862 7863If you define @code{SIZE_ASM_OP}, a default definition of this macro is 7864provided. The default assumes that the assembler recognizes a special 7865@samp{.} symbol as referring to the current address, and can calculate 7866the difference between this and another symbol. If your assembler does 7867not recognize @samp{.} or cannot do calculations with it, you will need 7868to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique. 7869@end defmac 7870 7871@defmac NO_DOLLAR_IN_LABEL 7872Define this macro if the assembler does not accept the character 7873@samp{$} in label names. By default constructors and destructors in 7874G++ have @samp{$} in the identifiers. If this macro is defined, 7875@samp{.} is used instead. 7876@end defmac 7877 7878@defmac NO_DOT_IN_LABEL 7879Define this macro if the assembler does not accept the character 7880@samp{.} in label names. By default constructors and destructors in G++ 7881have names that use @samp{.}. If this macro is defined, these names 7882are rewritten to avoid @samp{.}. 7883@end defmac 7884 7885@defmac TYPE_ASM_OP 7886A C string containing the appropriate assembler directive to specify the 7887type of a symbol, without any arguments. On systems that use ELF, the 7888default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other 7889systems, the default is not to define this macro. 7890 7891Define this macro only if it is correct to use the default definition of 7892@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own 7893custom definition of this macro, or if you do not need explicit symbol 7894types at all, do not define this macro. 7895@end defmac 7896 7897@defmac TYPE_OPERAND_FMT 7898A C string which specifies (using @code{printf} syntax) the format of 7899the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the 7900default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems, 7901the default is not to define this macro. 7902 7903Define this macro only if it is correct to use the default definition of 7904@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own 7905custom definition of this macro, or if you do not need explicit symbol 7906types at all, do not define this macro. 7907@end defmac 7908 7909@defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type}) 7910A C statement (sans semicolon) to output to the stdio stream 7911@var{stream} a directive telling the assembler that the type of the 7912symbol @var{name} is @var{type}. @var{type} is a C string; currently, 7913that string is always either @samp{"function"} or @samp{"object"}, but 7914you should not count on this. 7915 7916If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default 7917definition of this macro is provided. 7918@end defmac 7919 7920@defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) 7921A C statement (sans semicolon) to output to the stdio stream 7922@var{stream} any text necessary for declaring the name @var{name} of a 7923function which is being defined. This macro is responsible for 7924outputting the label definition (perhaps using 7925@code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the 7926@code{FUNCTION_DECL} tree node representing the function. 7927 7928If this macro is not defined, then the function name is defined in the 7929usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}). 7930 7931You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition 7932of this macro. 7933@end defmac 7934 7935@defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) 7936A C statement (sans semicolon) to output to the stdio stream 7937@var{stream} any text necessary for declaring the size of a function 7938which is being defined. The argument @var{name} is the name of the 7939function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node 7940representing the function. 7941 7942If this macro is not defined, then the function size is not defined. 7943 7944You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition 7945of this macro. 7946@end defmac 7947 7948@defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl}) 7949A C statement (sans semicolon) to output to the stdio stream 7950@var{stream} any text necessary for declaring the name @var{name} of an 7951initialized variable which is being defined. This macro must output the 7952label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument 7953@var{decl} is the @code{VAR_DECL} tree node representing the variable. 7954 7955If this macro is not defined, then the variable name is defined in the 7956usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 7957 7958You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or 7959@code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro. 7960@end defmac 7961 7962@deftypefn {Target Hook} void TARGET_ASM_DECLARE_CONSTANT_NAME (FILE *@var{file}, const char *@var{name}, const_tree @var{expr}, HOST_WIDE_INT @var{size}) 7963A target hook to output to the stdio stream @var{file} any text necessary 7964for declaring the name @var{name} of a constant which is being defined. This 7965target hook is responsible for outputting the label definition (perhaps using 7966@code{assemble_label}). The argument @var{exp} is the value of the constant, 7967and @var{size} is the size of the constant in bytes. The @var{name} 7968will be an internal label. 7969 7970The default version of this target hook, define the @var{name} in the 7971usual manner as a label (by means of @code{assemble_label}). 7972 7973You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook. 7974@end deftypefn 7975 7976@defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name}) 7977A C statement (sans semicolon) to output to the stdio stream 7978@var{stream} any text necessary for claiming a register @var{regno} 7979for a global variable @var{decl} with name @var{name}. 7980 7981If you don't define this macro, that is equivalent to defining it to do 7982nothing. 7983@end defmac 7984 7985@defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend}) 7986A C statement (sans semicolon) to finish up declaring a variable name 7987once the compiler has processed its initializer fully and thus has had a 7988chance to determine the size of an array when controlled by an 7989initializer. This is used on systems where it's necessary to declare 7990something about the size of the object. 7991 7992If you don't define this macro, that is equivalent to defining it to do 7993nothing. 7994 7995You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or 7996@code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro. 7997@end defmac 7998 7999@deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name}) 8000This target hook is a function to output to the stdio stream 8001@var{stream} some commands that will make the label @var{name} global; 8002that is, available for reference from other files. 8003 8004The default implementation relies on a proper definition of 8005@code{GLOBAL_ASM_OP}. 8006@end deftypefn 8007 8008@deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl}) 8009This target hook is a function to output to the stdio stream 8010@var{stream} some commands that will make the name associated with @var{decl} 8011global; that is, available for reference from other files. 8012 8013The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook. 8014@end deftypefn 8015 8016@deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_UNDEFINED_DECL (FILE *@var{stream}, const char *@var{name}, const_tree @var{decl}) 8017This target hook is a function to output to the stdio stream 8018@var{stream} some commands that will declare the name associated with 8019@var{decl} which is not defined in the current translation unit. Most 8020assemblers do not require anything to be output in this case. 8021@end deftypefn 8022 8023@defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name}) 8024A C statement (sans semicolon) to output to the stdio stream 8025@var{stream} some commands that will make the label @var{name} weak; 8026that is, available for reference from other files but only used if 8027no other definition is available. Use the expression 8028@code{assemble_name (@var{stream}, @var{name})} to output the name 8029itself; before and after that, output the additional assembler syntax 8030for making that name weak, and a newline. 8031 8032If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not 8033support weak symbols and you should not define the @code{SUPPORTS_WEAK} 8034macro. 8035@end defmac 8036 8037@defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value}) 8038Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and 8039@code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function 8040or variable decl. If @var{value} is not @code{NULL}, this C statement 8041should output to the stdio stream @var{stream} assembler code which 8042defines (equates) the weak symbol @var{name} to have the value 8043@var{value}. If @var{value} is @code{NULL}, it should output commands 8044to make @var{name} weak. 8045@end defmac 8046 8047@defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value}) 8048Outputs a directive that enables @var{name} to be used to refer to 8049symbol @var{value} with weak-symbol semantics. @code{decl} is the 8050declaration of @code{name}. 8051@end defmac 8052 8053@defmac SUPPORTS_WEAK 8054A preprocessor constant expression which evaluates to true if the target 8055supports weak symbols. 8056 8057If you don't define this macro, @file{defaults.h} provides a default 8058definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL} 8059is defined, the default definition is @samp{1}; otherwise, it is @samp{0}. 8060@end defmac 8061 8062@defmac TARGET_SUPPORTS_WEAK 8063A C expression which evaluates to true if the target supports weak symbols. 8064 8065If you don't define this macro, @file{defaults.h} provides a default 8066definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define 8067this macro if you want to control weak symbol support with a compiler 8068flag such as @option{-melf}. 8069@end defmac 8070 8071@defmac MAKE_DECL_ONE_ONLY (@var{decl}) 8072A C statement (sans semicolon) to mark @var{decl} to be emitted as a 8073public symbol such that extra copies in multiple translation units will 8074be discarded by the linker. Define this macro if your object file 8075format provides support for this concept, such as the @samp{COMDAT} 8076section flags in the Microsoft Windows PE/COFF format, and this support 8077requires changes to @var{decl}, such as putting it in a separate section. 8078@end defmac 8079 8080@defmac SUPPORTS_ONE_ONLY 8081A C expression which evaluates to true if the target supports one-only 8082semantics. 8083 8084If you don't define this macro, @file{varasm.c} provides a default 8085definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default 8086definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if 8087you want to control one-only symbol support with a compiler flag, or if 8088setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to 8089be emitted as one-only. 8090@end defmac 8091 8092@deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility}) 8093This target hook is a function to output to @var{asm_out_file} some 8094commands that will make the symbol(s) associated with @var{decl} have 8095hidden, protected or internal visibility as specified by @var{visibility}. 8096@end deftypefn 8097 8098@defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC 8099A C expression that evaluates to true if the target's linker expects 8100that weak symbols do not appear in a static archive's table of contents. 8101The default is @code{0}. 8102 8103Leaving weak symbols out of an archive's table of contents means that, 8104if a symbol will only have a definition in one translation unit and 8105will have undefined references from other translation units, that 8106symbol should not be weak. Defining this macro to be nonzero will 8107thus have the effect that certain symbols that would normally be weak 8108(explicit template instantiations, and vtables for polymorphic classes 8109with noninline key methods) will instead be nonweak. 8110 8111The C++ ABI requires this macro to be zero. Define this macro for 8112targets where full C++ ABI compliance is impossible and where linker 8113restrictions require weak symbols to be left out of a static archive's 8114table of contents. 8115@end defmac 8116 8117@defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name}) 8118A C statement (sans semicolon) to output to the stdio stream 8119@var{stream} any text necessary for declaring the name of an external 8120symbol named @var{name} which is referenced in this compilation but 8121not defined. The value of @var{decl} is the tree node for the 8122declaration. 8123 8124This macro need not be defined if it does not need to output anything. 8125The GNU assembler and most Unix assemblers don't require anything. 8126@end defmac 8127 8128@deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref}) 8129This target hook is a function to output to @var{asm_out_file} an assembler 8130pseudo-op to declare a library function name external. The name of the 8131library function is given by @var{symref}, which is a @code{symbol_ref}. 8132@end deftypefn 8133 8134@deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol}) 8135This target hook is a function to output to @var{asm_out_file} an assembler 8136directive to annotate @var{symbol} as used. The Darwin target uses the 8137.no_dead_code_strip directive. 8138@end deftypefn 8139 8140@defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name}) 8141A C statement (sans semicolon) to output to the stdio stream 8142@var{stream} a reference in assembler syntax to a label named 8143@var{name}. This should add @samp{_} to the front of the name, if that 8144is customary on your operating system, as it is in most Berkeley Unix 8145systems. This macro is used in @code{assemble_name}. 8146@end defmac 8147 8148@deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name}) 8149Given a symbol @var{name}, perform same mangling as @code{varasm.c}'s @code{assemble_name}, but in memory rather than to a file stream, returning result as an @code{IDENTIFIER_NODE}. Required for correct LTO symtabs. The default implementation calls the @code{TARGET_STRIP_NAME_ENCODING} hook and then prepends the @code{USER_LABEL_PREFIX}, if any. 8150@end deftypefn 8151 8152@defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym}) 8153A C statement (sans semicolon) to output a reference to 8154@code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name} 8155will be used to output the name of the symbol. This macro may be used 8156to modify the way a symbol is referenced depending on information 8157encoded by @code{TARGET_ENCODE_SECTION_INFO}. 8158@end defmac 8159 8160@defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf}) 8161A C statement (sans semicolon) to output a reference to @var{buf}, the 8162result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined, 8163@code{assemble_name} will be used to output the name of the symbol. 8164This macro is not used by @code{output_asm_label}, or the @code{%l} 8165specifier that calls it; the intention is that this macro should be set 8166when it is necessary to output a label differently when its address is 8167being taken. 8168@end defmac 8169 8170@deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno}) 8171A function to output to the stdio stream @var{stream} a label whose 8172name is made from the string @var{prefix} and the number @var{labelno}. 8173 8174It is absolutely essential that these labels be distinct from the labels 8175used for user-level functions and variables. Otherwise, certain programs 8176will have name conflicts with internal labels. 8177 8178It is desirable to exclude internal labels from the symbol table of the 8179object file. Most assemblers have a naming convention for labels that 8180should be excluded; on many systems, the letter @samp{L} at the 8181beginning of a label has this effect. You should find out what 8182convention your system uses, and follow it. 8183 8184The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}. 8185@end deftypefn 8186 8187@defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num}) 8188A C statement to output to the stdio stream @var{stream} a debug info 8189label whose name is made from the string @var{prefix} and the number 8190@var{num}. This is useful for VLIW targets, where debug info labels 8191may need to be treated differently than branch target labels. On some 8192systems, branch target labels must be at the beginning of instruction 8193bundles, but debug info labels can occur in the middle of instruction 8194bundles. 8195 8196If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be 8197used. 8198@end defmac 8199 8200@defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num}) 8201A C statement to store into the string @var{string} a label whose name 8202is made from the string @var{prefix} and the number @var{num}. 8203 8204This string, when output subsequently by @code{assemble_name}, should 8205produce the output that @code{(*targetm.asm_out.internal_label)} would produce 8206with the same @var{prefix} and @var{num}. 8207 8208If the string begins with @samp{*}, then @code{assemble_name} will 8209output the rest of the string unchanged. It is often convenient for 8210@code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the 8211string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets 8212to output the string, and may change it. (Of course, 8213@code{ASM_OUTPUT_LABELREF} is also part of your machine description, so 8214you should know what it does on your machine.) 8215@end defmac 8216 8217@defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number}) 8218A C expression to assign to @var{outvar} (which is a variable of type 8219@code{char *}) a newly allocated string made from the string 8220@var{name} and the number @var{number}, with some suitable punctuation 8221added. Use @code{alloca} to get space for the string. 8222 8223The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to 8224produce an assembler label for an internal static variable whose name is 8225@var{name}. Therefore, the string must be such as to result in valid 8226assembler code. The argument @var{number} is different each time this 8227macro is executed; it prevents conflicts between similarly-named 8228internal static variables in different scopes. 8229 8230Ideally this string should not be a valid C identifier, to prevent any 8231conflict with the user's own symbols. Most assemblers allow periods 8232or percent signs in assembler symbols; putting at least one of these 8233between the name and the number will suffice. 8234 8235If this macro is not defined, a default definition will be provided 8236which is correct for most systems. 8237@end defmac 8238 8239@defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value}) 8240A C statement to output to the stdio stream @var{stream} assembler code 8241which defines (equates) the symbol @var{name} to have the value @var{value}. 8242 8243@findex SET_ASM_OP 8244If @code{SET_ASM_OP} is defined, a default definition is provided which is 8245correct for most systems. 8246@end defmac 8247 8248@defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value}) 8249A C statement to output to the stdio stream @var{stream} assembler code 8250which defines (equates) the symbol whose tree node is @var{decl_of_name} 8251to have the value of the tree node @var{decl_of_value}. This macro will 8252be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if 8253the tree nodes are available. 8254 8255@findex SET_ASM_OP 8256If @code{SET_ASM_OP} is defined, a default definition is provided which is 8257correct for most systems. 8258@end defmac 8259 8260@defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value}) 8261A C statement that evaluates to true if the assembler code which defines 8262(equates) the symbol whose tree node is @var{decl_of_name} to have the value 8263of the tree node @var{decl_of_value} should be emitted near the end of the 8264current compilation unit. The default is to not defer output of defines. 8265This macro affects defines output by @samp{ASM_OUTPUT_DEF} and 8266@samp{ASM_OUTPUT_DEF_FROM_DECLS}. 8267@end defmac 8268 8269@defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value}) 8270A C statement to output to the stdio stream @var{stream} assembler code 8271which defines (equates) the weak symbol @var{name} to have the value 8272@var{value}. If @var{value} is @code{NULL}, it defines @var{name} as 8273an undefined weak symbol. 8274 8275Define this macro if the target only supports weak aliases; define 8276@code{ASM_OUTPUT_DEF} instead if possible. 8277@end defmac 8278 8279@defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name}) 8280Define this macro to override the default assembler names used for 8281Objective-C methods. 8282 8283The default name is a unique method number followed by the name of the 8284class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of 8285the category is also included in the assembler name (e.g.@: 8286@samp{_1_Foo_Bar}). 8287 8288These names are safe on most systems, but make debugging difficult since 8289the method's selector is not present in the name. Therefore, particular 8290systems define other ways of computing names. 8291 8292@var{buf} is an expression of type @code{char *} which gives you a 8293buffer in which to store the name; its length is as long as 8294@var{class_name}, @var{cat_name} and @var{sel_name} put together, plus 829550 characters extra. 8296 8297The argument @var{is_inst} specifies whether the method is an instance 8298method or a class method; @var{class_name} is the name of the class; 8299@var{cat_name} is the name of the category (or @code{NULL} if the method is not 8300in a category); and @var{sel_name} is the name of the selector. 8301 8302On systems where the assembler can handle quoted names, you can use this 8303macro to provide more human-readable names. 8304@end defmac 8305 8306@node Initialization 8307@subsection How Initialization Functions Are Handled 8308@cindex initialization routines 8309@cindex termination routines 8310@cindex constructors, output of 8311@cindex destructors, output of 8312 8313The compiled code for certain languages includes @dfn{constructors} 8314(also called @dfn{initialization routines})---functions to initialize 8315data in the program when the program is started. These functions need 8316to be called before the program is ``started''---that is to say, before 8317@code{main} is called. 8318 8319Compiling some languages generates @dfn{destructors} (also called 8320@dfn{termination routines}) that should be called when the program 8321terminates. 8322 8323To make the initialization and termination functions work, the compiler 8324must output something in the assembler code to cause those functions to 8325be called at the appropriate time. When you port the compiler to a new 8326system, you need to specify how to do this. 8327 8328There are two major ways that GCC currently supports the execution of 8329initialization and termination functions. Each way has two variants. 8330Much of the structure is common to all four variations. 8331 8332@findex __CTOR_LIST__ 8333@findex __DTOR_LIST__ 8334The linker must build two lists of these functions---a list of 8335initialization functions, called @code{__CTOR_LIST__}, and a list of 8336termination functions, called @code{__DTOR_LIST__}. 8337 8338Each list always begins with an ignored function pointer (which may hold 83390, @minus{}1, or a count of the function pointers after it, depending on 8340the environment). This is followed by a series of zero or more function 8341pointers to constructors (or destructors), followed by a function 8342pointer containing zero. 8343 8344Depending on the operating system and its executable file format, either 8345@file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup 8346time and exit time. Constructors are called in reverse order of the 8347list; destructors in forward order. 8348 8349The best way to handle static constructors works only for object file 8350formats which provide arbitrarily-named sections. A section is set 8351aside for a list of constructors, and another for a list of destructors. 8352Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each 8353object file that defines an initialization function also puts a word in 8354the constructor section to point to that function. The linker 8355accumulates all these words into one contiguous @samp{.ctors} section. 8356Termination functions are handled similarly. 8357 8358This method will be chosen as the default by @file{target-def.h} if 8359@code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not 8360support arbitrary sections, but does support special designated 8361constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP} 8362and @code{DTORS_SECTION_ASM_OP} to achieve the same effect. 8363 8364When arbitrary sections are available, there are two variants, depending 8365upon how the code in @file{crtstuff.c} is called. On systems that 8366support a @dfn{.init} section which is executed at program startup, 8367parts of @file{crtstuff.c} are compiled into that section. The 8368program is linked by the @command{gcc} driver like this: 8369 8370@smallexample 8371ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o 8372@end smallexample 8373 8374The prologue of a function (@code{__init}) appears in the @code{.init} 8375section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise 8376for the function @code{__fini} in the @dfn{.fini} section. Normally these 8377files are provided by the operating system or by the GNU C library, but 8378are provided by GCC for a few targets. 8379 8380The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets) 8381compiled from @file{crtstuff.c}. They contain, among other things, code 8382fragments within the @code{.init} and @code{.fini} sections that branch 8383to routines in the @code{.text} section. The linker will pull all parts 8384of a section together, which results in a complete @code{__init} function 8385that invokes the routines we need at startup. 8386 8387To use this variant, you must define the @code{INIT_SECTION_ASM_OP} 8388macro properly. 8389 8390If no init section is available, when GCC compiles any function called 8391@code{main} (or more accurately, any function designated as a program 8392entry point by the language front end calling @code{expand_main_function}), 8393it inserts a procedure call to @code{__main} as the first executable code 8394after the function prologue. The @code{__main} function is defined 8395in @file{libgcc2.c} and runs the global constructors. 8396 8397In file formats that don't support arbitrary sections, there are again 8398two variants. In the simplest variant, the GNU linker (GNU @code{ld}) 8399and an `a.out' format must be used. In this case, 8400@code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs} 8401entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__}, 8402and with the address of the void function containing the initialization 8403code as its value. The GNU linker recognizes this as a request to add 8404the value to a @dfn{set}; the values are accumulated, and are eventually 8405placed in the executable as a vector in the format described above, with 8406a leading (ignored) count and a trailing zero element. 8407@code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init 8408section is available, the absence of @code{INIT_SECTION_ASM_OP} causes 8409the compilation of @code{main} to call @code{__main} as above, starting 8410the initialization process. 8411 8412The last variant uses neither arbitrary sections nor the GNU linker. 8413This is preferable when you want to do dynamic linking and when using 8414file formats which the GNU linker does not support, such as `ECOFF'@. In 8415this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and 8416termination functions are recognized simply by their names. This requires 8417an extra program in the linkage step, called @command{collect2}. This program 8418pretends to be the linker, for use with GCC; it does its job by running 8419the ordinary linker, but also arranges to include the vectors of 8420initialization and termination functions. These functions are called 8421via @code{__main} as described above. In order to use this method, 8422@code{use_collect2} must be defined in the target in @file{config.gcc}. 8423 8424@ifinfo 8425The following section describes the specific macros that control and 8426customize the handling of initialization and termination functions. 8427@end ifinfo 8428 8429@node Macros for Initialization 8430@subsection Macros Controlling Initialization Routines 8431 8432Here are the macros that control how the compiler handles initialization 8433and termination functions: 8434 8435@defmac INIT_SECTION_ASM_OP 8436If defined, a C string constant, including spacing, for the assembler 8437operation to identify the following data as initialization code. If not 8438defined, GCC will assume such a section does not exist. When you are 8439using special sections for initialization and termination functions, this 8440macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to 8441run the initialization functions. 8442@end defmac 8443 8444@defmac HAS_INIT_SECTION 8445If defined, @code{main} will not call @code{__main} as described above. 8446This macro should be defined for systems that control start-up code 8447on a symbol-by-symbol basis, such as OSF/1, and should not 8448be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}. 8449@end defmac 8450 8451@defmac LD_INIT_SWITCH 8452If defined, a C string constant for a switch that tells the linker that 8453the following symbol is an initialization routine. 8454@end defmac 8455 8456@defmac LD_FINI_SWITCH 8457If defined, a C string constant for a switch that tells the linker that 8458the following symbol is a finalization routine. 8459@end defmac 8460 8461@defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func}) 8462If defined, a C statement that will write a function that can be 8463automatically called when a shared library is loaded. The function 8464should call @var{func}, which takes no arguments. If not defined, and 8465the object format requires an explicit initialization function, then a 8466function called @code{_GLOBAL__DI} will be generated. 8467 8468This function and the following one are used by collect2 when linking a 8469shared library that needs constructors or destructors, or has DWARF2 8470exception tables embedded in the code. 8471@end defmac 8472 8473@defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func}) 8474If defined, a C statement that will write a function that can be 8475automatically called when a shared library is unloaded. The function 8476should call @var{func}, which takes no arguments. If not defined, and 8477the object format requires an explicit finalization function, then a 8478function called @code{_GLOBAL__DD} will be generated. 8479@end defmac 8480 8481@defmac INVOKE__main 8482If defined, @code{main} will call @code{__main} despite the presence of 8483@code{INIT_SECTION_ASM_OP}. This macro should be defined for systems 8484where the init section is not actually run automatically, but is still 8485useful for collecting the lists of constructors and destructors. 8486@end defmac 8487 8488@defmac SUPPORTS_INIT_PRIORITY 8489If nonzero, the C++ @code{init_priority} attribute is supported and the 8490compiler should emit instructions to control the order of initialization 8491of objects. If zero, the compiler will issue an error message upon 8492encountering an @code{init_priority} attribute. 8493@end defmac 8494 8495@deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS 8496This value is true if the target supports some ``native'' method of 8497collecting constructors and destructors to be run at startup and exit. 8498It is false if we must use @command{collect2}. 8499@end deftypevr 8500 8501@deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority}) 8502If defined, a function that outputs assembler code to arrange to call 8503the function referenced by @var{symbol} at initialization time. 8504 8505Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking 8506no arguments and with no return value. If the target supports initialization 8507priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY}; 8508otherwise it must be @code{DEFAULT_INIT_PRIORITY}. 8509 8510If this macro is not defined by the target, a suitable default will 8511be chosen if (1) the target supports arbitrary section names, (2) the 8512target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2} 8513is not defined. 8514@end deftypefn 8515 8516@deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority}) 8517This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination 8518functions rather than initialization functions. 8519@end deftypefn 8520 8521If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine 8522generated for the generated object file will have static linkage. 8523 8524If your system uses @command{collect2} as the means of processing 8525constructors, then that program normally uses @command{nm} to scan 8526an object file for constructor functions to be called. 8527 8528On certain kinds of systems, you can define this macro to make 8529@command{collect2} work faster (and, in some cases, make it work at all): 8530 8531@defmac OBJECT_FORMAT_COFF 8532Define this macro if the system uses COFF (Common Object File Format) 8533object files, so that @command{collect2} can assume this format and scan 8534object files directly for dynamic constructor/destructor functions. 8535 8536This macro is effective only in a native compiler; @command{collect2} as 8537part of a cross compiler always uses @command{nm} for the target machine. 8538@end defmac 8539 8540@defmac REAL_NM_FILE_NAME 8541Define this macro as a C string constant containing the file name to use 8542to execute @command{nm}. The default is to search the path normally for 8543@command{nm}. 8544@end defmac 8545 8546@defmac NM_FLAGS 8547@command{collect2} calls @command{nm} to scan object files for static 8548constructors and destructors and LTO info. By default, @option{-n} is 8549passed. Define @code{NM_FLAGS} to a C string constant if other options 8550are needed to get the same output format as GNU @command{nm -n} 8551produces. 8552@end defmac 8553 8554If your system supports shared libraries and has a program to list the 8555dynamic dependencies of a given library or executable, you can define 8556these macros to enable support for running initialization and 8557termination functions in shared libraries: 8558 8559@defmac LDD_SUFFIX 8560Define this macro to a C string constant containing the name of the program 8561which lists dynamic dependencies, like @command{ldd} under SunOS 4. 8562@end defmac 8563 8564@defmac PARSE_LDD_OUTPUT (@var{ptr}) 8565Define this macro to be C code that extracts filenames from the output 8566of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable 8567of type @code{char *} that points to the beginning of a line of output 8568from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the 8569code must advance @var{ptr} to the beginning of the filename on that 8570line. Otherwise, it must set @var{ptr} to @code{NULL}. 8571@end defmac 8572 8573@defmac SHLIB_SUFFIX 8574Define this macro to a C string constant containing the default shared 8575library extension of the target (e.g., @samp{".so"}). @command{collect2} 8576strips version information after this suffix when generating global 8577constructor and destructor names. This define is only needed on targets 8578that use @command{collect2} to process constructors and destructors. 8579@end defmac 8580 8581@node Instruction Output 8582@subsection Output of Assembler Instructions 8583 8584@c prevent bad page break with this line 8585This describes assembler instruction output. 8586 8587@defmac REGISTER_NAMES 8588A C initializer containing the assembler's names for the machine 8589registers, each one as a C string constant. This is what translates 8590register numbers in the compiler into assembler language. 8591@end defmac 8592 8593@defmac ADDITIONAL_REGISTER_NAMES 8594If defined, a C initializer for an array of structures containing a name 8595and a register number. This macro defines additional names for hard 8596registers, thus allowing the @code{asm} option in declarations to refer 8597to registers using alternate names. 8598@end defmac 8599 8600@defmac OVERLAPPING_REGISTER_NAMES 8601If defined, a C initializer for an array of structures containing a 8602name, a register number and a count of the number of consecutive 8603machine registers the name overlaps. This macro defines additional 8604names for hard registers, thus allowing the @code{asm} option in 8605declarations to refer to registers using alternate names. Unlike 8606@code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the 8607register name implies multiple underlying registers. 8608 8609This macro should be used when it is important that a clobber in an 8610@code{asm} statement clobbers all the underlying values implied by the 8611register name. For example, on ARM, clobbering the double-precision 8612VFP register ``d0'' implies clobbering both single-precision registers 8613``s0'' and ``s1''. 8614@end defmac 8615 8616@defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr}) 8617Define this macro if you are using an unusual assembler that 8618requires different names for the machine instructions. 8619 8620The definition is a C statement or statements which output an 8621assembler instruction opcode to the stdio stream @var{stream}. The 8622macro-operand @var{ptr} is a variable of type @code{char *} which 8623points to the opcode name in its ``internal'' form---the form that is 8624written in the machine description. The definition should output the 8625opcode name to @var{stream}, performing any translation you desire, and 8626increment the variable @var{ptr} to point at the end of the opcode 8627so that it will not be output twice. 8628 8629In fact, your macro definition may process less than the entire opcode 8630name, or more than the opcode name; but if you want to process text 8631that includes @samp{%}-sequences to substitute operands, you must take 8632care of the substitution yourself. Just be sure to increment 8633@var{ptr} over whatever text should not be output normally. 8634 8635@findex recog_data.operand 8636If you need to look at the operand values, they can be found as the 8637elements of @code{recog_data.operand}. 8638 8639If the macro definition does nothing, the instruction is output 8640in the usual way. 8641@end defmac 8642 8643@defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands}) 8644If defined, a C statement to be executed just prior to the output of 8645assembler code for @var{insn}, to modify the extracted operands so 8646they will be output differently. 8647 8648Here the argument @var{opvec} is the vector containing the operands 8649extracted from @var{insn}, and @var{noperands} is the number of 8650elements of the vector which contain meaningful data for this insn. 8651The contents of this vector are what will be used to convert the insn 8652template into assembler code, so you can change the assembler output 8653by changing the contents of the vector. 8654 8655This macro is useful when various assembler syntaxes share a single 8656file of instruction patterns; by defining this macro differently, you 8657can cause a large class of instructions to be output differently (such 8658as with rearranged operands). Naturally, variations in assembler 8659syntax affecting individual insn patterns ought to be handled by 8660writing conditional output routines in those patterns. 8661 8662If this macro is not defined, it is equivalent to a null statement. 8663@end defmac 8664 8665@deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx_insn *@var{insn}, rtx *@var{opvec}, int @var{noperands}) 8666If defined, this target hook is a function which is executed just after the 8667output of assembler code for @var{insn}, to change the mode of the assembler 8668if necessary. 8669 8670Here the argument @var{opvec} is the vector containing the operands 8671extracted from @var{insn}, and @var{noperands} is the number of 8672elements of the vector which contain meaningful data for this insn. 8673The contents of this vector are what was used to convert the insn 8674template into assembler code, so you can change the assembler mode 8675by checking the contents of the vector. 8676@end deftypefn 8677 8678@defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code}) 8679A C compound statement to output to stdio stream @var{stream} the 8680assembler syntax for an instruction operand @var{x}. @var{x} is an 8681RTL expression. 8682 8683@var{code} is a value that can be used to specify one of several ways 8684of printing the operand. It is used when identical operands must be 8685printed differently depending on the context. @var{code} comes from 8686the @samp{%} specification that was used to request printing of the 8687operand. If the specification was just @samp{%@var{digit}} then 8688@var{code} is 0; if the specification was @samp{%@var{ltr} 8689@var{digit}} then @var{code} is the ASCII code for @var{ltr}. 8690 8691@findex reg_names 8692If @var{x} is a register, this macro should print the register's name. 8693The names can be found in an array @code{reg_names} whose type is 8694@code{char *[]}. @code{reg_names} is initialized from 8695@code{REGISTER_NAMES}. 8696 8697When the machine description has a specification @samp{%@var{punct}} 8698(a @samp{%} followed by a punctuation character), this macro is called 8699with a null pointer for @var{x} and the punctuation character for 8700@var{code}. 8701@end defmac 8702 8703@defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code}) 8704A C expression which evaluates to true if @var{code} is a valid 8705punctuation character for use in the @code{PRINT_OPERAND} macro. If 8706@code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no 8707punctuation characters (except for the standard one, @samp{%}) are used 8708in this way. 8709@end defmac 8710 8711@defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x}) 8712A C compound statement to output to stdio stream @var{stream} the 8713assembler syntax for an instruction operand that is a memory reference 8714whose address is @var{x}. @var{x} is an RTL expression. 8715 8716@cindex @code{TARGET_ENCODE_SECTION_INFO} usage 8717On some machines, the syntax for a symbolic address depends on the 8718section that the address refers to. On these machines, define the hook 8719@code{TARGET_ENCODE_SECTION_INFO} to store the information into the 8720@code{symbol_ref}, and then check for it here. @xref{Assembler 8721Format}. 8722@end defmac 8723 8724@findex dbr_sequence_length 8725@defmac DBR_OUTPUT_SEQEND (@var{file}) 8726A C statement, to be executed after all slot-filler instructions have 8727been output. If necessary, call @code{dbr_sequence_length} to 8728determine the number of slots filled in a sequence (zero if not 8729currently outputting a sequence), to decide how many no-ops to output, 8730or whatever. 8731 8732Don't define this macro if it has nothing to do, but it is helpful in 8733reading assembly output if the extent of the delay sequence is made 8734explicit (e.g.@: with white space). 8735@end defmac 8736 8737@findex final_sequence 8738Note that output routines for instructions with delay slots must be 8739prepared to deal with not being output as part of a sequence 8740(i.e.@: when the scheduling pass is not run, or when no slot fillers could be 8741found.) The variable @code{final_sequence} is null when not 8742processing a sequence, otherwise it contains the @code{sequence} rtx 8743being output. 8744 8745@findex asm_fprintf 8746@defmac REGISTER_PREFIX 8747@defmacx LOCAL_LABEL_PREFIX 8748@defmacx USER_LABEL_PREFIX 8749@defmacx IMMEDIATE_PREFIX 8750If defined, C string expressions to be used for the @samp{%R}, @samp{%L}, 8751@samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see 8752@file{final.c}). These are useful when a single @file{md} file must 8753support multiple assembler formats. In that case, the various @file{tm.h} 8754files can define these macros differently. 8755@end defmac 8756 8757@defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format}) 8758If defined this macro should expand to a series of @code{case} 8759statements which will be parsed inside the @code{switch} statement of 8760the @code{asm_fprintf} function. This allows targets to define extra 8761printf formats which may useful when generating their assembler 8762statements. Note that uppercase letters are reserved for future 8763generic extensions to asm_fprintf, and so are not available to target 8764specific code. The output file is given by the parameter @var{file}. 8765The varargs input pointer is @var{argptr} and the rest of the format 8766string, starting the character after the one that is being switched 8767upon, is pointed to by @var{format}. 8768@end defmac 8769 8770@defmac ASSEMBLER_DIALECT 8771If your target supports multiple dialects of assembler language (such as 8772different opcodes), define this macro as a C expression that gives the 8773numeric index of the assembler language dialect to use, with zero as the 8774first variant. 8775 8776If this macro is defined, you may use constructs of the form 8777@smallexample 8778@samp{@{option0|option1|option2@dots{}@}} 8779@end smallexample 8780@noindent 8781in the output templates of patterns (@pxref{Output Template}) or in the 8782first argument of @code{asm_fprintf}. This construct outputs 8783@samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of 8784@code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters 8785within these strings retain their usual meaning. If there are fewer 8786alternatives within the braces than the value of 8787@code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed 8788to print curly braces or @samp{|} character in assembler output directly, 8789@samp{%@{}, @samp{%@}} and @samp{%|} can be used. 8790 8791If you do not define this macro, the characters @samp{@{}, @samp{|} and 8792@samp{@}} do not have any special meaning when used in templates or 8793operands to @code{asm_fprintf}. 8794 8795Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX}, 8796@code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express 8797the variations in assembler language syntax with that mechanism. Define 8798@code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax 8799if the syntax variant are larger and involve such things as different 8800opcodes or operand order. 8801@end defmac 8802 8803@defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno}) 8804A C expression to output to @var{stream} some assembler code 8805which will push hard register number @var{regno} onto the stack. 8806The code need not be optimal, since this macro is used only when 8807profiling. 8808@end defmac 8809 8810@defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno}) 8811A C expression to output to @var{stream} some assembler code 8812which will pop hard register number @var{regno} off of the stack. 8813The code need not be optimal, since this macro is used only when 8814profiling. 8815@end defmac 8816 8817@node Dispatch Tables 8818@subsection Output of Dispatch Tables 8819 8820@c prevent bad page break with this line 8821This concerns dispatch tables. 8822 8823@cindex dispatch table 8824@defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel}) 8825A C statement to output to the stdio stream @var{stream} an assembler 8826pseudo-instruction to generate a difference between two labels. 8827@var{value} and @var{rel} are the numbers of two internal labels. The 8828definitions of these labels are output using 8829@code{(*targetm.asm_out.internal_label)}, and they must be printed in the same 8830way here. For example, 8831 8832@smallexample 8833fprintf (@var{stream}, "\t.word L%d-L%d\n", 8834 @var{value}, @var{rel}) 8835@end smallexample 8836 8837You must provide this macro on machines where the addresses in a 8838dispatch table are relative to the table's own address. If defined, GCC 8839will also use this macro on all machines when producing PIC@. 8840@var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the 8841mode and flags can be read. 8842@end defmac 8843 8844@defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value}) 8845This macro should be provided on machines where the addresses 8846in a dispatch table are absolute. 8847 8848The definition should be a C statement to output to the stdio stream 8849@var{stream} an assembler pseudo-instruction to generate a reference to 8850a label. @var{value} is the number of an internal label whose 8851definition is output using @code{(*targetm.asm_out.internal_label)}. 8852For example, 8853 8854@smallexample 8855fprintf (@var{stream}, "\t.word L%d\n", @var{value}) 8856@end smallexample 8857@end defmac 8858 8859@defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table}) 8860Define this if the label before a jump-table needs to be output 8861specially. The first three arguments are the same as for 8862@code{(*targetm.asm_out.internal_label)}; the fourth argument is the 8863jump-table which follows (a @code{jump_table_data} containing an 8864@code{addr_vec} or @code{addr_diff_vec}). 8865 8866This feature is used on system V to output a @code{swbeg} statement 8867for the table. 8868 8869If this macro is not defined, these labels are output with 8870@code{(*targetm.asm_out.internal_label)}. 8871@end defmac 8872 8873@defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table}) 8874Define this if something special must be output at the end of a 8875jump-table. The definition should be a C statement to be executed 8876after the assembler code for the table is written. It should write 8877the appropriate code to stdio stream @var{stream}. The argument 8878@var{table} is the jump-table insn, and @var{num} is the label-number 8879of the preceding label. 8880 8881If this macro is not defined, nothing special is output at the end of 8882the jump-table. 8883@end defmac 8884 8885@deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty}) 8886This target hook emits a label at the beginning of each FDE@. It 8887should be defined on targets where FDEs need special labels, and it 8888should write the appropriate label, for the FDE associated with the 8889function declaration @var{decl}, to the stdio stream @var{stream}. 8890The third argument, @var{for_eh}, is a boolean: true if this is for an 8891exception table. The fourth argument, @var{empty}, is a boolean: 8892true if this is a placeholder label for an omitted FDE@. 8893 8894The default is that FDEs are not given nonlocal labels. 8895@end deftypefn 8896 8897@deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream}) 8898This target hook emits a label at the beginning of the exception table. 8899It should be defined on targets where it is desirable for the table 8900to be broken up according to function. 8901 8902The default is that no label is emitted. 8903@end deftypefn 8904 8905@deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality}) 8906If the target implements @code{TARGET_ASM_UNWIND_EMIT}, this hook may be used to emit a directive to install a personality hook into the unwind info. This hook should not be used if dwarf2 unwind info is used. 8907@end deftypefn 8908 8909@deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx_insn *@var{insn}) 8910This target hook emits assembly directives required to unwind the 8911given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO} 8912returns @code{UI_TARGET}. 8913@end deftypefn 8914 8915@deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN 8916True if the @code{TARGET_ASM_UNWIND_EMIT} hook should be called before the assembly for @var{insn} has been emitted, false if the hook should be called afterward. 8917@end deftypevr 8918 8919@node Exception Region Output 8920@subsection Assembler Commands for Exception Regions 8921 8922@c prevent bad page break with this line 8923 8924This describes commands marking the start and the end of an exception 8925region. 8926 8927@defmac EH_FRAME_SECTION_NAME 8928If defined, a C string constant for the name of the section containing 8929exception handling frame unwind information. If not defined, GCC will 8930provide a default definition if the target supports named sections. 8931@file{crtstuff.c} uses this macro to switch to the appropriate section. 8932 8933You should define this symbol if your target supports DWARF 2 frame 8934unwind information and the default definition does not work. 8935@end defmac 8936 8937@defmac EH_FRAME_IN_DATA_SECTION 8938If defined, DWARF 2 frame unwind information will be placed in the 8939data section even though the target supports named sections. This 8940might be necessary, for instance, if the system linker does garbage 8941collection and sections cannot be marked as not to be collected. 8942 8943Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is 8944also defined. 8945@end defmac 8946 8947@defmac EH_TABLES_CAN_BE_READ_ONLY 8948Define this macro to 1 if your target is such that no frame unwind 8949information encoding used with non-PIC code will ever require a 8950runtime relocation, but the linker may not support merging read-only 8951and read-write sections into a single read-write section. 8952@end defmac 8953 8954@defmac MASK_RETURN_ADDR 8955An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so 8956that it does not contain any extraneous set bits in it. 8957@end defmac 8958 8959@defmac DWARF2_UNWIND_INFO 8960Define this macro to 0 if your target supports DWARF 2 frame unwind 8961information, but it does not yet work with exception handling. 8962Otherwise, if your target supports this information (if it defines 8963@code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}), 8964GCC will provide a default definition of 1. 8965@end defmac 8966 8967@deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts}) 8968This hook defines the mechanism that will be used for exception handling 8969by the target. If the target has ABI specified unwind tables, the hook 8970should return @code{UI_TARGET}. If the target is to use the 8971@code{setjmp}/@code{longjmp}-based exception handling scheme, the hook 8972should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind 8973information, the hook should return @code{UI_DWARF2}. 8974 8975A target may, if exceptions are disabled, choose to return @code{UI_NONE}. 8976This may end up simplifying other parts of target-specific code. The 8977default implementation of this hook never returns @code{UI_NONE}. 8978 8979Note that the value returned by this hook should be constant. It should 8980not depend on anything except the command-line switches described by 8981@var{opts}. In particular, the 8982setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor 8983macros and builtin functions related to exception handling are set up 8984depending on this setting. 8985 8986The default implementation of the hook first honors the 8987@option{--enable-sjlj-exceptions} configure option, then 8988@code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If 8989@code{DWARF2_UNWIND_INFO} depends on command-line options, the target 8990must define this hook so that @var{opts} is used correctly. 8991@end deftypefn 8992 8993@deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT 8994This variable should be set to @code{true} if the target ABI requires unwinding 8995tables even when exceptions are not used. It must not be modified by 8996command-line option processing. 8997@end deftypevr 8998 8999@defmac DONT_USE_BUILTIN_SETJMP 9000Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme 9001should use the @code{setjmp}/@code{longjmp} functions from the C library 9002instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery. 9003@end defmac 9004 9005@defmac JMP_BUF_SIZE 9006This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also 9007defined. Define this macro if the default size of @code{jmp_buf} buffer 9008for the @code{setjmp}/@code{longjmp}-based exception handling mechanism 9009is not large enough, or if it is much too large. 9010The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}. 9011@end defmac 9012 9013@defmac DWARF_CIE_DATA_ALIGNMENT 9014This macro need only be defined if the target might save registers in the 9015function prologue at an offset to the stack pointer that is not aligned to 9016@code{UNITS_PER_WORD}. The definition should be the negative minimum 9017alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive 9018minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if 9019the target supports DWARF 2 frame unwind information. 9020@end defmac 9021 9022@deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO 9023Contains the value true if the target should add a zero word onto the 9024end of a Dwarf-2 frame info section when used for exception handling. 9025Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and 9026true otherwise. 9027@end deftypevr 9028 9029@deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg}) 9030Given a register, this hook should return a parallel of registers to 9031represent where to find the register pieces. Define this hook if the 9032register and its mode are represented in Dwarf in non-contiguous 9033locations, or if the register should be represented in more than one 9034register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}. 9035If not defined, the default is to return @code{NULL_RTX}. 9036@end deftypefn 9037 9038@deftypefn {Target Hook} machine_mode TARGET_DWARF_FRAME_REG_MODE (int @var{regno}) 9039Given a register, this hook should return the mode which the 9040corresponding Dwarf frame register should have. This is normally 9041used to return a smaller mode than the raw mode to prevent call 9042clobbered parts of a register altering the frame register size 9043@end deftypefn 9044 9045@deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address}) 9046If some registers are represented in Dwarf-2 unwind information in 9047multiple pieces, define this hook to fill in information about the 9048sizes of those pieces in the table used by the unwinder at runtime. 9049It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after 9050filling in a single size corresponding to each hard register; 9051@var{address} is the address of the table. 9052@end deftypefn 9053 9054@deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym}) 9055This hook is used to output a reference from a frame unwinding table to 9056the type_info object identified by @var{sym}. It should return @code{true} 9057if the reference was output. Returning @code{false} will cause the 9058reference to be output using the normal Dwarf2 routines. 9059@end deftypefn 9060 9061@deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER 9062This flag should be set to @code{true} on targets that use an ARM EABI 9063based unwinding library, and @code{false} on other targets. This effects 9064the format of unwinding tables, and how the unwinder in entered after 9065running a cleanup. The default is @code{false}. 9066@end deftypevr 9067 9068@node Alignment Output 9069@subsection Assembler Commands for Alignment 9070 9071@c prevent bad page break with this line 9072This describes commands for alignment. 9073 9074@defmac JUMP_ALIGN (@var{label}) 9075The alignment (log base 2) to put in front of @var{label}, which is 9076a common destination of jumps and has no fallthru incoming edge. 9077 9078This macro need not be defined if you don't want any special alignment 9079to be done at such a time. Most machine descriptions do not currently 9080define the macro. 9081 9082Unless it's necessary to inspect the @var{label} parameter, it is better 9083to set the variable @var{align_jumps} in the target's 9084@code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's 9085selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation. 9086@end defmac 9087 9088@deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx_insn *@var{label}) 9089The maximum number of bytes to skip before @var{label} when applying 9090@code{JUMP_ALIGN}. This works only if 9091@code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 9092@end deftypefn 9093 9094@defmac LABEL_ALIGN_AFTER_BARRIER (@var{label}) 9095The alignment (log base 2) to put in front of @var{label}, which follows 9096a @code{BARRIER}. 9097 9098This macro need not be defined if you don't want any special alignment 9099to be done at such a time. Most machine descriptions do not currently 9100define the macro. 9101@end defmac 9102 9103@deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx_insn *@var{label}) 9104The maximum number of bytes to skip before @var{label} when applying 9105@code{LABEL_ALIGN_AFTER_BARRIER}. This works only if 9106@code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 9107@end deftypefn 9108 9109@defmac LOOP_ALIGN (@var{label}) 9110The alignment (log base 2) to put in front of @var{label} that heads 9111a frequently executed basic block (usually the header of a loop). 9112 9113This macro need not be defined if you don't want any special alignment 9114to be done at such a time. Most machine descriptions do not currently 9115define the macro. 9116 9117Unless it's necessary to inspect the @var{label} parameter, it is better 9118to set the variable @code{align_loops} in the target's 9119@code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's 9120selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation. 9121@end defmac 9122 9123@deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx_insn *@var{label}) 9124The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to 9125@var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is 9126defined. 9127@end deftypefn 9128 9129@defmac LABEL_ALIGN (@var{label}) 9130The alignment (log base 2) to put in front of @var{label}. 9131If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment, 9132the maximum of the specified values is used. 9133 9134Unless it's necessary to inspect the @var{label} parameter, it is better 9135to set the variable @code{align_labels} in the target's 9136@code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's 9137selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation. 9138@end defmac 9139 9140@deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx_insn *@var{label}) 9141The maximum number of bytes to skip when applying @code{LABEL_ALIGN} 9142to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} 9143is defined. 9144@end deftypefn 9145 9146@defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes}) 9147A C statement to output to the stdio stream @var{stream} an assembler 9148instruction to advance the location counter by @var{nbytes} bytes. 9149Those bytes should be zero when loaded. @var{nbytes} will be a C 9150expression of type @code{unsigned HOST_WIDE_INT}. 9151@end defmac 9152 9153@defmac ASM_NO_SKIP_IN_TEXT 9154Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the 9155text section because it fails to put zeros in the bytes that are skipped. 9156This is true on many Unix systems, where the pseudo--op to skip bytes 9157produces no-op instructions rather than zeros when used in the text 9158section. 9159@end defmac 9160 9161@defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power}) 9162A C statement to output to the stdio stream @var{stream} an assembler 9163command to advance the location counter to a multiple of 2 to the 9164@var{power} bytes. @var{power} will be a C expression of type @code{int}. 9165@end defmac 9166 9167@defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power}) 9168Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used 9169for padding, if necessary. 9170@end defmac 9171 9172@defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip}) 9173A C statement to output to the stdio stream @var{stream} an assembler 9174command to advance the location counter to a multiple of 2 to the 9175@var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to 9176satisfy the alignment request. @var{power} and @var{max_skip} will be 9177a C expression of type @code{int}. 9178@end defmac 9179 9180@need 3000 9181@node Debugging Info 9182@section Controlling Debugging Information Format 9183 9184@c prevent bad page break with this line 9185This describes how to specify debugging information. 9186 9187@menu 9188* All Debuggers:: Macros that affect all debugging formats uniformly. 9189* DBX Options:: Macros enabling specific options in DBX format. 9190* DBX Hooks:: Hook macros for varying DBX format. 9191* File Names and DBX:: Macros controlling output of file names in DBX format. 9192* SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. 9193* VMS Debug:: Macros for VMS debug format. 9194@end menu 9195 9196@node All Debuggers 9197@subsection Macros Affecting All Debugging Formats 9198 9199@c prevent bad page break with this line 9200These macros affect all debugging formats. 9201 9202@defmac DBX_REGISTER_NUMBER (@var{regno}) 9203A C expression that returns the DBX register number for the compiler 9204register number @var{regno}. In the default macro provided, the value 9205of this expression will be @var{regno} itself. But sometimes there are 9206some registers that the compiler knows about and DBX does not, or vice 9207versa. In such cases, some register may need to have one number in the 9208compiler and another for DBX@. 9209 9210If two registers have consecutive numbers inside GCC, and they can be 9211used as a pair to hold a multiword value, then they @emph{must} have 9212consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}. 9213Otherwise, debuggers will be unable to access such a pair, because they 9214expect register pairs to be consecutive in their own numbering scheme. 9215 9216If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that 9217does not preserve register pairs, then what you must do instead is 9218redefine the actual register numbering scheme. 9219@end defmac 9220 9221@defmac DEBUGGER_AUTO_OFFSET (@var{x}) 9222A C expression that returns the integer offset value for an automatic 9223variable having address @var{x} (an RTL expression). The default 9224computation assumes that @var{x} is based on the frame-pointer and 9225gives the offset from the frame-pointer. This is required for targets 9226that produce debugging output for DBX or COFF-style debugging output 9227for SDB and allow the frame-pointer to be eliminated when the 9228@option{-g} options is used. 9229@end defmac 9230 9231@defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x}) 9232A C expression that returns the integer offset value for an argument 9233having address @var{x} (an RTL expression). The nominal offset is 9234@var{offset}. 9235@end defmac 9236 9237@defmac PREFERRED_DEBUGGING_TYPE 9238A C expression that returns the type of debugging output GCC should 9239produce when the user specifies just @option{-g}. Define 9240this if you have arranged for GCC to support more than one format of 9241debugging output. Currently, the allowable values are @code{DBX_DEBUG}, 9242@code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, 9243@code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}. 9244 9245When the user specifies @option{-ggdb}, GCC normally also uses the 9246value of this macro to select the debugging output format, but with two 9247exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the 9248value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is 9249defined, GCC uses @code{DBX_DEBUG}. 9250 9251The value of this macro only affects the default debugging output; the 9252user can always get a specific type of output by using @option{-gstabs}, 9253@option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}. 9254@end defmac 9255 9256@node DBX Options 9257@subsection Specific Options for DBX Output 9258 9259@c prevent bad page break with this line 9260These are specific options for DBX output. 9261 9262@defmac DBX_DEBUGGING_INFO 9263Define this macro if GCC should produce debugging output for DBX 9264in response to the @option{-g} option. 9265@end defmac 9266 9267@defmac XCOFF_DEBUGGING_INFO 9268Define this macro if GCC should produce XCOFF format debugging output 9269in response to the @option{-g} option. This is a variant of DBX format. 9270@end defmac 9271 9272@defmac DEFAULT_GDB_EXTENSIONS 9273Define this macro to control whether GCC should by default generate 9274GDB's extended version of DBX debugging information (assuming DBX-format 9275debugging information is enabled at all). If you don't define the 9276macro, the default is 1: always generate the extended information 9277if there is any occasion to. 9278@end defmac 9279 9280@defmac DEBUG_SYMS_TEXT 9281Define this macro if all @code{.stabs} commands should be output while 9282in the text section. 9283@end defmac 9284 9285@defmac ASM_STABS_OP 9286A C string constant, including spacing, naming the assembler pseudo op to 9287use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol. 9288If you don't define this macro, @code{"\t.stabs\t"} is used. This macro 9289applies only to DBX debugging information format. 9290@end defmac 9291 9292@defmac ASM_STABD_OP 9293A C string constant, including spacing, naming the assembler pseudo op to 9294use instead of @code{"\t.stabd\t"} to define a debugging symbol whose 9295value is the current location. If you don't define this macro, 9296@code{"\t.stabd\t"} is used. This macro applies only to DBX debugging 9297information format. 9298@end defmac 9299 9300@defmac ASM_STABN_OP 9301A C string constant, including spacing, naming the assembler pseudo op to 9302use instead of @code{"\t.stabn\t"} to define a debugging symbol with no 9303name. If you don't define this macro, @code{"\t.stabn\t"} is used. This 9304macro applies only to DBX debugging information format. 9305@end defmac 9306 9307@defmac DBX_NO_XREFS 9308Define this macro if DBX on your system does not support the construct 9309@samp{xs@var{tagname}}. On some systems, this construct is used to 9310describe a forward reference to a structure named @var{tagname}. 9311On other systems, this construct is not supported at all. 9312@end defmac 9313 9314@defmac DBX_CONTIN_LENGTH 9315A symbol name in DBX-format debugging information is normally 9316continued (split into two separate @code{.stabs} directives) when it 9317exceeds a certain length (by default, 80 characters). On some 9318operating systems, DBX requires this splitting; on others, splitting 9319must not be done. You can inhibit splitting by defining this macro 9320with the value zero. You can override the default splitting-length by 9321defining this macro as an expression for the length you desire. 9322@end defmac 9323 9324@defmac DBX_CONTIN_CHAR 9325Normally continuation is indicated by adding a @samp{\} character to 9326the end of a @code{.stabs} string when a continuation follows. To use 9327a different character instead, define this macro as a character 9328constant for the character you want to use. Do not define this macro 9329if backslash is correct for your system. 9330@end defmac 9331 9332@defmac DBX_STATIC_STAB_DATA_SECTION 9333Define this macro if it is necessary to go to the data section before 9334outputting the @samp{.stabs} pseudo-op for a non-global static 9335variable. 9336@end defmac 9337 9338@defmac DBX_TYPE_DECL_STABS_CODE 9339The value to use in the ``code'' field of the @code{.stabs} directive 9340for a typedef. The default is @code{N_LSYM}. 9341@end defmac 9342 9343@defmac DBX_STATIC_CONST_VAR_CODE 9344The value to use in the ``code'' field of the @code{.stabs} directive 9345for a static variable located in the text section. DBX format does not 9346provide any ``right'' way to do this. The default is @code{N_FUN}. 9347@end defmac 9348 9349@defmac DBX_REGPARM_STABS_CODE 9350The value to use in the ``code'' field of the @code{.stabs} directive 9351for a parameter passed in registers. DBX format does not provide any 9352``right'' way to do this. The default is @code{N_RSYM}. 9353@end defmac 9354 9355@defmac DBX_REGPARM_STABS_LETTER 9356The letter to use in DBX symbol data to identify a symbol as a parameter 9357passed in registers. DBX format does not customarily provide any way to 9358do this. The default is @code{'P'}. 9359@end defmac 9360 9361@defmac DBX_FUNCTION_FIRST 9362Define this macro if the DBX information for a function and its 9363arguments should precede the assembler code for the function. Normally, 9364in DBX format, the debugging information entirely follows the assembler 9365code. 9366@end defmac 9367 9368@defmac DBX_BLOCKS_FUNCTION_RELATIVE 9369Define this macro, with value 1, if the value of a symbol describing 9370the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be 9371relative to the start of the enclosing function. Normally, GCC uses 9372an absolute address. 9373@end defmac 9374 9375@defmac DBX_LINES_FUNCTION_RELATIVE 9376Define this macro, with value 1, if the value of a symbol indicating 9377the current line number (@code{N_SLINE}) should be relative to the 9378start of the enclosing function. Normally, GCC uses an absolute address. 9379@end defmac 9380 9381@defmac DBX_USE_BINCL 9382Define this macro if GCC should generate @code{N_BINCL} and 9383@code{N_EINCL} stabs for included header files, as on Sun systems. This 9384macro also directs GCC to output a type number as a pair of a file 9385number and a type number within the file. Normally, GCC does not 9386generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single 9387number for a type number. 9388@end defmac 9389 9390@node DBX Hooks 9391@subsection Open-Ended Hooks for DBX Format 9392 9393@c prevent bad page break with this line 9394These are hooks for DBX format. 9395 9396@defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter}) 9397A C statement to output DBX debugging information before code for line 9398number @var{line} of the current source file to the stdio stream 9399@var{stream}. @var{counter} is the number of time the macro was 9400invoked, including the current invocation; it is intended to generate 9401unique labels in the assembly output. 9402 9403This macro should not be defined if the default output is correct, or 9404if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}. 9405@end defmac 9406 9407@defmac NO_DBX_FUNCTION_END 9408Some stabs encapsulation formats (in particular ECOFF), cannot handle the 9409@code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct. 9410On those machines, define this macro to turn this feature off without 9411disturbing the rest of the gdb extensions. 9412@end defmac 9413 9414@defmac NO_DBX_BNSYM_ENSYM 9415Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx 9416extension construct. On those machines, define this macro to turn this 9417feature off without disturbing the rest of the gdb extensions. 9418@end defmac 9419 9420@node File Names and DBX 9421@subsection File Names in DBX Format 9422 9423@c prevent bad page break with this line 9424This describes file names in DBX format. 9425 9426@defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name}) 9427A C statement to output DBX debugging information to the stdio stream 9428@var{stream}, which indicates that file @var{name} is the main source 9429file---the file specified as the input file for compilation. 9430This macro is called only once, at the beginning of compilation. 9431 9432This macro need not be defined if the standard form of output 9433for DBX debugging information is appropriate. 9434 9435It may be necessary to refer to a label equal to the beginning of the 9436text section. You can use @samp{assemble_name (stream, ltext_label_name)} 9437to do so. If you do this, you must also set the variable 9438@var{used_ltext_label_name} to @code{true}. 9439@end defmac 9440 9441@defmac NO_DBX_MAIN_SOURCE_DIRECTORY 9442Define this macro, with value 1, if GCC should not emit an indication 9443of the current directory for compilation and current source language at 9444the beginning of the file. 9445@end defmac 9446 9447@defmac NO_DBX_GCC_MARKER 9448Define this macro, with value 1, if GCC should not emit an indication 9449that this object file was compiled by GCC@. The default is to emit 9450an @code{N_OPT} stab at the beginning of every source file, with 9451@samp{gcc2_compiled.} for the string and value 0. 9452@end defmac 9453 9454@defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name}) 9455A C statement to output DBX debugging information at the end of 9456compilation of the main source file @var{name}. Output should be 9457written to the stdio stream @var{stream}. 9458 9459If you don't define this macro, nothing special is output at the end 9460of compilation, which is correct for most machines. 9461@end defmac 9462 9463@defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END 9464Define this macro @emph{instead of} defining 9465@code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at 9466the end of compilation is an @code{N_SO} stab with an empty string, 9467whose value is the highest absolute text address in the file. 9468@end defmac 9469 9470@need 2000 9471@node SDB and DWARF 9472@subsection Macros for SDB and DWARF Output 9473 9474@c prevent bad page break with this line 9475Here are macros for SDB and DWARF output. 9476 9477@defmac SDB_DEBUGGING_INFO 9478Define this macro if GCC should produce COFF-style debugging output 9479for SDB in response to the @option{-g} option. 9480@end defmac 9481 9482@defmac DWARF2_DEBUGGING_INFO 9483Define this macro if GCC should produce dwarf version 2 format 9484debugging output in response to the @option{-g} option. 9485 9486@deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function}) 9487Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to 9488be emitted for each function. Instead of an integer return the enum 9489value for the @code{DW_CC_} tag. 9490@end deftypefn 9491 9492To support optional call frame debugging information, you must also 9493define @code{INCOMING_RETURN_ADDR_RTX} and either set 9494@code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the 9495prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save} 9496as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't. 9497@end defmac 9498 9499@defmac DWARF2_FRAME_INFO 9500Define this macro to a nonzero value if GCC should always output 9501Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO} 9502(@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and 9503exceptions are enabled, GCC will output this information not matter 9504how you define @code{DWARF2_FRAME_INFO}. 9505@end defmac 9506 9507@deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void) 9508This hook defines the mechanism that will be used for describing frame 9509unwind information to the debugger. Normally the hook will return 9510@code{UI_DWARF2} if DWARF 2 debug information is enabled, and 9511return @code{UI_NONE} otherwise. 9512 9513A target may return @code{UI_DWARF2} even when DWARF 2 debug information 9514is disabled in order to always output DWARF 2 frame information. 9515 9516A target may return @code{UI_TARGET} if it has ABI specified unwind tables. 9517This will suppress generation of the normal debug frame unwind information. 9518@end deftypefn 9519 9520@defmac DWARF2_ASM_LINE_DEBUG_INFO 9521Define this macro to be a nonzero value if the assembler can generate Dwarf 2 9522line debug info sections. This will result in much more compact line number 9523tables, and hence is desirable if it works. 9524@end defmac 9525 9526@deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS 9527True if the @code{.debug_pubtypes} and @code{.debug_pubnames} sections should be emitted. These sections are not used on most platforms, and in particular GDB does not use them. 9528@end deftypevr 9529 9530@deftypevr {Target Hook} bool TARGET_FORCE_AT_COMP_DIR 9531True if the @code{DW_AT_comp_dir} attribute should be emitted for each compilation unit. This attribute is required for the darwin linker to emit debug information. 9532@end deftypevr 9533 9534@deftypevr {Target Hook} bool TARGET_DELAY_SCHED2 9535True if sched2 is not to be run at its normal place. 9536This usually means it will be run as part of machine-specific reorg. 9537@end deftypevr 9538 9539@deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK 9540True if vartrack is not to be run at its normal place. 9541This usually means it will be run as part of machine-specific reorg. 9542@end deftypevr 9543 9544@deftypevr {Target Hook} bool TARGET_NO_REGISTER_ALLOCATION 9545True if register allocation and the passes 9546following it should not be run. Usually true only for virtual assembler 9547targets. 9548@end deftypevr 9549 9550@defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) 9551A C statement to issue assembly directives that create a difference 9552@var{lab1} minus @var{lab2}, using an integer of the given @var{size}. 9553@end defmac 9554 9555@defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) 9556A C statement to issue assembly directives that create a difference 9557between the two given labels in system defined units, e.g. instruction 9558slots on IA64 VMS, using an integer of the given size. 9559@end defmac 9560 9561@defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section}) 9562A C statement to issue assembly directives that create a 9563section-relative reference to the given @var{label}, using an integer of the 9564given @var{size}. The label is known to be defined in the given @var{section}. 9565@end defmac 9566 9567@defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label}) 9568A C statement to issue assembly directives that create a self-relative 9569reference to the given @var{label}, using an integer of the given @var{size}. 9570@end defmac 9571 9572@defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label}) 9573A C statement to issue assembly directives that create a reference to 9574the DWARF table identifier @var{label} from the current section. This 9575is used on some systems to avoid garbage collecting a DWARF table which 9576is referenced by a function. 9577@end defmac 9578 9579@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x}) 9580If defined, this target hook is a function which outputs a DTP-relative 9581reference to the given TLS symbol of the specified size. 9582@end deftypefn 9583 9584@defmac PUT_SDB_@dots{} 9585Define these macros to override the assembler syntax for the special 9586SDB assembler directives. See @file{sdbout.c} for a list of these 9587macros and their arguments. If the standard syntax is used, you need 9588not define them yourself. 9589@end defmac 9590 9591@defmac SDB_DELIM 9592Some assemblers do not support a semicolon as a delimiter, even between 9593SDB assembler directives. In that case, define this macro to be the 9594delimiter to use (usually @samp{\n}). It is not necessary to define 9595a new set of @code{PUT_SDB_@var{op}} macros if this is the only change 9596required. 9597@end defmac 9598 9599@defmac SDB_ALLOW_UNKNOWN_REFERENCES 9600Define this macro to allow references to unknown structure, 9601union, or enumeration tags to be emitted. Standard COFF does not 9602allow handling of unknown references, MIPS ECOFF has support for 9603it. 9604@end defmac 9605 9606@defmac SDB_ALLOW_FORWARD_REFERENCES 9607Define this macro to allow references to structure, union, or 9608enumeration tags that have not yet been seen to be handled. Some 9609assemblers choke if forward tags are used, while some require it. 9610@end defmac 9611 9612@defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}) 9613A C statement to output SDB debugging information before code for line 9614number @var{line} of the current source file to the stdio stream 9615@var{stream}. The default is to emit an @code{.ln} directive. 9616@end defmac 9617 9618@need 2000 9619@node VMS Debug 9620@subsection Macros for VMS Debug Format 9621 9622@c prevent bad page break with this line 9623Here are macros for VMS debug format. 9624 9625@defmac VMS_DEBUGGING_INFO 9626Define this macro if GCC should produce debugging output for VMS 9627in response to the @option{-g} option. The default behavior for VMS 9628is to generate minimal debug info for a traceback in the absence of 9629@option{-g} unless explicitly overridden with @option{-g0}. This 9630behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and 9631@code{TARGET_OPTION_OVERRIDE}. 9632@end defmac 9633 9634@node Floating Point 9635@section Cross Compilation and Floating Point 9636@cindex cross compilation and floating point 9637@cindex floating point and cross compilation 9638 9639While all modern machines use twos-complement representation for integers, 9640there are a variety of representations for floating point numbers. This 9641means that in a cross-compiler the representation of floating point numbers 9642in the compiled program may be different from that used in the machine 9643doing the compilation. 9644 9645Because different representation systems may offer different amounts of 9646range and precision, all floating point constants must be represented in 9647the target machine's format. Therefore, the cross compiler cannot 9648safely use the host machine's floating point arithmetic; it must emulate 9649the target's arithmetic. To ensure consistency, GCC always uses 9650emulation to work with floating point values, even when the host and 9651target floating point formats are identical. 9652 9653The following macros are provided by @file{real.h} for the compiler to 9654use. All parts of the compiler which generate or optimize 9655floating-point calculations must use these macros. They may evaluate 9656their operands more than once, so operands must not have side effects. 9657 9658@defmac REAL_VALUE_TYPE 9659The C data type to be used to hold a floating point value in the target 9660machine's format. Typically this is a @code{struct} containing an 9661array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque 9662quantity. 9663@end defmac 9664 9665@deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 9666Compares for equality the two values, @var{x} and @var{y}. If the target 9667floating point format supports negative zeroes and/or NaNs, 9668@samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and 9669@samp{REAL_VALUES_EQUAL (NaN, NaN)} is false. 9670@end deftypefn 9671 9672@deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 9673Tests whether @var{x} is less than @var{y}. 9674@end deftypefn 9675 9676@deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x}) 9677Truncates @var{x} to a signed integer, rounding toward zero. 9678@end deftypefn 9679 9680@deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x}) 9681Truncates @var{x} to an unsigned integer, rounding toward zero. If 9682@var{x} is negative, returns zero. 9683@end deftypefn 9684 9685@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode}) 9686Converts @var{string} into a floating point number in the target machine's 9687representation for mode @var{mode}. This routine can handle both 9688decimal and hexadecimal floating point constants, using the syntax 9689defined by the C language for both. 9690@end deftypefn 9691 9692@deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x}) 9693Returns 1 if @var{x} is negative (including negative zero), 0 otherwise. 9694@end deftypefn 9695 9696@deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x}) 9697Determines whether @var{x} represents infinity (positive or negative). 9698@end deftypefn 9699 9700@deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x}) 9701Determines whether @var{x} represents a ``NaN'' (not-a-number). 9702@end deftypefn 9703 9704@deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 9705Calculates an arithmetic operation on the two floating point values 9706@var{x} and @var{y}, storing the result in @var{output} (which must be a 9707variable). 9708 9709The operation to be performed is specified by @var{code}. Only the 9710following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR}, 9711@code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}. 9712 9713If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the 9714target's floating point format cannot represent infinity, it will call 9715@code{abort}. Callers should check for this situation first, using 9716@code{MODE_HAS_INFINITIES}. @xref{Storage Layout}. 9717@end deftypefn 9718 9719@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x}) 9720Returns the negative of the floating point value @var{x}. 9721@end deftypefn 9722 9723@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x}) 9724Returns the absolute value of @var{x}. 9725@end deftypefn 9726 9727@node Mode Switching 9728@section Mode Switching Instructions 9729@cindex mode switching 9730The following macros control mode switching optimizations: 9731 9732@defmac OPTIMIZE_MODE_SWITCHING (@var{entity}) 9733Define this macro if the port needs extra instructions inserted for mode 9734switching in an optimizing compilation. 9735 9736For an example, the SH4 can perform both single and double precision 9737floating point operations, but to perform a single precision operation, 9738the FPSCR PR bit has to be cleared, while for a double precision 9739operation, this bit has to be set. Changing the PR bit requires a general 9740purpose register as a scratch register, hence these FPSCR sets have to 9741be inserted before reload, i.e.@: you can't put this into instruction emitting 9742or @code{TARGET_MACHINE_DEPENDENT_REORG}. 9743 9744You can have multiple entities that are mode-switched, and select at run time 9745which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should 9746return nonzero for any @var{entity} that needs mode-switching. 9747If you define this macro, you also have to define 9748@code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED}, 9749@code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}. 9750@code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT} 9751are optional. 9752@end defmac 9753 9754@defmac NUM_MODES_FOR_MODE_SWITCHING 9755If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as 9756initializer for an array of integers. Each initializer element 9757N refers to an entity that needs mode switching, and specifies the number 9758of different modes that might need to be set for this entity. 9759The position of the initializer in the initializer---starting counting at 9760zero---determines the integer that is used to refer to the mode-switched 9761entity in question. 9762In macros that take mode arguments / yield a mode result, modes are 9763represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode 9764switch is needed / supplied. 9765@end defmac 9766 9767@deftypefn {Target Hook} void TARGET_MODE_EMIT (int @var{entity}, int @var{mode}, int @var{prev_mode}, HARD_REG_SET @var{regs_live}) 9768Generate one or more insns to set @var{entity} to @var{mode}. @var{hard_reg_live} is the set of hard registers live at the point where the insn(s) are to be inserted. @var{prev_moxde} indicates the mode to switch from. Sets of a lower numbered entity will be emitted before sets of a higher numbered entity to a mode of the same or lower priority. 9769@end deftypefn 9770 9771@deftypefn {Target Hook} int TARGET_MODE_NEEDED (int @var{entity}, rtx_insn *@var{insn}) 9772@var{entity} is an integer specifying a mode-switched entity. If @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to return an integer value not larger than the corresponding element in @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must be switched into prior to the execution of @var{insn}. 9773@end deftypefn 9774 9775@deftypefn {Target Hook} int TARGET_MODE_AFTER (int @var{entity}, int @var{mode}, rtx_insn *@var{insn}) 9776@var{entity} is an integer specifying a mode-switched entity. If this macro is defined, it is evaluated for every @var{insn} during mode switching. It determines the mode that an insn results in (if different from the incoming mode). 9777@end deftypefn 9778 9779@deftypefn {Target Hook} int TARGET_MODE_ENTRY (int @var{entity}) 9780If this macro is defined, it is evaluated for every @var{entity} that needs mode switching. It should evaluate to an integer, which is a mode that @var{entity} is assumed to be switched to at function entry. If @code{TARGET_MODE_ENTRY} is defined then @code{TARGET_MODE_EXIT} must be defined. 9781@end deftypefn 9782 9783@deftypefn {Target Hook} int TARGET_MODE_EXIT (int @var{entity}) 9784If this macro is defined, it is evaluated for every @var{entity} that needs mode switching. It should evaluate to an integer, which is a mode that @var{entity} is assumed to be switched to at function exit. If @code{TARGET_MODE_EXIT} is defined then @code{TARGET_MODE_ENTRY} must be defined. 9785@end deftypefn 9786 9787@deftypefn {Target Hook} int TARGET_MODE_PRIORITY (int @var{entity}, int @var{n}) 9788This macro specifies the order in which modes for @var{entity} are processed. 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the lowest. The value of the macro should be an integer designating a mode for @var{entity}. For any fixed @var{entity}, @code{mode_priority} (@var{entity}, @var{n}) shall be a bijection in 0 @dots{} @code{num_modes_for_mode_switching[@var{entity}] - 1}. 9789@end deftypefn 9790 9791@node Target Attributes 9792@section Defining target-specific uses of @code{__attribute__} 9793@cindex target attributes 9794@cindex machine attributes 9795@cindex attributes, target-specific 9796 9797Target-specific attributes may be defined for functions, data and types. 9798These are described using the following target hooks; they also need to 9799be documented in @file{extend.texi}. 9800 9801@deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE 9802If defined, this target hook points to an array of @samp{struct 9803attribute_spec} (defined in @file{tree.h}) specifying the machine 9804specific attributes for this target and some of the restrictions on the 9805entities to which these attributes are applied and the arguments they 9806take. 9807@end deftypevr 9808 9809@deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name}) 9810If defined, this target hook is a function which returns true if the 9811machine-specific attribute named @var{name} expects an identifier 9812given as its first argument to be passed on as a plain identifier, not 9813subjected to name lookup. If this is not defined, the default is 9814false for all machine-specific attributes. 9815@end deftypefn 9816 9817@deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2}) 9818If defined, this target hook is a function which returns zero if the attributes on 9819@var{type1} and @var{type2} are incompatible, one if they are compatible, 9820and two if they are nearly compatible (which causes a warning to be 9821generated). If this is not defined, machine-specific attributes are 9822supposed always to be compatible. 9823@end deftypefn 9824 9825@deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type}) 9826If defined, this target hook is a function which assigns default attributes to 9827the newly defined @var{type}. 9828@end deftypefn 9829 9830@deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) 9831Define this target hook if the merging of type attributes needs special 9832handling. If defined, the result is a list of the combined 9833@code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed 9834that @code{comptypes} has already been called and returned 1. This 9835function may call @code{merge_attributes} to handle machine-independent 9836merging. 9837@end deftypefn 9838 9839@deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl}) 9840Define this target hook if the merging of decl attributes needs special 9841handling. If defined, the result is a list of the combined 9842@code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}. 9843@var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of 9844when this is needed are when one attribute overrides another, or when an 9845attribute is nullified by a subsequent definition. This function may 9846call @code{merge_attributes} to handle machine-independent merging. 9847 9848@findex TARGET_DLLIMPORT_DECL_ATTRIBUTES 9849If the only target-specific handling you require is @samp{dllimport} 9850for Microsoft Windows targets, you should define the macro 9851@code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler 9852will then define a function called 9853@code{merge_dllimport_decl_attributes} which can then be defined as 9854the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also 9855add @code{handle_dll_attribute} in the attribute table for your port 9856to perform initial processing of the @samp{dllimport} and 9857@samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and 9858@file{i386/i386.c}, for example. 9859@end deftypefn 9860 9861@deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl}) 9862@var{decl} is a variable or function with @code{__attribute__((dllimport))} specified. Use this hook if the target needs to add extra validation checks to @code{handle_dll_attribute}. 9863@end deftypefn 9864 9865@defmac TARGET_DECLSPEC 9866Define this macro to a nonzero value if you want to treat 9867@code{__declspec(X)} as equivalent to @code{__attribute((X))}. By 9868default, this behavior is enabled only for targets that define 9869@code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation 9870of @code{__declspec} is via a built-in macro, but you should not rely 9871on this implementation detail. 9872@end defmac 9873 9874@deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr}) 9875Define this target hook if you want to be able to add attributes to a decl 9876when it is being created. This is normally useful for back ends which 9877wish to implement a pragma by using the attributes which correspond to 9878the pragma's effect. The @var{node} argument is the decl which is being 9879created. The @var{attr_ptr} argument is a pointer to the attribute list 9880for this decl. The list itself should not be modified, since it may be 9881shared with other decls, but attributes may be chained on the head of 9882the list and @code{*@var{attr_ptr}} modified to point to the new 9883attributes, or a copy of the list may be made if further changes are 9884needed. 9885@end deftypefn 9886 9887@deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl}) 9888@cindex inlining 9889This target hook returns @code{true} if it is OK to inline @var{fndecl} 9890into the current function, despite its having target-specific 9891attributes, @code{false} otherwise. By default, if a function has a 9892target specific attribute attached to it, it will not be inlined. 9893@end deftypefn 9894 9895@deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags}) 9896This hook is called to parse @code{attribute(target("..."))}, which 9897allows setting target-specific options on individual functions. 9898These function-specific options may differ 9899from the options specified on the command line. The hook should return 9900@code{true} if the options are valid. 9901 9902The hook should set the @code{DECL_FUNCTION_SPECIFIC_TARGET} field in 9903the function declaration to hold a pointer to a target-specific 9904@code{struct cl_target_option} structure. 9905@end deftypefn 9906 9907@deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr}, struct gcc_options *@var{opts}) 9908This hook is called to save any additional target-specific information 9909in the @code{struct cl_target_option} structure for function-specific 9910options from the @code{struct gcc_options} structure. 9911@xref{Option file format}. 9912@end deftypefn 9913 9914@deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct gcc_options *@var{opts}, struct cl_target_option *@var{ptr}) 9915This hook is called to restore any additional target-specific 9916information in the @code{struct cl_target_option} structure for 9917function-specific options to the @code{struct gcc_options} structure. 9918@end deftypefn 9919 9920@deftypefn {Target Hook} void TARGET_OPTION_POST_STREAM_IN (struct cl_target_option *@var{ptr}) 9921This hook is called to update target-specific information in the 9922@code{struct cl_target_option} structure after it is streamed in from 9923LTO bytecode. 9924@end deftypefn 9925 9926@deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr}) 9927This hook is called to print any additional target-specific 9928information in the @code{struct cl_target_option} structure for 9929function-specific options. 9930@end deftypefn 9931 9932@deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target}) 9933This target hook parses the options for @code{#pragma GCC target}, which 9934sets the target-specific options for functions that occur later in the 9935input stream. The options accepted should be the same as those handled by the 9936@code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook. 9937@end deftypefn 9938 9939@deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void) 9940Sometimes certain combinations of command options do not make sense on 9941a particular target machine. You can override the hook 9942@code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called 9943once just after all the command options have been parsed. 9944 9945Don't use this hook to turn on various extra optimizations for 9946@option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for. 9947 9948If you need to do something whenever the optimization level is 9949changed via the optimize attribute or pragma, see 9950@code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE} 9951@end deftypefn 9952 9953@deftypefn {Target Hook} bool TARGET_OPTION_FUNCTION_VERSIONS (tree @var{decl1}, tree @var{decl2}) 9954This target hook returns @code{true} if @var{DECL1} and @var{DECL2} are 9955versions of the same function. @var{DECL1} and @var{DECL2} are function 9956versions if and only if they have the same function signature and 9957different target specific attributes, that is, they are compiled for 9958different target machines. 9959@end deftypefn 9960 9961@deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee}) 9962This target hook returns @code{false} if the @var{caller} function 9963cannot inline @var{callee}, based on target specific information. By 9964default, inlining is not allowed if the callee function has function 9965specific target options and the caller does not use the same options. 9966@end deftypefn 9967 9968@node Emulated TLS 9969@section Emulating TLS 9970@cindex Emulated TLS 9971 9972For targets whose psABI does not provide Thread Local Storage via 9973specific relocations and instruction sequences, an emulation layer is 9974used. A set of target hooks allows this emulation layer to be 9975configured for the requirements of a particular target. For instance 9976the psABI may in fact specify TLS support in terms of an emulation 9977layer. 9978 9979The emulation layer works by creating a control object for every TLS 9980object. To access the TLS object, a lookup function is provided 9981which, when given the address of the control object, will return the 9982address of the current thread's instance of the TLS object. 9983 9984@deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS 9985Contains the name of the helper function that uses a TLS control 9986object to locate a TLS instance. The default causes libgcc's 9987emulated TLS helper function to be used. 9988@end deftypevr 9989 9990@deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON 9991Contains the name of the helper function that should be used at 9992program startup to register TLS objects that are implicitly 9993initialized to zero. If this is @code{NULL}, all TLS objects will 9994have explicit initializers. The default causes libgcc's emulated TLS 9995registration function to be used. 9996@end deftypevr 9997 9998@deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION 9999Contains the name of the section in which TLS control variables should 10000be placed. The default of @code{NULL} allows these to be placed in 10001any section. 10002@end deftypevr 10003 10004@deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION 10005Contains the name of the section in which TLS initializers should be 10006placed. The default of @code{NULL} allows these to be placed in any 10007section. 10008@end deftypevr 10009 10010@deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX 10011Contains the prefix to be prepended to TLS control variable names. 10012The default of @code{NULL} uses a target-specific prefix. 10013@end deftypevr 10014 10015@deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX 10016Contains the prefix to be prepended to TLS initializer objects. The 10017default of @code{NULL} uses a target-specific prefix. 10018@end deftypevr 10019 10020@deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name}) 10021Specifies a function that generates the FIELD_DECLs for a TLS control 10022object type. @var{type} is the RECORD_TYPE the fields are for and 10023@var{name} should be filled with the structure tag, if the default of 10024@code{__emutls_object} is unsuitable. The default creates a type suitable 10025for libgcc's emulated TLS function. 10026@end deftypefn 10027 10028@deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr}) 10029Specifies a function that generates the CONSTRUCTOR to initialize a 10030TLS control object. @var{var} is the TLS control object, @var{decl} 10031is the TLS object and @var{tmpl_addr} is the address of the 10032initializer. The default initializes libgcc's emulated TLS control object. 10033@end deftypefn 10034 10035@deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED 10036Specifies whether the alignment of TLS control variable objects is 10037fixed and should not be increased as some backends may do to optimize 10038single objects. The default is false. 10039@end deftypevr 10040 10041@deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS 10042Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor 10043may be used to describe emulated TLS control objects. 10044@end deftypevr 10045 10046@node MIPS Coprocessors 10047@section Defining coprocessor specifics for MIPS targets. 10048@cindex MIPS coprocessor-definition macros 10049 10050The MIPS specification allows MIPS implementations to have as many as 4 10051coprocessors, each with as many as 32 private registers. GCC supports 10052accessing these registers and transferring values between the registers 10053and memory using asm-ized variables. For example: 10054 10055@smallexample 10056 register unsigned int cp0count asm ("c0r1"); 10057 unsigned int d; 10058 10059 d = cp0count + 3; 10060@end smallexample 10061 10062(``c0r1'' is the default name of register 1 in coprocessor 0; alternate 10063names may be added as described below, or the default names may be 10064overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.) 10065 10066Coprocessor registers are assumed to be epilogue-used; sets to them will 10067be preserved even if it does not appear that the register is used again 10068later in the function. 10069 10070Another note: according to the MIPS spec, coprocessor 1 (if present) is 10071the FPU@. One accesses COP1 registers through standard mips 10072floating-point support; they are not included in this mechanism. 10073 10074@node PCH Target 10075@section Parameters for Precompiled Header Validity Checking 10076@cindex parameters, precompiled headers 10077 10078@deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz}) 10079This hook returns a pointer to the data needed by 10080@code{TARGET_PCH_VALID_P} and sets 10081@samp{*@var{sz}} to the size of the data in bytes. 10082@end deftypefn 10083 10084@deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz}) 10085This hook checks whether the options used to create a PCH file are 10086compatible with the current settings. It returns @code{NULL} 10087if so and a suitable error message if not. Error messages will 10088be presented to the user and must be localized using @samp{_(@var{msg})}. 10089 10090@var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY} 10091when the PCH file was created and @var{sz} is the size of that data in bytes. 10092It's safe to assume that the data was created by the same version of the 10093compiler, so no format checking is needed. 10094 10095The default definition of @code{default_pch_valid_p} should be 10096suitable for most targets. 10097@end deftypefn 10098 10099@deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags}) 10100If this hook is nonnull, the default implementation of 10101@code{TARGET_PCH_VALID_P} will use it to check for compatible values 10102of @code{target_flags}. @var{pch_flags} specifies the value that 10103@code{target_flags} had when the PCH file was created. The return 10104value is the same as for @code{TARGET_PCH_VALID_P}. 10105@end deftypefn 10106 10107@deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void) 10108Called before writing out a PCH file. If the target has some 10109garbage-collected data that needs to be in a particular state on PCH loads, 10110it can use this hook to enforce that state. Very few targets need 10111to do anything here. 10112@end deftypefn 10113 10114@node C++ ABI 10115@section C++ ABI parameters 10116@cindex parameters, c++ abi 10117 10118@deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void) 10119Define this hook to override the integer type used for guard variables. 10120These are used to implement one-time construction of static objects. The 10121default is long_long_integer_type_node. 10122@end deftypefn 10123 10124@deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void) 10125This hook determines how guard variables are used. It should return 10126@code{false} (the default) if the first byte should be used. A return value of 10127@code{true} indicates that only the least significant bit should be used. 10128@end deftypefn 10129 10130@deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type}) 10131This hook returns the size of the cookie to use when allocating an array 10132whose elements have the indicated @var{type}. Assumes that it is already 10133known that a cookie is needed. The default is 10134@code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the 10135IA64/Generic C++ ABI@. 10136@end deftypefn 10137 10138@deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void) 10139This hook should return @code{true} if the element size should be stored in 10140array cookies. The default is to return @code{false}. 10141@end deftypefn 10142 10143@deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export}) 10144If defined by a backend this hook allows the decision made to export 10145class @var{type} to be overruled. Upon entry @var{import_export} 10146will contain 1 if the class is going to be exported, @minus{}1 if it is going 10147to be imported and 0 otherwise. This function should return the 10148modified value and perform any other actions necessary to support the 10149backend's targeted operating system. 10150@end deftypefn 10151 10152@deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void) 10153This hook should return @code{true} if constructors and destructors return 10154the address of the object created/destroyed. The default is to return 10155@code{false}. 10156@end deftypefn 10157 10158@deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void) 10159This hook returns true if the key method for a class (i.e., the method 10160which, if defined in the current translation unit, causes the virtual 10161table to be emitted) may be an inline function. Under the standard 10162Itanium C++ ABI the key method may be an inline function so long as 10163the function is not declared inline in the class definition. Under 10164some variants of the ABI, an inline function can never be the key 10165method. The default is to return @code{true}. 10166@end deftypefn 10167 10168@deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl}) 10169@var{decl} is a virtual table, virtual table table, typeinfo object, or other similar implicit class data object that will be emitted with external linkage in this translation unit. No ELF visibility has been explicitly specified. If the target needs to specify a visibility other than that of the containing class, use this hook to set @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}. 10170@end deftypefn 10171 10172@deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void) 10173This hook returns true (the default) if virtual tables and other 10174similar implicit class data objects are always COMDAT if they have 10175external linkage. If this hook returns false, then class data for 10176classes whose virtual table will be emitted in only one translation 10177unit will not be COMDAT. 10178@end deftypefn 10179 10180@deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void) 10181This hook returns true (the default) if the RTTI information for 10182the basic types which is defined in the C++ runtime should always 10183be COMDAT, false if it should not be COMDAT. 10184@end deftypefn 10185 10186@deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void) 10187This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI) 10188should be used to register static destructors when @option{-fuse-cxa-atexit} 10189is in effect. The default is to return false to use @code{__cxa_atexit}. 10190@end deftypefn 10191 10192@deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void) 10193This hook returns true if the target @code{atexit} function can be used 10194in the same manner as @code{__cxa_atexit} to register C++ static 10195destructors. This requires that @code{atexit}-registered functions in 10196shared libraries are run in the correct order when the libraries are 10197unloaded. The default is to return false. 10198@end deftypefn 10199 10200@deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type}) 10201@var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been defined. Use this hook to make adjustments to the class (eg, tweak visibility or perform any other required target modifications). 10202@end deftypefn 10203 10204@deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl}) 10205Return target-specific mangling context of @var{decl} or @code{NULL_TREE}. 10206@end deftypefn 10207 10208@node Named Address Spaces 10209@section Adding support for named address spaces 10210@cindex named address spaces 10211 10212The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275 10213standards committee, @cite{Programming Languages - C - Extensions to 10214support embedded processors}, specifies a syntax for embedded 10215processors to specify alternate address spaces. You can configure a 10216GCC port to support section 5.1 of the draft report to add support for 10217address spaces other than the default address space. These address 10218spaces are new keywords that are similar to the @code{volatile} and 10219@code{const} type attributes. 10220 10221Pointers to named address spaces can have a different size than 10222pointers to the generic address space. 10223 10224For example, the SPU port uses the @code{__ea} address space to refer 10225to memory in the host processor, rather than memory local to the SPU 10226processor. Access to memory in the @code{__ea} address space involves 10227issuing DMA operations to move data between the host processor and the 10228local processor memory address space. Pointers in the @code{__ea} 10229address space are either 32 bits or 64 bits based on the 10230@option{-mea32} or @option{-mea64} switches (native SPU pointers are 10231always 32 bits). 10232 10233Internally, address spaces are represented as a small integer in the 10234range 0 to 15 with address space 0 being reserved for the generic 10235address space. 10236 10237To register a named address space qualifier keyword with the C front end, 10238the target may call the @code{c_register_addr_space} routine. For example, 10239the SPU port uses the following to declare @code{__ea} as the keyword for 10240named address space #1: 10241@smallexample 10242#define ADDR_SPACE_EA 1 10243c_register_addr_space ("__ea", ADDR_SPACE_EA); 10244@end smallexample 10245 10246@deftypefn {Target Hook} machine_mode TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space}) 10247Define this to return the machine mode to use for pointers to 10248@var{address_space} if the target supports named address spaces. 10249The default version of this hook returns @code{ptr_mode} for the 10250generic address space only. 10251@end deftypefn 10252 10253@deftypefn {Target Hook} machine_mode TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space}) 10254Define this to return the machine mode to use for addresses in 10255@var{address_space} if the target supports named address spaces. 10256The default version of this hook returns @code{Pmode} for the 10257generic address space only. 10258@end deftypefn 10259 10260@deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (machine_mode @var{mode}, addr_space_t @var{as}) 10261Define this to return nonzero if the port can handle pointers 10262with machine mode @var{mode} to address space @var{as}. This target 10263hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook, 10264except that it includes explicit named address space support. The default 10265version of this hook returns true for the modes returned by either the 10266@code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE} 10267target hooks for the given address space. 10268@end deftypefn 10269 10270@deftypefn {Target Hook} bool TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{exp}, bool @var{strict}, addr_space_t @var{as}) 10271Define this to return true if @var{exp} is a valid address for mode 10272@var{mode} in the named address space @var{as}. The @var{strict} 10273parameter says whether strict addressing is in effect after reload has 10274finished. This target hook is the same as the 10275@code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes 10276explicit named address space support. 10277@end deftypefn 10278 10279@deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode}, addr_space_t @var{as}) 10280Define this to modify an invalid address @var{x} to be a valid address 10281with mode @var{mode} in the named address space @var{as}. This target 10282hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook, 10283except that it includes explicit named address space support. 10284@end deftypefn 10285 10286@deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset}) 10287Define this to return whether the @var{subset} named address space is 10288contained within the @var{superset} named address space. Pointers to 10289a named address space that is a subset of another named address space 10290will be converted automatically without a cast if used together in 10291arithmetic operations. Pointers to a superset address space can be 10292converted to pointers to a subset address space via explicit casts. 10293@end deftypefn 10294 10295@deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type}) 10296Define this to convert the pointer expression represented by the RTL 10297@var{op} with type @var{from_type} that points to a named address 10298space to a new pointer expression with type @var{to_type} that points 10299to a different named address space. When this hook it called, it is 10300guaranteed that one of the two address spaces is a subset of the other, 10301as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook. 10302@end deftypefn 10303 10304@node Misc 10305@section Miscellaneous Parameters 10306@cindex parameters, miscellaneous 10307 10308@c prevent bad page break with this line 10309Here are several miscellaneous parameters. 10310 10311@defmac HAS_LONG_COND_BRANCH 10312Define this boolean macro to indicate whether or not your architecture 10313has conditional branches that can span all of memory. It is used in 10314conjunction with an optimization that partitions hot and cold basic 10315blocks into separate sections of the executable. If this macro is 10316set to false, gcc will convert any conditional branches that attempt 10317to cross between sections into unconditional branches or indirect jumps. 10318@end defmac 10319 10320@defmac HAS_LONG_UNCOND_BRANCH 10321Define this boolean macro to indicate whether or not your architecture 10322has unconditional branches that can span all of memory. It is used in 10323conjunction with an optimization that partitions hot and cold basic 10324blocks into separate sections of the executable. If this macro is 10325set to false, gcc will convert any unconditional branches that attempt 10326to cross between sections into indirect jumps. 10327@end defmac 10328 10329@defmac CASE_VECTOR_MODE 10330An alias for a machine mode name. This is the machine mode that 10331elements of a jump-table should have. 10332@end defmac 10333 10334@defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body}) 10335Optional: return the preferred mode for an @code{addr_diff_vec} 10336when the minimum and maximum offset are known. If you define this, 10337it enables extra code in branch shortening to deal with @code{addr_diff_vec}. 10338To make this work, you also have to define @code{INSN_ALIGN} and 10339make the alignment for @code{addr_diff_vec} explicit. 10340The @var{body} argument is provided so that the offset_unsigned and scale 10341flags can be updated. 10342@end defmac 10343 10344@defmac CASE_VECTOR_PC_RELATIVE 10345Define this macro to be a C expression to indicate when jump-tables 10346should contain relative addresses. You need not define this macro if 10347jump-tables never contain relative addresses, or jump-tables should 10348contain relative addresses only when @option{-fPIC} or @option{-fPIC} 10349is in effect. 10350@end defmac 10351 10352@deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void) 10353This function return the smallest number of different values for which it 10354is best to use a jump-table instead of a tree of conditional branches. 10355The default is four for machines with a @code{casesi} instruction and 10356five otherwise. This is best for most machines. 10357@end deftypefn 10358 10359@defmac WORD_REGISTER_OPERATIONS 10360Define this macro if operations between registers with integral mode 10361smaller than a word are always performed on the entire register. 10362Most RISC machines have this property and most CISC machines do not. 10363@end defmac 10364 10365@defmac LOAD_EXTEND_OP (@var{mem_mode}) 10366Define this macro to be a C expression indicating when insns that read 10367memory in @var{mem_mode}, an integral mode narrower than a word, set the 10368bits outside of @var{mem_mode} to be either the sign-extension or the 10369zero-extension of the data read. Return @code{SIGN_EXTEND} for values 10370of @var{mem_mode} for which the 10371insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and 10372@code{UNKNOWN} for other modes. 10373 10374This macro is not called with @var{mem_mode} non-integral or with a width 10375greater than or equal to @code{BITS_PER_WORD}, so you may return any 10376value in this case. Do not define this macro if it would always return 10377@code{UNKNOWN}. On machines where this macro is defined, you will normally 10378define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}. 10379 10380You may return a non-@code{UNKNOWN} value even if for some hard registers 10381the sign extension is not performed, if for the @code{REGNO_REG_CLASS} 10382of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero 10383when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any 10384integral mode larger than this but not larger than @code{word_mode}. 10385 10386You must return @code{UNKNOWN} if for some hard registers that allow this 10387mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to 10388@code{word_mode}, but that they can change to another integral mode that 10389is larger then @var{mem_mode} but still smaller than @code{word_mode}. 10390@end defmac 10391 10392@defmac SHORT_IMMEDIATES_SIGN_EXTEND 10393Define this macro if loading short immediate values into registers sign 10394extends. 10395@end defmac 10396 10397@deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (machine_mode @var{mode}) 10398When @option{-ffast-math} is in effect, GCC tries to optimize 10399divisions by the same divisor, by turning them into multiplications by 10400the reciprocal. This target hook specifies the minimum number of divisions 10401that should be there for GCC to perform the optimization for a variable 10402of mode @var{mode}. The default implementation returns 3 if the machine 10403has an instruction for the division, and 2 if it does not. 10404@end deftypefn 10405 10406@defmac MOVE_MAX 10407The maximum number of bytes that a single instruction can move quickly 10408between memory and registers or between two memory locations. 10409@end defmac 10410 10411@defmac MAX_MOVE_MAX 10412The maximum number of bytes that a single instruction can move quickly 10413between memory and registers or between two memory locations. If this 10414is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the 10415constant value that is the largest value that @code{MOVE_MAX} can have 10416at run-time. 10417@end defmac 10418 10419@defmac SHIFT_COUNT_TRUNCATED 10420A C expression that is nonzero if on this machine the number of bits 10421actually used for the count of a shift operation is equal to the number 10422of bits needed to represent the size of the object being shifted. When 10423this macro is nonzero, the compiler will assume that it is safe to omit 10424a sign-extend, zero-extend, and certain bitwise `and' instructions that 10425truncates the count of a shift operation. On machines that have 10426instructions that act on bit-fields at variable positions, which may 10427include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED} 10428also enables deletion of truncations of the values that serve as 10429arguments to bit-field instructions. 10430 10431If both types of instructions truncate the count (for shifts) and 10432position (for bit-field operations), or if no variable-position bit-field 10433instructions exist, you should define this macro. 10434 10435However, on some machines, such as the 80386 and the 680x0, truncation 10436only applies to shift operations and not the (real or pretended) 10437bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on 10438such machines. Instead, add patterns to the @file{md} file that include 10439the implied truncation of the shift instructions. 10440 10441You need not define this macro if it would always have the value of zero. 10442@end defmac 10443 10444@anchor{TARGET_SHIFT_TRUNCATION_MASK} 10445@deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (machine_mode @var{mode}) 10446This function describes how the standard shift patterns for @var{mode} 10447deal with shifts by negative amounts or by more than the width of the mode. 10448@xref{shift patterns}. 10449 10450On many machines, the shift patterns will apply a mask @var{m} to the 10451shift count, meaning that a fixed-width shift of @var{x} by @var{y} is 10452equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If 10453this is true for mode @var{mode}, the function should return @var{m}, 10454otherwise it should return 0. A return value of 0 indicates that no 10455particular behavior is guaranteed. 10456 10457Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does 10458@emph{not} apply to general shift rtxes; it applies only to instructions 10459that are generated by the named shift patterns. 10460 10461The default implementation of this function returns 10462@code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED} 10463and 0 otherwise. This definition is always safe, but if 10464@code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns 10465nevertheless truncate the shift count, you may get better code 10466by overriding it. 10467@end deftypefn 10468 10469@defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec}) 10470A C expression which is nonzero if on this machine it is safe to 10471``convert'' an integer of @var{inprec} bits to one of @var{outprec} 10472bits (where @var{outprec} is smaller than @var{inprec}) by merely 10473operating on it as if it had only @var{outprec} bits. 10474 10475On many machines, this expression can be 1. 10476 10477@c rearranged this, removed the phrase "it is reported that". this was 10478@c to fix an overfull hbox. --mew 10feb93 10479When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for 10480modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result. 10481If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in 10482such cases may improve things. 10483@end defmac 10484 10485@deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (machine_mode @var{mode}, machine_mode @var{rep_mode}) 10486The representation of an integral mode can be such that the values 10487are always extended to a wider integral mode. Return 10488@code{SIGN_EXTEND} if values of @var{mode} are represented in 10489sign-extended form to @var{rep_mode}. Return @code{UNKNOWN} 10490otherwise. (Currently, none of the targets use zero-extended 10491representation this way so unlike @code{LOAD_EXTEND_OP}, 10492@code{TARGET_MODE_REP_EXTENDED} is expected to return either 10493@code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends 10494@var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next 10495widest integral mode and currently we take advantage of this fact.) 10496 10497Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN} 10498value even if the extension is not performed on certain hard registers 10499as long as for the @code{REGNO_REG_CLASS} of these hard registers 10500@code{CANNOT_CHANGE_MODE_CLASS} returns nonzero. 10501 10502Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP} 10503describe two related properties. If you define 10504@code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want 10505to define @code{LOAD_EXTEND_OP (mode)} to return the same type of 10506extension. 10507 10508In order to enforce the representation of @code{mode}, 10509@code{TRULY_NOOP_TRUNCATION} should return false when truncating to 10510@code{mode}. 10511@end deftypefn 10512 10513@defmac STORE_FLAG_VALUE 10514A C expression describing the value returned by a comparison operator 10515with an integral mode and stored by a store-flag instruction 10516(@samp{cstore@var{mode}4}) when the condition is true. This description must 10517apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the 10518comparison operators whose results have a @code{MODE_INT} mode. 10519 10520A value of 1 or @minus{}1 means that the instruction implementing the 10521comparison operator returns exactly 1 or @minus{}1 when the comparison is true 10522and 0 when the comparison is false. Otherwise, the value indicates 10523which bits of the result are guaranteed to be 1 when the comparison is 10524true. This value is interpreted in the mode of the comparison 10525operation, which is given by the mode of the first operand in the 10526@samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of 10527@code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by 10528the compiler. 10529 10530If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will 10531generate code that depends only on the specified bits. It can also 10532replace comparison operators with equivalent operations if they cause 10533the required bits to be set, even if the remaining bits are undefined. 10534For example, on a machine whose comparison operators return an 10535@code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as 10536@samp{0x80000000}, saying that just the sign bit is relevant, the 10537expression 10538 10539@smallexample 10540(ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0)) 10541@end smallexample 10542 10543@noindent 10544can be converted to 10545 10546@smallexample 10547(ashift:SI @var{x} (const_int @var{n})) 10548@end smallexample 10549 10550@noindent 10551where @var{n} is the appropriate shift count to move the bit being 10552tested into the sign bit. 10553 10554There is no way to describe a machine that always sets the low-order bit 10555for a true value, but does not guarantee the value of any other bits, 10556but we do not know of any machine that has such an instruction. If you 10557are trying to port GCC to such a machine, include an instruction to 10558perform a logical-and of the result with 1 in the pattern for the 10559comparison operators and let us know at @email{gcc@@gcc.gnu.org}. 10560 10561Often, a machine will have multiple instructions that obtain a value 10562from a comparison (or the condition codes). Here are rules to guide the 10563choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions 10564to be used: 10565 10566@itemize @bullet 10567@item 10568Use the shortest sequence that yields a valid definition for 10569@code{STORE_FLAG_VALUE}. It is more efficient for the compiler to 10570``normalize'' the value (convert it to, e.g., 1 or 0) than for the 10571comparison operators to do so because there may be opportunities to 10572combine the normalization with other operations. 10573 10574@item 10575For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being 10576slightly preferred on machines with expensive jumps and 1 preferred on 10577other machines. 10578 10579@item 10580As a second choice, choose a value of @samp{0x80000001} if instructions 10581exist that set both the sign and low-order bits but do not define the 10582others. 10583 10584@item 10585Otherwise, use a value of @samp{0x80000000}. 10586@end itemize 10587 10588Many machines can produce both the value chosen for 10589@code{STORE_FLAG_VALUE} and its negation in the same number of 10590instructions. On those machines, you should also define a pattern for 10591those cases, e.g., one matching 10592 10593@smallexample 10594(set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C}))) 10595@end smallexample 10596 10597Some machines can also perform @code{and} or @code{plus} operations on 10598condition code values with less instructions than the corresponding 10599@samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those 10600machines, define the appropriate patterns. Use the names @code{incscc} 10601and @code{decscc}, respectively, for the patterns which perform 10602@code{plus} or @code{minus} operations on condition code values. See 10603@file{rs6000.md} for some examples. The GNU Superoptimizer can be used to 10604find such instruction sequences on other machines. 10605 10606If this macro is not defined, the default value, 1, is used. You need 10607not define @code{STORE_FLAG_VALUE} if the machine has no store-flag 10608instructions, or if the value generated by these instructions is 1. 10609@end defmac 10610 10611@defmac FLOAT_STORE_FLAG_VALUE (@var{mode}) 10612A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is 10613returned when comparison operators with floating-point results are true. 10614Define this macro on machines that have comparison operations that return 10615floating-point values. If there are no such operations, do not define 10616this macro. 10617@end defmac 10618 10619@defmac VECTOR_STORE_FLAG_VALUE (@var{mode}) 10620A C expression that gives a rtx representing the nonzero true element 10621for vector comparisons. The returned rtx should be valid for the inner 10622mode of @var{mode} which is guaranteed to be a vector mode. Define 10623this macro on machines that have vector comparison operations that 10624return a vector result. If there are no such operations, do not define 10625this macro. Typically, this macro is defined as @code{const1_rtx} or 10626@code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent 10627the compiler optimizing such vector comparison operations for the 10628given mode. 10629@end defmac 10630 10631@defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) 10632@defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) 10633A C expression that indicates whether the architecture defines a value 10634for @code{clz} or @code{ctz} with a zero operand. 10635A result of @code{0} indicates the value is undefined. 10636If the value is defined for only the RTL expression, the macro should 10637evaluate to @code{1}; if the value applies also to the corresponding optab 10638entry (which is normally the case if it expands directly into 10639the corresponding RTL), then the macro should evaluate to @code{2}. 10640In the cases where the value is defined, @var{value} should be set to 10641this value. 10642 10643If this macro is not defined, the value of @code{clz} or 10644@code{ctz} at zero is assumed to be undefined. 10645 10646This macro must be defined if the target's expansion for @code{ffs} 10647relies on a particular value to get correct results. Otherwise it 10648is not necessary, though it may be used to optimize some corner cases, and 10649to provide a default expansion for the @code{ffs} optab. 10650 10651Note that regardless of this macro the ``definedness'' of @code{clz} 10652and @code{ctz} at zero do @emph{not} extend to the builtin functions 10653visible to the user. Thus one may be free to adjust the value at will 10654to match the target expansion of these operations without fear of 10655breaking the API@. 10656@end defmac 10657 10658@defmac Pmode 10659An alias for the machine mode for pointers. On most machines, define 10660this to be the integer mode corresponding to the width of a hardware 10661pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines. 10662On some machines you must define this to be one of the partial integer 10663modes, such as @code{PSImode}. 10664 10665The width of @code{Pmode} must be at least as large as the value of 10666@code{POINTER_SIZE}. If it is not equal, you must define the macro 10667@code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended 10668to @code{Pmode}. 10669@end defmac 10670 10671@defmac FUNCTION_MODE 10672An alias for the machine mode used for memory references to functions 10673being called, in @code{call} RTL expressions. On most CISC machines, 10674where an instruction can begin at any byte address, this should be 10675@code{QImode}. On most RISC machines, where all instructions have fixed 10676size and alignment, this should be a mode with the same size and alignment 10677as the machine instruction words - typically @code{SImode} or @code{HImode}. 10678@end defmac 10679 10680@defmac STDC_0_IN_SYSTEM_HEADERS 10681In normal operation, the preprocessor expands @code{__STDC__} to the 10682constant 1, to signify that GCC conforms to ISO Standard C@. On some 10683hosts, like Solaris, the system compiler uses a different convention, 10684where @code{__STDC__} is normally 0, but is 1 if the user specifies 10685strict conformance to the C Standard. 10686 10687Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host 10688convention when processing system header files, but when processing user 10689files @code{__STDC__} will always expand to 1. 10690@end defmac 10691 10692@deftypefn {C Target Hook} {const char *} TARGET_C_PREINCLUDE (void) 10693Define this hook to return the name of a header file to be included at the start of all compilations, as if it had been included with @code{#include <@var{file}>}. If this hook returns @code{NULL}, or is not defined, or the header is not found, or if the user specifies @option{-ffreestanding} or @option{-nostdinc}, no header is included. 10694 10695 This hook can be used together with a header provided by the system C library to implement ISO C requirements for certain macros to be predefined that describe properties of the whole implementation rather than just the compiler. 10696@end deftypefn 10697 10698@deftypefn {C Target Hook} bool TARGET_CXX_IMPLICIT_EXTERN_C (const char*@var{}) 10699Define this hook to add target-specific C++ implicit extern C functions. If this function returns true for the name of a file-scope function, that function implicitly gets extern "C" linkage rather than whatever language linkage the declaration would normally have. An example of such function is WinMain on Win32 targets. 10700@end deftypefn 10701 10702@defmac NO_IMPLICIT_EXTERN_C 10703Define this macro if the system header files support C++ as well as C@. 10704This macro inhibits the usual method of using system header files in 10705C++, which is to pretend that the file's contents are enclosed in 10706@samp{extern "C" @{@dots{}@}}. 10707@end defmac 10708 10709@findex #pragma 10710@findex pragma 10711@defmac REGISTER_TARGET_PRAGMAS () 10712Define this macro if you want to implement any target-specific pragmas. 10713If defined, it is a C expression which makes a series of calls to 10714@code{c_register_pragma} or @code{c_register_pragma_with_expansion} 10715for each pragma. The macro may also do any 10716setup required for the pragmas. 10717 10718The primary reason to define this macro is to provide compatibility with 10719other compilers for the same target. In general, we discourage 10720definition of target-specific pragmas for GCC@. 10721 10722If the pragma can be implemented by attributes then you should consider 10723defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well. 10724 10725Preprocessor macros that appear on pragma lines are not expanded. All 10726@samp{#pragma} directives that do not match any registered pragma are 10727silently ignored, unless the user specifies @option{-Wunknown-pragmas}. 10728@end defmac 10729 10730@deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) 10731@deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) 10732 10733Each call to @code{c_register_pragma} or 10734@code{c_register_pragma_with_expansion} establishes one pragma. The 10735@var{callback} routine will be called when the preprocessor encounters a 10736pragma of the form 10737 10738@smallexample 10739#pragma [@var{space}] @var{name} @dots{} 10740@end smallexample 10741 10742@var{space} is the case-sensitive namespace of the pragma, or 10743@code{NULL} to put the pragma in the global namespace. The callback 10744routine receives @var{pfile} as its first argument, which can be passed 10745on to cpplib's functions if necessary. You can lex tokens after the 10746@var{name} by calling @code{pragma_lex}. Tokens that are not read by the 10747callback will be silently ignored. The end of the line is indicated by 10748a token of type @code{CPP_EOF}. Macro expansion occurs on the 10749arguments of pragmas registered with 10750@code{c_register_pragma_with_expansion} but not on the arguments of 10751pragmas registered with @code{c_register_pragma}. 10752 10753Note that the use of @code{pragma_lex} is specific to the C and C++ 10754compilers. It will not work in the Java or Fortran compilers, or any 10755other language compilers for that matter. Thus if @code{pragma_lex} is going 10756to be called from target-specific code, it must only be done so when 10757building the C and C++ compilers. This can be done by defining the 10758variables @code{c_target_objs} and @code{cxx_target_objs} in the 10759target entry in the @file{config.gcc} file. These variables should name 10760the target-specific, language-specific object file which contains the 10761code that uses @code{pragma_lex}. Note it will also be necessary to add a 10762rule to the makefile fragment pointed to by @code{tmake_file} that shows 10763how to build this object file. 10764@end deftypefun 10765 10766@defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION 10767Define this macro if macros should be expanded in the 10768arguments of @samp{#pragma pack}. 10769@end defmac 10770 10771@defmac TARGET_DEFAULT_PACK_STRUCT 10772If your target requires a structure packing default other than 0 (meaning 10773the machine default), define this macro to the necessary value (in bytes). 10774This must be a value that would also be valid to use with 10775@samp{#pragma pack()} (that is, a small power of two). 10776@end defmac 10777 10778@defmac DOLLARS_IN_IDENTIFIERS 10779Define this macro to control use of the character @samp{$} in 10780identifier names for the C family of languages. 0 means @samp{$} is 10781not allowed by default; 1 means it is allowed. 1 is the default; 10782there is no need to define this macro in that case. 10783@end defmac 10784 10785@defmac INSN_SETS_ARE_DELAYED (@var{insn}) 10786Define this macro as a C expression that is nonzero if it is safe for the 10787delay slot scheduler to place instructions in the delay slot of @var{insn}, 10788even if they appear to use a resource set or clobbered in @var{insn}. 10789@var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that 10790every @code{call_insn} has this behavior. On machines where some @code{insn} 10791or @code{jump_insn} is really a function call and hence has this behavior, 10792you should define this macro. 10793 10794You need not define this macro if it would always return zero. 10795@end defmac 10796 10797@defmac INSN_REFERENCES_ARE_DELAYED (@var{insn}) 10798Define this macro as a C expression that is nonzero if it is safe for the 10799delay slot scheduler to place instructions in the delay slot of @var{insn}, 10800even if they appear to set or clobber a resource referenced in @var{insn}. 10801@var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where 10802some @code{insn} or @code{jump_insn} is really a function call and its operands 10803are registers whose use is actually in the subroutine it calls, you should 10804define this macro. Doing so allows the delay slot scheduler to move 10805instructions which copy arguments into the argument registers into the delay 10806slot of @var{insn}. 10807 10808You need not define this macro if it would always return zero. 10809@end defmac 10810 10811@defmac MULTIPLE_SYMBOL_SPACES 10812Define this macro as a C expression that is nonzero if, in some cases, 10813global symbols from one translation unit may not be bound to undefined 10814symbols in another translation unit without user intervention. For 10815instance, under Microsoft Windows symbols must be explicitly imported 10816from shared libraries (DLLs). 10817 10818You need not define this macro if it would always evaluate to zero. 10819@end defmac 10820 10821@deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers}) 10822This target hook should add to @var{clobbers} @code{STRING_CST} trees for 10823any hard regs the port wishes to automatically clobber for an asm. 10824It should return the result of the last @code{tree_cons} used to add a 10825clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the 10826corresponding parameters to the asm and may be inspected to avoid 10827clobbering a register that is an input or output of the asm. You can use 10828@code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test 10829for overlap with regards to asm-declared registers. 10830@end deftypefn 10831 10832@defmac MATH_LIBRARY 10833Define this macro as a C string constant for the linker argument to link 10834in the system math library, minus the initial @samp{"-l"}, or 10835@samp{""} if the target does not have a 10836separate math library. 10837 10838You need only define this macro if the default of @samp{"m"} is wrong. 10839@end defmac 10840 10841@defmac LIBRARY_PATH_ENV 10842Define this macro as a C string constant for the environment variable that 10843specifies where the linker should look for libraries. 10844 10845You need only define this macro if the default of @samp{"LIBRARY_PATH"} 10846is wrong. 10847@end defmac 10848 10849@defmac TARGET_POSIX_IO 10850Define this macro if the target supports the following POSIX@ file 10851functions, access, mkdir and file locking with fcntl / F_SETLKW@. 10852Defining @code{TARGET_POSIX_IO} will enable the test coverage code 10853to use file locking when exiting a program, which avoids race conditions 10854if the program has forked. It will also create directories at run-time 10855for cross-profiling. 10856@end defmac 10857 10858@defmac MAX_CONDITIONAL_EXECUTE 10859 10860A C expression for the maximum number of instructions to execute via 10861conditional execution instructions instead of a branch. A value of 10862@code{BRANCH_COST}+1 is the default if the machine does not use cc0, and 108631 if it does use cc0. 10864@end defmac 10865 10866@defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr}) 10867Used if the target needs to perform machine-dependent modifications on the 10868conditionals used for turning basic blocks into conditionally executed code. 10869@var{ce_info} points to a data structure, @code{struct ce_if_block}, which 10870contains information about the currently processed blocks. @var{true_expr} 10871and @var{false_expr} are the tests that are used for converting the 10872then-block and the else-block, respectively. Set either @var{true_expr} or 10873@var{false_expr} to a null pointer if the tests cannot be converted. 10874@end defmac 10875 10876@defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr}) 10877Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated 10878if-statements into conditions combined by @code{and} and @code{or} operations. 10879@var{bb} contains the basic block that contains the test that is currently 10880being processed and about to be turned into a condition. 10881@end defmac 10882 10883@defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn}) 10884A C expression to modify the @var{PATTERN} of an @var{INSN} that is to 10885be converted to conditional execution format. @var{ce_info} points to 10886a data structure, @code{struct ce_if_block}, which contains information 10887about the currently processed blocks. 10888@end defmac 10889 10890@defmac IFCVT_MODIFY_FINAL (@var{ce_info}) 10891A C expression to perform any final machine dependent modifications in 10892converting code to conditional execution. The involved basic blocks 10893can be found in the @code{struct ce_if_block} structure that is pointed 10894to by @var{ce_info}. 10895@end defmac 10896 10897@defmac IFCVT_MODIFY_CANCEL (@var{ce_info}) 10898A C expression to cancel any machine dependent modifications in 10899converting code to conditional execution. The involved basic blocks 10900can be found in the @code{struct ce_if_block} structure that is pointed 10901to by @var{ce_info}. 10902@end defmac 10903 10904@defmac IFCVT_MACHDEP_INIT (@var{ce_info}) 10905A C expression to initialize any machine specific data for if-conversion 10906of the if-block in the @code{struct ce_if_block} structure that is pointed 10907to by @var{ce_info}. 10908@end defmac 10909 10910@deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void) 10911If non-null, this hook performs a target-specific pass over the 10912instruction stream. The compiler will run it at all optimization levels, 10913just before the point at which it normally does delayed-branch scheduling. 10914 10915The exact purpose of the hook varies from target to target. Some use 10916it to do transformations that are necessary for correctness, such as 10917laying out in-function constant pools or avoiding hardware hazards. 10918Others use it as an opportunity to do some machine-dependent optimizations. 10919 10920You need not implement the hook if it has nothing to do. The default 10921definition is null. 10922@end deftypefn 10923 10924@deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void) 10925Define this hook if you have any machine-specific built-in functions 10926that need to be defined. It should be a function that performs the 10927necessary setup. 10928 10929Machine specific built-in functions can be useful to expand special machine 10930instructions that would otherwise not normally be generated because 10931they have no equivalent in the source language (for example, SIMD vector 10932instructions or prefetch instructions). 10933 10934To create a built-in function, call the function 10935@code{lang_hooks.builtin_function} 10936which is defined by the language front end. You can use any type nodes set 10937up by @code{build_common_tree_nodes}; 10938only language front ends that use those two functions will call 10939@samp{TARGET_INIT_BUILTINS}. 10940@end deftypefn 10941 10942@deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p}) 10943Define this hook if you have any machine-specific built-in functions 10944that need to be defined. It should be a function that returns the 10945builtin function declaration for the builtin function code @var{code}. 10946If there is no such builtin and it cannot be initialized at this time 10947if @var{initialize_p} is true the function should return @code{NULL_TREE}. 10948If @var{code} is out of range the function should return 10949@code{error_mark_node}. 10950@end deftypefn 10951 10952@deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, machine_mode @var{mode}, int @var{ignore}) 10953 10954Expand a call to a machine specific built-in function that was set up by 10955@samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the 10956function call; the result should go to @var{target} if that is 10957convenient, and have mode @var{mode} if that is convenient. 10958@var{subtarget} may be used as the target for computing one of 10959@var{exp}'s operands. @var{ignore} is nonzero if the value is to be 10960ignored. This function should return the result of the call to the 10961built-in function. 10962@end deftypefn 10963 10964@deftypefn {Target Hook} tree TARGET_BUILTIN_CHKP_FUNCTION (unsigned @var{fcode}) 10965This hook allows target to redefine built-in functions used by 10966Pointer Bounds Checker for code instrumentation. Hook should return 10967fndecl of function implementing generic builtin whose code is 10968passed in @var{fcode}. Currently following built-in functions are 10969obtained using this hook: 10970@deftypefn {Built-in Function} __bounds_type __chkp_bndmk (const void *@var{lb}, size_t @var{size}) 10971Function code - BUILT_IN_CHKP_BNDMK. This built-in function is used 10972by Pointer Bounds Checker to create bound values. @var{lb} holds low 10973bound of the resulting bounds. @var{size} holds size of created bounds. 10974@end deftypefn 10975 10976@deftypefn {Built-in Function} void __chkp_bndstx (const void *@var{ptr}, __bounds_type @var{b}, const void **@var{loc}) 10977Function code - @code{BUILT_IN_CHKP_BNDSTX}. This built-in function is used 10978by Pointer Bounds Checker to store bounds @var{b} for pointer @var{ptr} 10979when @var{ptr} is stored by address @var{loc}. 10980@end deftypefn 10981 10982@deftypefn {Built-in Function} __bounds_type __chkp_bndldx (const void **@var{loc}, const void *@var{ptr}) 10983Function code - @code{BUILT_IN_CHKP_BNDLDX}. This built-in function is used 10984by Pointer Bounds Checker to get bounds of pointer @var{ptr} loaded by 10985address @var{loc}. 10986@end deftypefn 10987 10988@deftypefn {Built-in Function} void __chkp_bndcl (const void *@var{ptr}, __bounds_type @var{b}) 10989Function code - @code{BUILT_IN_CHKP_BNDCL}. This built-in function is used 10990by Pointer Bounds Checker to perform check for pointer @var{ptr} against 10991lower bound of bounds @var{b}. 10992@end deftypefn 10993 10994@deftypefn {Built-in Function} void __chkp_bndcu (const void *@var{ptr}, __bounds_type @var{b}) 10995Function code - @code{BUILT_IN_CHKP_BNDCU}. This built-in function is used 10996by Pointer Bounds Checker to perform check for pointer @var{ptr} against 10997upper bound of bounds @var{b}. 10998@end deftypefn 10999 11000@deftypefn {Built-in Function} __bounds_type __chkp_bndret (void *@var{ptr}) 11001Function code - @code{BUILT_IN_CHKP_BNDRET}. This built-in function is used 11002by Pointer Bounds Checker to obtain bounds returned by a call statement. 11003@var{ptr} passed to built-in is @code{SSA_NAME} returned by the call. 11004@end deftypefn 11005 11006@deftypefn {Built-in Function} __bounds_type __chkp_intersect (__bounds_type @var{b1}, __bounds_type @var{b2}) 11007Function code - @code{BUILT_IN_CHKP_INTERSECT}. This built-in function 11008returns intersection of bounds @var{b1} and @var{b2}. 11009@end deftypefn 11010 11011@deftypefn {Built-in Function} __bounds_type __chkp_narrow (const void *@var{ptr}, __bounds_type @var{b}, size_t @var{s}) 11012Function code - @code{BUILT_IN_CHKP_NARROW}. This built-in function 11013returns intersection of bounds @var{b} and 11014[@var{ptr}, @var{ptr} + @var{s} - @code{1}]. 11015@end deftypefn 11016 11017@deftypefn {Built-in Function} size_t __chkp_sizeof (const void *@var{ptr}) 11018Function code - @code{BUILT_IN_CHKP_SIZEOF}. This built-in function 11019returns size of object referenced by @var{ptr}. @var{ptr} is always 11020@code{ADDR_EXPR} of @code{VAR_DECL}. This built-in is used by 11021Pointer Bounds Checker when bounds of object cannot be computed statically 11022(e.g. object has incomplete type). 11023@end deftypefn 11024 11025@deftypefn {Built-in Function} const void *__chkp_extract_lower (__bounds_type @var{b}) 11026Function code - @code{BUILT_IN_CHKP_EXTRACT_LOWER}. This built-in function 11027returns lower bound of bounds @var{b}. 11028@end deftypefn 11029 11030@deftypefn {Built-in Function} const void *__chkp_extract_upper (__bounds_type @var{b}) 11031Function code - @code{BUILT_IN_CHKP_EXTRACT_UPPER}. This built-in function 11032returns upper bound of bounds @var{b}. 11033@end deftypefn 11034@end deftypefn 11035@deftypefn {Target Hook} tree TARGET_CHKP_BOUND_TYPE (void) 11036Return type to be used for bounds 11037@end deftypefn 11038@deftypefn {Target Hook} {enum machine_mode} TARGET_CHKP_BOUND_MODE (void) 11039Return mode to be used for bounds. 11040@end deftypefn 11041@deftypefn {Target Hook} tree TARGET_CHKP_MAKE_BOUNDS_CONSTANT (HOST_WIDE_INT @var{lb}, HOST_WIDE_INT @var{ub}) 11042Return constant used to statically initialize constant bounds 11043with specified lower bound @var{lb} and upper bounds @var{ub}. 11044@end deftypefn 11045@deftypefn {Target Hook} int TARGET_CHKP_INITIALIZE_BOUNDS (tree @var{var}, tree @var{lb}, tree @var{ub}, tree *@var{stmts}) 11046Generate a list of statements @var{stmts} to initialize pointer 11047bounds variable @var{var} with bounds @var{lb} and @var{ub}. Return 11048the number of generated statements. 11049@end deftypefn 11050 11051@deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist}) 11052Select a replacement for a machine specific built-in function that 11053was set up by @samp{TARGET_INIT_BUILTINS}. This is done 11054@emph{before} regular type checking, and so allows the target to 11055implement a crude form of function overloading. @var{fndecl} is the 11056declaration of the built-in function. @var{arglist} is the list of 11057arguments passed to the built-in function. The result is a 11058complete expression that implements the operation, usually 11059another @code{CALL_EXPR}. 11060@var{arglist} really has type @samp{VEC(tree,gc)*} 11061@end deftypefn 11062 11063@deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore}) 11064Fold a call to a machine specific built-in function that was set up by 11065@samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the 11066built-in function. @var{n_args} is the number of arguments passed to 11067the function; the arguments themselves are pointed to by @var{argp}. 11068The result is another tree, valid for both GIMPLE and GENERIC, 11069containing a simplified expression for the call's result. If 11070@var{ignore} is true the value will be ignored. 11071@end deftypefn 11072 11073@deftypefn {Target Hook} bool TARGET_GIMPLE_FOLD_BUILTIN (gimple_stmt_iterator *@var{gsi}) 11074Fold a call to a machine specific built-in function that was set up 11075by @samp{TARGET_INIT_BUILTINS}. @var{gsi} points to the gimple 11076statement holding the function call. Returns true if any change 11077was made to the GIMPLE stream. 11078@end deftypefn 11079 11080@deftypefn {Target Hook} int TARGET_COMPARE_VERSION_PRIORITY (tree @var{decl1}, tree @var{decl2}) 11081This hook is used to compare the target attributes in two functions to 11082determine which function's features get higher priority. This is used 11083during function multi-versioning to figure out the order in which two 11084versions must be dispatched. A function version with a higher priority 11085is checked for dispatching earlier. @var{decl1} and @var{decl2} are 11086 the two function decls that will be compared. 11087@end deftypefn 11088 11089@deftypefn {Target Hook} tree TARGET_GET_FUNCTION_VERSIONS_DISPATCHER (void *@var{decl}) 11090This hook is used to get the dispatcher function for a set of function 11091versions. The dispatcher function is called to invoke the right function 11092version at run-time. @var{decl} is one version from a set of semantically 11093identical versions. 11094@end deftypefn 11095 11096@deftypefn {Target Hook} tree TARGET_GENERATE_VERSION_DISPATCHER_BODY (void *@var{arg}) 11097This hook is used to generate the dispatcher logic to invoke the right 11098function version at run-time for a given set of function versions. 11099@var{arg} points to the callgraph node of the dispatcher function whose 11100body must be generated. 11101@end deftypefn 11102 11103@deftypefn {Target Hook} bool TARGET_CAN_USE_DOLOOP_P (const widest_int @var{&iterations}, const widest_int @var{&iterations_max}, unsigned int @var{loop_depth}, bool @var{entered_at_top}) 11104Return true if it is possible to use low-overhead loops (@code{doloop_end} 11105and @code{doloop_begin}) for a particular loop. @var{iterations} gives the 11106exact number of iterations, or 0 if not known. @var{iterations_max} gives 11107the maximum number of iterations, or 0 if not known. @var{loop_depth} is 11108the nesting depth of the loop, with 1 for innermost loops, 2 for loops that 11109contain innermost loops, and so on. @var{entered_at_top} is true if the 11110loop is only entered from the top. 11111 11112This hook is only used if @code{doloop_end} is available. The default 11113implementation returns true. You can use @code{can_use_doloop_if_innermost} 11114if the loop must be the innermost, and if there are no other restrictions. 11115@end deftypefn 11116 11117@deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const rtx_insn *@var{insn}) 11118 11119Take an instruction in @var{insn} and return NULL if it is valid within a 11120low-overhead loop, otherwise return a string explaining why doloop 11121could not be applied. 11122 11123Many targets use special registers for low-overhead looping. For any 11124instruction that clobbers these this function should return a string indicating 11125the reason why the doloop could not be applied. 11126By default, the RTL loop optimizer does not use a present doloop pattern for 11127loops containing function calls or branch on table instructions. 11128@end deftypefn 11129 11130@deftypefn {Target Hook} bool TARGET_LEGITIMATE_COMBINED_INSN (rtx_insn *@var{insn}) 11131Take an instruction in @var{insn} and return @code{false} if the instruction is not appropriate as a combination of two or more instructions. The default is to accept all instructions. 11132@end deftypefn 11133 11134@deftypefn {Target Hook} bool TARGET_CAN_FOLLOW_JUMP (const rtx_insn *@var{follower}, const rtx_insn *@var{followee}) 11135FOLLOWER and FOLLOWEE are JUMP_INSN instructions; return true if FOLLOWER may be modified to follow FOLLOWEE; false, if it can't. For example, on some targets, certain kinds of branches can't be made to follow through a hot/cold partitioning. 11136@end deftypefn 11137 11138@deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code}) 11139This target hook returns @code{true} if @var{x} is considered to be commutative. 11140Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider 11141PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code 11142of the enclosing rtl, if known, otherwise it is UNKNOWN. 11143@end deftypefn 11144 11145@deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg}) 11146 11147When the initial value of a hard register has been copied in a pseudo 11148register, it is often not necessary to actually allocate another register 11149to this pseudo register, because the original hard register or a stack slot 11150it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE} 11151is called at the start of register allocation once for each hard register 11152that had its initial value copied by using 11153@code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}. 11154Possible values are @code{NULL_RTX}, if you don't want 11155to do any special allocation, a @code{REG} rtx---that would typically be 11156the hard register itself, if it is known not to be clobbered---or a 11157@code{MEM}. 11158If you are returning a @code{MEM}, this is only a hint for the allocator; 11159it might decide to use another register anyways. 11160You may use @code{current_function_is_leaf} or 11161@code{REG_N_SETS} in the hook to determine if the hard 11162register in question will not be clobbered. 11163The default value of this hook is @code{NULL}, which disables any special 11164allocation. 11165@end deftypefn 11166 11167@deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags}) 11168This target hook returns nonzero if @var{x}, an @code{unspec} or 11169@code{unspec_volatile} operation, might cause a trap. Targets can use 11170this hook to enhance precision of analysis for @code{unspec} and 11171@code{unspec_volatile} operations. You may call @code{may_trap_p_1} 11172to analyze inner elements of @var{x} in which case @var{flags} should be 11173passed along. 11174@end deftypefn 11175 11176@deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl}) 11177The compiler invokes this hook whenever it changes its current function 11178context (@code{cfun}). You can define this function if 11179the back end needs to perform any initialization or reset actions on a 11180per-function basis. For example, it may be used to implement function 11181attributes that affect register usage or code generation patterns. 11182The argument @var{decl} is the declaration for the new function context, 11183and may be null to indicate that the compiler has left a function context 11184and is returning to processing at the top level. 11185The default hook function does nothing. 11186 11187GCC sets @code{cfun} to a dummy function context during initialization of 11188some parts of the back end. The hook function is not invoked in this 11189situation; you need not worry about the hook being invoked recursively, 11190or when the back end is in a partially-initialized state. 11191@code{cfun} might be @code{NULL} to indicate processing at top level, 11192outside of any function scope. 11193@end deftypefn 11194 11195@defmac TARGET_OBJECT_SUFFIX 11196Define this macro to be a C string representing the suffix for object 11197files on your target machine. If you do not define this macro, GCC will 11198use @samp{.o} as the suffix for object files. 11199@end defmac 11200 11201@defmac TARGET_EXECUTABLE_SUFFIX 11202Define this macro to be a C string representing the suffix to be 11203automatically added to executable files on your target machine. If you 11204do not define this macro, GCC will use the null string as the suffix for 11205executable files. 11206@end defmac 11207 11208@defmac COLLECT_EXPORT_LIST 11209If defined, @code{collect2} will scan the individual object files 11210specified on its command line and create an export list for the linker. 11211Define this macro for systems like AIX, where the linker discards 11212object files that are not referenced from @code{main} and uses export 11213lists. 11214@end defmac 11215 11216@defmac MODIFY_JNI_METHOD_CALL (@var{mdecl}) 11217Define this macro to a C expression representing a variant of the 11218method call @var{mdecl}, if Java Native Interface (JNI) methods 11219must be invoked differently from other methods on your target. 11220For example, on 32-bit Microsoft Windows, JNI methods must be invoked using 11221the @code{stdcall} calling convention and this macro is then 11222defined as this expression: 11223 11224@smallexample 11225build_type_attribute_variant (@var{mdecl}, 11226 build_tree_list 11227 (get_identifier ("stdcall"), 11228 NULL)) 11229@end smallexample 11230@end defmac 11231 11232@deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void) 11233This target hook returns @code{true} past the point in which new jump 11234instructions could be created. On machines that require a register for 11235every jump such as the SHmedia ISA of SH5, this point would typically be 11236reload, so this target hook should be defined to a function such as: 11237 11238@smallexample 11239static bool 11240cannot_modify_jumps_past_reload_p () 11241@{ 11242 return (reload_completed || reload_in_progress); 11243@} 11244@end smallexample 11245@end deftypefn 11246 11247@deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void) 11248This target hook returns a register class for which branch target register 11249optimizations should be applied. All registers in this class should be 11250usable interchangeably. After reload, registers in this class will be 11251re-allocated and loads will be hoisted out of loops and be subjected 11252to inter-block scheduling. 11253@end deftypefn 11254 11255@deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen}) 11256Branch target register optimization will by default exclude callee-saved 11257registers 11258that are not already live during the current function; if this target hook 11259returns true, they will be included. The target code must than make sure 11260that all target registers in the class returned by 11261@samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are 11262saved. @var{after_prologue_epilogue_gen} indicates if prologues and 11263epilogues have already been generated. Note, even if you only return 11264true when @var{after_prologue_epilogue_gen} is false, you still are likely 11265to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET} 11266to reserve space for caller-saved target registers. 11267@end deftypefn 11268 11269@deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void) 11270This target hook returns true if the target supports conditional execution. 11271This target hook is required only when the target has several different 11272modes and they have different conditional execution capability, such as ARM. 11273@end deftypefn 11274 11275@deftypefn {Target Hook} rtx TARGET_GEN_CCMP_FIRST (rtx *@var{prep_seq}, rtx *@var{gen_seq}, int @var{code}, tree @var{op0}, tree @var{op1}) 11276This function prepares to emit a comparison insn for the first compare in a 11277 sequence of conditional comparisions. It returns a appropriate @code{CC} 11278 for passing to @code{gen_ccmp_next} or @code{cbranch_optab}. The insns to 11279 prepare the compare are saved in @var{prep_seq} and the compare insns are 11280 saved in @var{gen_seq}. They will be emitted when all the compares in the 11281 the conditional comparision are generated without error. @var{code} is 11282 the @code{rtx_code} of the compare for @var{op0} and @var{op1}. 11283@end deftypefn 11284 11285@deftypefn {Target Hook} rtx TARGET_GEN_CCMP_NEXT (rtx *@var{prep_seq}, rtx *@var{gen_seq}, rtx @var{prev}, int @var{cmp_code}, tree @var{op0}, tree @var{op1}, int @var{bit_code}) 11286This function prepare to emit a conditional comparison within a sequence of 11287 conditional comparisons. It returns a appropriate @code{CC} for passing to 11288 @code{gen_ccmp_next} or @code{cbranch_optab}. The insns to prepare the 11289 compare are saved in @var{prep_seq} and the compare insns are saved in 11290 @var{gen_seq}. They will be emitted when all the compares in the conditional 11291 comparision are generated without error. The @var{prev} expression is the 11292 result of a prior call to @code{gen_ccmp_first} or @code{gen_ccmp_next}. It 11293 may return @code{NULL} if the combination of @var{prev} and this comparison is 11294 not supported, otherwise the result must be appropriate for passing to 11295 @code{gen_ccmp_next} or @code{cbranch_optab}. @var{code} is the 11296 @code{rtx_code} of the compare for @var{op0} and @var{op1}. @var{bit_code} 11297 is @code{AND} or @code{IOR}, which is the op on the two compares. 11298@end deftypefn 11299 11300@deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop}) 11301This target hook returns a new value for the number of times @var{loop} 11302should be unrolled. The parameter @var{nunroll} is the number of times 11303the loop is to be unrolled. The parameter @var{loop} is a pointer to 11304the loop, which is going to be checked for unrolling. This target hook 11305is required only when the target has special constraints like maximum 11306number of memory accesses. 11307@end deftypefn 11308 11309@defmac POWI_MAX_MULTS 11310If defined, this macro is interpreted as a signed integer C expression 11311that specifies the maximum number of floating point multiplications 11312that should be emitted when expanding exponentiation by an integer 11313constant inline. When this value is defined, exponentiation requiring 11314more than this number of multiplications is implemented by calling the 11315system library's @code{pow}, @code{powf} or @code{powl} routines. 11316The default value places no upper bound on the multiplication count. 11317@end defmac 11318 11319@deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc}) 11320This target hook should register any extra include files for the 11321target. The parameter @var{stdinc} indicates if normal include files 11322are present. The parameter @var{sysroot} is the system root directory. 11323The parameter @var{iprefix} is the prefix for the gcc directory. 11324@end deftypefn 11325 11326@deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc}) 11327This target hook should register any extra include files for the 11328target before any standard headers. The parameter @var{stdinc} 11329indicates if normal include files are present. The parameter 11330@var{sysroot} is the system root directory. The parameter 11331@var{iprefix} is the prefix for the gcc directory. 11332@end deftypefn 11333 11334@deftypefn Macro void TARGET_OPTF (char *@var{path}) 11335This target hook should register special include paths for the target. 11336The parameter @var{path} is the include to register. On Darwin 11337systems, this is used for Framework includes, which have semantics 11338that are different from @option{-I}. 11339@end deftypefn 11340 11341@defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl}) 11342This target macro returns @code{true} if it is safe to use a local alias 11343for a virtual function @var{fndecl} when constructing thunks, 11344@code{false} otherwise. By default, the macro returns @code{true} for all 11345functions, if a target supports aliases (i.e.@: defines 11346@code{ASM_OUTPUT_DEF}), @code{false} otherwise, 11347@end defmac 11348 11349@defmac TARGET_FORMAT_TYPES 11350If defined, this macro is the name of a global variable containing 11351target-specific format checking information for the @option{-Wformat} 11352option. The default is to have no target-specific format checks. 11353@end defmac 11354 11355@defmac TARGET_N_FORMAT_TYPES 11356If defined, this macro is the number of entries in 11357@code{TARGET_FORMAT_TYPES}. 11358@end defmac 11359 11360@defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES 11361If defined, this macro is the name of a global variable containing 11362target-specific format overrides for the @option{-Wformat} option. The 11363default is to have no target-specific format overrides. If defined, 11364@code{TARGET_FORMAT_TYPES} must be defined, too. 11365@end defmac 11366 11367@defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT 11368If defined, this macro specifies the number of entries in 11369@code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}. 11370@end defmac 11371 11372@defmac TARGET_OVERRIDES_FORMAT_INIT 11373If defined, this macro specifies the optional initialization 11374routine for target specific customizations of the system printf 11375and scanf formatter settings. 11376@end defmac 11377 11378@deftypevr {Target Hook} bool TARGET_RELAXED_ORDERING 11379If set to @code{true}, means that the target's memory model does not 11380guarantee that loads which do not depend on one another will access 11381main memory in the order of the instruction stream; if ordering is 11382important, an explicit memory barrier must be used. This is true of 11383many recent processors which implement a policy of ``relaxed,'' 11384``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC, 11385and ia64. The default is @code{false}. 11386@end deftypevr 11387 11388@deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val}) 11389If defined, this macro returns the diagnostic message when it is 11390illegal to pass argument @var{val} to function @var{funcdecl} 11391with prototype @var{typelist}. 11392@end deftypefn 11393 11394@deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype}) 11395If defined, this macro returns the diagnostic message when it is 11396invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL} 11397if validity should be determined by the front end. 11398@end deftypefn 11399 11400@deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type}) 11401If defined, this macro returns the diagnostic message when it is 11402invalid to apply operation @var{op} (where unary plus is denoted by 11403@code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL} 11404if validity should be determined by the front end. 11405@end deftypefn 11406 11407@deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2}) 11408If defined, this macro returns the diagnostic message when it is 11409invalid to apply operation @var{op} to operands of types @var{type1} 11410and @var{type2}, or @code{NULL} if validity should be determined by 11411the front end. 11412@end deftypefn 11413 11414@deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (const_tree @var{type}) 11415If defined, this macro returns the diagnostic message when it is 11416invalid for functions to include parameters of type @var{type}, 11417or @code{NULL} if validity should be determined by 11418the front end. This is currently used only by the C and C++ front ends. 11419@end deftypefn 11420 11421@deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (const_tree @var{type}) 11422If defined, this macro returns the diagnostic message when it is 11423invalid for functions to have return type @var{type}, 11424or @code{NULL} if validity should be determined by 11425the front end. This is currently used only by the C and C++ front ends. 11426@end deftypefn 11427 11428@deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type}) 11429If defined, this target hook returns the type to which values of 11430@var{type} should be promoted when they appear in expressions, 11431analogous to the integer promotions, or @code{NULL_TREE} to use the 11432front end's normal promotion rules. This hook is useful when there are 11433target-specific types with special promotion rules. 11434This is currently used only by the C and C++ front ends. 11435@end deftypefn 11436 11437@deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr}) 11438If defined, this hook returns the result of converting @var{expr} to 11439@var{type}. It should return the converted expression, 11440or @code{NULL_TREE} to apply the front end's normal conversion rules. 11441This hook is useful when there are target-specific types with special 11442conversion rules. 11443This is currently used only by the C and C++ front ends. 11444@end deftypefn 11445 11446@defmac TARGET_USE_JCR_SECTION 11447This macro determines whether to use the JCR section to register Java 11448classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both 11449SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0. 11450@end defmac 11451 11452@defmac OBJC_JBLEN 11453This macro determines the size of the objective C jump buffer for the 11454NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value. 11455@end defmac 11456 11457@defmac LIBGCC2_UNWIND_ATTRIBUTE 11458Define this macro if any target-specific attributes need to be attached 11459to the functions in @file{libgcc} that provide low-level support for 11460call stack unwinding. It is used in declarations in @file{unwind-generic.h} 11461and the associated definitions of those functions. 11462@end defmac 11463 11464@deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void) 11465Define this macro to update the current function stack boundary if 11466necessary. 11467@end deftypefn 11468 11469@deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void) 11470This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a 11471different argument pointer register is needed to access the function's 11472argument list due to stack realignment. Return @code{NULL} if no DRAP 11473is needed. 11474@end deftypefn 11475 11476@deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void) 11477When optimization is disabled, this hook indicates whether or not 11478arguments should be allocated to stack slots. Normally, GCC allocates 11479stacks slots for arguments when not optimizing in order to make 11480debugging easier. However, when a function is declared with 11481@code{__attribute__((naked))}, there is no stack frame, and the compiler 11482cannot safely move arguments from the registers in which they are passed 11483to the stack. Therefore, this hook should return true in general, but 11484false for naked functions. The default implementation always returns true. 11485@end deftypefn 11486 11487@deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR 11488On some architectures it can take multiple instructions to synthesize 11489a constant. If there is another constant already in a register that 11490is close enough in value then it is preferable that the new constant 11491is computed from this register using immediate addition or 11492subtraction. We accomplish this through CSE. Besides the value of 11493the constant we also add a lower and an upper constant anchor to the 11494available expressions. These are then queried when encountering new 11495constants. The anchors are computed by rounding the constant up and 11496down to a multiple of the value of @code{TARGET_CONST_ANCHOR}. 11497@code{TARGET_CONST_ANCHOR} should be the maximum positive value 11498accepted by immediate-add plus one. We currently assume that the 11499value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on 11500MIPS, where add-immediate takes a 16-bit signed value, 11501@code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value 11502is zero, which disables this optimization. 11503@end deftypevr 11504 11505@deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_ASAN_SHADOW_OFFSET (void) 11506Return the offset bitwise ored into shifted address to get corresponding 11507Address Sanitizer shadow memory address. NULL if Address Sanitizer is not 11508supported by the target. 11509@end deftypefn 11510 11511@deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_MEMMODEL_CHECK (unsigned HOST_WIDE_INT @var{val}) 11512Validate target specific memory model mask bits. When NULL no target specific 11513memory model bits are allowed. 11514@end deftypefn 11515 11516@deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL 11517This value should be set if the result written by @code{atomic_test_and_set} is not exactly 1, i.e. the @code{bool} @code{true}. 11518@end deftypevr 11519 11520@deftypefn {Target Hook} bool TARGET_HAS_IFUNC_P (void) 11521It returns true if the target supports GNU indirect functions. 11522The support includes the assembler, linker and dynamic linker. 11523The default value of this hook is based on target's libc. 11524@end deftypefn 11525 11526@deftypefn {Target Hook} {unsigned int} TARGET_ATOMIC_ALIGN_FOR_MODE (machine_mode @var{mode}) 11527If defined, this function returns an appropriate alignment in bits for an atomic object of machine_mode @var{mode}. If 0 is returned then the default alignment for the specified mode is used. 11528@end deftypefn 11529 11530@deftypefn {Target Hook} void TARGET_ATOMIC_ASSIGN_EXPAND_FENV (tree *@var{hold}, tree *@var{clear}, tree *@var{update}) 11531ISO C11 requires atomic compound assignments that may raise floating-point exceptions to raise exceptions corresponding to the arithmetic operation whose result was successfully stored in a compare-and-exchange sequence. This requires code equivalent to calls to @code{feholdexcept}, @code{feclearexcept} and @code{feupdateenv} to be generated at appropriate points in the compare-and-exchange sequence. This hook should set @code{*@var{hold}} to an expression equivalent to the call to @code{feholdexcept}, @code{*@var{clear}} to an expression equivalent to the call to @code{feclearexcept} and @code{*@var{update}} to an expression equivalent to the call to @code{feupdateenv}. The three expressions are @code{NULL_TREE} on entry to the hook and may be left as @code{NULL_TREE} if no code is required in a particular place. The default implementation leaves all three expressions as @code{NULL_TREE}. The @code{__atomic_feraiseexcept} function from @code{libatomic} may be of use as part of the code generated in @code{*@var{update}}. 11532@end deftypefn 11533 11534@deftypefn {Target Hook} void TARGET_RECORD_OFFLOAD_SYMBOL (tree) 11535Used when offloaded functions are seen in the compilation unit and no named 11536sections are available. It is called once for each symbol that must be 11537recorded in the offload function and variable table. 11538@end deftypefn 11539 11540@deftypefn {Target Hook} {char *} TARGET_OFFLOAD_OPTIONS (void) 11541Used when writing out the list of options into an LTO file. It should 11542translate any relevant target-specific options (such as the ABI in use) 11543into one of the @option{-foffload} options that exist as a common interface 11544to express such options. It should return a string containing these options, 11545separated by spaces, which the caller will free. 11546 11547@end deftypefn 11548 11549@defmac TARGET_SUPPORTS_WIDE_INT 11550 11551On older ports, large integers are stored in @code{CONST_DOUBLE} rtl 11552objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero 11553to indicate that large integers are stored in 11554@code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows 11555very large integer constants to be represented. @code{CONST_DOUBLE} 11556is limited to twice the size of the host's @code{HOST_WIDE_INT} 11557representation. 11558 11559Converting a port mostly requires looking for the places where 11560@code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that 11561code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i 11562const_double"} at the port level gets you to 95% of the changes that 11563need to be made. There are a few places that require a deeper look. 11564 11565@itemize @bullet 11566@item 11567There is no equivalent to @code{hval} and @code{lval} for 11568@code{CONST_WIDE_INT}s. This would be difficult to express in the md 11569language since there are a variable number of elements. 11570 11571Most ports only check that @code{hval} is either 0 or -1 to see if the 11572value is small. As mentioned above, this will no longer be necessary 11573since small constants are always @code{CONST_INT}. Of course there 11574are still a few exceptions, the alpha's constraint used by the zap 11575instruction certainly requires careful examination by C code. 11576However, all the current code does is pass the hval and lval to C 11577code, so evolving the c code to look at the @code{CONST_WIDE_INT} is 11578not really a large change. 11579 11580@item 11581Because there is no standard template that ports use to materialize 11582constants, there is likely to be some futzing that is unique to each 11583port in this code. 11584 11585@item 11586The rtx costs may have to be adjusted to properly account for larger 11587constants that are represented as @code{CONST_WIDE_INT}. 11588@end itemize 11589 11590All and all it does not take long to convert ports that the 11591maintainer is familiar with. 11592 11593@end defmac 11594