1 @c Copyright (C) 1988-2013 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 * Old Constraints:: The old way to define machine-specific constraints.
36 * Stack and Calling:: Defining which way the stack grows and by how much.
37 * Varargs:: Defining the varargs macros.
38 * Trampolines:: Code set up at run time to enter a nested function.
39 * Library Calls:: Controlling how library routines are implicitly called.
40 * Addressing Modes:: Defining addressing modes valid for memory operands.
41 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
42 * Condition Code:: Defining how insns update the condition code.
43 * Costs:: Defining relative costs of different operations.
44 * Scheduling:: Adjusting the behavior of the instruction scheduler.
45 * Sections:: Dividing storage into text, data, and other sections.
46 * PIC:: Macros for position independent code.
47 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
48 * Debugging Info:: Defining the format of debugging output.
49 * Floating Point:: Handling floating point for cross-compilers.
50 * Mode Switching:: Insertion of mode-switching instructions.
51 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
52 * Emulated TLS:: Emulated TLS support.
53 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
54 * PCH Target:: Validity checking for precompiled headers.
55 * C++ ABI:: Controlling C++ ABI changes.
56 * Named Address Spaces:: Adding support for named address spaces
57 * Misc:: Everything else.
60 @node Target Structure
61 @section The Global @code{targetm} Variable
63 @cindex target functions
65 @deftypevar {struct gcc_target} targetm
66 The target @file{.c} file must define the global @code{targetm} variable
67 which contains pointers to functions and data relating to the target
68 machine. The variable is declared in @file{target.h};
69 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
70 used to initialize the variable, and macros for the default initializers
71 for elements of the structure. The @file{.c} file should override those
72 macros for which the default definition is inappropriate. For example:
75 #include "target-def.h"
77 /* @r{Initialize the GCC target structure.} */
79 #undef TARGET_COMP_TYPE_ATTRIBUTES
80 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
82 struct gcc_target targetm = TARGET_INITIALIZER;
86 Where a macro should be defined in the @file{.c} file in this manner to
87 form part of the @code{targetm} structure, it is documented below as a
88 ``Target Hook'' with a prototype. Many macros will change in future
89 from being defined in the @file{.h} file to being part of the
90 @code{targetm} structure.
92 Similarly, there is a @code{targetcm} variable for hooks that are
93 specific to front ends for C-family languages, documented as ``C
94 Target Hook''. This is declared in @file{c-family/c-target.h}, the
95 initializer @code{TARGETCM_INITIALIZER} in
96 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
97 themselves, they should set @code{target_has_targetcm=yes} in
98 @file{config.gcc}; otherwise a default definition is used.
100 Similarly, there is a @code{targetm_common} variable for hooks that
101 are shared between the compiler driver and the compilers proper,
102 documented as ``Common Target Hook''. This is declared in
103 @file{common/common-target.h}, the initializer
104 @code{TARGETM_COMMON_INITIALIZER} in
105 @file{common/common-target-def.h}. If targets initialize
106 @code{targetm_common} themselves, they should set
107 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
108 default definition is used.
111 @section Controlling the Compilation Driver, @file{gcc}
113 @cindex controlling the compilation driver
115 @c prevent bad page break with this line
116 You can control the compilation driver.
118 @defmac DRIVER_SELF_SPECS
119 A list of specs for the driver itself. It should be a suitable
120 initializer for an array of strings, with no surrounding braces.
122 The driver applies these specs to its own command line between loading
123 default @file{specs} files (but not command-line specified ones) and
124 choosing the multilib directory or running any subcommands. It
125 applies them in the order given, so each spec can depend on the
126 options added by earlier ones. It is also possible to remove options
127 using @samp{%<@var{option}} in the usual way.
129 This macro can be useful when a port has several interdependent target
130 options. It provides a way of standardizing the command line so
131 that the other specs are easier to write.
133 Do not define this macro if it does not need to do anything.
136 @defmac OPTION_DEFAULT_SPECS
137 A list of specs used to support configure-time default options (i.e.@:
138 @option{--with} options) in the driver. It should be a suitable initializer
139 for an array of structures, each containing two strings, without the
140 outermost pair of surrounding braces.
142 The first item in the pair is the name of the default. This must match
143 the code in @file{config.gcc} for the target. The second item is a spec
144 to apply if a default with this name was specified. The string
145 @samp{%(VALUE)} in the spec will be replaced by the value of the default
146 everywhere it occurs.
148 The driver will apply these specs to its own command line between loading
149 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
150 the same mechanism as @code{DRIVER_SELF_SPECS}.
152 Do not define this macro if it does not need to do anything.
156 A C string constant that tells the GCC driver program options to
157 pass to CPP@. It can also specify how to translate options you
158 give to GCC into options for GCC to pass to the CPP@.
160 Do not define this macro if it does not need to do anything.
163 @defmac CPLUSPLUS_CPP_SPEC
164 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
165 than C@. If you do not define this macro, then the value of
166 @code{CPP_SPEC} (if any) will be used instead.
170 A C string constant that tells the GCC driver program options to
171 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
173 It can also specify how to translate options you give to GCC into options
174 for GCC to pass to front ends.
176 Do not define this macro if it does not need to do anything.
180 A C string constant that tells the GCC driver program options to
181 pass to @code{cc1plus}. It can also specify how to translate options you
182 give to GCC into options for GCC to pass to the @code{cc1plus}.
184 Do not define this macro if it does not need to do anything.
185 Note that everything defined in CC1_SPEC is already passed to
186 @code{cc1plus} so there is no need to duplicate the contents of
187 CC1_SPEC in CC1PLUS_SPEC@.
191 A C string constant that tells the GCC driver program options to
192 pass to the assembler. It can also specify how to translate options
193 you give to GCC into options for GCC to pass to the assembler.
194 See the file @file{sun3.h} for an example of this.
196 Do not define this macro if it does not need to do anything.
199 @defmac ASM_FINAL_SPEC
200 A C string constant that tells the GCC driver program how to
201 run any programs which cleanup after the normal assembler.
202 Normally, this is not needed. See the file @file{mips.h} for
205 Do not define this macro if it does not need to do anything.
208 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
209 Define this macro, with no value, if the driver should give the assembler
210 an argument consisting of a single dash, @option{-}, to instruct it to
211 read from its standard input (which will be a pipe connected to the
212 output of the compiler proper). This argument is given after any
213 @option{-o} option specifying the name of the output file.
215 If you do not define this macro, the assembler is assumed to read its
216 standard input if given no non-option arguments. If your assembler
217 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
218 see @file{mips.h} for instance.
222 A C string constant that tells the GCC driver program options to
223 pass to the linker. It can also specify how to translate options you
224 give to GCC into options for GCC to pass to the linker.
226 Do not define this macro if it does not need to do anything.
230 Another C string constant used much like @code{LINK_SPEC}. The difference
231 between the two is that @code{LIB_SPEC} is used at the end of the
232 command given to the linker.
234 If this macro is not defined, a default is provided that
235 loads the standard C library from the usual place. See @file{gcc.c}.
239 Another C string constant that tells the GCC driver program
240 how and when to place a reference to @file{libgcc.a} into the
241 linker command line. This constant is placed both before and after
242 the value of @code{LIB_SPEC}.
244 If this macro is not defined, the GCC driver provides a default that
245 passes the string @option{-lgcc} to the linker.
248 @defmac REAL_LIBGCC_SPEC
249 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
250 @code{LIBGCC_SPEC} is not directly used by the driver program but is
251 instead modified to refer to different versions of @file{libgcc.a}
252 depending on the values of the command line flags @option{-static},
253 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
254 targets where these modifications are inappropriate, define
255 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
256 driver how to place a reference to @file{libgcc} on the link command
257 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
260 @defmac USE_LD_AS_NEEDED
261 A macro that controls the modifications to @code{LIBGCC_SPEC}
262 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
263 generated that uses @option{--as-needed} or equivalent options and the
264 shared @file{libgcc} in place of the
265 static exception handler library, when linking without any of
266 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
270 If defined, this C string constant is added to @code{LINK_SPEC}.
271 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
272 the modifications to @code{LIBGCC_SPEC} mentioned in
273 @code{REAL_LIBGCC_SPEC}.
276 @defmac STARTFILE_SPEC
277 Another C string constant used much like @code{LINK_SPEC}. The
278 difference between the two is that @code{STARTFILE_SPEC} is used at
279 the very beginning of the command given to the linker.
281 If this macro is not defined, a default is provided that loads the
282 standard C startup file from the usual place. See @file{gcc.c}.
286 Another C string constant used much like @code{LINK_SPEC}. The
287 difference between the two is that @code{ENDFILE_SPEC} is used at
288 the very end of the command given to the linker.
290 Do not define this macro if it does not need to do anything.
293 @defmac THREAD_MODEL_SPEC
294 GCC @code{-v} will print the thread model GCC was configured to use.
295 However, this doesn't work on platforms that are multilibbed on thread
296 models, such as AIX 4.3. On such platforms, define
297 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
298 blanks that names one of the recognized thread models. @code{%*}, the
299 default value of this macro, will expand to the value of
300 @code{thread_file} set in @file{config.gcc}.
303 @defmac SYSROOT_SUFFIX_SPEC
304 Define this macro to add a suffix to the target sysroot when GCC is
305 configured with a sysroot. This will cause GCC to search for usr/lib,
306 et al, within sysroot+suffix.
309 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
310 Define this macro to add a headers_suffix to the target sysroot when
311 GCC is configured with a sysroot. This will cause GCC to pass the
312 updated sysroot+headers_suffix to CPP, causing it to search for
313 usr/include, et al, within sysroot+headers_suffix.
317 Define this macro to provide additional specifications to put in the
318 @file{specs} file that can be used in various specifications like
321 The definition should be an initializer for an array of structures,
322 containing a string constant, that defines the specification name, and a
323 string constant that provides the specification.
325 Do not define this macro if it does not need to do anything.
327 @code{EXTRA_SPECS} is useful when an architecture contains several
328 related targets, which have various @code{@dots{}_SPECS} which are similar
329 to each other, and the maintainer would like one central place to keep
332 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
333 define either @code{_CALL_SYSV} when the System V calling sequence is
334 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
337 The @file{config/rs6000/rs6000.h} target file defines:
340 #define EXTRA_SPECS \
341 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
343 #define CPP_SYS_DEFAULT ""
346 The @file{config/rs6000/sysv.h} target file defines:
350 "%@{posix: -D_POSIX_SOURCE @} \
351 %@{mcall-sysv: -D_CALL_SYSV @} \
352 %@{!mcall-sysv: %(cpp_sysv_default) @} \
353 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
355 #undef CPP_SYSV_DEFAULT
356 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
359 while the @file{config/rs6000/eabiaix.h} target file defines
360 @code{CPP_SYSV_DEFAULT} as:
363 #undef CPP_SYSV_DEFAULT
364 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
368 @defmac LINK_LIBGCC_SPECIAL_1
369 Define this macro if the driver program should find the library
370 @file{libgcc.a}. If you do not define this macro, the driver program will pass
371 the argument @option{-lgcc} to tell the linker to do the search.
374 @defmac LINK_GCC_C_SEQUENCE_SPEC
375 The sequence in which libgcc and libc are specified to the linker.
376 By default this is @code{%G %L %G}.
379 @defmac LINK_COMMAND_SPEC
380 A C string constant giving the complete command line need to execute the
381 linker. When you do this, you will need to update your port each time a
382 change is made to the link command line within @file{gcc.c}. Therefore,
383 define this macro only if you need to completely redefine the command
384 line for invoking the linker and there is no other way to accomplish
385 the effect you need. Overriding this macro may be avoidable by overriding
386 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
389 @hook TARGET_ALWAYS_STRIP_DOTDOT
391 @defmac MULTILIB_DEFAULTS
392 Define this macro as a C expression for the initializer of an array of
393 string to tell the driver program which options are defaults for this
394 target and thus do not need to be handled specially when using
395 @code{MULTILIB_OPTIONS}.
397 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
398 the target makefile fragment or if none of the options listed in
399 @code{MULTILIB_OPTIONS} are set by default.
400 @xref{Target Fragment}.
403 @defmac RELATIVE_PREFIX_NOT_LINKDIR
404 Define this macro to tell @command{gcc} that it should only translate
405 a @option{-B} prefix into a @option{-L} linker option if the prefix
406 indicates an absolute file name.
409 @defmac MD_EXEC_PREFIX
410 If defined, this macro is an additional prefix to try after
411 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
412 when the compiler is built as a cross
413 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
414 to the list of directories used to find the assembler in @file{configure.in}.
417 @defmac STANDARD_STARTFILE_PREFIX
418 Define this macro as a C string constant if you wish to override the
419 standard choice of @code{libdir} as the default prefix to
420 try when searching for startup files such as @file{crt0.o}.
421 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
422 is built as a cross compiler.
425 @defmac STANDARD_STARTFILE_PREFIX_1
426 Define this macro as a C string constant if you wish to override the
427 standard choice of @code{/lib} as a prefix to try after the default prefix
428 when searching for startup files such as @file{crt0.o}.
429 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
430 is built as a cross compiler.
433 @defmac STANDARD_STARTFILE_PREFIX_2
434 Define this macro as a C string constant if you wish to override the
435 standard choice of @code{/lib} as yet another prefix to try after the
436 default prefix when searching for startup files such as @file{crt0.o}.
437 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
438 is built as a cross compiler.
441 @defmac MD_STARTFILE_PREFIX
442 If defined, this macro supplies an additional prefix to try after the
443 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
444 compiler is built as a cross compiler.
447 @defmac MD_STARTFILE_PREFIX_1
448 If defined, this macro supplies yet another prefix to try after the
449 standard prefixes. It is not searched when the compiler is built as a
453 @defmac INIT_ENVIRONMENT
454 Define this macro as a C string constant if you wish to set environment
455 variables for programs called by the driver, such as the assembler and
456 loader. The driver passes the value of this macro to @code{putenv} to
457 initialize the necessary environment variables.
460 @defmac LOCAL_INCLUDE_DIR
461 Define this macro as a C string constant if you wish to override the
462 standard choice of @file{/usr/local/include} as the default prefix to
463 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
464 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
465 @file{config.gcc}, normally @file{/usr/include}) in the search order.
467 Cross compilers do not search either @file{/usr/local/include} or its
471 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
472 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
473 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
474 If you do not define this macro, no component is used.
477 @defmac INCLUDE_DEFAULTS
478 Define this macro if you wish to override the entire default search path
479 for include files. For a native compiler, the default search path
480 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
481 @code{GPLUSPLUS_INCLUDE_DIR}, and
482 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
483 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
484 and specify private search areas for GCC@. The directory
485 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
487 The definition should be an initializer for an array of structures.
488 Each array element should have four elements: the directory name (a
489 string constant), the component name (also a string constant), a flag
490 for C++-only directories,
491 and a flag showing that the includes in the directory don't need to be
492 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
493 the array with a null element.
495 The component name denotes what GNU package the include file is part of,
496 if any, in all uppercase letters. For example, it might be @samp{GCC}
497 or @samp{BINUTILS}. If the package is part of a vendor-supplied
498 operating system, code the component name as @samp{0}.
500 For example, here is the definition used for VAX/VMS:
503 #define INCLUDE_DEFAULTS \
505 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
506 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
507 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
514 Here is the order of prefixes tried for exec files:
518 Any prefixes specified by the user with @option{-B}.
521 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
522 is not set and the compiler has not been installed in the configure-time
523 @var{prefix}, the location in which the compiler has actually been installed.
526 The directories specified by the environment variable @code{COMPILER_PATH}.
529 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
530 in the configured-time @var{prefix}.
533 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
536 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
539 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
543 Here is the order of prefixes tried for startfiles:
547 Any prefixes specified by the user with @option{-B}.
550 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
551 value based on the installed toolchain location.
554 The directories specified by the environment variable @code{LIBRARY_PATH}
555 (or port-specific name; native only, cross compilers do not use this).
558 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
559 in the configured @var{prefix} or this is a native compiler.
562 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
565 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
569 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
570 native compiler, or we have a target system root.
573 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
574 native compiler, or we have a target system root.
577 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
578 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
579 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
582 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
583 compiler, or we have a target system root. The default for this macro is
587 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
588 compiler, or we have a target system root. The default for this macro is
592 @node Run-time Target
593 @section Run-time Target Specification
594 @cindex run-time target specification
595 @cindex predefined macros
596 @cindex target specifications
598 @c prevent bad page break with this line
599 Here are run-time target specifications.
601 @defmac TARGET_CPU_CPP_BUILTINS ()
602 This function-like macro expands to a block of code that defines
603 built-in preprocessor macros and assertions for the target CPU, using
604 the functions @code{builtin_define}, @code{builtin_define_std} and
605 @code{builtin_assert}. When the front end
606 calls this macro it provides a trailing semicolon, and since it has
607 finished command line option processing your code can use those
610 @code{builtin_assert} takes a string in the form you pass to the
611 command-line option @option{-A}, such as @code{cpu=mips}, and creates
612 the assertion. @code{builtin_define} takes a string in the form
613 accepted by option @option{-D} and unconditionally defines the macro.
615 @code{builtin_define_std} takes a string representing the name of an
616 object-like macro. If it doesn't lie in the user's namespace,
617 @code{builtin_define_std} defines it unconditionally. Otherwise, it
618 defines a version with two leading underscores, and another version
619 with two leading and trailing underscores, and defines the original
620 only if an ISO standard was not requested on the command line. For
621 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
622 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
623 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
624 defines only @code{_ABI64}.
626 You can also test for the C dialect being compiled. The variable
627 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
628 or @code{clk_objective_c}. Note that if we are preprocessing
629 assembler, this variable will be @code{clk_c} but the function-like
630 macro @code{preprocessing_asm_p()} will return true, so you might want
631 to check for that first. If you need to check for strict ANSI, the
632 variable @code{flag_iso} can be used. The function-like macro
633 @code{preprocessing_trad_p()} can be used to check for traditional
637 @defmac TARGET_OS_CPP_BUILTINS ()
638 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
639 and is used for the target operating system instead.
642 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
643 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
644 and is used for the target object format. @file{elfos.h} uses this
645 macro to define @code{__ELF__}, so you probably do not need to define
649 @deftypevar {extern int} target_flags
650 This variable is declared in @file{options.h}, which is included before
651 any target-specific headers.
654 @hook TARGET_DEFAULT_TARGET_FLAGS
655 This variable specifies the initial value of @code{target_flags}.
656 Its default setting is 0.
659 @cindex optional hardware or system features
660 @cindex features, optional, in system conventions
662 @hook TARGET_HANDLE_OPTION
663 This hook is called whenever the user specifies one of the
664 target-specific options described by the @file{.opt} definition files
665 (@pxref{Options}). It has the opportunity to do some option-specific
666 processing and should return true if the option is valid. The default
667 definition does nothing but return true.
669 @var{decoded} specifies the option and its arguments. @var{opts} and
670 @var{opts_set} are the @code{gcc_options} structures to be used for
671 storing option state, and @var{loc} is the location at which the
672 option was passed (@code{UNKNOWN_LOCATION} except for options passed
676 @hook TARGET_HANDLE_C_OPTION
677 This target hook is called whenever the user specifies one of the
678 target-specific C language family options described by the @file{.opt}
679 definition files(@pxref{Options}). It has the opportunity to do some
680 option-specific processing and should return true if the option is
681 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
682 default definition does nothing but return false.
684 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
685 options. However, if processing an option requires routines that are
686 only available in the C (and related language) front ends, then you
687 should use @code{TARGET_HANDLE_C_OPTION} instead.
690 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
692 @hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE
694 @hook TARGET_OBJC_DECLARE_CLASS_DEFINITION
696 @hook TARGET_STRING_OBJECT_REF_TYPE_P
698 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
700 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
702 @defmac C_COMMON_OVERRIDE_OPTIONS
703 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
704 but is only used in the C
705 language frontends (C, Objective-C, C++, Objective-C++) and so can be
706 used to alter option flag variables which only exist in those
710 @hook TARGET_OPTION_OPTIMIZATION_TABLE
711 Some machines may desire to change what optimizations are performed for
712 various optimization levels. This variable, if defined, describes
713 options to enable at particular sets of optimization levels. These
714 options are processed once
715 just after the optimization level is determined and before the remainder
716 of the command options have been parsed, so may be overridden by other
717 options passed explicitly.
719 This processing is run once at program startup and when the optimization
720 options are changed via @code{#pragma GCC optimize} or by using the
721 @code{optimize} attribute.
724 @hook TARGET_OPTION_INIT_STRUCT
726 @hook TARGET_OPTION_DEFAULT_PARAMS
728 @defmac SWITCHABLE_TARGET
729 Some targets need to switch between substantially different subtargets
730 during compilation. For example, the MIPS target has one subtarget for
731 the traditional MIPS architecture and another for MIPS16. Source code
732 can switch between these two subarchitectures using the @code{mips16}
733 and @code{nomips16} attributes.
735 Such subtargets can differ in things like the set of available
736 registers, the set of available instructions, the costs of various
737 operations, and so on. GCC caches a lot of this type of information
738 in global variables, and recomputing them for each subtarget takes a
739 significant amount of time. The compiler therefore provides a facility
740 for maintaining several versions of the global variables and quickly
741 switching between them; see @file{target-globals.h} for details.
743 Define this macro to 1 if your target needs this facility. The default
747 @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
749 @node Per-Function Data
750 @section Defining data structures for per-function information.
751 @cindex per-function data
752 @cindex data structures
754 If the target needs to store information on a per-function basis, GCC
755 provides a macro and a couple of variables to allow this. Note, just
756 using statics to store the information is a bad idea, since GCC supports
757 nested functions, so you can be halfway through encoding one function
758 when another one comes along.
760 GCC defines a data structure called @code{struct function} which
761 contains all of the data specific to an individual function. This
762 structure contains a field called @code{machine} whose type is
763 @code{struct machine_function *}, which can be used by targets to point
764 to their own specific data.
766 If a target needs per-function specific data it should define the type
767 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
768 This macro should be used to initialize the function pointer
769 @code{init_machine_status}. This pointer is explained below.
771 One typical use of per-function, target specific data is to create an
772 RTX to hold the register containing the function's return address. This
773 RTX can then be used to implement the @code{__builtin_return_address}
774 function, for level 0.
776 Note---earlier implementations of GCC used a single data area to hold
777 all of the per-function information. Thus when processing of a nested
778 function began the old per-function data had to be pushed onto a
779 stack, and when the processing was finished, it had to be popped off the
780 stack. GCC used to provide function pointers called
781 @code{save_machine_status} and @code{restore_machine_status} to handle
782 the saving and restoring of the target specific information. Since the
783 single data area approach is no longer used, these pointers are no
786 @defmac INIT_EXPANDERS
787 Macro called to initialize any target specific information. This macro
788 is called once per function, before generation of any RTL has begun.
789 The intention of this macro is to allow the initialization of the
790 function pointer @code{init_machine_status}.
793 @deftypevar {void (*)(struct function *)} init_machine_status
794 If this function pointer is non-@code{NULL} it will be called once per
795 function, before function compilation starts, in order to allow the
796 target to perform any target specific initialization of the
797 @code{struct function} structure. It is intended that this would be
798 used to initialize the @code{machine} of that structure.
800 @code{struct machine_function} structures are expected to be freed by GC@.
801 Generally, any memory that they reference must be allocated by using
802 GC allocation, including the structure itself.
806 @section Storage Layout
807 @cindex storage layout
809 Note that the definitions of the macros in this table which are sizes or
810 alignments measured in bits do not need to be constant. They can be C
811 expressions that refer to static variables, such as the @code{target_flags}.
812 @xref{Run-time Target}.
814 @defmac BITS_BIG_ENDIAN
815 Define this macro to have the value 1 if the most significant bit in a
816 byte has the lowest number; otherwise define it to have the value zero.
817 This means that bit-field instructions count from the most significant
818 bit. If the machine has no bit-field instructions, then this must still
819 be defined, but it doesn't matter which value it is defined to. This
820 macro need not be a constant.
822 This macro does not affect the way structure fields are packed into
823 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
826 @defmac BYTES_BIG_ENDIAN
827 Define this macro to have the value 1 if the most significant byte in a
828 word has the lowest number. This macro need not be a constant.
831 @defmac WORDS_BIG_ENDIAN
832 Define this macro to have the value 1 if, in a multiword object, the
833 most significant word has the lowest number. This applies to both
834 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
835 order of words in memory is not the same as the order in registers. This
836 macro need not be a constant.
839 @defmac REG_WORDS_BIG_ENDIAN
840 On some machines, the order of words in a multiword object differs between
841 registers in memory. In such a situation, define this macro to describe
842 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
843 the order of words in memory.
846 @defmac FLOAT_WORDS_BIG_ENDIAN
847 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
848 @code{TFmode} floating point numbers are stored in memory with the word
849 containing the sign bit at the lowest address; otherwise define it to
850 have the value 0. This macro need not be a constant.
852 You need not define this macro if the ordering is the same as for
856 @defmac BITS_PER_UNIT
857 Define this macro to be the number of bits in an addressable storage
858 unit (byte). If you do not define this macro the default is 8.
861 @defmac BITS_PER_WORD
862 Number of bits in a word. If you do not define this macro, the default
863 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
866 @defmac MAX_BITS_PER_WORD
867 Maximum number of bits in a word. If this is undefined, the default is
868 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
869 largest value that @code{BITS_PER_WORD} can have at run-time.
872 @defmac UNITS_PER_WORD
873 Number of storage units in a word; normally the size of a general-purpose
874 register, a power of two from 1 or 8.
877 @defmac MIN_UNITS_PER_WORD
878 Minimum number of units in a word. If this is undefined, the default is
879 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
880 smallest value that @code{UNITS_PER_WORD} can have at run-time.
884 Width of a pointer, in bits. You must specify a value no wider than the
885 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
886 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
887 a value the default is @code{BITS_PER_WORD}.
890 @defmac POINTERS_EXTEND_UNSIGNED
891 A C expression that determines how pointers should be extended from
892 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
893 greater than zero if pointers should be zero-extended, zero if they
894 should be sign-extended, and negative if some other sort of conversion
895 is needed. In the last case, the extension is done by the target's
896 @code{ptr_extend} instruction.
898 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
899 and @code{word_mode} are all the same width.
902 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
903 A macro to update @var{m} and @var{unsignedp} when an object whose type
904 is @var{type} and which has the specified mode and signedness is to be
905 stored in a register. This macro is only called when @var{type} is a
908 On most RISC machines, which only have operations that operate on a full
909 register, define this macro to set @var{m} to @code{word_mode} if
910 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
911 cases, only integer modes should be widened because wider-precision
912 floating-point operations are usually more expensive than their narrower
915 For most machines, the macro definition does not change @var{unsignedp}.
916 However, some machines, have instructions that preferentially handle
917 either signed or unsigned quantities of certain modes. For example, on
918 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
919 sign-extend the result to 64 bits. On such machines, set
920 @var{unsignedp} according to which kind of extension is more efficient.
922 Do not define this macro if it would never modify @var{m}.
925 @hook TARGET_PROMOTE_FUNCTION_MODE
927 @defmac PARM_BOUNDARY
928 Normal alignment required for function parameters on the stack, in
929 bits. All stack parameters receive at least this much alignment
930 regardless of data type. On most machines, this is the same as the
934 @defmac STACK_BOUNDARY
935 Define this macro to the minimum alignment enforced by hardware for the
936 stack pointer on this machine. The definition is a C expression for the
937 desired alignment (measured in bits). This value is used as a default
938 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
939 this should be the same as @code{PARM_BOUNDARY}.
942 @defmac PREFERRED_STACK_BOUNDARY
943 Define this macro if you wish to preserve a certain alignment for the
944 stack pointer, greater than what the hardware enforces. The definition
945 is a C expression for the desired alignment (measured in bits). This
946 macro must evaluate to a value equal to or larger than
947 @code{STACK_BOUNDARY}.
950 @defmac INCOMING_STACK_BOUNDARY
951 Define this macro if the incoming stack boundary may be different
952 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
953 to a value equal to or larger than @code{STACK_BOUNDARY}.
956 @defmac FUNCTION_BOUNDARY
957 Alignment required for a function entry point, in bits.
960 @defmac BIGGEST_ALIGNMENT
961 Biggest alignment that any data type can require on this machine, in
962 bits. Note that this is not the biggest alignment that is supported,
963 just the biggest alignment that, when violated, may cause a fault.
966 @defmac MALLOC_ABI_ALIGNMENT
967 Alignment, in bits, a C conformant malloc implementation has to
968 provide. If not defined, the default value is @code{BITS_PER_WORD}.
971 @defmac ATTRIBUTE_ALIGNED_VALUE
972 Alignment used by the @code{__attribute__ ((aligned))} construct. If
973 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
976 @defmac MINIMUM_ATOMIC_ALIGNMENT
977 If defined, the smallest alignment, in bits, that can be given to an
978 object that can be referenced in one operation, without disturbing any
979 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
980 on machines that don't have byte or half-word store operations.
983 @defmac BIGGEST_FIELD_ALIGNMENT
984 Biggest alignment that any structure or union field can require on this
985 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
986 structure and union fields only, unless the field alignment has been set
987 by the @code{__attribute__ ((aligned (@var{n})))} construct.
990 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
991 An expression for the alignment of a structure field @var{field} if the
992 alignment computed in the usual way (including applying of
993 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
994 alignment) is @var{computed}. It overrides alignment only if the
995 field alignment has not been set by the
996 @code{__attribute__ ((aligned (@var{n})))} construct.
999 @defmac MAX_STACK_ALIGNMENT
1000 Biggest stack alignment guaranteed by the backend. Use this macro
1001 to specify the maximum alignment of a variable on stack.
1003 If not defined, the default value is @code{STACK_BOUNDARY}.
1005 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1006 @c But the fix for PR 32893 indicates that we can only guarantee
1007 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1008 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1011 @defmac MAX_OFILE_ALIGNMENT
1012 Biggest alignment supported by the object file format of this machine.
1013 Use this macro to limit the alignment which can be specified using the
1014 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1015 the default value is @code{BIGGEST_ALIGNMENT}.
1017 On systems that use ELF, the default (in @file{config/elfos.h}) is
1018 the largest supported 32-bit ELF section alignment representable on
1019 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1020 On 32-bit ELF the largest supported section alignment in bits is
1021 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1024 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1025 If defined, a C expression to compute the alignment for a variable in
1026 the static store. @var{type} is the data type, and @var{basic-align} is
1027 the alignment that the object would ordinarily have. The value of this
1028 macro is used instead of that alignment to align the object.
1030 If this macro is not defined, then @var{basic-align} is used.
1033 One use of this macro is to increase alignment of medium-size data to
1034 make it all fit in fewer cache lines. Another is to cause character
1035 arrays to be word-aligned so that @code{strcpy} calls that copy
1036 constants to character arrays can be done inline.
1039 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1040 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1041 some alignment increase, instead of optimization only purposes. E.g.@
1042 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1043 must be aligned to 16 byte boundaries.
1045 If this macro is not defined, then @var{basic-align} is used.
1048 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1049 If defined, a C expression to compute the alignment given to a constant
1050 that is being placed in memory. @var{constant} is the constant and
1051 @var{basic-align} is the alignment that the object would ordinarily
1052 have. The value of this macro is used instead of that alignment to
1055 If this macro is not defined, then @var{basic-align} is used.
1057 The typical use of this macro is to increase alignment for string
1058 constants to be word aligned so that @code{strcpy} calls that copy
1059 constants can be done inline.
1062 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1063 If defined, a C expression to compute the alignment for a variable in
1064 the local store. @var{type} is the data type, and @var{basic-align} is
1065 the alignment that the object would ordinarily have. The value of this
1066 macro is used instead of that alignment to align the object.
1068 If this macro is not defined, then @var{basic-align} is used.
1070 One use of this macro is to increase alignment of medium-size data to
1071 make it all fit in fewer cache lines.
1073 If the value of this macro has a type, it should be an unsigned type.
1076 @hook TARGET_VECTOR_ALIGNMENT
1078 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1079 If defined, a C expression to compute the alignment for stack slot.
1080 @var{type} is the data type, @var{mode} is the widest mode available,
1081 and @var{basic-align} is the alignment that the slot would ordinarily
1082 have. The value of this macro is used instead of that alignment to
1085 If this macro is not defined, then @var{basic-align} is used when
1086 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1089 This macro is to set alignment of stack slot to the maximum alignment
1090 of all possible modes which the slot may have.
1092 If the value of this macro has a type, it should be an unsigned type.
1095 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1096 If defined, a C expression to compute the alignment for a local
1097 variable @var{decl}.
1099 If this macro is not defined, then
1100 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1103 One use of this macro is to increase alignment of medium-size data to
1104 make it all fit in fewer cache lines.
1106 If the value of this macro has a type, it should be an unsigned type.
1109 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1110 If defined, a C expression to compute the minimum required alignment
1111 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1112 @var{mode}, assuming normal alignment @var{align}.
1114 If this macro is not defined, then @var{align} will be used.
1117 @defmac EMPTY_FIELD_BOUNDARY
1118 Alignment in bits to be given to a structure bit-field that follows an
1119 empty field such as @code{int : 0;}.
1121 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1124 @defmac STRUCTURE_SIZE_BOUNDARY
1125 Number of bits which any structure or union's size must be a multiple of.
1126 Each structure or union's size is rounded up to a multiple of this.
1128 If you do not define this macro, the default is the same as
1129 @code{BITS_PER_UNIT}.
1132 @defmac STRICT_ALIGNMENT
1133 Define this macro to be the value 1 if instructions will fail to work
1134 if given data not on the nominal alignment. If instructions will merely
1135 go slower in that case, define this macro as 0.
1138 @defmac PCC_BITFIELD_TYPE_MATTERS
1139 Define this if you wish to imitate the way many other C compilers handle
1140 alignment of bit-fields and the structures that contain them.
1142 The behavior is that the type written for a named bit-field (@code{int},
1143 @code{short}, or other integer type) imposes an alignment for the entire
1144 structure, as if the structure really did contain an ordinary field of
1145 that type. In addition, the bit-field is placed within the structure so
1146 that it would fit within such a field, not crossing a boundary for it.
1148 Thus, on most machines, a named bit-field whose type is written as
1149 @code{int} would not cross a four-byte boundary, and would force
1150 four-byte alignment for the whole structure. (The alignment used may
1151 not be four bytes; it is controlled by the other alignment parameters.)
1153 An unnamed bit-field will not affect the alignment of the containing
1156 If the macro is defined, its definition should be a C expression;
1157 a nonzero value for the expression enables this behavior.
1159 Note that if this macro is not defined, or its value is zero, some
1160 bit-fields may cross more than one alignment boundary. The compiler can
1161 support such references if there are @samp{insv}, @samp{extv}, and
1162 @samp{extzv} insns that can directly reference memory.
1164 The other known way of making bit-fields work is to define
1165 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1166 Then every structure can be accessed with fullwords.
1168 Unless the machine has bit-field instructions or you define
1169 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1170 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1172 If your aim is to make GCC use the same conventions for laying out
1173 bit-fields as are used by another compiler, here is how to investigate
1174 what the other compiler does. Compile and run this program:
1193 printf ("Size of foo1 is %d\n",
1194 sizeof (struct foo1));
1195 printf ("Size of foo2 is %d\n",
1196 sizeof (struct foo2));
1201 If this prints 2 and 5, then the compiler's behavior is what you would
1202 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1205 @defmac BITFIELD_NBYTES_LIMITED
1206 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1207 to aligning a bit-field within the structure.
1210 @hook TARGET_ALIGN_ANON_BITFIELD
1212 @hook TARGET_NARROW_VOLATILE_BITFIELD
1214 @hook TARGET_MEMBER_TYPE_FORCES_BLK
1216 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1217 Define this macro as an expression for the alignment of a type (given
1218 by @var{type} as a tree node) if the alignment computed in the usual
1219 way is @var{computed} and the alignment explicitly specified was
1222 The default is to use @var{specified} if it is larger; otherwise, use
1223 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1226 @defmac MAX_FIXED_MODE_SIZE
1227 An integer expression for the size in bits of the largest integer
1228 machine mode that should actually be used. All integer machine modes of
1229 this size or smaller can be used for structures and unions with the
1230 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1231 (DImode)} is assumed.
1234 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1235 If defined, an expression of type @code{enum machine_mode} that
1236 specifies the mode of the save area operand of a
1237 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1238 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1239 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1240 having its mode specified.
1242 You need not define this macro if it always returns @code{Pmode}. You
1243 would most commonly define this macro if the
1244 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1248 @defmac STACK_SIZE_MODE
1249 If defined, an expression of type @code{enum machine_mode} that
1250 specifies the mode of the size increment operand of an
1251 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1253 You need not define this macro if it always returns @code{word_mode}.
1254 You would most commonly define this macro if the @code{allocate_stack}
1255 pattern needs to support both a 32- and a 64-bit mode.
1258 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1260 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1262 @hook TARGET_UNWIND_WORD_MODE
1264 @defmac ROUND_TOWARDS_ZERO
1265 If defined, this macro should be true if the prevailing rounding
1266 mode is towards zero.
1268 Defining this macro only affects the way @file{libgcc.a} emulates
1269 floating-point arithmetic.
1271 Not defining this macro is equivalent to returning zero.
1274 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1275 This macro should return true if floats with @var{size}
1276 bits do not have a NaN or infinity representation, but use the largest
1277 exponent for normal numbers instead.
1279 Defining this macro only affects the way @file{libgcc.a} emulates
1280 floating-point arithmetic.
1282 The default definition of this macro returns false for all sizes.
1285 @hook TARGET_MS_BITFIELD_LAYOUT_P
1287 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1289 @hook TARGET_FIXED_POINT_SUPPORTED_P
1291 @hook TARGET_EXPAND_TO_RTL_HOOK
1293 @hook TARGET_INSTANTIATE_DECLS
1295 @hook TARGET_MANGLE_TYPE
1298 @section Layout of Source Language Data Types
1300 These macros define the sizes and other characteristics of the standard
1301 basic data types used in programs being compiled. Unlike the macros in
1302 the previous section, these apply to specific features of C and related
1303 languages, rather than to fundamental aspects of storage layout.
1305 @defmac INT_TYPE_SIZE
1306 A C expression for the size in bits of the type @code{int} on the
1307 target machine. If you don't define this, the default is one word.
1310 @defmac SHORT_TYPE_SIZE
1311 A C expression for the size in bits of the type @code{short} on the
1312 target machine. If you don't define this, the default is half a word.
1313 (If this would be less than one storage unit, it is rounded up to one
1317 @defmac LONG_TYPE_SIZE
1318 A C expression for the size in bits of the type @code{long} on the
1319 target machine. If you don't define this, the default is one word.
1322 @defmac ADA_LONG_TYPE_SIZE
1323 On some machines, the size used for the Ada equivalent of the type
1324 @code{long} by a native Ada compiler differs from that used by C@. In
1325 that situation, define this macro to be a C expression to be used for
1326 the size of that type. If you don't define this, the default is the
1327 value of @code{LONG_TYPE_SIZE}.
1330 @defmac LONG_LONG_TYPE_SIZE
1331 A C expression for the size in bits of the type @code{long long} on the
1332 target machine. If you don't define this, the default is two
1333 words. If you want to support GNU Ada on your machine, the value of this
1334 macro must be at least 64.
1337 @defmac CHAR_TYPE_SIZE
1338 A C expression for the size in bits of the type @code{char} on the
1339 target machine. If you don't define this, the default is
1340 @code{BITS_PER_UNIT}.
1343 @defmac BOOL_TYPE_SIZE
1344 A C expression for the size in bits of the C++ type @code{bool} and
1345 C99 type @code{_Bool} on the target machine. If you don't define
1346 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1349 @defmac FLOAT_TYPE_SIZE
1350 A C expression for the size in bits of the type @code{float} on the
1351 target machine. If you don't define this, the default is one word.
1354 @defmac DOUBLE_TYPE_SIZE
1355 A C expression for the size in bits of the type @code{double} on the
1356 target machine. If you don't define this, the default is two
1360 @defmac LONG_DOUBLE_TYPE_SIZE
1361 A C expression for the size in bits of the type @code{long double} on
1362 the target machine. If you don't define this, the default is two
1366 @defmac SHORT_FRACT_TYPE_SIZE
1367 A C expression for the size in bits of the type @code{short _Fract} on
1368 the target machine. If you don't define this, the default is
1369 @code{BITS_PER_UNIT}.
1372 @defmac FRACT_TYPE_SIZE
1373 A C expression for the size in bits of the type @code{_Fract} on
1374 the target machine. If you don't define this, the default is
1375 @code{BITS_PER_UNIT * 2}.
1378 @defmac LONG_FRACT_TYPE_SIZE
1379 A C expression for the size in bits of the type @code{long _Fract} on
1380 the target machine. If you don't define this, the default is
1381 @code{BITS_PER_UNIT * 4}.
1384 @defmac LONG_LONG_FRACT_TYPE_SIZE
1385 A C expression for the size in bits of the type @code{long long _Fract} on
1386 the target machine. If you don't define this, the default is
1387 @code{BITS_PER_UNIT * 8}.
1390 @defmac SHORT_ACCUM_TYPE_SIZE
1391 A C expression for the size in bits of the type @code{short _Accum} on
1392 the target machine. If you don't define this, the default is
1393 @code{BITS_PER_UNIT * 2}.
1396 @defmac ACCUM_TYPE_SIZE
1397 A C expression for the size in bits of the type @code{_Accum} on
1398 the target machine. If you don't define this, the default is
1399 @code{BITS_PER_UNIT * 4}.
1402 @defmac LONG_ACCUM_TYPE_SIZE
1403 A C expression for the size in bits of the type @code{long _Accum} on
1404 the target machine. If you don't define this, the default is
1405 @code{BITS_PER_UNIT * 8}.
1408 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1409 A C expression for the size in bits of the type @code{long long _Accum} on
1410 the target machine. If you don't define this, the default is
1411 @code{BITS_PER_UNIT * 16}.
1414 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1415 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1416 if you want routines in @file{libgcc2.a} for a size other than
1417 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1418 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1421 @defmac LIBGCC2_HAS_DF_MODE
1422 Define this macro if neither @code{DOUBLE_TYPE_SIZE} nor
1423 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1424 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1425 anyway. If you don't define this and either @code{DOUBLE_TYPE_SIZE}
1426 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1430 @defmac LIBGCC2_HAS_XF_MODE
1431 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1432 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1433 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1434 is 80 then the default is 1, otherwise it is 0.
1437 @defmac LIBGCC2_HAS_TF_MODE
1438 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1439 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1440 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1441 is 128 then the default is 1, otherwise it is 0.
1444 @defmac LIBGCC2_GNU_PREFIX
1445 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1446 hook and should be defined if that hook is overriden to be true. It
1447 causes function names in libgcc to be changed to use a @code{__gnu_}
1448 prefix for their name rather than the default @code{__}. A port which
1449 uses this macro should also arrange to use @file{t-gnu-prefix} in
1450 the libgcc @file{config.host}.
1457 Define these macros to be the size in bits of the mantissa of
1458 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1459 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1460 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1461 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1462 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1463 @code{DOUBLE_TYPE_SIZE} or
1464 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1467 @defmac TARGET_FLT_EVAL_METHOD
1468 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1469 assuming, if applicable, that the floating-point control word is in its
1470 default state. If you do not define this macro the value of
1471 @code{FLT_EVAL_METHOD} will be zero.
1474 @defmac WIDEST_HARDWARE_FP_SIZE
1475 A C expression for the size in bits of the widest floating-point format
1476 supported by the hardware. If you define this macro, you must specify a
1477 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1478 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1482 @defmac DEFAULT_SIGNED_CHAR
1483 An expression whose value is 1 or 0, according to whether the type
1484 @code{char} should be signed or unsigned by default. The user can
1485 always override this default with the options @option{-fsigned-char}
1486 and @option{-funsigned-char}.
1489 @hook TARGET_DEFAULT_SHORT_ENUMS
1492 A C expression for a string describing the name of the data type to use
1493 for size values. The typedef name @code{size_t} is defined using the
1494 contents of the string.
1496 The string can contain more than one keyword. If so, separate them with
1497 spaces, and write first any length keyword, then @code{unsigned} if
1498 appropriate, and finally @code{int}. The string must exactly match one
1499 of the data type names defined in the function
1500 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1501 You may not omit @code{int} or change the order---that would cause the
1502 compiler to crash on startup.
1504 If you don't define this macro, the default is @code{"long unsigned
1509 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1510 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1511 dealing with size. This macro is a C expression for a string describing
1512 the name of the data type from which the precision of @code{sizetype}
1515 The string has the same restrictions as @code{SIZE_TYPE} string.
1517 If you don't define this macro, the default is @code{SIZE_TYPE}.
1520 @defmac PTRDIFF_TYPE
1521 A C expression for a string describing the name of the data type to use
1522 for the result of subtracting two pointers. The typedef name
1523 @code{ptrdiff_t} is defined using the contents of the string. See
1524 @code{SIZE_TYPE} above for more information.
1526 If you don't define this macro, the default is @code{"long int"}.
1530 A C expression for a string describing the name of the data type to use
1531 for wide characters. The typedef name @code{wchar_t} is defined using
1532 the contents of the string. See @code{SIZE_TYPE} above for more
1535 If you don't define this macro, the default is @code{"int"}.
1538 @defmac WCHAR_TYPE_SIZE
1539 A C expression for the size in bits of the data type for wide
1540 characters. This is used in @code{cpp}, which cannot make use of
1545 A C expression for a string describing the name of the data type to
1546 use for wide characters passed to @code{printf} and returned from
1547 @code{getwc}. The typedef name @code{wint_t} is defined using the
1548 contents of the string. See @code{SIZE_TYPE} above for more
1551 If you don't define this macro, the default is @code{"unsigned int"}.
1555 A C expression for a string describing the name of the data type that
1556 can represent any value of any standard or extended signed integer type.
1557 The typedef name @code{intmax_t} is defined using the contents of the
1558 string. See @code{SIZE_TYPE} above for more information.
1560 If you don't define this macro, the default is the first of
1561 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1562 much precision as @code{long long int}.
1565 @defmac UINTMAX_TYPE
1566 A C expression for a string describing the name of the data type that
1567 can represent any value of any standard or extended unsigned integer
1568 type. The typedef name @code{uintmax_t} is defined using the contents
1569 of the string. See @code{SIZE_TYPE} above for more information.
1571 If you don't define this macro, the default is the first of
1572 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1573 unsigned int"} that has as much precision as @code{long long unsigned
1577 @defmac SIG_ATOMIC_TYPE
1583 @defmacx UINT16_TYPE
1584 @defmacx UINT32_TYPE
1585 @defmacx UINT64_TYPE
1586 @defmacx INT_LEAST8_TYPE
1587 @defmacx INT_LEAST16_TYPE
1588 @defmacx INT_LEAST32_TYPE
1589 @defmacx INT_LEAST64_TYPE
1590 @defmacx UINT_LEAST8_TYPE
1591 @defmacx UINT_LEAST16_TYPE
1592 @defmacx UINT_LEAST32_TYPE
1593 @defmacx UINT_LEAST64_TYPE
1594 @defmacx INT_FAST8_TYPE
1595 @defmacx INT_FAST16_TYPE
1596 @defmacx INT_FAST32_TYPE
1597 @defmacx INT_FAST64_TYPE
1598 @defmacx UINT_FAST8_TYPE
1599 @defmacx UINT_FAST16_TYPE
1600 @defmacx UINT_FAST32_TYPE
1601 @defmacx UINT_FAST64_TYPE
1602 @defmacx INTPTR_TYPE
1603 @defmacx UINTPTR_TYPE
1604 C expressions for the standard types @code{sig_atomic_t},
1605 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1606 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1607 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1608 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1609 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1610 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1611 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1612 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1613 @code{SIZE_TYPE} above for more information.
1615 If any of these macros evaluates to a null pointer, the corresponding
1616 type is not supported; if GCC is configured to provide
1617 @code{<stdint.h>} in such a case, the header provided may not conform
1618 to C99, depending on the type in question. The defaults for all of
1619 these macros are null pointers.
1622 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1623 The C++ compiler represents a pointer-to-member-function with a struct
1630 ptrdiff_t vtable_index;
1637 The C++ compiler must use one bit to indicate whether the function that
1638 will be called through a pointer-to-member-function is virtual.
1639 Normally, we assume that the low-order bit of a function pointer must
1640 always be zero. Then, by ensuring that the vtable_index is odd, we can
1641 distinguish which variant of the union is in use. But, on some
1642 platforms function pointers can be odd, and so this doesn't work. In
1643 that case, we use the low-order bit of the @code{delta} field, and shift
1644 the remainder of the @code{delta} field to the left.
1646 GCC will automatically make the right selection about where to store
1647 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1648 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1649 set such that functions always start at even addresses, but the lowest
1650 bit of pointers to functions indicate whether the function at that
1651 address is in ARM or Thumb mode. If this is the case of your
1652 architecture, you should define this macro to
1653 @code{ptrmemfunc_vbit_in_delta}.
1655 In general, you should not have to define this macro. On architectures
1656 in which function addresses are always even, according to
1657 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1658 @code{ptrmemfunc_vbit_in_pfn}.
1661 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1662 Normally, the C++ compiler uses function pointers in vtables. This
1663 macro allows the target to change to use ``function descriptors''
1664 instead. Function descriptors are found on targets for whom a
1665 function pointer is actually a small data structure. Normally the
1666 data structure consists of the actual code address plus a data
1667 pointer to which the function's data is relative.
1669 If vtables are used, the value of this macro should be the number
1670 of words that the function descriptor occupies.
1673 @defmac TARGET_VTABLE_ENTRY_ALIGN
1674 By default, the vtable entries are void pointers, the so the alignment
1675 is the same as pointer alignment. The value of this macro specifies
1676 the alignment of the vtable entry in bits. It should be defined only
1677 when special alignment is necessary. */
1680 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1681 There are a few non-descriptor entries in the vtable at offsets below
1682 zero. If these entries must be padded (say, to preserve the alignment
1683 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1684 of words in each data entry.
1688 @section Register Usage
1689 @cindex register usage
1691 This section explains how to describe what registers the target machine
1692 has, and how (in general) they can be used.
1694 The description of which registers a specific instruction can use is
1695 done with register classes; see @ref{Register Classes}. For information
1696 on using registers to access a stack frame, see @ref{Frame Registers}.
1697 For passing values in registers, see @ref{Register Arguments}.
1698 For returning values in registers, see @ref{Scalar Return}.
1701 * Register Basics:: Number and kinds of registers.
1702 * Allocation Order:: Order in which registers are allocated.
1703 * Values in Registers:: What kinds of values each reg can hold.
1704 * Leaf Functions:: Renumbering registers for leaf functions.
1705 * Stack Registers:: Handling a register stack such as 80387.
1708 @node Register Basics
1709 @subsection Basic Characteristics of Registers
1711 @c prevent bad page break with this line
1712 Registers have various characteristics.
1714 @defmac FIRST_PSEUDO_REGISTER
1715 Number of hardware registers known to the compiler. They receive
1716 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1717 pseudo register's number really is assigned the number
1718 @code{FIRST_PSEUDO_REGISTER}.
1721 @defmac FIXED_REGISTERS
1722 @cindex fixed register
1723 An initializer that says which registers are used for fixed purposes
1724 all throughout the compiled code and are therefore not available for
1725 general allocation. These would include the stack pointer, the frame
1726 pointer (except on machines where that can be used as a general
1727 register when no frame pointer is needed), the program counter on
1728 machines where that is considered one of the addressable registers,
1729 and any other numbered register with a standard use.
1731 This information is expressed as a sequence of numbers, separated by
1732 commas and surrounded by braces. The @var{n}th number is 1 if
1733 register @var{n} is fixed, 0 otherwise.
1735 The table initialized from this macro, and the table initialized by
1736 the following one, may be overridden at run time either automatically,
1737 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1738 the user with the command options @option{-ffixed-@var{reg}},
1739 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1742 @defmac CALL_USED_REGISTERS
1743 @cindex call-used register
1744 @cindex call-clobbered register
1745 @cindex call-saved register
1746 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1747 clobbered (in general) by function calls as well as for fixed
1748 registers. This macro therefore identifies the registers that are not
1749 available for general allocation of values that must live across
1752 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1753 automatically saves it on function entry and restores it on function
1754 exit, if the register is used within the function.
1757 @defmac CALL_REALLY_USED_REGISTERS
1758 @cindex call-used register
1759 @cindex call-clobbered register
1760 @cindex call-saved register
1761 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1762 that the entire set of @code{FIXED_REGISTERS} be included.
1763 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1764 This macro is optional. If not specified, it defaults to the value
1765 of @code{CALL_USED_REGISTERS}.
1768 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1769 @cindex call-used register
1770 @cindex call-clobbered register
1771 @cindex call-saved register
1772 A C expression that is nonzero if it is not permissible to store a
1773 value of mode @var{mode} in hard register number @var{regno} across a
1774 call without some part of it being clobbered. For most machines this
1775 macro need not be defined. It is only required for machines that do not
1776 preserve the entire contents of a register across a call.
1780 @findex call_used_regs
1783 @findex reg_class_contents
1784 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1786 @defmac INCOMING_REGNO (@var{out})
1787 Define this macro if the target machine has register windows. This C
1788 expression returns the register number as seen by the called function
1789 corresponding to the register number @var{out} as seen by the calling
1790 function. Return @var{out} if register number @var{out} is not an
1794 @defmac OUTGOING_REGNO (@var{in})
1795 Define this macro if the target machine has register windows. This C
1796 expression returns the register number as seen by the calling function
1797 corresponding to the register number @var{in} as seen by the called
1798 function. Return @var{in} if register number @var{in} is not an inbound
1802 @defmac LOCAL_REGNO (@var{regno})
1803 Define this macro if the target machine has register windows. This C
1804 expression returns true if the register is call-saved but is in the
1805 register window. Unlike most call-saved registers, such registers
1806 need not be explicitly restored on function exit or during non-local
1811 If the program counter has a register number, define this as that
1812 register number. Otherwise, do not define it.
1815 @node Allocation Order
1816 @subsection Order of Allocation of Registers
1817 @cindex order of register allocation
1818 @cindex register allocation order
1820 @c prevent bad page break with this line
1821 Registers are allocated in order.
1823 @defmac REG_ALLOC_ORDER
1824 If defined, an initializer for a vector of integers, containing the
1825 numbers of hard registers in the order in which GCC should prefer
1826 to use them (from most preferred to least).
1828 If this macro is not defined, registers are used lowest numbered first
1829 (all else being equal).
1831 One use of this macro is on machines where the highest numbered
1832 registers must always be saved and the save-multiple-registers
1833 instruction supports only sequences of consecutive registers. On such
1834 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1835 the highest numbered allocable register first.
1838 @defmac ADJUST_REG_ALLOC_ORDER
1839 A C statement (sans semicolon) to choose the order in which to allocate
1840 hard registers for pseudo-registers local to a basic block.
1842 Store the desired register order in the array @code{reg_alloc_order}.
1843 Element 0 should be the register to allocate first; element 1, the next
1844 register; and so on.
1846 The macro body should not assume anything about the contents of
1847 @code{reg_alloc_order} before execution of the macro.
1849 On most machines, it is not necessary to define this macro.
1852 @defmac HONOR_REG_ALLOC_ORDER
1853 Normally, IRA tries to estimate the costs for saving a register in the
1854 prologue and restoring it in the epilogue. This discourages it from
1855 using call-saved registers. If a machine wants to ensure that IRA
1856 allocates registers in the order given by REG_ALLOC_ORDER even if some
1857 call-saved registers appear earlier than call-used ones, this macro
1861 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1862 In some case register allocation order is not enough for the
1863 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1864 If this macro is defined, it should return a floating point value
1865 based on @var{regno}. The cost of using @var{regno} for a pseudo will
1866 be increased by approximately the pseudo's usage frequency times the
1867 value returned by this macro. Not defining this macro is equivalent
1868 to having it always return @code{0.0}.
1870 On most machines, it is not necessary to define this macro.
1873 @node Values in Registers
1874 @subsection How Values Fit in Registers
1876 This section discusses the macros that describe which kinds of values
1877 (specifically, which machine modes) each register can hold, and how many
1878 consecutive registers are needed for a given mode.
1880 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
1881 A C expression for the number of consecutive hard registers, starting
1882 at register number @var{regno}, required to hold a value of mode
1883 @var{mode}. This macro must never return zero, even if a register
1884 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
1885 and/or CANNOT_CHANGE_MODE_CLASS instead.
1887 On a machine where all registers are exactly one word, a suitable
1888 definition of this macro is
1891 #define HARD_REGNO_NREGS(REGNO, MODE) \
1892 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1897 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1898 A C expression that is nonzero if a value of mode @var{mode}, stored
1899 in memory, ends with padding that causes it to take up more space than
1900 in registers starting at register number @var{regno} (as determined by
1901 multiplying GCC's notion of the size of the register when containing
1902 this mode by the number of registers returned by
1903 @code{HARD_REGNO_NREGS}). By default this is zero.
1905 For example, if a floating-point value is stored in three 32-bit
1906 registers but takes up 128 bits in memory, then this would be
1909 This macros only needs to be defined if there are cases where
1910 @code{subreg_get_info}
1911 would otherwise wrongly determine that a @code{subreg} can be
1912 represented by an offset to the register number, when in fact such a
1913 @code{subreg} would contain some of the padding not stored in
1914 registers and so not be representable.
1917 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1918 For values of @var{regno} and @var{mode} for which
1919 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1920 returning the greater number of registers required to hold the value
1921 including any padding. In the example above, the value would be four.
1924 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1925 Define this macro if the natural size of registers that hold values
1926 of mode @var{mode} is not the word size. It is a C expression that
1927 should give the natural size in bytes for the specified mode. It is
1928 used by the register allocator to try to optimize its results. This
1929 happens for example on SPARC 64-bit where the natural size of
1930 floating-point registers is still 32-bit.
1933 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1934 A C expression that is nonzero if it is permissible to store a value
1935 of mode @var{mode} in hard register number @var{regno} (or in several
1936 registers starting with that one). For a machine where all registers
1937 are equivalent, a suitable definition is
1940 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1943 You need not include code to check for the numbers of fixed registers,
1944 because the allocation mechanism considers them to be always occupied.
1946 @cindex register pairs
1947 On some machines, double-precision values must be kept in even/odd
1948 register pairs. You can implement that by defining this macro to reject
1949 odd register numbers for such modes.
1951 The minimum requirement for a mode to be OK in a register is that the
1952 @samp{mov@var{mode}} instruction pattern support moves between the
1953 register and other hard register in the same class and that moving a
1954 value into the register and back out not alter it.
1956 Since the same instruction used to move @code{word_mode} will work for
1957 all narrower integer modes, it is not necessary on any machine for
1958 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1959 you define patterns @samp{movhi}, etc., to take advantage of this. This
1960 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1961 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1964 Many machines have special registers for floating point arithmetic.
1965 Often people assume that floating point machine modes are allowed only
1966 in floating point registers. This is not true. Any registers that
1967 can hold integers can safely @emph{hold} a floating point machine
1968 mode, whether or not floating arithmetic can be done on it in those
1969 registers. Integer move instructions can be used to move the values.
1971 On some machines, though, the converse is true: fixed-point machine
1972 modes may not go in floating registers. This is true if the floating
1973 registers normalize any value stored in them, because storing a
1974 non-floating value there would garble it. In this case,
1975 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1976 floating registers. But if the floating registers do not automatically
1977 normalize, if you can store any bit pattern in one and retrieve it
1978 unchanged without a trap, then any machine mode may go in a floating
1979 register, so you can define this macro to say so.
1981 The primary significance of special floating registers is rather that
1982 they are the registers acceptable in floating point arithmetic
1983 instructions. However, this is of no concern to
1984 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1985 constraints for those instructions.
1987 On some machines, the floating registers are especially slow to access,
1988 so that it is better to store a value in a stack frame than in such a
1989 register if floating point arithmetic is not being done. As long as the
1990 floating registers are not in class @code{GENERAL_REGS}, they will not
1991 be used unless some pattern's constraint asks for one.
1994 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1995 A C expression that is nonzero if it is OK to rename a hard register
1996 @var{from} to another hard register @var{to}.
1998 One common use of this macro is to prevent renaming of a register to
1999 another register that is not saved by a prologue in an interrupt
2002 The default is always nonzero.
2005 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2006 A C expression that is nonzero if a value of mode
2007 @var{mode1} is accessible in mode @var{mode2} without copying.
2009 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2010 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2011 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2012 should be nonzero. If they differ for any @var{r}, you should define
2013 this macro to return zero unless some other mechanism ensures the
2014 accessibility of the value in a narrower mode.
2016 You should define this macro to return nonzero in as many cases as
2017 possible since doing so will allow GCC to perform better register
2021 @hook TARGET_HARD_REGNO_SCRATCH_OK
2023 @defmac AVOID_CCMODE_COPIES
2024 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2025 registers. You should only define this macro if support for copying to/from
2026 @code{CCmode} is incomplete.
2029 @node Leaf Functions
2030 @subsection Handling Leaf Functions
2032 @cindex leaf functions
2033 @cindex functions, leaf
2034 On some machines, a leaf function (i.e., one which makes no calls) can run
2035 more efficiently if it does not make its own register window. Often this
2036 means it is required to receive its arguments in the registers where they
2037 are passed by the caller, instead of the registers where they would
2040 The special treatment for leaf functions generally applies only when
2041 other conditions are met; for example, often they may use only those
2042 registers for its own variables and temporaries. We use the term ``leaf
2043 function'' to mean a function that is suitable for this special
2044 handling, so that functions with no calls are not necessarily ``leaf
2047 GCC assigns register numbers before it knows whether the function is
2048 suitable for leaf function treatment. So it needs to renumber the
2049 registers in order to output a leaf function. The following macros
2052 @defmac LEAF_REGISTERS
2053 Name of a char vector, indexed by hard register number, which
2054 contains 1 for a register that is allowable in a candidate for leaf
2057 If leaf function treatment involves renumbering the registers, then the
2058 registers marked here should be the ones before renumbering---those that
2059 GCC would ordinarily allocate. The registers which will actually be
2060 used in the assembler code, after renumbering, should not be marked with 1
2063 Define this macro only if the target machine offers a way to optimize
2064 the treatment of leaf functions.
2067 @defmac LEAF_REG_REMAP (@var{regno})
2068 A C expression whose value is the register number to which @var{regno}
2069 should be renumbered, when a function is treated as a leaf function.
2071 If @var{regno} is a register number which should not appear in a leaf
2072 function before renumbering, then the expression should yield @minus{}1, which
2073 will cause the compiler to abort.
2075 Define this macro only if the target machine offers a way to optimize the
2076 treatment of leaf functions, and registers need to be renumbered to do
2080 @findex current_function_is_leaf
2081 @findex current_function_uses_only_leaf_regs
2082 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2083 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2084 specially. They can test the C variable @code{current_function_is_leaf}
2085 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2086 set prior to local register allocation and is valid for the remaining
2087 compiler passes. They can also test the C variable
2088 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2089 functions which only use leaf registers.
2090 @code{current_function_uses_only_leaf_regs} is valid after all passes
2091 that modify the instructions have been run and is only useful if
2092 @code{LEAF_REGISTERS} is defined.
2093 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2094 @c of the next paragraph?! --mew 2feb93
2096 @node Stack Registers
2097 @subsection Registers That Form a Stack
2099 There are special features to handle computers where some of the
2100 ``registers'' form a stack. Stack registers are normally written by
2101 pushing onto the stack, and are numbered relative to the top of the
2104 Currently, GCC can only handle one group of stack-like registers, and
2105 they must be consecutively numbered. Furthermore, the existing
2106 support for stack-like registers is specific to the 80387 floating
2107 point coprocessor. If you have a new architecture that uses
2108 stack-like registers, you will need to do substantial work on
2109 @file{reg-stack.c} and write your machine description to cooperate
2110 with it, as well as defining these macros.
2113 Define this if the machine has any stack-like registers.
2116 @defmac STACK_REG_COVER_CLASS
2117 This is a cover class containing the stack registers. Define this if
2118 the machine has any stack-like registers.
2121 @defmac FIRST_STACK_REG
2122 The number of the first stack-like register. This one is the top
2126 @defmac LAST_STACK_REG
2127 The number of the last stack-like register. This one is the bottom of
2131 @node Register Classes
2132 @section Register Classes
2133 @cindex register class definitions
2134 @cindex class definitions, register
2136 On many machines, the numbered registers are not all equivalent.
2137 For example, certain registers may not be allowed for indexed addressing;
2138 certain registers may not be allowed in some instructions. These machine
2139 restrictions are described to the compiler using @dfn{register classes}.
2141 You define a number of register classes, giving each one a name and saying
2142 which of the registers belong to it. Then you can specify register classes
2143 that are allowed as operands to particular instruction patterns.
2147 In general, each register will belong to several classes. In fact, one
2148 class must be named @code{ALL_REGS} and contain all the registers. Another
2149 class must be named @code{NO_REGS} and contain no registers. Often the
2150 union of two classes will be another class; however, this is not required.
2152 @findex GENERAL_REGS
2153 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2154 terribly special about the name, but the operand constraint letters
2155 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2156 the same as @code{ALL_REGS}, just define it as a macro which expands
2159 Order the classes so that if class @var{x} is contained in class @var{y}
2160 then @var{x} has a lower class number than @var{y}.
2162 The way classes other than @code{GENERAL_REGS} are specified in operand
2163 constraints is through machine-dependent operand constraint letters.
2164 You can define such letters to correspond to various classes, then use
2165 them in operand constraints.
2167 You must define the narrowest register classes for allocatable
2168 registers, so that each class either has no subclasses, or that for
2169 some mode, the move cost between registers within the class is
2170 cheaper than moving a register in the class to or from memory
2173 You should define a class for the union of two classes whenever some
2174 instruction allows both classes. For example, if an instruction allows
2175 either a floating point (coprocessor) register or a general register for a
2176 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2177 which includes both of them. Otherwise you will get suboptimal code,
2178 or even internal compiler errors when reload cannot find a register in the
2179 class computed via @code{reg_class_subunion}.
2181 You must also specify certain redundant information about the register
2182 classes: for each class, which classes contain it and which ones are
2183 contained in it; for each pair of classes, the largest class contained
2186 When a value occupying several consecutive registers is expected in a
2187 certain class, all the registers used must belong to that class.
2188 Therefore, register classes cannot be used to enforce a requirement for
2189 a register pair to start with an even-numbered register. The way to
2190 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2192 Register classes used for input-operands of bitwise-and or shift
2193 instructions have a special requirement: each such class must have, for
2194 each fixed-point machine mode, a subclass whose registers can transfer that
2195 mode to or from memory. For example, on some machines, the operations for
2196 single-byte values (@code{QImode}) are limited to certain registers. When
2197 this is so, each register class that is used in a bitwise-and or shift
2198 instruction must have a subclass consisting of registers from which
2199 single-byte values can be loaded or stored. This is so that
2200 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2202 @deftp {Data type} {enum reg_class}
2203 An enumerated type that must be defined with all the register class names
2204 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2205 must be the last register class, followed by one more enumerated value,
2206 @code{LIM_REG_CLASSES}, which is not a register class but rather
2207 tells how many classes there are.
2209 Each register class has a number, which is the value of casting
2210 the class name to type @code{int}. The number serves as an index
2211 in many of the tables described below.
2214 @defmac N_REG_CLASSES
2215 The number of distinct register classes, defined as follows:
2218 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2222 @defmac REG_CLASS_NAMES
2223 An initializer containing the names of the register classes as C string
2224 constants. These names are used in writing some of the debugging dumps.
2227 @defmac REG_CLASS_CONTENTS
2228 An initializer containing the contents of the register classes, as integers
2229 which are bit masks. The @var{n}th integer specifies the contents of class
2230 @var{n}. The way the integer @var{mask} is interpreted is that
2231 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2233 When the machine has more than 32 registers, an integer does not suffice.
2234 Then the integers are replaced by sub-initializers, braced groupings containing
2235 several integers. Each sub-initializer must be suitable as an initializer
2236 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2237 In this situation, the first integer in each sub-initializer corresponds to
2238 registers 0 through 31, the second integer to registers 32 through 63, and
2242 @defmac REGNO_REG_CLASS (@var{regno})
2243 A C expression whose value is a register class containing hard register
2244 @var{regno}. In general there is more than one such class; choose a class
2245 which is @dfn{minimal}, meaning that no smaller class also contains the
2249 @defmac BASE_REG_CLASS
2250 A macro whose definition is the name of the class to which a valid
2251 base register must belong. A base register is one used in an address
2252 which is the register value plus a displacement.
2255 @defmac MODE_BASE_REG_CLASS (@var{mode})
2256 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2257 the selection of a base register in a mode dependent manner. If
2258 @var{mode} is VOIDmode then it should return the same value as
2259 @code{BASE_REG_CLASS}.
2262 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2263 A C expression whose value is the register class to which a valid
2264 base register must belong in order to be used in a base plus index
2265 register address. You should define this macro if base plus index
2266 addresses have different requirements than other base register uses.
2269 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2270 A C expression whose value is the register class to which a valid
2271 base register for a memory reference in mode @var{mode} to address
2272 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2273 define the context in which the base register occurs. @var{outer_code} is
2274 the code of the immediately enclosing expression (@code{MEM} for the top level
2275 of an address, @code{ADDRESS} for something that occurs in an
2276 @code{address_operand}). @var{index_code} is the code of the corresponding
2277 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2280 @defmac INDEX_REG_CLASS
2281 A macro whose definition is the name of the class to which a valid
2282 index register must belong. An index register is one used in an
2283 address where its value is either multiplied by a scale factor or
2284 added to another register (as well as added to a displacement).
2287 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2288 A C expression which is nonzero if register number @var{num} is
2289 suitable for use as a base register in operand addresses.
2292 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2293 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2294 that expression may examine the mode of the memory reference in
2295 @var{mode}. You should define this macro if the mode of the memory
2296 reference affects whether a register may be used as a base register. If
2297 you define this macro, the compiler will use it instead of
2298 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2299 addresses that appear outside a @code{MEM}, i.e., as an
2300 @code{address_operand}.
2303 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2304 A C expression which is nonzero if register number @var{num} is suitable for
2305 use as a base register in base plus index operand addresses, accessing
2306 memory in mode @var{mode}. It may be either a suitable hard register or a
2307 pseudo register that has been allocated such a hard register. You should
2308 define this macro if base plus index addresses have different requirements
2309 than other base register uses.
2311 Use of this macro is deprecated; please use the more general
2312 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2315 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2316 A C expression which is nonzero if register number @var{num} is
2317 suitable for use as a base register in operand addresses, accessing
2318 memory in mode @var{mode} in address space @var{address_space}.
2319 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2320 that that expression may examine the context in which the register
2321 appears in the memory reference. @var{outer_code} is the code of the
2322 immediately enclosing expression (@code{MEM} if at the top level of the
2323 address, @code{ADDRESS} for something that occurs in an
2324 @code{address_operand}). @var{index_code} is the code of the
2325 corresponding index expression if @var{outer_code} is @code{PLUS};
2326 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2327 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2330 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2331 A C expression which is nonzero if register number @var{num} is
2332 suitable for use as an index register in operand addresses. It may be
2333 either a suitable hard register or a pseudo register that has been
2334 allocated such a hard register.
2336 The difference between an index register and a base register is that
2337 the index register may be scaled. If an address involves the sum of
2338 two registers, neither one of them scaled, then either one may be
2339 labeled the ``base'' and the other the ``index''; but whichever
2340 labeling is used must fit the machine's constraints of which registers
2341 may serve in each capacity. The compiler will try both labelings,
2342 looking for one that is valid, and will reload one or both registers
2343 only if neither labeling works.
2346 @hook TARGET_PREFERRED_RENAME_CLASS
2348 @hook TARGET_PREFERRED_RELOAD_CLASS
2350 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2351 A C expression that places additional restrictions on the register class
2352 to use when it is necessary to copy value @var{x} into a register in class
2353 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2354 another, smaller class. On many machines, the following definition is
2358 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2361 Sometimes returning a more restrictive class makes better code. For
2362 example, on the 68000, when @var{x} is an integer constant that is in range
2363 for a @samp{moveq} instruction, the value of this macro is always
2364 @code{DATA_REGS} as long as @var{class} includes the data registers.
2365 Requiring a data register guarantees that a @samp{moveq} will be used.
2367 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2368 @var{class} is if @var{x} is a legitimate constant which cannot be
2369 loaded into some register class. By returning @code{NO_REGS} you can
2370 force @var{x} into a memory location. For example, rs6000 can load
2371 immediate values into general-purpose registers, but does not have an
2372 instruction for loading an immediate value into a floating-point
2373 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2374 @var{x} is a floating-point constant. If the constant can't be loaded
2375 into any kind of register, code generation will be better if
2376 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2377 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2379 If an insn has pseudos in it after register allocation, reload will go
2380 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2381 to find the best one. Returning @code{NO_REGS}, in this case, makes
2382 reload add a @code{!} in front of the constraint: the x86 back-end uses
2383 this feature to discourage usage of 387 registers when math is done in
2384 the SSE registers (and vice versa).
2387 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2389 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2390 A C expression that places additional restrictions on the register class
2391 to use when it is necessary to be able to hold a value of mode
2392 @var{mode} in a reload register for which class @var{class} would
2395 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2396 there are certain modes that simply can't go in certain reload classes.
2398 The value is a register class; perhaps @var{class}, or perhaps another,
2401 Don't define this macro unless the target machine has limitations which
2402 require the macro to do something nontrivial.
2405 @hook TARGET_SECONDARY_RELOAD
2407 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2408 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2409 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2410 These macros are obsolete, new ports should use the target hook
2411 @code{TARGET_SECONDARY_RELOAD} instead.
2413 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2414 target hook. Older ports still define these macros to indicate to the
2415 reload phase that it may
2416 need to allocate at least one register for a reload in addition to the
2417 register to contain the data. Specifically, if copying @var{x} to a
2418 register @var{class} in @var{mode} requires an intermediate register,
2419 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2420 largest register class all of whose registers can be used as
2421 intermediate registers or scratch registers.
2423 If copying a register @var{class} in @var{mode} to @var{x} requires an
2424 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2425 was supposed to be defined be defined to return the largest register
2426 class required. If the
2427 requirements for input and output reloads were the same, the macro
2428 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2431 The values returned by these macros are often @code{GENERAL_REGS}.
2432 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2433 can be directly copied to or from a register of @var{class} in
2434 @var{mode} without requiring a scratch register. Do not define this
2435 macro if it would always return @code{NO_REGS}.
2437 If a scratch register is required (either with or without an
2438 intermediate register), you were supposed to define patterns for
2439 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2440 (@pxref{Standard Names}. These patterns, which were normally
2441 implemented with a @code{define_expand}, should be similar to the
2442 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2445 These patterns need constraints for the reload register and scratch
2447 contain a single register class. If the original reload register (whose
2448 class is @var{class}) can meet the constraint given in the pattern, the
2449 value returned by these macros is used for the class of the scratch
2450 register. Otherwise, two additional reload registers are required.
2451 Their classes are obtained from the constraints in the insn pattern.
2453 @var{x} might be a pseudo-register or a @code{subreg} of a
2454 pseudo-register, which could either be in a hard register or in memory.
2455 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2456 in memory and the hard register number if it is in a register.
2458 These macros should not be used in the case where a particular class of
2459 registers can only be copied to memory and not to another class of
2460 registers. In that case, secondary reload registers are not needed and
2461 would not be helpful. Instead, a stack location must be used to perform
2462 the copy and the @code{mov@var{m}} pattern should use memory as an
2463 intermediate storage. This case often occurs between floating-point and
2467 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2468 Certain machines have the property that some registers cannot be copied
2469 to some other registers without using memory. Define this macro on
2470 those machines to be a C expression that is nonzero if objects of mode
2471 @var{m} in registers of @var{class1} can only be copied to registers of
2472 class @var{class2} by storing a register of @var{class1} into memory
2473 and loading that memory location into a register of @var{class2}.
2475 Do not define this macro if its value would always be zero.
2478 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2479 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2480 allocates a stack slot for a memory location needed for register copies.
2481 If this macro is defined, the compiler instead uses the memory location
2482 defined by this macro.
2484 Do not define this macro if you do not define
2485 @code{SECONDARY_MEMORY_NEEDED}.
2488 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2489 When the compiler needs a secondary memory location to copy between two
2490 registers of mode @var{mode}, it normally allocates sufficient memory to
2491 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2492 load operations in a mode that many bits wide and whose class is the
2493 same as that of @var{mode}.
2495 This is right thing to do on most machines because it ensures that all
2496 bits of the register are copied and prevents accesses to the registers
2497 in a narrower mode, which some machines prohibit for floating-point
2500 However, this default behavior is not correct on some machines, such as
2501 the DEC Alpha, that store short integers in floating-point registers
2502 differently than in integer registers. On those machines, the default
2503 widening will not work correctly and you must define this macro to
2504 suppress that widening in some cases. See the file @file{alpha.h} for
2507 Do not define this macro if you do not define
2508 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2509 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2512 @hook TARGET_CLASS_LIKELY_SPILLED_P
2514 @hook TARGET_CLASS_MAX_NREGS
2516 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2517 A C expression for the maximum number of consecutive registers
2518 of class @var{class} needed to hold a value of mode @var{mode}.
2520 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2521 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2522 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2523 @var{mode})} for all @var{regno} values in the class @var{class}.
2525 This macro helps control the handling of multiple-word values
2529 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2530 If defined, a C expression that returns nonzero for a @var{class} for which
2531 a change from mode @var{from} to mode @var{to} is invalid.
2533 For the example, loading 32-bit integer or floating-point objects into
2534 floating-point registers on the Alpha extends them to 64 bits.
2535 Therefore loading a 64-bit object and then storing it as a 32-bit object
2536 does not store the low-order 32 bits, as would be the case for a normal
2537 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2541 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2542 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2543 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2549 @hook TARGET_REGISTER_PRIORITY
2551 @hook TARGET_REGISTER_USAGE_LEVELING_P
2553 @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2555 @hook TARGET_SPILL_CLASS
2557 @hook TARGET_CSTORE_MODE
2559 @node Old Constraints
2560 @section Obsolete Macros for Defining Constraints
2561 @cindex defining constraints, obsolete method
2562 @cindex constraints, defining, obsolete method
2564 Machine-specific constraints can be defined with these macros instead
2565 of the machine description constructs described in @ref{Define
2566 Constraints}. This mechanism is obsolete. New ports should not use
2567 it; old ports should convert to the new mechanism.
2569 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2570 For the constraint at the start of @var{str}, which starts with the letter
2571 @var{c}, return the length. This allows you to have register class /
2572 constant / extra constraints that are longer than a single letter;
2573 you don't need to define this macro if you can do with single-letter
2574 constraints only. The definition of this macro should use
2575 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2576 to handle specially.
2577 There are some sanity checks in genoutput.c that check the constraint lengths
2578 for the md file, so you can also use this macro to help you while you are
2579 transitioning from a byzantine single-letter-constraint scheme: when you
2580 return a negative length for a constraint you want to re-use, genoutput
2581 will complain about every instance where it is used in the md file.
2584 @defmac REG_CLASS_FROM_LETTER (@var{char})
2585 A C expression which defines the machine-dependent operand constraint
2586 letters for register classes. If @var{char} is such a letter, the
2587 value should be the register class corresponding to it. Otherwise,
2588 the value should be @code{NO_REGS}. The register letter @samp{r},
2589 corresponding to class @code{GENERAL_REGS}, will not be passed
2590 to this macro; you do not need to handle it.
2593 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2594 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2595 passed in @var{str}, so that you can use suffixes to distinguish between
2599 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2600 A C expression that defines the machine-dependent operand constraint
2601 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2602 particular ranges of integer values. If @var{c} is one of those
2603 letters, the expression should check that @var{value}, an integer, is in
2604 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2605 not one of those letters, the value should be 0 regardless of
2609 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2610 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2611 string passed in @var{str}, so that you can use suffixes to distinguish
2612 between different variants.
2615 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2616 A C expression that defines the machine-dependent operand constraint
2617 letters that specify particular ranges of @code{const_double} values
2618 (@samp{G} or @samp{H}).
2620 If @var{c} is one of those letters, the expression should check that
2621 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2622 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2623 letters, the value should be 0 regardless of @var{value}.
2625 @code{const_double} is used for all floating-point constants and for
2626 @code{DImode} fixed-point constants. A given letter can accept either
2627 or both kinds of values. It can use @code{GET_MODE} to distinguish
2628 between these kinds.
2631 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2632 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2633 string passed in @var{str}, so that you can use suffixes to distinguish
2634 between different variants.
2637 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2638 A C expression that defines the optional machine-dependent constraint
2639 letters that can be used to segregate specific types of operands, usually
2640 memory references, for the target machine. Any letter that is not
2641 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2642 @code{REG_CLASS_FROM_CONSTRAINT}
2643 may be used. Normally this macro will not be defined.
2645 If it is required for a particular target machine, it should return 1
2646 if @var{value} corresponds to the operand type represented by the
2647 constraint letter @var{c}. If @var{c} is not defined as an extra
2648 constraint, the value returned should be 0 regardless of @var{value}.
2650 For example, on the ROMP, load instructions cannot have their output
2651 in r0 if the memory reference contains a symbolic address. Constraint
2652 letter @samp{Q} is defined as representing a memory address that does
2653 @emph{not} contain a symbolic address. An alternative is specified with
2654 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2655 alternative specifies @samp{m} on the input and a register class that
2656 does not include r0 on the output.
2659 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2660 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2661 in @var{str}, so that you can use suffixes to distinguish between different
2665 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2666 A C expression that defines the optional machine-dependent constraint
2667 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2668 be treated like memory constraints by the reload pass.
2670 It should return 1 if the operand type represented by the constraint
2671 at the start of @var{str}, the first letter of which is the letter @var{c},
2672 comprises a subset of all memory references including
2673 all those whose address is simply a base register. This allows the reload
2674 pass to reload an operand, if it does not directly correspond to the operand
2675 type of @var{c}, by copying its address into a base register.
2677 For example, on the S/390, some instructions do not accept arbitrary
2678 memory references, but only those that do not make use of an index
2679 register. The constraint letter @samp{Q} is defined via
2680 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2681 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2682 a @samp{Q} constraint can handle any memory operand, because the
2683 reload pass knows it can be reloaded by copying the memory address
2684 into a base register if required. This is analogous to the way
2685 an @samp{o} constraint can handle any memory operand.
2688 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2689 A C expression that defines the optional machine-dependent constraint
2690 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2691 @code{EXTRA_CONSTRAINT_STR}, that should
2692 be treated like address constraints by the reload pass.
2694 It should return 1 if the operand type represented by the constraint
2695 at the start of @var{str}, which starts with the letter @var{c}, comprises
2696 a subset of all memory addresses including
2697 all those that consist of just a base register. This allows the reload
2698 pass to reload an operand, if it does not directly correspond to the operand
2699 type of @var{str}, by copying it into a base register.
2701 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2702 be used with the @code{address_operand} predicate. It is treated
2703 analogously to the @samp{p} constraint.
2706 @node Stack and Calling
2707 @section Stack Layout and Calling Conventions
2708 @cindex calling conventions
2710 @c prevent bad page break with this line
2711 This describes the stack layout and calling conventions.
2715 * Exception Handling::
2720 * Register Arguments::
2722 * Aggregate Return::
2727 * Stack Smashing Protection::
2731 @subsection Basic Stack Layout
2732 @cindex stack frame layout
2733 @cindex frame layout
2735 @c prevent bad page break with this line
2736 Here is the basic stack layout.
2738 @defmac STACK_GROWS_DOWNWARD
2739 Define this macro if pushing a word onto the stack moves the stack
2740 pointer to a smaller address.
2742 When we say, ``define this macro if @dots{}'', it means that the
2743 compiler checks this macro only with @code{#ifdef} so the precise
2744 definition used does not matter.
2747 @defmac STACK_PUSH_CODE
2748 This macro defines the operation used when something is pushed
2749 on the stack. In RTL, a push operation will be
2750 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2752 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2753 and @code{POST_INC}. Which of these is correct depends on
2754 the stack direction and on whether the stack pointer points
2755 to the last item on the stack or whether it points to the
2756 space for the next item on the stack.
2758 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2759 defined, which is almost always right, and @code{PRE_INC} otherwise,
2760 which is often wrong.
2763 @defmac FRAME_GROWS_DOWNWARD
2764 Define this macro to nonzero value if the addresses of local variable slots
2765 are at negative offsets from the frame pointer.
2768 @defmac ARGS_GROW_DOWNWARD
2769 Define this macro if successive arguments to a function occupy decreasing
2770 addresses on the stack.
2773 @defmac STARTING_FRAME_OFFSET
2774 Offset from the frame pointer to the first local variable slot to be allocated.
2776 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2777 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2778 Otherwise, it is found by adding the length of the first slot to the
2779 value @code{STARTING_FRAME_OFFSET}.
2780 @c i'm not sure if the above is still correct.. had to change it to get
2781 @c rid of an overfull. --mew 2feb93
2784 @defmac STACK_ALIGNMENT_NEEDED
2785 Define to zero to disable final alignment of the stack during reload.
2786 The nonzero default for this macro is suitable for most ports.
2788 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2789 is a register save block following the local block that doesn't require
2790 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2791 stack alignment and do it in the backend.
2794 @defmac STACK_POINTER_OFFSET
2795 Offset from the stack pointer register to the first location at which
2796 outgoing arguments are placed. If not specified, the default value of
2797 zero is used. This is the proper value for most machines.
2799 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2800 the first location at which outgoing arguments are placed.
2803 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2804 Offset from the argument pointer register to the first argument's
2805 address. On some machines it may depend on the data type of the
2808 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2809 the first argument's address.
2812 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2813 Offset from the stack pointer register to an item dynamically allocated
2814 on the stack, e.g., by @code{alloca}.
2816 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2817 length of the outgoing arguments. The default is correct for most
2818 machines. See @file{function.c} for details.
2821 @defmac INITIAL_FRAME_ADDRESS_RTX
2822 A C expression whose value is RTL representing the address of the initial
2823 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2824 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2825 default value will be used. Define this macro in order to make frame pointer
2826 elimination work in the presence of @code{__builtin_frame_address (count)} and
2827 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2830 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2831 A C expression whose value is RTL representing the address in a stack
2832 frame where the pointer to the caller's frame is stored. Assume that
2833 @var{frameaddr} is an RTL expression for the address of the stack frame
2836 If you don't define this macro, the default is to return the value
2837 of @var{frameaddr}---that is, the stack frame address is also the
2838 address of the stack word that points to the previous frame.
2841 @defmac SETUP_FRAME_ADDRESSES
2842 If defined, a C expression that produces the machine-specific code to
2843 setup the stack so that arbitrary frames can be accessed. For example,
2844 on the SPARC, we must flush all of the register windows to the stack
2845 before we can access arbitrary stack frames. You will seldom need to
2849 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2851 @defmac FRAME_ADDR_RTX (@var{frameaddr})
2852 A C expression whose value is RTL representing the value of the frame
2853 address for the current frame. @var{frameaddr} is the frame pointer
2854 of the current frame. This is used for __builtin_frame_address.
2855 You need only define this macro if the frame address is not the same
2856 as the frame pointer. Most machines do not need to define it.
2859 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2860 A C expression whose value is RTL representing the value of the return
2861 address for the frame @var{count} steps up from the current frame, after
2862 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2863 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2864 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2866 The value of the expression must always be the correct address when
2867 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2868 determine the return address of other frames.
2871 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2872 Define this if the return address of a particular stack frame is accessed
2873 from the frame pointer of the previous stack frame.
2876 @defmac INCOMING_RETURN_ADDR_RTX
2877 A C expression whose value is RTL representing the location of the
2878 incoming return address at the beginning of any function, before the
2879 prologue. This RTL is either a @code{REG}, indicating that the return
2880 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2883 You only need to define this macro if you want to support call frame
2884 debugging information like that provided by DWARF 2.
2886 If this RTL is a @code{REG}, you should also define
2887 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2890 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2891 A C expression whose value is an integer giving a DWARF 2 column
2892 number that may be used as an alternative return column. The column
2893 must not correspond to any gcc hard register (that is, it must not
2894 be in the range of @code{DWARF_FRAME_REGNUM}).
2896 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2897 general register, but an alternative column needs to be used for signal
2898 frames. Some targets have also used different frame return columns
2902 @defmac DWARF_ZERO_REG
2903 A C expression whose value is an integer giving a DWARF 2 register
2904 number that is considered to always have the value zero. This should
2905 only be defined if the target has an architected zero register, and
2906 someone decided it was a good idea to use that register number to
2907 terminate the stack backtrace. New ports should avoid this.
2910 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2912 @defmac INCOMING_FRAME_SP_OFFSET
2913 A C expression whose value is an integer giving the offset, in bytes,
2914 from the value of the stack pointer register to the top of the stack
2915 frame at the beginning of any function, before the prologue. The top of
2916 the frame is defined to be the value of the stack pointer in the
2917 previous frame, just before the call instruction.
2919 You only need to define this macro if you want to support call frame
2920 debugging information like that provided by DWARF 2.
2923 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2924 A C expression whose value is an integer giving the offset, in bytes,
2925 from the argument pointer to the canonical frame address (cfa). The
2926 final value should coincide with that calculated by
2927 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2928 during virtual register instantiation.
2930 The default value for this macro is
2931 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2932 which is correct for most machines; in general, the arguments are found
2933 immediately before the stack frame. Note that this is not the case on
2934 some targets that save registers into the caller's frame, such as SPARC
2935 and rs6000, and so such targets need to define this macro.
2937 You only need to define this macro if the default is incorrect, and you
2938 want to support call frame debugging information like that provided by
2942 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2943 If defined, a C expression whose value is an integer giving the offset
2944 in bytes from the frame pointer to the canonical frame address (cfa).
2945 The final value should coincide with that calculated by
2946 @code{INCOMING_FRAME_SP_OFFSET}.
2948 Normally the CFA is calculated as an offset from the argument pointer,
2949 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2950 variable due to the ABI, this may not be possible. If this macro is
2951 defined, it implies that the virtual register instantiation should be
2952 based on the frame pointer instead of the argument pointer. Only one
2953 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2957 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2958 If defined, a C expression whose value is an integer giving the offset
2959 in bytes from the canonical frame address (cfa) to the frame base used
2960 in DWARF 2 debug information. The default is zero. A different value
2961 may reduce the size of debug information on some ports.
2964 @node Exception Handling
2965 @subsection Exception Handling Support
2966 @cindex exception handling
2968 @defmac EH_RETURN_DATA_REGNO (@var{N})
2969 A C expression whose value is the @var{N}th register number used for
2970 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2971 @var{N} registers are usable.
2973 The exception handling library routines communicate with the exception
2974 handlers via a set of agreed upon registers. Ideally these registers
2975 should be call-clobbered; it is possible to use call-saved registers,
2976 but may negatively impact code size. The target must support at least
2977 2 data registers, but should define 4 if there are enough free registers.
2979 You must define this macro if you want to support call frame exception
2980 handling like that provided by DWARF 2.
2983 @defmac EH_RETURN_STACKADJ_RTX
2984 A C expression whose value is RTL representing a location in which
2985 to store a stack adjustment to be applied before function return.
2986 This is used to unwind the stack to an exception handler's call frame.
2987 It will be assigned zero on code paths that return normally.
2989 Typically this is a call-clobbered hard register that is otherwise
2990 untouched by the epilogue, but could also be a stack slot.
2992 Do not define this macro if the stack pointer is saved and restored
2993 by the regular prolog and epilog code in the call frame itself; in
2994 this case, the exception handling library routines will update the
2995 stack location to be restored in place. Otherwise, you must define
2996 this macro if you want to support call frame exception handling like
2997 that provided by DWARF 2.
3000 @defmac EH_RETURN_HANDLER_RTX
3001 A C expression whose value is RTL representing a location in which
3002 to store the address of an exception handler to which we should
3003 return. It will not be assigned on code paths that return normally.
3005 Typically this is the location in the call frame at which the normal
3006 return address is stored. For targets that return by popping an
3007 address off the stack, this might be a memory address just below
3008 the @emph{target} call frame rather than inside the current call
3009 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3010 been assigned, so it may be used to calculate the location of the
3013 Some targets have more complex requirements than storing to an
3014 address calculable during initial code generation. In that case
3015 the @code{eh_return} instruction pattern should be used instead.
3017 If you want to support call frame exception handling, you must
3018 define either this macro or the @code{eh_return} instruction pattern.
3021 @defmac RETURN_ADDR_OFFSET
3022 If defined, an integer-valued C expression for which rtl will be generated
3023 to add it to the exception handler address before it is searched in the
3024 exception handling tables, and to subtract it again from the address before
3025 using it to return to the exception handler.
3028 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3029 This macro chooses the encoding of pointers embedded in the exception
3030 handling sections. If at all possible, this should be defined such
3031 that the exception handling section will not require dynamic relocations,
3032 and so may be read-only.
3034 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3035 @var{global} is true if the symbol may be affected by dynamic relocations.
3036 The macro should return a combination of the @code{DW_EH_PE_*} defines
3037 as found in @file{dwarf2.h}.
3039 If this macro is not defined, pointers will not be encoded but
3040 represented directly.
3043 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3044 This macro allows the target to emit whatever special magic is required
3045 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3046 Generic code takes care of pc-relative and indirect encodings; this must
3047 be defined if the target uses text-relative or data-relative encodings.
3049 This is a C statement that branches to @var{done} if the format was
3050 handled. @var{encoding} is the format chosen, @var{size} is the number
3051 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3055 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3056 This macro allows the target to add CPU and operating system specific
3057 code to the call-frame unwinder for use when there is no unwind data
3058 available. The most common reason to implement this macro is to unwind
3059 through signal frames.
3061 This macro is called from @code{uw_frame_state_for} in
3062 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3063 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3064 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3065 for the address of the code being executed and @code{context->cfa} for
3066 the stack pointer value. If the frame can be decoded, the register
3067 save addresses should be updated in @var{fs} and the macro should
3068 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3069 the macro should evaluate to @code{_URC_END_OF_STACK}.
3071 For proper signal handling in Java this macro is accompanied by
3072 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3075 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3076 This macro allows the target to add operating system specific code to the
3077 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3078 usually used for signal or interrupt frames.
3080 This macro is called from @code{uw_update_context} in libgcc's
3081 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3082 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3083 for the abi and context in the @code{.unwabi} directive. If the
3084 @code{.unwabi} directive can be handled, the register save addresses should
3085 be updated in @var{fs}.
3088 @defmac TARGET_USES_WEAK_UNWIND_INFO
3089 A C expression that evaluates to true if the target requires unwind
3090 info to be given comdat linkage. Define it to be @code{1} if comdat
3091 linkage is necessary. The default is @code{0}.
3094 @node Stack Checking
3095 @subsection Specifying How Stack Checking is Done
3097 GCC will check that stack references are within the boundaries of the
3098 stack, if the option @option{-fstack-check} is specified, in one of
3103 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3104 will assume that you have arranged for full stack checking to be done
3105 at appropriate places in the configuration files. GCC will not do
3106 other special processing.
3109 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3110 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3111 that you have arranged for static stack checking (checking of the
3112 static stack frame of functions) to be done at appropriate places
3113 in the configuration files. GCC will only emit code to do dynamic
3114 stack checking (checking on dynamic stack allocations) using the third
3118 If neither of the above are true, GCC will generate code to periodically
3119 ``probe'' the stack pointer using the values of the macros defined below.
3122 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3123 GCC will change its allocation strategy for large objects if the option
3124 @option{-fstack-check} is specified: they will always be allocated
3125 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3127 @defmac STACK_CHECK_BUILTIN
3128 A nonzero value if stack checking is done by the configuration files in a
3129 machine-dependent manner. You should define this macro if stack checking
3130 is required by the ABI of your machine or if you would like to do stack
3131 checking in some more efficient way than the generic approach. The default
3132 value of this macro is zero.
3135 @defmac STACK_CHECK_STATIC_BUILTIN
3136 A nonzero value if static stack checking is done by the configuration files
3137 in a machine-dependent manner. You should define this macro if you would
3138 like to do static stack checking in some more efficient way than the generic
3139 approach. The default value of this macro is zero.
3142 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3143 An integer specifying the interval at which GCC must generate stack probe
3144 instructions, defined as 2 raised to this integer. You will normally
3145 define this macro so that the interval be no larger than the size of
3146 the ``guard pages'' at the end of a stack area. The default value
3147 of 12 (4096-byte interval) is suitable for most systems.
3150 @defmac STACK_CHECK_MOVING_SP
3151 An integer which is nonzero if GCC should move the stack pointer page by page
3152 when doing probes. This can be necessary on systems where the stack pointer
3153 contains the bottom address of the memory area accessible to the executing
3154 thread at any point in time. In this situation an alternate signal stack
3155 is required in order to be able to recover from a stack overflow. The
3156 default value of this macro is zero.
3159 @defmac STACK_CHECK_PROTECT
3160 The number of bytes of stack needed to recover from a stack overflow, for
3161 languages where such a recovery is supported. The default value of 75 words
3162 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3163 8192 bytes with other exception handling mechanisms should be adequate for
3167 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3168 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3169 in the opposite case.
3171 @defmac STACK_CHECK_MAX_FRAME_SIZE
3172 The maximum size of a stack frame, in bytes. GCC will generate probe
3173 instructions in non-leaf functions to ensure at least this many bytes of
3174 stack are available. If a stack frame is larger than this size, stack
3175 checking will not be reliable and GCC will issue a warning. The
3176 default is chosen so that GCC only generates one instruction on most
3177 systems. You should normally not change the default value of this macro.
3180 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3181 GCC uses this value to generate the above warning message. It
3182 represents the amount of fixed frame used by a function, not including
3183 space for any callee-saved registers, temporaries and user variables.
3184 You need only specify an upper bound for this amount and will normally
3185 use the default of four words.
3188 @defmac STACK_CHECK_MAX_VAR_SIZE
3189 The maximum size, in bytes, of an object that GCC will place in the
3190 fixed area of the stack frame when the user specifies
3191 @option{-fstack-check}.
3192 GCC computed the default from the values of the above macros and you will
3193 normally not need to override that default.
3197 @node Frame Registers
3198 @subsection Registers That Address the Stack Frame
3200 @c prevent bad page break with this line
3201 This discusses registers that address the stack frame.
3203 @defmac STACK_POINTER_REGNUM
3204 The register number of the stack pointer register, which must also be a
3205 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3206 the hardware determines which register this is.
3209 @defmac FRAME_POINTER_REGNUM
3210 The register number of the frame pointer register, which is used to
3211 access automatic variables in the stack frame. On some machines, the
3212 hardware determines which register this is. On other machines, you can
3213 choose any register you wish for this purpose.
3216 @defmac HARD_FRAME_POINTER_REGNUM
3217 On some machines the offset between the frame pointer and starting
3218 offset of the automatic variables is not known until after register
3219 allocation has been done (for example, because the saved registers are
3220 between these two locations). On those machines, define
3221 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3222 be used internally until the offset is known, and define
3223 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3224 used for the frame pointer.
3226 You should define this macro only in the very rare circumstances when it
3227 is not possible to calculate the offset between the frame pointer and
3228 the automatic variables until after register allocation has been
3229 completed. When this macro is defined, you must also indicate in your
3230 definition of @code{ELIMINABLE_REGS} how to eliminate
3231 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3232 or @code{STACK_POINTER_REGNUM}.
3234 Do not define this macro if it would be the same as
3235 @code{FRAME_POINTER_REGNUM}.
3238 @defmac ARG_POINTER_REGNUM
3239 The register number of the arg pointer register, which is used to access
3240 the function's argument list. On some machines, this is the same as the
3241 frame pointer register. On some machines, the hardware determines which
3242 register this is. On other machines, you can choose any register you
3243 wish for this purpose. If this is not the same register as the frame
3244 pointer register, then you must mark it as a fixed register according to
3245 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3246 (@pxref{Elimination}).
3249 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3250 Define this to a preprocessor constant that is nonzero if
3251 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3252 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3253 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3254 definition is not suitable for use in preprocessor conditionals.
3257 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3258 Define this to a preprocessor constant that is nonzero if
3259 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3260 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3261 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3262 definition is not suitable for use in preprocessor conditionals.
3265 @defmac RETURN_ADDRESS_POINTER_REGNUM
3266 The register number of the return address pointer register, which is used to
3267 access the current function's return address from the stack. On some
3268 machines, the return address is not at a fixed offset from the frame
3269 pointer or stack pointer or argument pointer. This register can be defined
3270 to point to the return address on the stack, and then be converted by
3271 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3273 Do not define this macro unless there is no other way to get the return
3274 address from the stack.
3277 @defmac STATIC_CHAIN_REGNUM
3278 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3279 Register numbers used for passing a function's static chain pointer. If
3280 register windows are used, the register number as seen by the called
3281 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3282 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3283 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3286 The static chain register need not be a fixed register.
3288 If the static chain is passed in memory, these macros should not be
3289 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3292 @hook TARGET_STATIC_CHAIN
3294 @defmac DWARF_FRAME_REGISTERS
3295 This macro specifies the maximum number of hard registers that can be
3296 saved in a call frame. This is used to size data structures used in
3297 DWARF2 exception handling.
3299 Prior to GCC 3.0, this macro was needed in order to establish a stable
3300 exception handling ABI in the face of adding new hard registers for ISA
3301 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3302 in the number of hard registers. Nevertheless, this macro can still be
3303 used to reduce the runtime memory requirements of the exception handling
3304 routines, which can be substantial if the ISA contains a lot of
3305 registers that are not call-saved.
3307 If this macro is not defined, it defaults to
3308 @code{FIRST_PSEUDO_REGISTER}.
3311 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3313 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3314 for backward compatibility in pre GCC 3.0 compiled code.
3316 If this macro is not defined, it defaults to
3317 @code{DWARF_FRAME_REGISTERS}.
3320 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3322 Define this macro if the target's representation for dwarf registers
3323 is different than the internal representation for unwind column.
3324 Given a dwarf register, this macro should return the internal unwind
3325 column number to use instead.
3327 See the PowerPC's SPE target for an example.
3330 @defmac DWARF_FRAME_REGNUM (@var{regno})
3332 Define this macro if the target's representation for dwarf registers
3333 used in .eh_frame or .debug_frame is different from that used in other
3334 debug info sections. Given a GCC hard register number, this macro
3335 should return the .eh_frame register number. The default is
3336 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3340 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3342 Define this macro to map register numbers held in the call frame info
3343 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3344 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3345 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3346 return @code{@var{regno}}.
3350 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3352 Define this macro if the target stores register values as
3353 @code{_Unwind_Word} type in unwind context. It should be defined if
3354 target register size is larger than the size of @code{void *}. The
3355 default is to store register values as @code{void *} type.
3359 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3361 Define this macro to be 1 if the target always uses extended unwind
3362 context with version, args_size and by_value fields. If it is undefined,
3363 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3364 defined and 0 otherwise.
3369 @subsection Eliminating Frame Pointer and Arg Pointer
3371 @c prevent bad page break with this line
3372 This is about eliminating the frame pointer and arg pointer.
3374 @hook TARGET_FRAME_POINTER_REQUIRED
3376 @findex get_frame_size
3377 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3378 A C statement to store in the variable @var{depth-var} the difference
3379 between the frame pointer and the stack pointer values immediately after
3380 the function prologue. The value would be computed from information
3381 such as the result of @code{get_frame_size ()} and the tables of
3382 registers @code{regs_ever_live} and @code{call_used_regs}.
3384 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3385 need not be defined. Otherwise, it must be defined even if
3386 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3387 case, you may set @var{depth-var} to anything.
3390 @defmac ELIMINABLE_REGS
3391 If defined, this macro specifies a table of register pairs used to
3392 eliminate unneeded registers that point into the stack frame. If it is not
3393 defined, the only elimination attempted by the compiler is to replace
3394 references to the frame pointer with references to the stack pointer.
3396 The definition of this macro is a list of structure initializations, each
3397 of which specifies an original and replacement register.
3399 On some machines, the position of the argument pointer is not known until
3400 the compilation is completed. In such a case, a separate hard register
3401 must be used for the argument pointer. This register can be eliminated by
3402 replacing it with either the frame pointer or the argument pointer,
3403 depending on whether or not the frame pointer has been eliminated.
3405 In this case, you might specify:
3407 #define ELIMINABLE_REGS \
3408 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3409 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3410 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3413 Note that the elimination of the argument pointer with the stack pointer is
3414 specified first since that is the preferred elimination.
3417 @hook TARGET_CAN_ELIMINATE
3419 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3420 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3421 specifies the initial difference between the specified pair of
3422 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3426 @node Stack Arguments
3427 @subsection Passing Function Arguments on the Stack
3428 @cindex arguments on stack
3429 @cindex stack arguments
3431 The macros in this section control how arguments are passed
3432 on the stack. See the following section for other macros that
3433 control passing certain arguments in registers.
3435 @hook TARGET_PROMOTE_PROTOTYPES
3438 A C expression. If nonzero, push insns will be used to pass
3440 If the target machine does not have a push instruction, set it to zero.
3441 That directs GCC to use an alternate strategy: to
3442 allocate the entire argument block and then store the arguments into
3443 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3446 @defmac PUSH_ARGS_REVERSED
3447 A C expression. If nonzero, function arguments will be evaluated from
3448 last to first, rather than from first to last. If this macro is not
3449 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3450 and args grow in opposite directions, and 0 otherwise.
3453 @defmac PUSH_ROUNDING (@var{npushed})
3454 A C expression that is the number of bytes actually pushed onto the
3455 stack when an instruction attempts to push @var{npushed} bytes.
3457 On some machines, the definition
3460 #define PUSH_ROUNDING(BYTES) (BYTES)
3464 will suffice. But on other machines, instructions that appear
3465 to push one byte actually push two bytes in an attempt to maintain
3466 alignment. Then the definition should be
3469 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3472 If the value of this macro has a type, it should be an unsigned type.
3475 @findex outgoing_args_size
3476 @findex crtl->outgoing_args_size
3477 @defmac ACCUMULATE_OUTGOING_ARGS
3478 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3479 will be computed and placed into
3480 @code{crtl->outgoing_args_size}. No space will be pushed
3481 onto the stack for each call; instead, the function prologue should
3482 increase the stack frame size by this amount.
3484 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3488 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3489 Define this macro if functions should assume that stack space has been
3490 allocated for arguments even when their values are passed in
3493 The value of this macro is the size, in bytes, of the area reserved for
3494 arguments passed in registers for the function represented by @var{fndecl},
3495 which can be zero if GCC is calling a library function.
3496 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3499 This space can be allocated by the caller, or be a part of the
3500 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3503 @c above is overfull. not sure what to do. --mew 5feb93 did
3504 @c something, not sure if it looks good. --mew 10feb93
3506 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3507 Define this to a nonzero value if it is the responsibility of the
3508 caller to allocate the area reserved for arguments passed in registers
3509 when calling a function of @var{fntype}. @var{fntype} may be NULL
3510 if the function called is a library function.
3512 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3513 whether the space for these arguments counts in the value of
3514 @code{crtl->outgoing_args_size}.
3517 @defmac STACK_PARMS_IN_REG_PARM_AREA
3518 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3519 stack parameters don't skip the area specified by it.
3520 @c i changed this, makes more sens and it should have taken care of the
3521 @c overfull.. not as specific, tho. --mew 5feb93
3523 Normally, when a parameter is not passed in registers, it is placed on the
3524 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3525 suppresses this behavior and causes the parameter to be passed on the
3526 stack in its natural location.
3529 @hook TARGET_RETURN_POPS_ARGS
3531 @defmac CALL_POPS_ARGS (@var{cum})
3532 A C expression that should indicate the number of bytes a call sequence
3533 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3534 when compiling a function call.
3536 @var{cum} is the variable in which all arguments to the called function
3537 have been accumulated.
3539 On certain architectures, such as the SH5, a call trampoline is used
3540 that pops certain registers off the stack, depending on the arguments
3541 that have been passed to the function. Since this is a property of the
3542 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3546 @node Register Arguments
3547 @subsection Passing Arguments in Registers
3548 @cindex arguments in registers
3549 @cindex registers arguments
3551 This section describes the macros which let you control how various
3552 types of arguments are passed in registers or how they are arranged in
3555 @hook TARGET_FUNCTION_ARG
3557 @hook TARGET_MUST_PASS_IN_STACK
3559 @hook TARGET_FUNCTION_INCOMING_ARG
3561 @hook TARGET_ARG_PARTIAL_BYTES
3563 @hook TARGET_PASS_BY_REFERENCE
3565 @hook TARGET_CALLEE_COPIES
3567 @defmac CUMULATIVE_ARGS
3568 A C type for declaring a variable that is used as the first argument
3569 of @code{TARGET_FUNCTION_ARG} and other related values. For some
3570 target machines, the type @code{int} suffices and can hold the number
3571 of bytes of argument so far.
3573 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3574 arguments that have been passed on the stack. The compiler has other
3575 variables to keep track of that. For target machines on which all
3576 arguments are passed on the stack, there is no need to store anything in
3577 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3578 should not be empty, so use @code{int}.
3581 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3582 If defined, this macro is called before generating any code for a
3583 function, but after the @var{cfun} descriptor for the function has been
3584 created. The back end may use this macro to update @var{cfun} to
3585 reflect an ABI other than that which would normally be used by default.
3586 If the compiler is generating code for a compiler-generated function,
3587 @var{fndecl} may be @code{NULL}.
3590 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3591 A C statement (sans semicolon) for initializing the variable
3592 @var{cum} for the state at the beginning of the argument list. The
3593 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3594 is the tree node for the data type of the function which will receive
3595 the args, or 0 if the args are to a compiler support library function.
3596 For direct calls that are not libcalls, @var{fndecl} contain the
3597 declaration node of the function. @var{fndecl} is also set when
3598 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3599 being compiled. @var{n_named_args} is set to the number of named
3600 arguments, including a structure return address if it is passed as a
3601 parameter, when making a call. When processing incoming arguments,
3602 @var{n_named_args} is set to @minus{}1.
3604 When processing a call to a compiler support library function,
3605 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3606 contains the name of the function, as a string. @var{libname} is 0 when
3607 an ordinary C function call is being processed. Thus, each time this
3608 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3609 never both of them at once.
3612 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3613 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3614 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3615 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3616 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3617 0)} is used instead.
3620 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3621 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3622 finding the arguments for the function being compiled. If this macro is
3623 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3625 The value passed for @var{libname} is always 0, since library routines
3626 with special calling conventions are never compiled with GCC@. The
3627 argument @var{libname} exists for symmetry with
3628 @code{INIT_CUMULATIVE_ARGS}.
3629 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3630 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3633 @hook TARGET_FUNCTION_ARG_ADVANCE
3635 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
3636 If defined, a C expression that is the number of bytes to add to the
3637 offset of the argument passed in memory. This is needed for the SPU,
3638 which passes @code{char} and @code{short} arguments in the preferred
3639 slot that is in the middle of the quad word instead of starting at the
3643 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3644 If defined, a C expression which determines whether, and in which direction,
3645 to pad out an argument with extra space. The value should be of type
3646 @code{enum direction}: either @code{upward} to pad above the argument,
3647 @code{downward} to pad below, or @code{none} to inhibit padding.
3649 The @emph{amount} of padding is not controlled by this macro, but by the
3650 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
3651 always just enough to reach the next multiple of that boundary.
3653 This macro has a default definition which is right for most systems.
3654 For little-endian machines, the default is to pad upward. For
3655 big-endian machines, the default is to pad downward for an argument of
3656 constant size shorter than an @code{int}, and upward otherwise.
3659 @defmac PAD_VARARGS_DOWN
3660 If defined, a C expression which determines whether the default
3661 implementation of va_arg will attempt to pad down before reading the
3662 next argument, if that argument is smaller than its aligned space as
3663 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3664 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3667 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3668 Specify padding for the last element of a block move between registers and
3669 memory. @var{first} is nonzero if this is the only element. Defining this
3670 macro allows better control of register function parameters on big-endian
3671 machines, without using @code{PARALLEL} rtl. In particular,
3672 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3673 registers, as there is no longer a "wrong" part of a register; For example,
3674 a three byte aggregate may be passed in the high part of a register if so
3678 @hook TARGET_FUNCTION_ARG_BOUNDARY
3680 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3682 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3683 A C expression that is nonzero if @var{regno} is the number of a hard
3684 register in which function arguments are sometimes passed. This does
3685 @emph{not} include implicit arguments such as the static chain and
3686 the structure-value address. On many machines, no registers can be
3687 used for this purpose since all function arguments are pushed on the
3691 @hook TARGET_SPLIT_COMPLEX_ARG
3693 @hook TARGET_BUILD_BUILTIN_VA_LIST
3695 @hook TARGET_ENUM_VA_LIST_P
3697 @hook TARGET_FN_ABI_VA_LIST
3699 @hook TARGET_FN_ABI_VA_LIST_BOUNDS_SIZE
3701 @hook TARGET_CANONICAL_VA_LIST_TYPE
3703 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
3705 @hook TARGET_VALID_POINTER_MODE
3707 @hook TARGET_REF_MAY_ALIAS_ERRNO
3709 @hook TARGET_SCALAR_MODE_SUPPORTED_P
3711 @hook TARGET_VECTOR_MODE_SUPPORTED_P
3713 @hook TARGET_ARRAY_MODE_SUPPORTED_P
3715 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3717 @hook TARGET_FLAGS_REGNUM
3720 @subsection How Scalar Function Values Are Returned
3721 @cindex return values in registers
3722 @cindex values, returned by functions
3723 @cindex scalars, returned as values
3725 This section discusses the macros that control returning scalars as
3726 values---values that can fit in registers.
3728 @hook TARGET_FUNCTION_VALUE
3730 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3731 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
3732 a new target instead.
3735 @defmac LIBCALL_VALUE (@var{mode})
3736 A C expression to create an RTX representing the place where a library
3737 function returns a value of mode @var{mode}.
3739 Note that ``library function'' in this context means a compiler
3740 support routine, used to perform arithmetic, whose name is known
3741 specially by the compiler and was not mentioned in the C code being
3745 @hook TARGET_LIBCALL_VALUE
3747 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3748 A C expression that is nonzero if @var{regno} is the number of a hard
3749 register in which the values of called function may come back.
3751 A register whose use for returning values is limited to serving as the
3752 second of a pair (for a value of type @code{double}, say) need not be
3753 recognized by this macro. So for most machines, this definition
3757 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3760 If the machine has register windows, so that the caller and the called
3761 function use different registers for the return value, this macro
3762 should recognize only the caller's register numbers.
3764 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3765 for a new target instead.
3768 @hook TARGET_FUNCTION_VALUE_REGNO_P
3770 @defmac APPLY_RESULT_SIZE
3771 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3772 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3773 saving and restoring an arbitrary return value.
3776 @hook TARGET_RETURN_IN_MSB
3778 @node Aggregate Return
3779 @subsection How Large Values Are Returned
3780 @cindex aggregates as return values
3781 @cindex large return values
3782 @cindex returning aggregate values
3783 @cindex structure value address
3785 When a function value's mode is @code{BLKmode} (and in some other
3786 cases), the value is not returned according to
3787 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
3788 caller passes the address of a block of memory in which the value
3789 should be stored. This address is called the @dfn{structure value
3792 This section describes how to control returning structure values in
3795 @hook TARGET_RETURN_IN_MEMORY
3797 @defmac DEFAULT_PCC_STRUCT_RETURN
3798 Define this macro to be 1 if all structure and union return values must be
3799 in memory. Since this results in slower code, this should be defined
3800 only if needed for compatibility with other compilers or with an ABI@.
3801 If you define this macro to be 0, then the conventions used for structure
3802 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3805 If not defined, this defaults to the value 1.
3808 @hook TARGET_STRUCT_VALUE_RTX
3810 @defmac PCC_STATIC_STRUCT_RETURN
3811 Define this macro if the usual system convention on the target machine
3812 for returning structures and unions is for the called function to return
3813 the address of a static variable containing the value.
3815 Do not define this if the usual system convention is for the caller to
3816 pass an address to the subroutine.
3818 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3819 nothing when you use @option{-freg-struct-return} mode.
3822 @hook TARGET_GET_RAW_RESULT_MODE
3824 @hook TARGET_GET_RAW_ARG_MODE
3827 @subsection Caller-Saves Register Allocation
3829 If you enable it, GCC can save registers around function calls. This
3830 makes it possible to use call-clobbered registers to hold variables that
3831 must live across calls.
3833 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3834 A C expression to determine whether it is worthwhile to consider placing
3835 a pseudo-register in a call-clobbered hard register and saving and
3836 restoring it around each function call. The expression should be 1 when
3837 this is worth doing, and 0 otherwise.
3839 If you don't define this macro, a default is used which is good on most
3840 machines: @code{4 * @var{calls} < @var{refs}}.
3843 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3844 A C expression specifying which mode is required for saving @var{nregs}
3845 of a pseudo-register in call-clobbered hard register @var{regno}. If
3846 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3847 returned. For most machines this macro need not be defined since GCC
3848 will select the smallest suitable mode.
3851 @node Function Entry
3852 @subsection Function Entry and Exit
3853 @cindex function entry and exit
3857 This section describes the macros that output function entry
3858 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3860 @hook TARGET_ASM_FUNCTION_PROLOGUE
3862 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
3864 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3866 @hook TARGET_ASM_FUNCTION_EPILOGUE
3870 @findex pretend_args_size
3871 @findex crtl->args.pretend_args_size
3872 A region of @code{crtl->args.pretend_args_size} bytes of
3873 uninitialized space just underneath the first argument arriving on the
3874 stack. (This may not be at the very start of the allocated stack region
3875 if the calling sequence has pushed anything else since pushing the stack
3876 arguments. But usually, on such machines, nothing else has been pushed
3877 yet, because the function prologue itself does all the pushing.) This
3878 region is used on machines where an argument may be passed partly in
3879 registers and partly in memory, and, in some cases to support the
3880 features in @code{<stdarg.h>}.
3883 An area of memory used to save certain registers used by the function.
3884 The size of this area, which may also include space for such things as
3885 the return address and pointers to previous stack frames, is
3886 machine-specific and usually depends on which registers have been used
3887 in the function. Machines with register windows often do not require
3891 A region of at least @var{size} bytes, possibly rounded up to an allocation
3892 boundary, to contain the local variables of the function. On some machines,
3893 this region and the save area may occur in the opposite order, with the
3894 save area closer to the top of the stack.
3897 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3898 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3899 @code{crtl->outgoing_args_size} bytes to be used for outgoing
3900 argument lists of the function. @xref{Stack Arguments}.
3903 @defmac EXIT_IGNORE_STACK
3904 Define this macro as a C expression that is nonzero if the return
3905 instruction or the function epilogue ignores the value of the stack
3906 pointer; in other words, if it is safe to delete an instruction to
3907 adjust the stack pointer before a return from the function. The
3910 Note that this macro's value is relevant only for functions for which
3911 frame pointers are maintained. It is never safe to delete a final
3912 stack adjustment in a function that has no frame pointer, and the
3913 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3916 @defmac EPILOGUE_USES (@var{regno})
3917 Define this macro as a C expression that is nonzero for registers that are
3918 used by the epilogue or the @samp{return} pattern. The stack and frame
3919 pointer registers are already assumed to be used as needed.
3922 @defmac EH_USES (@var{regno})
3923 Define this macro as a C expression that is nonzero for registers that are
3924 used by the exception handling mechanism, and so should be considered live
3925 on entry to an exception edge.
3928 @hook TARGET_ASM_OUTPUT_MI_THUNK
3930 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3933 @subsection Generating Code for Profiling
3934 @cindex profiling, code generation
3936 These macros will help you generate code for profiling.
3938 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3939 A C statement or compound statement to output to @var{file} some
3940 assembler code to call the profiling subroutine @code{mcount}.
3943 The details of how @code{mcount} expects to be called are determined by
3944 your operating system environment, not by GCC@. To figure them out,
3945 compile a small program for profiling using the system's installed C
3946 compiler and look at the assembler code that results.
3948 Older implementations of @code{mcount} expect the address of a counter
3949 variable to be loaded into some register. The name of this variable is
3950 @samp{LP} followed by the number @var{labelno}, so you would generate
3951 the name using @samp{LP%d} in a @code{fprintf}.
3954 @defmac PROFILE_HOOK
3955 A C statement or compound statement to output to @var{file} some assembly
3956 code to call the profiling subroutine @code{mcount} even the target does
3957 not support profiling.
3960 @defmac NO_PROFILE_COUNTERS
3961 Define this macro to be an expression with a nonzero value if the
3962 @code{mcount} subroutine on your system does not need a counter variable
3963 allocated for each function. This is true for almost all modern
3964 implementations. If you define this macro, you must not use the
3965 @var{labelno} argument to @code{FUNCTION_PROFILER}.
3968 @defmac PROFILE_BEFORE_PROLOGUE
3969 Define this macro if the code for function profiling should come before
3970 the function prologue. Normally, the profiling code comes after.
3974 @subsection Permitting tail calls
3977 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
3979 @hook TARGET_EXTRA_LIVE_ON_ENTRY
3981 @hook TARGET_SET_UP_BY_PROLOGUE
3983 @hook TARGET_WARN_FUNC_RETURN
3985 @node Stack Smashing Protection
3986 @subsection Stack smashing protection
3987 @cindex stack smashing protection
3989 @hook TARGET_STACK_PROTECT_GUARD
3991 @hook TARGET_STACK_PROTECT_FAIL
3993 @hook TARGET_SUPPORTS_SPLIT_STACK
3996 @section Implementing the Varargs Macros
3997 @cindex varargs implementation
3999 GCC comes with an implementation of @code{<varargs.h>} and
4000 @code{<stdarg.h>} that work without change on machines that pass arguments
4001 on the stack. Other machines require their own implementations of
4002 varargs, and the two machine independent header files must have
4003 conditionals to include it.
4005 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4006 the calling convention for @code{va_start}. The traditional
4007 implementation takes just one argument, which is the variable in which
4008 to store the argument pointer. The ISO implementation of
4009 @code{va_start} takes an additional second argument. The user is
4010 supposed to write the last named argument of the function here.
4012 However, @code{va_start} should not use this argument. The way to find
4013 the end of the named arguments is with the built-in functions described
4016 @defmac __builtin_saveregs ()
4017 Use this built-in function to save the argument registers in memory so
4018 that the varargs mechanism can access them. Both ISO and traditional
4019 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4020 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4022 On some machines, @code{__builtin_saveregs} is open-coded under the
4023 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4024 other machines, it calls a routine written in assembler language,
4025 found in @file{libgcc2.c}.
4027 Code generated for the call to @code{__builtin_saveregs} appears at the
4028 beginning of the function, as opposed to where the call to
4029 @code{__builtin_saveregs} is written, regardless of what the code is.
4030 This is because the registers must be saved before the function starts
4031 to use them for its own purposes.
4032 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4036 @defmac __builtin_next_arg (@var{lastarg})
4037 This builtin returns the address of the first anonymous stack
4038 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4039 returns the address of the location above the first anonymous stack
4040 argument. Use it in @code{va_start} to initialize the pointer for
4041 fetching arguments from the stack. Also use it in @code{va_start} to
4042 verify that the second parameter @var{lastarg} is the last named argument
4043 of the current function.
4046 @defmac __builtin_classify_type (@var{object})
4047 Since each machine has its own conventions for which data types are
4048 passed in which kind of register, your implementation of @code{va_arg}
4049 has to embody these conventions. The easiest way to categorize the
4050 specified data type is to use @code{__builtin_classify_type} together
4051 with @code{sizeof} and @code{__alignof__}.
4053 @code{__builtin_classify_type} ignores the value of @var{object},
4054 considering only its data type. It returns an integer describing what
4055 kind of type that is---integer, floating, pointer, structure, and so on.
4057 The file @file{typeclass.h} defines an enumeration that you can use to
4058 interpret the values of @code{__builtin_classify_type}.
4061 These machine description macros help implement varargs:
4063 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
4065 @hook TARGET_SETUP_INCOMING_VARARGS
4067 @hook TARGET_STRICT_ARGUMENT_NAMING
4069 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4071 @hook TARGET_LOAD_BOUNDS_FOR_ARG
4073 @hook TARGET_STORE_BOUNDS_FOR_ARG
4076 @section Trampolines for Nested Functions
4077 @cindex trampolines for nested functions
4078 @cindex nested functions, trampolines for
4080 A @dfn{trampoline} is a small piece of code that is created at run time
4081 when the address of a nested function is taken. It normally resides on
4082 the stack, in the stack frame of the containing function. These macros
4083 tell GCC how to generate code to allocate and initialize a
4086 The instructions in the trampoline must do two things: load a constant
4087 address into the static chain register, and jump to the real address of
4088 the nested function. On CISC machines such as the m68k, this requires
4089 two instructions, a move immediate and a jump. Then the two addresses
4090 exist in the trampoline as word-long immediate operands. On RISC
4091 machines, it is often necessary to load each address into a register in
4092 two parts. Then pieces of each address form separate immediate
4095 The code generated to initialize the trampoline must store the variable
4096 parts---the static chain value and the function address---into the
4097 immediate operands of the instructions. On a CISC machine, this is
4098 simply a matter of copying each address to a memory reference at the
4099 proper offset from the start of the trampoline. On a RISC machine, it
4100 may be necessary to take out pieces of the address and store them
4103 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
4105 @defmac TRAMPOLINE_SECTION
4106 Return the section into which the trampoline template is to be placed
4107 (@pxref{Sections}). The default value is @code{readonly_data_section}.
4110 @defmac TRAMPOLINE_SIZE
4111 A C expression for the size in bytes of the trampoline, as an integer.
4114 @defmac TRAMPOLINE_ALIGNMENT
4115 Alignment required for trampolines, in bits.
4117 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
4118 is used for aligning trampolines.
4121 @hook TARGET_TRAMPOLINE_INIT
4123 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
4125 Implementing trampolines is difficult on many machines because they have
4126 separate instruction and data caches. Writing into a stack location
4127 fails to clear the memory in the instruction cache, so when the program
4128 jumps to that location, it executes the old contents.
4130 Here are two possible solutions. One is to clear the relevant parts of
4131 the instruction cache whenever a trampoline is set up. The other is to
4132 make all trampolines identical, by having them jump to a standard
4133 subroutine. The former technique makes trampoline execution faster; the
4134 latter makes initialization faster.
4136 To clear the instruction cache when a trampoline is initialized, define
4137 the following macro.
4139 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4140 If defined, expands to a C expression clearing the @emph{instruction
4141 cache} in the specified interval. The definition of this macro would
4142 typically be a series of @code{asm} statements. Both @var{beg} and
4143 @var{end} are both pointer expressions.
4146 To use a standard subroutine, define the following macro. In addition,
4147 you must make sure that the instructions in a trampoline fill an entire
4148 cache line with identical instructions, or else ensure that the
4149 beginning of the trampoline code is always aligned at the same point in
4150 its cache line. Look in @file{m68k.h} as a guide.
4152 @defmac TRANSFER_FROM_TRAMPOLINE
4153 Define this macro if trampolines need a special subroutine to do their
4154 work. The macro should expand to a series of @code{asm} statements
4155 which will be compiled with GCC@. They go in a library function named
4156 @code{__transfer_from_trampoline}.
4158 If you need to avoid executing the ordinary prologue code of a compiled
4159 C function when you jump to the subroutine, you can do so by placing a
4160 special label of your own in the assembler code. Use one @code{asm}
4161 statement to generate an assembler label, and another to make the label
4162 global. Then trampolines can use that label to jump directly to your
4163 special assembler code.
4167 @section Implicit Calls to Library Routines
4168 @cindex library subroutine names
4169 @cindex @file{libgcc.a}
4171 @c prevent bad page break with this line
4172 Here is an explanation of implicit calls to library routines.
4174 @defmac DECLARE_LIBRARY_RENAMES
4175 This macro, if defined, should expand to a piece of C code that will get
4176 expanded when compiling functions for libgcc.a. It can be used to
4177 provide alternate names for GCC's internal library functions if there
4178 are ABI-mandated names that the compiler should provide.
4181 @findex set_optab_libfunc
4182 @findex init_one_libfunc
4183 @hook TARGET_INIT_LIBFUNCS
4185 @hook TARGET_LIBFUNC_GNU_PREFIX
4187 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4188 This macro should return @code{true} if the library routine that
4189 implements the floating point comparison operator @var{comparison} in
4190 mode @var{mode} will return a boolean, and @var{false} if it will
4193 GCC's own floating point libraries return tristates from the
4194 comparison operators, so the default returns false always. Most ports
4195 don't need to define this macro.
4198 @defmac TARGET_LIB_INT_CMP_BIASED
4199 This macro should evaluate to @code{true} if the integer comparison
4200 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4201 operand is smaller than the second, 1 to indicate that they are equal,
4202 and 2 to indicate that the first operand is greater than the second.
4203 If this macro evaluates to @code{false} the comparison functions return
4204 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4205 in @file{libgcc.a}, you do not need to define this macro.
4208 @cindex @code{EDOM}, implicit usage
4211 The value of @code{EDOM} on the target machine, as a C integer constant
4212 expression. If you don't define this macro, GCC does not attempt to
4213 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4214 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4217 If you do not define @code{TARGET_EDOM}, then compiled code reports
4218 domain errors by calling the library function and letting it report the
4219 error. If mathematical functions on your system use @code{matherr} when
4220 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4221 that @code{matherr} is used normally.
4224 @cindex @code{errno}, implicit usage
4225 @defmac GEN_ERRNO_RTX
4226 Define this macro as a C expression to create an rtl expression that
4227 refers to the global ``variable'' @code{errno}. (On certain systems,
4228 @code{errno} may not actually be a variable.) If you don't define this
4229 macro, a reasonable default is used.
4232 @hook TARGET_LIBC_HAS_FUNCTION
4234 @defmac NEXT_OBJC_RUNTIME
4235 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
4236 by default. This calling convention involves passing the object, the selector
4237 and the method arguments all at once to the method-lookup library function.
4238 This is the usual setting when targeting Darwin/Mac OS X systems, which have
4239 the NeXT runtime installed.
4241 If the macro is set to 0, the "GNU" Objective-C message sending convention
4242 will be used by default. This convention passes just the object and the
4243 selector to the method-lookup function, which returns a pointer to the method.
4245 In either case, it remains possible to select code-generation for the alternate
4246 scheme, by means of compiler command line switches.
4249 @node Addressing Modes
4250 @section Addressing Modes
4251 @cindex addressing modes
4253 @c prevent bad page break with this line
4254 This is about addressing modes.
4256 @defmac HAVE_PRE_INCREMENT
4257 @defmacx HAVE_PRE_DECREMENT
4258 @defmacx HAVE_POST_INCREMENT
4259 @defmacx HAVE_POST_DECREMENT
4260 A C expression that is nonzero if the machine supports pre-increment,
4261 pre-decrement, post-increment, or post-decrement addressing respectively.
4264 @defmac HAVE_PRE_MODIFY_DISP
4265 @defmacx HAVE_POST_MODIFY_DISP
4266 A C expression that is nonzero if the machine supports pre- or
4267 post-address side-effect generation involving constants other than
4268 the size of the memory operand.
4271 @defmac HAVE_PRE_MODIFY_REG
4272 @defmacx HAVE_POST_MODIFY_REG
4273 A C expression that is nonzero if the machine supports pre- or
4274 post-address side-effect generation involving a register displacement.
4277 @defmac CONSTANT_ADDRESS_P (@var{x})
4278 A C expression that is 1 if the RTX @var{x} is a constant which
4279 is a valid address. On most machines the default definition of
4280 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4281 is acceptable, but a few machines are more restrictive as to which
4282 constant addresses are supported.
4285 @defmac CONSTANT_P (@var{x})
4286 @code{CONSTANT_P}, which is defined by target-independent code,
4287 accepts integer-values expressions whose values are not explicitly
4288 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4289 expressions and @code{const} arithmetic expressions, in addition to
4290 @code{const_int} and @code{const_double} expressions.
4293 @defmac MAX_REGS_PER_ADDRESS
4294 A number, the maximum number of registers that can appear in a valid
4295 memory address. Note that it is up to you to specify a value equal to
4296 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4300 @hook TARGET_LEGITIMATE_ADDRESS_P
4302 @defmac TARGET_MEM_CONSTRAINT
4303 A single character to be used instead of the default @code{'m'}
4304 character for general memory addresses. This defines the constraint
4305 letter which matches the memory addresses accepted by
4306 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
4307 support new address formats in your back end without changing the
4308 semantics of the @code{'m'} constraint. This is necessary in order to
4309 preserve functionality of inline assembly constructs using the
4310 @code{'m'} constraint.
4313 @defmac FIND_BASE_TERM (@var{x})
4314 A C expression to determine the base term of address @var{x},
4315 or to provide a simplified version of @var{x} from which @file{alias.c}
4316 can easily find the base term. This macro is used in only two places:
4317 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
4319 It is always safe for this macro to not be defined. It exists so
4320 that alias analysis can understand machine-dependent addresses.
4322 The typical use of this macro is to handle addresses containing
4323 a label_ref or symbol_ref within an UNSPEC@.
4326 @hook TARGET_LEGITIMIZE_ADDRESS
4328 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4329 A C compound statement that attempts to replace @var{x}, which is an address
4330 that needs reloading, with a valid memory address for an operand of mode
4331 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4332 It is not necessary to define this macro, but it might be useful for
4333 performance reasons.
4335 For example, on the i386, it is sometimes possible to use a single
4336 reload register instead of two by reloading a sum of two pseudo
4337 registers into a register. On the other hand, for number of RISC
4338 processors offsets are limited so that often an intermediate address
4339 needs to be generated in order to address a stack slot. By defining
4340 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4341 generated for adjacent some stack slots can be made identical, and thus
4344 @emph{Note}: This macro should be used with caution. It is necessary
4345 to know something of how reload works in order to effectively use this,
4346 and it is quite easy to produce macros that build in too much knowledge
4347 of reload internals.
4349 @emph{Note}: This macro must be able to reload an address created by a
4350 previous invocation of this macro. If it fails to handle such addresses
4351 then the compiler may generate incorrect code or abort.
4354 The macro definition should use @code{push_reload} to indicate parts that
4355 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4356 suitable to be passed unaltered to @code{push_reload}.
4358 The code generated by this macro must not alter the substructure of
4359 @var{x}. If it transforms @var{x} into a more legitimate form, it
4360 should assign @var{x} (which will always be a C variable) a new value.
4361 This also applies to parts that you change indirectly by calling
4364 @findex strict_memory_address_p
4365 The macro definition may use @code{strict_memory_address_p} to test if
4366 the address has become legitimate.
4369 If you want to change only a part of @var{x}, one standard way of doing
4370 this is to use @code{copy_rtx}. Note, however, that it unshares only a
4371 single level of rtl. Thus, if the part to be changed is not at the
4372 top level, you'll need to replace first the top level.
4373 It is not necessary for this macro to come up with a legitimate
4374 address; but often a machine-dependent strategy can generate better code.
4377 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
4379 @hook TARGET_LEGITIMATE_CONSTANT_P
4381 @hook TARGET_DELEGITIMIZE_ADDRESS
4383 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4385 @hook TARGET_CANNOT_FORCE_CONST_MEM
4387 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4389 @hook TARGET_USE_BLOCKS_FOR_DECL_P
4391 @hook TARGET_BUILTIN_RECIPROCAL
4393 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4395 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4397 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4399 @hook TARGET_VECTORIZE_VEC_PERM_CONST_OK
4401 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
4403 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4405 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4407 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4409 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
4411 @hook TARGET_VECTORIZE_INIT_COST
4413 @hook TARGET_VECTORIZE_ADD_STMT_COST
4415 @hook TARGET_VECTORIZE_FINISH_COST
4417 @hook TARGET_VECTORIZE_DESTROY_COST_DATA
4419 @hook TARGET_VECTORIZE_BUILTIN_TM_LOAD
4421 @hook TARGET_VECTORIZE_BUILTIN_TM_STORE
4423 @hook TARGET_VECTORIZE_BUILTIN_GATHER
4425 @node Anchored Addresses
4426 @section Anchored Addresses
4427 @cindex anchored addresses
4428 @cindex @option{-fsection-anchors}
4430 GCC usually addresses every static object as a separate entity.
4431 For example, if we have:
4435 int foo (void) @{ return a + b + c; @}
4438 the code for @code{foo} will usually calculate three separate symbolic
4439 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
4440 it would be better to calculate just one symbolic address and access
4441 the three variables relative to it. The equivalent pseudocode would
4447 register int *xr = &x;
4448 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4452 (which isn't valid C). We refer to shared addresses like @code{x} as
4453 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
4455 The hooks below describe the target properties that GCC needs to know
4456 in order to make effective use of section anchors. It won't use
4457 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4458 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4460 @hook TARGET_MIN_ANCHOR_OFFSET
4462 @hook TARGET_MAX_ANCHOR_OFFSET
4464 @hook TARGET_ASM_OUTPUT_ANCHOR
4466 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4468 @node Condition Code
4469 @section Condition Code Status
4470 @cindex condition code status
4472 The macros in this section can be split in two families, according to the
4473 two ways of representing condition codes in GCC.
4475 The first representation is the so called @code{(cc0)} representation
4476 (@pxref{Jump Patterns}), where all instructions can have an implicit
4477 clobber of the condition codes. The second is the condition code
4478 register representation, which provides better schedulability for
4479 architectures that do have a condition code register, but on which
4480 most instructions do not affect it. The latter category includes
4483 The implicit clobbering poses a strong restriction on the placement of
4484 the definition and use of the condition code, which need to be in adjacent
4485 insns for machines using @code{(cc0)}. This can prevent important
4486 optimizations on some machines. For example, on the IBM RS/6000, there
4487 is a delay for taken branches unless the condition code register is set
4488 three instructions earlier than the conditional branch. The instruction
4489 scheduler cannot perform this optimization if it is not permitted to
4490 separate the definition and use of the condition code register.
4492 For this reason, it is possible and suggested to use a register to
4493 represent the condition code for new ports. If there is a specific
4494 condition code register in the machine, use a hard register. If the
4495 condition code or comparison result can be placed in any general register,
4496 or if there are multiple condition registers, use a pseudo register.
4497 Registers used to store the condition code value will usually have a mode
4498 that is in class @code{MODE_CC}.
4500 Alternatively, you can use @code{BImode} if the comparison operator is
4501 specified already in the compare instruction. In this case, you are not
4502 interested in most macros in this section.
4505 * CC0 Condition Codes:: Old style representation of condition codes.
4506 * MODE_CC Condition Codes:: Modern representation of condition codes.
4509 @node CC0 Condition Codes
4510 @subsection Representation of condition codes using @code{(cc0)}
4514 The file @file{conditions.h} defines a variable @code{cc_status} to
4515 describe how the condition code was computed (in case the interpretation of
4516 the condition code depends on the instruction that it was set by). This
4517 variable contains the RTL expressions on which the condition code is
4518 currently based, and several standard flags.
4520 Sometimes additional machine-specific flags must be defined in the machine
4521 description header file. It can also add additional machine-specific
4522 information by defining @code{CC_STATUS_MDEP}.
4524 @defmac CC_STATUS_MDEP
4525 C code for a data type which is used for declaring the @code{mdep}
4526 component of @code{cc_status}. It defaults to @code{int}.
4528 This macro is not used on machines that do not use @code{cc0}.
4531 @defmac CC_STATUS_MDEP_INIT
4532 A C expression to initialize the @code{mdep} field to ``empty''.
4533 The default definition does nothing, since most machines don't use
4534 the field anyway. If you want to use the field, you should probably
4535 define this macro to initialize it.
4537 This macro is not used on machines that do not use @code{cc0}.
4540 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4541 A C compound statement to set the components of @code{cc_status}
4542 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4543 this macro's responsibility to recognize insns that set the condition
4544 code as a byproduct of other activity as well as those that explicitly
4547 This macro is not used on machines that do not use @code{cc0}.
4549 If there are insns that do not set the condition code but do alter
4550 other machine registers, this macro must check to see whether they
4551 invalidate the expressions that the condition code is recorded as
4552 reflecting. For example, on the 68000, insns that store in address
4553 registers do not set the condition code, which means that usually
4554 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4555 insns. But suppose that the previous insn set the condition code
4556 based on location @samp{a4@@(102)} and the current insn stores a new
4557 value in @samp{a4}. Although the condition code is not changed by
4558 this, it will no longer be true that it reflects the contents of
4559 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4560 @code{cc_status} in this case to say that nothing is known about the
4561 condition code value.
4563 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4564 with the results of peephole optimization: insns whose patterns are
4565 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4566 constants which are just the operands. The RTL structure of these
4567 insns is not sufficient to indicate what the insns actually do. What
4568 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4569 @code{CC_STATUS_INIT}.
4571 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4572 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4573 @samp{cc}. This avoids having detailed information about patterns in
4574 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4577 @node MODE_CC Condition Codes
4578 @subsection Representation of condition codes using registers
4582 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4583 On many machines, the condition code may be produced by other instructions
4584 than compares, for example the branch can use directly the condition
4585 code set by a subtract instruction. However, on some machines
4586 when the condition code is set this way some bits (such as the overflow
4587 bit) are not set in the same way as a test instruction, so that a different
4588 branch instruction must be used for some conditional branches. When
4589 this happens, use the machine mode of the condition code register to
4590 record different formats of the condition code register. Modes can
4591 also be used to record which compare instruction (e.g. a signed or an
4592 unsigned comparison) produced the condition codes.
4594 If other modes than @code{CCmode} are required, add them to
4595 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4596 a mode given an operand of a compare. This is needed because the modes
4597 have to be chosen not only during RTL generation but also, for example,
4598 by instruction combination. The result of @code{SELECT_CC_MODE} should
4599 be consistent with the mode used in the patterns; for example to support
4600 the case of the add on the SPARC discussed above, we have the pattern
4604 [(set (reg:CC_NOOV 0)
4606 (plus:SI (match_operand:SI 0 "register_operand" "%r")
4607 (match_operand:SI 1 "arith_operand" "rI"))
4614 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
4615 for comparisons whose argument is a @code{plus}:
4618 #define SELECT_CC_MODE(OP,X,Y) \
4619 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4620 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4621 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4622 || GET_CODE (X) == NEG) \
4623 ? CC_NOOVmode : CCmode))
4626 Another reason to use modes is to retain information on which operands
4627 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4630 You should define this macro if and only if you define extra CC modes
4631 in @file{@var{machine}-modes.def}.
4634 @hook TARGET_CANONICALIZE_COMPARISON
4636 @defmac REVERSIBLE_CC_MODE (@var{mode})
4637 A C expression whose value is one if it is always safe to reverse a
4638 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4639 can ever return @var{mode} for a floating-point inequality comparison,
4640 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4642 You need not define this macro if it would always returns zero or if the
4643 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4644 For example, here is the definition used on the SPARC, where floating-point
4645 inequality comparisons are always given @code{CCFPEmode}:
4648 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4652 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
4653 A C expression whose value is reversed condition code of the @var{code} for
4654 comparison done in CC_MODE @var{mode}. The macro is used only in case
4655 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4656 machine has some non-standard way how to reverse certain conditionals. For
4657 instance in case all floating point conditions are non-trapping, compiler may
4658 freely convert unordered compares to ordered one. Then definition may look
4662 #define REVERSE_CONDITION(CODE, MODE) \
4663 ((MODE) != CCFPmode ? reverse_condition (CODE) \
4664 : reverse_condition_maybe_unordered (CODE))
4668 @hook TARGET_FIXED_CONDITION_CODE_REGS
4670 @hook TARGET_CC_MODES_COMPATIBLE
4673 @section Describing Relative Costs of Operations
4674 @cindex costs of instructions
4675 @cindex relative costs
4676 @cindex speed of instructions
4678 These macros let you describe the relative speed of various operations
4679 on the target machine.
4681 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4682 A C expression for the cost of moving data of mode @var{mode} from a
4683 register in class @var{from} to one in class @var{to}. The classes are
4684 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4685 value of 2 is the default; other values are interpreted relative to
4688 It is not required that the cost always equal 2 when @var{from} is the
4689 same as @var{to}; on some machines it is expensive to move between
4690 registers if they are not general registers.
4692 If reload sees an insn consisting of a single @code{set} between two
4693 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4694 classes returns a value of 2, reload does not check to ensure that the
4695 constraints of the insn are met. Setting a cost of other than 2 will
4696 allow reload to verify that the constraints are met. You should do this
4697 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4699 These macros are obsolete, new ports should use the target hook
4700 @code{TARGET_REGISTER_MOVE_COST} instead.
4703 @hook TARGET_REGISTER_MOVE_COST
4705 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4706 A C expression for the cost of moving data of mode @var{mode} between a
4707 register of class @var{class} and memory; @var{in} is zero if the value
4708 is to be written to memory, nonzero if it is to be read in. This cost
4709 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4710 registers and memory is more expensive than between two registers, you
4711 should define this macro to express the relative cost.
4713 If you do not define this macro, GCC uses a default cost of 4 plus
4714 the cost of copying via a secondary reload register, if one is
4715 needed. If your machine requires a secondary reload register to copy
4716 between memory and a register of @var{class} but the reload mechanism is
4717 more complex than copying via an intermediate, define this macro to
4718 reflect the actual cost of the move.
4720 GCC defines the function @code{memory_move_secondary_cost} if
4721 secondary reloads are needed. It computes the costs due to copying via
4722 a secondary register. If your machine copies from memory using a
4723 secondary register in the conventional way but the default base value of
4724 4 is not correct for your machine, define this macro to add some other
4725 value to the result of that function. The arguments to that function
4726 are the same as to this macro.
4728 These macros are obsolete, new ports should use the target hook
4729 @code{TARGET_MEMORY_MOVE_COST} instead.
4732 @hook TARGET_MEMORY_MOVE_COST
4734 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4735 A C expression for the cost of a branch instruction. A value of 1 is
4736 the default; other values are interpreted relative to that. Parameter
4737 @var{speed_p} is true when the branch in question should be optimized
4738 for speed. When it is false, @code{BRANCH_COST} should return a value
4739 optimal for code size rather than performance. @var{predictable_p} is
4740 true for well-predicted branches. On many architectures the
4741 @code{BRANCH_COST} can be reduced then.
4744 Here are additional macros which do not specify precise relative costs,
4745 but only that certain actions are more expensive than GCC would
4748 @defmac SLOW_BYTE_ACCESS
4749 Define this macro as a C expression which is nonzero if accessing less
4750 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4751 faster than accessing a word of memory, i.e., if such access
4752 require more than one instruction or if there is no difference in cost
4753 between byte and (aligned) word loads.
4755 When this macro is not defined, the compiler will access a field by
4756 finding the smallest containing object; when it is defined, a fullword
4757 load will be used if alignment permits. Unless bytes accesses are
4758 faster than word accesses, using word accesses is preferable since it
4759 may eliminate subsequent memory access if subsequent accesses occur to
4760 other fields in the same word of the structure, but to different bytes.
4763 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
4764 Define this macro to be the value 1 if memory accesses described by the
4765 @var{mode} and @var{alignment} parameters have a cost many times greater
4766 than aligned accesses, for example if they are emulated in a trap
4769 When this macro is nonzero, the compiler will act as if
4770 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
4771 moves. This can cause significantly more instructions to be produced.
4772 Therefore, do not set this macro nonzero if unaligned accesses only add a
4773 cycle or two to the time for a memory access.
4775 If the value of this macro is always zero, it need not be defined. If
4776 this macro is defined, it should produce a nonzero value when
4777 @code{STRICT_ALIGNMENT} is nonzero.
4780 @defmac MOVE_RATIO (@var{speed})
4781 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4782 which a sequence of insns should be generated instead of a
4783 string move insn or a library call. Increasing the value will always
4784 make code faster, but eventually incurs high cost in increased code size.
4786 Note that on machines where the corresponding move insn is a
4787 @code{define_expand} that emits a sequence of insns, this macro counts
4788 the number of such sequences.
4790 The parameter @var{speed} is true if the code is currently being
4791 optimized for speed rather than size.
4793 If you don't define this, a reasonable default is used.
4796 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
4797 A C expression used to determine whether @code{move_by_pieces} will be used to
4798 copy a chunk of memory, or whether some other block move mechanism
4799 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4800 than @code{MOVE_RATIO}.
4803 @defmac MOVE_MAX_PIECES
4804 A C expression used by @code{move_by_pieces} to determine the largest unit
4805 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4808 @defmac CLEAR_RATIO (@var{speed})
4809 The threshold of number of scalar move insns, @emph{below} which a sequence
4810 of insns should be generated to clear memory instead of a string clear insn
4811 or a library call. Increasing the value will always make code faster, but
4812 eventually incurs high cost in increased code size.
4814 The parameter @var{speed} is true if the code is currently being
4815 optimized for speed rather than size.
4817 If you don't define this, a reasonable default is used.
4820 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
4821 A C expression used to determine whether @code{clear_by_pieces} will be used
4822 to clear a chunk of memory, or whether some other block clear mechanism
4823 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4824 than @code{CLEAR_RATIO}.
4827 @defmac SET_RATIO (@var{speed})
4828 The threshold of number of scalar move insns, @emph{below} which a sequence
4829 of insns should be generated to set memory to a constant value, instead of
4830 a block set insn or a library call.
4831 Increasing the value will always make code faster, but
4832 eventually incurs high cost in increased code size.
4834 The parameter @var{speed} is true if the code is currently being
4835 optimized for speed rather than size.
4837 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
4840 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
4841 A C expression used to determine whether @code{store_by_pieces} will be
4842 used to set a chunk of memory to a constant value, or whether some
4843 other mechanism will be used. Used by @code{__builtin_memset} when
4844 storing values other than constant zero.
4845 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4846 than @code{SET_RATIO}.
4849 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
4850 A C expression used to determine whether @code{store_by_pieces} will be
4851 used to set a chunk of memory to a constant string value, or whether some
4852 other mechanism will be used. Used by @code{__builtin_strcpy} when
4853 called with a constant source string.
4854 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4855 than @code{MOVE_RATIO}.
4858 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
4859 A C expression used to determine whether a load postincrement is a good
4860 thing to use for a given mode. Defaults to the value of
4861 @code{HAVE_POST_INCREMENT}.
4864 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
4865 A C expression used to determine whether a load postdecrement is a good
4866 thing to use for a given mode. Defaults to the value of
4867 @code{HAVE_POST_DECREMENT}.
4870 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4871 A C expression used to determine whether a load preincrement is a good
4872 thing to use for a given mode. Defaults to the value of
4873 @code{HAVE_PRE_INCREMENT}.
4876 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4877 A C expression used to determine whether a load predecrement is a good
4878 thing to use for a given mode. Defaults to the value of
4879 @code{HAVE_PRE_DECREMENT}.
4882 @defmac USE_STORE_POST_INCREMENT (@var{mode})
4883 A C expression used to determine whether a store postincrement is a good
4884 thing to use for a given mode. Defaults to the value of
4885 @code{HAVE_POST_INCREMENT}.
4888 @defmac USE_STORE_POST_DECREMENT (@var{mode})
4889 A C expression used to determine whether a store postdecrement is a good
4890 thing to use for a given mode. Defaults to the value of
4891 @code{HAVE_POST_DECREMENT}.
4894 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
4895 This macro is used to determine whether a store preincrement is a good
4896 thing to use for a given mode. Defaults to the value of
4897 @code{HAVE_PRE_INCREMENT}.
4900 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
4901 This macro is used to determine whether a store predecrement is a good
4902 thing to use for a given mode. Defaults to the value of
4903 @code{HAVE_PRE_DECREMENT}.
4906 @defmac NO_FUNCTION_CSE
4907 Define this macro if it is as good or better to call a constant
4908 function address than to call an address kept in a register.
4911 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4912 Define this macro if a non-short-circuit operation produced by
4913 @samp{fold_range_test ()} is optimal. This macro defaults to true if
4914 @code{BRANCH_COST} is greater than or equal to the value 2.
4917 @hook TARGET_RTX_COSTS
4919 @hook TARGET_ADDRESS_COST
4922 @section Adjusting the Instruction Scheduler
4924 The instruction scheduler may need a fair amount of machine-specific
4925 adjustment in order to produce good code. GCC provides several target
4926 hooks for this purpose. It is usually enough to define just a few of
4927 them: try the first ones in this list first.
4929 @hook TARGET_SCHED_ISSUE_RATE
4931 @hook TARGET_SCHED_VARIABLE_ISSUE
4933 @hook TARGET_SCHED_ADJUST_COST
4935 @hook TARGET_SCHED_ADJUST_PRIORITY
4937 @hook TARGET_SCHED_REORDER
4939 @hook TARGET_SCHED_REORDER2
4941 @hook TARGET_SCHED_MACRO_FUSION_P
4943 @hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4945 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4947 @hook TARGET_SCHED_INIT
4949 @hook TARGET_SCHED_FINISH
4951 @hook TARGET_SCHED_INIT_GLOBAL
4953 @hook TARGET_SCHED_FINISH_GLOBAL
4955 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4957 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4959 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4961 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4963 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4965 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4967 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4969 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4971 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4973 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4975 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4977 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4979 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4981 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4983 @hook TARGET_SCHED_DFA_NEW_CYCLE
4985 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4987 @hook TARGET_SCHED_H_I_D_EXTENDED
4989 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4991 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
4993 @hook TARGET_SCHED_SET_SCHED_CONTEXT
4995 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4997 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
4999 @hook TARGET_SCHED_SPECULATE_INSN
5001 @hook TARGET_SCHED_NEEDS_BLOCK_P
5003 @hook TARGET_SCHED_GEN_SPEC_CHECK
5005 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC
5007 @hook TARGET_SCHED_SET_SCHED_FLAGS
5009 @hook TARGET_SCHED_SMS_RES_MII
5011 @hook TARGET_SCHED_DISPATCH
5013 @hook TARGET_SCHED_DISPATCH_DO
5015 @hook TARGET_SCHED_EXPOSED_PIPELINE
5017 @hook TARGET_SCHED_REASSOCIATION_WIDTH
5020 @section Dividing the Output into Sections (Texts, Data, @dots{})
5021 @c the above section title is WAY too long. maybe cut the part between
5022 @c the (...)? --mew 10feb93
5024 An object file is divided into sections containing different types of
5025 data. In the most common case, there are three sections: the @dfn{text
5026 section}, which holds instructions and read-only data; the @dfn{data
5027 section}, which holds initialized writable data; and the @dfn{bss
5028 section}, which holds uninitialized data. Some systems have other kinds
5031 @file{varasm.c} provides several well-known sections, such as
5032 @code{text_section}, @code{data_section} and @code{bss_section}.
5033 The normal way of controlling a @code{@var{foo}_section} variable
5034 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
5035 as described below. The macros are only read once, when @file{varasm.c}
5036 initializes itself, so their values must be run-time constants.
5037 They may however depend on command-line flags.
5039 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
5040 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
5041 to be string literals.
5043 Some assemblers require a different string to be written every time a
5044 section is selected. If your assembler falls into this category, you
5045 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
5046 @code{get_unnamed_section} to set up the sections.
5048 You must always create a @code{text_section}, either by defining
5049 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
5050 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
5051 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
5052 create a distinct @code{readonly_data_section}, the default is to
5053 reuse @code{text_section}.
5055 All the other @file{varasm.c} sections are optional, and are null
5056 if the target does not provide them.
5058 @defmac TEXT_SECTION_ASM_OP
5059 A C expression whose value is a string, including spacing, containing the
5060 assembler operation that should precede instructions and read-only data.
5061 Normally @code{"\t.text"} is right.
5064 @defmac HOT_TEXT_SECTION_NAME
5065 If defined, a C string constant for the name of the section containing most
5066 frequently executed functions of the program. If not defined, GCC will provide
5067 a default definition if the target supports named sections.
5070 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5071 If defined, a C string constant for the name of the section containing unlikely
5072 executed functions in the program.
5075 @defmac DATA_SECTION_ASM_OP
5076 A C expression whose value is a string, including spacing, containing the
5077 assembler operation to identify the following data as writable initialized
5078 data. Normally @code{"\t.data"} is right.
5081 @defmac SDATA_SECTION_ASM_OP
5082 If defined, a C expression whose value is a string, including spacing,
5083 containing the assembler operation to identify the following data as
5084 initialized, writable small data.
5087 @defmac READONLY_DATA_SECTION_ASM_OP
5088 A C expression whose value is a string, including spacing, containing the
5089 assembler operation to identify the following data as read-only initialized
5093 @defmac BSS_SECTION_ASM_OP
5094 If defined, a C expression whose value is a string, including spacing,
5095 containing the assembler operation to identify the following data as
5096 uninitialized global data. If not defined, and
5097 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
5098 uninitialized global data will be output in the data section if
5099 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5103 @defmac SBSS_SECTION_ASM_OP
5104 If defined, a C expression whose value is a string, including spacing,
5105 containing the assembler operation to identify the following data as
5106 uninitialized, writable small data.
5109 @defmac TLS_COMMON_ASM_OP
5110 If defined, a C expression whose value is a string containing the
5111 assembler operation to identify the following data as thread-local
5112 common data. The default is @code{".tls_common"}.
5115 @defmac TLS_SECTION_ASM_FLAG
5116 If defined, a C expression whose value is a character constant
5117 containing the flag used to mark a section as a TLS section. The
5118 default is @code{'T'}.
5121 @defmac INIT_SECTION_ASM_OP
5122 If defined, a C expression whose value is a string, including spacing,
5123 containing the assembler operation to identify the following data as
5124 initialization code. If not defined, GCC will assume such a section does
5125 not exist. This section has no corresponding @code{init_section}
5126 variable; it is used entirely in runtime code.
5129 @defmac FINI_SECTION_ASM_OP
5130 If defined, a C expression whose value is a string, including spacing,
5131 containing the assembler operation to identify the following data as
5132 finalization code. If not defined, GCC will assume such a section does
5133 not exist. This section has no corresponding @code{fini_section}
5134 variable; it is used entirely in runtime code.
5137 @defmac INIT_ARRAY_SECTION_ASM_OP
5138 If defined, a C expression whose value is a string, including spacing,
5139 containing the assembler operation to identify the following data as
5140 part of the @code{.init_array} (or equivalent) section. If not
5141 defined, GCC will assume such a section does not exist. Do not define
5142 both this macro and @code{INIT_SECTION_ASM_OP}.
5145 @defmac FINI_ARRAY_SECTION_ASM_OP
5146 If defined, a C expression whose value is a string, including spacing,
5147 containing the assembler operation to identify the following data as
5148 part of the @code{.fini_array} (or equivalent) section. If not
5149 defined, GCC will assume such a section does not exist. Do not define
5150 both this macro and @code{FINI_SECTION_ASM_OP}.
5153 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5154 If defined, an ASM statement that switches to a different section
5155 via @var{section_op}, calls @var{function}, and switches back to
5156 the text section. This is used in @file{crtstuff.c} if
5157 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5158 to initialization and finalization functions from the init and fini
5159 sections. By default, this macro uses a simple function call. Some
5160 ports need hand-crafted assembly code to avoid dependencies on
5161 registers initialized in the function prologue or to ensure that
5162 constant pools don't end up too far way in the text section.
5165 @defmac TARGET_LIBGCC_SDATA_SECTION
5166 If defined, a string which names the section into which small
5167 variables defined in crtstuff and libgcc should go. This is useful
5168 when the target has options for optimizing access to small data, and
5169 you want the crtstuff and libgcc routines to be conservative in what
5170 they expect of your application yet liberal in what your application
5171 expects. For example, for targets with a @code{.sdata} section (like
5172 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
5173 require small data support from your application, but use this macro
5174 to put small data into @code{.sdata} so that your application can
5175 access these variables whether it uses small data or not.
5178 @defmac FORCE_CODE_SECTION_ALIGN
5179 If defined, an ASM statement that aligns a code section to some
5180 arbitrary boundary. This is used to force all fragments of the
5181 @code{.init} and @code{.fini} sections to have to same alignment
5182 and thus prevent the linker from having to add any padding.
5185 @defmac JUMP_TABLES_IN_TEXT_SECTION
5186 Define this macro to be an expression with a nonzero value if jump
5187 tables (for @code{tablejump} insns) should be output in the text
5188 section, along with the assembler instructions. Otherwise, the
5189 readonly data section is used.
5191 This macro is irrelevant if there is no separate readonly data section.
5194 @hook TARGET_ASM_INIT_SECTIONS
5196 @hook TARGET_ASM_RELOC_RW_MASK
5198 @hook TARGET_ASM_SELECT_SECTION
5200 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
5201 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
5202 for @code{FUNCTION_DECL}s as well as for variables and constants.
5204 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
5205 function has been determined to be likely to be called, and nonzero if
5206 it is unlikely to be called.
5209 @hook TARGET_ASM_UNIQUE_SECTION
5211 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
5213 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
5215 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
5217 @hook TARGET_ASM_SELECT_RTX_SECTION
5219 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
5221 @hook TARGET_ENCODE_SECTION_INFO
5223 @hook TARGET_STRIP_NAME_ENCODING
5225 @hook TARGET_IN_SMALL_DATA_P
5227 @hook TARGET_HAVE_SRODATA_SECTION
5229 @hook TARGET_PROFILE_BEFORE_PROLOGUE
5231 @hook TARGET_BINDS_LOCAL_P
5233 @hook TARGET_HAVE_TLS
5237 @section Position Independent Code
5238 @cindex position independent code
5241 This section describes macros that help implement generation of position
5242 independent code. Simply defining these macros is not enough to
5243 generate valid PIC; you must also add support to the hook
5244 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
5245 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
5246 must modify the definition of @samp{movsi} to do something appropriate
5247 when the source operand contains a symbolic address. You may also
5248 need to alter the handling of switch statements so that they use
5250 @c i rearranged the order of the macros above to try to force one of
5251 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5253 @defmac PIC_OFFSET_TABLE_REGNUM
5254 The register number of the register used to address a table of static
5255 data addresses in memory. In some cases this register is defined by a
5256 processor's ``application binary interface'' (ABI)@. When this macro
5257 is defined, RTL is generated for this register once, as with the stack
5258 pointer and frame pointer registers. If this macro is not defined, it
5259 is up to the machine-dependent files to allocate such a register (if
5260 necessary). Note that this register must be fixed when in use (e.g.@:
5261 when @code{flag_pic} is true).
5264 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5265 A C expression that is nonzero if the register defined by
5266 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
5267 the default is zero. Do not define
5268 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5271 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
5272 A C expression that is nonzero if @var{x} is a legitimate immediate
5273 operand on the target machine when generating position independent code.
5274 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5275 check this. You can also assume @var{flag_pic} is true, so you need not
5276 check it either. You need not define this macro if all constants
5277 (including @code{SYMBOL_REF}) can be immediate operands when generating
5278 position independent code.
5281 @node Assembler Format
5282 @section Defining the Output Assembler Language
5284 This section describes macros whose principal purpose is to describe how
5285 to write instructions in assembler language---rather than what the
5289 * File Framework:: Structural information for the assembler file.
5290 * Data Output:: Output of constants (numbers, strings, addresses).
5291 * Uninitialized Data:: Output of uninitialized variables.
5292 * Label Output:: Output and generation of labels.
5293 * Initialization:: General principles of initialization
5294 and termination routines.
5295 * Macros for Initialization::
5296 Specific macros that control the handling of
5297 initialization and termination routines.
5298 * Instruction Output:: Output of actual instructions.
5299 * Dispatch Tables:: Output of jump tables.
5300 * Exception Region Output:: Output of exception region code.
5301 * Alignment Output:: Pseudo ops for alignment and skipping data.
5304 @node File Framework
5305 @subsection The Overall Framework of an Assembler File
5306 @cindex assembler format
5307 @cindex output of assembler code
5309 @c prevent bad page break with this line
5310 This describes the overall framework of an assembly file.
5312 @findex default_file_start
5313 @hook TARGET_ASM_FILE_START
5315 @hook TARGET_ASM_FILE_START_APP_OFF
5317 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5319 @hook TARGET_ASM_FILE_END
5321 @deftypefun void file_end_indicate_exec_stack ()
5322 Some systems use a common convention, the @samp{.note.GNU-stack}
5323 special section, to indicate whether or not an object file relies on
5324 the stack being executable. If your system uses this convention, you
5325 should define @code{TARGET_ASM_FILE_END} to this function. If you
5326 need to do other things in that hook, have your hook function call
5330 @hook TARGET_ASM_LTO_START
5332 @hook TARGET_ASM_LTO_END
5334 @hook TARGET_ASM_CODE_END
5336 @defmac ASM_COMMENT_START
5337 A C string constant describing how to begin a comment in the target
5338 assembler language. The compiler assumes that the comment will end at
5339 the end of the line.
5343 A C string constant for text to be output before each @code{asm}
5344 statement or group of consecutive ones. Normally this is
5345 @code{"#APP"}, which is a comment that has no effect on most
5346 assemblers but tells the GNU assembler that it must check the lines
5347 that follow for all valid assembler constructs.
5351 A C string constant for text to be output after each @code{asm}
5352 statement or group of consecutive ones. Normally this is
5353 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5354 time-saving assumptions that are valid for ordinary compiler output.
5357 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5358 A C statement to output COFF information or DWARF debugging information
5359 which indicates that filename @var{name} is the current source file to
5360 the stdio stream @var{stream}.
5362 This macro need not be defined if the standard form of output
5363 for the file format in use is appropriate.
5366 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5368 @hook TARGET_ASM_OUTPUT_IDENT
5370 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5371 A C statement to output the string @var{string} to the stdio stream
5372 @var{stream}. If you do not call the function @code{output_quoted_string}
5373 in your config files, GCC will only call it to output filenames to
5374 the assembler source. So you can use it to canonicalize the format
5375 of the filename using this macro.
5378 @hook TARGET_ASM_NAMED_SECTION
5380 @hook TARGET_ASM_FUNCTION_SECTION
5382 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5384 @hook TARGET_HAVE_NAMED_SECTIONS
5385 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5386 It must not be modified by command-line option processing.
5389 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5390 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5392 @hook TARGET_SECTION_TYPE_FLAGS
5394 @hook TARGET_ASM_RECORD_GCC_SWITCHES
5396 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5400 @subsection Output of Data
5403 @hook TARGET_ASM_BYTE_OP
5405 @hook TARGET_ASM_INTEGER
5407 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5409 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5410 A C statement to output to the stdio stream @var{stream} an assembler
5411 instruction to assemble a string constant containing the @var{len}
5412 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5413 @code{char *} and @var{len} a C expression of type @code{int}.
5415 If the assembler has a @code{.ascii} pseudo-op as found in the
5416 Berkeley Unix assembler, do not define the macro
5417 @code{ASM_OUTPUT_ASCII}.
5420 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5421 A C statement to output word @var{n} of a function descriptor for
5422 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5423 is defined, and is otherwise unused.
5426 @defmac CONSTANT_POOL_BEFORE_FUNCTION
5427 You may define this macro as a C expression. You should define the
5428 expression to have a nonzero value if GCC should output the constant
5429 pool for a function before the code for the function, or a zero value if
5430 GCC should output the constant pool after the function. If you do
5431 not define this macro, the usual case, GCC will output the constant
5432 pool before the function.
5435 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5436 A C statement to output assembler commands to define the start of the
5437 constant pool for a function. @var{funname} is a string giving
5438 the name of the function. Should the return type of the function
5439 be required, it can be obtained via @var{fundecl}. @var{size}
5440 is the size, in bytes, of the constant pool that will be written
5441 immediately after this call.
5443 If no constant-pool prefix is required, the usual case, this macro need
5447 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5448 A C statement (with or without semicolon) to output a constant in the
5449 constant pool, if it needs special treatment. (This macro need not do
5450 anything for RTL expressions that can be output normally.)
5452 The argument @var{file} is the standard I/O stream to output the
5453 assembler code on. @var{x} is the RTL expression for the constant to
5454 output, and @var{mode} is the machine mode (in case @var{x} is a
5455 @samp{const_int}). @var{align} is the required alignment for the value
5456 @var{x}; you should output an assembler directive to force this much
5459 The argument @var{labelno} is a number to use in an internal label for
5460 the address of this pool entry. The definition of this macro is
5461 responsible for outputting the label definition at the proper place.
5462 Here is how to do this:
5465 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5468 When you output a pool entry specially, you should end with a
5469 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5470 entry from being output a second time in the usual manner.
5472 You need not define this macro if it would do nothing.
5475 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5476 A C statement to output assembler commands to at the end of the constant
5477 pool for a function. @var{funname} is a string giving the name of the
5478 function. Should the return type of the function be required, you can
5479 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5480 constant pool that GCC wrote immediately before this call.
5482 If no constant-pool epilogue is required, the usual case, you need not
5486 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5487 Define this macro as a C expression which is nonzero if @var{C} is
5488 used as a logical line separator by the assembler. @var{STR} points
5489 to the position in the string where @var{C} was found; this can be used if
5490 a line separator uses multiple characters.
5492 If you do not define this macro, the default is that only
5493 the character @samp{;} is treated as a logical line separator.
5496 @hook TARGET_ASM_OPEN_PAREN
5498 These macros are provided by @file{real.h} for writing the definitions
5499 of @code{ASM_OUTPUT_DOUBLE} and the like:
5501 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5502 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5503 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5504 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5505 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5506 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5507 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5508 target's floating point representation, and store its bit pattern in
5509 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5510 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5511 simple @code{long int}. For the others, it should be an array of
5512 @code{long int}. The number of elements in this array is determined
5513 by the size of the desired target floating point data type: 32 bits of
5514 it go in each @code{long int} array element. Each array element holds
5515 32 bits of the result, even if @code{long int} is wider than 32 bits
5516 on the host machine.
5518 The array element values are designed so that you can print them out
5519 using @code{fprintf} in the order they should appear in the target
5523 @node Uninitialized Data
5524 @subsection Output of Uninitialized Variables
5526 Each of the macros in this section is used to do the whole job of
5527 outputting a single uninitialized variable.
5529 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5530 A C statement (sans semicolon) to output to the stdio stream
5531 @var{stream} the assembler definition of a common-label named
5532 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5533 is the size rounded up to whatever alignment the caller wants. It is
5534 possible that @var{size} may be zero, for instance if a struct with no
5535 other member than a zero-length array is defined. In this case, the
5536 backend must output a symbol definition that allocates at least one
5537 byte, both so that the address of the resulting object does not compare
5538 equal to any other, and because some object formats cannot even express
5539 the concept of a zero-sized common symbol, as that is how they represent
5540 an ordinary undefined external.
5542 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5543 output the name itself; before and after that, output the additional
5544 assembler syntax for defining the name, and a newline.
5546 This macro controls how the assembler definitions of uninitialized
5547 common global variables are output.
5550 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5551 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5552 separate, explicit argument. If you define this macro, it is used in
5553 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5554 handling the required alignment of the variable. The alignment is specified
5555 as the number of bits.
5558 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5559 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5560 variable to be output, if there is one, or @code{NULL_TREE} if there
5561 is no corresponding variable. If you define this macro, GCC will use it
5562 in place of both @code{ASM_OUTPUT_COMMON} and
5563 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5564 the variable's decl in order to chose what to output.
5567 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5568 A C statement (sans semicolon) to output to the stdio stream
5569 @var{stream} the assembler definition of uninitialized global @var{decl} named
5570 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
5571 is the alignment specified as the number of bits.
5573 Try to use function @code{asm_output_aligned_bss} defined in file
5574 @file{varasm.c} when defining this macro. If unable, use the expression
5575 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5576 before and after that, output the additional assembler syntax for defining
5577 the name, and a newline.
5579 There are two ways of handling global BSS@. One is to define this macro.
5580 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5581 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5582 You do not need to do both.
5584 Some languages do not have @code{common} data, and require a
5585 non-common form of global BSS in order to handle uninitialized globals
5586 efficiently. C++ is one example of this. However, if the target does
5587 not support global BSS, the front end may choose to make globals
5588 common in order to save space in the object file.
5591 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5592 A C statement (sans semicolon) to output to the stdio stream
5593 @var{stream} the assembler definition of a local-common-label named
5594 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5595 is the size rounded up to whatever alignment the caller wants.
5597 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5598 output the name itself; before and after that, output the additional
5599 assembler syntax for defining the name, and a newline.
5601 This macro controls how the assembler definitions of uninitialized
5602 static variables are output.
5605 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5606 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5607 separate, explicit argument. If you define this macro, it is used in
5608 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5609 handling the required alignment of the variable. The alignment is specified
5610 as the number of bits.
5613 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5614 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5615 variable to be output, if there is one, or @code{NULL_TREE} if there
5616 is no corresponding variable. If you define this macro, GCC will use it
5617 in place of both @code{ASM_OUTPUT_DECL} and
5618 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5619 the variable's decl in order to chose what to output.
5623 @subsection Output and Generation of Labels
5625 @c prevent bad page break with this line
5626 This is about outputting labels.
5628 @findex assemble_name
5629 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5630 A C statement (sans semicolon) to output to the stdio stream
5631 @var{stream} the assembler definition of a label named @var{name}.
5632 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5633 output the name itself; before and after that, output the additional
5634 assembler syntax for defining the name, and a newline. A default
5635 definition of this macro is provided which is correct for most systems.
5638 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5639 A C statement (sans semicolon) to output to the stdio stream
5640 @var{stream} the assembler definition of a label named @var{name} of
5642 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5643 output the name itself; before and after that, output the additional
5644 assembler syntax for defining the name, and a newline. A default
5645 definition of this macro is provided which is correct for most systems.
5647 If this macro is not defined, then the function name is defined in the
5648 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5651 @findex assemble_name_raw
5652 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5653 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5654 to refer to a compiler-generated label. The default definition uses
5655 @code{assemble_name_raw}, which is like @code{assemble_name} except
5656 that it is more efficient.
5660 A C string containing the appropriate assembler directive to specify the
5661 size of a symbol, without any arguments. On systems that use ELF, the
5662 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5663 systems, the default is not to define this macro.
5665 Define this macro only if it is correct to use the default definitions
5666 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5667 for your system. If you need your own custom definitions of those
5668 macros, or if you do not need explicit symbol sizes at all, do not
5672 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5673 A C statement (sans semicolon) to output to the stdio stream
5674 @var{stream} a directive telling the assembler that the size of the
5675 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
5676 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5680 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5681 A C statement (sans semicolon) to output to the stdio stream
5682 @var{stream} a directive telling the assembler to calculate the size of
5683 the symbol @var{name} by subtracting its address from the current
5686 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5687 provided. The default assumes that the assembler recognizes a special
5688 @samp{.} symbol as referring to the current address, and can calculate
5689 the difference between this and another symbol. If your assembler does
5690 not recognize @samp{.} or cannot do calculations with it, you will need
5691 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5694 @defmac NO_DOLLAR_IN_LABEL
5695 Define this macro if the assembler does not accept the character
5696 @samp{$} in label names. By default constructors and destructors in
5697 G++ have @samp{$} in the identifiers. If this macro is defined,
5698 @samp{.} is used instead.
5701 @defmac NO_DOT_IN_LABEL
5702 Define this macro if the assembler does not accept the character
5703 @samp{.} in label names. By default constructors and destructors in G++
5704 have names that use @samp{.}. If this macro is defined, these names
5705 are rewritten to avoid @samp{.}.
5709 A C string containing the appropriate assembler directive to specify the
5710 type of a symbol, without any arguments. On systems that use ELF, the
5711 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5712 systems, the default is not to define this macro.
5714 Define this macro only if it is correct to use the default definition of
5715 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5716 custom definition of this macro, or if you do not need explicit symbol
5717 types at all, do not define this macro.
5720 @defmac TYPE_OPERAND_FMT
5721 A C string which specifies (using @code{printf} syntax) the format of
5722 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
5723 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5724 the default is not to define this macro.
5726 Define this macro only if it is correct to use the default definition of
5727 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5728 custom definition of this macro, or if you do not need explicit symbol
5729 types at all, do not define this macro.
5732 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5733 A C statement (sans semicolon) to output to the stdio stream
5734 @var{stream} a directive telling the assembler that the type of the
5735 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
5736 that string is always either @samp{"function"} or @samp{"object"}, but
5737 you should not count on this.
5739 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5740 definition of this macro is provided.
5743 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5744 A C statement (sans semicolon) to output to the stdio stream
5745 @var{stream} any text necessary for declaring the name @var{name} of a
5746 function which is being defined. This macro is responsible for
5747 outputting the label definition (perhaps using
5748 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5749 @code{FUNCTION_DECL} tree node representing the function.
5751 If this macro is not defined, then the function name is defined in the
5752 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5754 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5758 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5759 A C statement (sans semicolon) to output to the stdio stream
5760 @var{stream} any text necessary for declaring the size of a function
5761 which is being defined. The argument @var{name} is the name of the
5762 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5763 representing the function.
5765 If this macro is not defined, then the function size is not defined.
5767 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5771 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5772 A C statement (sans semicolon) to output to the stdio stream
5773 @var{stream} any text necessary for declaring the name @var{name} of an
5774 initialized variable which is being defined. This macro must output the
5775 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5776 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5778 If this macro is not defined, then the variable name is defined in the
5779 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5781 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5782 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5785 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
5787 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5788 A C statement (sans semicolon) to output to the stdio stream
5789 @var{stream} any text necessary for claiming a register @var{regno}
5790 for a global variable @var{decl} with name @var{name}.
5792 If you don't define this macro, that is equivalent to defining it to do
5796 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5797 A C statement (sans semicolon) to finish up declaring a variable name
5798 once the compiler has processed its initializer fully and thus has had a
5799 chance to determine the size of an array when controlled by an
5800 initializer. This is used on systems where it's necessary to declare
5801 something about the size of the object.
5803 If you don't define this macro, that is equivalent to defining it to do
5806 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5807 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5810 @hook TARGET_ASM_GLOBALIZE_LABEL
5812 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
5814 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5815 A C statement (sans semicolon) to output to the stdio stream
5816 @var{stream} some commands that will make the label @var{name} weak;
5817 that is, available for reference from other files but only used if
5818 no other definition is available. Use the expression
5819 @code{assemble_name (@var{stream}, @var{name})} to output the name
5820 itself; before and after that, output the additional assembler syntax
5821 for making that name weak, and a newline.
5823 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5824 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5828 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5829 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5830 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5831 or variable decl. If @var{value} is not @code{NULL}, this C statement
5832 should output to the stdio stream @var{stream} assembler code which
5833 defines (equates) the weak symbol @var{name} to have the value
5834 @var{value}. If @var{value} is @code{NULL}, it should output commands
5835 to make @var{name} weak.
5838 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5839 Outputs a directive that enables @var{name} to be used to refer to
5840 symbol @var{value} with weak-symbol semantics. @code{decl} is the
5841 declaration of @code{name}.
5844 @defmac SUPPORTS_WEAK
5845 A preprocessor constant expression which evaluates to true if the target
5846 supports weak symbols.
5848 If you don't define this macro, @file{defaults.h} provides a default
5849 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5850 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5853 @defmac TARGET_SUPPORTS_WEAK
5854 A C expression which evaluates to true if the target supports weak symbols.
5856 If you don't define this macro, @file{defaults.h} provides a default
5857 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
5858 this macro if you want to control weak symbol support with a compiler
5859 flag such as @option{-melf}.
5862 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
5863 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5864 public symbol such that extra copies in multiple translation units will
5865 be discarded by the linker. Define this macro if your object file
5866 format provides support for this concept, such as the @samp{COMDAT}
5867 section flags in the Microsoft Windows PE/COFF format, and this support
5868 requires changes to @var{decl}, such as putting it in a separate section.
5871 @defmac SUPPORTS_ONE_ONLY
5872 A C expression which evaluates to true if the target supports one-only
5875 If you don't define this macro, @file{varasm.c} provides a default
5876 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5877 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5878 you want to control one-only symbol support with a compiler flag, or if
5879 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5880 be emitted as one-only.
5883 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
5885 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5886 A C expression that evaluates to true if the target's linker expects
5887 that weak symbols do not appear in a static archive's table of contents.
5888 The default is @code{0}.
5890 Leaving weak symbols out of an archive's table of contents means that,
5891 if a symbol will only have a definition in one translation unit and
5892 will have undefined references from other translation units, that
5893 symbol should not be weak. Defining this macro to be nonzero will
5894 thus have the effect that certain symbols that would normally be weak
5895 (explicit template instantiations, and vtables for polymorphic classes
5896 with noninline key methods) will instead be nonweak.
5898 The C++ ABI requires this macro to be zero. Define this macro for
5899 targets where full C++ ABI compliance is impossible and where linker
5900 restrictions require weak symbols to be left out of a static archive's
5904 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5905 A C statement (sans semicolon) to output to the stdio stream
5906 @var{stream} any text necessary for declaring the name of an external
5907 symbol named @var{name} which is referenced in this compilation but
5908 not defined. The value of @var{decl} is the tree node for the
5911 This macro need not be defined if it does not need to output anything.
5912 The GNU assembler and most Unix assemblers don't require anything.
5915 @hook TARGET_ASM_EXTERNAL_LIBCALL
5917 @hook TARGET_ASM_MARK_DECL_PRESERVED
5919 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5920 A C statement (sans semicolon) to output to the stdio stream
5921 @var{stream} a reference in assembler syntax to a label named
5922 @var{name}. This should add @samp{_} to the front of the name, if that
5923 is customary on your operating system, as it is in most Berkeley Unix
5924 systems. This macro is used in @code{assemble_name}.
5927 @hook TARGET_MANGLE_ASSEMBLER_NAME
5929 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5930 A C statement (sans semicolon) to output a reference to
5931 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
5932 will be used to output the name of the symbol. This macro may be used
5933 to modify the way a symbol is referenced depending on information
5934 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5937 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5938 A C statement (sans semicolon) to output a reference to @var{buf}, the
5939 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
5940 @code{assemble_name} will be used to output the name of the symbol.
5941 This macro is not used by @code{output_asm_label}, or the @code{%l}
5942 specifier that calls it; the intention is that this macro should be set
5943 when it is necessary to output a label differently when its address is
5947 @hook TARGET_ASM_INTERNAL_LABEL
5949 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5950 A C statement to output to the stdio stream @var{stream} a debug info
5951 label whose name is made from the string @var{prefix} and the number
5952 @var{num}. This is useful for VLIW targets, where debug info labels
5953 may need to be treated differently than branch target labels. On some
5954 systems, branch target labels must be at the beginning of instruction
5955 bundles, but debug info labels can occur in the middle of instruction
5958 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5962 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5963 A C statement to store into the string @var{string} a label whose name
5964 is made from the string @var{prefix} and the number @var{num}.
5966 This string, when output subsequently by @code{assemble_name}, should
5967 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5968 with the same @var{prefix} and @var{num}.
5970 If the string begins with @samp{*}, then @code{assemble_name} will
5971 output the rest of the string unchanged. It is often convenient for
5972 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5973 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5974 to output the string, and may change it. (Of course,
5975 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5976 you should know what it does on your machine.)
5979 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5980 A C expression to assign to @var{outvar} (which is a variable of type
5981 @code{char *}) a newly allocated string made from the string
5982 @var{name} and the number @var{number}, with some suitable punctuation
5983 added. Use @code{alloca} to get space for the string.
5985 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5986 produce an assembler label for an internal static variable whose name is
5987 @var{name}. Therefore, the string must be such as to result in valid
5988 assembler code. The argument @var{number} is different each time this
5989 macro is executed; it prevents conflicts between similarly-named
5990 internal static variables in different scopes.
5992 Ideally this string should not be a valid C identifier, to prevent any
5993 conflict with the user's own symbols. Most assemblers allow periods
5994 or percent signs in assembler symbols; putting at least one of these
5995 between the name and the number will suffice.
5997 If this macro is not defined, a default definition will be provided
5998 which is correct for most systems.
6001 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6002 A C statement to output to the stdio stream @var{stream} assembler code
6003 which defines (equates) the symbol @var{name} to have the value @var{value}.
6006 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6007 correct for most systems.
6010 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6011 A C statement to output to the stdio stream @var{stream} assembler code
6012 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6013 to have the value of the tree node @var{decl_of_value}. This macro will
6014 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6015 the tree nodes are available.
6018 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6019 correct for most systems.
6022 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
6023 A C statement that evaluates to true if the assembler code which defines
6024 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
6025 of the tree node @var{decl_of_value} should be emitted near the end of the
6026 current compilation unit. The default is to not defer output of defines.
6027 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
6028 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
6031 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6032 A C statement to output to the stdio stream @var{stream} assembler code
6033 which defines (equates) the weak symbol @var{name} to have the value
6034 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6035 an undefined weak symbol.
6037 Define this macro if the target only supports weak aliases; define
6038 @code{ASM_OUTPUT_DEF} instead if possible.
6041 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6042 Define this macro to override the default assembler names used for
6043 Objective-C methods.
6045 The default name is a unique method number followed by the name of the
6046 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6047 the category is also included in the assembler name (e.g.@:
6050 These names are safe on most systems, but make debugging difficult since
6051 the method's selector is not present in the name. Therefore, particular
6052 systems define other ways of computing names.
6054 @var{buf} is an expression of type @code{char *} which gives you a
6055 buffer in which to store the name; its length is as long as
6056 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6057 50 characters extra.
6059 The argument @var{is_inst} specifies whether the method is an instance
6060 method or a class method; @var{class_name} is the name of the class;
6061 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6062 in a category); and @var{sel_name} is the name of the selector.
6064 On systems where the assembler can handle quoted names, you can use this
6065 macro to provide more human-readable names.
6068 @node Initialization
6069 @subsection How Initialization Functions Are Handled
6070 @cindex initialization routines
6071 @cindex termination routines
6072 @cindex constructors, output of
6073 @cindex destructors, output of
6075 The compiled code for certain languages includes @dfn{constructors}
6076 (also called @dfn{initialization routines})---functions to initialize
6077 data in the program when the program is started. These functions need
6078 to be called before the program is ``started''---that is to say, before
6079 @code{main} is called.
6081 Compiling some languages generates @dfn{destructors} (also called
6082 @dfn{termination routines}) that should be called when the program
6085 To make the initialization and termination functions work, the compiler
6086 must output something in the assembler code to cause those functions to
6087 be called at the appropriate time. When you port the compiler to a new
6088 system, you need to specify how to do this.
6090 There are two major ways that GCC currently supports the execution of
6091 initialization and termination functions. Each way has two variants.
6092 Much of the structure is common to all four variations.
6094 @findex __CTOR_LIST__
6095 @findex __DTOR_LIST__
6096 The linker must build two lists of these functions---a list of
6097 initialization functions, called @code{__CTOR_LIST__}, and a list of
6098 termination functions, called @code{__DTOR_LIST__}.
6100 Each list always begins with an ignored function pointer (which may hold
6101 0, @minus{}1, or a count of the function pointers after it, depending on
6102 the environment). This is followed by a series of zero or more function
6103 pointers to constructors (or destructors), followed by a function
6104 pointer containing zero.
6106 Depending on the operating system and its executable file format, either
6107 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6108 time and exit time. Constructors are called in reverse order of the
6109 list; destructors in forward order.
6111 The best way to handle static constructors works only for object file
6112 formats which provide arbitrarily-named sections. A section is set
6113 aside for a list of constructors, and another for a list of destructors.
6114 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6115 object file that defines an initialization function also puts a word in
6116 the constructor section to point to that function. The linker
6117 accumulates all these words into one contiguous @samp{.ctors} section.
6118 Termination functions are handled similarly.
6120 This method will be chosen as the default by @file{target-def.h} if
6121 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6122 support arbitrary sections, but does support special designated
6123 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6124 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6126 When arbitrary sections are available, there are two variants, depending
6127 upon how the code in @file{crtstuff.c} is called. On systems that
6128 support a @dfn{.init} section which is executed at program startup,
6129 parts of @file{crtstuff.c} are compiled into that section. The
6130 program is linked by the @command{gcc} driver like this:
6133 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6136 The prologue of a function (@code{__init}) appears in the @code{.init}
6137 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6138 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6139 files are provided by the operating system or by the GNU C library, but
6140 are provided by GCC for a few targets.
6142 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6143 compiled from @file{crtstuff.c}. They contain, among other things, code
6144 fragments within the @code{.init} and @code{.fini} sections that branch
6145 to routines in the @code{.text} section. The linker will pull all parts
6146 of a section together, which results in a complete @code{__init} function
6147 that invokes the routines we need at startup.
6149 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6152 If no init section is available, when GCC compiles any function called
6153 @code{main} (or more accurately, any function designated as a program
6154 entry point by the language front end calling @code{expand_main_function}),
6155 it inserts a procedure call to @code{__main} as the first executable code
6156 after the function prologue. The @code{__main} function is defined
6157 in @file{libgcc2.c} and runs the global constructors.
6159 In file formats that don't support arbitrary sections, there are again
6160 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6161 and an `a.out' format must be used. In this case,
6162 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6163 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6164 and with the address of the void function containing the initialization
6165 code as its value. The GNU linker recognizes this as a request to add
6166 the value to a @dfn{set}; the values are accumulated, and are eventually
6167 placed in the executable as a vector in the format described above, with
6168 a leading (ignored) count and a trailing zero element.
6169 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6170 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6171 the compilation of @code{main} to call @code{__main} as above, starting
6172 the initialization process.
6174 The last variant uses neither arbitrary sections nor the GNU linker.
6175 This is preferable when you want to do dynamic linking and when using
6176 file formats which the GNU linker does not support, such as `ECOFF'@. In
6177 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6178 termination functions are recognized simply by their names. This requires
6179 an extra program in the linkage step, called @command{collect2}. This program
6180 pretends to be the linker, for use with GCC; it does its job by running
6181 the ordinary linker, but also arranges to include the vectors of
6182 initialization and termination functions. These functions are called
6183 via @code{__main} as described above. In order to use this method,
6184 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6187 The following section describes the specific macros that control and
6188 customize the handling of initialization and termination functions.
6191 @node Macros for Initialization
6192 @subsection Macros Controlling Initialization Routines
6194 Here are the macros that control how the compiler handles initialization
6195 and termination functions:
6197 @defmac INIT_SECTION_ASM_OP
6198 If defined, a C string constant, including spacing, for the assembler
6199 operation to identify the following data as initialization code. If not
6200 defined, GCC will assume such a section does not exist. When you are
6201 using special sections for initialization and termination functions, this
6202 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6203 run the initialization functions.
6206 @defmac HAS_INIT_SECTION
6207 If defined, @code{main} will not call @code{__main} as described above.
6208 This macro should be defined for systems that control start-up code
6209 on a symbol-by-symbol basis, such as OSF/1, and should not
6210 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6213 @defmac LD_INIT_SWITCH
6214 If defined, a C string constant for a switch that tells the linker that
6215 the following symbol is an initialization routine.
6218 @defmac LD_FINI_SWITCH
6219 If defined, a C string constant for a switch that tells the linker that
6220 the following symbol is a finalization routine.
6223 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6224 If defined, a C statement that will write a function that can be
6225 automatically called when a shared library is loaded. The function
6226 should call @var{func}, which takes no arguments. If not defined, and
6227 the object format requires an explicit initialization function, then a
6228 function called @code{_GLOBAL__DI} will be generated.
6230 This function and the following one are used by collect2 when linking a
6231 shared library that needs constructors or destructors, or has DWARF2
6232 exception tables embedded in the code.
6235 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6236 If defined, a C statement that will write a function that can be
6237 automatically called when a shared library is unloaded. The function
6238 should call @var{func}, which takes no arguments. If not defined, and
6239 the object format requires an explicit finalization function, then a
6240 function called @code{_GLOBAL__DD} will be generated.
6243 @defmac INVOKE__main
6244 If defined, @code{main} will call @code{__main} despite the presence of
6245 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6246 where the init section is not actually run automatically, but is still
6247 useful for collecting the lists of constructors and destructors.
6250 @defmac SUPPORTS_INIT_PRIORITY
6251 If nonzero, the C++ @code{init_priority} attribute is supported and the
6252 compiler should emit instructions to control the order of initialization
6253 of objects. If zero, the compiler will issue an error message upon
6254 encountering an @code{init_priority} attribute.
6257 @hook TARGET_HAVE_CTORS_DTORS
6259 @hook TARGET_ASM_CONSTRUCTOR
6261 @hook TARGET_ASM_DESTRUCTOR
6263 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6264 generated for the generated object file will have static linkage.
6266 If your system uses @command{collect2} as the means of processing
6267 constructors, then that program normally uses @command{nm} to scan
6268 an object file for constructor functions to be called.
6270 On certain kinds of systems, you can define this macro to make
6271 @command{collect2} work faster (and, in some cases, make it work at all):
6273 @defmac OBJECT_FORMAT_COFF
6274 Define this macro if the system uses COFF (Common Object File Format)
6275 object files, so that @command{collect2} can assume this format and scan
6276 object files directly for dynamic constructor/destructor functions.
6278 This macro is effective only in a native compiler; @command{collect2} as
6279 part of a cross compiler always uses @command{nm} for the target machine.
6282 @defmac REAL_NM_FILE_NAME
6283 Define this macro as a C string constant containing the file name to use
6284 to execute @command{nm}. The default is to search the path normally for
6289 @command{collect2} calls @command{nm} to scan object files for static
6290 constructors and destructors and LTO info. By default, @option{-n} is
6291 passed. Define @code{NM_FLAGS} to a C string constant if other options
6292 are needed to get the same output format as GNU @command{nm -n}
6296 If your system supports shared libraries and has a program to list the
6297 dynamic dependencies of a given library or executable, you can define
6298 these macros to enable support for running initialization and
6299 termination functions in shared libraries:
6302 Define this macro to a C string constant containing the name of the program
6303 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6306 @defmac PARSE_LDD_OUTPUT (@var{ptr})
6307 Define this macro to be C code that extracts filenames from the output
6308 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6309 of type @code{char *} that points to the beginning of a line of output
6310 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6311 code must advance @var{ptr} to the beginning of the filename on that
6312 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6315 @defmac SHLIB_SUFFIX
6316 Define this macro to a C string constant containing the default shared
6317 library extension of the target (e.g., @samp{".so"}). @command{collect2}
6318 strips version information after this suffix when generating global
6319 constructor and destructor names. This define is only needed on targets
6320 that use @command{collect2} to process constructors and destructors.
6323 @node Instruction Output
6324 @subsection Output of Assembler Instructions
6326 @c prevent bad page break with this line
6327 This describes assembler instruction output.
6329 @defmac REGISTER_NAMES
6330 A C initializer containing the assembler's names for the machine
6331 registers, each one as a C string constant. This is what translates
6332 register numbers in the compiler into assembler language.
6335 @defmac ADDITIONAL_REGISTER_NAMES
6336 If defined, a C initializer for an array of structures containing a name
6337 and a register number. This macro defines additional names for hard
6338 registers, thus allowing the @code{asm} option in declarations to refer
6339 to registers using alternate names.
6342 @defmac OVERLAPPING_REGISTER_NAMES
6343 If defined, a C initializer for an array of structures containing a
6344 name, a register number and a count of the number of consecutive
6345 machine registers the name overlaps. This macro defines additional
6346 names for hard registers, thus allowing the @code{asm} option in
6347 declarations to refer to registers using alternate names. Unlike
6348 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6349 register name implies multiple underlying registers.
6351 This macro should be used when it is important that a clobber in an
6352 @code{asm} statement clobbers all the underlying values implied by the
6353 register name. For example, on ARM, clobbering the double-precision
6354 VFP register ``d0'' implies clobbering both single-precision registers
6358 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6359 Define this macro if you are using an unusual assembler that
6360 requires different names for the machine instructions.
6362 The definition is a C statement or statements which output an
6363 assembler instruction opcode to the stdio stream @var{stream}. The
6364 macro-operand @var{ptr} is a variable of type @code{char *} which
6365 points to the opcode name in its ``internal'' form---the form that is
6366 written in the machine description. The definition should output the
6367 opcode name to @var{stream}, performing any translation you desire, and
6368 increment the variable @var{ptr} to point at the end of the opcode
6369 so that it will not be output twice.
6371 In fact, your macro definition may process less than the entire opcode
6372 name, or more than the opcode name; but if you want to process text
6373 that includes @samp{%}-sequences to substitute operands, you must take
6374 care of the substitution yourself. Just be sure to increment
6375 @var{ptr} over whatever text should not be output normally.
6377 @findex recog_data.operand
6378 If you need to look at the operand values, they can be found as the
6379 elements of @code{recog_data.operand}.
6381 If the macro definition does nothing, the instruction is output
6385 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6386 If defined, a C statement to be executed just prior to the output of
6387 assembler code for @var{insn}, to modify the extracted operands so
6388 they will be output differently.
6390 Here the argument @var{opvec} is the vector containing the operands
6391 extracted from @var{insn}, and @var{noperands} is the number of
6392 elements of the vector which contain meaningful data for this insn.
6393 The contents of this vector are what will be used to convert the insn
6394 template into assembler code, so you can change the assembler output
6395 by changing the contents of the vector.
6397 This macro is useful when various assembler syntaxes share a single
6398 file of instruction patterns; by defining this macro differently, you
6399 can cause a large class of instructions to be output differently (such
6400 as with rearranged operands). Naturally, variations in assembler
6401 syntax affecting individual insn patterns ought to be handled by
6402 writing conditional output routines in those patterns.
6404 If this macro is not defined, it is equivalent to a null statement.
6407 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
6409 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6410 A C compound statement to output to stdio stream @var{stream} the
6411 assembler syntax for an instruction operand @var{x}. @var{x} is an
6414 @var{code} is a value that can be used to specify one of several ways
6415 of printing the operand. It is used when identical operands must be
6416 printed differently depending on the context. @var{code} comes from
6417 the @samp{%} specification that was used to request printing of the
6418 operand. If the specification was just @samp{%@var{digit}} then
6419 @var{code} is 0; if the specification was @samp{%@var{ltr}
6420 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6423 If @var{x} is a register, this macro should print the register's name.
6424 The names can be found in an array @code{reg_names} whose type is
6425 @code{char *[]}. @code{reg_names} is initialized from
6426 @code{REGISTER_NAMES}.
6428 When the machine description has a specification @samp{%@var{punct}}
6429 (a @samp{%} followed by a punctuation character), this macro is called
6430 with a null pointer for @var{x} and the punctuation character for
6434 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6435 A C expression which evaluates to true if @var{code} is a valid
6436 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6437 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6438 punctuation characters (except for the standard one, @samp{%}) are used
6442 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6443 A C compound statement to output to stdio stream @var{stream} the
6444 assembler syntax for an instruction operand that is a memory reference
6445 whose address is @var{x}. @var{x} is an RTL expression.
6447 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6448 On some machines, the syntax for a symbolic address depends on the
6449 section that the address refers to. On these machines, define the hook
6450 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6451 @code{symbol_ref}, and then check for it here. @xref{Assembler
6455 @findex dbr_sequence_length
6456 @defmac DBR_OUTPUT_SEQEND (@var{file})
6457 A C statement, to be executed after all slot-filler instructions have
6458 been output. If necessary, call @code{dbr_sequence_length} to
6459 determine the number of slots filled in a sequence (zero if not
6460 currently outputting a sequence), to decide how many no-ops to output,
6463 Don't define this macro if it has nothing to do, but it is helpful in
6464 reading assembly output if the extent of the delay sequence is made
6465 explicit (e.g.@: with white space).
6468 @findex final_sequence
6469 Note that output routines for instructions with delay slots must be
6470 prepared to deal with not being output as part of a sequence
6471 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6472 found.) The variable @code{final_sequence} is null when not
6473 processing a sequence, otherwise it contains the @code{sequence} rtx
6477 @defmac REGISTER_PREFIX
6478 @defmacx LOCAL_LABEL_PREFIX
6479 @defmacx USER_LABEL_PREFIX
6480 @defmacx IMMEDIATE_PREFIX
6481 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6482 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6483 @file{final.c}). These are useful when a single @file{md} file must
6484 support multiple assembler formats. In that case, the various @file{tm.h}
6485 files can define these macros differently.
6488 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6489 If defined this macro should expand to a series of @code{case}
6490 statements which will be parsed inside the @code{switch} statement of
6491 the @code{asm_fprintf} function. This allows targets to define extra
6492 printf formats which may useful when generating their assembler
6493 statements. Note that uppercase letters are reserved for future
6494 generic extensions to asm_fprintf, and so are not available to target
6495 specific code. The output file is given by the parameter @var{file}.
6496 The varargs input pointer is @var{argptr} and the rest of the format
6497 string, starting the character after the one that is being switched
6498 upon, is pointed to by @var{format}.
6501 @defmac ASSEMBLER_DIALECT
6502 If your target supports multiple dialects of assembler language (such as
6503 different opcodes), define this macro as a C expression that gives the
6504 numeric index of the assembler language dialect to use, with zero as the
6507 If this macro is defined, you may use constructs of the form
6509 @samp{@{option0|option1|option2@dots{}@}}
6512 in the output templates of patterns (@pxref{Output Template}) or in the
6513 first argument of @code{asm_fprintf}. This construct outputs
6514 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6515 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
6516 within these strings retain their usual meaning. If there are fewer
6517 alternatives within the braces than the value of
6518 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6519 to print curly braces or @samp{|} character in assembler output directly,
6520 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6522 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6523 @samp{@}} do not have any special meaning when used in templates or
6524 operands to @code{asm_fprintf}.
6526 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6527 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6528 the variations in assembler language syntax with that mechanism. Define
6529 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6530 if the syntax variant are larger and involve such things as different
6531 opcodes or operand order.
6534 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6535 A C expression to output to @var{stream} some assembler code
6536 which will push hard register number @var{regno} onto the stack.
6537 The code need not be optimal, since this macro is used only when
6541 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6542 A C expression to output to @var{stream} some assembler code
6543 which will pop hard register number @var{regno} off of the stack.
6544 The code need not be optimal, since this macro is used only when
6548 @node Dispatch Tables
6549 @subsection Output of Dispatch Tables
6551 @c prevent bad page break with this line
6552 This concerns dispatch tables.
6554 @cindex dispatch table
6555 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6556 A C statement to output to the stdio stream @var{stream} an assembler
6557 pseudo-instruction to generate a difference between two labels.
6558 @var{value} and @var{rel} are the numbers of two internal labels. The
6559 definitions of these labels are output using
6560 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6561 way here. For example,
6564 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6565 @var{value}, @var{rel})
6568 You must provide this macro on machines where the addresses in a
6569 dispatch table are relative to the table's own address. If defined, GCC
6570 will also use this macro on all machines when producing PIC@.
6571 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6572 mode and flags can be read.
6575 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6576 This macro should be provided on machines where the addresses
6577 in a dispatch table are absolute.
6579 The definition should be a C statement to output to the stdio stream
6580 @var{stream} an assembler pseudo-instruction to generate a reference to
6581 a label. @var{value} is the number of an internal label whose
6582 definition is output using @code{(*targetm.asm_out.internal_label)}.
6586 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6590 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6591 Define this if the label before a jump-table needs to be output
6592 specially. The first three arguments are the same as for
6593 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6594 jump-table which follows (a @code{jump_table_data} containing an
6595 @code{addr_vec} or @code{addr_diff_vec}).
6597 This feature is used on system V to output a @code{swbeg} statement
6600 If this macro is not defined, these labels are output with
6601 @code{(*targetm.asm_out.internal_label)}.
6604 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6605 Define this if something special must be output at the end of a
6606 jump-table. The definition should be a C statement to be executed
6607 after the assembler code for the table is written. It should write
6608 the appropriate code to stdio stream @var{stream}. The argument
6609 @var{table} is the jump-table insn, and @var{num} is the label-number
6610 of the preceding label.
6612 If this macro is not defined, nothing special is output at the end of
6616 @hook TARGET_ASM_EMIT_UNWIND_LABEL
6618 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6620 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6622 @hook TARGET_ASM_UNWIND_EMIT
6624 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6626 @node Exception Region Output
6627 @subsection Assembler Commands for Exception Regions
6629 @c prevent bad page break with this line
6631 This describes commands marking the start and the end of an exception
6634 @defmac EH_FRAME_SECTION_NAME
6635 If defined, a C string constant for the name of the section containing
6636 exception handling frame unwind information. If not defined, GCC will
6637 provide a default definition if the target supports named sections.
6638 @file{crtstuff.c} uses this macro to switch to the appropriate section.
6640 You should define this symbol if your target supports DWARF 2 frame
6641 unwind information and the default definition does not work.
6644 @defmac EH_FRAME_IN_DATA_SECTION
6645 If defined, DWARF 2 frame unwind information will be placed in the
6646 data section even though the target supports named sections. This
6647 might be necessary, for instance, if the system linker does garbage
6648 collection and sections cannot be marked as not to be collected.
6650 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
6654 @defmac EH_TABLES_CAN_BE_READ_ONLY
6655 Define this macro to 1 if your target is such that no frame unwind
6656 information encoding used with non-PIC code will ever require a
6657 runtime relocation, but the linker may not support merging read-only
6658 and read-write sections into a single read-write section.
6661 @defmac MASK_RETURN_ADDR
6662 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6663 that it does not contain any extraneous set bits in it.
6666 @defmac DWARF2_UNWIND_INFO
6667 Define this macro to 0 if your target supports DWARF 2 frame unwind
6668 information, but it does not yet work with exception handling.
6669 Otherwise, if your target supports this information (if it defines
6670 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6671 GCC will provide a default definition of 1.
6674 @hook TARGET_EXCEPT_UNWIND_INFO
6675 This hook defines the mechanism that will be used for exception handling
6676 by the target. If the target has ABI specified unwind tables, the hook
6677 should return @code{UI_TARGET}. If the target is to use the
6678 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6679 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
6680 information, the hook should return @code{UI_DWARF2}.
6682 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6683 This may end up simplifying other parts of target-specific code. The
6684 default implementation of this hook never returns @code{UI_NONE}.
6686 Note that the value returned by this hook should be constant. It should
6687 not depend on anything except the command-line switches described by
6688 @var{opts}. In particular, the
6689 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6690 macros and builtin functions related to exception handling are set up
6691 depending on this setting.
6693 The default implementation of the hook first honors the
6694 @option{--enable-sjlj-exceptions} configure option, then
6695 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
6696 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6697 must define this hook so that @var{opts} is used correctly.
6700 @hook TARGET_UNWIND_TABLES_DEFAULT
6701 This variable should be set to @code{true} if the target ABI requires unwinding
6702 tables even when exceptions are not used. It must not be modified by
6703 command-line option processing.
6706 @defmac DONT_USE_BUILTIN_SETJMP
6707 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6708 should use the @code{setjmp}/@code{longjmp} functions from the C library
6709 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6712 @defmac JMP_BUF_SIZE
6713 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6714 defined. Define this macro if the default size of @code{jmp_buf} buffer
6715 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6716 is not large enough, or if it is much too large.
6717 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6720 @defmac DWARF_CIE_DATA_ALIGNMENT
6721 This macro need only be defined if the target might save registers in the
6722 function prologue at an offset to the stack pointer that is not aligned to
6723 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6724 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
6725 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
6726 the target supports DWARF 2 frame unwind information.
6729 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6731 @hook TARGET_DWARF_REGISTER_SPAN
6733 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6735 @hook TARGET_ASM_TTYPE
6737 @hook TARGET_ARM_EABI_UNWINDER
6739 @node Alignment Output
6740 @subsection Assembler Commands for Alignment
6742 @c prevent bad page break with this line
6743 This describes commands for alignment.
6745 @defmac JUMP_ALIGN (@var{label})
6746 The alignment (log base 2) to put in front of @var{label}, which is
6747 a common destination of jumps and has no fallthru incoming edge.
6749 This macro need not be defined if you don't want any special alignment
6750 to be done at such a time. Most machine descriptions do not currently
6753 Unless it's necessary to inspect the @var{label} parameter, it is better
6754 to set the variable @var{align_jumps} in the target's
6755 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6756 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6759 @hook TARGET_ASM_JUMP_ALIGN_MAX_SKIP
6761 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6762 The alignment (log base 2) to put in front of @var{label}, which follows
6765 This macro need not be defined if you don't want any special alignment
6766 to be done at such a time. Most machine descriptions do not currently
6770 @hook TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6772 @defmac LOOP_ALIGN (@var{label})
6773 The alignment (log base 2) to put in front of @var{label} that heads
6774 a frequently executed basic block (usually the header of a loop).
6776 This macro need not be defined if you don't want any special alignment
6777 to be done at such a time. Most machine descriptions do not currently
6780 Unless it's necessary to inspect the @var{label} parameter, it is better
6781 to set the variable @code{align_loops} in the target's
6782 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6783 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6786 @hook TARGET_ASM_LOOP_ALIGN_MAX_SKIP
6788 @defmac LABEL_ALIGN (@var{label})
6789 The alignment (log base 2) to put in front of @var{label}.
6790 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6791 the maximum of the specified values is used.
6793 Unless it's necessary to inspect the @var{label} parameter, it is better
6794 to set the variable @code{align_labels} in the target's
6795 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6796 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6799 @hook TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6801 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6802 A C statement to output to the stdio stream @var{stream} an assembler
6803 instruction to advance the location counter by @var{nbytes} bytes.
6804 Those bytes should be zero when loaded. @var{nbytes} will be a C
6805 expression of type @code{unsigned HOST_WIDE_INT}.
6808 @defmac ASM_NO_SKIP_IN_TEXT
6809 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6810 text section because it fails to put zeros in the bytes that are skipped.
6811 This is true on many Unix systems, where the pseudo--op to skip bytes
6812 produces no-op instructions rather than zeros when used in the text
6816 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6817 A C statement to output to the stdio stream @var{stream} an assembler
6818 command to advance the location counter to a multiple of 2 to the
6819 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6822 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6823 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6824 for padding, if necessary.
6827 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6828 A C statement to output to the stdio stream @var{stream} an assembler
6829 command to advance the location counter to a multiple of 2 to the
6830 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6831 satisfy the alignment request. @var{power} and @var{max_skip} will be
6832 a C expression of type @code{int}.
6836 @node Debugging Info
6837 @section Controlling Debugging Information Format
6839 @c prevent bad page break with this line
6840 This describes how to specify debugging information.
6843 * All Debuggers:: Macros that affect all debugging formats uniformly.
6844 * DBX Options:: Macros enabling specific options in DBX format.
6845 * DBX Hooks:: Hook macros for varying DBX format.
6846 * File Names and DBX:: Macros controlling output of file names in DBX format.
6847 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6848 * VMS Debug:: Macros for VMS debug format.
6852 @subsection Macros Affecting All Debugging Formats
6854 @c prevent bad page break with this line
6855 These macros affect all debugging formats.
6857 @defmac DBX_REGISTER_NUMBER (@var{regno})
6858 A C expression that returns the DBX register number for the compiler
6859 register number @var{regno}. In the default macro provided, the value
6860 of this expression will be @var{regno} itself. But sometimes there are
6861 some registers that the compiler knows about and DBX does not, or vice
6862 versa. In such cases, some register may need to have one number in the
6863 compiler and another for DBX@.
6865 If two registers have consecutive numbers inside GCC, and they can be
6866 used as a pair to hold a multiword value, then they @emph{must} have
6867 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6868 Otherwise, debuggers will be unable to access such a pair, because they
6869 expect register pairs to be consecutive in their own numbering scheme.
6871 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6872 does not preserve register pairs, then what you must do instead is
6873 redefine the actual register numbering scheme.
6876 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
6877 A C expression that returns the integer offset value for an automatic
6878 variable having address @var{x} (an RTL expression). The default
6879 computation assumes that @var{x} is based on the frame-pointer and
6880 gives the offset from the frame-pointer. This is required for targets
6881 that produce debugging output for DBX or COFF-style debugging output
6882 for SDB and allow the frame-pointer to be eliminated when the
6883 @option{-g} options is used.
6886 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6887 A C expression that returns the integer offset value for an argument
6888 having address @var{x} (an RTL expression). The nominal offset is
6892 @defmac PREFERRED_DEBUGGING_TYPE
6893 A C expression that returns the type of debugging output GCC should
6894 produce when the user specifies just @option{-g}. Define
6895 this if you have arranged for GCC to support more than one format of
6896 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6897 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
6898 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
6900 When the user specifies @option{-ggdb}, GCC normally also uses the
6901 value of this macro to select the debugging output format, but with two
6902 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6903 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6904 defined, GCC uses @code{DBX_DEBUG}.
6906 The value of this macro only affects the default debugging output; the
6907 user can always get a specific type of output by using @option{-gstabs},
6908 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
6912 @subsection Specific Options for DBX Output
6914 @c prevent bad page break with this line
6915 These are specific options for DBX output.
6917 @defmac DBX_DEBUGGING_INFO
6918 Define this macro if GCC should produce debugging output for DBX
6919 in response to the @option{-g} option.
6922 @defmac XCOFF_DEBUGGING_INFO
6923 Define this macro if GCC should produce XCOFF format debugging output
6924 in response to the @option{-g} option. This is a variant of DBX format.
6927 @defmac DEFAULT_GDB_EXTENSIONS
6928 Define this macro to control whether GCC should by default generate
6929 GDB's extended version of DBX debugging information (assuming DBX-format
6930 debugging information is enabled at all). If you don't define the
6931 macro, the default is 1: always generate the extended information
6932 if there is any occasion to.
6935 @defmac DEBUG_SYMS_TEXT
6936 Define this macro if all @code{.stabs} commands should be output while
6937 in the text section.
6940 @defmac ASM_STABS_OP
6941 A C string constant, including spacing, naming the assembler pseudo op to
6942 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6943 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
6944 applies only to DBX debugging information format.
6947 @defmac ASM_STABD_OP
6948 A C string constant, including spacing, naming the assembler pseudo op to
6949 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6950 value is the current location. If you don't define this macro,
6951 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
6955 @defmac ASM_STABN_OP
6956 A C string constant, including spacing, naming the assembler pseudo op to
6957 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6958 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
6959 macro applies only to DBX debugging information format.
6962 @defmac DBX_NO_XREFS
6963 Define this macro if DBX on your system does not support the construct
6964 @samp{xs@var{tagname}}. On some systems, this construct is used to
6965 describe a forward reference to a structure named @var{tagname}.
6966 On other systems, this construct is not supported at all.
6969 @defmac DBX_CONTIN_LENGTH
6970 A symbol name in DBX-format debugging information is normally
6971 continued (split into two separate @code{.stabs} directives) when it
6972 exceeds a certain length (by default, 80 characters). On some
6973 operating systems, DBX requires this splitting; on others, splitting
6974 must not be done. You can inhibit splitting by defining this macro
6975 with the value zero. You can override the default splitting-length by
6976 defining this macro as an expression for the length you desire.
6979 @defmac DBX_CONTIN_CHAR
6980 Normally continuation is indicated by adding a @samp{\} character to
6981 the end of a @code{.stabs} string when a continuation follows. To use
6982 a different character instead, define this macro as a character
6983 constant for the character you want to use. Do not define this macro
6984 if backslash is correct for your system.
6987 @defmac DBX_STATIC_STAB_DATA_SECTION
6988 Define this macro if it is necessary to go to the data section before
6989 outputting the @samp{.stabs} pseudo-op for a non-global static
6993 @defmac DBX_TYPE_DECL_STABS_CODE
6994 The value to use in the ``code'' field of the @code{.stabs} directive
6995 for a typedef. The default is @code{N_LSYM}.
6998 @defmac DBX_STATIC_CONST_VAR_CODE
6999 The value to use in the ``code'' field of the @code{.stabs} directive
7000 for a static variable located in the text section. DBX format does not
7001 provide any ``right'' way to do this. The default is @code{N_FUN}.
7004 @defmac DBX_REGPARM_STABS_CODE
7005 The value to use in the ``code'' field of the @code{.stabs} directive
7006 for a parameter passed in registers. DBX format does not provide any
7007 ``right'' way to do this. The default is @code{N_RSYM}.
7010 @defmac DBX_REGPARM_STABS_LETTER
7011 The letter to use in DBX symbol data to identify a symbol as a parameter
7012 passed in registers. DBX format does not customarily provide any way to
7013 do this. The default is @code{'P'}.
7016 @defmac DBX_FUNCTION_FIRST
7017 Define this macro if the DBX information for a function and its
7018 arguments should precede the assembler code for the function. Normally,
7019 in DBX format, the debugging information entirely follows the assembler
7023 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7024 Define this macro, with value 1, if the value of a symbol describing
7025 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
7026 relative to the start of the enclosing function. Normally, GCC uses
7027 an absolute address.
7030 @defmac DBX_LINES_FUNCTION_RELATIVE
7031 Define this macro, with value 1, if the value of a symbol indicating
7032 the current line number (@code{N_SLINE}) should be relative to the
7033 start of the enclosing function. Normally, GCC uses an absolute address.
7036 @defmac DBX_USE_BINCL
7037 Define this macro if GCC should generate @code{N_BINCL} and
7038 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7039 macro also directs GCC to output a type number as a pair of a file
7040 number and a type number within the file. Normally, GCC does not
7041 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7042 number for a type number.
7046 @subsection Open-Ended Hooks for DBX Format
7048 @c prevent bad page break with this line
7049 These are hooks for DBX format.
7051 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
7052 A C statement to output DBX debugging information before code for line
7053 number @var{line} of the current source file to the stdio stream
7054 @var{stream}. @var{counter} is the number of time the macro was
7055 invoked, including the current invocation; it is intended to generate
7056 unique labels in the assembly output.
7058 This macro should not be defined if the default output is correct, or
7059 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
7062 @defmac NO_DBX_FUNCTION_END
7063 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7064 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7065 On those machines, define this macro to turn this feature off without
7066 disturbing the rest of the gdb extensions.
7069 @defmac NO_DBX_BNSYM_ENSYM
7070 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
7071 extension construct. On those machines, define this macro to turn this
7072 feature off without disturbing the rest of the gdb extensions.
7075 @node File Names and DBX
7076 @subsection File Names in DBX Format
7078 @c prevent bad page break with this line
7079 This describes file names in DBX format.
7081 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7082 A C statement to output DBX debugging information to the stdio stream
7083 @var{stream}, which indicates that file @var{name} is the main source
7084 file---the file specified as the input file for compilation.
7085 This macro is called only once, at the beginning of compilation.
7087 This macro need not be defined if the standard form of output
7088 for DBX debugging information is appropriate.
7090 It may be necessary to refer to a label equal to the beginning of the
7091 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
7092 to do so. If you do this, you must also set the variable
7093 @var{used_ltext_label_name} to @code{true}.
7096 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
7097 Define this macro, with value 1, if GCC should not emit an indication
7098 of the current directory for compilation and current source language at
7099 the beginning of the file.
7102 @defmac NO_DBX_GCC_MARKER
7103 Define this macro, with value 1, if GCC should not emit an indication
7104 that this object file was compiled by GCC@. The default is to emit
7105 an @code{N_OPT} stab at the beginning of every source file, with
7106 @samp{gcc2_compiled.} for the string and value 0.
7109 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7110 A C statement to output DBX debugging information at the end of
7111 compilation of the main source file @var{name}. Output should be
7112 written to the stdio stream @var{stream}.
7114 If you don't define this macro, nothing special is output at the end
7115 of compilation, which is correct for most machines.
7118 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
7119 Define this macro @emph{instead of} defining
7120 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
7121 the end of compilation is an @code{N_SO} stab with an empty string,
7122 whose value is the highest absolute text address in the file.
7127 @subsection Macros for SDB and DWARF Output
7129 @c prevent bad page break with this line
7130 Here are macros for SDB and DWARF output.
7132 @defmac SDB_DEBUGGING_INFO
7133 Define this macro if GCC should produce COFF-style debugging output
7134 for SDB in response to the @option{-g} option.
7137 @defmac DWARF2_DEBUGGING_INFO
7138 Define this macro if GCC should produce dwarf version 2 format
7139 debugging output in response to the @option{-g} option.
7141 @hook TARGET_DWARF_CALLING_CONVENTION
7143 To support optional call frame debugging information, you must also
7144 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7145 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7146 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7147 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
7150 @defmac DWARF2_FRAME_INFO
7151 Define this macro to a nonzero value if GCC should always output
7152 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
7153 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
7154 exceptions are enabled, GCC will output this information not matter
7155 how you define @code{DWARF2_FRAME_INFO}.
7158 @hook TARGET_DEBUG_UNWIND_INFO
7160 @defmac DWARF2_ASM_LINE_DEBUG_INFO
7161 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7162 line debug info sections. This will result in much more compact line number
7163 tables, and hence is desirable if it works.
7166 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
7168 @hook TARGET_FORCE_AT_COMP_DIR
7170 @hook TARGET_DELAY_SCHED2
7172 @hook TARGET_DELAY_VARTRACK
7174 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7175 A C statement to issue assembly directives that create a difference
7176 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
7179 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7180 A C statement to issue assembly directives that create a difference
7181 between the two given labels in system defined units, e.g. instruction
7182 slots on IA64 VMS, using an integer of the given size.
7185 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
7186 A C statement to issue assembly directives that create a
7187 section-relative reference to the given @var{label}, using an integer of the
7188 given @var{size}. The label is known to be defined in the given @var{section}.
7191 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
7192 A C statement to issue assembly directives that create a self-relative
7193 reference to the given @var{label}, using an integer of the given @var{size}.
7196 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
7197 A C statement to issue assembly directives that create a reference to
7198 the DWARF table identifier @var{label} from the current section. This
7199 is used on some systems to avoid garbage collecting a DWARF table which
7200 is referenced by a function.
7203 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
7205 @defmac PUT_SDB_@dots{}
7206 Define these macros to override the assembler syntax for the special
7207 SDB assembler directives. See @file{sdbout.c} for a list of these
7208 macros and their arguments. If the standard syntax is used, you need
7209 not define them yourself.
7213 Some assemblers do not support a semicolon as a delimiter, even between
7214 SDB assembler directives. In that case, define this macro to be the
7215 delimiter to use (usually @samp{\n}). It is not necessary to define
7216 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7220 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
7221 Define this macro to allow references to unknown structure,
7222 union, or enumeration tags to be emitted. Standard COFF does not
7223 allow handling of unknown references, MIPS ECOFF has support for
7227 @defmac SDB_ALLOW_FORWARD_REFERENCES
7228 Define this macro to allow references to structure, union, or
7229 enumeration tags that have not yet been seen to be handled. Some
7230 assemblers choke if forward tags are used, while some require it.
7233 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
7234 A C statement to output SDB debugging information before code for line
7235 number @var{line} of the current source file to the stdio stream
7236 @var{stream}. The default is to emit an @code{.ln} directive.
7241 @subsection Macros for VMS Debug Format
7243 @c prevent bad page break with this line
7244 Here are macros for VMS debug format.
7246 @defmac VMS_DEBUGGING_INFO
7247 Define this macro if GCC should produce debugging output for VMS
7248 in response to the @option{-g} option. The default behavior for VMS
7249 is to generate minimal debug info for a traceback in the absence of
7250 @option{-g} unless explicitly overridden with @option{-g0}. This
7251 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
7252 @code{TARGET_OPTION_OVERRIDE}.
7255 @node Floating Point
7256 @section Cross Compilation and Floating Point
7257 @cindex cross compilation and floating point
7258 @cindex floating point and cross compilation
7260 While all modern machines use twos-complement representation for integers,
7261 there are a variety of representations for floating point numbers. This
7262 means that in a cross-compiler the representation of floating point numbers
7263 in the compiled program may be different from that used in the machine
7264 doing the compilation.
7266 Because different representation systems may offer different amounts of
7267 range and precision, all floating point constants must be represented in
7268 the target machine's format. Therefore, the cross compiler cannot
7269 safely use the host machine's floating point arithmetic; it must emulate
7270 the target's arithmetic. To ensure consistency, GCC always uses
7271 emulation to work with floating point values, even when the host and
7272 target floating point formats are identical.
7274 The following macros are provided by @file{real.h} for the compiler to
7275 use. All parts of the compiler which generate or optimize
7276 floating-point calculations must use these macros. They may evaluate
7277 their operands more than once, so operands must not have side effects.
7279 @defmac REAL_VALUE_TYPE
7280 The C data type to be used to hold a floating point value in the target
7281 machine's format. Typically this is a @code{struct} containing an
7282 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7286 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7287 Compares for equality the two values, @var{x} and @var{y}. If the target
7288 floating point format supports negative zeroes and/or NaNs,
7289 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
7290 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
7293 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7294 Tests whether @var{x} is less than @var{y}.
7297 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7298 Truncates @var{x} to a signed integer, rounding toward zero.
7301 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7302 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7303 @var{x} is negative, returns zero.
7306 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
7307 Converts @var{string} into a floating point number in the target machine's
7308 representation for mode @var{mode}. This routine can handle both
7309 decimal and hexadecimal floating point constants, using the syntax
7310 defined by the C language for both.
7313 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7314 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7317 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7318 Determines whether @var{x} represents infinity (positive or negative).
7321 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7322 Determines whether @var{x} represents a ``NaN'' (not-a-number).
7325 @deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7326 Calculates an arithmetic operation on the two floating point values
7327 @var{x} and @var{y}, storing the result in @var{output} (which must be a
7330 The operation to be performed is specified by @var{code}. Only the
7331 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
7332 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
7334 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
7335 target's floating point format cannot represent infinity, it will call
7336 @code{abort}. Callers should check for this situation first, using
7337 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
7340 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
7341 Returns the negative of the floating point value @var{x}.
7344 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
7345 Returns the absolute value of @var{x}.
7348 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
7349 Converts a floating point value @var{x} into a double-precision integer
7350 which is then stored into @var{low} and @var{high}. If the value is not
7351 integral, it is truncated.
7354 @deftypefn Macro void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE @var{x}, HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, enum machine_mode @var{mode})
7355 Converts a double-precision integer found in @var{low} and @var{high},
7356 into a floating point value which is then stored into @var{x}. The
7357 value is truncated to fit in mode @var{mode}.
7360 @node Mode Switching
7361 @section Mode Switching Instructions
7362 @cindex mode switching
7363 The following macros control mode switching optimizations:
7365 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
7366 Define this macro if the port needs extra instructions inserted for mode
7367 switching in an optimizing compilation.
7369 For an example, the SH4 can perform both single and double precision
7370 floating point operations, but to perform a single precision operation,
7371 the FPSCR PR bit has to be cleared, while for a double precision
7372 operation, this bit has to be set. Changing the PR bit requires a general
7373 purpose register as a scratch register, hence these FPSCR sets have to
7374 be inserted before reload, i.e.@: you can't put this into instruction emitting
7375 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
7377 You can have multiple entities that are mode-switched, and select at run time
7378 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7379 return nonzero for any @var{entity} that needs mode-switching.
7380 If you define this macro, you also have to define
7381 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
7382 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
7383 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
7387 @defmac NUM_MODES_FOR_MODE_SWITCHING
7388 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7389 initializer for an array of integers. Each initializer element
7390 N refers to an entity that needs mode switching, and specifies the number
7391 of different modes that might need to be set for this entity.
7392 The position of the initializer in the initializer---starting counting at
7393 zero---determines the integer that is used to refer to the mode-switched
7395 In macros that take mode arguments / yield a mode result, modes are
7396 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
7397 switch is needed / supplied.
7400 @defmac MODE_NEEDED (@var{entity}, @var{insn})
7401 @var{entity} is an integer specifying a mode-switched entity. If
7402 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
7403 return an integer value not larger than the corresponding element in
7404 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
7405 be switched into prior to the execution of @var{insn}.
7408 @defmac MODE_AFTER (@var{entity}, @var{mode}, @var{insn})
7409 @var{entity} is an integer specifying a mode-switched entity. If
7410 this macro is defined, it is evaluated for every @var{insn} during
7411 mode switching. It determines the mode that an insn results in (if
7412 different from the incoming mode).
7415 @defmac MODE_ENTRY (@var{entity})
7416 If this macro is defined, it is evaluated for every @var{entity} that needs
7417 mode switching. It should evaluate to an integer, which is a mode that
7418 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
7419 is defined then @code{MODE_EXIT} must be defined.
7422 @defmac MODE_EXIT (@var{entity})
7423 If this macro is defined, it is evaluated for every @var{entity} that needs
7424 mode switching. It should evaluate to an integer, which is a mode that
7425 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
7426 is defined then @code{MODE_ENTRY} must be defined.
7429 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
7430 This macro specifies the order in which modes for @var{entity} are processed.
7431 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
7432 lowest. The value of the macro should be an integer designating a mode
7433 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
7434 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
7435 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
7438 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
7439 Generate one or more insns to set @var{entity} to @var{mode}.
7440 @var{hard_reg_live} is the set of hard registers live at the point where
7441 the insn(s) are to be inserted.
7444 @node Target Attributes
7445 @section Defining target-specific uses of @code{__attribute__}
7446 @cindex target attributes
7447 @cindex machine attributes
7448 @cindex attributes, target-specific
7450 Target-specific attributes may be defined for functions, data and types.
7451 These are described using the following target hooks; they also need to
7452 be documented in @file{extend.texi}.
7454 @hook TARGET_ATTRIBUTE_TABLE
7456 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
7458 @hook TARGET_COMP_TYPE_ATTRIBUTES
7460 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
7462 @hook TARGET_MERGE_TYPE_ATTRIBUTES
7464 @hook TARGET_MERGE_DECL_ATTRIBUTES
7466 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7468 @defmac TARGET_DECLSPEC
7469 Define this macro to a nonzero value if you want to treat
7470 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
7471 default, this behavior is enabled only for targets that define
7472 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
7473 of @code{__declspec} is via a built-in macro, but you should not rely
7474 on this implementation detail.
7477 @hook TARGET_INSERT_ATTRIBUTES
7479 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7481 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
7483 @hook TARGET_OPTION_SAVE
7485 @hook TARGET_OPTION_RESTORE
7487 @hook TARGET_OPTION_PRINT
7489 @hook TARGET_OPTION_PRAGMA_PARSE
7491 @hook TARGET_OPTION_OVERRIDE
7493 @hook TARGET_OPTION_FUNCTION_VERSIONS
7495 @hook TARGET_CAN_INLINE_P
7498 @section Emulating TLS
7499 @cindex Emulated TLS
7501 For targets whose psABI does not provide Thread Local Storage via
7502 specific relocations and instruction sequences, an emulation layer is
7503 used. A set of target hooks allows this emulation layer to be
7504 configured for the requirements of a particular target. For instance
7505 the psABI may in fact specify TLS support in terms of an emulation
7508 The emulation layer works by creating a control object for every TLS
7509 object. To access the TLS object, a lookup function is provided
7510 which, when given the address of the control object, will return the
7511 address of the current thread's instance of the TLS object.
7513 @hook TARGET_EMUTLS_GET_ADDRESS
7515 @hook TARGET_EMUTLS_REGISTER_COMMON
7517 @hook TARGET_EMUTLS_VAR_SECTION
7519 @hook TARGET_EMUTLS_TMPL_SECTION
7521 @hook TARGET_EMUTLS_VAR_PREFIX
7523 @hook TARGET_EMUTLS_TMPL_PREFIX
7525 @hook TARGET_EMUTLS_VAR_FIELDS
7527 @hook TARGET_EMUTLS_VAR_INIT
7529 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7531 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7533 @node MIPS Coprocessors
7534 @section Defining coprocessor specifics for MIPS targets.
7535 @cindex MIPS coprocessor-definition macros
7537 The MIPS specification allows MIPS implementations to have as many as 4
7538 coprocessors, each with as many as 32 private registers. GCC supports
7539 accessing these registers and transferring values between the registers
7540 and memory using asm-ized variables. For example:
7543 register unsigned int cp0count asm ("c0r1");
7549 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7550 names may be added as described below, or the default names may be
7551 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7553 Coprocessor registers are assumed to be epilogue-used; sets to them will
7554 be preserved even if it does not appear that the register is used again
7555 later in the function.
7557 Another note: according to the MIPS spec, coprocessor 1 (if present) is
7558 the FPU@. One accesses COP1 registers through standard mips
7559 floating-point support; they are not included in this mechanism.
7561 There is one macro used in defining the MIPS coprocessor interface which
7562 you may want to override in subtargets; it is described below.
7565 @section Parameters for Precompiled Header Validity Checking
7566 @cindex parameters, precompiled headers
7568 @hook TARGET_GET_PCH_VALIDITY
7570 @hook TARGET_PCH_VALID_P
7572 @hook TARGET_CHECK_PCH_TARGET_FLAGS
7574 @hook TARGET_PREPARE_PCH_SAVE
7577 @section C++ ABI parameters
7578 @cindex parameters, c++ abi
7580 @hook TARGET_CXX_GUARD_TYPE
7582 @hook TARGET_CXX_GUARD_MASK_BIT
7584 @hook TARGET_CXX_GET_COOKIE_SIZE
7586 @hook TARGET_CXX_COOKIE_HAS_SIZE
7588 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
7590 @hook TARGET_CXX_CDTOR_RETURNS_THIS
7592 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7594 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7596 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7598 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7600 @hook TARGET_CXX_USE_AEABI_ATEXIT
7602 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7604 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7606 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
7608 @node Named Address Spaces
7609 @section Adding support for named address spaces
7610 @cindex named address spaces
7612 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7613 standards committee, @cite{Programming Languages - C - Extensions to
7614 support embedded processors}, specifies a syntax for embedded
7615 processors to specify alternate address spaces. You can configure a
7616 GCC port to support section 5.1 of the draft report to add support for
7617 address spaces other than the default address space. These address
7618 spaces are new keywords that are similar to the @code{volatile} and
7619 @code{const} type attributes.
7621 Pointers to named address spaces can have a different size than
7622 pointers to the generic address space.
7624 For example, the SPU port uses the @code{__ea} address space to refer
7625 to memory in the host processor, rather than memory local to the SPU
7626 processor. Access to memory in the @code{__ea} address space involves
7627 issuing DMA operations to move data between the host processor and the
7628 local processor memory address space. Pointers in the @code{__ea}
7629 address space are either 32 bits or 64 bits based on the
7630 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
7633 Internally, address spaces are represented as a small integer in the
7634 range 0 to 15 with address space 0 being reserved for the generic
7637 To register a named address space qualifier keyword with the C front end,
7638 the target may call the @code{c_register_addr_space} routine. For example,
7639 the SPU port uses the following to declare @code{__ea} as the keyword for
7640 named address space #1:
7642 #define ADDR_SPACE_EA 1
7643 c_register_addr_space ("__ea", ADDR_SPACE_EA);
7646 @hook TARGET_ADDR_SPACE_POINTER_MODE
7648 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
7650 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7652 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7654 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7656 @hook TARGET_ADDR_SPACE_SUBSET_P
7658 @hook TARGET_ADDR_SPACE_CONVERT
7661 @section Miscellaneous Parameters
7662 @cindex parameters, miscellaneous
7664 @c prevent bad page break with this line
7665 Here are several miscellaneous parameters.
7667 @defmac HAS_LONG_COND_BRANCH
7668 Define this boolean macro to indicate whether or not your architecture
7669 has conditional branches that can span all of memory. It is used in
7670 conjunction with an optimization that partitions hot and cold basic
7671 blocks into separate sections of the executable. If this macro is
7672 set to false, gcc will convert any conditional branches that attempt
7673 to cross between sections into unconditional branches or indirect jumps.
7676 @defmac HAS_LONG_UNCOND_BRANCH
7677 Define this boolean macro to indicate whether or not your architecture
7678 has unconditional branches that can span all of memory. It is used in
7679 conjunction with an optimization that partitions hot and cold basic
7680 blocks into separate sections of the executable. If this macro is
7681 set to false, gcc will convert any unconditional branches that attempt
7682 to cross between sections into indirect jumps.
7685 @defmac CASE_VECTOR_MODE
7686 An alias for a machine mode name. This is the machine mode that
7687 elements of a jump-table should have.
7690 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7691 Optional: return the preferred mode for an @code{addr_diff_vec}
7692 when the minimum and maximum offset are known. If you define this,
7693 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7694 To make this work, you also have to define @code{INSN_ALIGN} and
7695 make the alignment for @code{addr_diff_vec} explicit.
7696 The @var{body} argument is provided so that the offset_unsigned and scale
7697 flags can be updated.
7700 @defmac CASE_VECTOR_PC_RELATIVE
7701 Define this macro to be a C expression to indicate when jump-tables
7702 should contain relative addresses. You need not define this macro if
7703 jump-tables never contain relative addresses, or jump-tables should
7704 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7708 @hook TARGET_CASE_VALUES_THRESHOLD
7710 @defmac WORD_REGISTER_OPERATIONS
7711 Define this macro if operations between registers with integral mode
7712 smaller than a word are always performed on the entire register.
7713 Most RISC machines have this property and most CISC machines do not.
7716 @defmac LOAD_EXTEND_OP (@var{mem_mode})
7717 Define this macro to be a C expression indicating when insns that read
7718 memory in @var{mem_mode}, an integral mode narrower than a word, set the
7719 bits outside of @var{mem_mode} to be either the sign-extension or the
7720 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7721 of @var{mem_mode} for which the
7722 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7723 @code{UNKNOWN} for other modes.
7725 This macro is not called with @var{mem_mode} non-integral or with a width
7726 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7727 value in this case. Do not define this macro if it would always return
7728 @code{UNKNOWN}. On machines where this macro is defined, you will normally
7729 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7731 You may return a non-@code{UNKNOWN} value even if for some hard registers
7732 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7733 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
7734 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7735 integral mode larger than this but not larger than @code{word_mode}.
7737 You must return @code{UNKNOWN} if for some hard registers that allow this
7738 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
7739 @code{word_mode}, but that they can change to another integral mode that
7740 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7743 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
7744 Define this macro if loading short immediate values into registers sign
7748 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7751 The maximum number of bytes that a single instruction can move quickly
7752 between memory and registers or between two memory locations.
7755 @defmac MAX_MOVE_MAX
7756 The maximum number of bytes that a single instruction can move quickly
7757 between memory and registers or between two memory locations. If this
7758 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7759 constant value that is the largest value that @code{MOVE_MAX} can have
7763 @defmac SHIFT_COUNT_TRUNCATED
7764 A C expression that is nonzero if on this machine the number of bits
7765 actually used for the count of a shift operation is equal to the number
7766 of bits needed to represent the size of the object being shifted. When
7767 this macro is nonzero, the compiler will assume that it is safe to omit
7768 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7769 truncates the count of a shift operation. On machines that have
7770 instructions that act on bit-fields at variable positions, which may
7771 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7772 also enables deletion of truncations of the values that serve as
7773 arguments to bit-field instructions.
7775 If both types of instructions truncate the count (for shifts) and
7776 position (for bit-field operations), or if no variable-position bit-field
7777 instructions exist, you should define this macro.
7779 However, on some machines, such as the 80386 and the 680x0, truncation
7780 only applies to shift operations and not the (real or pretended)
7781 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7782 such machines. Instead, add patterns to the @file{md} file that include
7783 the implied truncation of the shift instructions.
7785 You need not define this macro if it would always have the value of zero.
7788 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
7789 @hook TARGET_SHIFT_TRUNCATION_MASK
7791 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7792 A C expression which is nonzero if on this machine it is safe to
7793 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7794 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7795 operating on it as if it had only @var{outprec} bits.
7797 On many machines, this expression can be 1.
7799 @c rearranged this, removed the phrase "it is reported that". this was
7800 @c to fix an overfull hbox. --mew 10feb93
7801 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7802 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7803 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7804 such cases may improve things.
7807 @hook TARGET_MODE_REP_EXTENDED
7809 @defmac STORE_FLAG_VALUE
7810 A C expression describing the value returned by a comparison operator
7811 with an integral mode and stored by a store-flag instruction
7812 (@samp{cstore@var{mode}4}) when the condition is true. This description must
7813 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7814 comparison operators whose results have a @code{MODE_INT} mode.
7816 A value of 1 or @minus{}1 means that the instruction implementing the
7817 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7818 and 0 when the comparison is false. Otherwise, the value indicates
7819 which bits of the result are guaranteed to be 1 when the comparison is
7820 true. This value is interpreted in the mode of the comparison
7821 operation, which is given by the mode of the first operand in the
7822 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
7823 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7826 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7827 generate code that depends only on the specified bits. It can also
7828 replace comparison operators with equivalent operations if they cause
7829 the required bits to be set, even if the remaining bits are undefined.
7830 For example, on a machine whose comparison operators return an
7831 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7832 @samp{0x80000000}, saying that just the sign bit is relevant, the
7836 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7843 (ashift:SI @var{x} (const_int @var{n}))
7847 where @var{n} is the appropriate shift count to move the bit being
7848 tested into the sign bit.
7850 There is no way to describe a machine that always sets the low-order bit
7851 for a true value, but does not guarantee the value of any other bits,
7852 but we do not know of any machine that has such an instruction. If you
7853 are trying to port GCC to such a machine, include an instruction to
7854 perform a logical-and of the result with 1 in the pattern for the
7855 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7857 Often, a machine will have multiple instructions that obtain a value
7858 from a comparison (or the condition codes). Here are rules to guide the
7859 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7864 Use the shortest sequence that yields a valid definition for
7865 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7866 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7867 comparison operators to do so because there may be opportunities to
7868 combine the normalization with other operations.
7871 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7872 slightly preferred on machines with expensive jumps and 1 preferred on
7876 As a second choice, choose a value of @samp{0x80000001} if instructions
7877 exist that set both the sign and low-order bits but do not define the
7881 Otherwise, use a value of @samp{0x80000000}.
7884 Many machines can produce both the value chosen for
7885 @code{STORE_FLAG_VALUE} and its negation in the same number of
7886 instructions. On those machines, you should also define a pattern for
7887 those cases, e.g., one matching
7890 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7893 Some machines can also perform @code{and} or @code{plus} operations on
7894 condition code values with less instructions than the corresponding
7895 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
7896 machines, define the appropriate patterns. Use the names @code{incscc}
7897 and @code{decscc}, respectively, for the patterns which perform
7898 @code{plus} or @code{minus} operations on condition code values. See
7899 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
7900 find such instruction sequences on other machines.
7902 If this macro is not defined, the default value, 1, is used. You need
7903 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7904 instructions, or if the value generated by these instructions is 1.
7907 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7908 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7909 returned when comparison operators with floating-point results are true.
7910 Define this macro on machines that have comparison operations that return
7911 floating-point values. If there are no such operations, do not define
7915 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7916 A C expression that gives a rtx representing the nonzero true element
7917 for vector comparisons. The returned rtx should be valid for the inner
7918 mode of @var{mode} which is guaranteed to be a vector mode. Define
7919 this macro on machines that have vector comparison operations that
7920 return a vector result. If there are no such operations, do not define
7921 this macro. Typically, this macro is defined as @code{const1_rtx} or
7922 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
7923 the compiler optimizing such vector comparison operations for the
7927 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7928 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7929 A C expression that indicates whether the architecture defines a value
7930 for @code{clz} or @code{ctz} with a zero operand.
7931 A result of @code{0} indicates the value is undefined.
7932 If the value is defined for only the RTL expression, the macro should
7933 evaluate to @code{1}; if the value applies also to the corresponding optab
7934 entry (which is normally the case if it expands directly into
7935 the corresponding RTL), then the macro should evaluate to @code{2}.
7936 In the cases where the value is defined, @var{value} should be set to
7939 If this macro is not defined, the value of @code{clz} or
7940 @code{ctz} at zero is assumed to be undefined.
7942 This macro must be defined if the target's expansion for @code{ffs}
7943 relies on a particular value to get correct results. Otherwise it
7944 is not necessary, though it may be used to optimize some corner cases, and
7945 to provide a default expansion for the @code{ffs} optab.
7947 Note that regardless of this macro the ``definedness'' of @code{clz}
7948 and @code{ctz} at zero do @emph{not} extend to the builtin functions
7949 visible to the user. Thus one may be free to adjust the value at will
7950 to match the target expansion of these operations without fear of
7955 An alias for the machine mode for pointers. On most machines, define
7956 this to be the integer mode corresponding to the width of a hardware
7957 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7958 On some machines you must define this to be one of the partial integer
7959 modes, such as @code{PSImode}.
7961 The width of @code{Pmode} must be at least as large as the value of
7962 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7963 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7967 @defmac FUNCTION_MODE
7968 An alias for the machine mode used for memory references to functions
7969 being called, in @code{call} RTL expressions. On most CISC machines,
7970 where an instruction can begin at any byte address, this should be
7971 @code{QImode}. On most RISC machines, where all instructions have fixed
7972 size and alignment, this should be a mode with the same size and alignment
7973 as the machine instruction words - typically @code{SImode} or @code{HImode}.
7976 @defmac STDC_0_IN_SYSTEM_HEADERS
7977 In normal operation, the preprocessor expands @code{__STDC__} to the
7978 constant 1, to signify that GCC conforms to ISO Standard C@. On some
7979 hosts, like Solaris, the system compiler uses a different convention,
7980 where @code{__STDC__} is normally 0, but is 1 if the user specifies
7981 strict conformance to the C Standard.
7983 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7984 convention when processing system header files, but when processing user
7985 files @code{__STDC__} will always expand to 1.
7988 @hook TARGET_C_PREINCLUDE
7990 @hook TARGET_CXX_IMPLICIT_EXTERN_C
7992 @defmac NO_IMPLICIT_EXTERN_C
7993 Define this macro if the system header files support C++ as well as C@.
7994 This macro inhibits the usual method of using system header files in
7995 C++, which is to pretend that the file's contents are enclosed in
7996 @samp{extern "C" @{@dots{}@}}.
8001 @defmac REGISTER_TARGET_PRAGMAS ()
8002 Define this macro if you want to implement any target-specific pragmas.
8003 If defined, it is a C expression which makes a series of calls to
8004 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
8005 for each pragma. The macro may also do any
8006 setup required for the pragmas.
8008 The primary reason to define this macro is to provide compatibility with
8009 other compilers for the same target. In general, we discourage
8010 definition of target-specific pragmas for GCC@.
8012 If the pragma can be implemented by attributes then you should consider
8013 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8015 Preprocessor macros that appear on pragma lines are not expanded. All
8016 @samp{#pragma} directives that do not match any registered pragma are
8017 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8020 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8021 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8023 Each call to @code{c_register_pragma} or
8024 @code{c_register_pragma_with_expansion} establishes one pragma. The
8025 @var{callback} routine will be called when the preprocessor encounters a
8029 #pragma [@var{space}] @var{name} @dots{}
8032 @var{space} is the case-sensitive namespace of the pragma, or
8033 @code{NULL} to put the pragma in the global namespace. The callback
8034 routine receives @var{pfile} as its first argument, which can be passed
8035 on to cpplib's functions if necessary. You can lex tokens after the
8036 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
8037 callback will be silently ignored. The end of the line is indicated by
8038 a token of type @code{CPP_EOF}. Macro expansion occurs on the
8039 arguments of pragmas registered with
8040 @code{c_register_pragma_with_expansion} but not on the arguments of
8041 pragmas registered with @code{c_register_pragma}.
8043 Note that the use of @code{pragma_lex} is specific to the C and C++
8044 compilers. It will not work in the Java or Fortran compilers, or any
8045 other language compilers for that matter. Thus if @code{pragma_lex} is going
8046 to be called from target-specific code, it must only be done so when
8047 building the C and C++ compilers. This can be done by defining the
8048 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8049 target entry in the @file{config.gcc} file. These variables should name
8050 the target-specific, language-specific object file which contains the
8051 code that uses @code{pragma_lex}. Note it will also be necessary to add a
8052 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8053 how to build this object file.
8056 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
8057 Define this macro if macros should be expanded in the
8058 arguments of @samp{#pragma pack}.
8061 @defmac TARGET_DEFAULT_PACK_STRUCT
8062 If your target requires a structure packing default other than 0 (meaning
8063 the machine default), define this macro to the necessary value (in bytes).
8064 This must be a value that would also be valid to use with
8065 @samp{#pragma pack()} (that is, a small power of two).
8068 @defmac DOLLARS_IN_IDENTIFIERS
8069 Define this macro to control use of the character @samp{$} in
8070 identifier names for the C family of languages. 0 means @samp{$} is
8071 not allowed by default; 1 means it is allowed. 1 is the default;
8072 there is no need to define this macro in that case.
8075 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
8076 Define this macro as a C expression that is nonzero if it is safe for the
8077 delay slot scheduler to place instructions in the delay slot of @var{insn},
8078 even if they appear to use a resource set or clobbered in @var{insn}.
8079 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8080 every @code{call_insn} has this behavior. On machines where some @code{insn}
8081 or @code{jump_insn} is really a function call and hence has this behavior,
8082 you should define this macro.
8084 You need not define this macro if it would always return zero.
8087 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
8088 Define this macro as a C expression that is nonzero if it is safe for the
8089 delay slot scheduler to place instructions in the delay slot of @var{insn},
8090 even if they appear to set or clobber a resource referenced in @var{insn}.
8091 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8092 some @code{insn} or @code{jump_insn} is really a function call and its operands
8093 are registers whose use is actually in the subroutine it calls, you should
8094 define this macro. Doing so allows the delay slot scheduler to move
8095 instructions which copy arguments into the argument registers into the delay
8098 You need not define this macro if it would always return zero.
8101 @defmac MULTIPLE_SYMBOL_SPACES
8102 Define this macro as a C expression that is nonzero if, in some cases,
8103 global symbols from one translation unit may not be bound to undefined
8104 symbols in another translation unit without user intervention. For
8105 instance, under Microsoft Windows symbols must be explicitly imported
8106 from shared libraries (DLLs).
8108 You need not define this macro if it would always evaluate to zero.
8111 @hook TARGET_MD_ASM_CLOBBERS
8113 @defmac MATH_LIBRARY
8114 Define this macro as a C string constant for the linker argument to link
8115 in the system math library, minus the initial @samp{"-l"}, or
8116 @samp{""} if the target does not have a
8117 separate math library.
8119 You need only define this macro if the default of @samp{"m"} is wrong.
8122 @defmac LIBRARY_PATH_ENV
8123 Define this macro as a C string constant for the environment variable that
8124 specifies where the linker should look for libraries.
8126 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8130 @defmac TARGET_POSIX_IO
8131 Define this macro if the target supports the following POSIX@ file
8132 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
8133 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
8134 to use file locking when exiting a program, which avoids race conditions
8135 if the program has forked. It will also create directories at run-time
8136 for cross-profiling.
8139 @defmac MAX_CONDITIONAL_EXECUTE
8141 A C expression for the maximum number of instructions to execute via
8142 conditional execution instructions instead of a branch. A value of
8143 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8144 1 if it does use cc0.
8147 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
8148 Used if the target needs to perform machine-dependent modifications on the
8149 conditionals used for turning basic blocks into conditionally executed code.
8150 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8151 contains information about the currently processed blocks. @var{true_expr}
8152 and @var{false_expr} are the tests that are used for converting the
8153 then-block and the else-block, respectively. Set either @var{true_expr} or
8154 @var{false_expr} to a null pointer if the tests cannot be converted.
8157 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
8158 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
8159 if-statements into conditions combined by @code{and} and @code{or} operations.
8160 @var{bb} contains the basic block that contains the test that is currently
8161 being processed and about to be turned into a condition.
8164 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
8165 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
8166 be converted to conditional execution format. @var{ce_info} points to
8167 a data structure, @code{struct ce_if_block}, which contains information
8168 about the currently processed blocks.
8171 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
8172 A C expression to perform any final machine dependent modifications in
8173 converting code to conditional execution. The involved basic blocks
8174 can be found in the @code{struct ce_if_block} structure that is pointed
8175 to by @var{ce_info}.
8178 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
8179 A C expression to cancel any machine dependent modifications in
8180 converting code to conditional execution. The involved basic blocks
8181 can be found in the @code{struct ce_if_block} structure that is pointed
8182 to by @var{ce_info}.
8185 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
8186 A C expression to initialize any machine specific data for if-conversion
8187 of the if-block in the @code{struct ce_if_block} structure that is pointed
8188 to by @var{ce_info}.
8191 @hook TARGET_MACHINE_DEPENDENT_REORG
8193 @hook TARGET_INIT_BUILTINS
8195 @hook TARGET_BUILTIN_DECL
8197 @hook TARGET_EXPAND_BUILTIN
8199 @hook TARGET_BUILTIN_CHKP_FUNCTION
8200 @hook TARGET_CHKP_BOUND_TYPE
8201 @hook TARGET_CHKP_BOUND_MODE
8203 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
8205 @hook TARGET_FOLD_BUILTIN
8207 @hook TARGET_GIMPLE_FOLD_BUILTIN
8209 @hook TARGET_COMPARE_VERSION_PRIORITY
8211 @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
8213 @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
8215 @hook TARGET_INVALID_WITHIN_DOLOOP
8217 @hook TARGET_LEGITIMATE_COMBINED_INSN
8219 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
8221 Take a branch insn in @var{branch1} and another in @var{branch2}.
8222 Return true if redirecting @var{branch1} to the destination of
8223 @var{branch2} is possible.
8225 On some targets, branches may have a limited range. Optimizing the
8226 filling of delay slots can result in branches being redirected, and this
8227 may in turn cause a branch offset to overflow.
8230 @hook TARGET_CAN_FOLLOW_JUMP
8232 @hook TARGET_COMMUTATIVE_P
8234 @hook TARGET_ALLOCATE_INITIAL_VALUE
8236 @hook TARGET_UNSPEC_MAY_TRAP_P
8238 @hook TARGET_SET_CURRENT_FUNCTION
8240 @defmac TARGET_OBJECT_SUFFIX
8241 Define this macro to be a C string representing the suffix for object
8242 files on your target machine. If you do not define this macro, GCC will
8243 use @samp{.o} as the suffix for object files.
8246 @defmac TARGET_EXECUTABLE_SUFFIX
8247 Define this macro to be a C string representing the suffix to be
8248 automatically added to executable files on your target machine. If you
8249 do not define this macro, GCC will use the null string as the suffix for
8253 @defmac COLLECT_EXPORT_LIST
8254 If defined, @code{collect2} will scan the individual object files
8255 specified on its command line and create an export list for the linker.
8256 Define this macro for systems like AIX, where the linker discards
8257 object files that are not referenced from @code{main} and uses export
8261 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
8262 Define this macro to a C expression representing a variant of the
8263 method call @var{mdecl}, if Java Native Interface (JNI) methods
8264 must be invoked differently from other methods on your target.
8265 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
8266 the @code{stdcall} calling convention and this macro is then
8267 defined as this expression:
8270 build_type_attribute_variant (@var{mdecl},
8272 (get_identifier ("stdcall"),
8277 @hook TARGET_CANNOT_MODIFY_JUMPS_P
8279 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
8281 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
8283 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
8285 @hook TARGET_LOOP_UNROLL_ADJUST
8287 @defmac POWI_MAX_MULTS
8288 If defined, this macro is interpreted as a signed integer C expression
8289 that specifies the maximum number of floating point multiplications
8290 that should be emitted when expanding exponentiation by an integer
8291 constant inline. When this value is defined, exponentiation requiring
8292 more than this number of multiplications is implemented by calling the
8293 system library's @code{pow}, @code{powf} or @code{powl} routines.
8294 The default value places no upper bound on the multiplication count.
8297 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8298 This target hook should register any extra include files for the
8299 target. The parameter @var{stdinc} indicates if normal include files
8300 are present. The parameter @var{sysroot} is the system root directory.
8301 The parameter @var{iprefix} is the prefix for the gcc directory.
8304 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8305 This target hook should register any extra include files for the
8306 target before any standard headers. The parameter @var{stdinc}
8307 indicates if normal include files are present. The parameter
8308 @var{sysroot} is the system root directory. The parameter
8309 @var{iprefix} is the prefix for the gcc directory.
8312 @deftypefn Macro void TARGET_OPTF (char *@var{path})
8313 This target hook should register special include paths for the target.
8314 The parameter @var{path} is the include to register. On Darwin
8315 systems, this is used for Framework includes, which have semantics
8316 that are different from @option{-I}.
8319 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
8320 This target macro returns @code{true} if it is safe to use a local alias
8321 for a virtual function @var{fndecl} when constructing thunks,
8322 @code{false} otherwise. By default, the macro returns @code{true} for all
8323 functions, if a target supports aliases (i.e.@: defines
8324 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
8327 @defmac TARGET_FORMAT_TYPES
8328 If defined, this macro is the name of a global variable containing
8329 target-specific format checking information for the @option{-Wformat}
8330 option. The default is to have no target-specific format checks.
8333 @defmac TARGET_N_FORMAT_TYPES
8334 If defined, this macro is the number of entries in
8335 @code{TARGET_FORMAT_TYPES}.
8338 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
8339 If defined, this macro is the name of a global variable containing
8340 target-specific format overrides for the @option{-Wformat} option. The
8341 default is to have no target-specific format overrides. If defined,
8342 @code{TARGET_FORMAT_TYPES} must be defined, too.
8345 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
8346 If defined, this macro specifies the number of entries in
8347 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
8350 @defmac TARGET_OVERRIDES_FORMAT_INIT
8351 If defined, this macro specifies the optional initialization
8352 routine for target specific customizations of the system printf
8353 and scanf formatter settings.
8356 @hook TARGET_RELAXED_ORDERING
8358 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
8360 @hook TARGET_INVALID_CONVERSION
8362 @hook TARGET_INVALID_UNARY_OP
8364 @hook TARGET_INVALID_BINARY_OP
8366 @hook TARGET_INVALID_PARAMETER_TYPE
8368 @hook TARGET_INVALID_RETURN_TYPE
8370 @hook TARGET_PROMOTED_TYPE
8372 @hook TARGET_CONVERT_TO_TYPE
8374 @defmac TARGET_USE_JCR_SECTION
8375 This macro determines whether to use the JCR section to register Java
8376 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
8377 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
8381 This macro determines the size of the objective C jump buffer for the
8382 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
8385 @defmac LIBGCC2_UNWIND_ATTRIBUTE
8386 Define this macro if any target-specific attributes need to be attached
8387 to the functions in @file{libgcc} that provide low-level support for
8388 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
8389 and the associated definitions of those functions.
8392 @hook TARGET_UPDATE_STACK_BOUNDARY
8394 @hook TARGET_GET_DRAP_RTX
8396 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
8398 @hook TARGET_CONST_ANCHOR
8400 @hook TARGET_ASAN_SHADOW_OFFSET
8402 @hook TARGET_MEMMODEL_CHECK
8404 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
8406 @hook TARGET_HAS_IFUNC_P
8408 @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV