1 @c Copyright (C) 1988-2017 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
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
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
15 The header file defines numerous macros that convey the information
16 about 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
20 source file defines a variable @code{targetm}, which is a structure
21 containing pointers to functions and data relating to the target
22 machine. @file{@var{machine}.c} should also contain their definitions,
23 if they are not defined elsewhere in GCC, and other functions called
24 through the macros defined in the @file{.h} file.
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.
59 @node Target Structure
60 @section The Global @code{targetm} Variable
62 @cindex target functions
64 @deftypevar {struct gcc_target} targetm
65 The target @file{.c} file must define the global @code{targetm} variable
66 which contains pointers to functions and data relating to the target
67 machine. The variable is declared in @file{target.h};
68 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
69 used to initialize the variable, and macros for the default initializers
70 for elements of the structure. The @file{.c} file should override those
71 macros for which the default definition is inappropriate. For example:
74 #include "target-def.h"
76 /* @r{Initialize the GCC target structure.} */
78 #undef TARGET_COMP_TYPE_ATTRIBUTES
79 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
81 struct gcc_target targetm = TARGET_INITIALIZER;
85 Where a macro should be defined in the @file{.c} file in this manner to
86 form part of the @code{targetm} structure, it is documented below as a
87 ``Target Hook'' with a prototype. Many macros will change in future
88 from being defined in the @file{.h} file to being part of the
89 @code{targetm} structure.
91 Similarly, there is a @code{targetcm} variable for hooks that are
92 specific to front ends for C-family languages, documented as ``C
93 Target Hook''. This is declared in @file{c-family/c-target.h}, the
94 initializer @code{TARGETCM_INITIALIZER} in
95 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
96 themselves, they should set @code{target_has_targetcm=yes} in
97 @file{config.gcc}; otherwise a default definition is used.
99 Similarly, there is a @code{targetm_common} variable for hooks that
100 are shared between the compiler driver and the compilers proper,
101 documented 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
107 default definition is used.
110 @section Controlling the Compilation Driver, @file{gcc}
112 @cindex controlling the compilation driver
114 @c prevent bad page break with this line
115 You can control the compilation driver.
117 @defmac DRIVER_SELF_SPECS
118 A list of specs for the driver itself. It should be a suitable
119 initializer for an array of strings, with no surrounding braces.
121 The driver applies these specs to its own command line between loading
122 default @file{specs} files (but not command-line specified ones) and
123 choosing the multilib directory or running any subcommands. It
124 applies them in the order given, so each spec can depend on the
125 options added by earlier ones. It is also possible to remove options
126 using @samp{%<@var{option}} in the usual way.
128 This macro can be useful when a port has several interdependent target
129 options. It provides a way of standardizing the command line so
130 that the other specs are easier to write.
132 Do not define this macro if it does not need to do anything.
135 @defmac OPTION_DEFAULT_SPECS
136 A 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
138 for an array of structures, each containing two strings, without the
139 outermost pair of surrounding braces.
141 The first item in the pair is the name of the default. This must match
142 the code in @file{config.gcc} for the target. The second item is a spec
143 to 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
145 everywhere it occurs.
147 The driver will apply these specs to its own command line between loading
148 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
149 the same mechanism as @code{DRIVER_SELF_SPECS}.
151 Do not define this macro if it does not need to do anything.
155 A C string constant that tells the GCC driver program options to
156 pass to CPP@. It can also specify how to translate options you
157 give to GCC into options for GCC to pass to the CPP@.
159 Do not define this macro if it does not need to do anything.
162 @defmac CPLUSPLUS_CPP_SPEC
163 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
164 than C@. If you do not define this macro, then the value of
165 @code{CPP_SPEC} (if any) will be used instead.
169 A C string constant that tells the GCC driver program options to
170 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
172 It can also specify how to translate options you give to GCC into options
173 for GCC to pass to front ends.
175 Do not define this macro if it does not need to do anything.
179 A C string constant that tells the GCC driver program options to
180 pass to @code{cc1plus}. It can also specify how to translate options you
181 give to GCC into options for GCC to pass to the @code{cc1plus}.
183 Do not define this macro if it does not need to do anything.
184 Note that everything defined in CC1_SPEC is already passed to
185 @code{cc1plus} so there is no need to duplicate the contents of
186 CC1_SPEC in CC1PLUS_SPEC@.
190 A C string constant that tells the GCC driver program options to
191 pass to the assembler. It can also specify how to translate options
192 you give to GCC into options for GCC to pass to the assembler.
193 See the file @file{sun3.h} for an example of this.
195 Do not define this macro if it does not need to do anything.
198 @defmac ASM_FINAL_SPEC
199 A C string constant that tells the GCC driver program how to
200 run any programs which cleanup after the normal assembler.
201 Normally, this is not needed. See the file @file{mips.h} for
204 Do not define this macro if it does not need to do anything.
207 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
208 Define this macro, with no value, if the driver should give the assembler
209 an argument consisting of a single dash, @option{-}, to instruct it to
210 read from its standard input (which will be a pipe connected to the
211 output of the compiler proper). This argument is given after any
212 @option{-o} option specifying the name of the output file.
214 If you do not define this macro, the assembler is assumed to read its
215 standard input if given no non-option arguments. If your assembler
216 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
217 see @file{mips.h} for instance.
221 A C string constant that tells the GCC driver program options to
222 pass to the linker. It can also specify how to translate options you
223 give to GCC into options for GCC to pass to the linker.
225 Do not define this macro if it does not need to do anything.
229 Another C string constant used much like @code{LINK_SPEC}. The difference
230 between the two is that @code{LIB_SPEC} is used at the end of the
231 command given to the linker.
233 If this macro is not defined, a default is provided that
234 loads the standard C library from the usual place. See @file{gcc.c}.
238 Another C string constant that tells the GCC driver program
239 how and when to place a reference to @file{libgcc.a} into the
240 linker command line. This constant is placed both before and after
241 the value of @code{LIB_SPEC}.
243 If this macro is not defined, the GCC driver provides a default that
244 passes the string @option{-lgcc} to the linker.
247 @defmac REAL_LIBGCC_SPEC
248 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
249 @code{LIBGCC_SPEC} is not directly used by the driver program but is
250 instead modified to refer to different versions of @file{libgcc.a}
251 depending on the values of the command line flags @option{-static},
252 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
253 targets where these modifications are inappropriate, define
254 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
255 driver how to place a reference to @file{libgcc} on the link command
256 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
259 @defmac USE_LD_AS_NEEDED
260 A macro that controls the modifications to @code{LIBGCC_SPEC}
261 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
262 generated that uses @option{--as-needed} or equivalent options and the
263 shared @file{libgcc} in place of the
264 static exception handler library, when linking without any of
265 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
269 If defined, this C string constant is added to @code{LINK_SPEC}.
270 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
271 the modifications to @code{LIBGCC_SPEC} mentioned in
272 @code{REAL_LIBGCC_SPEC}.
275 @defmac STARTFILE_SPEC
276 Another C string constant used much like @code{LINK_SPEC}. The
277 difference between the two is that @code{STARTFILE_SPEC} is used at
278 the very beginning of the command given to the linker.
280 If this macro is not defined, a default is provided that loads the
281 standard C startup file from the usual place. See @file{gcc.c}.
285 Another C string constant used much like @code{LINK_SPEC}. The
286 difference between the two is that @code{ENDFILE_SPEC} is used at
287 the very end of the command given to the linker.
289 Do not define this macro if it does not need to do anything.
292 @defmac THREAD_MODEL_SPEC
293 GCC @code{-v} will print the thread model GCC was configured to use.
294 However, this doesn't work on platforms that are multilibbed on thread
295 models, such as AIX 4.3. On such platforms, define
296 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
297 blanks that names one of the recognized thread models. @code{%*}, the
298 default value of this macro, will expand to the value of
299 @code{thread_file} set in @file{config.gcc}.
302 @defmac SYSROOT_SUFFIX_SPEC
303 Define this macro to add a suffix to the target sysroot when GCC is
304 configured with a sysroot. This will cause GCC to search for usr/lib,
305 et al, within sysroot+suffix.
308 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
309 Define this macro to add a headers_suffix to the target sysroot when
310 GCC is configured with a sysroot. This will cause GCC to pass the
311 updated sysroot+headers_suffix to CPP, causing it to search for
312 usr/include, et al, within sysroot+headers_suffix.
316 Define this macro to provide additional specifications to put in the
317 @file{specs} file that can be used in various specifications like
320 The definition should be an initializer for an array of structures,
321 containing a string constant, that defines the specification name, and a
322 string constant that provides the specification.
324 Do not define this macro if it does not need to do anything.
326 @code{EXTRA_SPECS} is useful when an architecture contains several
327 related targets, which have various @code{@dots{}_SPECS} which are similar
328 to each other, and the maintainer would like one central place to keep
331 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
332 define either @code{_CALL_SYSV} when the System V calling sequence is
333 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
336 The @file{config/rs6000/rs6000.h} target file defines:
339 #define EXTRA_SPECS \
340 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
342 #define CPP_SYS_DEFAULT ""
345 The @file{config/rs6000/sysv.h} target file defines:
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@}"
354 #undef CPP_SYSV_DEFAULT
355 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
358 while the @file{config/rs6000/eabiaix.h} target file defines
359 @code{CPP_SYSV_DEFAULT} as:
362 #undef CPP_SYSV_DEFAULT
363 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
367 @defmac LINK_LIBGCC_SPECIAL_1
368 Define 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
370 the argument @option{-lgcc} to tell the linker to do the search.
373 @defmac LINK_GCC_C_SEQUENCE_SPEC
374 The sequence in which libgcc and libc are specified to the linker.
375 By default this is @code{%G %L %G}.
378 @defmac POST_LINK_SPEC
379 Define this macro to add additional steps to be executed after linker.
380 The default value of this macro is empty string.
383 @defmac LINK_COMMAND_SPEC
384 A C string constant giving the complete command line need to execute the
385 linker. When you do this, you will need to update your port each time a
386 change is made to the link command line within @file{gcc.c}. Therefore,
387 define this macro only if you need to completely redefine the command
388 line for invoking the linker and there is no other way to accomplish
389 the effect you need. Overriding this macro may be avoidable by overriding
390 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
393 @hook TARGET_ALWAYS_STRIP_DOTDOT
395 @defmac MULTILIB_DEFAULTS
396 Define this macro as a C expression for the initializer of an array of
397 string to tell the driver program which options are defaults for this
398 target and thus do not need to be handled specially when using
399 @code{MULTILIB_OPTIONS}.
401 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
402 the target makefile fragment or if none of the options listed in
403 @code{MULTILIB_OPTIONS} are set by default.
404 @xref{Target Fragment}.
407 @defmac RELATIVE_PREFIX_NOT_LINKDIR
408 Define this macro to tell @command{gcc} that it should only translate
409 a @option{-B} prefix into a @option{-L} linker option if the prefix
410 indicates an absolute file name.
413 @defmac MD_EXEC_PREFIX
414 If defined, this macro is an additional prefix to try after
415 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
416 when the compiler is built as a cross
417 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
418 to the list of directories used to find the assembler in @file{configure.ac}.
421 @defmac STANDARD_STARTFILE_PREFIX
422 Define this macro as a C string constant if you wish to override the
423 standard choice of @code{libdir} as the default prefix to
424 try when searching for startup files such as @file{crt0.o}.
425 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
426 is built as a cross compiler.
429 @defmac STANDARD_STARTFILE_PREFIX_1
430 Define this macro as a C string constant if you wish to override the
431 standard choice of @code{/lib} as a prefix to try after the default prefix
432 when searching for startup files such as @file{crt0.o}.
433 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
434 is built as a cross compiler.
437 @defmac STANDARD_STARTFILE_PREFIX_2
438 Define this macro as a C string constant if you wish to override the
439 standard choice of @code{/lib} as yet another prefix to try after the
440 default prefix when searching for startup files such as @file{crt0.o}.
441 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
442 is built as a cross compiler.
445 @defmac MD_STARTFILE_PREFIX
446 If defined, this macro supplies an additional prefix to try after the
447 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
448 compiler is built as a cross compiler.
451 @defmac MD_STARTFILE_PREFIX_1
452 If defined, this macro supplies yet another prefix to try after the
453 standard prefixes. It is not searched when the compiler is built as a
457 @defmac INIT_ENVIRONMENT
458 Define this macro as a C string constant if you wish to set environment
459 variables for programs called by the driver, such as the assembler and
460 loader. The driver passes the value of this macro to @code{putenv} to
461 initialize the necessary environment variables.
464 @defmac LOCAL_INCLUDE_DIR
465 Define this macro as a C string constant if you wish to override the
466 standard choice of @file{/usr/local/include} as the default prefix to
467 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
468 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
469 @file{config.gcc}, normally @file{/usr/include}) in the search order.
471 Cross compilers do not search either @file{/usr/local/include} or its
475 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
476 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
477 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
478 If you do not define this macro, no component is used.
481 @defmac INCLUDE_DEFAULTS
482 Define this macro if you wish to override the entire default search path
483 for include files. For a native compiler, the default search path
484 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
485 @code{GPLUSPLUS_INCLUDE_DIR}, and
486 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
487 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
488 and specify private search areas for GCC@. The directory
489 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
491 The definition should be an initializer for an array of structures.
492 Each array element should have four elements: the directory name (a
493 string constant), the component name (also a string constant), a flag
494 for C++-only directories,
495 and a flag showing that the includes in the directory don't need to be
496 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
497 the array with a null element.
499 The component name denotes what GNU package the include file is part of,
500 if any, in all uppercase letters. For example, it might be @samp{GCC}
501 or @samp{BINUTILS}. If the package is part of a vendor-supplied
502 operating system, code the component name as @samp{0}.
504 For example, here is the definition used for VAX/VMS:
507 #define INCLUDE_DEFAULTS \
509 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
510 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
511 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
518 Here is the order of prefixes tried for exec files:
522 Any prefixes specified by the user with @option{-B}.
525 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
526 is not set and the compiler has not been installed in the configure-time
527 @var{prefix}, the location in which the compiler has actually been installed.
530 The directories specified by the environment variable @code{COMPILER_PATH}.
533 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
534 in the configured-time @var{prefix}.
537 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
540 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
543 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
547 Here is the order of prefixes tried for startfiles:
551 Any prefixes specified by the user with @option{-B}.
554 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
555 value based on the installed toolchain location.
558 The directories specified by the environment variable @code{LIBRARY_PATH}
559 (or port-specific name; native only, cross compilers do not use this).
562 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
563 in the configured @var{prefix} or this is a native compiler.
566 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
569 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
573 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
574 native compiler, or we have a target system root.
577 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
578 native compiler, or we have a target system root.
581 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
582 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
583 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
586 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
587 compiler, or we have a target system root. The default for this macro is
591 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
592 compiler, or we have a target system root. The default for this macro is
596 @node Run-time Target
597 @section Run-time Target Specification
598 @cindex run-time target specification
599 @cindex predefined macros
600 @cindex target specifications
602 @c prevent bad page break with this line
603 Here are run-time target specifications.
605 @defmac TARGET_CPU_CPP_BUILTINS ()
606 This function-like macro expands to a block of code that defines
607 built-in preprocessor macros and assertions for the target CPU, using
608 the functions @code{builtin_define}, @code{builtin_define_std} and
609 @code{builtin_assert}. When the front end
610 calls this macro it provides a trailing semicolon, and since it has
611 finished command line option processing your code can use those
614 @code{builtin_assert} takes a string in the form you pass to the
615 command-line option @option{-A}, such as @code{cpu=mips}, and creates
616 the assertion. @code{builtin_define} takes a string in the form
617 accepted by option @option{-D} and unconditionally defines the macro.
619 @code{builtin_define_std} takes a string representing the name of an
620 object-like macro. If it doesn't lie in the user's namespace,
621 @code{builtin_define_std} defines it unconditionally. Otherwise, it
622 defines a version with two leading underscores, and another version
623 with two leading and trailing underscores, and defines the original
624 only if an ISO standard was not requested on the command line. For
625 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
626 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
627 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
628 defines only @code{_ABI64}.
630 You can also test for the C dialect being compiled. The variable
631 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
632 or @code{clk_objective_c}. Note that if we are preprocessing
633 assembler, this variable will be @code{clk_c} but the function-like
634 macro @code{preprocessing_asm_p()} will return true, so you might want
635 to check for that first. If you need to check for strict ANSI, the
636 variable @code{flag_iso} can be used. The function-like macro
637 @code{preprocessing_trad_p()} can be used to check for traditional
641 @defmac TARGET_OS_CPP_BUILTINS ()
642 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
643 and is used for the target operating system instead.
646 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
647 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
648 and is used for the target object format. @file{elfos.h} uses this
649 macro to define @code{__ELF__}, so you probably do not need to define
653 @deftypevar {extern int} target_flags
654 This variable is declared in @file{options.h}, which is included before
655 any target-specific headers.
658 @hook TARGET_DEFAULT_TARGET_FLAGS
659 This variable specifies the initial value of @code{target_flags}.
660 Its default setting is 0.
663 @cindex optional hardware or system features
664 @cindex features, optional, in system conventions
666 @hook TARGET_HANDLE_OPTION
667 This hook is called whenever the user specifies one of the
668 target-specific options described by the @file{.opt} definition files
669 (@pxref{Options}). It has the opportunity to do some option-specific
670 processing and should return true if the option is valid. The default
671 definition does nothing but return true.
673 @var{decoded} specifies the option and its arguments. @var{opts} and
674 @var{opts_set} are the @code{gcc_options} structures to be used for
675 storing option state, and @var{loc} is the location at which the
676 option was passed (@code{UNKNOWN_LOCATION} except for options passed
680 @hook TARGET_HANDLE_C_OPTION
681 This target hook is called whenever the user specifies one of the
682 target-specific C language family options described by the @file{.opt}
683 definition files(@pxref{Options}). It has the opportunity to do some
684 option-specific processing and should return true if the option is
685 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
686 default definition does nothing but return false.
688 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
689 options. However, if processing an option requires routines that are
690 only available in the C (and related language) front ends, then you
691 should use @code{TARGET_HANDLE_C_OPTION} instead.
694 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
696 @hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE
698 @hook TARGET_OBJC_DECLARE_CLASS_DEFINITION
700 @hook TARGET_STRING_OBJECT_REF_TYPE_P
702 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
704 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
706 @defmac C_COMMON_OVERRIDE_OPTIONS
707 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
708 but is only used in the C
709 language frontends (C, Objective-C, C++, Objective-C++) and so can be
710 used to alter option flag variables which only exist in those
714 @hook TARGET_OPTION_OPTIMIZATION_TABLE
715 Some machines may desire to change what optimizations are performed for
716 various optimization levels. This variable, if defined, describes
717 options to enable at particular sets of optimization levels. These
718 options are processed once
719 just after the optimization level is determined and before the remainder
720 of the command options have been parsed, so may be overridden by other
721 options passed explicitly.
723 This processing is run once at program startup and when the optimization
724 options are changed via @code{#pragma GCC optimize} or by using the
725 @code{optimize} attribute.
728 @hook TARGET_OPTION_INIT_STRUCT
730 @hook TARGET_OPTION_DEFAULT_PARAMS
732 @defmac SWITCHABLE_TARGET
733 Some targets need to switch between substantially different subtargets
734 during compilation. For example, the MIPS target has one subtarget for
735 the traditional MIPS architecture and another for MIPS16. Source code
736 can switch between these two subarchitectures using the @code{mips16}
737 and @code{nomips16} attributes.
739 Such subtargets can differ in things like the set of available
740 registers, the set of available instructions, the costs of various
741 operations, and so on. GCC caches a lot of this type of information
742 in global variables, and recomputing them for each subtarget takes a
743 significant amount of time. The compiler therefore provides a facility
744 for maintaining several versions of the global variables and quickly
745 switching between them; see @file{target-globals.h} for details.
747 Define this macro to 1 if your target needs this facility. The default
751 @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
753 @node Per-Function Data
754 @section Defining data structures for per-function information.
755 @cindex per-function data
756 @cindex data structures
758 If the target needs to store information on a per-function basis, GCC
759 provides a macro and a couple of variables to allow this. Note, just
760 using statics to store the information is a bad idea, since GCC supports
761 nested functions, so you can be halfway through encoding one function
762 when another one comes along.
764 GCC defines a data structure called @code{struct function} which
765 contains all of the data specific to an individual function. This
766 structure contains a field called @code{machine} whose type is
767 @code{struct machine_function *}, which can be used by targets to point
768 to their own specific data.
770 If a target needs per-function specific data it should define the type
771 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
772 This macro should be used to initialize the function pointer
773 @code{init_machine_status}. This pointer is explained below.
775 One typical use of per-function, target specific data is to create an
776 RTX to hold the register containing the function's return address. This
777 RTX can then be used to implement the @code{__builtin_return_address}
778 function, for level 0.
780 Note---earlier implementations of GCC used a single data area to hold
781 all of the per-function information. Thus when processing of a nested
782 function began the old per-function data had to be pushed onto a
783 stack, and when the processing was finished, it had to be popped off the
784 stack. GCC used to provide function pointers called
785 @code{save_machine_status} and @code{restore_machine_status} to handle
786 the saving and restoring of the target specific information. Since the
787 single data area approach is no longer used, these pointers are no
790 @defmac INIT_EXPANDERS
791 Macro called to initialize any target specific information. This macro
792 is called once per function, before generation of any RTL has begun.
793 The intention of this macro is to allow the initialization of the
794 function pointer @code{init_machine_status}.
797 @deftypevar {void (*)(struct function *)} init_machine_status
798 If this function pointer is non-@code{NULL} it will be called once per
799 function, before function compilation starts, in order to allow the
800 target to perform any target specific initialization of the
801 @code{struct function} structure. It is intended that this would be
802 used to initialize the @code{machine} of that structure.
804 @code{struct machine_function} structures are expected to be freed by GC@.
805 Generally, any memory that they reference must be allocated by using
806 GC allocation, including the structure itself.
810 @section Storage Layout
811 @cindex storage layout
813 Note that the definitions of the macros in this table which are sizes or
814 alignments measured in bits do not need to be constant. They can be C
815 expressions that refer to static variables, such as the @code{target_flags}.
816 @xref{Run-time Target}.
818 @defmac BITS_BIG_ENDIAN
819 Define this macro to have the value 1 if the most significant bit in a
820 byte has the lowest number; otherwise define it to have the value zero.
821 This means that bit-field instructions count from the most significant
822 bit. If the machine has no bit-field instructions, then this must still
823 be defined, but it doesn't matter which value it is defined to. This
824 macro need not be a constant.
826 This macro does not affect the way structure fields are packed into
827 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
830 @defmac BYTES_BIG_ENDIAN
831 Define this macro to have the value 1 if the most significant byte in a
832 word has the lowest number. This macro need not be a constant.
835 @defmac WORDS_BIG_ENDIAN
836 Define this macro to have the value 1 if, in a multiword object, the
837 most significant word has the lowest number. This applies to both
838 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
839 order of words in memory is not the same as the order in registers. This
840 macro need not be a constant.
843 @defmac REG_WORDS_BIG_ENDIAN
844 On some machines, the order of words in a multiword object differs between
845 registers in memory. In such a situation, define this macro to describe
846 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
847 the order of words in memory.
850 @defmac FLOAT_WORDS_BIG_ENDIAN
851 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
852 @code{TFmode} floating point numbers are stored in memory with the word
853 containing the sign bit at the lowest address; otherwise define it to
854 have the value 0. This macro need not be a constant.
856 You need not define this macro if the ordering is the same as for
860 @defmac BITS_PER_WORD
861 Number of bits in a word. If you do not define this macro, the default
862 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
865 @defmac MAX_BITS_PER_WORD
866 Maximum number of bits in a word. If this is undefined, the default is
867 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
868 largest value that @code{BITS_PER_WORD} can have at run-time.
871 @defmac UNITS_PER_WORD
872 Number of storage units in a word; normally the size of a general-purpose
873 register, a power of two from 1 or 8.
876 @defmac MIN_UNITS_PER_WORD
877 Minimum number of units in a word. If this is undefined, the default is
878 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
879 smallest value that @code{UNITS_PER_WORD} can have at run-time.
883 Width of a pointer, in bits. You must specify a value no wider than the
884 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
885 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
886 a value the default is @code{BITS_PER_WORD}.
889 @defmac POINTERS_EXTEND_UNSIGNED
890 A C expression that determines how pointers should be extended from
891 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
892 greater than zero if pointers should be zero-extended, zero if they
893 should be sign-extended, and negative if some other sort of conversion
894 is needed. In the last case, the extension is done by the target's
895 @code{ptr_extend} instruction.
897 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
898 and @code{word_mode} are all the same width.
901 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
902 A macro to update @var{m} and @var{unsignedp} when an object whose type
903 is @var{type} and which has the specified mode and signedness is to be
904 stored in a register. This macro is only called when @var{type} is a
907 On most RISC machines, which only have operations that operate on a full
908 register, define this macro to set @var{m} to @code{word_mode} if
909 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
910 cases, only integer modes should be widened because wider-precision
911 floating-point operations are usually more expensive than their narrower
914 For most machines, the macro definition does not change @var{unsignedp}.
915 However, some machines, have instructions that preferentially handle
916 either signed or unsigned quantities of certain modes. For example, on
917 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
918 sign-extend the result to 64 bits. On such machines, set
919 @var{unsignedp} according to which kind of extension is more efficient.
921 Do not define this macro if it would never modify @var{m}.
924 @hook TARGET_C_EXCESS_PRECISION
926 @hook TARGET_PROMOTE_FUNCTION_MODE
928 @defmac PARM_BOUNDARY
929 Normal alignment required for function parameters on the stack, in
930 bits. All stack parameters receive at least this much alignment
931 regardless of data type. On most machines, this is the same as the
935 @defmac STACK_BOUNDARY
936 Define this macro to the minimum alignment enforced by hardware for the
937 stack pointer on this machine. The definition is a C expression for the
938 desired alignment (measured in bits). This value is used as a default
939 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
940 this should be the same as @code{PARM_BOUNDARY}.
943 @defmac PREFERRED_STACK_BOUNDARY
944 Define this macro if you wish to preserve a certain alignment for the
945 stack pointer, greater than what the hardware enforces. The definition
946 is a C expression for the desired alignment (measured in bits). This
947 macro must evaluate to a value equal to or larger than
948 @code{STACK_BOUNDARY}.
951 @defmac INCOMING_STACK_BOUNDARY
952 Define this macro if the incoming stack boundary may be different
953 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
954 to a value equal to or larger than @code{STACK_BOUNDARY}.
957 @defmac FUNCTION_BOUNDARY
958 Alignment required for a function entry point, in bits.
961 @defmac BIGGEST_ALIGNMENT
962 Biggest alignment that any data type can require on this machine, in
963 bits. Note that this is not the biggest alignment that is supported,
964 just the biggest alignment that, when violated, may cause a fault.
967 @hook TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
969 @defmac MALLOC_ABI_ALIGNMENT
970 Alignment, in bits, a C conformant malloc implementation has to
971 provide. If not defined, the default value is @code{BITS_PER_WORD}.
974 @defmac ATTRIBUTE_ALIGNED_VALUE
975 Alignment used by the @code{__attribute__ ((aligned))} construct. If
976 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
979 @defmac MINIMUM_ATOMIC_ALIGNMENT
980 If defined, the smallest alignment, in bits, that can be given to an
981 object that can be referenced in one operation, without disturbing any
982 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
983 on machines that don't have byte or half-word store operations.
986 @defmac BIGGEST_FIELD_ALIGNMENT
987 Biggest alignment that any structure or union field can require on this
988 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
989 structure and union fields only, unless the field alignment has been set
990 by the @code{__attribute__ ((aligned (@var{n})))} construct.
993 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
994 An expression for the alignment of a structure field @var{field} if the
995 alignment computed in the usual way (including applying of
996 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
997 alignment) is @var{computed}. It overrides alignment only if the
998 field alignment has not been set by the
999 @code{__attribute__ ((aligned (@var{n})))} construct.
1002 @defmac MAX_STACK_ALIGNMENT
1003 Biggest stack alignment guaranteed by the backend. Use this macro
1004 to specify the maximum alignment of a variable on stack.
1006 If not defined, the default value is @code{STACK_BOUNDARY}.
1008 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1009 @c But the fix for PR 32893 indicates that we can only guarantee
1010 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1011 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1014 @defmac MAX_OFILE_ALIGNMENT
1015 Biggest alignment supported by the object file format of this machine.
1016 Use this macro to limit the alignment which can be specified using the
1017 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1018 the default value is @code{BIGGEST_ALIGNMENT}.
1020 On systems that use ELF, the default (in @file{config/elfos.h}) is
1021 the largest supported 32-bit ELF section alignment representable on
1022 a 32-bit host e.g. @samp{(((uint64_t) 1 << 28) * 8)}.
1023 On 32-bit ELF the largest supported section alignment in bits is
1024 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1027 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1028 If defined, a C expression to compute the alignment for a variable in
1029 the static store. @var{type} is the data type, and @var{basic-align} is
1030 the alignment that the object would ordinarily have. The value of this
1031 macro is used instead of that alignment to align the object.
1033 If this macro is not defined, then @var{basic-align} is used.
1036 One use of this macro is to increase alignment of medium-size data to
1037 make it all fit in fewer cache lines. Another is to cause character
1038 arrays to be word-aligned so that @code{strcpy} calls that copy
1039 constants to character arrays can be done inline.
1042 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1043 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1044 some alignment increase, instead of optimization only purposes. E.g.@
1045 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1046 must be aligned to 16 byte boundaries.
1048 If this macro is not defined, then @var{basic-align} is used.
1051 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1052 If defined, a C expression to compute the alignment given to a constant
1053 that is being placed in memory. @var{constant} is the constant and
1054 @var{basic-align} is the alignment that the object would ordinarily
1055 have. The value of this macro is used instead of that alignment to
1058 The default definition just returns @var{basic-align}.
1060 The typical use of this macro is to increase alignment for string
1061 constants to be word aligned so that @code{strcpy} calls that copy
1062 constants can be done inline.
1065 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1066 If defined, a C expression to compute the alignment for a variable in
1067 the local store. @var{type} is the data type, and @var{basic-align} is
1068 the alignment that the object would ordinarily have. The value of this
1069 macro is used instead of that alignment to align the object.
1071 If this macro is not defined, then @var{basic-align} is used.
1073 One use of this macro is to increase alignment of medium-size data to
1074 make it all fit in fewer cache lines.
1076 If the value of this macro has a type, it should be an unsigned type.
1079 @hook TARGET_VECTOR_ALIGNMENT
1081 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1082 If defined, a C expression to compute the alignment for stack slot.
1083 @var{type} is the data type, @var{mode} is the widest mode available,
1084 and @var{basic-align} is the alignment that the slot would ordinarily
1085 have. The value of this macro is used instead of that alignment to
1088 If this macro is not defined, then @var{basic-align} is used when
1089 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1092 This macro is to set alignment of stack slot to the maximum alignment
1093 of all possible modes which the slot may have.
1095 If the value of this macro has a type, it should be an unsigned type.
1098 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1099 If defined, a C expression to compute the alignment for a local
1100 variable @var{decl}.
1102 If this macro is not defined, then
1103 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1106 One use of this macro is to increase alignment of medium-size data to
1107 make it all fit in fewer cache lines.
1109 If the value of this macro has a type, it should be an unsigned type.
1112 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1113 If defined, a C expression to compute the minimum required alignment
1114 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1115 @var{mode}, assuming normal alignment @var{align}.
1117 If this macro is not defined, then @var{align} will be used.
1120 @defmac EMPTY_FIELD_BOUNDARY
1121 Alignment in bits to be given to a structure bit-field that follows an
1122 empty field such as @code{int : 0;}.
1124 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1127 @defmac STRUCTURE_SIZE_BOUNDARY
1128 Number of bits which any structure or union's size must be a multiple of.
1129 Each structure or union's size is rounded up to a multiple of this.
1131 If you do not define this macro, the default is the same as
1132 @code{BITS_PER_UNIT}.
1135 @defmac STRICT_ALIGNMENT
1136 Define this macro to be the value 1 if instructions will fail to work
1137 if given data not on the nominal alignment. If instructions will merely
1138 go slower in that case, define this macro as 0.
1141 @defmac PCC_BITFIELD_TYPE_MATTERS
1142 Define this if you wish to imitate the way many other C compilers handle
1143 alignment of bit-fields and the structures that contain them.
1145 The behavior is that the type written for a named bit-field (@code{int},
1146 @code{short}, or other integer type) imposes an alignment for the entire
1147 structure, as if the structure really did contain an ordinary field of
1148 that type. In addition, the bit-field is placed within the structure so
1149 that it would fit within such a field, not crossing a boundary for it.
1151 Thus, on most machines, a named bit-field whose type is written as
1152 @code{int} would not cross a four-byte boundary, and would force
1153 four-byte alignment for the whole structure. (The alignment used may
1154 not be four bytes; it is controlled by the other alignment parameters.)
1156 An unnamed bit-field will not affect the alignment of the containing
1159 If the macro is defined, its definition should be a C expression;
1160 a nonzero value for the expression enables this behavior.
1162 Note that if this macro is not defined, or its value is zero, some
1163 bit-fields may cross more than one alignment boundary. The compiler can
1164 support such references if there are @samp{insv}, @samp{extv}, and
1165 @samp{extzv} insns that can directly reference memory.
1167 The other known way of making bit-fields work is to define
1168 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1169 Then every structure can be accessed with fullwords.
1171 Unless the machine has bit-field instructions or you define
1172 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1173 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1175 If your aim is to make GCC use the same conventions for laying out
1176 bit-fields as are used by another compiler, here is how to investigate
1177 what the other compiler does. Compile and run this program:
1196 printf ("Size of foo1 is %d\n",
1197 sizeof (struct foo1));
1198 printf ("Size of foo2 is %d\n",
1199 sizeof (struct foo2));
1204 If this prints 2 and 5, then the compiler's behavior is what you would
1205 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1208 @defmac BITFIELD_NBYTES_LIMITED
1209 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1210 to aligning a bit-field within the structure.
1213 @hook TARGET_ALIGN_ANON_BITFIELD
1215 @hook TARGET_NARROW_VOLATILE_BITFIELD
1217 @hook TARGET_MEMBER_TYPE_FORCES_BLK
1219 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1220 Define this macro as an expression for the alignment of a type (given
1221 by @var{type} as a tree node) if the alignment computed in the usual
1222 way is @var{computed} and the alignment explicitly specified was
1225 The default is to use @var{specified} if it is larger; otherwise, use
1226 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1229 @defmac MAX_FIXED_MODE_SIZE
1230 An integer expression for the size in bits of the largest integer
1231 machine mode that should actually be used. All integer machine modes of
1232 this size or smaller can be used for structures and unions with the
1233 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1234 (DImode)} is assumed.
1237 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1238 If defined, an expression of type @code{machine_mode} that
1239 specifies the mode of the save area operand of a
1240 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1241 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1242 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1243 having its mode specified.
1245 You need not define this macro if it always returns @code{Pmode}. You
1246 would most commonly define this macro if the
1247 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1251 @defmac STACK_SIZE_MODE
1252 If defined, an expression of type @code{machine_mode} that
1253 specifies the mode of the size increment operand of an
1254 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1256 You need not define this macro if it always returns @code{word_mode}.
1257 You would most commonly define this macro if the @code{allocate_stack}
1258 pattern needs to support both a 32- and a 64-bit mode.
1261 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1263 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1265 @hook TARGET_UNWIND_WORD_MODE
1267 @hook TARGET_MS_BITFIELD_LAYOUT_P
1269 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1271 @hook TARGET_FIXED_POINT_SUPPORTED_P
1273 @hook TARGET_EXPAND_TO_RTL_HOOK
1275 @hook TARGET_INSTANTIATE_DECLS
1277 @hook TARGET_MANGLE_TYPE
1280 @section Layout of Source Language Data Types
1282 These macros define the sizes and other characteristics of the standard
1283 basic data types used in programs being compiled. Unlike the macros in
1284 the previous section, these apply to specific features of C and related
1285 languages, rather than to fundamental aspects of storage layout.
1287 @defmac INT_TYPE_SIZE
1288 A C expression for the size in bits of the type @code{int} on the
1289 target machine. If you don't define this, the default is one word.
1292 @defmac SHORT_TYPE_SIZE
1293 A C expression for the size in bits of the type @code{short} on the
1294 target machine. If you don't define this, the default is half a word.
1295 (If this would be less than one storage unit, it is rounded up to one
1299 @defmac LONG_TYPE_SIZE
1300 A C expression for the size in bits of the type @code{long} on the
1301 target machine. If you don't define this, the default is one word.
1304 @defmac ADA_LONG_TYPE_SIZE
1305 On some machines, the size used for the Ada equivalent of the type
1306 @code{long} by a native Ada compiler differs from that used by C@. In
1307 that situation, define this macro to be a C expression to be used for
1308 the size of that type. If you don't define this, the default is the
1309 value of @code{LONG_TYPE_SIZE}.
1312 @defmac LONG_LONG_TYPE_SIZE
1313 A C expression for the size in bits of the type @code{long long} on the
1314 target machine. If you don't define this, the default is two
1315 words. If you want to support GNU Ada on your machine, the value of this
1316 macro must be at least 64.
1319 @defmac CHAR_TYPE_SIZE
1320 A C expression for the size in bits of the type @code{char} on the
1321 target machine. If you don't define this, the default is
1322 @code{BITS_PER_UNIT}.
1325 @defmac BOOL_TYPE_SIZE
1326 A C expression for the size in bits of the C++ type @code{bool} and
1327 C99 type @code{_Bool} on the target machine. If you don't define
1328 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1331 @defmac FLOAT_TYPE_SIZE
1332 A C expression for the size in bits of the type @code{float} on the
1333 target machine. If you don't define this, the default is one word.
1336 @defmac DOUBLE_TYPE_SIZE
1337 A C expression for the size in bits of the type @code{double} on the
1338 target machine. If you don't define this, the default is two
1342 @defmac LONG_DOUBLE_TYPE_SIZE
1343 A C expression for the size in bits of the type @code{long double} on
1344 the target machine. If you don't define this, the default is two
1348 @defmac SHORT_FRACT_TYPE_SIZE
1349 A C expression for the size in bits of the type @code{short _Fract} on
1350 the target machine. If you don't define this, the default is
1351 @code{BITS_PER_UNIT}.
1354 @defmac FRACT_TYPE_SIZE
1355 A C expression for the size in bits of the type @code{_Fract} on
1356 the target machine. If you don't define this, the default is
1357 @code{BITS_PER_UNIT * 2}.
1360 @defmac LONG_FRACT_TYPE_SIZE
1361 A C expression for the size in bits of the type @code{long _Fract} on
1362 the target machine. If you don't define this, the default is
1363 @code{BITS_PER_UNIT * 4}.
1366 @defmac LONG_LONG_FRACT_TYPE_SIZE
1367 A C expression for the size in bits of the type @code{long long _Fract} on
1368 the target machine. If you don't define this, the default is
1369 @code{BITS_PER_UNIT * 8}.
1372 @defmac SHORT_ACCUM_TYPE_SIZE
1373 A C expression for the size in bits of the type @code{short _Accum} on
1374 the target machine. If you don't define this, the default is
1375 @code{BITS_PER_UNIT * 2}.
1378 @defmac ACCUM_TYPE_SIZE
1379 A C expression for the size in bits of the type @code{_Accum} on
1380 the target machine. If you don't define this, the default is
1381 @code{BITS_PER_UNIT * 4}.
1384 @defmac LONG_ACCUM_TYPE_SIZE
1385 A C expression for the size in bits of the type @code{long _Accum} on
1386 the target machine. If you don't define this, the default is
1387 @code{BITS_PER_UNIT * 8}.
1390 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1391 A C expression for the size in bits of the type @code{long long _Accum} on
1392 the target machine. If you don't define this, the default is
1393 @code{BITS_PER_UNIT * 16}.
1396 @defmac LIBGCC2_GNU_PREFIX
1397 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1398 hook and should be defined if that hook is overriden to be true. It
1399 causes function names in libgcc to be changed to use a @code{__gnu_}
1400 prefix for their name rather than the default @code{__}. A port which
1401 uses this macro should also arrange to use @file{t-gnu-prefix} in
1402 the libgcc @file{config.host}.
1405 @defmac WIDEST_HARDWARE_FP_SIZE
1406 A C expression for the size in bits of the widest floating-point format
1407 supported by the hardware. If you define this macro, you must specify a
1408 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1409 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1413 @defmac DEFAULT_SIGNED_CHAR
1414 An expression whose value is 1 or 0, according to whether the type
1415 @code{char} should be signed or unsigned by default. The user can
1416 always override this default with the options @option{-fsigned-char}
1417 and @option{-funsigned-char}.
1420 @hook TARGET_DEFAULT_SHORT_ENUMS
1423 A C expression for a string describing the name of the data type to use
1424 for size values. The typedef name @code{size_t} is defined using the
1425 contents of the string.
1427 The string can contain more than one keyword. If so, separate them with
1428 spaces, and write first any length keyword, then @code{unsigned} if
1429 appropriate, and finally @code{int}. The string must exactly match one
1430 of the data type names defined in the function
1431 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1432 You may not omit @code{int} or change the order---that would cause the
1433 compiler to crash on startup.
1435 If you don't define this macro, the default is @code{"long unsigned
1440 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1441 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1442 dealing with size. This macro is a C expression for a string describing
1443 the name of the data type from which the precision of @code{sizetype}
1446 The string has the same restrictions as @code{SIZE_TYPE} string.
1448 If you don't define this macro, the default is @code{SIZE_TYPE}.
1451 @defmac PTRDIFF_TYPE
1452 A C expression for a string describing the name of the data type to use
1453 for the result of subtracting two pointers. The typedef name
1454 @code{ptrdiff_t} is defined using the contents of the string. See
1455 @code{SIZE_TYPE} above for more information.
1457 If you don't define this macro, the default is @code{"long int"}.
1461 A C expression for a string describing the name of the data type to use
1462 for wide characters. The typedef name @code{wchar_t} is defined using
1463 the contents of the string. See @code{SIZE_TYPE} above for more
1466 If you don't define this macro, the default is @code{"int"}.
1469 @defmac WCHAR_TYPE_SIZE
1470 A C expression for the size in bits of the data type for wide
1471 characters. This is used in @code{cpp}, which cannot make use of
1476 A C expression for a string describing the name of the data type to
1477 use for wide characters passed to @code{printf} and returned from
1478 @code{getwc}. The typedef name @code{wint_t} is defined using the
1479 contents of the string. See @code{SIZE_TYPE} above for more
1482 If you don't define this macro, the default is @code{"unsigned int"}.
1486 A C expression for a string describing the name of the data type that
1487 can represent any value of any standard or extended signed integer type.
1488 The typedef name @code{intmax_t} is defined using the contents of the
1489 string. See @code{SIZE_TYPE} above for more information.
1491 If you don't define this macro, the default is the first of
1492 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1493 much precision as @code{long long int}.
1496 @defmac UINTMAX_TYPE
1497 A C expression for a string describing the name of the data type that
1498 can represent any value of any standard or extended unsigned integer
1499 type. The typedef name @code{uintmax_t} is defined using the contents
1500 of the string. See @code{SIZE_TYPE} above for more information.
1502 If you don't define this macro, the default is the first of
1503 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1504 unsigned int"} that has as much precision as @code{long long unsigned
1508 @defmac SIG_ATOMIC_TYPE
1514 @defmacx UINT16_TYPE
1515 @defmacx UINT32_TYPE
1516 @defmacx UINT64_TYPE
1517 @defmacx INT_LEAST8_TYPE
1518 @defmacx INT_LEAST16_TYPE
1519 @defmacx INT_LEAST32_TYPE
1520 @defmacx INT_LEAST64_TYPE
1521 @defmacx UINT_LEAST8_TYPE
1522 @defmacx UINT_LEAST16_TYPE
1523 @defmacx UINT_LEAST32_TYPE
1524 @defmacx UINT_LEAST64_TYPE
1525 @defmacx INT_FAST8_TYPE
1526 @defmacx INT_FAST16_TYPE
1527 @defmacx INT_FAST32_TYPE
1528 @defmacx INT_FAST64_TYPE
1529 @defmacx UINT_FAST8_TYPE
1530 @defmacx UINT_FAST16_TYPE
1531 @defmacx UINT_FAST32_TYPE
1532 @defmacx UINT_FAST64_TYPE
1533 @defmacx INTPTR_TYPE
1534 @defmacx UINTPTR_TYPE
1535 C expressions for the standard types @code{sig_atomic_t},
1536 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1537 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1538 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1539 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1540 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1541 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1542 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1543 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1544 @code{SIZE_TYPE} above for more information.
1546 If any of these macros evaluates to a null pointer, the corresponding
1547 type is not supported; if GCC is configured to provide
1548 @code{<stdint.h>} in such a case, the header provided may not conform
1549 to C99, depending on the type in question. The defaults for all of
1550 these macros are null pointers.
1553 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1554 The C++ compiler represents a pointer-to-member-function with a struct
1561 ptrdiff_t vtable_index;
1568 The C++ compiler must use one bit to indicate whether the function that
1569 will be called through a pointer-to-member-function is virtual.
1570 Normally, we assume that the low-order bit of a function pointer must
1571 always be zero. Then, by ensuring that the vtable_index is odd, we can
1572 distinguish which variant of the union is in use. But, on some
1573 platforms function pointers can be odd, and so this doesn't work. In
1574 that case, we use the low-order bit of the @code{delta} field, and shift
1575 the remainder of the @code{delta} field to the left.
1577 GCC will automatically make the right selection about where to store
1578 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1579 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1580 set such that functions always start at even addresses, but the lowest
1581 bit of pointers to functions indicate whether the function at that
1582 address is in ARM or Thumb mode. If this is the case of your
1583 architecture, you should define this macro to
1584 @code{ptrmemfunc_vbit_in_delta}.
1586 In general, you should not have to define this macro. On architectures
1587 in which function addresses are always even, according to
1588 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1589 @code{ptrmemfunc_vbit_in_pfn}.
1592 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1593 Normally, the C++ compiler uses function pointers in vtables. This
1594 macro allows the target to change to use ``function descriptors''
1595 instead. Function descriptors are found on targets for whom a
1596 function pointer is actually a small data structure. Normally the
1597 data structure consists of the actual code address plus a data
1598 pointer to which the function's data is relative.
1600 If vtables are used, the value of this macro should be the number
1601 of words that the function descriptor occupies.
1604 @defmac TARGET_VTABLE_ENTRY_ALIGN
1605 By default, the vtable entries are void pointers, the so the alignment
1606 is the same as pointer alignment. The value of this macro specifies
1607 the alignment of the vtable entry in bits. It should be defined only
1608 when special alignment is necessary. */
1611 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1612 There are a few non-descriptor entries in the vtable at offsets below
1613 zero. If these entries must be padded (say, to preserve the alignment
1614 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1615 of words in each data entry.
1619 @section Register Usage
1620 @cindex register usage
1622 This section explains how to describe what registers the target machine
1623 has, and how (in general) they can be used.
1625 The description of which registers a specific instruction can use is
1626 done with register classes; see @ref{Register Classes}. For information
1627 on using registers to access a stack frame, see @ref{Frame Registers}.
1628 For passing values in registers, see @ref{Register Arguments}.
1629 For returning values in registers, see @ref{Scalar Return}.
1632 * Register Basics:: Number and kinds of registers.
1633 * Allocation Order:: Order in which registers are allocated.
1634 * Values in Registers:: What kinds of values each reg can hold.
1635 * Leaf Functions:: Renumbering registers for leaf functions.
1636 * Stack Registers:: Handling a register stack such as 80387.
1639 @node Register Basics
1640 @subsection Basic Characteristics of Registers
1642 @c prevent bad page break with this line
1643 Registers have various characteristics.
1645 @defmac FIRST_PSEUDO_REGISTER
1646 Number of hardware registers known to the compiler. They receive
1647 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1648 pseudo register's number really is assigned the number
1649 @code{FIRST_PSEUDO_REGISTER}.
1652 @defmac FIXED_REGISTERS
1653 @cindex fixed register
1654 An initializer that says which registers are used for fixed purposes
1655 all throughout the compiled code and are therefore not available for
1656 general allocation. These would include the stack pointer, the frame
1657 pointer (except on machines where that can be used as a general
1658 register when no frame pointer is needed), the program counter on
1659 machines where that is considered one of the addressable registers,
1660 and any other numbered register with a standard use.
1662 This information is expressed as a sequence of numbers, separated by
1663 commas and surrounded by braces. The @var{n}th number is 1 if
1664 register @var{n} is fixed, 0 otherwise.
1666 The table initialized from this macro, and the table initialized by
1667 the following one, may be overridden at run time either automatically,
1668 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1669 the user with the command options @option{-ffixed-@var{reg}},
1670 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1673 @defmac CALL_USED_REGISTERS
1674 @cindex call-used register
1675 @cindex call-clobbered register
1676 @cindex call-saved register
1677 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1678 clobbered (in general) by function calls as well as for fixed
1679 registers. This macro therefore identifies the registers that are not
1680 available for general allocation of values that must live across
1683 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1684 automatically saves it on function entry and restores it on function
1685 exit, if the register is used within the function.
1688 @defmac CALL_REALLY_USED_REGISTERS
1689 @cindex call-used register
1690 @cindex call-clobbered register
1691 @cindex call-saved register
1692 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1693 that the entire set of @code{FIXED_REGISTERS} be included.
1694 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1695 This macro is optional. If not specified, it defaults to the value
1696 of @code{CALL_USED_REGISTERS}.
1699 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1700 @cindex call-used register
1701 @cindex call-clobbered register
1702 @cindex call-saved register
1703 A C expression that is nonzero if it is not permissible to store a
1704 value of mode @var{mode} in hard register number @var{regno} across a
1705 call without some part of it being clobbered. For most machines this
1706 macro need not be defined. It is only required for machines that do not
1707 preserve the entire contents of a register across a call.
1711 @findex call_used_regs
1714 @findex reg_class_contents
1715 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1717 @defmac INCOMING_REGNO (@var{out})
1718 Define this macro if the target machine has register windows. This C
1719 expression returns the register number as seen by the called function
1720 corresponding to the register number @var{out} as seen by the calling
1721 function. Return @var{out} if register number @var{out} is not an
1725 @defmac OUTGOING_REGNO (@var{in})
1726 Define this macro if the target machine has register windows. This C
1727 expression returns the register number as seen by the calling function
1728 corresponding to the register number @var{in} as seen by the called
1729 function. Return @var{in} if register number @var{in} is not an inbound
1733 @defmac LOCAL_REGNO (@var{regno})
1734 Define this macro if the target machine has register windows. This C
1735 expression returns true if the register is call-saved but is in the
1736 register window. Unlike most call-saved registers, such registers
1737 need not be explicitly restored on function exit or during non-local
1742 If the program counter has a register number, define this as that
1743 register number. Otherwise, do not define it.
1746 @node Allocation Order
1747 @subsection Order of Allocation of Registers
1748 @cindex order of register allocation
1749 @cindex register allocation order
1751 @c prevent bad page break with this line
1752 Registers are allocated in order.
1754 @defmac REG_ALLOC_ORDER
1755 If defined, an initializer for a vector of integers, containing the
1756 numbers of hard registers in the order in which GCC should prefer
1757 to use them (from most preferred to least).
1759 If this macro is not defined, registers are used lowest numbered first
1760 (all else being equal).
1762 One use of this macro is on machines where the highest numbered
1763 registers must always be saved and the save-multiple-registers
1764 instruction supports only sequences of consecutive registers. On such
1765 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1766 the highest numbered allocable register first.
1769 @defmac ADJUST_REG_ALLOC_ORDER
1770 A C statement (sans semicolon) to choose the order in which to allocate
1771 hard registers for pseudo-registers local to a basic block.
1773 Store the desired register order in the array @code{reg_alloc_order}.
1774 Element 0 should be the register to allocate first; element 1, the next
1775 register; and so on.
1777 The macro body should not assume anything about the contents of
1778 @code{reg_alloc_order} before execution of the macro.
1780 On most machines, it is not necessary to define this macro.
1783 @defmac HONOR_REG_ALLOC_ORDER
1784 Normally, IRA tries to estimate the costs for saving a register in the
1785 prologue and restoring it in the epilogue. This discourages it from
1786 using call-saved registers. If a machine wants to ensure that IRA
1787 allocates registers in the order given by REG_ALLOC_ORDER even if some
1788 call-saved registers appear earlier than call-used ones, then define this
1789 macro as a C expression to nonzero. Default is 0.
1792 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1793 In some case register allocation order is not enough for the
1794 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1795 If this macro is defined, it should return a floating point value
1796 based on @var{regno}. The cost of using @var{regno} for a pseudo will
1797 be increased by approximately the pseudo's usage frequency times the
1798 value returned by this macro. Not defining this macro is equivalent
1799 to having it always return @code{0.0}.
1801 On most machines, it is not necessary to define this macro.
1804 @node Values in Registers
1805 @subsection How Values Fit in Registers
1807 This section discusses the macros that describe which kinds of values
1808 (specifically, which machine modes) each register can hold, and how many
1809 consecutive registers are needed for a given mode.
1811 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
1812 A C expression for the number of consecutive hard registers, starting
1813 at register number @var{regno}, required to hold a value of mode
1814 @var{mode}. This macro must never return zero, even if a register
1815 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
1816 and/or CANNOT_CHANGE_MODE_CLASS instead.
1818 On a machine where all registers are exactly one word, a suitable
1819 definition of this macro is
1822 #define HARD_REGNO_NREGS(REGNO, MODE) \
1823 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1828 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1829 A C expression that is nonzero if a value of mode @var{mode}, stored
1830 in memory, ends with padding that causes it to take up more space than
1831 in registers starting at register number @var{regno} (as determined by
1832 multiplying GCC's notion of the size of the register when containing
1833 this mode by the number of registers returned by
1834 @code{HARD_REGNO_NREGS}). By default this is zero.
1836 For example, if a floating-point value is stored in three 32-bit
1837 registers but takes up 128 bits in memory, then this would be
1840 This macros only needs to be defined if there are cases where
1841 @code{subreg_get_info}
1842 would otherwise wrongly determine that a @code{subreg} can be
1843 represented by an offset to the register number, when in fact such a
1844 @code{subreg} would contain some of the padding not stored in
1845 registers and so not be representable.
1848 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1849 For values of @var{regno} and @var{mode} for which
1850 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1851 returning the greater number of registers required to hold the value
1852 including any padding. In the example above, the value would be four.
1855 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1856 Define this macro if the natural size of registers that hold values
1857 of mode @var{mode} is not the word size. It is a C expression that
1858 should give the natural size in bytes for the specified mode. It is
1859 used by the register allocator to try to optimize its results. This
1860 happens for example on SPARC 64-bit where the natural size of
1861 floating-point registers is still 32-bit.
1864 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1865 A C expression that is nonzero if it is permissible to store a value
1866 of mode @var{mode} in hard register number @var{regno} (or in several
1867 registers starting with that one). For a machine where all registers
1868 are equivalent, a suitable definition is
1871 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1874 You need not include code to check for the numbers of fixed registers,
1875 because the allocation mechanism considers them to be always occupied.
1877 @cindex register pairs
1878 On some machines, double-precision values must be kept in even/odd
1879 register pairs. You can implement that by defining this macro to reject
1880 odd register numbers for such modes.
1882 The minimum requirement for a mode to be OK in a register is that the
1883 @samp{mov@var{mode}} instruction pattern support moves between the
1884 register and other hard register in the same class and that moving a
1885 value into the register and back out not alter it.
1887 Since the same instruction used to move @code{word_mode} will work for
1888 all narrower integer modes, it is not necessary on any machine for
1889 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1890 you define patterns @samp{movhi}, etc., to take advantage of this. This
1891 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1892 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1895 Many machines have special registers for floating point arithmetic.
1896 Often people assume that floating point machine modes are allowed only
1897 in floating point registers. This is not true. Any registers that
1898 can hold integers can safely @emph{hold} a floating point machine
1899 mode, whether or not floating arithmetic can be done on it in those
1900 registers. Integer move instructions can be used to move the values.
1902 On some machines, though, the converse is true: fixed-point machine
1903 modes may not go in floating registers. This is true if the floating
1904 registers normalize any value stored in them, because storing a
1905 non-floating value there would garble it. In this case,
1906 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1907 floating registers. But if the floating registers do not automatically
1908 normalize, if you can store any bit pattern in one and retrieve it
1909 unchanged without a trap, then any machine mode may go in a floating
1910 register, so you can define this macro to say so.
1912 The primary significance of special floating registers is rather that
1913 they are the registers acceptable in floating point arithmetic
1914 instructions. However, this is of no concern to
1915 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1916 constraints for those instructions.
1918 On some machines, the floating registers are especially slow to access,
1919 so that it is better to store a value in a stack frame than in such a
1920 register if floating point arithmetic is not being done. As long as the
1921 floating registers are not in class @code{GENERAL_REGS}, they will not
1922 be used unless some pattern's constraint asks for one.
1925 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1926 A C expression that is nonzero if it is OK to rename a hard register
1927 @var{from} to another hard register @var{to}.
1929 One common use of this macro is to prevent renaming of a register to
1930 another register that is not saved by a prologue in an interrupt
1933 The default is always nonzero.
1936 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1937 A C expression that is nonzero if a value of mode
1938 @var{mode1} is accessible in mode @var{mode2} without copying.
1940 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1941 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1942 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1943 should be nonzero. If they differ for any @var{r}, you should define
1944 this macro to return zero unless some other mechanism ensures the
1945 accessibility of the value in a narrower mode.
1947 You should define this macro to return nonzero in as many cases as
1948 possible since doing so will allow GCC to perform better register
1952 @hook TARGET_HARD_REGNO_SCRATCH_OK
1954 @defmac AVOID_CCMODE_COPIES
1955 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1956 registers. You should only define this macro if support for copying to/from
1957 @code{CCmode} is incomplete.
1960 @node Leaf Functions
1961 @subsection Handling Leaf Functions
1963 @cindex leaf functions
1964 @cindex functions, leaf
1965 On some machines, a leaf function (i.e., one which makes no calls) can run
1966 more efficiently if it does not make its own register window. Often this
1967 means it is required to receive its arguments in the registers where they
1968 are passed by the caller, instead of the registers where they would
1971 The special treatment for leaf functions generally applies only when
1972 other conditions are met; for example, often they may use only those
1973 registers for its own variables and temporaries. We use the term ``leaf
1974 function'' to mean a function that is suitable for this special
1975 handling, so that functions with no calls are not necessarily ``leaf
1978 GCC assigns register numbers before it knows whether the function is
1979 suitable for leaf function treatment. So it needs to renumber the
1980 registers in order to output a leaf function. The following macros
1983 @defmac LEAF_REGISTERS
1984 Name of a char vector, indexed by hard register number, which
1985 contains 1 for a register that is allowable in a candidate for leaf
1988 If leaf function treatment involves renumbering the registers, then the
1989 registers marked here should be the ones before renumbering---those that
1990 GCC would ordinarily allocate. The registers which will actually be
1991 used in the assembler code, after renumbering, should not be marked with 1
1994 Define this macro only if the target machine offers a way to optimize
1995 the treatment of leaf functions.
1998 @defmac LEAF_REG_REMAP (@var{regno})
1999 A C expression whose value is the register number to which @var{regno}
2000 should be renumbered, when a function is treated as a leaf function.
2002 If @var{regno} is a register number which should not appear in a leaf
2003 function before renumbering, then the expression should yield @minus{}1, which
2004 will cause the compiler to abort.
2006 Define this macro only if the target machine offers a way to optimize the
2007 treatment of leaf functions, and registers need to be renumbered to do
2011 @findex current_function_is_leaf
2012 @findex current_function_uses_only_leaf_regs
2013 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2014 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2015 specially. They can test the C variable @code{current_function_is_leaf}
2016 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2017 set prior to local register allocation and is valid for the remaining
2018 compiler passes. They can also test the C variable
2019 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2020 functions which only use leaf registers.
2021 @code{current_function_uses_only_leaf_regs} is valid after all passes
2022 that modify the instructions have been run and is only useful if
2023 @code{LEAF_REGISTERS} is defined.
2024 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2025 @c of the next paragraph?! --mew 2feb93
2027 @node Stack Registers
2028 @subsection Registers That Form a Stack
2030 There are special features to handle computers where some of the
2031 ``registers'' form a stack. Stack registers are normally written by
2032 pushing onto the stack, and are numbered relative to the top of the
2035 Currently, GCC can only handle one group of stack-like registers, and
2036 they must be consecutively numbered. Furthermore, the existing
2037 support for stack-like registers is specific to the 80387 floating
2038 point coprocessor. If you have a new architecture that uses
2039 stack-like registers, you will need to do substantial work on
2040 @file{reg-stack.c} and write your machine description to cooperate
2041 with it, as well as defining these macros.
2044 Define this if the machine has any stack-like registers.
2047 @defmac STACK_REG_COVER_CLASS
2048 This is a cover class containing the stack registers. Define this if
2049 the machine has any stack-like registers.
2052 @defmac FIRST_STACK_REG
2053 The number of the first stack-like register. This one is the top
2057 @defmac LAST_STACK_REG
2058 The number of the last stack-like register. This one is the bottom of
2062 @node Register Classes
2063 @section Register Classes
2064 @cindex register class definitions
2065 @cindex class definitions, register
2067 On many machines, the numbered registers are not all equivalent.
2068 For example, certain registers may not be allowed for indexed addressing;
2069 certain registers may not be allowed in some instructions. These machine
2070 restrictions are described to the compiler using @dfn{register classes}.
2072 You define a number of register classes, giving each one a name and saying
2073 which of the registers belong to it. Then you can specify register classes
2074 that are allowed as operands to particular instruction patterns.
2078 In general, each register will belong to several classes. In fact, one
2079 class must be named @code{ALL_REGS} and contain all the registers. Another
2080 class must be named @code{NO_REGS} and contain no registers. Often the
2081 union of two classes will be another class; however, this is not required.
2083 @findex GENERAL_REGS
2084 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2085 terribly special about the name, but the operand constraint letters
2086 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2087 the same as @code{ALL_REGS}, just define it as a macro which expands
2090 Order the classes so that if class @var{x} is contained in class @var{y}
2091 then @var{x} has a lower class number than @var{y}.
2093 The way classes other than @code{GENERAL_REGS} are specified in operand
2094 constraints is through machine-dependent operand constraint letters.
2095 You can define such letters to correspond to various classes, then use
2096 them in operand constraints.
2098 You must define the narrowest register classes for allocatable
2099 registers, so that each class either has no subclasses, or that for
2100 some mode, the move cost between registers within the class is
2101 cheaper than moving a register in the class to or from memory
2104 You should define a class for the union of two classes whenever some
2105 instruction allows both classes. For example, if an instruction allows
2106 either a floating point (coprocessor) register or a general register for a
2107 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2108 which includes both of them. Otherwise you will get suboptimal code,
2109 or even internal compiler errors when reload cannot find a register in the
2110 class computed via @code{reg_class_subunion}.
2112 You must also specify certain redundant information about the register
2113 classes: for each class, which classes contain it and which ones are
2114 contained in it; for each pair of classes, the largest class contained
2117 When a value occupying several consecutive registers is expected in a
2118 certain class, all the registers used must belong to that class.
2119 Therefore, register classes cannot be used to enforce a requirement for
2120 a register pair to start with an even-numbered register. The way to
2121 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2123 Register classes used for input-operands of bitwise-and or shift
2124 instructions have a special requirement: each such class must have, for
2125 each fixed-point machine mode, a subclass whose registers can transfer that
2126 mode to or from memory. For example, on some machines, the operations for
2127 single-byte values (@code{QImode}) are limited to certain registers. When
2128 this is so, each register class that is used in a bitwise-and or shift
2129 instruction must have a subclass consisting of registers from which
2130 single-byte values can be loaded or stored. This is so that
2131 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2133 @deftp {Data type} {enum reg_class}
2134 An enumerated type that must be defined with all the register class names
2135 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2136 must be the last register class, followed by one more enumerated value,
2137 @code{LIM_REG_CLASSES}, which is not a register class but rather
2138 tells how many classes there are.
2140 Each register class has a number, which is the value of casting
2141 the class name to type @code{int}. The number serves as an index
2142 in many of the tables described below.
2145 @defmac N_REG_CLASSES
2146 The number of distinct register classes, defined as follows:
2149 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2153 @defmac REG_CLASS_NAMES
2154 An initializer containing the names of the register classes as C string
2155 constants. These names are used in writing some of the debugging dumps.
2158 @defmac REG_CLASS_CONTENTS
2159 An initializer containing the contents of the register classes, as integers
2160 which are bit masks. The @var{n}th integer specifies the contents of class
2161 @var{n}. The way the integer @var{mask} is interpreted is that
2162 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2164 When the machine has more than 32 registers, an integer does not suffice.
2165 Then the integers are replaced by sub-initializers, braced groupings containing
2166 several integers. Each sub-initializer must be suitable as an initializer
2167 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2168 In this situation, the first integer in each sub-initializer corresponds to
2169 registers 0 through 31, the second integer to registers 32 through 63, and
2173 @defmac REGNO_REG_CLASS (@var{regno})
2174 A C expression whose value is a register class containing hard register
2175 @var{regno}. In general there is more than one such class; choose a class
2176 which is @dfn{minimal}, meaning that no smaller class also contains the
2180 @defmac BASE_REG_CLASS
2181 A macro whose definition is the name of the class to which a valid
2182 base register must belong. A base register is one used in an address
2183 which is the register value plus a displacement.
2186 @defmac MODE_BASE_REG_CLASS (@var{mode})
2187 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2188 the selection of a base register in a mode dependent manner. If
2189 @var{mode} is VOIDmode then it should return the same value as
2190 @code{BASE_REG_CLASS}.
2193 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2194 A C expression whose value is the register class to which a valid
2195 base register must belong in order to be used in a base plus index
2196 register address. You should define this macro if base plus index
2197 addresses have different requirements than other base register uses.
2200 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2201 A C expression whose value is the register class to which a valid
2202 base register for a memory reference in mode @var{mode} to address
2203 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2204 define the context in which the base register occurs. @var{outer_code} is
2205 the code of the immediately enclosing expression (@code{MEM} for the top level
2206 of an address, @code{ADDRESS} for something that occurs in an
2207 @code{address_operand}). @var{index_code} is the code of the corresponding
2208 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2211 @defmac INDEX_REG_CLASS
2212 A macro whose definition is the name of the class to which a valid
2213 index register must belong. An index register is one used in an
2214 address where its value is either multiplied by a scale factor or
2215 added to another register (as well as added to a displacement).
2218 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2219 A C expression which is nonzero if register number @var{num} is
2220 suitable for use as a base register in operand addresses.
2223 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2224 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2225 that expression may examine the mode of the memory reference in
2226 @var{mode}. You should define this macro if the mode of the memory
2227 reference affects whether a register may be used as a base register. If
2228 you define this macro, the compiler will use it instead of
2229 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2230 addresses that appear outside a @code{MEM}, i.e., as an
2231 @code{address_operand}.
2234 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2235 A C expression which is nonzero if register number @var{num} is suitable for
2236 use as a base register in base plus index operand addresses, accessing
2237 memory in mode @var{mode}. It may be either a suitable hard register or a
2238 pseudo register that has been allocated such a hard register. You should
2239 define this macro if base plus index addresses have different requirements
2240 than other base register uses.
2242 Use of this macro is deprecated; please use the more general
2243 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2246 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2247 A C expression which is nonzero if register number @var{num} is
2248 suitable for use as a base register in operand addresses, accessing
2249 memory in mode @var{mode} in address space @var{address_space}.
2250 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2251 that that expression may examine the context in which the register
2252 appears in the memory reference. @var{outer_code} is the code of the
2253 immediately enclosing expression (@code{MEM} if at the top level of the
2254 address, @code{ADDRESS} for something that occurs in an
2255 @code{address_operand}). @var{index_code} is the code of the
2256 corresponding index expression if @var{outer_code} is @code{PLUS};
2257 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2258 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2261 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2262 A C expression which is nonzero if register number @var{num} is
2263 suitable for use as an index register in operand addresses. It may be
2264 either a suitable hard register or a pseudo register that has been
2265 allocated such a hard register.
2267 The difference between an index register and a base register is that
2268 the index register may be scaled. If an address involves the sum of
2269 two registers, neither one of them scaled, then either one may be
2270 labeled the ``base'' and the other the ``index''; but whichever
2271 labeling is used must fit the machine's constraints of which registers
2272 may serve in each capacity. The compiler will try both labelings,
2273 looking for one that is valid, and will reload one or both registers
2274 only if neither labeling works.
2277 @hook TARGET_PREFERRED_RENAME_CLASS
2279 @hook TARGET_PREFERRED_RELOAD_CLASS
2281 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2282 A C expression that places additional restrictions on the register class
2283 to use when it is necessary to copy value @var{x} into a register in class
2284 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2285 another, smaller class. On many machines, the following definition is
2289 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2292 Sometimes returning a more restrictive class makes better code. For
2293 example, on the 68000, when @var{x} is an integer constant that is in range
2294 for a @samp{moveq} instruction, the value of this macro is always
2295 @code{DATA_REGS} as long as @var{class} includes the data registers.
2296 Requiring a data register guarantees that a @samp{moveq} will be used.
2298 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2299 @var{class} is if @var{x} is a legitimate constant which cannot be
2300 loaded into some register class. By returning @code{NO_REGS} you can
2301 force @var{x} into a memory location. For example, rs6000 can load
2302 immediate values into general-purpose registers, but does not have an
2303 instruction for loading an immediate value into a floating-point
2304 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2305 @var{x} is a floating-point constant. If the constant can't be loaded
2306 into any kind of register, code generation will be better if
2307 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2308 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2310 If an insn has pseudos in it after register allocation, reload will go
2311 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2312 to find the best one. Returning @code{NO_REGS}, in this case, makes
2313 reload add a @code{!} in front of the constraint: the x86 back-end uses
2314 this feature to discourage usage of 387 registers when math is done in
2315 the SSE registers (and vice versa).
2318 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2320 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2321 A C expression that places additional restrictions on the register class
2322 to use when it is necessary to be able to hold a value of mode
2323 @var{mode} in a reload register for which class @var{class} would
2326 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2327 there are certain modes that simply can't go in certain reload classes.
2329 The value is a register class; perhaps @var{class}, or perhaps another,
2332 Don't define this macro unless the target machine has limitations which
2333 require the macro to do something nontrivial.
2336 @hook TARGET_SECONDARY_RELOAD
2338 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2339 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2340 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2341 These macros are obsolete, new ports should use the target hook
2342 @code{TARGET_SECONDARY_RELOAD} instead.
2344 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2345 target hook. Older ports still define these macros to indicate to the
2346 reload phase that it may
2347 need to allocate at least one register for a reload in addition to the
2348 register to contain the data. Specifically, if copying @var{x} to a
2349 register @var{class} in @var{mode} requires an intermediate register,
2350 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2351 largest register class all of whose registers can be used as
2352 intermediate registers or scratch registers.
2354 If copying a register @var{class} in @var{mode} to @var{x} requires an
2355 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2356 was supposed to be defined be defined to return the largest register
2357 class required. If the
2358 requirements for input and output reloads were the same, the macro
2359 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2362 The values returned by these macros are often @code{GENERAL_REGS}.
2363 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2364 can be directly copied to or from a register of @var{class} in
2365 @var{mode} without requiring a scratch register. Do not define this
2366 macro if it would always return @code{NO_REGS}.
2368 If a scratch register is required (either with or without an
2369 intermediate register), you were supposed to define patterns for
2370 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2371 (@pxref{Standard Names}. These patterns, which were normally
2372 implemented with a @code{define_expand}, should be similar to the
2373 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2376 These patterns need constraints for the reload register and scratch
2378 contain a single register class. If the original reload register (whose
2379 class is @var{class}) can meet the constraint given in the pattern, the
2380 value returned by these macros is used for the class of the scratch
2381 register. Otherwise, two additional reload registers are required.
2382 Their classes are obtained from the constraints in the insn pattern.
2384 @var{x} might be a pseudo-register or a @code{subreg} of a
2385 pseudo-register, which could either be in a hard register or in memory.
2386 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2387 in memory and the hard register number if it is in a register.
2389 These macros should not be used in the case where a particular class of
2390 registers can only be copied to memory and not to another class of
2391 registers. In that case, secondary reload registers are not needed and
2392 would not be helpful. Instead, a stack location must be used to perform
2393 the copy and the @code{mov@var{m}} pattern should use memory as an
2394 intermediate storage. This case often occurs between floating-point and
2398 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2399 Certain machines have the property that some registers cannot be copied
2400 to some other registers without using memory. Define this macro on
2401 those machines to be a C expression that is nonzero if objects of mode
2402 @var{m} in registers of @var{class1} can only be copied to registers of
2403 class @var{class2} by storing a register of @var{class1} into memory
2404 and loading that memory location into a register of @var{class2}.
2406 Do not define this macro if its value would always be zero.
2409 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2410 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2411 allocates a stack slot for a memory location needed for register copies.
2412 If this macro is defined, the compiler instead uses the memory location
2413 defined by this macro.
2415 Do not define this macro if you do not define
2416 @code{SECONDARY_MEMORY_NEEDED}.
2419 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2420 When the compiler needs a secondary memory location to copy between two
2421 registers of mode @var{mode}, it normally allocates sufficient memory to
2422 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2423 load operations in a mode that many bits wide and whose class is the
2424 same as that of @var{mode}.
2426 This is right thing to do on most machines because it ensures that all
2427 bits of the register are copied and prevents accesses to the registers
2428 in a narrower mode, which some machines prohibit for floating-point
2431 However, this default behavior is not correct on some machines, such as
2432 the DEC Alpha, that store short integers in floating-point registers
2433 differently than in integer registers. On those machines, the default
2434 widening will not work correctly and you must define this macro to
2435 suppress that widening in some cases. See the file @file{alpha.h} for
2438 Do not define this macro if you do not define
2439 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2440 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2443 @hook TARGET_CLASS_LIKELY_SPILLED_P
2445 @hook TARGET_CLASS_MAX_NREGS
2447 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2448 A C expression for the maximum number of consecutive registers
2449 of class @var{class} needed to hold a value of mode @var{mode}.
2451 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2452 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2453 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2454 @var{mode})} for all @var{regno} values in the class @var{class}.
2456 This macro helps control the handling of multiple-word values
2460 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2461 If defined, a C expression that returns nonzero for a @var{class} for which
2462 a change from mode @var{from} to mode @var{to} is invalid.
2464 For example, loading 32-bit integer or floating-point objects into
2465 floating-point registers on Alpha extends them to 64 bits.
2466 Therefore loading a 64-bit object and then storing it as a 32-bit object
2467 does not store the low-order 32 bits, as would be the case for a normal
2468 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2472 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2473 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2474 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2477 Even if storing from a register in mode @var{to} would be valid,
2478 if both @var{from} and @code{raw_reg_mode} for @var{class} are wider
2479 than @code{word_mode}, then we must prevent @var{to} narrowing the
2480 mode. This happens when the middle-end assumes that it can load
2481 or store pieces of an @var{N}-word pseudo, and that the pseudo will
2482 eventually be allocated to @var{N} @code{word_mode} hard registers.
2483 Failure to prevent this kind of mode change will result in the
2484 entire @code{raw_reg_mode} being modified instead of the partial
2485 value that the middle-end intended.
2489 @hook TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS
2493 @hook TARGET_REGISTER_PRIORITY
2495 @hook TARGET_REGISTER_USAGE_LEVELING_P
2497 @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2499 @hook TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P
2501 @hook TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT
2503 @hook TARGET_SPILL_CLASS
2505 @hook TARGET_ADDITIONAL_ALLOCNO_CLASS_P
2507 @hook TARGET_CSTORE_MODE
2509 @hook TARGET_COMPUTE_PRESSURE_CLASSES
2511 @node Stack and Calling
2512 @section Stack Layout and Calling Conventions
2513 @cindex calling conventions
2515 @c prevent bad page break with this line
2516 This describes the stack layout and calling conventions.
2520 * Exception Handling::
2525 * Register Arguments::
2527 * Aggregate Return::
2532 * Shrink-wrapping separate components::
2533 * Stack Smashing Protection::
2534 * Miscellaneous Register Hooks::
2538 @subsection Basic Stack Layout
2539 @cindex stack frame layout
2540 @cindex frame layout
2542 @c prevent bad page break with this line
2543 Here is the basic stack layout.
2545 @defmac STACK_GROWS_DOWNWARD
2546 Define this macro to be true if pushing a word onto the stack moves the stack
2547 pointer to a smaller address, and false otherwise.
2550 @defmac STACK_PUSH_CODE
2551 This macro defines the operation used when something is pushed
2552 on the stack. In RTL, a push operation will be
2553 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2555 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2556 and @code{POST_INC}. Which of these is correct depends on
2557 the stack direction and on whether the stack pointer points
2558 to the last item on the stack or whether it points to the
2559 space for the next item on the stack.
2561 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2562 true, which is almost always right, and @code{PRE_INC} otherwise,
2563 which is often wrong.
2566 @defmac FRAME_GROWS_DOWNWARD
2567 Define this macro to nonzero value if the addresses of local variable slots
2568 are at negative offsets from the frame pointer.
2571 @defmac ARGS_GROW_DOWNWARD
2572 Define this macro if successive arguments to a function occupy decreasing
2573 addresses on the stack.
2576 @defmac STARTING_FRAME_OFFSET
2577 Offset from the frame pointer to the first local variable slot to be allocated.
2579 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2580 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2581 Otherwise, it is found by adding the length of the first slot to the
2582 value @code{STARTING_FRAME_OFFSET}.
2583 @c i'm not sure if the above is still correct.. had to change it to get
2584 @c rid of an overfull. --mew 2feb93
2587 @defmac STACK_ALIGNMENT_NEEDED
2588 Define to zero to disable final alignment of the stack during reload.
2589 The nonzero default for this macro is suitable for most ports.
2591 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2592 is a register save block following the local block that doesn't require
2593 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2594 stack alignment and do it in the backend.
2597 @defmac STACK_POINTER_OFFSET
2598 Offset from the stack pointer register to the first location at which
2599 outgoing arguments are placed. If not specified, the default value of
2600 zero is used. This is the proper value for most machines.
2602 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2603 the first location at which outgoing arguments are placed.
2606 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2607 Offset from the argument pointer register to the first argument's
2608 address. On some machines it may depend on the data type of the
2611 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2612 the first argument's address.
2615 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2616 Offset from the stack pointer register to an item dynamically allocated
2617 on the stack, e.g., by @code{alloca}.
2619 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2620 length of the outgoing arguments. The default is correct for most
2621 machines. See @file{function.c} for details.
2624 @defmac INITIAL_FRAME_ADDRESS_RTX
2625 A C expression whose value is RTL representing the address of the initial
2626 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2627 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2628 default value will be used. Define this macro in order to make frame pointer
2629 elimination work in the presence of @code{__builtin_frame_address (count)} and
2630 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2633 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2634 A C expression whose value is RTL representing the address in a stack
2635 frame where the pointer to the caller's frame is stored. Assume that
2636 @var{frameaddr} is an RTL expression for the address of the stack frame
2639 If you don't define this macro, the default is to return the value
2640 of @var{frameaddr}---that is, the stack frame address is also the
2641 address of the stack word that points to the previous frame.
2644 @defmac SETUP_FRAME_ADDRESSES
2645 A C expression that produces the machine-specific code to
2646 setup the stack so that arbitrary frames can be accessed. For example,
2647 on the SPARC, we must flush all of the register windows to the stack
2648 before we can access arbitrary stack frames. You will seldom need to
2649 define this macro. The default is to do nothing.
2652 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2654 @defmac FRAME_ADDR_RTX (@var{frameaddr})
2655 A C expression whose value is RTL representing the value of the frame
2656 address for the current frame. @var{frameaddr} is the frame pointer
2657 of the current frame. This is used for __builtin_frame_address.
2658 You need only define this macro if the frame address is not the same
2659 as the frame pointer. Most machines do not need to define it.
2662 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2663 A C expression whose value is RTL representing the value of the return
2664 address for the frame @var{count} steps up from the current frame, after
2665 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2666 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2667 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
2669 The value of the expression must always be the correct address when
2670 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2671 determine the return address of other frames.
2674 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2675 Define this macro to nonzero value if the return address of a particular
2676 stack frame is accessed from the frame pointer of the previous stack
2677 frame. The zero default for this macro is suitable for most ports.
2680 @defmac INCOMING_RETURN_ADDR_RTX
2681 A C expression whose value is RTL representing the location of the
2682 incoming return address at the beginning of any function, before the
2683 prologue. This RTL is either a @code{REG}, indicating that the return
2684 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2687 You only need to define this macro if you want to support call frame
2688 debugging information like that provided by DWARF 2.
2690 If this RTL is a @code{REG}, you should also define
2691 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2694 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2695 A C expression whose value is an integer giving a DWARF 2 column
2696 number that may be used as an alternative return column. The column
2697 must not correspond to any gcc hard register (that is, it must not
2698 be in the range of @code{DWARF_FRAME_REGNUM}).
2700 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2701 general register, but an alternative column needs to be used for signal
2702 frames. Some targets have also used different frame return columns
2706 @defmac DWARF_ZERO_REG
2707 A C expression whose value is an integer giving a DWARF 2 register
2708 number that is considered to always have the value zero. This should
2709 only be defined if the target has an architected zero register, and
2710 someone decided it was a good idea to use that register number to
2711 terminate the stack backtrace. New ports should avoid this.
2714 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2716 @defmac INCOMING_FRAME_SP_OFFSET
2717 A C expression whose value is an integer giving the offset, in bytes,
2718 from the value of the stack pointer register to the top of the stack
2719 frame at the beginning of any function, before the prologue. The top of
2720 the frame is defined to be the value of the stack pointer in the
2721 previous frame, just before the call instruction.
2723 You only need to define this macro if you want to support call frame
2724 debugging information like that provided by DWARF 2.
2727 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2728 A C expression whose value is an integer giving the offset, in bytes,
2729 from the argument pointer to the canonical frame address (cfa). The
2730 final value should coincide with that calculated by
2731 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2732 during virtual register instantiation.
2734 The default value for this macro is
2735 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2736 which is correct for most machines; in general, the arguments are found
2737 immediately before the stack frame. Note that this is not the case on
2738 some targets that save registers into the caller's frame, such as SPARC
2739 and rs6000, and so such targets need to define this macro.
2741 You only need to define this macro if the default is incorrect, and you
2742 want to support call frame debugging information like that provided by
2746 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2747 If defined, a C expression whose value is an integer giving the offset
2748 in bytes from the frame pointer to the canonical frame address (cfa).
2749 The final value should coincide with that calculated by
2750 @code{INCOMING_FRAME_SP_OFFSET}.
2752 Normally the CFA is calculated as an offset from the argument pointer,
2753 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2754 variable due to the ABI, this may not be possible. If this macro is
2755 defined, it implies that the virtual register instantiation should be
2756 based on the frame pointer instead of the argument pointer. Only one
2757 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2761 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2762 If defined, a C expression whose value is an integer giving the offset
2763 in bytes from the canonical frame address (cfa) to the frame base used
2764 in DWARF 2 debug information. The default is zero. A different value
2765 may reduce the size of debug information on some ports.
2768 @node Exception Handling
2769 @subsection Exception Handling Support
2770 @cindex exception handling
2772 @defmac EH_RETURN_DATA_REGNO (@var{N})
2773 A C expression whose value is the @var{N}th register number used for
2774 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2775 @var{N} registers are usable.
2777 The exception handling library routines communicate with the exception
2778 handlers via a set of agreed upon registers. Ideally these registers
2779 should be call-clobbered; it is possible to use call-saved registers,
2780 but may negatively impact code size. The target must support at least
2781 2 data registers, but should define 4 if there are enough free registers.
2783 You must define this macro if you want to support call frame exception
2784 handling like that provided by DWARF 2.
2787 @defmac EH_RETURN_STACKADJ_RTX
2788 A C expression whose value is RTL representing a location in which
2789 to store a stack adjustment to be applied before function return.
2790 This is used to unwind the stack to an exception handler's call frame.
2791 It will be assigned zero on code paths that return normally.
2793 Typically this is a call-clobbered hard register that is otherwise
2794 untouched by the epilogue, but could also be a stack slot.
2796 Do not define this macro if the stack pointer is saved and restored
2797 by the regular prolog and epilog code in the call frame itself; in
2798 this case, the exception handling library routines will update the
2799 stack location to be restored in place. Otherwise, you must define
2800 this macro if you want to support call frame exception handling like
2801 that provided by DWARF 2.
2804 @defmac EH_RETURN_HANDLER_RTX
2805 A C expression whose value is RTL representing a location in which
2806 to store the address of an exception handler to which we should
2807 return. It will not be assigned on code paths that return normally.
2809 Typically this is the location in the call frame at which the normal
2810 return address is stored. For targets that return by popping an
2811 address off the stack, this might be a memory address just below
2812 the @emph{target} call frame rather than inside the current call
2813 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2814 been assigned, so it may be used to calculate the location of the
2817 Some targets have more complex requirements than storing to an
2818 address calculable during initial code generation. In that case
2819 the @code{eh_return} instruction pattern should be used instead.
2821 If you want to support call frame exception handling, you must
2822 define either this macro or the @code{eh_return} instruction pattern.
2825 @defmac RETURN_ADDR_OFFSET
2826 If defined, an integer-valued C expression for which rtl will be generated
2827 to add it to the exception handler address before it is searched in the
2828 exception handling tables, and to subtract it again from the address before
2829 using it to return to the exception handler.
2832 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2833 This macro chooses the encoding of pointers embedded in the exception
2834 handling sections. If at all possible, this should be defined such
2835 that the exception handling section will not require dynamic relocations,
2836 and so may be read-only.
2838 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2839 @var{global} is true if the symbol may be affected by dynamic relocations.
2840 The macro should return a combination of the @code{DW_EH_PE_*} defines
2841 as found in @file{dwarf2.h}.
2843 If this macro is not defined, pointers will not be encoded but
2844 represented directly.
2847 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2848 This macro allows the target to emit whatever special magic is required
2849 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2850 Generic code takes care of pc-relative and indirect encodings; this must
2851 be defined if the target uses text-relative or data-relative encodings.
2853 This is a C statement that branches to @var{done} if the format was
2854 handled. @var{encoding} is the format chosen, @var{size} is the number
2855 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2859 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
2860 This macro allows the target to add CPU and operating system specific
2861 code to the call-frame unwinder for use when there is no unwind data
2862 available. The most common reason to implement this macro is to unwind
2863 through signal frames.
2865 This macro is called from @code{uw_frame_state_for} in
2866 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
2867 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2868 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2869 for the address of the code being executed and @code{context->cfa} for
2870 the stack pointer value. If the frame can be decoded, the register
2871 save addresses should be updated in @var{fs} and the macro should
2872 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
2873 the macro should evaluate to @code{_URC_END_OF_STACK}.
2875 For proper signal handling in Java this macro is accompanied by
2876 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
2879 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
2880 This macro allows the target to add operating system specific code to the
2881 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
2882 usually used for signal or interrupt frames.
2884 This macro is called from @code{uw_update_context} in libgcc's
2885 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2886 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
2887 for the abi and context in the @code{.unwabi} directive. If the
2888 @code{.unwabi} directive can be handled, the register save addresses should
2889 be updated in @var{fs}.
2892 @defmac TARGET_USES_WEAK_UNWIND_INFO
2893 A C expression that evaluates to true if the target requires unwind
2894 info to be given comdat linkage. Define it to be @code{1} if comdat
2895 linkage is necessary. The default is @code{0}.
2898 @node Stack Checking
2899 @subsection Specifying How Stack Checking is Done
2901 GCC will check that stack references are within the boundaries of the
2902 stack, if the option @option{-fstack-check} is specified, in one of
2907 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2908 will assume that you have arranged for full stack checking to be done
2909 at appropriate places in the configuration files. GCC will not do
2910 other special processing.
2913 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
2914 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
2915 that you have arranged for static stack checking (checking of the
2916 static stack frame of functions) to be done at appropriate places
2917 in the configuration files. GCC will only emit code to do dynamic
2918 stack checking (checking on dynamic stack allocations) using the third
2922 If neither of the above are true, GCC will generate code to periodically
2923 ``probe'' the stack pointer using the values of the macros defined below.
2926 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
2927 GCC will change its allocation strategy for large objects if the option
2928 @option{-fstack-check} is specified: they will always be allocated
2929 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
2931 @defmac STACK_CHECK_BUILTIN
2932 A nonzero value if stack checking is done by the configuration files in a
2933 machine-dependent manner. You should define this macro if stack checking
2934 is required by the ABI of your machine or if you would like to do stack
2935 checking in some more efficient way than the generic approach. The default
2936 value of this macro is zero.
2939 @defmac STACK_CHECK_STATIC_BUILTIN
2940 A nonzero value if static stack checking is done by the configuration files
2941 in a machine-dependent manner. You should define this macro if you would
2942 like to do static stack checking in some more efficient way than the generic
2943 approach. The default value of this macro is zero.
2946 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
2947 An integer specifying the interval at which GCC must generate stack probe
2948 instructions, defined as 2 raised to this integer. You will normally
2949 define this macro so that the interval be no larger than the size of
2950 the ``guard pages'' at the end of a stack area. The default value
2951 of 12 (4096-byte interval) is suitable for most systems.
2954 @defmac STACK_CHECK_MOVING_SP
2955 An integer which is nonzero if GCC should move the stack pointer page by page
2956 when doing probes. This can be necessary on systems where the stack pointer
2957 contains the bottom address of the memory area accessible to the executing
2958 thread at any point in time. In this situation an alternate signal stack
2959 is required in order to be able to recover from a stack overflow. The
2960 default value of this macro is zero.
2963 @defmac STACK_CHECK_PROTECT
2964 The number of bytes of stack needed to recover from a stack overflow, for
2965 languages where such a recovery is supported. The default value of 4KB/8KB
2966 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
2967 8KB/12KB with other exception handling mechanisms should be adequate for most
2968 architectures and operating systems.
2971 The following macros are relevant only if neither STACK_CHECK_BUILTIN
2972 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
2973 in the opposite case.
2975 @defmac STACK_CHECK_MAX_FRAME_SIZE
2976 The maximum size of a stack frame, in bytes. GCC will generate probe
2977 instructions in non-leaf functions to ensure at least this many bytes of
2978 stack are available. If a stack frame is larger than this size, stack
2979 checking will not be reliable and GCC will issue a warning. The
2980 default is chosen so that GCC only generates one instruction on most
2981 systems. You should normally not change the default value of this macro.
2984 @defmac STACK_CHECK_FIXED_FRAME_SIZE
2985 GCC uses this value to generate the above warning message. It
2986 represents the amount of fixed frame used by a function, not including
2987 space for any callee-saved registers, temporaries and user variables.
2988 You need only specify an upper bound for this amount and will normally
2989 use the default of four words.
2992 @defmac STACK_CHECK_MAX_VAR_SIZE
2993 The maximum size, in bytes, of an object that GCC will place in the
2994 fixed area of the stack frame when the user specifies
2995 @option{-fstack-check}.
2996 GCC computed the default from the values of the above macros and you will
2997 normally not need to override that default.
3001 @node Frame Registers
3002 @subsection Registers That Address the Stack Frame
3004 @c prevent bad page break with this line
3005 This discusses registers that address the stack frame.
3007 @defmac STACK_POINTER_REGNUM
3008 The register number of the stack pointer register, which must also be a
3009 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3010 the hardware determines which register this is.
3013 @defmac FRAME_POINTER_REGNUM
3014 The register number of the frame pointer register, which is used to
3015 access automatic variables in the stack frame. On some machines, the
3016 hardware determines which register this is. On other machines, you can
3017 choose any register you wish for this purpose.
3020 @defmac HARD_FRAME_POINTER_REGNUM
3021 On some machines the offset between the frame pointer and starting
3022 offset of the automatic variables is not known until after register
3023 allocation has been done (for example, because the saved registers are
3024 between these two locations). On those machines, define
3025 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3026 be used internally until the offset is known, and define
3027 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3028 used for the frame pointer.
3030 You should define this macro only in the very rare circumstances when it
3031 is not possible to calculate the offset between the frame pointer and
3032 the automatic variables until after register allocation has been
3033 completed. When this macro is defined, you must also indicate in your
3034 definition of @code{ELIMINABLE_REGS} how to eliminate
3035 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3036 or @code{STACK_POINTER_REGNUM}.
3038 Do not define this macro if it would be the same as
3039 @code{FRAME_POINTER_REGNUM}.
3042 @defmac ARG_POINTER_REGNUM
3043 The register number of the arg pointer register, which is used to access
3044 the function's argument list. On some machines, this is the same as the
3045 frame pointer register. On some machines, the hardware determines which
3046 register this is. On other machines, you can choose any register you
3047 wish for this purpose. If this is not the same register as the frame
3048 pointer register, then you must mark it as a fixed register according to
3049 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3050 (@pxref{Elimination}).
3053 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3054 Define this to a preprocessor constant that is nonzero if
3055 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3056 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3057 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3058 definition is not suitable for use in preprocessor conditionals.
3061 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3062 Define this to a preprocessor constant that is nonzero if
3063 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3064 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3065 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3066 definition is not suitable for use in preprocessor conditionals.
3069 @defmac RETURN_ADDRESS_POINTER_REGNUM
3070 The register number of the return address pointer register, which is used to
3071 access the current function's return address from the stack. On some
3072 machines, the return address is not at a fixed offset from the frame
3073 pointer or stack pointer or argument pointer. This register can be defined
3074 to point to the return address on the stack, and then be converted by
3075 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3077 Do not define this macro unless there is no other way to get the return
3078 address from the stack.
3081 @defmac STATIC_CHAIN_REGNUM
3082 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3083 Register numbers used for passing a function's static chain pointer. If
3084 register windows are used, the register number as seen by the called
3085 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3086 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3087 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3090 The static chain register need not be a fixed register.
3092 If the static chain is passed in memory, these macros should not be
3093 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3096 @hook TARGET_STATIC_CHAIN
3098 @defmac DWARF_FRAME_REGISTERS
3099 This macro specifies the maximum number of hard registers that can be
3100 saved in a call frame. This is used to size data structures used in
3101 DWARF2 exception handling.
3103 Prior to GCC 3.0, this macro was needed in order to establish a stable
3104 exception handling ABI in the face of adding new hard registers for ISA
3105 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3106 in the number of hard registers. Nevertheless, this macro can still be
3107 used to reduce the runtime memory requirements of the exception handling
3108 routines, which can be substantial if the ISA contains a lot of
3109 registers that are not call-saved.
3111 If this macro is not defined, it defaults to
3112 @code{FIRST_PSEUDO_REGISTER}.
3115 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3117 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3118 for backward compatibility in pre GCC 3.0 compiled code.
3120 If this macro is not defined, it defaults to
3121 @code{DWARF_FRAME_REGISTERS}.
3124 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3126 Define this macro if the target's representation for dwarf registers
3127 is different than the internal representation for unwind column.
3128 Given a dwarf register, this macro should return the internal unwind
3129 column number to use instead.
3131 See the PowerPC's SPE target for an example.
3134 @defmac DWARF_FRAME_REGNUM (@var{regno})
3136 Define this macro if the target's representation for dwarf registers
3137 used in .eh_frame or .debug_frame is different from that used in other
3138 debug info sections. Given a GCC hard register number, this macro
3139 should return the .eh_frame register number. The default is
3140 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3144 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3146 Define this macro to map register numbers held in the call frame info
3147 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3148 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3149 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3150 return @code{@var{regno}}.
3154 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3156 Define this macro if the target stores register values as
3157 @code{_Unwind_Word} type in unwind context. It should be defined if
3158 target register size is larger than the size of @code{void *}. The
3159 default is to store register values as @code{void *} type.
3163 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3165 Define this macro to be 1 if the target always uses extended unwind
3166 context with version, args_size and by_value fields. If it is undefined,
3167 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3168 defined and 0 otherwise.
3173 @subsection Eliminating Frame Pointer and Arg Pointer
3175 @c prevent bad page break with this line
3176 This is about eliminating the frame pointer and arg pointer.
3178 @hook TARGET_FRAME_POINTER_REQUIRED
3180 @defmac ELIMINABLE_REGS
3181 This macro specifies a table of register pairs used to eliminate
3182 unneeded registers that point into the stack frame.
3184 The definition of this macro is a list of structure initializations, each
3185 of which specifies an original and replacement register.
3187 On some machines, the position of the argument pointer is not known until
3188 the compilation is completed. In such a case, a separate hard register
3189 must be used for the argument pointer. This register can be eliminated by
3190 replacing it with either the frame pointer or the argument pointer,
3191 depending on whether or not the frame pointer has been eliminated.
3193 In this case, you might specify:
3195 #define ELIMINABLE_REGS \
3196 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3197 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3198 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3201 Note that the elimination of the argument pointer with the stack pointer is
3202 specified first since that is the preferred elimination.
3205 @hook TARGET_CAN_ELIMINATE
3207 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3208 This macro returns the initial difference between the specified pair
3209 of registers. The value would be computed from information
3210 such as the result of @code{get_frame_size ()} and the tables of
3211 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3214 @node Stack Arguments
3215 @subsection Passing Function Arguments on the Stack
3216 @cindex arguments on stack
3217 @cindex stack arguments
3219 The macros in this section control how arguments are passed
3220 on the stack. See the following section for other macros that
3221 control passing certain arguments in registers.
3223 @hook TARGET_PROMOTE_PROTOTYPES
3226 A C expression. If nonzero, push insns will be used to pass
3228 If the target machine does not have a push instruction, set it to zero.
3229 That directs GCC to use an alternate strategy: to
3230 allocate the entire argument block and then store the arguments into
3231 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3234 @defmac PUSH_ARGS_REVERSED
3235 A C expression. If nonzero, function arguments will be evaluated from
3236 last to first, rather than from first to last. If this macro is not
3237 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3238 and args grow in opposite directions, and 0 otherwise.
3241 @defmac PUSH_ROUNDING (@var{npushed})
3242 A C expression that is the number of bytes actually pushed onto the
3243 stack when an instruction attempts to push @var{npushed} bytes.
3245 On some machines, the definition
3248 #define PUSH_ROUNDING(BYTES) (BYTES)
3252 will suffice. But on other machines, instructions that appear
3253 to push one byte actually push two bytes in an attempt to maintain
3254 alignment. Then the definition should be
3257 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3260 If the value of this macro has a type, it should be an unsigned type.
3263 @findex outgoing_args_size
3264 @findex crtl->outgoing_args_size
3265 @defmac ACCUMULATE_OUTGOING_ARGS
3266 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3267 will be computed and placed into
3268 @code{crtl->outgoing_args_size}. No space will be pushed
3269 onto the stack for each call; instead, the function prologue should
3270 increase the stack frame size by this amount.
3272 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3276 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3277 Define this macro if functions should assume that stack space has been
3278 allocated for arguments even when their values are passed in
3281 The value of this macro is the size, in bytes, of the area reserved for
3282 arguments passed in registers for the function represented by @var{fndecl},
3283 which can be zero if GCC is calling a library function.
3284 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3287 This space can be allocated by the caller, or be a part of the
3288 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3291 @c above is overfull. not sure what to do. --mew 5feb93 did
3292 @c something, not sure if it looks good. --mew 10feb93
3294 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3295 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3296 Define this macro if space guaranteed when compiling a function body
3297 is different to space required when making a call, a situation that
3298 can arise with K&R style function definitions.
3301 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3302 Define this to a nonzero value if it is the responsibility of the
3303 caller to allocate the area reserved for arguments passed in registers
3304 when calling a function of @var{fntype}. @var{fntype} may be NULL
3305 if the function called is a library function.
3307 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3308 whether the space for these arguments counts in the value of
3309 @code{crtl->outgoing_args_size}.
3312 @defmac STACK_PARMS_IN_REG_PARM_AREA
3313 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3314 stack parameters don't skip the area specified by it.
3315 @c i changed this, makes more sens and it should have taken care of the
3316 @c overfull.. not as specific, tho. --mew 5feb93
3318 Normally, when a parameter is not passed in registers, it is placed on the
3319 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3320 suppresses this behavior and causes the parameter to be passed on the
3321 stack in its natural location.
3324 @hook TARGET_RETURN_POPS_ARGS
3326 @defmac CALL_POPS_ARGS (@var{cum})
3327 A C expression that should indicate the number of bytes a call sequence
3328 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3329 when compiling a function call.
3331 @var{cum} is the variable in which all arguments to the called function
3332 have been accumulated.
3334 On certain architectures, such as the SH5, a call trampoline is used
3335 that pops certain registers off the stack, depending on the arguments
3336 that have been passed to the function. Since this is a property of the
3337 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3341 @node Register Arguments
3342 @subsection Passing Arguments in Registers
3343 @cindex arguments in registers
3344 @cindex registers arguments
3346 This section describes the macros which let you control how various
3347 types of arguments are passed in registers or how they are arranged in
3350 @hook TARGET_FUNCTION_ARG
3352 @hook TARGET_MUST_PASS_IN_STACK
3354 @hook TARGET_FUNCTION_INCOMING_ARG
3356 @hook TARGET_USE_PSEUDO_PIC_REG
3358 @hook TARGET_INIT_PIC_REG
3360 @hook TARGET_ARG_PARTIAL_BYTES
3362 @hook TARGET_PASS_BY_REFERENCE
3364 @hook TARGET_CALLEE_COPIES
3366 @defmac CUMULATIVE_ARGS
3367 A C type for declaring a variable that is used as the first argument
3368 of @code{TARGET_FUNCTION_ARG} and other related values. For some
3369 target machines, the type @code{int} suffices and can hold the number
3370 of bytes of argument so far.
3372 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3373 arguments that have been passed on the stack. The compiler has other
3374 variables to keep track of that. For target machines on which all
3375 arguments are passed on the stack, there is no need to store anything in
3376 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3377 should not be empty, so use @code{int}.
3380 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3381 If defined, this macro is called before generating any code for a
3382 function, but after the @var{cfun} descriptor for the function has been
3383 created. The back end may use this macro to update @var{cfun} to
3384 reflect an ABI other than that which would normally be used by default.
3385 If the compiler is generating code for a compiler-generated function,
3386 @var{fndecl} may be @code{NULL}.
3389 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3390 A C statement (sans semicolon) for initializing the variable
3391 @var{cum} for the state at the beginning of the argument list. The
3392 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3393 is the tree node for the data type of the function which will receive
3394 the args, or 0 if the args are to a compiler support library function.
3395 For direct calls that are not libcalls, @var{fndecl} contain the
3396 declaration node of the function. @var{fndecl} is also set when
3397 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3398 being compiled. @var{n_named_args} is set to the number of named
3399 arguments, including a structure return address if it is passed as a
3400 parameter, when making a call. When processing incoming arguments,
3401 @var{n_named_args} is set to @minus{}1.
3403 When processing a call to a compiler support library function,
3404 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3405 contains the name of the function, as a string. @var{libname} is 0 when
3406 an ordinary C function call is being processed. Thus, each time this
3407 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3408 never both of them at once.
3411 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3412 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3413 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3414 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3415 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3416 0)} is used instead.
3419 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3420 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3421 finding the arguments for the function being compiled. If this macro is
3422 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3424 The value passed for @var{libname} is always 0, since library routines
3425 with special calling conventions are never compiled with GCC@. The
3426 argument @var{libname} exists for symmetry with
3427 @code{INIT_CUMULATIVE_ARGS}.
3428 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3429 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3432 @hook TARGET_FUNCTION_ARG_ADVANCE
3434 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
3435 If defined, a C expression that is the number of bytes to add to the
3436 offset of the argument passed in memory. This is needed for the SPU,
3437 which passes @code{char} and @code{short} arguments in the preferred
3438 slot that is in the middle of the quad word instead of starting at the
3442 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3443 If defined, a C expression which determines whether, and in which direction,
3444 to pad out an argument with extra space. The value should be of type
3445 @code{enum direction}: either @code{upward} to pad above the argument,
3446 @code{downward} to pad below, or @code{none} to inhibit padding.
3448 The @emph{amount} of padding is not controlled by this macro, but by the
3449 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
3450 always just enough to reach the next multiple of that boundary.
3452 This macro has a default definition which is right for most systems.
3453 For little-endian machines, the default is to pad upward. For
3454 big-endian machines, the default is to pad downward for an argument of
3455 constant size shorter than an @code{int}, and upward otherwise.
3458 @defmac PAD_VARARGS_DOWN
3459 If defined, a C expression which determines whether the default
3460 implementation of va_arg will attempt to pad down before reading the
3461 next argument, if that argument is smaller than its aligned space as
3462 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3463 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3466 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3467 Specify padding for the last element of a block move between registers and
3468 memory. @var{first} is nonzero if this is the only element. Defining this
3469 macro allows better control of register function parameters on big-endian
3470 machines, without using @code{PARALLEL} rtl. In particular,
3471 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3472 registers, as there is no longer a "wrong" part of a register; For example,
3473 a three byte aggregate may be passed in the high part of a register if so
3477 @hook TARGET_FUNCTION_ARG_BOUNDARY
3479 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3481 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3482 A C expression that is nonzero if @var{regno} is the number of a hard
3483 register in which function arguments are sometimes passed. This does
3484 @emph{not} include implicit arguments such as the static chain and
3485 the structure-value address. On many machines, no registers can be
3486 used for this purpose since all function arguments are pushed on the
3490 @hook TARGET_SPLIT_COMPLEX_ARG
3492 @hook TARGET_BUILD_BUILTIN_VA_LIST
3494 @hook TARGET_ENUM_VA_LIST_P
3496 @hook TARGET_FN_ABI_VA_LIST
3498 @hook TARGET_CANONICAL_VA_LIST_TYPE
3500 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
3502 @hook TARGET_VALID_POINTER_MODE
3504 @hook TARGET_REF_MAY_ALIAS_ERRNO
3506 @hook TARGET_SCALAR_MODE_SUPPORTED_P
3508 @hook TARGET_VECTOR_MODE_SUPPORTED_P
3510 @hook TARGET_ARRAY_MODE_SUPPORTED_P
3512 @hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P
3514 @hook TARGET_FLOATN_MODE
3516 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3519 @subsection How Scalar Function Values Are Returned
3520 @cindex return values in registers
3521 @cindex values, returned by functions
3522 @cindex scalars, returned as values
3524 This section discusses the macros that control returning scalars as
3525 values---values that can fit in registers.
3527 @hook TARGET_FUNCTION_VALUE
3529 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3530 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
3531 a new target instead.
3534 @defmac LIBCALL_VALUE (@var{mode})
3535 A C expression to create an RTX representing the place where a library
3536 function returns a value of mode @var{mode}.
3538 Note that ``library function'' in this context means a compiler
3539 support routine, used to perform arithmetic, whose name is known
3540 specially by the compiler and was not mentioned in the C code being
3544 @hook TARGET_LIBCALL_VALUE
3546 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3547 A C expression that is nonzero if @var{regno} is the number of a hard
3548 register in which the values of called function may come back.
3550 A register whose use for returning values is limited to serving as the
3551 second of a pair (for a value of type @code{double}, say) need not be
3552 recognized by this macro. So for most machines, this definition
3556 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3559 If the machine has register windows, so that the caller and the called
3560 function use different registers for the return value, this macro
3561 should recognize only the caller's register numbers.
3563 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3564 for a new target instead.
3567 @hook TARGET_FUNCTION_VALUE_REGNO_P
3569 @defmac APPLY_RESULT_SIZE
3570 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3571 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3572 saving and restoring an arbitrary return value.
3575 @hook TARGET_OMIT_STRUCT_RETURN_REG
3577 @hook TARGET_RETURN_IN_MSB
3579 @node Aggregate Return
3580 @subsection How Large Values Are Returned
3581 @cindex aggregates as return values
3582 @cindex large return values
3583 @cindex returning aggregate values
3584 @cindex structure value address
3586 When a function value's mode is @code{BLKmode} (and in some other
3587 cases), the value is not returned according to
3588 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
3589 caller passes the address of a block of memory in which the value
3590 should be stored. This address is called the @dfn{structure value
3593 This section describes how to control returning structure values in
3596 @hook TARGET_RETURN_IN_MEMORY
3598 @defmac DEFAULT_PCC_STRUCT_RETURN
3599 Define this macro to be 1 if all structure and union return values must be
3600 in memory. Since this results in slower code, this should be defined
3601 only if needed for compatibility with other compilers or with an ABI@.
3602 If you define this macro to be 0, then the conventions used for structure
3603 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3606 If not defined, this defaults to the value 1.
3609 @hook TARGET_STRUCT_VALUE_RTX
3611 @defmac PCC_STATIC_STRUCT_RETURN
3612 Define this macro if the usual system convention on the target machine
3613 for returning structures and unions is for the called function to return
3614 the address of a static variable containing the value.
3616 Do not define this if the usual system convention is for the caller to
3617 pass an address to the subroutine.
3619 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3620 nothing when you use @option{-freg-struct-return} mode.
3623 @hook TARGET_GET_RAW_RESULT_MODE
3625 @hook TARGET_GET_RAW_ARG_MODE
3628 @subsection Caller-Saves Register Allocation
3630 If you enable it, GCC can save registers around function calls. This
3631 makes it possible to use call-clobbered registers to hold variables that
3632 must live across calls.
3634 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3635 A C expression specifying which mode is required for saving @var{nregs}
3636 of a pseudo-register in call-clobbered hard register @var{regno}. If
3637 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3638 returned. For most machines this macro need not be defined since GCC
3639 will select the smallest suitable mode.
3642 @node Function Entry
3643 @subsection Function Entry and Exit
3644 @cindex function entry and exit
3648 This section describes the macros that output function entry
3649 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3651 @hook TARGET_ASM_FUNCTION_PROLOGUE
3653 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
3655 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3657 @hook TARGET_ASM_FUNCTION_EPILOGUE
3661 @findex pretend_args_size
3662 @findex crtl->args.pretend_args_size
3663 A region of @code{crtl->args.pretend_args_size} bytes of
3664 uninitialized space just underneath the first argument arriving on the
3665 stack. (This may not be at the very start of the allocated stack region
3666 if the calling sequence has pushed anything else since pushing the stack
3667 arguments. But usually, on such machines, nothing else has been pushed
3668 yet, because the function prologue itself does all the pushing.) This
3669 region is used on machines where an argument may be passed partly in
3670 registers and partly in memory, and, in some cases to support the
3671 features in @code{<stdarg.h>}.
3674 An area of memory used to save certain registers used by the function.
3675 The size of this area, which may also include space for such things as
3676 the return address and pointers to previous stack frames, is
3677 machine-specific and usually depends on which registers have been used
3678 in the function. Machines with register windows often do not require
3682 A region of at least @var{size} bytes, possibly rounded up to an allocation
3683 boundary, to contain the local variables of the function. On some machines,
3684 this region and the save area may occur in the opposite order, with the
3685 save area closer to the top of the stack.
3688 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3689 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3690 @code{crtl->outgoing_args_size} bytes to be used for outgoing
3691 argument lists of the function. @xref{Stack Arguments}.
3694 @defmac EXIT_IGNORE_STACK
3695 Define this macro as a C expression that is nonzero if the return
3696 instruction or the function epilogue ignores the value of the stack
3697 pointer; in other words, if it is safe to delete an instruction to
3698 adjust the stack pointer before a return from the function. The
3701 Note that this macro's value is relevant only for functions for which
3702 frame pointers are maintained. It is never safe to delete a final
3703 stack adjustment in a function that has no frame pointer, and the
3704 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3707 @defmac EPILOGUE_USES (@var{regno})
3708 Define this macro as a C expression that is nonzero for registers that are
3709 used by the epilogue or the @samp{return} pattern. The stack and frame
3710 pointer registers are already assumed to be used as needed.
3713 @defmac EH_USES (@var{regno})
3714 Define this macro as a C expression that is nonzero for registers that are
3715 used by the exception handling mechanism, and so should be considered live
3716 on entry to an exception edge.
3719 @hook TARGET_ASM_OUTPUT_MI_THUNK
3721 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3724 @subsection Generating Code for Profiling
3725 @cindex profiling, code generation
3727 These macros will help you generate code for profiling.
3729 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3730 A C statement or compound statement to output to @var{file} some
3731 assembler code to call the profiling subroutine @code{mcount}.
3734 The details of how @code{mcount} expects to be called are determined by
3735 your operating system environment, not by GCC@. To figure them out,
3736 compile a small program for profiling using the system's installed C
3737 compiler and look at the assembler code that results.
3739 Older implementations of @code{mcount} expect the address of a counter
3740 variable to be loaded into some register. The name of this variable is
3741 @samp{LP} followed by the number @var{labelno}, so you would generate
3742 the name using @samp{LP%d} in a @code{fprintf}.
3745 @defmac PROFILE_HOOK
3746 A C statement or compound statement to output to @var{file} some assembly
3747 code to call the profiling subroutine @code{mcount} even the target does
3748 not support profiling.
3751 @defmac NO_PROFILE_COUNTERS
3752 Define this macro to be an expression with a nonzero value if the
3753 @code{mcount} subroutine on your system does not need a counter variable
3754 allocated for each function. This is true for almost all modern
3755 implementations. If you define this macro, you must not use the
3756 @var{labelno} argument to @code{FUNCTION_PROFILER}.
3759 @defmac PROFILE_BEFORE_PROLOGUE
3760 Define this macro if the code for function profiling should come before
3761 the function prologue. Normally, the profiling code comes after.
3764 @hook TARGET_KEEP_LEAF_WHEN_PROFILED
3767 @subsection Permitting tail calls
3770 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
3772 @hook TARGET_EXTRA_LIVE_ON_ENTRY
3774 @hook TARGET_SET_UP_BY_PROLOGUE
3776 @hook TARGET_WARN_FUNC_RETURN
3778 @node Shrink-wrapping separate components
3779 @subsection Shrink-wrapping separate components
3780 @cindex shrink-wrapping separate components
3782 The prologue may perform a variety of target dependent tasks such as
3783 saving callee-saved registers, saving the return address, aligning the
3784 stack, creating a stack frame, initializing the PIC register, setting
3785 up the static chain, etc.
3787 On some targets some of these tasks may be independent of others and
3788 thus may be shrink-wrapped separately. These independent tasks are
3789 referred to as components and are handled generically by the target
3790 independent parts of GCC.
3792 Using the following hooks those prologue or epilogue components can be
3793 shrink-wrapped separately, so that the initialization (and possibly
3794 teardown) those components do is not done as frequently on execution
3795 paths where this would unnecessary.
3797 What exactly those components are is up to the target code; the generic
3798 code treats them abstractly, as a bit in an @code{sbitmap}. These
3799 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
3800 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
3803 @hook TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
3805 @hook TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
3807 @hook TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
3809 @hook TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS
3811 @hook TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS
3813 @hook TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
3815 @node Stack Smashing Protection
3816 @subsection Stack smashing protection
3817 @cindex stack smashing protection
3819 @hook TARGET_STACK_PROTECT_GUARD
3821 @hook TARGET_STACK_PROTECT_FAIL
3823 @hook TARGET_STACK_PROTECT_RUNTIME_ENABLED_P
3825 @hook TARGET_SUPPORTS_SPLIT_STACK
3827 @node Miscellaneous Register Hooks
3828 @subsection Miscellaneous register hooks
3829 @cindex miscellaneous register hooks
3831 @hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
3834 @section Implementing the Varargs Macros
3835 @cindex varargs implementation
3837 GCC comes with an implementation of @code{<varargs.h>} and
3838 @code{<stdarg.h>} that work without change on machines that pass arguments
3839 on the stack. Other machines require their own implementations of
3840 varargs, and the two machine independent header files must have
3841 conditionals to include it.
3843 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
3844 the calling convention for @code{va_start}. The traditional
3845 implementation takes just one argument, which is the variable in which
3846 to store the argument pointer. The ISO implementation of
3847 @code{va_start} takes an additional second argument. The user is
3848 supposed to write the last named argument of the function here.
3850 However, @code{va_start} should not use this argument. The way to find
3851 the end of the named arguments is with the built-in functions described
3854 @defmac __builtin_saveregs ()
3855 Use this built-in function to save the argument registers in memory so
3856 that the varargs mechanism can access them. Both ISO and traditional
3857 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3858 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
3860 On some machines, @code{__builtin_saveregs} is open-coded under the
3861 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
3862 other machines, it calls a routine written in assembler language,
3863 found in @file{libgcc2.c}.
3865 Code generated for the call to @code{__builtin_saveregs} appears at the
3866 beginning of the function, as opposed to where the call to
3867 @code{__builtin_saveregs} is written, regardless of what the code is.
3868 This is because the registers must be saved before the function starts
3869 to use them for its own purposes.
3870 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3874 @defmac __builtin_next_arg (@var{lastarg})
3875 This builtin returns the address of the first anonymous stack
3876 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3877 returns the address of the location above the first anonymous stack
3878 argument. Use it in @code{va_start} to initialize the pointer for
3879 fetching arguments from the stack. Also use it in @code{va_start} to
3880 verify that the second parameter @var{lastarg} is the last named argument
3881 of the current function.
3884 @defmac __builtin_classify_type (@var{object})
3885 Since each machine has its own conventions for which data types are
3886 passed in which kind of register, your implementation of @code{va_arg}
3887 has to embody these conventions. The easiest way to categorize the
3888 specified data type is to use @code{__builtin_classify_type} together
3889 with @code{sizeof} and @code{__alignof__}.
3891 @code{__builtin_classify_type} ignores the value of @var{object},
3892 considering only its data type. It returns an integer describing what
3893 kind of type that is---integer, floating, pointer, structure, and so on.
3895 The file @file{typeclass.h} defines an enumeration that you can use to
3896 interpret the values of @code{__builtin_classify_type}.
3899 These machine description macros help implement varargs:
3901 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
3903 @hook TARGET_SETUP_INCOMING_VARARGS
3905 @hook TARGET_STRICT_ARGUMENT_NAMING
3907 @hook TARGET_CALL_ARGS
3909 @hook TARGET_END_CALL_ARGS
3911 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
3913 @hook TARGET_LOAD_BOUNDS_FOR_ARG
3915 @hook TARGET_STORE_BOUNDS_FOR_ARG
3917 @hook TARGET_LOAD_RETURNED_BOUNDS
3919 @hook TARGET_STORE_RETURNED_BOUNDS
3921 @hook TARGET_CHKP_FUNCTION_VALUE_BOUNDS
3923 @hook TARGET_SETUP_INCOMING_VARARG_BOUNDS
3926 @section Trampolines for Nested Functions
3927 @cindex trampolines for nested functions
3928 @cindex nested functions, trampolines for
3930 A @dfn{trampoline} is a small piece of code that is created at run time
3931 when the address of a nested function is taken. It normally resides on
3932 the stack, in the stack frame of the containing function. These macros
3933 tell GCC how to generate code to allocate and initialize a
3936 The instructions in the trampoline must do two things: load a constant
3937 address into the static chain register, and jump to the real address of
3938 the nested function. On CISC machines such as the m68k, this requires
3939 two instructions, a move immediate and a jump. Then the two addresses
3940 exist in the trampoline as word-long immediate operands. On RISC
3941 machines, it is often necessary to load each address into a register in
3942 two parts. Then pieces of each address form separate immediate
3945 The code generated to initialize the trampoline must store the variable
3946 parts---the static chain value and the function address---into the
3947 immediate operands of the instructions. On a CISC machine, this is
3948 simply a matter of copying each address to a memory reference at the
3949 proper offset from the start of the trampoline. On a RISC machine, it
3950 may be necessary to take out pieces of the address and store them
3953 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
3955 @defmac TRAMPOLINE_SECTION
3956 Return the section into which the trampoline template is to be placed
3957 (@pxref{Sections}). The default value is @code{readonly_data_section}.
3960 @defmac TRAMPOLINE_SIZE
3961 A C expression for the size in bytes of the trampoline, as an integer.
3964 @defmac TRAMPOLINE_ALIGNMENT
3965 Alignment required for trampolines, in bits.
3967 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
3968 is used for aligning trampolines.
3971 @hook TARGET_TRAMPOLINE_INIT
3973 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
3975 @hook TARGET_CUSTOM_FUNCTION_DESCRIPTORS
3977 Implementing trampolines is difficult on many machines because they have
3978 separate instruction and data caches. Writing into a stack location
3979 fails to clear the memory in the instruction cache, so when the program
3980 jumps to that location, it executes the old contents.
3982 Here are two possible solutions. One is to clear the relevant parts of
3983 the instruction cache whenever a trampoline is set up. The other is to
3984 make all trampolines identical, by having them jump to a standard
3985 subroutine. The former technique makes trampoline execution faster; the
3986 latter makes initialization faster.
3988 To clear the instruction cache when a trampoline is initialized, define
3989 the following macro.
3991 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
3992 If defined, expands to a C expression clearing the @emph{instruction
3993 cache} in the specified interval. The definition of this macro would
3994 typically be a series of @code{asm} statements. Both @var{beg} and
3995 @var{end} are both pointer expressions.
3998 To use a standard subroutine, define the following macro. In addition,
3999 you must make sure that the instructions in a trampoline fill an entire
4000 cache line with identical instructions, or else ensure that the
4001 beginning of the trampoline code is always aligned at the same point in
4002 its cache line. Look in @file{m68k.h} as a guide.
4004 @defmac TRANSFER_FROM_TRAMPOLINE
4005 Define this macro if trampolines need a special subroutine to do their
4006 work. The macro should expand to a series of @code{asm} statements
4007 which will be compiled with GCC@. They go in a library function named
4008 @code{__transfer_from_trampoline}.
4010 If you need to avoid executing the ordinary prologue code of a compiled
4011 C function when you jump to the subroutine, you can do so by placing a
4012 special label of your own in the assembler code. Use one @code{asm}
4013 statement to generate an assembler label, and another to make the label
4014 global. Then trampolines can use that label to jump directly to your
4015 special assembler code.
4019 @section Implicit Calls to Library Routines
4020 @cindex library subroutine names
4021 @cindex @file{libgcc.a}
4023 @c prevent bad page break with this line
4024 Here is an explanation of implicit calls to library routines.
4026 @defmac DECLARE_LIBRARY_RENAMES
4027 This macro, if defined, should expand to a piece of C code that will get
4028 expanded when compiling functions for libgcc.a. It can be used to
4029 provide alternate names for GCC's internal library functions if there
4030 are ABI-mandated names that the compiler should provide.
4033 @findex set_optab_libfunc
4034 @findex init_one_libfunc
4035 @hook TARGET_INIT_LIBFUNCS
4037 @hook TARGET_LIBFUNC_GNU_PREFIX
4039 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4040 This macro should return @code{true} if the library routine that
4041 implements the floating point comparison operator @var{comparison} in
4042 mode @var{mode} will return a boolean, and @var{false} if it will
4045 GCC's own floating point libraries return tristates from the
4046 comparison operators, so the default returns false always. Most ports
4047 don't need to define this macro.
4050 @defmac TARGET_LIB_INT_CMP_BIASED
4051 This macro should evaluate to @code{true} if the integer comparison
4052 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4053 operand is smaller than the second, 1 to indicate that they are equal,
4054 and 2 to indicate that the first operand is greater than the second.
4055 If this macro evaluates to @code{false} the comparison functions return
4056 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4057 in @file{libgcc.a}, you do not need to define this macro.
4060 @defmac TARGET_HAS_NO_HW_DIVIDE
4061 This macro should be defined if the target has no hardware divide
4062 instructions. If this macro is defined, GCC will use an algorithm which
4063 make use of simple logical and arithmetic operations for 64-bit
4064 division. If the macro is not defined, GCC will use an algorithm which
4065 make use of a 64-bit by 32-bit divide primitive.
4068 @cindex @code{EDOM}, implicit usage
4071 The value of @code{EDOM} on the target machine, as a C integer constant
4072 expression. If you don't define this macro, GCC does not attempt to
4073 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4074 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4077 If you do not define @code{TARGET_EDOM}, then compiled code reports
4078 domain errors by calling the library function and letting it report the
4079 error. If mathematical functions on your system use @code{matherr} when
4080 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4081 that @code{matherr} is used normally.
4084 @cindex @code{errno}, implicit usage
4085 @defmac GEN_ERRNO_RTX
4086 Define this macro as a C expression to create an rtl expression that
4087 refers to the global ``variable'' @code{errno}. (On certain systems,
4088 @code{errno} may not actually be a variable.) If you don't define this
4089 macro, a reasonable default is used.
4092 @hook TARGET_LIBC_HAS_FUNCTION
4094 @defmac NEXT_OBJC_RUNTIME
4095 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
4096 by default. This calling convention involves passing the object, the selector
4097 and the method arguments all at once to the method-lookup library function.
4098 This is the usual setting when targeting Darwin/Mac OS X systems, which have
4099 the NeXT runtime installed.
4101 If the macro is set to 0, the "GNU" Objective-C message sending convention
4102 will be used by default. This convention passes just the object and the
4103 selector to the method-lookup function, which returns a pointer to the method.
4105 In either case, it remains possible to select code-generation for the alternate
4106 scheme, by means of compiler command line switches.
4109 @node Addressing Modes
4110 @section Addressing Modes
4111 @cindex addressing modes
4113 @c prevent bad page break with this line
4114 This is about addressing modes.
4116 @defmac HAVE_PRE_INCREMENT
4117 @defmacx HAVE_PRE_DECREMENT
4118 @defmacx HAVE_POST_INCREMENT
4119 @defmacx HAVE_POST_DECREMENT
4120 A C expression that is nonzero if the machine supports pre-increment,
4121 pre-decrement, post-increment, or post-decrement addressing respectively.
4124 @defmac HAVE_PRE_MODIFY_DISP
4125 @defmacx HAVE_POST_MODIFY_DISP
4126 A C expression that is nonzero if the machine supports pre- or
4127 post-address side-effect generation involving constants other than
4128 the size of the memory operand.
4131 @defmac HAVE_PRE_MODIFY_REG
4132 @defmacx HAVE_POST_MODIFY_REG
4133 A C expression that is nonzero if the machine supports pre- or
4134 post-address side-effect generation involving a register displacement.
4137 @defmac CONSTANT_ADDRESS_P (@var{x})
4138 A C expression that is 1 if the RTX @var{x} is a constant which
4139 is a valid address. On most machines the default definition of
4140 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4141 is acceptable, but a few machines are more restrictive as to which
4142 constant addresses are supported.
4145 @defmac CONSTANT_P (@var{x})
4146 @code{CONSTANT_P}, which is defined by target-independent code,
4147 accepts integer-values expressions whose values are not explicitly
4148 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4149 expressions and @code{const} arithmetic expressions, in addition to
4150 @code{const_int} and @code{const_double} expressions.
4153 @defmac MAX_REGS_PER_ADDRESS
4154 A number, the maximum number of registers that can appear in a valid
4155 memory address. Note that it is up to you to specify a value equal to
4156 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4160 @hook TARGET_LEGITIMATE_ADDRESS_P
4162 @defmac TARGET_MEM_CONSTRAINT
4163 A single character to be used instead of the default @code{'m'}
4164 character for general memory addresses. This defines the constraint
4165 letter which matches the memory addresses accepted by
4166 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
4167 support new address formats in your back end without changing the
4168 semantics of the @code{'m'} constraint. This is necessary in order to
4169 preserve functionality of inline assembly constructs using the
4170 @code{'m'} constraint.
4173 @defmac FIND_BASE_TERM (@var{x})
4174 A C expression to determine the base term of address @var{x},
4175 or to provide a simplified version of @var{x} from which @file{alias.c}
4176 can easily find the base term. This macro is used in only two places:
4177 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
4179 It is always safe for this macro to not be defined. It exists so
4180 that alias analysis can understand machine-dependent addresses.
4182 The typical use of this macro is to handle addresses containing
4183 a label_ref or symbol_ref within an UNSPEC@.
4186 @hook TARGET_LEGITIMIZE_ADDRESS
4188 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4189 A C compound statement that attempts to replace @var{x}, which is an address
4190 that needs reloading, with a valid memory address for an operand of mode
4191 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4192 It is not necessary to define this macro, but it might be useful for
4193 performance reasons.
4195 For example, on the i386, it is sometimes possible to use a single
4196 reload register instead of two by reloading a sum of two pseudo
4197 registers into a register. On the other hand, for number of RISC
4198 processors offsets are limited so that often an intermediate address
4199 needs to be generated in order to address a stack slot. By defining
4200 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4201 generated for adjacent some stack slots can be made identical, and thus
4204 @emph{Note}: This macro should be used with caution. It is necessary
4205 to know something of how reload works in order to effectively use this,
4206 and it is quite easy to produce macros that build in too much knowledge
4207 of reload internals.
4209 @emph{Note}: This macro must be able to reload an address created by a
4210 previous invocation of this macro. If it fails to handle such addresses
4211 then the compiler may generate incorrect code or abort.
4214 The macro definition should use @code{push_reload} to indicate parts that
4215 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4216 suitable to be passed unaltered to @code{push_reload}.
4218 The code generated by this macro must not alter the substructure of
4219 @var{x}. If it transforms @var{x} into a more legitimate form, it
4220 should assign @var{x} (which will always be a C variable) a new value.
4221 This also applies to parts that you change indirectly by calling
4224 @findex strict_memory_address_p
4225 The macro definition may use @code{strict_memory_address_p} to test if
4226 the address has become legitimate.
4229 If you want to change only a part of @var{x}, one standard way of doing
4230 this is to use @code{copy_rtx}. Note, however, that it unshares only a
4231 single level of rtl. Thus, if the part to be changed is not at the
4232 top level, you'll need to replace first the top level.
4233 It is not necessary for this macro to come up with a legitimate
4234 address; but often a machine-dependent strategy can generate better code.
4237 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
4239 @hook TARGET_LEGITIMATE_CONSTANT_P
4241 @hook TARGET_DELEGITIMIZE_ADDRESS
4243 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4245 @hook TARGET_CANNOT_FORCE_CONST_MEM
4247 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4249 @hook TARGET_USE_BLOCKS_FOR_DECL_P
4251 @hook TARGET_BUILTIN_RECIPROCAL
4253 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4255 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4257 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4259 @hook TARGET_VECTORIZE_VEC_PERM_CONST_OK
4261 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
4263 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4265 @hook TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION
4267 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4269 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4271 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
4273 @hook TARGET_VECTORIZE_GET_MASK_MODE
4275 @hook TARGET_VECTORIZE_INIT_COST
4277 @hook TARGET_VECTORIZE_ADD_STMT_COST
4279 @hook TARGET_VECTORIZE_FINISH_COST
4281 @hook TARGET_VECTORIZE_DESTROY_COST_DATA
4283 @hook TARGET_VECTORIZE_BUILTIN_GATHER
4285 @hook TARGET_VECTORIZE_BUILTIN_SCATTER
4287 @hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN
4289 @hook TARGET_SIMD_CLONE_ADJUST
4291 @hook TARGET_SIMD_CLONE_USABLE
4293 @hook TARGET_SIMT_VF
4295 @hook TARGET_GOACC_VALIDATE_DIMS
4297 @hook TARGET_GOACC_DIM_LIMIT
4299 @hook TARGET_GOACC_FORK_JOIN
4301 @hook TARGET_GOACC_REDUCTION
4303 @node Anchored Addresses
4304 @section Anchored Addresses
4305 @cindex anchored addresses
4306 @cindex @option{-fsection-anchors}
4308 GCC usually addresses every static object as a separate entity.
4309 For example, if we have:
4313 int foo (void) @{ return a + b + c; @}
4316 the code for @code{foo} will usually calculate three separate symbolic
4317 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
4318 it would be better to calculate just one symbolic address and access
4319 the three variables relative to it. The equivalent pseudocode would
4325 register int *xr = &x;
4326 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4330 (which isn't valid C). We refer to shared addresses like @code{x} as
4331 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
4333 The hooks below describe the target properties that GCC needs to know
4334 in order to make effective use of section anchors. It won't use
4335 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4336 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4338 @hook TARGET_MIN_ANCHOR_OFFSET
4340 @hook TARGET_MAX_ANCHOR_OFFSET
4342 @hook TARGET_ASM_OUTPUT_ANCHOR
4344 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4346 @node Condition Code
4347 @section Condition Code Status
4348 @cindex condition code status
4350 The macros in this section can be split in two families, according to the
4351 two ways of representing condition codes in GCC.
4353 The first representation is the so called @code{(cc0)} representation
4354 (@pxref{Jump Patterns}), where all instructions can have an implicit
4355 clobber of the condition codes. The second is the condition code
4356 register representation, which provides better schedulability for
4357 architectures that do have a condition code register, but on which
4358 most instructions do not affect it. The latter category includes
4361 The implicit clobbering poses a strong restriction on the placement of
4362 the definition and use of the condition code. In the past the definition
4363 and use were always adjacent. However, recent changes to support trapping
4364 arithmatic may result in the definition and user being in different blocks.
4365 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
4366 the definition may be the source of exception handling edges.
4368 These restrictions can prevent important
4369 optimizations on some machines. For example, on the IBM RS/6000, there
4370 is a delay for taken branches unless the condition code register is set
4371 three instructions earlier than the conditional branch. The instruction
4372 scheduler cannot perform this optimization if it is not permitted to
4373 separate the definition and use of the condition code register.
4375 For this reason, it is possible and suggested to use a register to
4376 represent the condition code for new ports. If there is a specific
4377 condition code register in the machine, use a hard register. If the
4378 condition code or comparison result can be placed in any general register,
4379 or if there are multiple condition registers, use a pseudo register.
4380 Registers used to store the condition code value will usually have a mode
4381 that is in class @code{MODE_CC}.
4383 Alternatively, you can use @code{BImode} if the comparison operator is
4384 specified already in the compare instruction. In this case, you are not
4385 interested in most macros in this section.
4388 * CC0 Condition Codes:: Old style representation of condition codes.
4389 * MODE_CC Condition Codes:: Modern representation of condition codes.
4392 @node CC0 Condition Codes
4393 @subsection Representation of condition codes using @code{(cc0)}
4397 The file @file{conditions.h} defines a variable @code{cc_status} to
4398 describe how the condition code was computed (in case the interpretation of
4399 the condition code depends on the instruction that it was set by). This
4400 variable contains the RTL expressions on which the condition code is
4401 currently based, and several standard flags.
4403 Sometimes additional machine-specific flags must be defined in the machine
4404 description header file. It can also add additional machine-specific
4405 information by defining @code{CC_STATUS_MDEP}.
4407 @defmac CC_STATUS_MDEP
4408 C code for a data type which is used for declaring the @code{mdep}
4409 component of @code{cc_status}. It defaults to @code{int}.
4411 This macro is not used on machines that do not use @code{cc0}.
4414 @defmac CC_STATUS_MDEP_INIT
4415 A C expression to initialize the @code{mdep} field to ``empty''.
4416 The default definition does nothing, since most machines don't use
4417 the field anyway. If you want to use the field, you should probably
4418 define this macro to initialize it.
4420 This macro is not used on machines that do not use @code{cc0}.
4423 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4424 A C compound statement to set the components of @code{cc_status}
4425 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4426 this macro's responsibility to recognize insns that set the condition
4427 code as a byproduct of other activity as well as those that explicitly
4430 This macro is not used on machines that do not use @code{cc0}.
4432 If there are insns that do not set the condition code but do alter
4433 other machine registers, this macro must check to see whether they
4434 invalidate the expressions that the condition code is recorded as
4435 reflecting. For example, on the 68000, insns that store in address
4436 registers do not set the condition code, which means that usually
4437 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4438 insns. But suppose that the previous insn set the condition code
4439 based on location @samp{a4@@(102)} and the current insn stores a new
4440 value in @samp{a4}. Although the condition code is not changed by
4441 this, it will no longer be true that it reflects the contents of
4442 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4443 @code{cc_status} in this case to say that nothing is known about the
4444 condition code value.
4446 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4447 with the results of peephole optimization: insns whose patterns are
4448 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4449 constants which are just the operands. The RTL structure of these
4450 insns is not sufficient to indicate what the insns actually do. What
4451 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4452 @code{CC_STATUS_INIT}.
4454 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4455 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4456 @samp{cc}. This avoids having detailed information about patterns in
4457 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4460 @node MODE_CC Condition Codes
4461 @subsection Representation of condition codes using registers
4465 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4466 On many machines, the condition code may be produced by other instructions
4467 than compares, for example the branch can use directly the condition
4468 code set by a subtract instruction. However, on some machines
4469 when the condition code is set this way some bits (such as the overflow
4470 bit) are not set in the same way as a test instruction, so that a different
4471 branch instruction must be used for some conditional branches. When
4472 this happens, use the machine mode of the condition code register to
4473 record different formats of the condition code register. Modes can
4474 also be used to record which compare instruction (e.g. a signed or an
4475 unsigned comparison) produced the condition codes.
4477 If other modes than @code{CCmode} are required, add them to
4478 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4479 a mode given an operand of a compare. This is needed because the modes
4480 have to be chosen not only during RTL generation but also, for example,
4481 by instruction combination. The result of @code{SELECT_CC_MODE} should
4482 be consistent with the mode used in the patterns; for example to support
4483 the case of the add on the SPARC discussed above, we have the pattern
4489 (plus:SI (match_operand:SI 0 "register_operand" "%r")
4490 (match_operand:SI 1 "arith_operand" "rI"))
4497 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
4498 for comparisons whose argument is a @code{plus}:
4501 #define SELECT_CC_MODE(OP,X,Y) \
4502 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4503 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
4504 ? CCFPEmode : CCFPmode) \
4505 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4506 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
4507 ? CCNZmode : CCmode))
4510 Another reason to use modes is to retain information on which operands
4511 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4514 You should define this macro if and only if you define extra CC modes
4515 in @file{@var{machine}-modes.def}.
4518 @hook TARGET_CANONICALIZE_COMPARISON
4520 @defmac REVERSIBLE_CC_MODE (@var{mode})
4521 A C expression whose value is one if it is always safe to reverse a
4522 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4523 can ever return @var{mode} for a floating-point inequality comparison,
4524 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4526 You need not define this macro if it would always returns zero or if the
4527 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4528 For example, here is the definition used on the SPARC, where floating-point
4529 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
4532 #define REVERSIBLE_CC_MODE(MODE) \
4533 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
4537 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
4538 A C expression whose value is reversed condition code of the @var{code} for
4539 comparison done in CC_MODE @var{mode}. The macro is used only in case
4540 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4541 machine has some non-standard way how to reverse certain conditionals. For
4542 instance in case all floating point conditions are non-trapping, compiler may
4543 freely convert unordered compares to ordered ones. Then definition may look
4547 #define REVERSE_CONDITION(CODE, MODE) \
4548 ((MODE) != CCFPmode ? reverse_condition (CODE) \
4549 : reverse_condition_maybe_unordered (CODE))
4553 @hook TARGET_FIXED_CONDITION_CODE_REGS
4555 @hook TARGET_CC_MODES_COMPATIBLE
4557 @hook TARGET_FLAGS_REGNUM
4560 @section Describing Relative Costs of Operations
4561 @cindex costs of instructions
4562 @cindex relative costs
4563 @cindex speed of instructions
4565 These macros let you describe the relative speed of various operations
4566 on the target machine.
4568 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4569 A C expression for the cost of moving data of mode @var{mode} from a
4570 register in class @var{from} to one in class @var{to}. The classes are
4571 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4572 value of 2 is the default; other values are interpreted relative to
4575 It is not required that the cost always equal 2 when @var{from} is the
4576 same as @var{to}; on some machines it is expensive to move between
4577 registers if they are not general registers.
4579 If reload sees an insn consisting of a single @code{set} between two
4580 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4581 classes returns a value of 2, reload does not check to ensure that the
4582 constraints of the insn are met. Setting a cost of other than 2 will
4583 allow reload to verify that the constraints are met. You should do this
4584 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4586 These macros are obsolete, new ports should use the target hook
4587 @code{TARGET_REGISTER_MOVE_COST} instead.
4590 @hook TARGET_REGISTER_MOVE_COST
4592 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4593 A C expression for the cost of moving data of mode @var{mode} between a
4594 register of class @var{class} and memory; @var{in} is zero if the value
4595 is to be written to memory, nonzero if it is to be read in. This cost
4596 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4597 registers and memory is more expensive than between two registers, you
4598 should define this macro to express the relative cost.
4600 If you do not define this macro, GCC uses a default cost of 4 plus
4601 the cost of copying via a secondary reload register, if one is
4602 needed. If your machine requires a secondary reload register to copy
4603 between memory and a register of @var{class} but the reload mechanism is
4604 more complex than copying via an intermediate, define this macro to
4605 reflect the actual cost of the move.
4607 GCC defines the function @code{memory_move_secondary_cost} if
4608 secondary reloads are needed. It computes the costs due to copying via
4609 a secondary register. If your machine copies from memory using a
4610 secondary register in the conventional way but the default base value of
4611 4 is not correct for your machine, define this macro to add some other
4612 value to the result of that function. The arguments to that function
4613 are the same as to this macro.
4615 These macros are obsolete, new ports should use the target hook
4616 @code{TARGET_MEMORY_MOVE_COST} instead.
4619 @hook TARGET_MEMORY_MOVE_COST
4621 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4622 A C expression for the cost of a branch instruction. A value of 1 is
4623 the default; other values are interpreted relative to that. Parameter
4624 @var{speed_p} is true when the branch in question should be optimized
4625 for speed. When it is false, @code{BRANCH_COST} should return a value
4626 optimal for code size rather than performance. @var{predictable_p} is
4627 true for well-predicted branches. On many architectures the
4628 @code{BRANCH_COST} can be reduced then.
4631 Here are additional macros which do not specify precise relative costs,
4632 but only that certain actions are more expensive than GCC would
4635 @defmac SLOW_BYTE_ACCESS
4636 Define this macro as a C expression which is nonzero if accessing less
4637 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4638 faster than accessing a word of memory, i.e., if such access
4639 require more than one instruction or if there is no difference in cost
4640 between byte and (aligned) word loads.
4642 When this macro is not defined, the compiler will access a field by
4643 finding the smallest containing object; when it is defined, a fullword
4644 load will be used if alignment permits. Unless bytes accesses are
4645 faster than word accesses, using word accesses is preferable since it
4646 may eliminate subsequent memory access if subsequent accesses occur to
4647 other fields in the same word of the structure, but to different bytes.
4650 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
4651 Define this macro to be the value 1 if memory accesses described by the
4652 @var{mode} and @var{alignment} parameters have a cost many times greater
4653 than aligned accesses, for example if they are emulated in a trap
4654 handler. This macro is invoked only for unaligned accesses, i.e. when
4655 @code{@var{alignment} < GET_MODE_ALIGNMENT (@var{mode})}.
4657 When this macro is nonzero, the compiler will act as if
4658 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
4659 moves. This can cause significantly more instructions to be produced.
4660 Therefore, do not set this macro nonzero if unaligned accesses only add a
4661 cycle or two to the time for a memory access.
4663 If the value of this macro is always zero, it need not be defined. If
4664 this macro is defined, it should produce a nonzero value when
4665 @code{STRICT_ALIGNMENT} is nonzero.
4668 @defmac MOVE_RATIO (@var{speed})
4669 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4670 which a sequence of insns should be generated instead of a
4671 string move insn or a library call. Increasing the value will always
4672 make code faster, but eventually incurs high cost in increased code size.
4674 Note that on machines where the corresponding move insn is a
4675 @code{define_expand} that emits a sequence of insns, this macro counts
4676 the number of such sequences.
4678 The parameter @var{speed} is true if the code is currently being
4679 optimized for speed rather than size.
4681 If you don't define this, a reasonable default is used.
4684 @hook TARGET_USE_BY_PIECES_INFRASTRUCTURE_P
4686 @hook TARGET_COMPARE_BY_PIECES_BRANCH_RATIO
4688 @defmac MOVE_MAX_PIECES
4689 A C expression used by @code{move_by_pieces} to determine the largest unit
4690 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4693 @defmac STORE_MAX_PIECES
4694 A C expression used by @code{store_by_pieces} to determine the largest unit
4695 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
4696 the size of @code{HOST_WIDE_INT}, whichever is smaller.
4699 @defmac COMPARE_MAX_PIECES
4700 A C expression used by @code{compare_by_pieces} to determine the largest unit
4701 a load or store used to compare memory is. Defaults to
4702 @code{MOVE_MAX_PIECES}.
4705 @defmac CLEAR_RATIO (@var{speed})
4706 The threshold of number of scalar move insns, @emph{below} which a sequence
4707 of insns should be generated to clear memory instead of a string clear insn
4708 or a library call. Increasing the value will always make code faster, but
4709 eventually incurs high cost in increased code size.
4711 The parameter @var{speed} is true if the code is currently being
4712 optimized for speed rather than size.
4714 If you don't define this, a reasonable default is used.
4717 @defmac SET_RATIO (@var{speed})
4718 The threshold of number of scalar move insns, @emph{below} which a sequence
4719 of insns should be generated to set memory to a constant value, instead of
4720 a block set insn or a library call.
4721 Increasing the value will always make code faster, but
4722 eventually incurs high cost in increased code size.
4724 The parameter @var{speed} is true if the code is currently being
4725 optimized for speed rather than size.
4727 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
4730 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
4731 A C expression used to determine whether a load postincrement is a good
4732 thing to use for a given mode. Defaults to the value of
4733 @code{HAVE_POST_INCREMENT}.
4736 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
4737 A C expression used to determine whether a load postdecrement is a good
4738 thing to use for a given mode. Defaults to the value of
4739 @code{HAVE_POST_DECREMENT}.
4742 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4743 A C expression used to determine whether a load preincrement is a good
4744 thing to use for a given mode. Defaults to the value of
4745 @code{HAVE_PRE_INCREMENT}.
4748 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4749 A C expression used to determine whether a load predecrement is a good
4750 thing to use for a given mode. Defaults to the value of
4751 @code{HAVE_PRE_DECREMENT}.
4754 @defmac USE_STORE_POST_INCREMENT (@var{mode})
4755 A C expression used to determine whether a store postincrement is a good
4756 thing to use for a given mode. Defaults to the value of
4757 @code{HAVE_POST_INCREMENT}.
4760 @defmac USE_STORE_POST_DECREMENT (@var{mode})
4761 A C expression used to determine whether a store postdecrement is a good
4762 thing to use for a given mode. Defaults to the value of
4763 @code{HAVE_POST_DECREMENT}.
4766 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
4767 This macro is used to determine whether a store preincrement is a good
4768 thing to use for a given mode. Defaults to the value of
4769 @code{HAVE_PRE_INCREMENT}.
4772 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
4773 This macro is used to determine whether a store predecrement is a good
4774 thing to use for a given mode. Defaults to the value of
4775 @code{HAVE_PRE_DECREMENT}.
4778 @defmac NO_FUNCTION_CSE
4779 Define this macro to be true if it is as good or better to call a constant
4780 function address than to call an address kept in a register.
4783 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4784 Define this macro if a non-short-circuit operation produced by
4785 @samp{fold_range_test ()} is optimal. This macro defaults to true if
4786 @code{BRANCH_COST} is greater than or equal to the value 2.
4789 @hook TARGET_OPTAB_SUPPORTED_P
4791 @hook TARGET_RTX_COSTS
4793 @hook TARGET_ADDRESS_COST
4795 @hook TARGET_MAX_NOCE_IFCVT_SEQ_COST
4797 @hook TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P
4800 @section Adjusting the Instruction Scheduler
4802 The instruction scheduler may need a fair amount of machine-specific
4803 adjustment in order to produce good code. GCC provides several target
4804 hooks for this purpose. It is usually enough to define just a few of
4805 them: try the first ones in this list first.
4807 @hook TARGET_SCHED_ISSUE_RATE
4809 @hook TARGET_SCHED_VARIABLE_ISSUE
4811 @hook TARGET_SCHED_ADJUST_COST
4813 @hook TARGET_SCHED_ADJUST_PRIORITY
4815 @hook TARGET_SCHED_REORDER
4817 @hook TARGET_SCHED_REORDER2
4819 @hook TARGET_SCHED_MACRO_FUSION_P
4821 @hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4823 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4825 @hook TARGET_SCHED_INIT
4827 @hook TARGET_SCHED_FINISH
4829 @hook TARGET_SCHED_INIT_GLOBAL
4831 @hook TARGET_SCHED_FINISH_GLOBAL
4833 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4835 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4837 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4839 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4841 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4843 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4845 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4847 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4849 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4851 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4853 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4855 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4857 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4859 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4861 @hook TARGET_SCHED_DFA_NEW_CYCLE
4863 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4865 @hook TARGET_SCHED_H_I_D_EXTENDED
4867 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4869 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
4871 @hook TARGET_SCHED_SET_SCHED_CONTEXT
4873 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4875 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
4877 @hook TARGET_SCHED_SPECULATE_INSN
4879 @hook TARGET_SCHED_NEEDS_BLOCK_P
4881 @hook TARGET_SCHED_GEN_SPEC_CHECK
4883 @hook TARGET_SCHED_SET_SCHED_FLAGS
4885 @hook TARGET_SCHED_SMS_RES_MII
4887 @hook TARGET_SCHED_DISPATCH
4889 @hook TARGET_SCHED_DISPATCH_DO
4891 @hook TARGET_SCHED_EXPOSED_PIPELINE
4893 @hook TARGET_SCHED_REASSOCIATION_WIDTH
4895 @hook TARGET_SCHED_FUSION_PRIORITY
4897 @hook TARGET_EXPAND_DIVMOD_LIBFUNC
4900 @section Dividing the Output into Sections (Texts, Data, @dots{})
4901 @c the above section title is WAY too long. maybe cut the part between
4902 @c the (...)? --mew 10feb93
4904 An object file is divided into sections containing different types of
4905 data. In the most common case, there are three sections: the @dfn{text
4906 section}, which holds instructions and read-only data; the @dfn{data
4907 section}, which holds initialized writable data; and the @dfn{bss
4908 section}, which holds uninitialized data. Some systems have other kinds
4911 @file{varasm.c} provides several well-known sections, such as
4912 @code{text_section}, @code{data_section} and @code{bss_section}.
4913 The normal way of controlling a @code{@var{foo}_section} variable
4914 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
4915 as described below. The macros are only read once, when @file{varasm.c}
4916 initializes itself, so their values must be run-time constants.
4917 They may however depend on command-line flags.
4919 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
4920 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
4921 to be string literals.
4923 Some assemblers require a different string to be written every time a
4924 section is selected. If your assembler falls into this category, you
4925 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
4926 @code{get_unnamed_section} to set up the sections.
4928 You must always create a @code{text_section}, either by defining
4929 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
4930 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
4931 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
4932 create a distinct @code{readonly_data_section}, the default is to
4933 reuse @code{text_section}.
4935 All the other @file{varasm.c} sections are optional, and are null
4936 if the target does not provide them.
4938 @defmac TEXT_SECTION_ASM_OP
4939 A C expression whose value is a string, including spacing, containing the
4940 assembler operation that should precede instructions and read-only data.
4941 Normally @code{"\t.text"} is right.
4944 @defmac HOT_TEXT_SECTION_NAME
4945 If defined, a C string constant for the name of the section containing most
4946 frequently executed functions of the program. If not defined, GCC will provide
4947 a default definition if the target supports named sections.
4950 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
4951 If defined, a C string constant for the name of the section containing unlikely
4952 executed functions in the program.
4955 @defmac DATA_SECTION_ASM_OP
4956 A C expression whose value is a string, including spacing, containing the
4957 assembler operation to identify the following data as writable initialized
4958 data. Normally @code{"\t.data"} is right.
4961 @defmac SDATA_SECTION_ASM_OP
4962 If defined, a C expression whose value is a string, including spacing,
4963 containing the assembler operation to identify the following data as
4964 initialized, writable small data.
4967 @defmac READONLY_DATA_SECTION_ASM_OP
4968 A C expression whose value is a string, including spacing, containing the
4969 assembler operation to identify the following data as read-only initialized
4973 @defmac BSS_SECTION_ASM_OP
4974 If defined, a C expression whose value is a string, including spacing,
4975 containing the assembler operation to identify the following data as
4976 uninitialized global data. If not defined, and
4977 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
4978 uninitialized global data will be output in the data section if
4979 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
4983 @defmac SBSS_SECTION_ASM_OP
4984 If defined, a C expression whose value is a string, including spacing,
4985 containing the assembler operation to identify the following data as
4986 uninitialized, writable small data.
4989 @defmac TLS_COMMON_ASM_OP
4990 If defined, a C expression whose value is a string containing the
4991 assembler operation to identify the following data as thread-local
4992 common data. The default is @code{".tls_common"}.
4995 @defmac TLS_SECTION_ASM_FLAG
4996 If defined, a C expression whose value is a character constant
4997 containing the flag used to mark a section as a TLS section. The
4998 default is @code{'T'}.
5001 @defmac INIT_SECTION_ASM_OP
5002 If defined, a C expression whose value is a string, including spacing,
5003 containing the assembler operation to identify the following data as
5004 initialization code. If not defined, GCC will assume such a section does
5005 not exist. This section has no corresponding @code{init_section}
5006 variable; it is used entirely in runtime code.
5009 @defmac FINI_SECTION_ASM_OP
5010 If defined, a C expression whose value is a string, including spacing,
5011 containing the assembler operation to identify the following data as
5012 finalization code. If not defined, GCC will assume such a section does
5013 not exist. This section has no corresponding @code{fini_section}
5014 variable; it is used entirely in runtime code.
5017 @defmac INIT_ARRAY_SECTION_ASM_OP
5018 If defined, a C expression whose value is a string, including spacing,
5019 containing the assembler operation to identify the following data as
5020 part of the @code{.init_array} (or equivalent) section. If not
5021 defined, GCC will assume such a section does not exist. Do not define
5022 both this macro and @code{INIT_SECTION_ASM_OP}.
5025 @defmac FINI_ARRAY_SECTION_ASM_OP
5026 If defined, a C expression whose value is a string, including spacing,
5027 containing the assembler operation to identify the following data as
5028 part of the @code{.fini_array} (or equivalent) section. If not
5029 defined, GCC will assume such a section does not exist. Do not define
5030 both this macro and @code{FINI_SECTION_ASM_OP}.
5033 @defmac MACH_DEP_SECTION_ASM_FLAG
5034 If defined, a C expression whose value is a character constant
5035 containing the flag used to mark a machine-dependent section. This
5036 corresponds to the @code{SECTION_MACH_DEP} section flag.
5039 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5040 If defined, an ASM statement that switches to a different section
5041 via @var{section_op}, calls @var{function}, and switches back to
5042 the text section. This is used in @file{crtstuff.c} if
5043 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5044 to initialization and finalization functions from the init and fini
5045 sections. By default, this macro uses a simple function call. Some
5046 ports need hand-crafted assembly code to avoid dependencies on
5047 registers initialized in the function prologue or to ensure that
5048 constant pools don't end up too far way in the text section.
5051 @defmac TARGET_LIBGCC_SDATA_SECTION
5052 If defined, a string which names the section into which small
5053 variables defined in crtstuff and libgcc should go. This is useful
5054 when the target has options for optimizing access to small data, and
5055 you want the crtstuff and libgcc routines to be conservative in what
5056 they expect of your application yet liberal in what your application
5057 expects. For example, for targets with a @code{.sdata} section (like
5058 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
5059 require small data support from your application, but use this macro
5060 to put small data into @code{.sdata} so that your application can
5061 access these variables whether it uses small data or not.
5064 @defmac FORCE_CODE_SECTION_ALIGN
5065 If defined, an ASM statement that aligns a code section to some
5066 arbitrary boundary. This is used to force all fragments of the
5067 @code{.init} and @code{.fini} sections to have to same alignment
5068 and thus prevent the linker from having to add any padding.
5071 @defmac JUMP_TABLES_IN_TEXT_SECTION
5072 Define this macro to be an expression with a nonzero value if jump
5073 tables (for @code{tablejump} insns) should be output in the text
5074 section, along with the assembler instructions. Otherwise, the
5075 readonly data section is used.
5077 This macro is irrelevant if there is no separate readonly data section.
5080 @hook TARGET_ASM_INIT_SECTIONS
5082 @hook TARGET_ASM_RELOC_RW_MASK
5084 @hook TARGET_ASM_SELECT_SECTION
5086 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
5087 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
5088 for @code{FUNCTION_DECL}s as well as for variables and constants.
5090 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
5091 function has been determined to be likely to be called, and nonzero if
5092 it is unlikely to be called.
5095 @hook TARGET_ASM_UNIQUE_SECTION
5097 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
5099 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
5101 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
5103 @hook TARGET_ASM_SELECT_RTX_SECTION
5105 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
5107 @hook TARGET_ENCODE_SECTION_INFO
5109 @hook TARGET_STRIP_NAME_ENCODING
5111 @hook TARGET_IN_SMALL_DATA_P
5113 @hook TARGET_HAVE_SRODATA_SECTION
5115 @hook TARGET_PROFILE_BEFORE_PROLOGUE
5117 @hook TARGET_BINDS_LOCAL_P
5119 @hook TARGET_HAVE_TLS
5123 @section Position Independent Code
5124 @cindex position independent code
5127 This section describes macros that help implement generation of position
5128 independent code. Simply defining these macros is not enough to
5129 generate valid PIC; you must also add support to the hook
5130 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
5131 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
5132 must modify the definition of @samp{movsi} to do something appropriate
5133 when the source operand contains a symbolic address. You may also
5134 need to alter the handling of switch statements so that they use
5136 @c i rearranged the order of the macros above to try to force one of
5137 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5139 @defmac PIC_OFFSET_TABLE_REGNUM
5140 The register number of the register used to address a table of static
5141 data addresses in memory. In some cases this register is defined by a
5142 processor's ``application binary interface'' (ABI)@. When this macro
5143 is defined, RTL is generated for this register once, as with the stack
5144 pointer and frame pointer registers. If this macro is not defined, it
5145 is up to the machine-dependent files to allocate such a register (if
5146 necessary). Note that this register must be fixed when in use (e.g.@:
5147 when @code{flag_pic} is true).
5150 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5151 A C expression that is nonzero if the register defined by
5152 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
5153 the default is zero. Do not define
5154 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5157 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
5158 A C expression that is nonzero if @var{x} is a legitimate immediate
5159 operand on the target machine when generating position independent code.
5160 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5161 check this. You can also assume @var{flag_pic} is true, so you need not
5162 check it either. You need not define this macro if all constants
5163 (including @code{SYMBOL_REF}) can be immediate operands when generating
5164 position independent code.
5167 @node Assembler Format
5168 @section Defining the Output Assembler Language
5170 This section describes macros whose principal purpose is to describe how
5171 to write instructions in assembler language---rather than what the
5175 * File Framework:: Structural information for the assembler file.
5176 * Data Output:: Output of constants (numbers, strings, addresses).
5177 * Uninitialized Data:: Output of uninitialized variables.
5178 * Label Output:: Output and generation of labels.
5179 * Initialization:: General principles of initialization
5180 and termination routines.
5181 * Macros for Initialization::
5182 Specific macros that control the handling of
5183 initialization and termination routines.
5184 * Instruction Output:: Output of actual instructions.
5185 * Dispatch Tables:: Output of jump tables.
5186 * Exception Region Output:: Output of exception region code.
5187 * Alignment Output:: Pseudo ops for alignment and skipping data.
5190 @node File Framework
5191 @subsection The Overall Framework of an Assembler File
5192 @cindex assembler format
5193 @cindex output of assembler code
5195 @c prevent bad page break with this line
5196 This describes the overall framework of an assembly file.
5198 @findex default_file_start
5199 @hook TARGET_ASM_FILE_START
5201 @hook TARGET_ASM_FILE_START_APP_OFF
5203 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5205 @hook TARGET_ASM_FILE_END
5207 @deftypefun void file_end_indicate_exec_stack ()
5208 Some systems use a common convention, the @samp{.note.GNU-stack}
5209 special section, to indicate whether or not an object file relies on
5210 the stack being executable. If your system uses this convention, you
5211 should define @code{TARGET_ASM_FILE_END} to this function. If you
5212 need to do other things in that hook, have your hook function call
5216 @hook TARGET_ASM_LTO_START
5218 @hook TARGET_ASM_LTO_END
5220 @hook TARGET_ASM_CODE_END
5222 @defmac ASM_COMMENT_START
5223 A C string constant describing how to begin a comment in the target
5224 assembler language. The compiler assumes that the comment will end at
5225 the end of the line.
5229 A C string constant for text to be output before each @code{asm}
5230 statement or group of consecutive ones. Normally this is
5231 @code{"#APP"}, which is a comment that has no effect on most
5232 assemblers but tells the GNU assembler that it must check the lines
5233 that follow for all valid assembler constructs.
5237 A C string constant for text to be output after each @code{asm}
5238 statement or group of consecutive ones. Normally this is
5239 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5240 time-saving assumptions that are valid for ordinary compiler output.
5243 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5244 A C statement to output COFF information or DWARF debugging information
5245 which indicates that filename @var{name} is the current source file to
5246 the stdio stream @var{stream}.
5248 This macro need not be defined if the standard form of output
5249 for the file format in use is appropriate.
5252 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5254 @hook TARGET_ASM_OUTPUT_IDENT
5256 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5257 A C statement to output the string @var{string} to the stdio stream
5258 @var{stream}. If you do not call the function @code{output_quoted_string}
5259 in your config files, GCC will only call it to output filenames to
5260 the assembler source. So you can use it to canonicalize the format
5261 of the filename using this macro.
5264 @hook TARGET_ASM_NAMED_SECTION
5266 @hook TARGET_ASM_ELF_FLAGS_NUMERIC
5268 @hook TARGET_ASM_FUNCTION_SECTION
5270 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5272 @hook TARGET_HAVE_NAMED_SECTIONS
5273 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5274 It must not be modified by command-line option processing.
5277 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5278 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5280 @hook TARGET_SECTION_TYPE_FLAGS
5282 @hook TARGET_ASM_RECORD_GCC_SWITCHES
5284 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5288 @subsection Output of Data
5291 @hook TARGET_ASM_BYTE_OP
5293 @hook TARGET_ASM_INTEGER
5295 @hook TARGET_ASM_DECL_END
5297 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5299 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5300 A C statement to output to the stdio stream @var{stream} an assembler
5301 instruction to assemble a string constant containing the @var{len}
5302 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5303 @code{char *} and @var{len} a C expression of type @code{int}.
5305 If the assembler has a @code{.ascii} pseudo-op as found in the
5306 Berkeley Unix assembler, do not define the macro
5307 @code{ASM_OUTPUT_ASCII}.
5310 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5311 A C statement to output word @var{n} of a function descriptor for
5312 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5313 is defined, and is otherwise unused.
5316 @defmac CONSTANT_POOL_BEFORE_FUNCTION
5317 You may define this macro as a C expression. You should define the
5318 expression to have a nonzero value if GCC should output the constant
5319 pool for a function before the code for the function, or a zero value if
5320 GCC should output the constant pool after the function. If you do
5321 not define this macro, the usual case, GCC will output the constant
5322 pool before the function.
5325 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5326 A C statement to output assembler commands to define the start of the
5327 constant pool for a function. @var{funname} is a string giving
5328 the name of the function. Should the return type of the function
5329 be required, it can be obtained via @var{fundecl}. @var{size}
5330 is the size, in bytes, of the constant pool that will be written
5331 immediately after this call.
5333 If no constant-pool prefix is required, the usual case, this macro need
5337 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5338 A C statement (with or without semicolon) to output a constant in the
5339 constant pool, if it needs special treatment. (This macro need not do
5340 anything for RTL expressions that can be output normally.)
5342 The argument @var{file} is the standard I/O stream to output the
5343 assembler code on. @var{x} is the RTL expression for the constant to
5344 output, and @var{mode} is the machine mode (in case @var{x} is a
5345 @samp{const_int}). @var{align} is the required alignment for the value
5346 @var{x}; you should output an assembler directive to force this much
5349 The argument @var{labelno} is a number to use in an internal label for
5350 the address of this pool entry. The definition of this macro is
5351 responsible for outputting the label definition at the proper place.
5352 Here is how to do this:
5355 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5358 When you output a pool entry specially, you should end with a
5359 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5360 entry from being output a second time in the usual manner.
5362 You need not define this macro if it would do nothing.
5365 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5366 A C statement to output assembler commands to at the end of the constant
5367 pool for a function. @var{funname} is a string giving the name of the
5368 function. Should the return type of the function be required, you can
5369 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5370 constant pool that GCC wrote immediately before this call.
5372 If no constant-pool epilogue is required, the usual case, you need not
5376 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5377 Define this macro as a C expression which is nonzero if @var{C} is
5378 used as a logical line separator by the assembler. @var{STR} points
5379 to the position in the string where @var{C} was found; this can be used if
5380 a line separator uses multiple characters.
5382 If you do not define this macro, the default is that only
5383 the character @samp{;} is treated as a logical line separator.
5386 @hook TARGET_ASM_OPEN_PAREN
5388 These macros are provided by @file{real.h} for writing the definitions
5389 of @code{ASM_OUTPUT_DOUBLE} and the like:
5391 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5392 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5393 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5394 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5395 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5396 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5397 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5398 target's floating point representation, and store its bit pattern in
5399 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5400 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5401 simple @code{long int}. For the others, it should be an array of
5402 @code{long int}. The number of elements in this array is determined
5403 by the size of the desired target floating point data type: 32 bits of
5404 it go in each @code{long int} array element. Each array element holds
5405 32 bits of the result, even if @code{long int} is wider than 32 bits
5406 on the host machine.
5408 The array element values are designed so that you can print them out
5409 using @code{fprintf} in the order they should appear in the target
5413 @node Uninitialized Data
5414 @subsection Output of Uninitialized Variables
5416 Each of the macros in this section is used to do the whole job of
5417 outputting a single uninitialized variable.
5419 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5420 A C statement (sans semicolon) to output to the stdio stream
5421 @var{stream} the assembler definition of a common-label named
5422 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5423 is the size rounded up to whatever alignment the caller wants. It is
5424 possible that @var{size} may be zero, for instance if a struct with no
5425 other member than a zero-length array is defined. In this case, the
5426 backend must output a symbol definition that allocates at least one
5427 byte, both so that the address of the resulting object does not compare
5428 equal to any other, and because some object formats cannot even express
5429 the concept of a zero-sized common symbol, as that is how they represent
5430 an ordinary undefined external.
5432 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5433 output the name itself; before and after that, output the additional
5434 assembler syntax for defining the name, and a newline.
5436 This macro controls how the assembler definitions of uninitialized
5437 common global variables are output.
5440 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5441 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5442 separate, explicit argument. If you define this macro, it is used in
5443 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5444 handling the required alignment of the variable. The alignment is specified
5445 as the number of bits.
5448 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5449 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5450 variable to be output, if there is one, or @code{NULL_TREE} if there
5451 is no corresponding variable. If you define this macro, GCC will use it
5452 in place of both @code{ASM_OUTPUT_COMMON} and
5453 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5454 the variable's decl in order to chose what to output.
5457 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5458 A C statement (sans semicolon) to output to the stdio stream
5459 @var{stream} the assembler definition of uninitialized global @var{decl} named
5460 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
5461 is the alignment specified as the number of bits.
5463 Try to use function @code{asm_output_aligned_bss} defined in file
5464 @file{varasm.c} when defining this macro. If unable, use the expression
5465 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5466 before and after that, output the additional assembler syntax for defining
5467 the name, and a newline.
5469 There are two ways of handling global BSS@. One is to define this macro.
5470 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5471 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5472 You do not need to do both.
5474 Some languages do not have @code{common} data, and require a
5475 non-common form of global BSS in order to handle uninitialized globals
5476 efficiently. C++ is one example of this. However, if the target does
5477 not support global BSS, the front end may choose to make globals
5478 common in order to save space in the object file.
5481 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5482 A C statement (sans semicolon) to output to the stdio stream
5483 @var{stream} the assembler definition of a local-common-label named
5484 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5485 is the size rounded up to whatever alignment the caller wants.
5487 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5488 output the name itself; before and after that, output the additional
5489 assembler syntax for defining the name, and a newline.
5491 This macro controls how the assembler definitions of uninitialized
5492 static variables are output.
5495 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5496 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5497 separate, explicit argument. If you define this macro, it is used in
5498 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5499 handling the required alignment of the variable. The alignment is specified
5500 as the number of bits.
5503 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5504 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5505 variable to be output, if there is one, or @code{NULL_TREE} if there
5506 is no corresponding variable. If you define this macro, GCC will use it
5507 in place of both @code{ASM_OUTPUT_DECL} and
5508 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5509 the variable's decl in order to chose what to output.
5513 @subsection Output and Generation of Labels
5515 @c prevent bad page break with this line
5516 This is about outputting labels.
5518 @findex assemble_name
5519 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5520 A C statement (sans semicolon) to output to the stdio stream
5521 @var{stream} the assembler definition of a label named @var{name}.
5522 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5523 output the name itself; before and after that, output the additional
5524 assembler syntax for defining the name, and a newline. A default
5525 definition of this macro is provided which is correct for most systems.
5528 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5529 A C statement (sans semicolon) to output to the stdio stream
5530 @var{stream} the assembler definition of a label named @var{name} of
5532 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5533 output the name itself; before and after that, output the additional
5534 assembler syntax for defining the name, and a newline. A default
5535 definition of this macro is provided which is correct for most systems.
5537 If this macro is not defined, then the function name is defined in the
5538 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5541 @findex assemble_name_raw
5542 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5543 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5544 to refer to a compiler-generated label. The default definition uses
5545 @code{assemble_name_raw}, which is like @code{assemble_name} except
5546 that it is more efficient.
5550 A C string containing the appropriate assembler directive to specify the
5551 size of a symbol, without any arguments. On systems that use ELF, the
5552 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5553 systems, the default is not to define this macro.
5555 Define this macro only if it is correct to use the default definitions
5556 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5557 for your system. If you need your own custom definitions of those
5558 macros, or if you do not need explicit symbol sizes at all, do not
5562 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5563 A C statement (sans semicolon) to output to the stdio stream
5564 @var{stream} a directive telling the assembler that the size of the
5565 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
5566 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5570 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5571 A C statement (sans semicolon) to output to the stdio stream
5572 @var{stream} a directive telling the assembler to calculate the size of
5573 the symbol @var{name} by subtracting its address from the current
5576 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5577 provided. The default assumes that the assembler recognizes a special
5578 @samp{.} symbol as referring to the current address, and can calculate
5579 the difference between this and another symbol. If your assembler does
5580 not recognize @samp{.} or cannot do calculations with it, you will need
5581 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5584 @defmac NO_DOLLAR_IN_LABEL
5585 Define this macro if the assembler does not accept the character
5586 @samp{$} in label names. By default constructors and destructors in
5587 G++ have @samp{$} in the identifiers. If this macro is defined,
5588 @samp{.} is used instead.
5591 @defmac NO_DOT_IN_LABEL
5592 Define this macro if the assembler does not accept the character
5593 @samp{.} in label names. By default constructors and destructors in G++
5594 have names that use @samp{.}. If this macro is defined, these names
5595 are rewritten to avoid @samp{.}.
5599 A C string containing the appropriate assembler directive to specify the
5600 type of a symbol, without any arguments. On systems that use ELF, the
5601 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5602 systems, the default is not to define this macro.
5604 Define this macro only if it is correct to use the default definition of
5605 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5606 custom definition of this macro, or if you do not need explicit symbol
5607 types at all, do not define this macro.
5610 @defmac TYPE_OPERAND_FMT
5611 A C string which specifies (using @code{printf} syntax) the format of
5612 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
5613 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5614 the default is not to define this macro.
5616 Define this macro only if it is correct to use the default definition of
5617 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5618 custom definition of this macro, or if you do not need explicit symbol
5619 types at all, do not define this macro.
5622 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5623 A C statement (sans semicolon) to output to the stdio stream
5624 @var{stream} a directive telling the assembler that the type of the
5625 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
5626 that string is always either @samp{"function"} or @samp{"object"}, but
5627 you should not count on this.
5629 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5630 definition of this macro is provided.
5633 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5634 A C statement (sans semicolon) to output to the stdio stream
5635 @var{stream} any text necessary for declaring the name @var{name} of a
5636 function which is being defined. This macro is responsible for
5637 outputting the label definition (perhaps using
5638 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5639 @code{FUNCTION_DECL} tree node representing the function.
5641 If this macro is not defined, then the function name is defined in the
5642 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5644 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5648 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5649 A C statement (sans semicolon) to output to the stdio stream
5650 @var{stream} any text necessary for declaring the size of a function
5651 which is being defined. The argument @var{name} is the name of the
5652 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5653 representing the function.
5655 If this macro is not defined, then the function size is not defined.
5657 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5661 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5662 A C statement (sans semicolon) to output to the stdio stream
5663 @var{stream} any text necessary for declaring the name @var{name} of a
5664 cold function partition which is being defined. This macro is responsible
5665 for outputting the label definition (perhaps using
5666 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5667 @code{FUNCTION_DECL} tree node representing the function.
5669 If this macro is not defined, then the cold partition name is defined in the
5670 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5672 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5676 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5677 A C statement (sans semicolon) to output to the stdio stream
5678 @var{stream} any text necessary for declaring the size of a cold function
5679 partition which is being defined. The argument @var{name} is the name of the
5680 cold partition of the function. The argument @var{decl} is the
5681 @code{FUNCTION_DECL} tree node representing the function.
5683 If this macro is not defined, then the partition size is not defined.
5685 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5689 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5690 A C statement (sans semicolon) to output to the stdio stream
5691 @var{stream} any text necessary for declaring the name @var{name} of an
5692 initialized variable which is being defined. This macro must output the
5693 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5694 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5696 If this macro is not defined, then the variable name is defined in the
5697 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5699 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5700 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5703 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
5705 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5706 A C statement (sans semicolon) to output to the stdio stream
5707 @var{stream} any text necessary for claiming a register @var{regno}
5708 for a global variable @var{decl} with name @var{name}.
5710 If you don't define this macro, that is equivalent to defining it to do
5714 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5715 A C statement (sans semicolon) to finish up declaring a variable name
5716 once the compiler has processed its initializer fully and thus has had a
5717 chance to determine the size of an array when controlled by an
5718 initializer. This is used on systems where it's necessary to declare
5719 something about the size of the object.
5721 If you don't define this macro, that is equivalent to defining it to do
5724 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5725 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5728 @hook TARGET_ASM_GLOBALIZE_LABEL
5730 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
5732 @hook TARGET_ASM_ASSEMBLE_UNDEFINED_DECL
5734 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5735 A C statement (sans semicolon) to output to the stdio stream
5736 @var{stream} some commands that will make the label @var{name} weak;
5737 that is, available for reference from other files but only used if
5738 no other definition is available. Use the expression
5739 @code{assemble_name (@var{stream}, @var{name})} to output the name
5740 itself; before and after that, output the additional assembler syntax
5741 for making that name weak, and a newline.
5743 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5744 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5748 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5749 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5750 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5751 or variable decl. If @var{value} is not @code{NULL}, this C statement
5752 should output to the stdio stream @var{stream} assembler code which
5753 defines (equates) the weak symbol @var{name} to have the value
5754 @var{value}. If @var{value} is @code{NULL}, it should output commands
5755 to make @var{name} weak.
5758 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5759 Outputs a directive that enables @var{name} to be used to refer to
5760 symbol @var{value} with weak-symbol semantics. @code{decl} is the
5761 declaration of @code{name}.
5764 @defmac SUPPORTS_WEAK
5765 A preprocessor constant expression which evaluates to true if the target
5766 supports weak symbols.
5768 If you don't define this macro, @file{defaults.h} provides a default
5769 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5770 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5773 @defmac TARGET_SUPPORTS_WEAK
5774 A C expression which evaluates to true if the target supports weak symbols.
5776 If you don't define this macro, @file{defaults.h} provides a default
5777 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
5778 this macro if you want to control weak symbol support with a compiler
5779 flag such as @option{-melf}.
5782 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
5783 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5784 public symbol such that extra copies in multiple translation units will
5785 be discarded by the linker. Define this macro if your object file
5786 format provides support for this concept, such as the @samp{COMDAT}
5787 section flags in the Microsoft Windows PE/COFF format, and this support
5788 requires changes to @var{decl}, such as putting it in a separate section.
5791 @defmac SUPPORTS_ONE_ONLY
5792 A C expression which evaluates to true if the target supports one-only
5795 If you don't define this macro, @file{varasm.c} provides a default
5796 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5797 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5798 you want to control one-only symbol support with a compiler flag, or if
5799 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5800 be emitted as one-only.
5803 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
5805 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5806 A C expression that evaluates to true if the target's linker expects
5807 that weak symbols do not appear in a static archive's table of contents.
5808 The default is @code{0}.
5810 Leaving weak symbols out of an archive's table of contents means that,
5811 if a symbol will only have a definition in one translation unit and
5812 will have undefined references from other translation units, that
5813 symbol should not be weak. Defining this macro to be nonzero will
5814 thus have the effect that certain symbols that would normally be weak
5815 (explicit template instantiations, and vtables for polymorphic classes
5816 with noninline key methods) will instead be nonweak.
5818 The C++ ABI requires this macro to be zero. Define this macro for
5819 targets where full C++ ABI compliance is impossible and where linker
5820 restrictions require weak symbols to be left out of a static archive's
5824 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5825 A C statement (sans semicolon) to output to the stdio stream
5826 @var{stream} any text necessary for declaring the name of an external
5827 symbol named @var{name} which is referenced in this compilation but
5828 not defined. The value of @var{decl} is the tree node for the
5831 This macro need not be defined if it does not need to output anything.
5832 The GNU assembler and most Unix assemblers don't require anything.
5835 @hook TARGET_ASM_EXTERNAL_LIBCALL
5837 @hook TARGET_ASM_MARK_DECL_PRESERVED
5839 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5840 A C statement (sans semicolon) to output to the stdio stream
5841 @var{stream} a reference in assembler syntax to a label named
5842 @var{name}. This should add @samp{_} to the front of the name, if that
5843 is customary on your operating system, as it is in most Berkeley Unix
5844 systems. This macro is used in @code{assemble_name}.
5847 @hook TARGET_MANGLE_ASSEMBLER_NAME
5849 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5850 A C statement (sans semicolon) to output a reference to
5851 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
5852 will be used to output the name of the symbol. This macro may be used
5853 to modify the way a symbol is referenced depending on information
5854 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5857 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5858 A C statement (sans semicolon) to output a reference to @var{buf}, the
5859 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
5860 @code{assemble_name} will be used to output the name of the symbol.
5861 This macro is not used by @code{output_asm_label}, or the @code{%l}
5862 specifier that calls it; the intention is that this macro should be set
5863 when it is necessary to output a label differently when its address is
5867 @hook TARGET_ASM_INTERNAL_LABEL
5869 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5870 A C statement to output to the stdio stream @var{stream} a debug info
5871 label whose name is made from the string @var{prefix} and the number
5872 @var{num}. This is useful for VLIW targets, where debug info labels
5873 may need to be treated differently than branch target labels. On some
5874 systems, branch target labels must be at the beginning of instruction
5875 bundles, but debug info labels can occur in the middle of instruction
5878 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5882 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5883 A C statement to store into the string @var{string} a label whose name
5884 is made from the string @var{prefix} and the number @var{num}.
5886 This string, when output subsequently by @code{assemble_name}, should
5887 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5888 with the same @var{prefix} and @var{num}.
5890 If the string begins with @samp{*}, then @code{assemble_name} will
5891 output the rest of the string unchanged. It is often convenient for
5892 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5893 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5894 to output the string, and may change it. (Of course,
5895 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5896 you should know what it does on your machine.)
5899 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5900 A C expression to assign to @var{outvar} (which is a variable of type
5901 @code{char *}) a newly allocated string made from the string
5902 @var{name} and the number @var{number}, with some suitable punctuation
5903 added. Use @code{alloca} to get space for the string.
5905 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5906 produce an assembler label for an internal static variable whose name is
5907 @var{name}. Therefore, the string must be such as to result in valid
5908 assembler code. The argument @var{number} is different each time this
5909 macro is executed; it prevents conflicts between similarly-named
5910 internal static variables in different scopes.
5912 Ideally this string should not be a valid C identifier, to prevent any
5913 conflict with the user's own symbols. Most assemblers allow periods
5914 or percent signs in assembler symbols; putting at least one of these
5915 between the name and the number will suffice.
5917 If this macro is not defined, a default definition will be provided
5918 which is correct for most systems.
5921 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5922 A C statement to output to the stdio stream @var{stream} assembler code
5923 which defines (equates) the symbol @var{name} to have the value @var{value}.
5926 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5927 correct for most systems.
5930 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5931 A C statement to output to the stdio stream @var{stream} assembler code
5932 which defines (equates) the symbol whose tree node is @var{decl_of_name}
5933 to have the value of the tree node @var{decl_of_value}. This macro will
5934 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5935 the tree nodes are available.
5938 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5939 correct for most systems.
5942 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
5943 A C statement that evaluates to true if the assembler code which defines
5944 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
5945 of the tree node @var{decl_of_value} should be emitted near the end of the
5946 current compilation unit. The default is to not defer output of defines.
5947 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
5948 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
5951 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5952 A C statement to output to the stdio stream @var{stream} assembler code
5953 which defines (equates) the weak symbol @var{name} to have the value
5954 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
5955 an undefined weak symbol.
5957 Define this macro if the target only supports weak aliases; define
5958 @code{ASM_OUTPUT_DEF} instead if possible.
5961 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5962 Define this macro to override the default assembler names used for
5963 Objective-C methods.
5965 The default name is a unique method number followed by the name of the
5966 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5967 the category is also included in the assembler name (e.g.@:
5970 These names are safe on most systems, but make debugging difficult since
5971 the method's selector is not present in the name. Therefore, particular
5972 systems define other ways of computing names.
5974 @var{buf} is an expression of type @code{char *} which gives you a
5975 buffer in which to store the name; its length is as long as
5976 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5977 50 characters extra.
5979 The argument @var{is_inst} specifies whether the method is an instance
5980 method or a class method; @var{class_name} is the name of the class;
5981 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
5982 in a category); and @var{sel_name} is the name of the selector.
5984 On systems where the assembler can handle quoted names, you can use this
5985 macro to provide more human-readable names.
5988 @node Initialization
5989 @subsection How Initialization Functions Are Handled
5990 @cindex initialization routines
5991 @cindex termination routines
5992 @cindex constructors, output of
5993 @cindex destructors, output of
5995 The compiled code for certain languages includes @dfn{constructors}
5996 (also called @dfn{initialization routines})---functions to initialize
5997 data in the program when the program is started. These functions need
5998 to be called before the program is ``started''---that is to say, before
5999 @code{main} is called.
6001 Compiling some languages generates @dfn{destructors} (also called
6002 @dfn{termination routines}) that should be called when the program
6005 To make the initialization and termination functions work, the compiler
6006 must output something in the assembler code to cause those functions to
6007 be called at the appropriate time. When you port the compiler to a new
6008 system, you need to specify how to do this.
6010 There are two major ways that GCC currently supports the execution of
6011 initialization and termination functions. Each way has two variants.
6012 Much of the structure is common to all four variations.
6014 @findex __CTOR_LIST__
6015 @findex __DTOR_LIST__
6016 The linker must build two lists of these functions---a list of
6017 initialization functions, called @code{__CTOR_LIST__}, and a list of
6018 termination functions, called @code{__DTOR_LIST__}.
6020 Each list always begins with an ignored function pointer (which may hold
6021 0, @minus{}1, or a count of the function pointers after it, depending on
6022 the environment). This is followed by a series of zero or more function
6023 pointers to constructors (or destructors), followed by a function
6024 pointer containing zero.
6026 Depending on the operating system and its executable file format, either
6027 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6028 time and exit time. Constructors are called in reverse order of the
6029 list; destructors in forward order.
6031 The best way to handle static constructors works only for object file
6032 formats which provide arbitrarily-named sections. A section is set
6033 aside for a list of constructors, and another for a list of destructors.
6034 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6035 object file that defines an initialization function also puts a word in
6036 the constructor section to point to that function. The linker
6037 accumulates all these words into one contiguous @samp{.ctors} section.
6038 Termination functions are handled similarly.
6040 This method will be chosen as the default by @file{target-def.h} if
6041 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6042 support arbitrary sections, but does support special designated
6043 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6044 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6046 When arbitrary sections are available, there are two variants, depending
6047 upon how the code in @file{crtstuff.c} is called. On systems that
6048 support a @dfn{.init} section which is executed at program startup,
6049 parts of @file{crtstuff.c} are compiled into that section. The
6050 program is linked by the @command{gcc} driver like this:
6053 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6056 The prologue of a function (@code{__init}) appears in the @code{.init}
6057 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6058 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6059 files are provided by the operating system or by the GNU C library, but
6060 are provided by GCC for a few targets.
6062 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6063 compiled from @file{crtstuff.c}. They contain, among other things, code
6064 fragments within the @code{.init} and @code{.fini} sections that branch
6065 to routines in the @code{.text} section. The linker will pull all parts
6066 of a section together, which results in a complete @code{__init} function
6067 that invokes the routines we need at startup.
6069 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6072 If no init section is available, when GCC compiles any function called
6073 @code{main} (or more accurately, any function designated as a program
6074 entry point by the language front end calling @code{expand_main_function}),
6075 it inserts a procedure call to @code{__main} as the first executable code
6076 after the function prologue. The @code{__main} function is defined
6077 in @file{libgcc2.c} and runs the global constructors.
6079 In file formats that don't support arbitrary sections, there are again
6080 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6081 and an `a.out' format must be used. In this case,
6082 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6083 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6084 and with the address of the void function containing the initialization
6085 code as its value. The GNU linker recognizes this as a request to add
6086 the value to a @dfn{set}; the values are accumulated, and are eventually
6087 placed in the executable as a vector in the format described above, with
6088 a leading (ignored) count and a trailing zero element.
6089 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6090 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6091 the compilation of @code{main} to call @code{__main} as above, starting
6092 the initialization process.
6094 The last variant uses neither arbitrary sections nor the GNU linker.
6095 This is preferable when you want to do dynamic linking and when using
6096 file formats which the GNU linker does not support, such as `ECOFF'@. In
6097 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6098 termination functions are recognized simply by their names. This requires
6099 an extra program in the linkage step, called @command{collect2}. This program
6100 pretends to be the linker, for use with GCC; it does its job by running
6101 the ordinary linker, but also arranges to include the vectors of
6102 initialization and termination functions. These functions are called
6103 via @code{__main} as described above. In order to use this method,
6104 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6107 The following section describes the specific macros that control and
6108 customize the handling of initialization and termination functions.
6111 @node Macros for Initialization
6112 @subsection Macros Controlling Initialization Routines
6114 Here are the macros that control how the compiler handles initialization
6115 and termination functions:
6117 @defmac INIT_SECTION_ASM_OP
6118 If defined, a C string constant, including spacing, for the assembler
6119 operation to identify the following data as initialization code. If not
6120 defined, GCC will assume such a section does not exist. When you are
6121 using special sections for initialization and termination functions, this
6122 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6123 run the initialization functions.
6126 @defmac HAS_INIT_SECTION
6127 If defined, @code{main} will not call @code{__main} as described above.
6128 This macro should be defined for systems that control start-up code
6129 on a symbol-by-symbol basis, such as OSF/1, and should not
6130 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6133 @defmac LD_INIT_SWITCH
6134 If defined, a C string constant for a switch that tells the linker that
6135 the following symbol is an initialization routine.
6138 @defmac LD_FINI_SWITCH
6139 If defined, a C string constant for a switch that tells the linker that
6140 the following symbol is a finalization routine.
6143 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6144 If defined, a C statement that will write a function that can be
6145 automatically called when a shared library is loaded. The function
6146 should call @var{func}, which takes no arguments. If not defined, and
6147 the object format requires an explicit initialization function, then a
6148 function called @code{_GLOBAL__DI} will be generated.
6150 This function and the following one are used by collect2 when linking a
6151 shared library that needs constructors or destructors, or has DWARF2
6152 exception tables embedded in the code.
6155 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6156 If defined, a C statement that will write a function that can be
6157 automatically called when a shared library is unloaded. The function
6158 should call @var{func}, which takes no arguments. If not defined, and
6159 the object format requires an explicit finalization function, then a
6160 function called @code{_GLOBAL__DD} will be generated.
6163 @defmac INVOKE__main
6164 If defined, @code{main} will call @code{__main} despite the presence of
6165 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6166 where the init section is not actually run automatically, but is still
6167 useful for collecting the lists of constructors and destructors.
6170 @defmac SUPPORTS_INIT_PRIORITY
6171 If nonzero, the C++ @code{init_priority} attribute is supported and the
6172 compiler should emit instructions to control the order of initialization
6173 of objects. If zero, the compiler will issue an error message upon
6174 encountering an @code{init_priority} attribute.
6177 @hook TARGET_HAVE_CTORS_DTORS
6179 @hook TARGET_ASM_CONSTRUCTOR
6181 @hook TARGET_ASM_DESTRUCTOR
6183 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6184 generated for the generated object file will have static linkage.
6186 If your system uses @command{collect2} as the means of processing
6187 constructors, then that program normally uses @command{nm} to scan
6188 an object file for constructor functions to be called.
6190 On certain kinds of systems, you can define this macro to make
6191 @command{collect2} work faster (and, in some cases, make it work at all):
6193 @defmac OBJECT_FORMAT_COFF
6194 Define this macro if the system uses COFF (Common Object File Format)
6195 object files, so that @command{collect2} can assume this format and scan
6196 object files directly for dynamic constructor/destructor functions.
6198 This macro is effective only in a native compiler; @command{collect2} as
6199 part of a cross compiler always uses @command{nm} for the target machine.
6202 @defmac REAL_NM_FILE_NAME
6203 Define this macro as a C string constant containing the file name to use
6204 to execute @command{nm}. The default is to search the path normally for
6209 @command{collect2} calls @command{nm} to scan object files for static
6210 constructors and destructors and LTO info. By default, @option{-n} is
6211 passed. Define @code{NM_FLAGS} to a C string constant if other options
6212 are needed to get the same output format as GNU @command{nm -n}
6216 If your system supports shared libraries and has a program to list the
6217 dynamic dependencies of a given library or executable, you can define
6218 these macros to enable support for running initialization and
6219 termination functions in shared libraries:
6222 Define this macro to a C string constant containing the name of the program
6223 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6226 @defmac PARSE_LDD_OUTPUT (@var{ptr})
6227 Define this macro to be C code that extracts filenames from the output
6228 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6229 of type @code{char *} that points to the beginning of a line of output
6230 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6231 code must advance @var{ptr} to the beginning of the filename on that
6232 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6235 @defmac SHLIB_SUFFIX
6236 Define this macro to a C string constant containing the default shared
6237 library extension of the target (e.g., @samp{".so"}). @command{collect2}
6238 strips version information after this suffix when generating global
6239 constructor and destructor names. This define is only needed on targets
6240 that use @command{collect2} to process constructors and destructors.
6243 @node Instruction Output
6244 @subsection Output of Assembler Instructions
6246 @c prevent bad page break with this line
6247 This describes assembler instruction output.
6249 @defmac REGISTER_NAMES
6250 A C initializer containing the assembler's names for the machine
6251 registers, each one as a C string constant. This is what translates
6252 register numbers in the compiler into assembler language.
6255 @defmac ADDITIONAL_REGISTER_NAMES
6256 If defined, a C initializer for an array of structures containing a name
6257 and a register number. This macro defines additional names for hard
6258 registers, thus allowing the @code{asm} option in declarations to refer
6259 to registers using alternate names.
6262 @defmac OVERLAPPING_REGISTER_NAMES
6263 If defined, a C initializer for an array of structures containing a
6264 name, a register number and a count of the number of consecutive
6265 machine registers the name overlaps. This macro defines additional
6266 names for hard registers, thus allowing the @code{asm} option in
6267 declarations to refer to registers using alternate names. Unlike
6268 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6269 register name implies multiple underlying registers.
6271 This macro should be used when it is important that a clobber in an
6272 @code{asm} statement clobbers all the underlying values implied by the
6273 register name. For example, on ARM, clobbering the double-precision
6274 VFP register ``d0'' implies clobbering both single-precision registers
6278 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6279 Define this macro if you are using an unusual assembler that
6280 requires different names for the machine instructions.
6282 The definition is a C statement or statements which output an
6283 assembler instruction opcode to the stdio stream @var{stream}. The
6284 macro-operand @var{ptr} is a variable of type @code{char *} which
6285 points to the opcode name in its ``internal'' form---the form that is
6286 written in the machine description. The definition should output the
6287 opcode name to @var{stream}, performing any translation you desire, and
6288 increment the variable @var{ptr} to point at the end of the opcode
6289 so that it will not be output twice.
6291 In fact, your macro definition may process less than the entire opcode
6292 name, or more than the opcode name; but if you want to process text
6293 that includes @samp{%}-sequences to substitute operands, you must take
6294 care of the substitution yourself. Just be sure to increment
6295 @var{ptr} over whatever text should not be output normally.
6297 @findex recog_data.operand
6298 If you need to look at the operand values, they can be found as the
6299 elements of @code{recog_data.operand}.
6301 If the macro definition does nothing, the instruction is output
6305 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6306 If defined, a C statement to be executed just prior to the output of
6307 assembler code for @var{insn}, to modify the extracted operands so
6308 they will be output differently.
6310 Here the argument @var{opvec} is the vector containing the operands
6311 extracted from @var{insn}, and @var{noperands} is the number of
6312 elements of the vector which contain meaningful data for this insn.
6313 The contents of this vector are what will be used to convert the insn
6314 template into assembler code, so you can change the assembler output
6315 by changing the contents of the vector.
6317 This macro is useful when various assembler syntaxes share a single
6318 file of instruction patterns; by defining this macro differently, you
6319 can cause a large class of instructions to be output differently (such
6320 as with rearranged operands). Naturally, variations in assembler
6321 syntax affecting individual insn patterns ought to be handled by
6322 writing conditional output routines in those patterns.
6324 If this macro is not defined, it is equivalent to a null statement.
6327 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
6329 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6330 A C compound statement to output to stdio stream @var{stream} the
6331 assembler syntax for an instruction operand @var{x}. @var{x} is an
6334 @var{code} is a value that can be used to specify one of several ways
6335 of printing the operand. It is used when identical operands must be
6336 printed differently depending on the context. @var{code} comes from
6337 the @samp{%} specification that was used to request printing of the
6338 operand. If the specification was just @samp{%@var{digit}} then
6339 @var{code} is 0; if the specification was @samp{%@var{ltr}
6340 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6343 If @var{x} is a register, this macro should print the register's name.
6344 The names can be found in an array @code{reg_names} whose type is
6345 @code{char *[]}. @code{reg_names} is initialized from
6346 @code{REGISTER_NAMES}.
6348 When the machine description has a specification @samp{%@var{punct}}
6349 (a @samp{%} followed by a punctuation character), this macro is called
6350 with a null pointer for @var{x} and the punctuation character for
6354 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6355 A C expression which evaluates to true if @var{code} is a valid
6356 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6357 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6358 punctuation characters (except for the standard one, @samp{%}) are used
6362 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6363 A C compound statement to output to stdio stream @var{stream} the
6364 assembler syntax for an instruction operand that is a memory reference
6365 whose address is @var{x}. @var{x} is an RTL expression.
6367 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6368 On some machines, the syntax for a symbolic address depends on the
6369 section that the address refers to. On these machines, define the hook
6370 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6371 @code{symbol_ref}, and then check for it here. @xref{Assembler
6375 @findex dbr_sequence_length
6376 @defmac DBR_OUTPUT_SEQEND (@var{file})
6377 A C statement, to be executed after all slot-filler instructions have
6378 been output. If necessary, call @code{dbr_sequence_length} to
6379 determine the number of slots filled in a sequence (zero if not
6380 currently outputting a sequence), to decide how many no-ops to output,
6383 Don't define this macro if it has nothing to do, but it is helpful in
6384 reading assembly output if the extent of the delay sequence is made
6385 explicit (e.g.@: with white space).
6388 @findex final_sequence
6389 Note that output routines for instructions with delay slots must be
6390 prepared to deal with not being output as part of a sequence
6391 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6392 found.) The variable @code{final_sequence} is null when not
6393 processing a sequence, otherwise it contains the @code{sequence} rtx
6397 @defmac REGISTER_PREFIX
6398 @defmacx LOCAL_LABEL_PREFIX
6399 @defmacx USER_LABEL_PREFIX
6400 @defmacx IMMEDIATE_PREFIX
6401 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6402 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6403 @file{final.c}). These are useful when a single @file{md} file must
6404 support multiple assembler formats. In that case, the various @file{tm.h}
6405 files can define these macros differently.
6408 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6409 If defined this macro should expand to a series of @code{case}
6410 statements which will be parsed inside the @code{switch} statement of
6411 the @code{asm_fprintf} function. This allows targets to define extra
6412 printf formats which may useful when generating their assembler
6413 statements. Note that uppercase letters are reserved for future
6414 generic extensions to asm_fprintf, and so are not available to target
6415 specific code. The output file is given by the parameter @var{file}.
6416 The varargs input pointer is @var{argptr} and the rest of the format
6417 string, starting the character after the one that is being switched
6418 upon, is pointed to by @var{format}.
6421 @defmac ASSEMBLER_DIALECT
6422 If your target supports multiple dialects of assembler language (such as
6423 different opcodes), define this macro as a C expression that gives the
6424 numeric index of the assembler language dialect to use, with zero as the
6427 If this macro is defined, you may use constructs of the form
6429 @samp{@{option0|option1|option2@dots{}@}}
6432 in the output templates of patterns (@pxref{Output Template}) or in the
6433 first argument of @code{asm_fprintf}. This construct outputs
6434 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6435 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
6436 within these strings retain their usual meaning. If there are fewer
6437 alternatives within the braces than the value of
6438 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6439 to print curly braces or @samp{|} character in assembler output directly,
6440 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6442 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6443 @samp{@}} do not have any special meaning when used in templates or
6444 operands to @code{asm_fprintf}.
6446 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6447 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6448 the variations in assembler language syntax with that mechanism. Define
6449 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6450 if the syntax variant are larger and involve such things as different
6451 opcodes or operand order.
6454 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6455 A C expression to output to @var{stream} some assembler code
6456 which will push hard register number @var{regno} onto the stack.
6457 The code need not be optimal, since this macro is used only when
6461 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6462 A C expression to output to @var{stream} some assembler code
6463 which will pop hard register number @var{regno} off of the stack.
6464 The code need not be optimal, since this macro is used only when
6468 @node Dispatch Tables
6469 @subsection Output of Dispatch Tables
6471 @c prevent bad page break with this line
6472 This concerns dispatch tables.
6474 @cindex dispatch table
6475 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6476 A C statement to output to the stdio stream @var{stream} an assembler
6477 pseudo-instruction to generate a difference between two labels.
6478 @var{value} and @var{rel} are the numbers of two internal labels. The
6479 definitions of these labels are output using
6480 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6481 way here. For example,
6484 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6485 @var{value}, @var{rel})
6488 You must provide this macro on machines where the addresses in a
6489 dispatch table are relative to the table's own address. If defined, GCC
6490 will also use this macro on all machines when producing PIC@.
6491 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6492 mode and flags can be read.
6495 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6496 This macro should be provided on machines where the addresses
6497 in a dispatch table are absolute.
6499 The definition should be a C statement to output to the stdio stream
6500 @var{stream} an assembler pseudo-instruction to generate a reference to
6501 a label. @var{value} is the number of an internal label whose
6502 definition is output using @code{(*targetm.asm_out.internal_label)}.
6506 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6510 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6511 Define this if the label before a jump-table needs to be output
6512 specially. The first three arguments are the same as for
6513 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6514 jump-table which follows (a @code{jump_table_data} containing an
6515 @code{addr_vec} or @code{addr_diff_vec}).
6517 This feature is used on system V to output a @code{swbeg} statement
6520 If this macro is not defined, these labels are output with
6521 @code{(*targetm.asm_out.internal_label)}.
6524 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6525 Define this if something special must be output at the end of a
6526 jump-table. The definition should be a C statement to be executed
6527 after the assembler code for the table is written. It should write
6528 the appropriate code to stdio stream @var{stream}. The argument
6529 @var{table} is the jump-table insn, and @var{num} is the label-number
6530 of the preceding label.
6532 If this macro is not defined, nothing special is output at the end of
6536 @hook TARGET_ASM_EMIT_UNWIND_LABEL
6538 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6540 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6542 @hook TARGET_ASM_UNWIND_EMIT
6544 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6546 @node Exception Region Output
6547 @subsection Assembler Commands for Exception Regions
6549 @c prevent bad page break with this line
6551 This describes commands marking the start and the end of an exception
6554 @defmac EH_FRAME_SECTION_NAME
6555 If defined, a C string constant for the name of the section containing
6556 exception handling frame unwind information. If not defined, GCC will
6557 provide a default definition if the target supports named sections.
6558 @file{crtstuff.c} uses this macro to switch to the appropriate section.
6560 You should define this symbol if your target supports DWARF 2 frame
6561 unwind information and the default definition does not work.
6564 @defmac EH_FRAME_THROUGH_COLLECT2
6565 If defined, DWARF 2 frame unwind information will identified by
6566 specially named labels. The collect2 process will locate these
6567 labels and generate code to register the frames.
6569 This might be necessary, for instance, if the system linker will not
6570 place the eh_frames in-between the sentinals from @file{crtstuff.c},
6571 or if the system linker does garbage collection and sections cannot
6572 be marked as not to be collected.
6575 @defmac EH_TABLES_CAN_BE_READ_ONLY
6576 Define this macro to 1 if your target is such that no frame unwind
6577 information encoding used with non-PIC code will ever require a
6578 runtime relocation, but the linker may not support merging read-only
6579 and read-write sections into a single read-write section.
6582 @defmac MASK_RETURN_ADDR
6583 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6584 that it does not contain any extraneous set bits in it.
6587 @defmac DWARF2_UNWIND_INFO
6588 Define this macro to 0 if your target supports DWARF 2 frame unwind
6589 information, but it does not yet work with exception handling.
6590 Otherwise, if your target supports this information (if it defines
6591 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6592 GCC will provide a default definition of 1.
6595 @hook TARGET_EXCEPT_UNWIND_INFO
6596 This hook defines the mechanism that will be used for exception handling
6597 by the target. If the target has ABI specified unwind tables, the hook
6598 should return @code{UI_TARGET}. If the target is to use the
6599 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6600 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
6601 information, the hook should return @code{UI_DWARF2}.
6603 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6604 This may end up simplifying other parts of target-specific code. The
6605 default implementation of this hook never returns @code{UI_NONE}.
6607 Note that the value returned by this hook should be constant. It should
6608 not depend on anything except the command-line switches described by
6609 @var{opts}. In particular, the
6610 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6611 macros and builtin functions related to exception handling are set up
6612 depending on this setting.
6614 The default implementation of the hook first honors the
6615 @option{--enable-sjlj-exceptions} configure option, then
6616 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
6617 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6618 must define this hook so that @var{opts} is used correctly.
6621 @hook TARGET_UNWIND_TABLES_DEFAULT
6622 This variable should be set to @code{true} if the target ABI requires unwinding
6623 tables even when exceptions are not used. It must not be modified by
6624 command-line option processing.
6627 @defmac DONT_USE_BUILTIN_SETJMP
6628 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6629 should use the @code{setjmp}/@code{longjmp} functions from the C library
6630 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6633 @defmac JMP_BUF_SIZE
6634 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6635 defined. Define this macro if the default size of @code{jmp_buf} buffer
6636 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6637 is not large enough, or if it is much too large.
6638 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6641 @defmac DWARF_CIE_DATA_ALIGNMENT
6642 This macro need only be defined if the target might save registers in the
6643 function prologue at an offset to the stack pointer that is not aligned to
6644 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6645 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
6646 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
6647 the target supports DWARF 2 frame unwind information.
6650 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6652 @hook TARGET_DWARF_REGISTER_SPAN
6654 @hook TARGET_DWARF_FRAME_REG_MODE
6656 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6658 @hook TARGET_ASM_TTYPE
6660 @hook TARGET_ARM_EABI_UNWINDER
6662 @node Alignment Output
6663 @subsection Assembler Commands for Alignment
6665 @c prevent bad page break with this line
6666 This describes commands for alignment.
6668 @defmac JUMP_ALIGN (@var{label})
6669 The alignment (log base 2) to put in front of @var{label}, which is
6670 a common destination of jumps and has no fallthru incoming edge.
6672 This macro need not be defined if you don't want any special alignment
6673 to be done at such a time. Most machine descriptions do not currently
6676 Unless it's necessary to inspect the @var{label} parameter, it is better
6677 to set the variable @var{align_jumps} in the target's
6678 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6679 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6682 @hook TARGET_ASM_JUMP_ALIGN_MAX_SKIP
6684 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6685 The alignment (log base 2) to put in front of @var{label}, which follows
6688 This macro need not be defined if you don't want any special alignment
6689 to be done at such a time. Most machine descriptions do not currently
6693 @hook TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6695 @defmac LOOP_ALIGN (@var{label})
6696 The alignment (log base 2) to put in front of @var{label} that heads
6697 a frequently executed basic block (usually the header of a loop).
6699 This macro need not be defined if you don't want any special alignment
6700 to be done at such a time. Most machine descriptions do not currently
6703 Unless it's necessary to inspect the @var{label} parameter, it is better
6704 to set the variable @code{align_loops} in the target's
6705 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6706 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6709 @hook TARGET_ASM_LOOP_ALIGN_MAX_SKIP
6711 @defmac LABEL_ALIGN (@var{label})
6712 The alignment (log base 2) to put in front of @var{label}.
6713 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6714 the maximum of the specified values is used.
6716 Unless it's necessary to inspect the @var{label} parameter, it is better
6717 to set the variable @code{align_labels} in the target's
6718 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6719 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6722 @hook TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6724 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6725 A C statement to output to the stdio stream @var{stream} an assembler
6726 instruction to advance the location counter by @var{nbytes} bytes.
6727 Those bytes should be zero when loaded. @var{nbytes} will be a C
6728 expression of type @code{unsigned HOST_WIDE_INT}.
6731 @defmac ASM_NO_SKIP_IN_TEXT
6732 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6733 text section because it fails to put zeros in the bytes that are skipped.
6734 This is true on many Unix systems, where the pseudo--op to skip bytes
6735 produces no-op instructions rather than zeros when used in the text
6739 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6740 A C statement to output to the stdio stream @var{stream} an assembler
6741 command to advance the location counter to a multiple of 2 to the
6742 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6745 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6746 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6747 for padding, if necessary.
6750 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6751 A C statement to output to the stdio stream @var{stream} an assembler
6752 command to advance the location counter to a multiple of 2 to the
6753 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6754 satisfy the alignment request. @var{power} and @var{max_skip} will be
6755 a C expression of type @code{int}.
6759 @node Debugging Info
6760 @section Controlling Debugging Information Format
6762 @c prevent bad page break with this line
6763 This describes how to specify debugging information.
6766 * All Debuggers:: Macros that affect all debugging formats uniformly.
6767 * DBX Options:: Macros enabling specific options in DBX format.
6768 * DBX Hooks:: Hook macros for varying DBX format.
6769 * File Names and DBX:: Macros controlling output of file names in DBX format.
6770 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6771 * VMS Debug:: Macros for VMS debug format.
6775 @subsection Macros Affecting All Debugging Formats
6777 @c prevent bad page break with this line
6778 These macros affect all debugging formats.
6780 @defmac DBX_REGISTER_NUMBER (@var{regno})
6781 A C expression that returns the DBX register number for the compiler
6782 register number @var{regno}. In the default macro provided, the value
6783 of this expression will be @var{regno} itself. But sometimes there are
6784 some registers that the compiler knows about and DBX does not, or vice
6785 versa. In such cases, some register may need to have one number in the
6786 compiler and another for DBX@.
6788 If two registers have consecutive numbers inside GCC, and they can be
6789 used as a pair to hold a multiword value, then they @emph{must} have
6790 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6791 Otherwise, debuggers will be unable to access such a pair, because they
6792 expect register pairs to be consecutive in their own numbering scheme.
6794 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6795 does not preserve register pairs, then what you must do instead is
6796 redefine the actual register numbering scheme.
6799 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
6800 A C expression that returns the integer offset value for an automatic
6801 variable having address @var{x} (an RTL expression). The default
6802 computation assumes that @var{x} is based on the frame-pointer and
6803 gives the offset from the frame-pointer. This is required for targets
6804 that produce debugging output for DBX or COFF-style debugging output
6805 for SDB and allow the frame-pointer to be eliminated when the
6806 @option{-g} options is used.
6809 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6810 A C expression that returns the integer offset value for an argument
6811 having address @var{x} (an RTL expression). The nominal offset is
6815 @defmac PREFERRED_DEBUGGING_TYPE
6816 A C expression that returns the type of debugging output GCC should
6817 produce when the user specifies just @option{-g}. Define
6818 this if you have arranged for GCC to support more than one format of
6819 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6820 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
6821 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
6823 When the user specifies @option{-ggdb}, GCC normally also uses the
6824 value of this macro to select the debugging output format, but with two
6825 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6826 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6827 defined, GCC uses @code{DBX_DEBUG}.
6829 The value of this macro only affects the default debugging output; the
6830 user can always get a specific type of output by using @option{-gstabs},
6831 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
6835 @subsection Specific Options for DBX Output
6837 @c prevent bad page break with this line
6838 These are specific options for DBX output.
6840 @defmac DBX_DEBUGGING_INFO
6841 Define this macro if GCC should produce debugging output for DBX
6842 in response to the @option{-g} option.
6845 @defmac XCOFF_DEBUGGING_INFO
6846 Define this macro if GCC should produce XCOFF format debugging output
6847 in response to the @option{-g} option. This is a variant of DBX format.
6850 @defmac DEFAULT_GDB_EXTENSIONS
6851 Define this macro to control whether GCC should by default generate
6852 GDB's extended version of DBX debugging information (assuming DBX-format
6853 debugging information is enabled at all). If you don't define the
6854 macro, the default is 1: always generate the extended information
6855 if there is any occasion to.
6858 @defmac DEBUG_SYMS_TEXT
6859 Define this macro if all @code{.stabs} commands should be output while
6860 in the text section.
6863 @defmac ASM_STABS_OP
6864 A C string constant, including spacing, naming the assembler pseudo op to
6865 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6866 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
6867 applies only to DBX debugging information format.
6870 @defmac ASM_STABD_OP
6871 A C string constant, including spacing, naming the assembler pseudo op to
6872 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6873 value is the current location. If you don't define this macro,
6874 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
6878 @defmac ASM_STABN_OP
6879 A C string constant, including spacing, naming the assembler pseudo op to
6880 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6881 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
6882 macro applies only to DBX debugging information format.
6885 @defmac DBX_NO_XREFS
6886 Define this macro if DBX on your system does not support the construct
6887 @samp{xs@var{tagname}}. On some systems, this construct is used to
6888 describe a forward reference to a structure named @var{tagname}.
6889 On other systems, this construct is not supported at all.
6892 @defmac DBX_CONTIN_LENGTH
6893 A symbol name in DBX-format debugging information is normally
6894 continued (split into two separate @code{.stabs} directives) when it
6895 exceeds a certain length (by default, 80 characters). On some
6896 operating systems, DBX requires this splitting; on others, splitting
6897 must not be done. You can inhibit splitting by defining this macro
6898 with the value zero. You can override the default splitting-length by
6899 defining this macro as an expression for the length you desire.
6902 @defmac DBX_CONTIN_CHAR
6903 Normally continuation is indicated by adding a @samp{\} character to
6904 the end of a @code{.stabs} string when a continuation follows. To use
6905 a different character instead, define this macro as a character
6906 constant for the character you want to use. Do not define this macro
6907 if backslash is correct for your system.
6910 @defmac DBX_STATIC_STAB_DATA_SECTION
6911 Define this macro if it is necessary to go to the data section before
6912 outputting the @samp{.stabs} pseudo-op for a non-global static
6916 @defmac DBX_TYPE_DECL_STABS_CODE
6917 The value to use in the ``code'' field of the @code{.stabs} directive
6918 for a typedef. The default is @code{N_LSYM}.
6921 @defmac DBX_STATIC_CONST_VAR_CODE
6922 The value to use in the ``code'' field of the @code{.stabs} directive
6923 for a static variable located in the text section. DBX format does not
6924 provide any ``right'' way to do this. The default is @code{N_FUN}.
6927 @defmac DBX_REGPARM_STABS_CODE
6928 The value to use in the ``code'' field of the @code{.stabs} directive
6929 for a parameter passed in registers. DBX format does not provide any
6930 ``right'' way to do this. The default is @code{N_RSYM}.
6933 @defmac DBX_REGPARM_STABS_LETTER
6934 The letter to use in DBX symbol data to identify a symbol as a parameter
6935 passed in registers. DBX format does not customarily provide any way to
6936 do this. The default is @code{'P'}.
6939 @defmac DBX_FUNCTION_FIRST
6940 Define this macro if the DBX information for a function and its
6941 arguments should precede the assembler code for the function. Normally,
6942 in DBX format, the debugging information entirely follows the assembler
6946 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
6947 Define this macro, with value 1, if the value of a symbol describing
6948 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
6949 relative to the start of the enclosing function. Normally, GCC uses
6950 an absolute address.
6953 @defmac DBX_LINES_FUNCTION_RELATIVE
6954 Define this macro, with value 1, if the value of a symbol indicating
6955 the current line number (@code{N_SLINE}) should be relative to the
6956 start of the enclosing function. Normally, GCC uses an absolute address.
6959 @defmac DBX_USE_BINCL
6960 Define this macro if GCC should generate @code{N_BINCL} and
6961 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6962 macro also directs GCC to output a type number as a pair of a file
6963 number and a type number within the file. Normally, GCC does not
6964 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6965 number for a type number.
6969 @subsection Open-Ended Hooks for DBX Format
6971 @c prevent bad page break with this line
6972 These are hooks for DBX format.
6974 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6975 A C statement to output DBX debugging information before code for line
6976 number @var{line} of the current source file to the stdio stream
6977 @var{stream}. @var{counter} is the number of time the macro was
6978 invoked, including the current invocation; it is intended to generate
6979 unique labels in the assembly output.
6981 This macro should not be defined if the default output is correct, or
6982 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
6985 @defmac NO_DBX_FUNCTION_END
6986 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6987 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
6988 On those machines, define this macro to turn this feature off without
6989 disturbing the rest of the gdb extensions.
6992 @defmac NO_DBX_BNSYM_ENSYM
6993 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
6994 extension construct. On those machines, define this macro to turn this
6995 feature off without disturbing the rest of the gdb extensions.
6998 @node File Names and DBX
6999 @subsection File Names in DBX Format
7001 @c prevent bad page break with this line
7002 This describes file names in DBX format.
7004 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7005 A C statement to output DBX debugging information to the stdio stream
7006 @var{stream}, which indicates that file @var{name} is the main source
7007 file---the file specified as the input file for compilation.
7008 This macro is called only once, at the beginning of compilation.
7010 This macro need not be defined if the standard form of output
7011 for DBX debugging information is appropriate.
7013 It may be necessary to refer to a label equal to the beginning of the
7014 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
7015 to do so. If you do this, you must also set the variable
7016 @var{used_ltext_label_name} to @code{true}.
7019 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
7020 Define this macro, with value 1, if GCC should not emit an indication
7021 of the current directory for compilation and current source language at
7022 the beginning of the file.
7025 @defmac NO_DBX_GCC_MARKER
7026 Define this macro, with value 1, if GCC should not emit an indication
7027 that this object file was compiled by GCC@. The default is to emit
7028 an @code{N_OPT} stab at the beginning of every source file, with
7029 @samp{gcc2_compiled.} for the string and value 0.
7032 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7033 A C statement to output DBX debugging information at the end of
7034 compilation of the main source file @var{name}. Output should be
7035 written to the stdio stream @var{stream}.
7037 If you don't define this macro, nothing special is output at the end
7038 of compilation, which is correct for most machines.
7041 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
7042 Define this macro @emph{instead of} defining
7043 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
7044 the end of compilation is an @code{N_SO} stab with an empty string,
7045 whose value is the highest absolute text address in the file.
7050 @subsection Macros for SDB and DWARF Output
7052 @c prevent bad page break with this line
7053 Here are macros for SDB and DWARF output.
7055 @defmac SDB_DEBUGGING_INFO
7056 Define this macro to 1 if GCC should produce COFF-style debugging output
7057 for SDB in response to the @option{-g} option.
7060 @defmac DWARF2_DEBUGGING_INFO
7061 Define this macro if GCC should produce dwarf version 2 format
7062 debugging output in response to the @option{-g} option.
7064 @hook TARGET_DWARF_CALLING_CONVENTION
7066 To support optional call frame debugging information, you must also
7067 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7068 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7069 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7070 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
7073 @defmac DWARF2_FRAME_INFO
7074 Define this macro to a nonzero value if GCC should always output
7075 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
7076 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
7077 exceptions are enabled, GCC will output this information not matter
7078 how you define @code{DWARF2_FRAME_INFO}.
7081 @hook TARGET_DEBUG_UNWIND_INFO
7083 @defmac DWARF2_ASM_LINE_DEBUG_INFO
7084 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7085 line debug info sections. This will result in much more compact line number
7086 tables, and hence is desirable if it works.
7089 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
7091 @hook TARGET_DELAY_SCHED2
7093 @hook TARGET_DELAY_VARTRACK
7095 @hook TARGET_NO_REGISTER_ALLOCATION
7097 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7098 A C statement to issue assembly directives that create a difference
7099 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
7102 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7103 A C statement to issue assembly directives that create a difference
7104 between the two given labels in system defined units, e.g. instruction
7105 slots on IA64 VMS, using an integer of the given size.
7108 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
7109 A C statement to issue assembly directives that create a
7110 section-relative reference to the given @var{label} plus @var{offset}, using
7111 an integer of the given @var{size}. The label is known to be defined in the
7112 given @var{section}.
7115 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
7116 A C statement to issue assembly directives that create a self-relative
7117 reference to the given @var{label}, using an integer of the given @var{size}.
7120 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
7121 A C statement to issue assembly directives that create a reference to the
7122 given @var{label} relative to the dbase, using an integer of the given @var{size}.
7125 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
7126 A C statement to issue assembly directives that create a reference to
7127 the DWARF table identifier @var{label} from the current section. This
7128 is used on some systems to avoid garbage collecting a DWARF table which
7129 is referenced by a function.
7132 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
7134 @defmac PUT_SDB_@dots{}
7135 Define these macros to override the assembler syntax for the special
7136 SDB assembler directives. See @file{sdbout.c} for a list of these
7137 macros and their arguments. If the standard syntax is used, you need
7138 not define them yourself.
7142 Some assemblers do not support a semicolon as a delimiter, even between
7143 SDB assembler directives. In that case, define this macro to be the
7144 delimiter to use (usually @samp{\n}). It is not necessary to define
7145 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7149 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
7150 Define this macro to allow references to unknown structure,
7151 union, or enumeration tags to be emitted. Standard COFF does not
7152 allow handling of unknown references, MIPS ECOFF has support for
7156 @defmac SDB_ALLOW_FORWARD_REFERENCES
7157 Define this macro to allow references to structure, union, or
7158 enumeration tags that have not yet been seen to be handled. Some
7159 assemblers choke if forward tags are used, while some require it.
7162 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
7163 A C statement to output SDB debugging information before code for line
7164 number @var{line} of the current source file to the stdio stream
7165 @var{stream}. The default is to emit an @code{.ln} directive.
7170 @subsection Macros for VMS Debug Format
7172 @c prevent bad page break with this line
7173 Here are macros for VMS debug format.
7175 @defmac VMS_DEBUGGING_INFO
7176 Define this macro if GCC should produce debugging output for VMS
7177 in response to the @option{-g} option. The default behavior for VMS
7178 is to generate minimal debug info for a traceback in the absence of
7179 @option{-g} unless explicitly overridden with @option{-g0}. This
7180 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
7181 @code{TARGET_OPTION_OVERRIDE}.
7184 @node Floating Point
7185 @section Cross Compilation and Floating Point
7186 @cindex cross compilation and floating point
7187 @cindex floating point and cross compilation
7189 While all modern machines use twos-complement representation for integers,
7190 there are a variety of representations for floating point numbers. This
7191 means that in a cross-compiler the representation of floating point numbers
7192 in the compiled program may be different from that used in the machine
7193 doing the compilation.
7195 Because different representation systems may offer different amounts of
7196 range and precision, all floating point constants must be represented in
7197 the target machine's format. Therefore, the cross compiler cannot
7198 safely use the host machine's floating point arithmetic; it must emulate
7199 the target's arithmetic. To ensure consistency, GCC always uses
7200 emulation to work with floating point values, even when the host and
7201 target floating point formats are identical.
7203 The following macros are provided by @file{real.h} for the compiler to
7204 use. All parts of the compiler which generate or optimize
7205 floating-point calculations must use these macros. They may evaluate
7206 their operands more than once, so operands must not have side effects.
7208 @defmac REAL_VALUE_TYPE
7209 The C data type to be used to hold a floating point value in the target
7210 machine's format. Typically this is a @code{struct} containing an
7211 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7215 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7216 Truncates @var{x} to a signed integer, rounding toward zero.
7219 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7220 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7221 @var{x} is negative, returns zero.
7224 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
7225 Converts @var{string} into a floating point number in the target machine's
7226 representation for mode @var{mode}. This routine can handle both
7227 decimal and hexadecimal floating point constants, using the syntax
7228 defined by the C language for both.
7231 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7232 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7235 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7236 Determines whether @var{x} represents infinity (positive or negative).
7239 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7240 Determines whether @var{x} represents a ``NaN'' (not-a-number).
7243 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
7244 Returns the negative of the floating point value @var{x}.
7247 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
7248 Returns the absolute value of @var{x}.
7251 @node Mode Switching
7252 @section Mode Switching Instructions
7253 @cindex mode switching
7254 The following macros control mode switching optimizations:
7256 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
7257 Define this macro if the port needs extra instructions inserted for mode
7258 switching in an optimizing compilation.
7260 For an example, the SH4 can perform both single and double precision
7261 floating point operations, but to perform a single precision operation,
7262 the FPSCR PR bit has to be cleared, while for a double precision
7263 operation, this bit has to be set. Changing the PR bit requires a general
7264 purpose register as a scratch register, hence these FPSCR sets have to
7265 be inserted before reload, i.e.@: you can't put this into instruction emitting
7266 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
7268 You can have multiple entities that are mode-switched, and select at run time
7269 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7270 return nonzero for any @var{entity} that needs mode-switching.
7271 If you define this macro, you also have to define
7272 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
7273 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
7274 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
7278 @defmac NUM_MODES_FOR_MODE_SWITCHING
7279 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7280 initializer for an array of integers. Each initializer element
7281 N refers to an entity that needs mode switching, and specifies the number
7282 of different modes that might need to be set for this entity.
7283 The position of the initializer in the initializer---starting counting at
7284 zero---determines the integer that is used to refer to the mode-switched
7286 In macros that take mode arguments / yield a mode result, modes are
7287 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
7288 switch is needed / supplied.
7291 @hook TARGET_MODE_EMIT
7293 @hook TARGET_MODE_NEEDED
7295 @hook TARGET_MODE_AFTER
7297 @hook TARGET_MODE_ENTRY
7299 @hook TARGET_MODE_EXIT
7301 @hook TARGET_MODE_PRIORITY
7303 @node Target Attributes
7304 @section Defining target-specific uses of @code{__attribute__}
7305 @cindex target attributes
7306 @cindex machine attributes
7307 @cindex attributes, target-specific
7309 Target-specific attributes may be defined for functions, data and types.
7310 These are described using the following target hooks; they also need to
7311 be documented in @file{extend.texi}.
7313 @hook TARGET_ATTRIBUTE_TABLE
7315 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
7317 @hook TARGET_COMP_TYPE_ATTRIBUTES
7319 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
7321 @hook TARGET_MERGE_TYPE_ATTRIBUTES
7323 @hook TARGET_MERGE_DECL_ATTRIBUTES
7325 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7327 @defmac TARGET_DECLSPEC
7328 Define this macro to a nonzero value if you want to treat
7329 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
7330 default, this behavior is enabled only for targets that define
7331 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
7332 of @code{__declspec} is via a built-in macro, but you should not rely
7333 on this implementation detail.
7336 @hook TARGET_INSERT_ATTRIBUTES
7338 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7340 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
7342 @hook TARGET_OPTION_SAVE
7344 @hook TARGET_OPTION_RESTORE
7346 @hook TARGET_OPTION_POST_STREAM_IN
7348 @hook TARGET_OPTION_PRINT
7350 @hook TARGET_OPTION_PRAGMA_PARSE
7352 @hook TARGET_OPTION_OVERRIDE
7354 @hook TARGET_OPTION_FUNCTION_VERSIONS
7356 @hook TARGET_CAN_INLINE_P
7358 @hook TARGET_RELAYOUT_FUNCTION
7361 @section Emulating TLS
7362 @cindex Emulated TLS
7364 For targets whose psABI does not provide Thread Local Storage via
7365 specific relocations and instruction sequences, an emulation layer is
7366 used. A set of target hooks allows this emulation layer to be
7367 configured for the requirements of a particular target. For instance
7368 the psABI may in fact specify TLS support in terms of an emulation
7371 The emulation layer works by creating a control object for every TLS
7372 object. To access the TLS object, a lookup function is provided
7373 which, when given the address of the control object, will return the
7374 address of the current thread's instance of the TLS object.
7376 @hook TARGET_EMUTLS_GET_ADDRESS
7378 @hook TARGET_EMUTLS_REGISTER_COMMON
7380 @hook TARGET_EMUTLS_VAR_SECTION
7382 @hook TARGET_EMUTLS_TMPL_SECTION
7384 @hook TARGET_EMUTLS_VAR_PREFIX
7386 @hook TARGET_EMUTLS_TMPL_PREFIX
7388 @hook TARGET_EMUTLS_VAR_FIELDS
7390 @hook TARGET_EMUTLS_VAR_INIT
7392 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7394 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7396 @node MIPS Coprocessors
7397 @section Defining coprocessor specifics for MIPS targets.
7398 @cindex MIPS coprocessor-definition macros
7400 The MIPS specification allows MIPS implementations to have as many as 4
7401 coprocessors, each with as many as 32 private registers. GCC supports
7402 accessing these registers and transferring values between the registers
7403 and memory using asm-ized variables. For example:
7406 register unsigned int cp0count asm ("c0r1");
7412 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7413 names may be added as described below, or the default names may be
7414 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7416 Coprocessor registers are assumed to be epilogue-used; sets to them will
7417 be preserved even if it does not appear that the register is used again
7418 later in the function.
7420 Another note: according to the MIPS spec, coprocessor 1 (if present) is
7421 the FPU@. One accesses COP1 registers through standard mips
7422 floating-point support; they are not included in this mechanism.
7425 @section Parameters for Precompiled Header Validity Checking
7426 @cindex parameters, precompiled headers
7428 @hook TARGET_GET_PCH_VALIDITY
7430 @hook TARGET_PCH_VALID_P
7432 @hook TARGET_CHECK_PCH_TARGET_FLAGS
7434 @hook TARGET_PREPARE_PCH_SAVE
7437 @section C++ ABI parameters
7438 @cindex parameters, c++ abi
7440 @hook TARGET_CXX_GUARD_TYPE
7442 @hook TARGET_CXX_GUARD_MASK_BIT
7444 @hook TARGET_CXX_GET_COOKIE_SIZE
7446 @hook TARGET_CXX_COOKIE_HAS_SIZE
7448 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
7450 @hook TARGET_CXX_CDTOR_RETURNS_THIS
7452 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7454 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7456 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7458 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7460 @hook TARGET_CXX_USE_AEABI_ATEXIT
7462 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7464 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7466 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
7468 @node Named Address Spaces
7469 @section Adding support for named address spaces
7470 @cindex named address spaces
7472 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7473 standards committee, @cite{Programming Languages - C - Extensions to
7474 support embedded processors}, specifies a syntax for embedded
7475 processors to specify alternate address spaces. You can configure a
7476 GCC port to support section 5.1 of the draft report to add support for
7477 address spaces other than the default address space. These address
7478 spaces are new keywords that are similar to the @code{volatile} and
7479 @code{const} type attributes.
7481 Pointers to named address spaces can have a different size than
7482 pointers to the generic address space.
7484 For example, the SPU port uses the @code{__ea} address space to refer
7485 to memory in the host processor, rather than memory local to the SPU
7486 processor. Access to memory in the @code{__ea} address space involves
7487 issuing DMA operations to move data between the host processor and the
7488 local processor memory address space. Pointers in the @code{__ea}
7489 address space are either 32 bits or 64 bits based on the
7490 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
7493 Internally, address spaces are represented as a small integer in the
7494 range 0 to 15 with address space 0 being reserved for the generic
7497 To register a named address space qualifier keyword with the C front end,
7498 the target may call the @code{c_register_addr_space} routine. For example,
7499 the SPU port uses the following to declare @code{__ea} as the keyword for
7500 named address space #1:
7502 #define ADDR_SPACE_EA 1
7503 c_register_addr_space ("__ea", ADDR_SPACE_EA);
7506 @hook TARGET_ADDR_SPACE_POINTER_MODE
7508 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
7510 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7512 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7514 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7516 @hook TARGET_ADDR_SPACE_SUBSET_P
7518 @hook TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID
7520 @hook TARGET_ADDR_SPACE_CONVERT
7522 @hook TARGET_ADDR_SPACE_DEBUG
7524 @hook TARGET_ADDR_SPACE_DIAGNOSE_USAGE
7527 @section Miscellaneous Parameters
7528 @cindex parameters, miscellaneous
7530 @c prevent bad page break with this line
7531 Here are several miscellaneous parameters.
7533 @defmac HAS_LONG_COND_BRANCH
7534 Define this boolean macro to indicate whether or not your architecture
7535 has conditional branches that can span all of memory. It is used in
7536 conjunction with an optimization that partitions hot and cold basic
7537 blocks into separate sections of the executable. If this macro is
7538 set to false, gcc will convert any conditional branches that attempt
7539 to cross between sections into unconditional branches or indirect jumps.
7542 @defmac HAS_LONG_UNCOND_BRANCH
7543 Define this boolean macro to indicate whether or not your architecture
7544 has unconditional branches that can span all of memory. It is used in
7545 conjunction with an optimization that partitions hot and cold basic
7546 blocks into separate sections of the executable. If this macro is
7547 set to false, gcc will convert any unconditional branches that attempt
7548 to cross between sections into indirect jumps.
7551 @defmac CASE_VECTOR_MODE
7552 An alias for a machine mode name. This is the machine mode that
7553 elements of a jump-table should have.
7556 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7557 Optional: return the preferred mode for an @code{addr_diff_vec}
7558 when the minimum and maximum offset are known. If you define this,
7559 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7560 To make this work, you also have to define @code{INSN_ALIGN} and
7561 make the alignment for @code{addr_diff_vec} explicit.
7562 The @var{body} argument is provided so that the offset_unsigned and scale
7563 flags can be updated.
7566 @defmac CASE_VECTOR_PC_RELATIVE
7567 Define this macro to be a C expression to indicate when jump-tables
7568 should contain relative addresses. You need not define this macro if
7569 jump-tables never contain relative addresses, or jump-tables should
7570 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7574 @hook TARGET_CASE_VALUES_THRESHOLD
7576 @defmac WORD_REGISTER_OPERATIONS
7577 Define this macro to 1 if operations between registers with integral mode
7578 smaller than a word are always performed on the entire register.
7579 Most RISC machines have this property and most CISC machines do not.
7582 @hook TARGET_MIN_ARITHMETIC_PRECISION
7584 @defmac LOAD_EXTEND_OP (@var{mem_mode})
7585 Define this macro to be a C expression indicating when insns that read
7586 memory in @var{mem_mode}, an integral mode narrower than a word, set the
7587 bits outside of @var{mem_mode} to be either the sign-extension or the
7588 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7589 of @var{mem_mode} for which the
7590 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7591 @code{UNKNOWN} for other modes.
7593 This macro is not called with @var{mem_mode} non-integral or with a width
7594 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7595 value in this case. Do not define this macro if it would always return
7596 @code{UNKNOWN}. On machines where this macro is defined, you will normally
7597 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7599 You may return a non-@code{UNKNOWN} value even if for some hard registers
7600 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7601 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
7602 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7603 integral mode larger than this but not larger than @code{word_mode}.
7605 You must return @code{UNKNOWN} if for some hard registers that allow this
7606 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
7607 @code{word_mode}, but that they can change to another integral mode that
7608 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7611 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
7612 Define this macro to 1 if loading short immediate values into registers sign
7616 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7619 The maximum number of bytes that a single instruction can move quickly
7620 between memory and registers or between two memory locations.
7623 @defmac MAX_MOVE_MAX
7624 The maximum number of bytes that a single instruction can move quickly
7625 between memory and registers or between two memory locations. If this
7626 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7627 constant value that is the largest value that @code{MOVE_MAX} can have
7631 @defmac SHIFT_COUNT_TRUNCATED
7632 A C expression that is nonzero if on this machine the number of bits
7633 actually used for the count of a shift operation is equal to the number
7634 of bits needed to represent the size of the object being shifted. When
7635 this macro is nonzero, the compiler will assume that it is safe to omit
7636 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7637 truncates the count of a shift operation. On machines that have
7638 instructions that act on bit-fields at variable positions, which may
7639 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7640 also enables deletion of truncations of the values that serve as
7641 arguments to bit-field instructions.
7643 If both types of instructions truncate the count (for shifts) and
7644 position (for bit-field operations), or if no variable-position bit-field
7645 instructions exist, you should define this macro.
7647 However, on some machines, such as the 80386 and the 680x0, truncation
7648 only applies to shift operations and not the (real or pretended)
7649 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7650 such machines. Instead, add patterns to the @file{md} file that include
7651 the implied truncation of the shift instructions.
7653 You need not define this macro if it would always have the value of zero.
7656 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
7657 @hook TARGET_SHIFT_TRUNCATION_MASK
7659 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7660 A C expression which is nonzero if on this machine it is safe to
7661 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7662 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7663 operating on it as if it had only @var{outprec} bits.
7665 On many machines, this expression can be 1.
7667 @c rearranged this, removed the phrase "it is reported that". this was
7668 @c to fix an overfull hbox. --mew 10feb93
7669 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7670 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7671 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7672 such cases may improve things.
7675 @hook TARGET_MODE_REP_EXTENDED
7677 @defmac STORE_FLAG_VALUE
7678 A C expression describing the value returned by a comparison operator
7679 with an integral mode and stored by a store-flag instruction
7680 (@samp{cstore@var{mode}4}) when the condition is true. This description must
7681 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7682 comparison operators whose results have a @code{MODE_INT} mode.
7684 A value of 1 or @minus{}1 means that the instruction implementing the
7685 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7686 and 0 when the comparison is false. Otherwise, the value indicates
7687 which bits of the result are guaranteed to be 1 when the comparison is
7688 true. This value is interpreted in the mode of the comparison
7689 operation, which is given by the mode of the first operand in the
7690 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
7691 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7694 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7695 generate code that depends only on the specified bits. It can also
7696 replace comparison operators with equivalent operations if they cause
7697 the required bits to be set, even if the remaining bits are undefined.
7698 For example, on a machine whose comparison operators return an
7699 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7700 @samp{0x80000000}, saying that just the sign bit is relevant, the
7704 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7711 (ashift:SI @var{x} (const_int @var{n}))
7715 where @var{n} is the appropriate shift count to move the bit being
7716 tested into the sign bit.
7718 There is no way to describe a machine that always sets the low-order bit
7719 for a true value, but does not guarantee the value of any other bits,
7720 but we do not know of any machine that has such an instruction. If you
7721 are trying to port GCC to such a machine, include an instruction to
7722 perform a logical-and of the result with 1 in the pattern for the
7723 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7725 Often, a machine will have multiple instructions that obtain a value
7726 from a comparison (or the condition codes). Here are rules to guide the
7727 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7732 Use the shortest sequence that yields a valid definition for
7733 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7734 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7735 comparison operators to do so because there may be opportunities to
7736 combine the normalization with other operations.
7739 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7740 slightly preferred on machines with expensive jumps and 1 preferred on
7744 As a second choice, choose a value of @samp{0x80000001} if instructions
7745 exist that set both the sign and low-order bits but do not define the
7749 Otherwise, use a value of @samp{0x80000000}.
7752 Many machines can produce both the value chosen for
7753 @code{STORE_FLAG_VALUE} and its negation in the same number of
7754 instructions. On those machines, you should also define a pattern for
7755 those cases, e.g., one matching
7758 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7761 Some machines can also perform @code{and} or @code{plus} operations on
7762 condition code values with less instructions than the corresponding
7763 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
7764 machines, define the appropriate patterns. Use the names @code{incscc}
7765 and @code{decscc}, respectively, for the patterns which perform
7766 @code{plus} or @code{minus} operations on condition code values. See
7767 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
7768 find such instruction sequences on other machines.
7770 If this macro is not defined, the default value, 1, is used. You need
7771 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7772 instructions, or if the value generated by these instructions is 1.
7775 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7776 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7777 returned when comparison operators with floating-point results are true.
7778 Define this macro on machines that have comparison operations that return
7779 floating-point values. If there are no such operations, do not define
7783 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7784 A C expression that gives a rtx representing the nonzero true element
7785 for vector comparisons. The returned rtx should be valid for the inner
7786 mode of @var{mode} which is guaranteed to be a vector mode. Define
7787 this macro on machines that have vector comparison operations that
7788 return a vector result. If there are no such operations, do not define
7789 this macro. Typically, this macro is defined as @code{const1_rtx} or
7790 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
7791 the compiler optimizing such vector comparison operations for the
7795 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7796 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7797 A C expression that indicates whether the architecture defines a value
7798 for @code{clz} or @code{ctz} with a zero operand.
7799 A result of @code{0} indicates the value is undefined.
7800 If the value is defined for only the RTL expression, the macro should
7801 evaluate to @code{1}; if the value applies also to the corresponding optab
7802 entry (which is normally the case if it expands directly into
7803 the corresponding RTL), then the macro should evaluate to @code{2}.
7804 In the cases where the value is defined, @var{value} should be set to
7807 If this macro is not defined, the value of @code{clz} or
7808 @code{ctz} at zero is assumed to be undefined.
7810 This macro must be defined if the target's expansion for @code{ffs}
7811 relies on a particular value to get correct results. Otherwise it
7812 is not necessary, though it may be used to optimize some corner cases, and
7813 to provide a default expansion for the @code{ffs} optab.
7815 Note that regardless of this macro the ``definedness'' of @code{clz}
7816 and @code{ctz} at zero do @emph{not} extend to the builtin functions
7817 visible to the user. Thus one may be free to adjust the value at will
7818 to match the target expansion of these operations without fear of
7823 An alias for the machine mode for pointers. On most machines, define
7824 this to be the integer mode corresponding to the width of a hardware
7825 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7826 On some machines you must define this to be one of the partial integer
7827 modes, such as @code{PSImode}.
7829 The width of @code{Pmode} must be at least as large as the value of
7830 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7831 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7835 @defmac FUNCTION_MODE
7836 An alias for the machine mode used for memory references to functions
7837 being called, in @code{call} RTL expressions. On most CISC machines,
7838 where an instruction can begin at any byte address, this should be
7839 @code{QImode}. On most RISC machines, where all instructions have fixed
7840 size and alignment, this should be a mode with the same size and alignment
7841 as the machine instruction words - typically @code{SImode} or @code{HImode}.
7844 @defmac STDC_0_IN_SYSTEM_HEADERS
7845 In normal operation, the preprocessor expands @code{__STDC__} to the
7846 constant 1, to signify that GCC conforms to ISO Standard C@. On some
7847 hosts, like Solaris, the system compiler uses a different convention,
7848 where @code{__STDC__} is normally 0, but is 1 if the user specifies
7849 strict conformance to the C Standard.
7851 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7852 convention when processing system header files, but when processing user
7853 files @code{__STDC__} will always expand to 1.
7856 @hook TARGET_C_PREINCLUDE
7858 @hook TARGET_CXX_IMPLICIT_EXTERN_C
7860 @defmac NO_IMPLICIT_EXTERN_C
7861 Define this macro if the system header files support C++ as well as C@.
7862 This macro inhibits the usual method of using system header files in
7863 C++, which is to pretend that the file's contents are enclosed in
7864 @samp{extern "C" @{@dots{}@}}.
7869 @defmac REGISTER_TARGET_PRAGMAS ()
7870 Define this macro if you want to implement any target-specific pragmas.
7871 If defined, it is a C expression which makes a series of calls to
7872 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
7873 for each pragma. The macro may also do any
7874 setup required for the pragmas.
7876 The primary reason to define this macro is to provide compatibility with
7877 other compilers for the same target. In general, we discourage
7878 definition of target-specific pragmas for GCC@.
7880 If the pragma can be implemented by attributes then you should consider
7881 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
7883 Preprocessor macros that appear on pragma lines are not expanded. All
7884 @samp{#pragma} directives that do not match any registered pragma are
7885 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
7888 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7889 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7891 Each call to @code{c_register_pragma} or
7892 @code{c_register_pragma_with_expansion} establishes one pragma. The
7893 @var{callback} routine will be called when the preprocessor encounters a
7897 #pragma [@var{space}] @var{name} @dots{}
7900 @var{space} is the case-sensitive namespace of the pragma, or
7901 @code{NULL} to put the pragma in the global namespace. The callback
7902 routine receives @var{pfile} as its first argument, which can be passed
7903 on to cpplib's functions if necessary. You can lex tokens after the
7904 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
7905 callback will be silently ignored. The end of the line is indicated by
7906 a token of type @code{CPP_EOF}. Macro expansion occurs on the
7907 arguments of pragmas registered with
7908 @code{c_register_pragma_with_expansion} but not on the arguments of
7909 pragmas registered with @code{c_register_pragma}.
7911 Note that the use of @code{pragma_lex} is specific to the C and C++
7912 compilers. It will not work in the Java or Fortran compilers, or any
7913 other language compilers for that matter. Thus if @code{pragma_lex} is going
7914 to be called from target-specific code, it must only be done so when
7915 building the C and C++ compilers. This can be done by defining the
7916 variables @code{c_target_objs} and @code{cxx_target_objs} in the
7917 target entry in the @file{config.gcc} file. These variables should name
7918 the target-specific, language-specific object file which contains the
7919 code that uses @code{pragma_lex}. Note it will also be necessary to add a
7920 rule to the makefile fragment pointed to by @code{tmake_file} that shows
7921 how to build this object file.
7924 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
7925 Define this macro if macros should be expanded in the
7926 arguments of @samp{#pragma pack}.
7929 @defmac TARGET_DEFAULT_PACK_STRUCT
7930 If your target requires a structure packing default other than 0 (meaning
7931 the machine default), define this macro to the necessary value (in bytes).
7932 This must be a value that would also be valid to use with
7933 @samp{#pragma pack()} (that is, a small power of two).
7936 @defmac DOLLARS_IN_IDENTIFIERS
7937 Define this macro to control use of the character @samp{$} in
7938 identifier names for the C family of languages. 0 means @samp{$} is
7939 not allowed by default; 1 means it is allowed. 1 is the default;
7940 there is no need to define this macro in that case.
7943 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
7944 Define this macro as a C expression that is nonzero if it is safe for the
7945 delay slot scheduler to place instructions in the delay slot of @var{insn},
7946 even if they appear to use a resource set or clobbered in @var{insn}.
7947 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7948 every @code{call_insn} has this behavior. On machines where some @code{insn}
7949 or @code{jump_insn} is really a function call and hence has this behavior,
7950 you should define this macro.
7952 You need not define this macro if it would always return zero.
7955 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
7956 Define this macro as a C expression that is nonzero if it is safe for the
7957 delay slot scheduler to place instructions in the delay slot of @var{insn},
7958 even if they appear to set or clobber a resource referenced in @var{insn}.
7959 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7960 some @code{insn} or @code{jump_insn} is really a function call and its operands
7961 are registers whose use is actually in the subroutine it calls, you should
7962 define this macro. Doing so allows the delay slot scheduler to move
7963 instructions which copy arguments into the argument registers into the delay
7966 You need not define this macro if it would always return zero.
7969 @defmac MULTIPLE_SYMBOL_SPACES
7970 Define this macro as a C expression that is nonzero if, in some cases,
7971 global symbols from one translation unit may not be bound to undefined
7972 symbols in another translation unit without user intervention. For
7973 instance, under Microsoft Windows symbols must be explicitly imported
7974 from shared libraries (DLLs).
7976 You need not define this macro if it would always evaluate to zero.
7979 @hook TARGET_MD_ASM_ADJUST
7981 @defmac MATH_LIBRARY
7982 Define this macro as a C string constant for the linker argument to link
7983 in the system math library, minus the initial @samp{"-l"}, or
7984 @samp{""} if the target does not have a
7985 separate math library.
7987 You need only define this macro if the default of @samp{"m"} is wrong.
7990 @defmac LIBRARY_PATH_ENV
7991 Define this macro as a C string constant for the environment variable that
7992 specifies where the linker should look for libraries.
7994 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7998 @defmac TARGET_POSIX_IO
7999 Define this macro if the target supports the following POSIX@ file
8000 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
8001 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
8002 to use file locking when exiting a program, which avoids race conditions
8003 if the program has forked. It will also create directories at run-time
8004 for cross-profiling.
8007 @defmac MAX_CONDITIONAL_EXECUTE
8009 A C expression for the maximum number of instructions to execute via
8010 conditional execution instructions instead of a branch. A value of
8011 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8012 1 if it does use cc0.
8015 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
8016 Used if the target needs to perform machine-dependent modifications on the
8017 conditionals used for turning basic blocks into conditionally executed code.
8018 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8019 contains information about the currently processed blocks. @var{true_expr}
8020 and @var{false_expr} are the tests that are used for converting the
8021 then-block and the else-block, respectively. Set either @var{true_expr} or
8022 @var{false_expr} to a null pointer if the tests cannot be converted.
8025 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
8026 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
8027 if-statements into conditions combined by @code{and} and @code{or} operations.
8028 @var{bb} contains the basic block that contains the test that is currently
8029 being processed and about to be turned into a condition.
8032 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
8033 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
8034 be converted to conditional execution format. @var{ce_info} points to
8035 a data structure, @code{struct ce_if_block}, which contains information
8036 about the currently processed blocks.
8039 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
8040 A C expression to perform any final machine dependent modifications in
8041 converting code to conditional execution. The involved basic blocks
8042 can be found in the @code{struct ce_if_block} structure that is pointed
8043 to by @var{ce_info}.
8046 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
8047 A C expression to cancel any machine dependent modifications in
8048 converting code to conditional execution. The involved basic blocks
8049 can be found in the @code{struct ce_if_block} structure that is pointed
8050 to by @var{ce_info}.
8053 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
8054 A C expression to initialize any machine specific data for if-conversion
8055 of the if-block in the @code{struct ce_if_block} structure that is pointed
8056 to by @var{ce_info}.
8059 @hook TARGET_MACHINE_DEPENDENT_REORG
8061 @hook TARGET_INIT_BUILTINS
8063 @hook TARGET_BUILTIN_DECL
8065 @hook TARGET_EXPAND_BUILTIN
8067 @hook TARGET_BUILTIN_CHKP_FUNCTION
8068 @hook TARGET_CHKP_BOUND_TYPE
8069 @hook TARGET_CHKP_BOUND_MODE
8070 @hook TARGET_CHKP_MAKE_BOUNDS_CONSTANT
8071 @hook TARGET_CHKP_INITIALIZE_BOUNDS
8073 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
8075 @hook TARGET_FOLD_BUILTIN
8077 @hook TARGET_GIMPLE_FOLD_BUILTIN
8079 @hook TARGET_COMPARE_VERSION_PRIORITY
8081 @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
8083 @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
8085 @hook TARGET_CAN_USE_DOLOOP_P
8087 @hook TARGET_INVALID_WITHIN_DOLOOP
8089 @hook TARGET_LEGITIMATE_COMBINED_INSN
8091 @hook TARGET_CAN_FOLLOW_JUMP
8093 @hook TARGET_COMMUTATIVE_P
8095 @hook TARGET_ALLOCATE_INITIAL_VALUE
8097 @hook TARGET_UNSPEC_MAY_TRAP_P
8099 @hook TARGET_SET_CURRENT_FUNCTION
8101 @defmac TARGET_OBJECT_SUFFIX
8102 Define this macro to be a C string representing the suffix for object
8103 files on your target machine. If you do not define this macro, GCC will
8104 use @samp{.o} as the suffix for object files.
8107 @defmac TARGET_EXECUTABLE_SUFFIX
8108 Define this macro to be a C string representing the suffix to be
8109 automatically added to executable files on your target machine. If you
8110 do not define this macro, GCC will use the null string as the suffix for
8114 @defmac COLLECT_EXPORT_LIST
8115 If defined, @code{collect2} will scan the individual object files
8116 specified on its command line and create an export list for the linker.
8117 Define this macro for systems like AIX, where the linker discards
8118 object files that are not referenced from @code{main} and uses export
8122 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
8123 Define this macro to a C expression representing a variant of the
8124 method call @var{mdecl}, if Java Native Interface (JNI) methods
8125 must be invoked differently from other methods on your target.
8126 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
8127 the @code{stdcall} calling convention and this macro is then
8128 defined as this expression:
8131 build_type_attribute_variant (@var{mdecl},
8133 (get_identifier ("stdcall"),
8138 @hook TARGET_CANNOT_MODIFY_JUMPS_P
8140 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
8142 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
8144 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
8146 @hook TARGET_GEN_CCMP_FIRST
8148 @hook TARGET_GEN_CCMP_NEXT
8150 @hook TARGET_LOOP_UNROLL_ADJUST
8152 @defmac POWI_MAX_MULTS
8153 If defined, this macro is interpreted as a signed integer C expression
8154 that specifies the maximum number of floating point multiplications
8155 that should be emitted when expanding exponentiation by an integer
8156 constant inline. When this value is defined, exponentiation requiring
8157 more than this number of multiplications is implemented by calling the
8158 system library's @code{pow}, @code{powf} or @code{powl} routines.
8159 The default value places no upper bound on the multiplication count.
8162 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8163 This target hook should register any extra include files for the
8164 target. The parameter @var{stdinc} indicates if normal include files
8165 are present. The parameter @var{sysroot} is the system root directory.
8166 The parameter @var{iprefix} is the prefix for the gcc directory.
8169 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8170 This target hook should register any extra include files for the
8171 target before any standard headers. The parameter @var{stdinc}
8172 indicates if normal include files are present. The parameter
8173 @var{sysroot} is the system root directory. The parameter
8174 @var{iprefix} is the prefix for the gcc directory.
8177 @deftypefn Macro void TARGET_OPTF (char *@var{path})
8178 This target hook should register special include paths for the target.
8179 The parameter @var{path} is the include to register. On Darwin
8180 systems, this is used for Framework includes, which have semantics
8181 that are different from @option{-I}.
8184 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
8185 This target macro returns @code{true} if it is safe to use a local alias
8186 for a virtual function @var{fndecl} when constructing thunks,
8187 @code{false} otherwise. By default, the macro returns @code{true} for all
8188 functions, if a target supports aliases (i.e.@: defines
8189 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
8192 @defmac TARGET_FORMAT_TYPES
8193 If defined, this macro is the name of a global variable containing
8194 target-specific format checking information for the @option{-Wformat}
8195 option. The default is to have no target-specific format checks.
8198 @defmac TARGET_N_FORMAT_TYPES
8199 If defined, this macro is the number of entries in
8200 @code{TARGET_FORMAT_TYPES}.
8203 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
8204 If defined, this macro is the name of a global variable containing
8205 target-specific format overrides for the @option{-Wformat} option. The
8206 default is to have no target-specific format overrides. If defined,
8207 @code{TARGET_FORMAT_TYPES} must be defined, too.
8210 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
8211 If defined, this macro specifies the number of entries in
8212 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
8215 @defmac TARGET_OVERRIDES_FORMAT_INIT
8216 If defined, this macro specifies the optional initialization
8217 routine for target specific customizations of the system printf
8218 and scanf formatter settings.
8221 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
8223 @hook TARGET_INVALID_CONVERSION
8225 @hook TARGET_INVALID_UNARY_OP
8227 @hook TARGET_INVALID_BINARY_OP
8229 @hook TARGET_PROMOTED_TYPE
8231 @hook TARGET_CONVERT_TO_TYPE
8234 This macro determines the size of the objective C jump buffer for the
8235 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
8238 @defmac LIBGCC2_UNWIND_ATTRIBUTE
8239 Define this macro if any target-specific attributes need to be attached
8240 to the functions in @file{libgcc} that provide low-level support for
8241 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
8242 and the associated definitions of those functions.
8245 @hook TARGET_UPDATE_STACK_BOUNDARY
8247 @hook TARGET_GET_DRAP_RTX
8249 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
8251 @hook TARGET_CONST_ANCHOR
8253 @hook TARGET_ASAN_SHADOW_OFFSET
8255 @hook TARGET_MEMMODEL_CHECK
8257 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
8259 @hook TARGET_HAS_IFUNC_P
8261 @hook TARGET_ATOMIC_ALIGN_FOR_MODE
8263 @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV
8265 @hook TARGET_RECORD_OFFLOAD_SYMBOL
8267 @hook TARGET_OFFLOAD_OPTIONS
8269 @defmac TARGET_SUPPORTS_WIDE_INT
8271 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
8272 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
8273 to indicate that large integers are stored in
8274 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
8275 very large integer constants to be represented. @code{CONST_DOUBLE}
8276 is limited to twice the size of the host's @code{HOST_WIDE_INT}
8279 Converting a port mostly requires looking for the places where
8280 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
8281 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
8282 const_double"} at the port level gets you to 95% of the changes that
8283 need to be made. There are a few places that require a deeper look.
8287 There is no equivalent to @code{hval} and @code{lval} for
8288 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
8289 language since there are a variable number of elements.
8291 Most ports only check that @code{hval} is either 0 or -1 to see if the
8292 value is small. As mentioned above, this will no longer be necessary
8293 since small constants are always @code{CONST_INT}. Of course there
8294 are still a few exceptions, the alpha's constraint used by the zap
8295 instruction certainly requires careful examination by C code.
8296 However, all the current code does is pass the hval and lval to C
8297 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
8298 not really a large change.
8301 Because there is no standard template that ports use to materialize
8302 constants, there is likely to be some futzing that is unique to each
8306 The rtx costs may have to be adjusted to properly account for larger
8307 constants that are represented as @code{CONST_WIDE_INT}.
8310 All and all it does not take long to convert ports that the
8311 maintainer is familiar with.
8315 @hook TARGET_RUN_TARGET_SELFTESTS