1 @c Copyright (C) 1988-2020 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Target Description Macros and Functions
7 @cindex machine description macros
8 @cindex target description macros
9 @cindex macros, target description
10 @cindex @file{tm.h} macros
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
15 The header file defines numerous macros that convey the information
16 about the target machine that does not fit into the scheme of the
17 @file{.md} file. The file @file{tm.h} should be a link to
18 @file{@var{machine}.h}. The header file @file{config.h} includes
19 @file{tm.h} and most compiler source files include @file{config.h}. The
20 source file defines a variable @code{targetm}, which is a structure
21 containing pointers to functions and data relating to the target
22 machine. @file{@var{machine}.c} should also contain their definitions,
23 if they are not defined elsewhere in GCC, and other functions called
24 through the macros defined in the @file{.h} file.
27 * Target Structure:: The @code{targetm} variable.
28 * Driver:: Controlling how the driver runs the compilation passes.
29 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
30 * Per-Function Data:: Defining data structures for per-function information.
31 * Storage Layout:: Defining sizes and alignments of data.
32 * Type Layout:: Defining sizes and properties of basic user data types.
33 * Registers:: Naming and describing the hardware registers.
34 * Register Classes:: Defining the classes of hardware registers.
35 * Stack and Calling:: Defining which way the stack grows and by how much.
36 * Varargs:: Defining the varargs macros.
37 * Trampolines:: Code set up at run time to enter a nested function.
38 * Library Calls:: Controlling how library routines are implicitly called.
39 * Addressing Modes:: Defining addressing modes valid for memory operands.
40 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
41 * Condition Code:: Defining how insns update the condition code.
42 * Costs:: Defining relative costs of different operations.
43 * Scheduling:: Adjusting the behavior of the instruction scheduler.
44 * Sections:: Dividing storage into text, data, and other sections.
45 * PIC:: Macros for position independent code.
46 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
47 * Debugging Info:: Defining the format of debugging output.
48 * Floating Point:: Handling floating point for cross-compilers.
49 * Mode Switching:: Insertion of mode-switching instructions.
50 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
51 * Emulated TLS:: Emulated TLS support.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * PCH Target:: Validity checking for precompiled headers.
54 * C++ ABI:: Controlling C++ ABI changes.
55 * D Language and ABI:: Controlling D 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.
110 Similarly, there is a @code{targetdm} variable for hooks that are
111 specific to the D language front end, documented as ``D Target Hook''.
112 This is declared in @file{d/d-target.h}, the initializer
113 @code{TARGETDM_INITIALIZER} in @file{d/d-target-def.h}. If targets
114 initialize @code{targetdm} themselves, they should set
115 @code{target_has_targetdm=yes} in @file{config.gcc}; otherwise a default
119 @section Controlling the Compilation Driver, @file{gcc}
121 @cindex controlling the compilation driver
123 @c prevent bad page break with this line
124 You can control the compilation driver.
126 @defmac DRIVER_SELF_SPECS
127 A list of specs for the driver itself. It should be a suitable
128 initializer for an array of strings, with no surrounding braces.
130 The driver applies these specs to its own command line between loading
131 default @file{specs} files (but not command-line specified ones) and
132 choosing the multilib directory or running any subcommands. It
133 applies them in the order given, so each spec can depend on the
134 options added by earlier ones. It is also possible to remove options
135 using @samp{%<@var{option}} in the usual way.
137 This macro can be useful when a port has several interdependent target
138 options. It provides a way of standardizing the command line so
139 that the other specs are easier to write.
141 Do not define this macro if it does not need to do anything.
144 @defmac OPTION_DEFAULT_SPECS
145 A list of specs used to support configure-time default options (i.e.@:
146 @option{--with} options) in the driver. It should be a suitable initializer
147 for an array of structures, each containing two strings, without the
148 outermost pair of surrounding braces.
150 The first item in the pair is the name of the default. This must match
151 the code in @file{config.gcc} for the target. The second item is a spec
152 to apply if a default with this name was specified. The string
153 @samp{%(VALUE)} in the spec will be replaced by the value of the default
154 everywhere it occurs.
156 The driver will apply these specs to its own command line between loading
157 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
158 the same mechanism as @code{DRIVER_SELF_SPECS}.
160 Do not define this macro if it does not need to do anything.
164 A C string constant that tells the GCC driver program options to
165 pass to CPP@. It can also specify how to translate options you
166 give to GCC into options for GCC to pass to the CPP@.
168 Do not define this macro if it does not need to do anything.
171 @defmac CPLUSPLUS_CPP_SPEC
172 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
173 than C@. If you do not define this macro, then the value of
174 @code{CPP_SPEC} (if any) will be used instead.
178 A C string constant that tells the GCC driver program options to
179 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
181 It can also specify how to translate options you give to GCC into options
182 for GCC to pass to front ends.
184 Do not define this macro if it does not need to do anything.
188 A C string constant that tells the GCC driver program options to
189 pass to @code{cc1plus}. It can also specify how to translate options you
190 give to GCC into options for GCC to pass to the @code{cc1plus}.
192 Do not define this macro if it does not need to do anything.
193 Note that everything defined in CC1_SPEC is already passed to
194 @code{cc1plus} so there is no need to duplicate the contents of
195 CC1_SPEC in CC1PLUS_SPEC@.
199 A C string constant that tells the GCC driver program options to
200 pass to the assembler. It can also specify how to translate options
201 you give to GCC into options for GCC to pass to the assembler.
202 See the file @file{sun3.h} for an example of this.
204 Do not define this macro if it does not need to do anything.
207 @defmac ASM_FINAL_SPEC
208 A C string constant that tells the GCC driver program how to
209 run any programs which cleanup after the normal assembler.
210 Normally, this is not needed. See the file @file{mips.h} for
213 Do not define this macro if it does not need to do anything.
216 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
217 Define this macro, with no value, if the driver should give the assembler
218 an argument consisting of a single dash, @option{-}, to instruct it to
219 read from its standard input (which will be a pipe connected to the
220 output of the compiler proper). This argument is given after any
221 @option{-o} option specifying the name of the output file.
223 If you do not define this macro, the assembler is assumed to read its
224 standard input if given no non-option arguments. If your assembler
225 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
226 see @file{mips.h} for instance.
230 A C string constant that tells the GCC driver program options to
231 pass to the linker. It can also specify how to translate options you
232 give to GCC into options for GCC to pass to the linker.
234 Do not define this macro if it does not need to do anything.
238 Another C string constant used much like @code{LINK_SPEC}. The difference
239 between the two is that @code{LIB_SPEC} is used at the end of the
240 command given to the linker.
242 If this macro is not defined, a default is provided that
243 loads the standard C library from the usual place. See @file{gcc.c}.
247 Another C string constant that tells the GCC driver program
248 how and when to place a reference to @file{libgcc.a} into the
249 linker command line. This constant is placed both before and after
250 the value of @code{LIB_SPEC}.
252 If this macro is not defined, the GCC driver provides a default that
253 passes the string @option{-lgcc} to the linker.
256 @defmac REAL_LIBGCC_SPEC
257 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
258 @code{LIBGCC_SPEC} is not directly used by the driver program but is
259 instead modified to refer to different versions of @file{libgcc.a}
260 depending on the values of the command line flags @option{-static},
261 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
262 targets where these modifications are inappropriate, define
263 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
264 driver how to place a reference to @file{libgcc} on the link command
265 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
268 @defmac USE_LD_AS_NEEDED
269 A macro that controls the modifications to @code{LIBGCC_SPEC}
270 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
271 generated that uses @option{--as-needed} or equivalent options and the
272 shared @file{libgcc} in place of the
273 static exception handler library, when linking without any of
274 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
278 If defined, this C string constant is added to @code{LINK_SPEC}.
279 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
280 the modifications to @code{LIBGCC_SPEC} mentioned in
281 @code{REAL_LIBGCC_SPEC}.
284 @defmac STARTFILE_SPEC
285 Another C string constant used much like @code{LINK_SPEC}. The
286 difference between the two is that @code{STARTFILE_SPEC} is used at
287 the very beginning of the command given to the linker.
289 If this macro is not defined, a default is provided that loads the
290 standard C startup file from the usual place. See @file{gcc.c}.
294 Another C string constant used much like @code{LINK_SPEC}. The
295 difference between the two is that @code{ENDFILE_SPEC} is used at
296 the very end of the command given to the linker.
298 Do not define this macro if it does not need to do anything.
301 @defmac THREAD_MODEL_SPEC
302 GCC @code{-v} will print the thread model GCC was configured to use.
303 However, this doesn't work on platforms that are multilibbed on thread
304 models, such as AIX 4.3. On such platforms, define
305 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
306 blanks that names one of the recognized thread models. @code{%*}, the
307 default value of this macro, will expand to the value of
308 @code{thread_file} set in @file{config.gcc}.
311 @defmac SYSROOT_SUFFIX_SPEC
312 Define this macro to add a suffix to the target sysroot when GCC is
313 configured with a sysroot. This will cause GCC to search for usr/lib,
314 et al, within sysroot+suffix.
317 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
318 Define this macro to add a headers_suffix to the target sysroot when
319 GCC is configured with a sysroot. This will cause GCC to pass the
320 updated sysroot+headers_suffix to CPP, causing it to search for
321 usr/include, et al, within sysroot+headers_suffix.
325 Define this macro to provide additional specifications to put in the
326 @file{specs} file that can be used in various specifications like
329 The definition should be an initializer for an array of structures,
330 containing a string constant, that defines the specification name, and a
331 string constant that provides the specification.
333 Do not define this macro if it does not need to do anything.
335 @code{EXTRA_SPECS} is useful when an architecture contains several
336 related targets, which have various @code{@dots{}_SPECS} which are similar
337 to each other, and the maintainer would like one central place to keep
340 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
341 define either @code{_CALL_SYSV} when the System V calling sequence is
342 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
345 The @file{config/rs6000/rs6000.h} target file defines:
348 #define EXTRA_SPECS \
349 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
351 #define CPP_SYS_DEFAULT ""
354 The @file{config/rs6000/sysv.h} target file defines:
358 "%@{posix: -D_POSIX_SOURCE @} \
359 %@{mcall-sysv: -D_CALL_SYSV @} \
360 %@{!mcall-sysv: %(cpp_sysv_default) @} \
361 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
363 #undef CPP_SYSV_DEFAULT
364 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
367 while the @file{config/rs6000/eabiaix.h} target file defines
368 @code{CPP_SYSV_DEFAULT} as:
371 #undef CPP_SYSV_DEFAULT
372 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
376 @defmac LINK_LIBGCC_SPECIAL_1
377 Define this macro if the driver program should find the library
378 @file{libgcc.a}. If you do not define this macro, the driver program will pass
379 the argument @option{-lgcc} to tell the linker to do the search.
382 @defmac LINK_GCC_C_SEQUENCE_SPEC
383 The sequence in which libgcc and libc are specified to the linker.
384 By default this is @code{%G %L %G}.
387 @defmac POST_LINK_SPEC
388 Define this macro to add additional steps to be executed after linker.
389 The default value of this macro is empty string.
392 @defmac LINK_COMMAND_SPEC
393 A C string constant giving the complete command line need to execute the
394 linker. When you do this, you will need to update your port each time a
395 change is made to the link command line within @file{gcc.c}. Therefore,
396 define this macro only if you need to completely redefine the command
397 line for invoking the linker and there is no other way to accomplish
398 the effect you need. Overriding this macro may be avoidable by overriding
399 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
402 @hook TARGET_ALWAYS_STRIP_DOTDOT
404 @defmac MULTILIB_DEFAULTS
405 Define this macro as a C expression for the initializer of an array of
406 string to tell the driver program which options are defaults for this
407 target and thus do not need to be handled specially when using
408 @code{MULTILIB_OPTIONS}.
410 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
411 the target makefile fragment or if none of the options listed in
412 @code{MULTILIB_OPTIONS} are set by default.
413 @xref{Target Fragment}.
416 @defmac RELATIVE_PREFIX_NOT_LINKDIR
417 Define this macro to tell @command{gcc} that it should only translate
418 a @option{-B} prefix into a @option{-L} linker option if the prefix
419 indicates an absolute file name.
422 @defmac MD_EXEC_PREFIX
423 If defined, this macro is an additional prefix to try after
424 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
425 when the compiler is built as a cross
426 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
427 to the list of directories used to find the assembler in @file{configure.ac}.
430 @defmac STANDARD_STARTFILE_PREFIX
431 Define this macro as a C string constant if you wish to override the
432 standard choice of @code{libdir} as the default prefix to
433 try when searching for startup files such as @file{crt0.o}.
434 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
435 is built as a cross compiler.
438 @defmac STANDARD_STARTFILE_PREFIX_1
439 Define this macro as a C string constant if you wish to override the
440 standard choice of @code{/lib} as a prefix to try after the default prefix
441 when searching for startup files such as @file{crt0.o}.
442 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
443 is built as a cross compiler.
446 @defmac STANDARD_STARTFILE_PREFIX_2
447 Define this macro as a C string constant if you wish to override the
448 standard choice of @code{/lib} as yet another prefix to try after the
449 default prefix when searching for startup files such as @file{crt0.o}.
450 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
451 is built as a cross compiler.
454 @defmac MD_STARTFILE_PREFIX
455 If defined, this macro supplies an additional prefix to try after the
456 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
457 compiler is built as a cross compiler.
460 @defmac MD_STARTFILE_PREFIX_1
461 If defined, this macro supplies yet another prefix to try after the
462 standard prefixes. It is not searched when the compiler is built as a
466 @defmac INIT_ENVIRONMENT
467 Define this macro as a C string constant if you wish to set environment
468 variables for programs called by the driver, such as the assembler and
469 loader. The driver passes the value of this macro to @code{putenv} to
470 initialize the necessary environment variables.
473 @defmac LOCAL_INCLUDE_DIR
474 Define this macro as a C string constant if you wish to override the
475 standard choice of @file{/usr/local/include} as the default prefix to
476 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
477 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
478 @file{config.gcc}, normally @file{/usr/include}) in the search order.
480 Cross compilers do not search either @file{/usr/local/include} or its
484 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
485 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
486 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
487 If you do not define this macro, no component is used.
490 @defmac INCLUDE_DEFAULTS
491 Define this macro if you wish to override the entire default search path
492 for include files. For a native compiler, the default search path
493 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
494 @code{GPLUSPLUS_INCLUDE_DIR}, and
495 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
496 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
497 and specify private search areas for GCC@. The directory
498 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
500 The definition should be an initializer for an array of structures.
501 Each array element should have four elements: the directory name (a
502 string constant), the component name (also a string constant), a flag
503 for C++-only directories,
504 and a flag showing that the includes in the directory don't need to be
505 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
506 the array with a null element.
508 The component name denotes what GNU package the include file is part of,
509 if any, in all uppercase letters. For example, it might be @samp{GCC}
510 or @samp{BINUTILS}. If the package is part of a vendor-supplied
511 operating system, code the component name as @samp{0}.
513 For example, here is the definition used for VAX/VMS:
516 #define INCLUDE_DEFAULTS \
518 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
519 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
520 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
527 Here is the order of prefixes tried for exec files:
531 Any prefixes specified by the user with @option{-B}.
534 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
535 is not set and the compiler has not been installed in the configure-time
536 @var{prefix}, the location in which the compiler has actually been installed.
539 The directories specified by the environment variable @code{COMPILER_PATH}.
542 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
543 in the configured-time @var{prefix}.
546 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
549 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
552 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
556 Here is the order of prefixes tried for startfiles:
560 Any prefixes specified by the user with @option{-B}.
563 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
564 value based on the installed toolchain location.
567 The directories specified by the environment variable @code{LIBRARY_PATH}
568 (or port-specific name; native only, cross compilers do not use this).
571 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
572 in the configured @var{prefix} or this is a native compiler.
575 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
578 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
582 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
583 native compiler, or we have a target system root.
586 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
587 native compiler, or we have a target system root.
590 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
591 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
592 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
595 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
596 compiler, or we have a target system root. The default for this macro is
600 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
601 compiler, or we have a target system root. The default for this macro is
605 @node Run-time Target
606 @section Run-time Target Specification
607 @cindex run-time target specification
608 @cindex predefined macros
609 @cindex target specifications
611 @c prevent bad page break with this line
612 Here are run-time target specifications.
614 @defmac TARGET_CPU_CPP_BUILTINS ()
615 This function-like macro expands to a block of code that defines
616 built-in preprocessor macros and assertions for the target CPU, using
617 the functions @code{builtin_define}, @code{builtin_define_std} and
618 @code{builtin_assert}. When the front end
619 calls this macro it provides a trailing semicolon, and since it has
620 finished command line option processing your code can use those
623 @code{builtin_assert} takes a string in the form you pass to the
624 command-line option @option{-A}, such as @code{cpu=mips}, and creates
625 the assertion. @code{builtin_define} takes a string in the form
626 accepted by option @option{-D} and unconditionally defines the macro.
628 @code{builtin_define_std} takes a string representing the name of an
629 object-like macro. If it doesn't lie in the user's namespace,
630 @code{builtin_define_std} defines it unconditionally. Otherwise, it
631 defines a version with two leading underscores, and another version
632 with two leading and trailing underscores, and defines the original
633 only if an ISO standard was not requested on the command line. For
634 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
635 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
636 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
637 defines only @code{_ABI64}.
639 You can also test for the C dialect being compiled. The variable
640 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
641 or @code{clk_objective_c}. Note that if we are preprocessing
642 assembler, this variable will be @code{clk_c} but the function-like
643 macro @code{preprocessing_asm_p()} will return true, so you might want
644 to check for that first. If you need to check for strict ANSI, the
645 variable @code{flag_iso} can be used. The function-like macro
646 @code{preprocessing_trad_p()} can be used to check for traditional
650 @defmac TARGET_OS_CPP_BUILTINS ()
651 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
652 and is used for the target operating system instead.
655 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
656 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
657 and is used for the target object format. @file{elfos.h} uses this
658 macro to define @code{__ELF__}, so you probably do not need to define
662 @deftypevar {extern int} target_flags
663 This variable is declared in @file{options.h}, which is included before
664 any target-specific headers.
667 @hook TARGET_DEFAULT_TARGET_FLAGS
668 This variable specifies the initial value of @code{target_flags}.
669 Its default setting is 0.
672 @cindex optional hardware or system features
673 @cindex features, optional, in system conventions
675 @hook TARGET_HANDLE_OPTION
676 This hook is called whenever the user specifies one of the
677 target-specific options described by the @file{.opt} definition files
678 (@pxref{Options}). It has the opportunity to do some option-specific
679 processing and should return true if the option is valid. The default
680 definition does nothing but return true.
682 @var{decoded} specifies the option and its arguments. @var{opts} and
683 @var{opts_set} are the @code{gcc_options} structures to be used for
684 storing option state, and @var{loc} is the location at which the
685 option was passed (@code{UNKNOWN_LOCATION} except for options passed
689 @hook TARGET_HANDLE_C_OPTION
690 This target hook is called whenever the user specifies one of the
691 target-specific C language family options described by the @file{.opt}
692 definition files(@pxref{Options}). It has the opportunity to do some
693 option-specific processing and should return true if the option is
694 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
695 default definition does nothing but return false.
697 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
698 options. However, if processing an option requires routines that are
699 only available in the C (and related language) front ends, then you
700 should use @code{TARGET_HANDLE_C_OPTION} instead.
703 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
705 @hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE
707 @hook TARGET_OBJC_DECLARE_CLASS_DEFINITION
709 @hook TARGET_STRING_OBJECT_REF_TYPE_P
711 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
713 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
715 @defmac C_COMMON_OVERRIDE_OPTIONS
716 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
717 but is only used in the C
718 language frontends (C, Objective-C, C++, Objective-C++) and so can be
719 used to alter option flag variables which only exist in those
723 @hook TARGET_OPTION_OPTIMIZATION_TABLE
724 Some machines may desire to change what optimizations are performed for
725 various optimization levels. This variable, if defined, describes
726 options to enable at particular sets of optimization levels. These
727 options are processed once
728 just after the optimization level is determined and before the remainder
729 of the command options have been parsed, so may be overridden by other
730 options passed explicitly.
732 This processing is run once at program startup and when the optimization
733 options are changed via @code{#pragma GCC optimize} or by using the
734 @code{optimize} attribute.
737 @hook TARGET_OPTION_INIT_STRUCT
739 @defmac SWITCHABLE_TARGET
740 Some targets need to switch between substantially different subtargets
741 during compilation. For example, the MIPS target has one subtarget for
742 the traditional MIPS architecture and another for MIPS16. Source code
743 can switch between these two subarchitectures using the @code{mips16}
744 and @code{nomips16} attributes.
746 Such subtargets can differ in things like the set of available
747 registers, the set of available instructions, the costs of various
748 operations, and so on. GCC caches a lot of this type of information
749 in global variables, and recomputing them for each subtarget takes a
750 significant amount of time. The compiler therefore provides a facility
751 for maintaining several versions of the global variables and quickly
752 switching between them; see @file{target-globals.h} for details.
754 Define this macro to 1 if your target needs this facility. The default
758 @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
760 @node Per-Function Data
761 @section Defining data structures for per-function information.
762 @cindex per-function data
763 @cindex data structures
765 If the target needs to store information on a per-function basis, GCC
766 provides a macro and a couple of variables to allow this. Note, just
767 using statics to store the information is a bad idea, since GCC supports
768 nested functions, so you can be halfway through encoding one function
769 when another one comes along.
771 GCC defines a data structure called @code{struct function} which
772 contains all of the data specific to an individual function. This
773 structure contains a field called @code{machine} whose type is
774 @code{struct machine_function *}, which can be used by targets to point
775 to their own specific data.
777 If a target needs per-function specific data it should define the type
778 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
779 This macro should be used to initialize the function pointer
780 @code{init_machine_status}. This pointer is explained below.
782 One typical use of per-function, target specific data is to create an
783 RTX to hold the register containing the function's return address. This
784 RTX can then be used to implement the @code{__builtin_return_address}
785 function, for level 0.
787 Note---earlier implementations of GCC used a single data area to hold
788 all of the per-function information. Thus when processing of a nested
789 function began the old per-function data had to be pushed onto a
790 stack, and when the processing was finished, it had to be popped off the
791 stack. GCC used to provide function pointers called
792 @code{save_machine_status} and @code{restore_machine_status} to handle
793 the saving and restoring of the target specific information. Since the
794 single data area approach is no longer used, these pointers are no
797 @defmac INIT_EXPANDERS
798 Macro called to initialize any target specific information. This macro
799 is called once per function, before generation of any RTL has begun.
800 The intention of this macro is to allow the initialization of the
801 function pointer @code{init_machine_status}.
804 @deftypevar {void (*)(struct function *)} init_machine_status
805 If this function pointer is non-@code{NULL} it will be called once per
806 function, before function compilation starts, in order to allow the
807 target to perform any target specific initialization of the
808 @code{struct function} structure. It is intended that this would be
809 used to initialize the @code{machine} of that structure.
811 @code{struct machine_function} structures are expected to be freed by GC@.
812 Generally, any memory that they reference must be allocated by using
813 GC allocation, including the structure itself.
817 @section Storage Layout
818 @cindex storage layout
820 Note that the definitions of the macros in this table which are sizes or
821 alignments measured in bits do not need to be constant. They can be C
822 expressions that refer to static variables, such as the @code{target_flags}.
823 @xref{Run-time Target}.
825 @defmac BITS_BIG_ENDIAN
826 Define this macro to have the value 1 if the most significant bit in a
827 byte has the lowest number; otherwise define it to have the value zero.
828 This means that bit-field instructions count from the most significant
829 bit. If the machine has no bit-field instructions, then this must still
830 be defined, but it doesn't matter which value it is defined to. This
831 macro need not be a constant.
833 This macro does not affect the way structure fields are packed into
834 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
837 @defmac BYTES_BIG_ENDIAN
838 Define this macro to have the value 1 if the most significant byte in a
839 word has the lowest number. This macro need not be a constant.
842 @defmac WORDS_BIG_ENDIAN
843 Define this macro to have the value 1 if, in a multiword object, the
844 most significant word has the lowest number. This applies to both
845 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
846 order of words in memory is not the same as the order in registers. This
847 macro need not be a constant.
850 @defmac REG_WORDS_BIG_ENDIAN
851 On some machines, the order of words in a multiword object differs between
852 registers in memory. In such a situation, define this macro to describe
853 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
854 the order of words in memory.
857 @defmac FLOAT_WORDS_BIG_ENDIAN
858 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
859 @code{TFmode} floating point numbers are stored in memory with the word
860 containing the sign bit at the lowest address; otherwise define it to
861 have the value 0. This macro need not be a constant.
863 You need not define this macro if the ordering is the same as for
867 @defmac BITS_PER_WORD
868 Number of bits in a word. If you do not define this macro, the default
869 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
872 @defmac MAX_BITS_PER_WORD
873 Maximum number of bits in a word. If this is undefined, the default is
874 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
875 largest value that @code{BITS_PER_WORD} can have at run-time.
878 @defmac UNITS_PER_WORD
879 Number of storage units in a word; normally the size of a general-purpose
880 register, a power of two from 1 or 8.
883 @defmac MIN_UNITS_PER_WORD
884 Minimum number of units in a word. If this is undefined, the default is
885 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
886 smallest value that @code{UNITS_PER_WORD} can have at run-time.
890 Width of a pointer, in bits. You must specify a value no wider than the
891 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
892 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
893 a value the default is @code{BITS_PER_WORD}.
896 @defmac POINTERS_EXTEND_UNSIGNED
897 A C expression that determines how pointers should be extended from
898 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
899 greater than zero if pointers should be zero-extended, zero if they
900 should be sign-extended, and negative if some other sort of conversion
901 is needed. In the last case, the extension is done by the target's
902 @code{ptr_extend} instruction.
904 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
905 and @code{word_mode} are all the same width.
908 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
909 A macro to update @var{m} and @var{unsignedp} when an object whose type
910 is @var{type} and which has the specified mode and signedness is to be
911 stored in a register. This macro is only called when @var{type} is a
914 On most RISC machines, which only have operations that operate on a full
915 register, define this macro to set @var{m} to @code{word_mode} if
916 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
917 cases, only integer modes should be widened because wider-precision
918 floating-point operations are usually more expensive than their narrower
921 For most machines, the macro definition does not change @var{unsignedp}.
922 However, some machines, have instructions that preferentially handle
923 either signed or unsigned quantities of certain modes. For example, on
924 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
925 sign-extend the result to 64 bits. On such machines, set
926 @var{unsignedp} according to which kind of extension is more efficient.
928 Do not define this macro if it would never modify @var{m}.
931 @hook TARGET_C_EXCESS_PRECISION
933 @hook TARGET_PROMOTE_FUNCTION_MODE
935 @defmac PARM_BOUNDARY
936 Normal alignment required for function parameters on the stack, in
937 bits. All stack parameters receive at least this much alignment
938 regardless of data type. On most machines, this is the same as the
942 @defmac STACK_BOUNDARY
943 Define this macro to the minimum alignment enforced by hardware for the
944 stack pointer on this machine. The definition is a C expression for the
945 desired alignment (measured in bits). This value is used as a default
946 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
947 this should be the same as @code{PARM_BOUNDARY}.
950 @defmac PREFERRED_STACK_BOUNDARY
951 Define this macro if you wish to preserve a certain alignment for the
952 stack pointer, greater than what the hardware enforces. The definition
953 is a C expression for the desired alignment (measured in bits). This
954 macro must evaluate to a value equal to or larger than
955 @code{STACK_BOUNDARY}.
958 @defmac INCOMING_STACK_BOUNDARY
959 Define this macro if the incoming stack boundary may be different
960 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
961 to a value equal to or larger than @code{STACK_BOUNDARY}.
964 @defmac FUNCTION_BOUNDARY
965 Alignment required for a function entry point, in bits.
968 @defmac BIGGEST_ALIGNMENT
969 Biggest alignment that any data type can require on this machine, in
970 bits. Note that this is not the biggest alignment that is supported,
971 just the biggest alignment that, when violated, may cause a fault.
974 @hook TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
976 @defmac MALLOC_ABI_ALIGNMENT
977 Alignment, in bits, a C conformant malloc implementation has to
978 provide. If not defined, the default value is @code{BITS_PER_WORD}.
981 @defmac ATTRIBUTE_ALIGNED_VALUE
982 Alignment used by the @code{__attribute__ ((aligned))} construct. If
983 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
986 @defmac MINIMUM_ATOMIC_ALIGNMENT
987 If defined, the smallest alignment, in bits, that can be given to an
988 object that can be referenced in one operation, without disturbing any
989 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
990 on machines that don't have byte or half-word store operations.
993 @defmac BIGGEST_FIELD_ALIGNMENT
994 Biggest alignment that any structure or union field can require on this
995 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
996 structure and union fields only, unless the field alignment has been set
997 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1000 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1001 An expression for the alignment of a structure field @var{field} of
1002 type @var{type} if the alignment computed in the usual way (including
1003 applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1004 alignment) is @var{computed}. It overrides alignment only if the
1005 field alignment has not been set by the
1006 @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field}
1007 may be @code{NULL_TREE} in case we just query for the minimum alignment
1008 of a field of type @var{type} in structure context.
1011 @defmac MAX_STACK_ALIGNMENT
1012 Biggest stack alignment guaranteed by the backend. Use this macro
1013 to specify the maximum alignment of a variable on stack.
1015 If not defined, the default value is @code{STACK_BOUNDARY}.
1017 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1018 @c But the fix for PR 32893 indicates that we can only guarantee
1019 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1020 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1023 @defmac MAX_OFILE_ALIGNMENT
1024 Biggest alignment supported by the object file format of this machine.
1025 Use this macro to limit the alignment which can be specified using the
1026 @code{__attribute__ ((aligned (@var{n})))} construct for functions and
1027 objects with static storage duration. The alignment of automatic
1028 objects may exceed the object file format maximum up to the maximum
1029 supported by GCC. If not defined, the default value is
1030 @code{BIGGEST_ALIGNMENT}.
1032 On systems that use ELF, the default (in @file{config/elfos.h}) is
1033 the largest supported 32-bit ELF section alignment representable on
1034 a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}.
1035 On 32-bit ELF the largest supported section alignment in bits is
1036 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1039 @hook TARGET_STATIC_RTX_ALIGNMENT
1041 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1042 If defined, a C expression to compute the alignment for a variable in
1043 the static store. @var{type} is the data type, and @var{basic-align} is
1044 the alignment that the object would ordinarily have. The value of this
1045 macro is used instead of that alignment to align the object.
1047 If this macro is not defined, then @var{basic-align} is used.
1050 One use of this macro is to increase alignment of medium-size data to
1051 make it all fit in fewer cache lines. Another is to cause character
1052 arrays to be word-aligned so that @code{strcpy} calls that copy
1053 constants to character arrays can be done inline.
1056 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1057 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1058 some alignment increase, instead of optimization only purposes. E.g.@
1059 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1060 must be aligned to 16 byte boundaries.
1062 If this macro is not defined, then @var{basic-align} is used.
1065 @hook TARGET_CONSTANT_ALIGNMENT
1067 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1068 If defined, a C expression to compute the alignment for a variable in
1069 the local store. @var{type} is the data type, and @var{basic-align} is
1070 the alignment that the object would ordinarily have. The value of this
1071 macro is used instead of that alignment to align the object.
1073 If this macro is not defined, then @var{basic-align} is used.
1075 One use of this macro is to increase alignment of medium-size data to
1076 make it all fit in fewer cache lines.
1078 If the value of this macro has a type, it should be an unsigned type.
1081 @hook TARGET_VECTOR_ALIGNMENT
1083 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1084 If defined, a C expression to compute the alignment for stack slot.
1085 @var{type} is the data type, @var{mode} is the widest mode available,
1086 and @var{basic-align} is the alignment that the slot would ordinarily
1087 have. The value of this macro is used instead of that alignment to
1090 If this macro is not defined, then @var{basic-align} is used when
1091 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1094 This macro is to set alignment of stack slot to the maximum alignment
1095 of all possible modes which the slot may have.
1097 If the value of this macro has a type, it should be an unsigned type.
1100 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1101 If defined, a C expression to compute the alignment for a local
1102 variable @var{decl}.
1104 If this macro is not defined, then
1105 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1108 One use of this macro is to increase alignment of medium-size data to
1109 make it all fit in fewer cache lines.
1111 If the value of this macro has a type, it should be an unsigned type.
1114 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1115 If defined, a C expression to compute the minimum required alignment
1116 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1117 @var{mode}, assuming normal alignment @var{align}.
1119 If this macro is not defined, then @var{align} will be used.
1122 @defmac EMPTY_FIELD_BOUNDARY
1123 Alignment in bits to be given to a structure bit-field that follows an
1124 empty field such as @code{int : 0;}.
1126 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1129 @defmac STRUCTURE_SIZE_BOUNDARY
1130 Number of bits which any structure or union's size must be a multiple of.
1131 Each structure or union's size is rounded up to a multiple of this.
1133 If you do not define this macro, the default is the same as
1134 @code{BITS_PER_UNIT}.
1137 @defmac STRICT_ALIGNMENT
1138 Define this macro to be the value 1 if instructions will fail to work
1139 if given data not on the nominal alignment. If instructions will merely
1140 go slower in that case, define this macro as 0.
1143 @defmac PCC_BITFIELD_TYPE_MATTERS
1144 Define this if you wish to imitate the way many other C compilers handle
1145 alignment of bit-fields and the structures that contain them.
1147 The behavior is that the type written for a named bit-field (@code{int},
1148 @code{short}, or other integer type) imposes an alignment for the entire
1149 structure, as if the structure really did contain an ordinary field of
1150 that type. In addition, the bit-field is placed within the structure so
1151 that it would fit within such a field, not crossing a boundary for it.
1153 Thus, on most machines, a named bit-field whose type is written as
1154 @code{int} would not cross a four-byte boundary, and would force
1155 four-byte alignment for the whole structure. (The alignment used may
1156 not be four bytes; it is controlled by the other alignment parameters.)
1158 An unnamed bit-field will not affect the alignment of the containing
1161 If the macro is defined, its definition should be a C expression;
1162 a nonzero value for the expression enables this behavior.
1164 Note that if this macro is not defined, or its value is zero, some
1165 bit-fields may cross more than one alignment boundary. The compiler can
1166 support such references if there are @samp{insv}, @samp{extv}, and
1167 @samp{extzv} insns that can directly reference memory.
1169 The other known way of making bit-fields work is to define
1170 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1171 Then every structure can be accessed with fullwords.
1173 Unless the machine has bit-field instructions or you define
1174 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1175 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1177 If your aim is to make GCC use the same conventions for laying out
1178 bit-fields as are used by another compiler, here is how to investigate
1179 what the other compiler does. Compile and run this program:
1198 printf ("Size of foo1 is %d\n",
1199 sizeof (struct foo1));
1200 printf ("Size of foo2 is %d\n",
1201 sizeof (struct foo2));
1206 If this prints 2 and 5, then the compiler's behavior is what you would
1207 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1210 @defmac BITFIELD_NBYTES_LIMITED
1211 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1212 to aligning a bit-field within the structure.
1215 @hook TARGET_ALIGN_ANON_BITFIELD
1217 @hook TARGET_NARROW_VOLATILE_BITFIELD
1219 @hook TARGET_MEMBER_TYPE_FORCES_BLK
1221 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1222 Define this macro as an expression for the alignment of a type (given
1223 by @var{type} as a tree node) if the alignment computed in the usual
1224 way is @var{computed} and the alignment explicitly specified was
1227 The default is to use @var{specified} if it is larger; otherwise, use
1228 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1231 @defmac MAX_FIXED_MODE_SIZE
1232 An integer expression for the size in bits of the largest integer
1233 machine mode that should actually be used. All integer machine modes of
1234 this size or smaller can be used for structures and unions with the
1235 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1236 (DImode)} is assumed.
1239 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1240 If defined, an expression of type @code{machine_mode} that
1241 specifies the mode of the save area operand of a
1242 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1243 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1244 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1245 having its mode specified.
1247 You need not define this macro if it always returns @code{Pmode}. You
1248 would most commonly define this macro if the
1249 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1253 @defmac STACK_SIZE_MODE
1254 If defined, an expression of type @code{machine_mode} that
1255 specifies the mode of the size increment operand of an
1256 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1258 You need not define this macro if it always returns @code{word_mode}.
1259 You would most commonly define this macro if the @code{allocate_stack}
1260 pattern needs to support both a 32- and a 64-bit mode.
1263 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1265 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1267 @hook TARGET_UNWIND_WORD_MODE
1269 @hook TARGET_MS_BITFIELD_LAYOUT_P
1271 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1273 @hook TARGET_FIXED_POINT_SUPPORTED_P
1275 @hook TARGET_EXPAND_TO_RTL_HOOK
1277 @hook TARGET_INSTANTIATE_DECLS
1279 @hook TARGET_MANGLE_TYPE
1282 @section Layout of Source Language Data Types
1284 These macros define the sizes and other characteristics of the standard
1285 basic data types used in programs being compiled. Unlike the macros in
1286 the previous section, these apply to specific features of C and related
1287 languages, rather than to fundamental aspects of storage layout.
1289 @defmac INT_TYPE_SIZE
1290 A C expression for the size in bits of the type @code{int} on the
1291 target machine. If you don't define this, the default is one word.
1294 @defmac SHORT_TYPE_SIZE
1295 A C expression for the size in bits of the type @code{short} on the
1296 target machine. If you don't define this, the default is half a word.
1297 (If this would be less than one storage unit, it is rounded up to one
1301 @defmac LONG_TYPE_SIZE
1302 A C expression for the size in bits of the type @code{long} on the
1303 target machine. If you don't define this, the default is one word.
1306 @defmac ADA_LONG_TYPE_SIZE
1307 On some machines, the size used for the Ada equivalent of the type
1308 @code{long} by a native Ada compiler differs from that used by C@. In
1309 that situation, define this macro to be a C expression to be used for
1310 the size of that type. If you don't define this, the default is the
1311 value of @code{LONG_TYPE_SIZE}.
1314 @defmac LONG_LONG_TYPE_SIZE
1315 A C expression for the size in bits of the type @code{long long} on the
1316 target machine. If you don't define this, the default is two
1317 words. If you want to support GNU Ada on your machine, the value of this
1318 macro must be at least 64.
1321 @defmac CHAR_TYPE_SIZE
1322 A C expression for the size in bits of the type @code{char} on the
1323 target machine. If you don't define this, the default is
1324 @code{BITS_PER_UNIT}.
1327 @defmac BOOL_TYPE_SIZE
1328 A C expression for the size in bits of the C++ type @code{bool} and
1329 C99 type @code{_Bool} on the target machine. If you don't define
1330 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1333 @defmac FLOAT_TYPE_SIZE
1334 A C expression for the size in bits of the type @code{float} on the
1335 target machine. If you don't define this, the default is one word.
1338 @defmac DOUBLE_TYPE_SIZE
1339 A C expression for the size in bits of the type @code{double} on the
1340 target machine. If you don't define this, the default is two
1344 @defmac LONG_DOUBLE_TYPE_SIZE
1345 A C expression for the size in bits of the type @code{long double} on
1346 the target machine. If you don't define this, the default is two
1350 @defmac SHORT_FRACT_TYPE_SIZE
1351 A C expression for the size in bits of the type @code{short _Fract} on
1352 the target machine. If you don't define this, the default is
1353 @code{BITS_PER_UNIT}.
1356 @defmac FRACT_TYPE_SIZE
1357 A C expression for the size in bits of the type @code{_Fract} on
1358 the target machine. If you don't define this, the default is
1359 @code{BITS_PER_UNIT * 2}.
1362 @defmac LONG_FRACT_TYPE_SIZE
1363 A C expression for the size in bits of the type @code{long _Fract} on
1364 the target machine. If you don't define this, the default is
1365 @code{BITS_PER_UNIT * 4}.
1368 @defmac LONG_LONG_FRACT_TYPE_SIZE
1369 A C expression for the size in bits of the type @code{long long _Fract} on
1370 the target machine. If you don't define this, the default is
1371 @code{BITS_PER_UNIT * 8}.
1374 @defmac SHORT_ACCUM_TYPE_SIZE
1375 A C expression for the size in bits of the type @code{short _Accum} on
1376 the target machine. If you don't define this, the default is
1377 @code{BITS_PER_UNIT * 2}.
1380 @defmac ACCUM_TYPE_SIZE
1381 A C expression for the size in bits of the type @code{_Accum} on
1382 the target machine. If you don't define this, the default is
1383 @code{BITS_PER_UNIT * 4}.
1386 @defmac LONG_ACCUM_TYPE_SIZE
1387 A C expression for the size in bits of the type @code{long _Accum} on
1388 the target machine. If you don't define this, the default is
1389 @code{BITS_PER_UNIT * 8}.
1392 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1393 A C expression for the size in bits of the type @code{long long _Accum} on
1394 the target machine. If you don't define this, the default is
1395 @code{BITS_PER_UNIT * 16}.
1398 @defmac LIBGCC2_GNU_PREFIX
1399 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1400 hook and should be defined if that hook is overriden to be true. It
1401 causes function names in libgcc to be changed to use a @code{__gnu_}
1402 prefix for their name rather than the default @code{__}. A port which
1403 uses this macro should also arrange to use @file{t-gnu-prefix} in
1404 the libgcc @file{config.host}.
1407 @defmac WIDEST_HARDWARE_FP_SIZE
1408 A C expression for the size in bits of the widest floating-point format
1409 supported by the hardware. If you define this macro, you must specify a
1410 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1411 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1415 @defmac DEFAULT_SIGNED_CHAR
1416 An expression whose value is 1 or 0, according to whether the type
1417 @code{char} should be signed or unsigned by default. The user can
1418 always override this default with the options @option{-fsigned-char}
1419 and @option{-funsigned-char}.
1422 @hook TARGET_DEFAULT_SHORT_ENUMS
1425 A C expression for a string describing the name of the data type to use
1426 for size values. The typedef name @code{size_t} is defined using the
1427 contents of the string.
1429 The string can contain more than one keyword. If so, separate them with
1430 spaces, and write first any length keyword, then @code{unsigned} if
1431 appropriate, and finally @code{int}. The string must exactly match one
1432 of the data type names defined in the function
1433 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1434 You may not omit @code{int} or change the order---that would cause the
1435 compiler to crash on startup.
1437 If you don't define this macro, the default is @code{"long unsigned
1442 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1443 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1444 dealing with size. This macro is a C expression for a string describing
1445 the name of the data type from which the precision of @code{sizetype}
1448 The string has the same restrictions as @code{SIZE_TYPE} string.
1450 If you don't define this macro, the default is @code{SIZE_TYPE}.
1453 @defmac PTRDIFF_TYPE
1454 A C expression for a string describing the name of the data type to use
1455 for the result of subtracting two pointers. The typedef name
1456 @code{ptrdiff_t} is defined using the contents of the string. See
1457 @code{SIZE_TYPE} above for more information.
1459 If you don't define this macro, the default is @code{"long int"}.
1463 A C expression for a string describing the name of the data type to use
1464 for wide characters. The typedef name @code{wchar_t} is defined using
1465 the contents of the string. See @code{SIZE_TYPE} above for more
1468 If you don't define this macro, the default is @code{"int"}.
1471 @defmac WCHAR_TYPE_SIZE
1472 A C expression for the size in bits of the data type for wide
1473 characters. This is used in @code{cpp}, which cannot make use of
1478 A C expression for a string describing the name of the data type to
1479 use for wide characters passed to @code{printf} and returned from
1480 @code{getwc}. The typedef name @code{wint_t} is defined using the
1481 contents of the string. See @code{SIZE_TYPE} above for more
1484 If you don't define this macro, the default is @code{"unsigned int"}.
1488 A C expression for a string describing the name of the data type that
1489 can represent any value of any standard or extended signed integer type.
1490 The typedef name @code{intmax_t} is defined using the contents of the
1491 string. See @code{SIZE_TYPE} above for more information.
1493 If you don't define this macro, the default is the first of
1494 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1495 much precision as @code{long long int}.
1498 @defmac UINTMAX_TYPE
1499 A C expression for a string describing the name of the data type that
1500 can represent any value of any standard or extended unsigned integer
1501 type. The typedef name @code{uintmax_t} is defined using the contents
1502 of the string. See @code{SIZE_TYPE} above for more information.
1504 If you don't define this macro, the default is the first of
1505 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1506 unsigned int"} that has as much precision as @code{long long unsigned
1510 @defmac SIG_ATOMIC_TYPE
1516 @defmacx UINT16_TYPE
1517 @defmacx UINT32_TYPE
1518 @defmacx UINT64_TYPE
1519 @defmacx INT_LEAST8_TYPE
1520 @defmacx INT_LEAST16_TYPE
1521 @defmacx INT_LEAST32_TYPE
1522 @defmacx INT_LEAST64_TYPE
1523 @defmacx UINT_LEAST8_TYPE
1524 @defmacx UINT_LEAST16_TYPE
1525 @defmacx UINT_LEAST32_TYPE
1526 @defmacx UINT_LEAST64_TYPE
1527 @defmacx INT_FAST8_TYPE
1528 @defmacx INT_FAST16_TYPE
1529 @defmacx INT_FAST32_TYPE
1530 @defmacx INT_FAST64_TYPE
1531 @defmacx UINT_FAST8_TYPE
1532 @defmacx UINT_FAST16_TYPE
1533 @defmacx UINT_FAST32_TYPE
1534 @defmacx UINT_FAST64_TYPE
1535 @defmacx INTPTR_TYPE
1536 @defmacx UINTPTR_TYPE
1537 C expressions for the standard types @code{sig_atomic_t},
1538 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1539 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1540 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1541 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1542 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1543 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1544 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1545 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1546 @code{SIZE_TYPE} above for more information.
1548 If any of these macros evaluates to a null pointer, the corresponding
1549 type is not supported; if GCC is configured to provide
1550 @code{<stdint.h>} in such a case, the header provided may not conform
1551 to C99, depending on the type in question. The defaults for all of
1552 these macros are null pointers.
1555 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1556 The C++ compiler represents a pointer-to-member-function with a struct
1563 ptrdiff_t vtable_index;
1570 The C++ compiler must use one bit to indicate whether the function that
1571 will be called through a pointer-to-member-function is virtual.
1572 Normally, we assume that the low-order bit of a function pointer must
1573 always be zero. Then, by ensuring that the vtable_index is odd, we can
1574 distinguish which variant of the union is in use. But, on some
1575 platforms function pointers can be odd, and so this doesn't work. In
1576 that case, we use the low-order bit of the @code{delta} field, and shift
1577 the remainder of the @code{delta} field to the left.
1579 GCC will automatically make the right selection about where to store
1580 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1581 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1582 set such that functions always start at even addresses, but the lowest
1583 bit of pointers to functions indicate whether the function at that
1584 address is in ARM or Thumb mode. If this is the case of your
1585 architecture, you should define this macro to
1586 @code{ptrmemfunc_vbit_in_delta}.
1588 In general, you should not have to define this macro. On architectures
1589 in which function addresses are always even, according to
1590 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1591 @code{ptrmemfunc_vbit_in_pfn}.
1594 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1595 Normally, the C++ compiler uses function pointers in vtables. This
1596 macro allows the target to change to use ``function descriptors''
1597 instead. Function descriptors are found on targets for whom a
1598 function pointer is actually a small data structure. Normally the
1599 data structure consists of the actual code address plus a data
1600 pointer to which the function's data is relative.
1602 If vtables are used, the value of this macro should be the number
1603 of words that the function descriptor occupies.
1606 @defmac TARGET_VTABLE_ENTRY_ALIGN
1607 By default, the vtable entries are void pointers, the so the alignment
1608 is the same as pointer alignment. The value of this macro specifies
1609 the alignment of the vtable entry in bits. It should be defined only
1610 when special alignment is necessary. */
1613 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1614 There are a few non-descriptor entries in the vtable at offsets below
1615 zero. If these entries must be padded (say, to preserve the alignment
1616 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1617 of words in each data entry.
1621 @section Register Usage
1622 @cindex register usage
1624 This section explains how to describe what registers the target machine
1625 has, and how (in general) they can be used.
1627 The description of which registers a specific instruction can use is
1628 done with register classes; see @ref{Register Classes}. For information
1629 on using registers to access a stack frame, see @ref{Frame Registers}.
1630 For passing values in registers, see @ref{Register Arguments}.
1631 For returning values in registers, see @ref{Scalar Return}.
1634 * Register Basics:: Number and kinds of registers.
1635 * Allocation Order:: Order in which registers are allocated.
1636 * Values in Registers:: What kinds of values each reg can hold.
1637 * Leaf Functions:: Renumbering registers for leaf functions.
1638 * Stack Registers:: Handling a register stack such as 80387.
1641 @node Register Basics
1642 @subsection Basic Characteristics of Registers
1644 @c prevent bad page break with this line
1645 Registers have various characteristics.
1647 @defmac FIRST_PSEUDO_REGISTER
1648 Number of hardware registers known to the compiler. They receive
1649 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1650 pseudo register's number really is assigned the number
1651 @code{FIRST_PSEUDO_REGISTER}.
1654 @defmac FIXED_REGISTERS
1655 @cindex fixed register
1656 An initializer that says which registers are used for fixed purposes
1657 all throughout the compiled code and are therefore not available for
1658 general allocation. These would include the stack pointer, the frame
1659 pointer (except on machines where that can be used as a general
1660 register when no frame pointer is needed), the program counter on
1661 machines where that is considered one of the addressable registers,
1662 and any other numbered register with a standard use.
1664 This information is expressed as a sequence of numbers, separated by
1665 commas and surrounded by braces. The @var{n}th number is 1 if
1666 register @var{n} is fixed, 0 otherwise.
1668 The table initialized from this macro, and the table initialized by
1669 the following one, may be overridden at run time either automatically,
1670 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1671 the user with the command options @option{-ffixed-@var{reg}},
1672 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1675 @defmac CALL_USED_REGISTERS
1676 @cindex call-used register
1677 @cindex call-clobbered register
1678 @cindex call-saved register
1679 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1680 clobbered (in general) by function calls as well as for fixed
1681 registers. This macro therefore identifies the registers that are not
1682 available for general allocation of values that must live across
1685 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1686 automatically saves it on function entry and restores it on function
1687 exit, if the register is used within the function.
1689 Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1690 must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1693 @defmac CALL_REALLY_USED_REGISTERS
1694 @cindex call-used register
1695 @cindex call-clobbered register
1696 @cindex call-saved register
1697 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1698 that the entire set of @code{FIXED_REGISTERS} be included.
1699 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1701 Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1702 must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1705 @cindex call-used register
1706 @cindex call-clobbered register
1707 @cindex call-saved register
1708 @hook TARGET_FNTYPE_ABI
1710 @hook TARGET_INSN_CALLEE_ABI
1712 @cindex call-used register
1713 @cindex call-clobbered register
1714 @cindex call-saved register
1715 @hook TARGET_HARD_REGNO_CALL_PART_CLOBBERED
1717 @hook TARGET_GET_MULTILIB_ABI_NAME
1720 @findex call_used_regs
1723 @findex reg_class_contents
1724 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1726 @defmac INCOMING_REGNO (@var{out})
1727 Define this macro if the target machine has register windows. This C
1728 expression returns the register number as seen by the called function
1729 corresponding to the register number @var{out} as seen by the calling
1730 function. Return @var{out} if register number @var{out} is not an
1734 @defmac OUTGOING_REGNO (@var{in})
1735 Define this macro if the target machine has register windows. This C
1736 expression returns the register number as seen by the calling function
1737 corresponding to the register number @var{in} as seen by the called
1738 function. Return @var{in} if register number @var{in} is not an inbound
1742 @defmac LOCAL_REGNO (@var{regno})
1743 Define this macro if the target machine has register windows. This C
1744 expression returns true if the register is call-saved but is in the
1745 register window. Unlike most call-saved registers, such registers
1746 need not be explicitly restored on function exit or during non-local
1751 If the program counter has a register number, define this as that
1752 register number. Otherwise, do not define it.
1755 @node Allocation Order
1756 @subsection Order of Allocation of Registers
1757 @cindex order of register allocation
1758 @cindex register allocation order
1760 @c prevent bad page break with this line
1761 Registers are allocated in order.
1763 @defmac REG_ALLOC_ORDER
1764 If defined, an initializer for a vector of integers, containing the
1765 numbers of hard registers in the order in which GCC should prefer
1766 to use them (from most preferred to least).
1768 If this macro is not defined, registers are used lowest numbered first
1769 (all else being equal).
1771 One use of this macro is on machines where the highest numbered
1772 registers must always be saved and the save-multiple-registers
1773 instruction supports only sequences of consecutive registers. On such
1774 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1775 the highest numbered allocable register first.
1778 @defmac ADJUST_REG_ALLOC_ORDER
1779 A C statement (sans semicolon) to choose the order in which to allocate
1780 hard registers for pseudo-registers local to a basic block.
1782 Store the desired register order in the array @code{reg_alloc_order}.
1783 Element 0 should be the register to allocate first; element 1, the next
1784 register; and so on.
1786 The macro body should not assume anything about the contents of
1787 @code{reg_alloc_order} before execution of the macro.
1789 On most machines, it is not necessary to define this macro.
1792 @defmac HONOR_REG_ALLOC_ORDER
1793 Normally, IRA tries to estimate the costs for saving a register in the
1794 prologue and restoring it in the epilogue. This discourages it from
1795 using call-saved registers. If a machine wants to ensure that IRA
1796 allocates registers in the order given by REG_ALLOC_ORDER even if some
1797 call-saved registers appear earlier than call-used ones, then define this
1798 macro as a C expression to nonzero. Default is 0.
1801 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1802 In some case register allocation order is not enough for the
1803 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1804 If this macro is defined, it should return a floating point value
1805 based on @var{regno}. The cost of using @var{regno} for a pseudo will
1806 be increased by approximately the pseudo's usage frequency times the
1807 value returned by this macro. Not defining this macro is equivalent
1808 to having it always return @code{0.0}.
1810 On most machines, it is not necessary to define this macro.
1813 @node Values in Registers
1814 @subsection How Values Fit in Registers
1816 This section discusses the macros that describe which kinds of values
1817 (specifically, which machine modes) each register can hold, and how many
1818 consecutive registers are needed for a given mode.
1820 @hook TARGET_HARD_REGNO_NREGS
1822 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1823 A C expression that is nonzero if a value of mode @var{mode}, stored
1824 in memory, ends with padding that causes it to take up more space than
1825 in registers starting at register number @var{regno} (as determined by
1826 multiplying GCC's notion of the size of the register when containing
1827 this mode by the number of registers returned by
1828 @code{TARGET_HARD_REGNO_NREGS}). By default this is zero.
1830 For example, if a floating-point value is stored in three 32-bit
1831 registers but takes up 128 bits in memory, then this would be
1834 This macros only needs to be defined if there are cases where
1835 @code{subreg_get_info}
1836 would otherwise wrongly determine that a @code{subreg} can be
1837 represented by an offset to the register number, when in fact such a
1838 @code{subreg} would contain some of the padding not stored in
1839 registers and so not be representable.
1842 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1843 For values of @var{regno} and @var{mode} for which
1844 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1845 returning the greater number of registers required to hold the value
1846 including any padding. In the example above, the value would be four.
1849 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1850 Define this macro if the natural size of registers that hold values
1851 of mode @var{mode} is not the word size. It is a C expression that
1852 should give the natural size in bytes for the specified mode. It is
1853 used by the register allocator to try to optimize its results. This
1854 happens for example on SPARC 64-bit where the natural size of
1855 floating-point registers is still 32-bit.
1858 @hook TARGET_HARD_REGNO_MODE_OK
1860 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1861 A C expression that is nonzero if it is OK to rename a hard register
1862 @var{from} to another hard register @var{to}.
1864 One common use of this macro is to prevent renaming of a register to
1865 another register that is not saved by a prologue in an interrupt
1868 The default is always nonzero.
1871 @hook TARGET_MODES_TIEABLE_P
1873 @hook TARGET_HARD_REGNO_SCRATCH_OK
1875 @defmac AVOID_CCMODE_COPIES
1876 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1877 registers. You should only define this macro if support for copying to/from
1878 @code{CCmode} is incomplete.
1881 @node Leaf Functions
1882 @subsection Handling Leaf Functions
1884 @cindex leaf functions
1885 @cindex functions, leaf
1886 On some machines, a leaf function (i.e., one which makes no calls) can run
1887 more efficiently if it does not make its own register window. Often this
1888 means it is required to receive its arguments in the registers where they
1889 are passed by the caller, instead of the registers where they would
1892 The special treatment for leaf functions generally applies only when
1893 other conditions are met; for example, often they may use only those
1894 registers for its own variables and temporaries. We use the term ``leaf
1895 function'' to mean a function that is suitable for this special
1896 handling, so that functions with no calls are not necessarily ``leaf
1899 GCC assigns register numbers before it knows whether the function is
1900 suitable for leaf function treatment. So it needs to renumber the
1901 registers in order to output a leaf function. The following macros
1904 @defmac LEAF_REGISTERS
1905 Name of a char vector, indexed by hard register number, which
1906 contains 1 for a register that is allowable in a candidate for leaf
1909 If leaf function treatment involves renumbering the registers, then the
1910 registers marked here should be the ones before renumbering---those that
1911 GCC would ordinarily allocate. The registers which will actually be
1912 used in the assembler code, after renumbering, should not be marked with 1
1915 Define this macro only if the target machine offers a way to optimize
1916 the treatment of leaf functions.
1919 @defmac LEAF_REG_REMAP (@var{regno})
1920 A C expression whose value is the register number to which @var{regno}
1921 should be renumbered, when a function is treated as a leaf function.
1923 If @var{regno} is a register number which should not appear in a leaf
1924 function before renumbering, then the expression should yield @minus{}1, which
1925 will cause the compiler to abort.
1927 Define this macro only if the target machine offers a way to optimize the
1928 treatment of leaf functions, and registers need to be renumbered to do
1932 @findex current_function_is_leaf
1933 @findex current_function_uses_only_leaf_regs
1934 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
1935 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
1936 specially. They can test the C variable @code{current_function_is_leaf}
1937 which is nonzero for leaf functions. @code{current_function_is_leaf} is
1938 set prior to local register allocation and is valid for the remaining
1939 compiler passes. They can also test the C variable
1940 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
1941 functions which only use leaf registers.
1942 @code{current_function_uses_only_leaf_regs} is valid after all passes
1943 that modify the instructions have been run and is only useful if
1944 @code{LEAF_REGISTERS} is defined.
1945 @c changed this to fix overfull. ALSO: why the "it" at the beginning
1946 @c of the next paragraph?! --mew 2feb93
1948 @node Stack Registers
1949 @subsection Registers That Form a Stack
1951 There are special features to handle computers where some of the
1952 ``registers'' form a stack. Stack registers are normally written by
1953 pushing onto the stack, and are numbered relative to the top of the
1956 Currently, GCC can only handle one group of stack-like registers, and
1957 they must be consecutively numbered. Furthermore, the existing
1958 support for stack-like registers is specific to the 80387 floating
1959 point coprocessor. If you have a new architecture that uses
1960 stack-like registers, you will need to do substantial work on
1961 @file{reg-stack.c} and write your machine description to cooperate
1962 with it, as well as defining these macros.
1965 Define this if the machine has any stack-like registers.
1968 @defmac STACK_REG_COVER_CLASS
1969 This is a cover class containing the stack registers. Define this if
1970 the machine has any stack-like registers.
1973 @defmac FIRST_STACK_REG
1974 The number of the first stack-like register. This one is the top
1978 @defmac LAST_STACK_REG
1979 The number of the last stack-like register. This one is the bottom of
1983 @node Register Classes
1984 @section Register Classes
1985 @cindex register class definitions
1986 @cindex class definitions, register
1988 On many machines, the numbered registers are not all equivalent.
1989 For example, certain registers may not be allowed for indexed addressing;
1990 certain registers may not be allowed in some instructions. These machine
1991 restrictions are described to the compiler using @dfn{register classes}.
1993 You define a number of register classes, giving each one a name and saying
1994 which of the registers belong to it. Then you can specify register classes
1995 that are allowed as operands to particular instruction patterns.
1999 In general, each register will belong to several classes. In fact, one
2000 class must be named @code{ALL_REGS} and contain all the registers. Another
2001 class must be named @code{NO_REGS} and contain no registers. Often the
2002 union of two classes will be another class; however, this is not required.
2004 @findex GENERAL_REGS
2005 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2006 terribly special about the name, but the operand constraint letters
2007 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2008 the same as @code{ALL_REGS}, just define it as a macro which expands
2011 Order the classes so that if class @var{x} is contained in class @var{y}
2012 then @var{x} has a lower class number than @var{y}.
2014 The way classes other than @code{GENERAL_REGS} are specified in operand
2015 constraints is through machine-dependent operand constraint letters.
2016 You can define such letters to correspond to various classes, then use
2017 them in operand constraints.
2019 You must define the narrowest register classes for allocatable
2020 registers, so that each class either has no subclasses, or that for
2021 some mode, the move cost between registers within the class is
2022 cheaper than moving a register in the class to or from memory
2025 You should define a class for the union of two classes whenever some
2026 instruction allows both classes. For example, if an instruction allows
2027 either a floating point (coprocessor) register or a general register for a
2028 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2029 which includes both of them. Otherwise you will get suboptimal code,
2030 or even internal compiler errors when reload cannot find a register in the
2031 class computed via @code{reg_class_subunion}.
2033 You must also specify certain redundant information about the register
2034 classes: for each class, which classes contain it and which ones are
2035 contained in it; for each pair of classes, the largest class contained
2038 When a value occupying several consecutive registers is expected in a
2039 certain class, all the registers used must belong to that class.
2040 Therefore, register classes cannot be used to enforce a requirement for
2041 a register pair to start with an even-numbered register. The way to
2042 specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2044 Register classes used for input-operands of bitwise-and or shift
2045 instructions have a special requirement: each such class must have, for
2046 each fixed-point machine mode, a subclass whose registers can transfer that
2047 mode to or from memory. For example, on some machines, the operations for
2048 single-byte values (@code{QImode}) are limited to certain registers. When
2049 this is so, each register class that is used in a bitwise-and or shift
2050 instruction must have a subclass consisting of registers from which
2051 single-byte values can be loaded or stored. This is so that
2052 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2054 @deftp {Data type} {enum reg_class}
2055 An enumerated type that must be defined with all the register class names
2056 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2057 must be the last register class, followed by one more enumerated value,
2058 @code{LIM_REG_CLASSES}, which is not a register class but rather
2059 tells how many classes there are.
2061 Each register class has a number, which is the value of casting
2062 the class name to type @code{int}. The number serves as an index
2063 in many of the tables described below.
2066 @defmac N_REG_CLASSES
2067 The number of distinct register classes, defined as follows:
2070 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2074 @defmac REG_CLASS_NAMES
2075 An initializer containing the names of the register classes as C string
2076 constants. These names are used in writing some of the debugging dumps.
2079 @defmac REG_CLASS_CONTENTS
2080 An initializer containing the contents of the register classes, as integers
2081 which are bit masks. The @var{n}th integer specifies the contents of class
2082 @var{n}. The way the integer @var{mask} is interpreted is that
2083 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2085 When the machine has more than 32 registers, an integer does not suffice.
2086 Then the integers are replaced by sub-initializers, braced groupings containing
2087 several integers. Each sub-initializer must be suitable as an initializer
2088 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2089 In this situation, the first integer in each sub-initializer corresponds to
2090 registers 0 through 31, the second integer to registers 32 through 63, and
2094 @defmac REGNO_REG_CLASS (@var{regno})
2095 A C expression whose value is a register class containing hard register
2096 @var{regno}. In general there is more than one such class; choose a class
2097 which is @dfn{minimal}, meaning that no smaller class also contains the
2101 @defmac BASE_REG_CLASS
2102 A macro whose definition is the name of the class to which a valid
2103 base register must belong. A base register is one used in an address
2104 which is the register value plus a displacement.
2107 @defmac MODE_BASE_REG_CLASS (@var{mode})
2108 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2109 the selection of a base register in a mode dependent manner. If
2110 @var{mode} is VOIDmode then it should return the same value as
2111 @code{BASE_REG_CLASS}.
2114 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2115 A C expression whose value is the register class to which a valid
2116 base register must belong in order to be used in a base plus index
2117 register address. You should define this macro if base plus index
2118 addresses have different requirements than other base register uses.
2121 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2122 A C expression whose value is the register class to which a valid
2123 base register for a memory reference in mode @var{mode} to address
2124 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2125 define the context in which the base register occurs. @var{outer_code} is
2126 the code of the immediately enclosing expression (@code{MEM} for the top level
2127 of an address, @code{ADDRESS} for something that occurs in an
2128 @code{address_operand}). @var{index_code} is the code of the corresponding
2129 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2132 @defmac INDEX_REG_CLASS
2133 A macro whose definition is the name of the class to which a valid
2134 index register must belong. An index register is one used in an
2135 address where its value is either multiplied by a scale factor or
2136 added to another register (as well as added to a displacement).
2139 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2140 A C expression which is nonzero if register number @var{num} is
2141 suitable for use as a base register in operand addresses.
2144 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2145 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2146 that expression may examine the mode of the memory reference in
2147 @var{mode}. You should define this macro if the mode of the memory
2148 reference affects whether a register may be used as a base register. If
2149 you define this macro, the compiler will use it instead of
2150 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2151 addresses that appear outside a @code{MEM}, i.e., as an
2152 @code{address_operand}.
2155 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2156 A C expression which is nonzero if register number @var{num} is suitable for
2157 use as a base register in base plus index operand addresses, accessing
2158 memory in mode @var{mode}. It may be either a suitable hard register or a
2159 pseudo register that has been allocated such a hard register. You should
2160 define this macro if base plus index addresses have different requirements
2161 than other base register uses.
2163 Use of this macro is deprecated; please use the more general
2164 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2167 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2168 A C expression which is nonzero if register number @var{num} is
2169 suitable for use as a base register in operand addresses, accessing
2170 memory in mode @var{mode} in address space @var{address_space}.
2171 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2172 that that expression may examine the context in which the register
2173 appears in the memory reference. @var{outer_code} is the code of the
2174 immediately enclosing expression (@code{MEM} if at the top level of the
2175 address, @code{ADDRESS} for something that occurs in an
2176 @code{address_operand}). @var{index_code} is the code of the
2177 corresponding index expression if @var{outer_code} is @code{PLUS};
2178 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2179 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2182 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2183 A C expression which is nonzero if register number @var{num} is
2184 suitable for use as an index register in operand addresses. It may be
2185 either a suitable hard register or a pseudo register that has been
2186 allocated such a hard register.
2188 The difference between an index register and a base register is that
2189 the index register may be scaled. If an address involves the sum of
2190 two registers, neither one of them scaled, then either one may be
2191 labeled the ``base'' and the other the ``index''; but whichever
2192 labeling is used must fit the machine's constraints of which registers
2193 may serve in each capacity. The compiler will try both labelings,
2194 looking for one that is valid, and will reload one or both registers
2195 only if neither labeling works.
2198 @hook TARGET_PREFERRED_RENAME_CLASS
2200 @hook TARGET_PREFERRED_RELOAD_CLASS
2202 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2203 A C expression that places additional restrictions on the register class
2204 to use when it is necessary to copy value @var{x} into a register in class
2205 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2206 another, smaller class. On many machines, the following definition is
2210 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2213 Sometimes returning a more restrictive class makes better code. For
2214 example, on the 68000, when @var{x} is an integer constant that is in range
2215 for a @samp{moveq} instruction, the value of this macro is always
2216 @code{DATA_REGS} as long as @var{class} includes the data registers.
2217 Requiring a data register guarantees that a @samp{moveq} will be used.
2219 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2220 @var{class} is if @var{x} is a legitimate constant which cannot be
2221 loaded into some register class. By returning @code{NO_REGS} you can
2222 force @var{x} into a memory location. For example, rs6000 can load
2223 immediate values into general-purpose registers, but does not have an
2224 instruction for loading an immediate value into a floating-point
2225 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2226 @var{x} is a floating-point constant. If the constant cannot be loaded
2227 into any kind of register, code generation will be better if
2228 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2229 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2231 If an insn has pseudos in it after register allocation, reload will go
2232 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2233 to find the best one. Returning @code{NO_REGS}, in this case, makes
2234 reload add a @code{!} in front of the constraint: the x86 back-end uses
2235 this feature to discourage usage of 387 registers when math is done in
2236 the SSE registers (and vice versa).
2239 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2241 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2242 A C expression that places additional restrictions on the register class
2243 to use when it is necessary to be able to hold a value of mode
2244 @var{mode} in a reload register for which class @var{class} would
2247 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2248 there are certain modes that simply cannot go in certain reload classes.
2250 The value is a register class; perhaps @var{class}, or perhaps another,
2253 Don't define this macro unless the target machine has limitations which
2254 require the macro to do something nontrivial.
2257 @hook TARGET_SECONDARY_RELOAD
2259 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2260 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2261 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2262 These macros are obsolete, new ports should use the target hook
2263 @code{TARGET_SECONDARY_RELOAD} instead.
2265 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2266 target hook. Older ports still define these macros to indicate to the
2267 reload phase that it may
2268 need to allocate at least one register for a reload in addition to the
2269 register to contain the data. Specifically, if copying @var{x} to a
2270 register @var{class} in @var{mode} requires an intermediate register,
2271 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2272 largest register class all of whose registers can be used as
2273 intermediate registers or scratch registers.
2275 If copying a register @var{class} in @var{mode} to @var{x} requires an
2276 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2277 was supposed to be defined be defined to return the largest register
2278 class required. If the
2279 requirements for input and output reloads were the same, the macro
2280 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2283 The values returned by these macros are often @code{GENERAL_REGS}.
2284 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2285 can be directly copied to or from a register of @var{class} in
2286 @var{mode} without requiring a scratch register. Do not define this
2287 macro if it would always return @code{NO_REGS}.
2289 If a scratch register is required (either with or without an
2290 intermediate register), you were supposed to define patterns for
2291 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2292 (@pxref{Standard Names}. These patterns, which were normally
2293 implemented with a @code{define_expand}, should be similar to the
2294 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2297 These patterns need constraints for the reload register and scratch
2299 contain a single register class. If the original reload register (whose
2300 class is @var{class}) can meet the constraint given in the pattern, the
2301 value returned by these macros is used for the class of the scratch
2302 register. Otherwise, two additional reload registers are required.
2303 Their classes are obtained from the constraints in the insn pattern.
2305 @var{x} might be a pseudo-register or a @code{subreg} of a
2306 pseudo-register, which could either be in a hard register or in memory.
2307 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2308 in memory and the hard register number if it is in a register.
2310 These macros should not be used in the case where a particular class of
2311 registers can only be copied to memory and not to another class of
2312 registers. In that case, secondary reload registers are not needed and
2313 would not be helpful. Instead, a stack location must be used to perform
2314 the copy and the @code{mov@var{m}} pattern should use memory as an
2315 intermediate storage. This case often occurs between floating-point and
2319 @hook TARGET_SECONDARY_MEMORY_NEEDED
2321 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2322 Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2323 allocates a stack slot for a memory location needed for register copies.
2324 If this macro is defined, the compiler instead uses the memory location
2325 defined by this macro.
2327 Do not define this macro if you do not define
2328 @code{TARGET_SECONDARY_MEMORY_NEEDED}.
2331 @hook TARGET_SECONDARY_MEMORY_NEEDED_MODE
2333 @hook TARGET_SELECT_EARLY_REMAT_MODES
2335 @hook TARGET_CLASS_LIKELY_SPILLED_P
2337 @hook TARGET_CLASS_MAX_NREGS
2339 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2340 A C expression for the maximum number of consecutive registers
2341 of class @var{class} needed to hold a value of mode @var{mode}.
2343 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact,
2344 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2345 should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2346 @var{mode})} for all @var{regno} values in the class @var{class}.
2348 This macro helps control the handling of multiple-word values
2352 @hook TARGET_CAN_CHANGE_MODE_CLASS
2354 @hook TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS
2358 @hook TARGET_REGISTER_PRIORITY
2360 @hook TARGET_REGISTER_USAGE_LEVELING_P
2362 @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2364 @hook TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P
2366 @hook TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT
2368 @hook TARGET_SPILL_CLASS
2370 @hook TARGET_ADDITIONAL_ALLOCNO_CLASS_P
2372 @hook TARGET_CSTORE_MODE
2374 @hook TARGET_COMPUTE_PRESSURE_CLASSES
2376 @node Stack and Calling
2377 @section Stack Layout and Calling Conventions
2378 @cindex calling conventions
2380 @c prevent bad page break with this line
2381 This describes the stack layout and calling conventions.
2385 * Exception Handling::
2390 * Register Arguments::
2392 * Aggregate Return::
2397 * Shrink-wrapping separate components::
2398 * Stack Smashing Protection::
2399 * Miscellaneous Register Hooks::
2403 @subsection Basic Stack Layout
2404 @cindex stack frame layout
2405 @cindex frame layout
2407 @c prevent bad page break with this line
2408 Here is the basic stack layout.
2410 @defmac STACK_GROWS_DOWNWARD
2411 Define this macro to be true if pushing a word onto the stack moves the stack
2412 pointer to a smaller address, and false otherwise.
2415 @defmac STACK_PUSH_CODE
2416 This macro defines the operation used when something is pushed
2417 on the stack. In RTL, a push operation will be
2418 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2420 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2421 and @code{POST_INC}. Which of these is correct depends on
2422 the stack direction and on whether the stack pointer points
2423 to the last item on the stack or whether it points to the
2424 space for the next item on the stack.
2426 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2427 true, which is almost always right, and @code{PRE_INC} otherwise,
2428 which is often wrong.
2431 @defmac FRAME_GROWS_DOWNWARD
2432 Define this macro to nonzero value if the addresses of local variable slots
2433 are at negative offsets from the frame pointer.
2436 @defmac ARGS_GROW_DOWNWARD
2437 Define this macro if successive arguments to a function occupy decreasing
2438 addresses on the stack.
2441 @hook TARGET_STARTING_FRAME_OFFSET
2443 @defmac STACK_ALIGNMENT_NEEDED
2444 Define to zero to disable final alignment of the stack during reload.
2445 The nonzero default for this macro is suitable for most ports.
2447 On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there
2448 is a register save block following the local block that doesn't require
2449 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2450 stack alignment and do it in the backend.
2453 @defmac STACK_POINTER_OFFSET
2454 Offset from the stack pointer register to the first location at which
2455 outgoing arguments are placed. If not specified, the default value of
2456 zero is used. This is the proper value for most machines.
2458 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2459 the first location at which outgoing arguments are placed.
2462 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2463 Offset from the argument pointer register to the first argument's
2464 address. On some machines it may depend on the data type of the
2467 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2468 the first argument's address.
2471 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2472 Offset from the stack pointer register to an item dynamically allocated
2473 on the stack, e.g., by @code{alloca}.
2475 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2476 length of the outgoing arguments. The default is correct for most
2477 machines. See @file{function.c} for details.
2480 @defmac INITIAL_FRAME_ADDRESS_RTX
2481 A C expression whose value is RTL representing the address of the initial
2482 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2483 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2484 default value will be used. Define this macro in order to make frame pointer
2485 elimination work in the presence of @code{__builtin_frame_address (count)} and
2486 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2489 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2490 A C expression whose value is RTL representing the address in a stack
2491 frame where the pointer to the caller's frame is stored. Assume that
2492 @var{frameaddr} is an RTL expression for the address of the stack frame
2495 If you don't define this macro, the default is to return the value
2496 of @var{frameaddr}---that is, the stack frame address is also the
2497 address of the stack word that points to the previous frame.
2500 @defmac SETUP_FRAME_ADDRESSES
2501 A C expression that produces the machine-specific code to
2502 setup the stack so that arbitrary frames can be accessed. For example,
2503 on the SPARC, we must flush all of the register windows to the stack
2504 before we can access arbitrary stack frames. You will seldom need to
2505 define this macro. The default is to do nothing.
2508 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2510 @defmac FRAME_ADDR_RTX (@var{frameaddr})
2511 A C expression whose value is RTL representing the value of the frame
2512 address for the current frame. @var{frameaddr} is the frame pointer
2513 of the current frame. This is used for __builtin_frame_address.
2514 You need only define this macro if the frame address is not the same
2515 as the frame pointer. Most machines do not need to define it.
2518 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2519 A C expression whose value is RTL representing the value of the return
2520 address for the frame @var{count} steps up from the current frame, after
2521 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2522 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2523 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
2525 The value of the expression must always be the correct address when
2526 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2527 determine the return address of other frames.
2530 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2531 Define this macro to nonzero value if the return address of a particular
2532 stack frame is accessed from the frame pointer of the previous stack
2533 frame. The zero default for this macro is suitable for most ports.
2536 @defmac INCOMING_RETURN_ADDR_RTX
2537 A C expression whose value is RTL representing the location of the
2538 incoming return address at the beginning of any function, before the
2539 prologue. This RTL is either a @code{REG}, indicating that the return
2540 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2543 You only need to define this macro if you want to support call frame
2544 debugging information like that provided by DWARF 2.
2546 If this RTL is a @code{REG}, you should also define
2547 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2550 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2551 A C expression whose value is an integer giving a DWARF 2 column
2552 number that may be used as an alternative return column. The column
2553 must not correspond to any gcc hard register (that is, it must not
2554 be in the range of @code{DWARF_FRAME_REGNUM}).
2556 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2557 general register, but an alternative column needs to be used for signal
2558 frames. Some targets have also used different frame return columns
2562 @defmac DWARF_ZERO_REG
2563 A C expression whose value is an integer giving a DWARF 2 register
2564 number that is considered to always have the value zero. This should
2565 only be defined if the target has an architected zero register, and
2566 someone decided it was a good idea to use that register number to
2567 terminate the stack backtrace. New ports should avoid this.
2570 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2572 @hook TARGET_DWARF_POLY_INDETERMINATE_VALUE
2574 @defmac INCOMING_FRAME_SP_OFFSET
2575 A C expression whose value is an integer giving the offset, in bytes,
2576 from the value of the stack pointer register to the top of the stack
2577 frame at the beginning of any function, before the prologue. The top of
2578 the frame is defined to be the value of the stack pointer in the
2579 previous frame, just before the call instruction.
2581 You only need to define this macro if you want to support call frame
2582 debugging information like that provided by DWARF 2.
2585 @defmac DEFAULT_INCOMING_FRAME_SP_OFFSET
2586 Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all
2587 functions of the same ABI, and when using GAS @code{.cfi_*} directives
2588 must also agree with the default CFI GAS emits. Define this macro
2589 only if @code{INCOMING_FRAME_SP_OFFSET} can have different values
2590 between different functions of the same ABI or when
2591 @code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI.
2594 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2595 A C expression whose value is an integer giving the offset, in bytes,
2596 from the argument pointer to the canonical frame address (cfa). The
2597 final value should coincide with that calculated by
2598 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2599 during virtual register instantiation.
2601 The default value for this macro is
2602 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2603 which is correct for most machines; in general, the arguments are found
2604 immediately before the stack frame. Note that this is not the case on
2605 some targets that save registers into the caller's frame, such as SPARC
2606 and rs6000, and so such targets need to define this macro.
2608 You only need to define this macro if the default is incorrect, and you
2609 want to support call frame debugging information like that provided by
2613 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2614 If defined, a C expression whose value is an integer giving the offset
2615 in bytes from the frame pointer to the canonical frame address (cfa).
2616 The final value should coincide with that calculated by
2617 @code{INCOMING_FRAME_SP_OFFSET}.
2619 Normally the CFA is calculated as an offset from the argument pointer,
2620 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2621 variable due to the ABI, this may not be possible. If this macro is
2622 defined, it implies that the virtual register instantiation should be
2623 based on the frame pointer instead of the argument pointer. Only one
2624 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2628 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2629 If defined, a C expression whose value is an integer giving the offset
2630 in bytes from the canonical frame address (cfa) to the frame base used
2631 in DWARF 2 debug information. The default is zero. A different value
2632 may reduce the size of debug information on some ports.
2635 @node Exception Handling
2636 @subsection Exception Handling Support
2637 @cindex exception handling
2639 @defmac EH_RETURN_DATA_REGNO (@var{N})
2640 A C expression whose value is the @var{N}th register number used for
2641 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2642 @var{N} registers are usable.
2644 The exception handling library routines communicate with the exception
2645 handlers via a set of agreed upon registers. Ideally these registers
2646 should be call-clobbered; it is possible to use call-saved registers,
2647 but may negatively impact code size. The target must support at least
2648 2 data registers, but should define 4 if there are enough free registers.
2650 You must define this macro if you want to support call frame exception
2651 handling like that provided by DWARF 2.
2654 @defmac EH_RETURN_STACKADJ_RTX
2655 A C expression whose value is RTL representing a location in which
2656 to store a stack adjustment to be applied before function return.
2657 This is used to unwind the stack to an exception handler's call frame.
2658 It will be assigned zero on code paths that return normally.
2660 Typically this is a call-clobbered hard register that is otherwise
2661 untouched by the epilogue, but could also be a stack slot.
2663 Do not define this macro if the stack pointer is saved and restored
2664 by the regular prolog and epilog code in the call frame itself; in
2665 this case, the exception handling library routines will update the
2666 stack location to be restored in place. Otherwise, you must define
2667 this macro if you want to support call frame exception handling like
2668 that provided by DWARF 2.
2671 @defmac EH_RETURN_HANDLER_RTX
2672 A C expression whose value is RTL representing a location in which
2673 to store the address of an exception handler to which we should
2674 return. It will not be assigned on code paths that return normally.
2676 Typically this is the location in the call frame at which the normal
2677 return address is stored. For targets that return by popping an
2678 address off the stack, this might be a memory address just below
2679 the @emph{target} call frame rather than inside the current call
2680 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2681 been assigned, so it may be used to calculate the location of the
2684 Some targets have more complex requirements than storing to an
2685 address calculable during initial code generation. In that case
2686 the @code{eh_return} instruction pattern should be used instead.
2688 If you want to support call frame exception handling, you must
2689 define either this macro or the @code{eh_return} instruction pattern.
2692 @defmac RETURN_ADDR_OFFSET
2693 If defined, an integer-valued C expression for which rtl will be generated
2694 to add it to the exception handler address before it is searched in the
2695 exception handling tables, and to subtract it again from the address before
2696 using it to return to the exception handler.
2699 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2700 This macro chooses the encoding of pointers embedded in the exception
2701 handling sections. If at all possible, this should be defined such
2702 that the exception handling section will not require dynamic relocations,
2703 and so may be read-only.
2705 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2706 @var{global} is true if the symbol may be affected by dynamic relocations.
2707 The macro should return a combination of the @code{DW_EH_PE_*} defines
2708 as found in @file{dwarf2.h}.
2710 If this macro is not defined, pointers will not be encoded but
2711 represented directly.
2714 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2715 This macro allows the target to emit whatever special magic is required
2716 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2717 Generic code takes care of pc-relative and indirect encodings; this must
2718 be defined if the target uses text-relative or data-relative encodings.
2720 This is a C statement that branches to @var{done} if the format was
2721 handled. @var{encoding} is the format chosen, @var{size} is the number
2722 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2726 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
2727 This macro allows the target to add CPU and operating system specific
2728 code to the call-frame unwinder for use when there is no unwind data
2729 available. The most common reason to implement this macro is to unwind
2730 through signal frames.
2732 This macro is called from @code{uw_frame_state_for} in
2733 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
2734 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2735 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2736 for the address of the code being executed and @code{context->cfa} for
2737 the stack pointer value. If the frame can be decoded, the register
2738 save addresses should be updated in @var{fs} and the macro should
2739 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
2740 the macro should evaluate to @code{_URC_END_OF_STACK}.
2742 For proper signal handling in Java this macro is accompanied by
2743 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
2746 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
2747 This macro allows the target to add operating system specific code to the
2748 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
2749 usually used for signal or interrupt frames.
2751 This macro is called from @code{uw_update_context} in libgcc's
2752 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2753 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
2754 for the abi and context in the @code{.unwabi} directive. If the
2755 @code{.unwabi} directive can be handled, the register save addresses should
2756 be updated in @var{fs}.
2759 @defmac TARGET_USES_WEAK_UNWIND_INFO
2760 A C expression that evaluates to true if the target requires unwind
2761 info to be given comdat linkage. Define it to be @code{1} if comdat
2762 linkage is necessary. The default is @code{0}.
2765 @node Stack Checking
2766 @subsection Specifying How Stack Checking is Done
2768 GCC will check that stack references are within the boundaries of the
2769 stack, if the option @option{-fstack-check} is specified, in one of
2774 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2775 will assume that you have arranged for full stack checking to be done
2776 at appropriate places in the configuration files. GCC will not do
2777 other special processing.
2780 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
2781 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
2782 that you have arranged for static stack checking (checking of the
2783 static stack frame of functions) to be done at appropriate places
2784 in the configuration files. GCC will only emit code to do dynamic
2785 stack checking (checking on dynamic stack allocations) using the third
2789 If neither of the above are true, GCC will generate code to periodically
2790 ``probe'' the stack pointer using the values of the macros defined below.
2793 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
2794 GCC will change its allocation strategy for large objects if the option
2795 @option{-fstack-check} is specified: they will always be allocated
2796 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
2798 @defmac STACK_CHECK_BUILTIN
2799 A nonzero value if stack checking is done by the configuration files in a
2800 machine-dependent manner. You should define this macro if stack checking
2801 is required by the ABI of your machine or if you would like to do stack
2802 checking in some more efficient way than the generic approach. The default
2803 value of this macro is zero.
2806 @defmac STACK_CHECK_STATIC_BUILTIN
2807 A nonzero value if static stack checking is done by the configuration files
2808 in a machine-dependent manner. You should define this macro if you would
2809 like to do static stack checking in some more efficient way than the generic
2810 approach. The default value of this macro is zero.
2813 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
2814 An integer specifying the interval at which GCC must generate stack probe
2815 instructions, defined as 2 raised to this integer. You will normally
2816 define this macro so that the interval be no larger than the size of
2817 the ``guard pages'' at the end of a stack area. The default value
2818 of 12 (4096-byte interval) is suitable for most systems.
2821 @defmac STACK_CHECK_MOVING_SP
2822 An integer which is nonzero if GCC should move the stack pointer page by page
2823 when doing probes. This can be necessary on systems where the stack pointer
2824 contains the bottom address of the memory area accessible to the executing
2825 thread at any point in time. In this situation an alternate signal stack
2826 is required in order to be able to recover from a stack overflow. The
2827 default value of this macro is zero.
2830 @defmac STACK_CHECK_PROTECT
2831 The number of bytes of stack needed to recover from a stack overflow, for
2832 languages where such a recovery is supported. The default value of 4KB/8KB
2833 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
2834 8KB/12KB with other exception handling mechanisms should be adequate for most
2835 architectures and operating systems.
2838 The following macros are relevant only if neither STACK_CHECK_BUILTIN
2839 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
2840 in the opposite case.
2842 @defmac STACK_CHECK_MAX_FRAME_SIZE
2843 The maximum size of a stack frame, in bytes. GCC will generate probe
2844 instructions in non-leaf functions to ensure at least this many bytes of
2845 stack are available. If a stack frame is larger than this size, stack
2846 checking will not be reliable and GCC will issue a warning. The
2847 default is chosen so that GCC only generates one instruction on most
2848 systems. You should normally not change the default value of this macro.
2851 @defmac STACK_CHECK_FIXED_FRAME_SIZE
2852 GCC uses this value to generate the above warning message. It
2853 represents the amount of fixed frame used by a function, not including
2854 space for any callee-saved registers, temporaries and user variables.
2855 You need only specify an upper bound for this amount and will normally
2856 use the default of four words.
2859 @defmac STACK_CHECK_MAX_VAR_SIZE
2860 The maximum size, in bytes, of an object that GCC will place in the
2861 fixed area of the stack frame when the user specifies
2862 @option{-fstack-check}.
2863 GCC computed the default from the values of the above macros and you will
2864 normally not need to override that default.
2867 @hook TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE
2870 @node Frame Registers
2871 @subsection Registers That Address the Stack Frame
2873 @c prevent bad page break with this line
2874 This discusses registers that address the stack frame.
2876 @defmac STACK_POINTER_REGNUM
2877 The register number of the stack pointer register, which must also be a
2878 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2879 the hardware determines which register this is.
2882 @defmac FRAME_POINTER_REGNUM
2883 The register number of the frame pointer register, which is used to
2884 access automatic variables in the stack frame. On some machines, the
2885 hardware determines which register this is. On other machines, you can
2886 choose any register you wish for this purpose.
2889 @defmac HARD_FRAME_POINTER_REGNUM
2890 On some machines the offset between the frame pointer and starting
2891 offset of the automatic variables is not known until after register
2892 allocation has been done (for example, because the saved registers are
2893 between these two locations). On those machines, define
2894 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2895 be used internally until the offset is known, and define
2896 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2897 used for the frame pointer.
2899 You should define this macro only in the very rare circumstances when it
2900 is not possible to calculate the offset between the frame pointer and
2901 the automatic variables until after register allocation has been
2902 completed. When this macro is defined, you must also indicate in your
2903 definition of @code{ELIMINABLE_REGS} how to eliminate
2904 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2905 or @code{STACK_POINTER_REGNUM}.
2907 Do not define this macro if it would be the same as
2908 @code{FRAME_POINTER_REGNUM}.
2911 @defmac ARG_POINTER_REGNUM
2912 The register number of the arg pointer register, which is used to access
2913 the function's argument list. On some machines, this is the same as the
2914 frame pointer register. On some machines, the hardware determines which
2915 register this is. On other machines, you can choose any register you
2916 wish for this purpose. If this is not the same register as the frame
2917 pointer register, then you must mark it as a fixed register according to
2918 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2919 (@pxref{Elimination}).
2922 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
2923 Define this to a preprocessor constant that is nonzero if
2924 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
2925 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
2926 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
2927 definition is not suitable for use in preprocessor conditionals.
2930 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
2931 Define this to a preprocessor constant that is nonzero if
2932 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
2933 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
2934 ARG_POINTER_REGNUM)}; you only need to define this macro if that
2935 definition is not suitable for use in preprocessor conditionals.
2938 @defmac RETURN_ADDRESS_POINTER_REGNUM
2939 The register number of the return address pointer register, which is used to
2940 access the current function's return address from the stack. On some
2941 machines, the return address is not at a fixed offset from the frame
2942 pointer or stack pointer or argument pointer. This register can be defined
2943 to point to the return address on the stack, and then be converted by
2944 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2946 Do not define this macro unless there is no other way to get the return
2947 address from the stack.
2950 @defmac STATIC_CHAIN_REGNUM
2951 @defmacx STATIC_CHAIN_INCOMING_REGNUM
2952 Register numbers used for passing a function's static chain pointer. If
2953 register windows are used, the register number as seen by the called
2954 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2955 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2956 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2959 The static chain register need not be a fixed register.
2961 If the static chain is passed in memory, these macros should not be
2962 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
2965 @hook TARGET_STATIC_CHAIN
2967 @defmac DWARF_FRAME_REGISTERS
2968 This macro specifies the maximum number of hard registers that can be
2969 saved in a call frame. This is used to size data structures used in
2970 DWARF2 exception handling.
2972 Prior to GCC 3.0, this macro was needed in order to establish a stable
2973 exception handling ABI in the face of adding new hard registers for ISA
2974 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
2975 in the number of hard registers. Nevertheless, this macro can still be
2976 used to reduce the runtime memory requirements of the exception handling
2977 routines, which can be substantial if the ISA contains a lot of
2978 registers that are not call-saved.
2980 If this macro is not defined, it defaults to
2981 @code{FIRST_PSEUDO_REGISTER}.
2984 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
2986 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
2987 for backward compatibility in pre GCC 3.0 compiled code.
2989 If this macro is not defined, it defaults to
2990 @code{DWARF_FRAME_REGISTERS}.
2993 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
2995 Define this macro if the target's representation for dwarf registers
2996 is different than the internal representation for unwind column.
2997 Given a dwarf register, this macro should return the internal unwind
2998 column number to use instead.
3001 @defmac DWARF_FRAME_REGNUM (@var{regno})
3003 Define this macro if the target's representation for dwarf registers
3004 used in .eh_frame or .debug_frame is different from that used in other
3005 debug info sections. Given a GCC hard register number, this macro
3006 should return the .eh_frame register number. The default is
3007 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3011 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3013 Define this macro to map register numbers held in the call frame info
3014 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3015 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3016 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3017 return @code{@var{regno}}.
3021 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3023 Define this macro if the target stores register values as
3024 @code{_Unwind_Word} type in unwind context. It should be defined if
3025 target register size is larger than the size of @code{void *}. The
3026 default is to store register values as @code{void *} type.
3030 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3032 Define this macro to be 1 if the target always uses extended unwind
3033 context with version, args_size and by_value fields. If it is undefined,
3034 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3035 defined and 0 otherwise.
3039 @defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value})
3040 Define this macro if the target has pseudo DWARF registers whose
3041 values need to be computed lazily on demand by the unwinder (such as when
3042 referenced in a CFA expression). The macro returns true if @var{regno}
3043 is such a register and stores its value in @samp{*@var{value}} if so.
3047 @subsection Eliminating Frame Pointer and Arg Pointer
3049 @c prevent bad page break with this line
3050 This is about eliminating the frame pointer and arg pointer.
3052 @hook TARGET_FRAME_POINTER_REQUIRED
3054 @defmac ELIMINABLE_REGS
3055 This macro specifies a table of register pairs used to eliminate
3056 unneeded registers that point into the stack frame.
3058 The definition of this macro is a list of structure initializations, each
3059 of which specifies an original and replacement register.
3061 On some machines, the position of the argument pointer is not known until
3062 the compilation is completed. In such a case, a separate hard register
3063 must be used for the argument pointer. This register can be eliminated by
3064 replacing it with either the frame pointer or the argument pointer,
3065 depending on whether or not the frame pointer has been eliminated.
3067 In this case, you might specify:
3069 #define ELIMINABLE_REGS \
3070 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3071 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3072 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3075 Note that the elimination of the argument pointer with the stack pointer is
3076 specified first since that is the preferred elimination.
3079 @hook TARGET_CAN_ELIMINATE
3081 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3082 This macro returns the initial difference between the specified pair
3083 of registers. The value would be computed from information
3084 such as the result of @code{get_frame_size ()} and the tables of
3085 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3088 @hook TARGET_COMPUTE_FRAME_LAYOUT
3090 @node Stack Arguments
3091 @subsection Passing Function Arguments on the Stack
3092 @cindex arguments on stack
3093 @cindex stack arguments
3095 The macros in this section control how arguments are passed
3096 on the stack. See the following section for other macros that
3097 control passing certain arguments in registers.
3099 @hook TARGET_PROMOTE_PROTOTYPES
3102 A C expression. If nonzero, push insns will be used to pass
3104 If the target machine does not have a push instruction, set it to zero.
3105 That directs GCC to use an alternate strategy: to
3106 allocate the entire argument block and then store the arguments into
3107 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3110 @defmac PUSH_ARGS_REVERSED
3111 A C expression. If nonzero, function arguments will be evaluated from
3112 last to first, rather than from first to last. If this macro is not
3113 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3114 and args grow in opposite directions, and 0 otherwise.
3117 @defmac PUSH_ROUNDING (@var{npushed})
3118 A C expression that is the number of bytes actually pushed onto the
3119 stack when an instruction attempts to push @var{npushed} bytes.
3121 On some machines, the definition
3124 #define PUSH_ROUNDING(BYTES) (BYTES)
3128 will suffice. But on other machines, instructions that appear
3129 to push one byte actually push two bytes in an attempt to maintain
3130 alignment. Then the definition should be
3133 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3136 If the value of this macro has a type, it should be an unsigned type.
3139 @findex outgoing_args_size
3140 @findex crtl->outgoing_args_size
3141 @defmac ACCUMULATE_OUTGOING_ARGS
3142 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3143 will be computed and placed into
3144 @code{crtl->outgoing_args_size}. No space will be pushed
3145 onto the stack for each call; instead, the function prologue should
3146 increase the stack frame size by this amount.
3148 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3152 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3153 Define this macro if functions should assume that stack space has been
3154 allocated for arguments even when their values are passed in
3157 The value of this macro is the size, in bytes, of the area reserved for
3158 arguments passed in registers for the function represented by @var{fndecl},
3159 which can be zero if GCC is calling a library function.
3160 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3163 This space can be allocated by the caller, or be a part of the
3164 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3167 @c above is overfull. not sure what to do. --mew 5feb93 did
3168 @c something, not sure if it looks good. --mew 10feb93
3170 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3171 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3172 Define this macro if space guaranteed when compiling a function body
3173 is different to space required when making a call, a situation that
3174 can arise with K&R style function definitions.
3177 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3178 Define this to a nonzero value if it is the responsibility of the
3179 caller to allocate the area reserved for arguments passed in registers
3180 when calling a function of @var{fntype}. @var{fntype} may be NULL
3181 if the function called is a library function.
3183 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3184 whether the space for these arguments counts in the value of
3185 @code{crtl->outgoing_args_size}.
3188 @defmac STACK_PARMS_IN_REG_PARM_AREA
3189 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3190 stack parameters don't skip the area specified by it.
3191 @c i changed this, makes more sens and it should have taken care of the
3192 @c overfull.. not as specific, tho. --mew 5feb93
3194 Normally, when a parameter is not passed in registers, it is placed on the
3195 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3196 suppresses this behavior and causes the parameter to be passed on the
3197 stack in its natural location.
3200 @hook TARGET_RETURN_POPS_ARGS
3202 @defmac CALL_POPS_ARGS (@var{cum})
3203 A C expression that should indicate the number of bytes a call sequence
3204 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3205 when compiling a function call.
3207 @var{cum} is the variable in which all arguments to the called function
3208 have been accumulated.
3210 On certain architectures, such as the SH5, a call trampoline is used
3211 that pops certain registers off the stack, depending on the arguments
3212 that have been passed to the function. Since this is a property of the
3213 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3217 @node Register Arguments
3218 @subsection Passing Arguments in Registers
3219 @cindex arguments in registers
3220 @cindex registers arguments
3222 This section describes the macros which let you control how various
3223 types of arguments are passed in registers or how they are arranged in
3226 @hook TARGET_FUNCTION_ARG
3228 @hook TARGET_MUST_PASS_IN_STACK
3230 @hook TARGET_FUNCTION_INCOMING_ARG
3232 @hook TARGET_USE_PSEUDO_PIC_REG
3234 @hook TARGET_INIT_PIC_REG
3236 @hook TARGET_ARG_PARTIAL_BYTES
3238 @hook TARGET_PASS_BY_REFERENCE
3240 @hook TARGET_CALLEE_COPIES
3242 @defmac CUMULATIVE_ARGS
3243 A C type for declaring a variable that is used as the first argument
3244 of @code{TARGET_FUNCTION_ARG} and other related values. For some
3245 target machines, the type @code{int} suffices and can hold the number
3246 of bytes of argument so far.
3248 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3249 arguments that have been passed on the stack. The compiler has other
3250 variables to keep track of that. For target machines on which all
3251 arguments are passed on the stack, there is no need to store anything in
3252 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3253 should not be empty, so use @code{int}.
3256 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3257 If defined, this macro is called before generating any code for a
3258 function, but after the @var{cfun} descriptor for the function has been
3259 created. The back end may use this macro to update @var{cfun} to
3260 reflect an ABI other than that which would normally be used by default.
3261 If the compiler is generating code for a compiler-generated function,
3262 @var{fndecl} may be @code{NULL}.
3265 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3266 A C statement (sans semicolon) for initializing the variable
3267 @var{cum} for the state at the beginning of the argument list. The
3268 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3269 is the tree node for the data type of the function which will receive
3270 the args, or 0 if the args are to a compiler support library function.
3271 For direct calls that are not libcalls, @var{fndecl} contain the
3272 declaration node of the function. @var{fndecl} is also set when
3273 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3274 being compiled. @var{n_named_args} is set to the number of named
3275 arguments, including a structure return address if it is passed as a
3276 parameter, when making a call. When processing incoming arguments,
3277 @var{n_named_args} is set to @minus{}1.
3279 When processing a call to a compiler support library function,
3280 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3281 contains the name of the function, as a string. @var{libname} is 0 when
3282 an ordinary C function call is being processed. Thus, each time this
3283 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3284 never both of them at once.
3287 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3288 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3289 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3290 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3291 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3292 0)} is used instead.
3295 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3296 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3297 finding the arguments for the function being compiled. If this macro is
3298 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3300 The value passed for @var{libname} is always 0, since library routines
3301 with special calling conventions are never compiled with GCC@. The
3302 argument @var{libname} exists for symmetry with
3303 @code{INIT_CUMULATIVE_ARGS}.
3304 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3305 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3308 @hook TARGET_FUNCTION_ARG_ADVANCE
3310 @hook TARGET_FUNCTION_ARG_OFFSET
3312 @hook TARGET_FUNCTION_ARG_PADDING
3314 @defmac PAD_VARARGS_DOWN
3315 If defined, a C expression which determines whether the default
3316 implementation of va_arg will attempt to pad down before reading the
3317 next argument, if that argument is smaller than its aligned space as
3318 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3319 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3322 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3323 Specify padding for the last element of a block move between registers and
3324 memory. @var{first} is nonzero if this is the only element. Defining this
3325 macro allows better control of register function parameters on big-endian
3326 machines, without using @code{PARALLEL} rtl. In particular,
3327 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3328 registers, as there is no longer a "wrong" part of a register; For example,
3329 a three byte aggregate may be passed in the high part of a register if so
3333 @hook TARGET_FUNCTION_ARG_BOUNDARY
3335 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3337 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3338 A C expression that is nonzero if @var{regno} is the number of a hard
3339 register in which function arguments are sometimes passed. This does
3340 @emph{not} include implicit arguments such as the static chain and
3341 the structure-value address. On many machines, no registers can be
3342 used for this purpose since all function arguments are pushed on the
3346 @hook TARGET_SPLIT_COMPLEX_ARG
3348 @hook TARGET_BUILD_BUILTIN_VA_LIST
3350 @hook TARGET_ENUM_VA_LIST_P
3352 @hook TARGET_FN_ABI_VA_LIST
3354 @hook TARGET_CANONICAL_VA_LIST_TYPE
3356 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
3358 @hook TARGET_VALID_POINTER_MODE
3360 @hook TARGET_REF_MAY_ALIAS_ERRNO
3362 @hook TARGET_TRANSLATE_MODE_ATTRIBUTE
3364 @hook TARGET_SCALAR_MODE_SUPPORTED_P
3366 @hook TARGET_VECTOR_MODE_SUPPORTED_P
3368 @hook TARGET_ARRAY_MODE
3370 @hook TARGET_ARRAY_MODE_SUPPORTED_P
3372 @hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P
3374 @hook TARGET_FLOATN_MODE
3376 @hook TARGET_FLOATN_BUILTIN_P
3378 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3381 @subsection How Scalar Function Values Are Returned
3382 @cindex return values in registers
3383 @cindex values, returned by functions
3384 @cindex scalars, returned as values
3386 This section discusses the macros that control returning scalars as
3387 values---values that can fit in registers.
3389 @hook TARGET_FUNCTION_VALUE
3391 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3392 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
3393 a new target instead.
3396 @defmac LIBCALL_VALUE (@var{mode})
3397 A C expression to create an RTX representing the place where a library
3398 function returns a value of mode @var{mode}.
3400 Note that ``library function'' in this context means a compiler
3401 support routine, used to perform arithmetic, whose name is known
3402 specially by the compiler and was not mentioned in the C code being
3406 @hook TARGET_LIBCALL_VALUE
3408 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3409 A C expression that is nonzero if @var{regno} is the number of a hard
3410 register in which the values of called function may come back.
3412 A register whose use for returning values is limited to serving as the
3413 second of a pair (for a value of type @code{double}, say) need not be
3414 recognized by this macro. So for most machines, this definition
3418 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3421 If the machine has register windows, so that the caller and the called
3422 function use different registers for the return value, this macro
3423 should recognize only the caller's register numbers.
3425 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3426 for a new target instead.
3429 @hook TARGET_FUNCTION_VALUE_REGNO_P
3431 @defmac APPLY_RESULT_SIZE
3432 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3433 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3434 saving and restoring an arbitrary return value.
3437 @hook TARGET_OMIT_STRUCT_RETURN_REG
3439 @hook TARGET_RETURN_IN_MSB
3441 @node Aggregate Return
3442 @subsection How Large Values Are Returned
3443 @cindex aggregates as return values
3444 @cindex large return values
3445 @cindex returning aggregate values
3446 @cindex structure value address
3448 When a function value's mode is @code{BLKmode} (and in some other
3449 cases), the value is not returned according to
3450 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
3451 caller passes the address of a block of memory in which the value
3452 should be stored. This address is called the @dfn{structure value
3455 This section describes how to control returning structure values in
3458 @hook TARGET_RETURN_IN_MEMORY
3460 @defmac DEFAULT_PCC_STRUCT_RETURN
3461 Define this macro to be 1 if all structure and union return values must be
3462 in memory. Since this results in slower code, this should be defined
3463 only if needed for compatibility with other compilers or with an ABI@.
3464 If you define this macro to be 0, then the conventions used for structure
3465 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3468 If not defined, this defaults to the value 1.
3471 @hook TARGET_STRUCT_VALUE_RTX
3473 @defmac PCC_STATIC_STRUCT_RETURN
3474 Define this macro if the usual system convention on the target machine
3475 for returning structures and unions is for the called function to return
3476 the address of a static variable containing the value.
3478 Do not define this if the usual system convention is for the caller to
3479 pass an address to the subroutine.
3481 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3482 nothing when you use @option{-freg-struct-return} mode.
3485 @hook TARGET_GET_RAW_RESULT_MODE
3487 @hook TARGET_GET_RAW_ARG_MODE
3489 @hook TARGET_EMPTY_RECORD_P
3491 @hook TARGET_WARN_PARAMETER_PASSING_ABI
3494 @subsection Caller-Saves Register Allocation
3496 If you enable it, GCC can save registers around function calls. This
3497 makes it possible to use call-clobbered registers to hold variables that
3498 must live across calls.
3500 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3501 A C expression specifying which mode is required for saving @var{nregs}
3502 of a pseudo-register in call-clobbered hard register @var{regno}. If
3503 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3504 returned. For most machines this macro need not be defined since GCC
3505 will select the smallest suitable mode.
3508 @node Function Entry
3509 @subsection Function Entry and Exit
3510 @cindex function entry and exit
3514 This section describes the macros that output function entry
3515 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3517 @hook TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY
3519 @hook TARGET_ASM_FUNCTION_PROLOGUE
3521 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
3523 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3525 @hook TARGET_ASM_FUNCTION_EPILOGUE
3529 @findex pretend_args_size
3530 @findex crtl->args.pretend_args_size
3531 A region of @code{crtl->args.pretend_args_size} bytes of
3532 uninitialized space just underneath the first argument arriving on the
3533 stack. (This may not be at the very start of the allocated stack region
3534 if the calling sequence has pushed anything else since pushing the stack
3535 arguments. But usually, on such machines, nothing else has been pushed
3536 yet, because the function prologue itself does all the pushing.) This
3537 region is used on machines where an argument may be passed partly in
3538 registers and partly in memory, and, in some cases to support the
3539 features in @code{<stdarg.h>}.
3542 An area of memory used to save certain registers used by the function.
3543 The size of this area, which may also include space for such things as
3544 the return address and pointers to previous stack frames, is
3545 machine-specific and usually depends on which registers have been used
3546 in the function. Machines with register windows often do not require
3550 A region of at least @var{size} bytes, possibly rounded up to an allocation
3551 boundary, to contain the local variables of the function. On some machines,
3552 this region and the save area may occur in the opposite order, with the
3553 save area closer to the top of the stack.
3556 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3557 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3558 @code{crtl->outgoing_args_size} bytes to be used for outgoing
3559 argument lists of the function. @xref{Stack Arguments}.
3562 @defmac EXIT_IGNORE_STACK
3563 Define this macro as a C expression that is nonzero if the return
3564 instruction or the function epilogue ignores the value of the stack
3565 pointer; in other words, if it is safe to delete an instruction to
3566 adjust the stack pointer before a return from the function. The
3569 Note that this macro's value is relevant only for functions for which
3570 frame pointers are maintained. It is never safe to delete a final
3571 stack adjustment in a function that has no frame pointer, and the
3572 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3575 @defmac EPILOGUE_USES (@var{regno})
3576 Define this macro as a C expression that is nonzero for registers that are
3577 used by the epilogue or the @samp{return} pattern. The stack and frame
3578 pointer registers are already assumed to be used as needed.
3581 @defmac EH_USES (@var{regno})
3582 Define this macro as a C expression that is nonzero for registers that are
3583 used by the exception handling mechanism, and so should be considered live
3584 on entry to an exception edge.
3587 @hook TARGET_ASM_OUTPUT_MI_THUNK
3589 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3592 @subsection Generating Code for Profiling
3593 @cindex profiling, code generation
3595 These macros will help you generate code for profiling.
3597 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3598 A C statement or compound statement to output to @var{file} some
3599 assembler code to call the profiling subroutine @code{mcount}.
3602 The details of how @code{mcount} expects to be called are determined by
3603 your operating system environment, not by GCC@. To figure them out,
3604 compile a small program for profiling using the system's installed C
3605 compiler and look at the assembler code that results.
3607 Older implementations of @code{mcount} expect the address of a counter
3608 variable to be loaded into some register. The name of this variable is
3609 @samp{LP} followed by the number @var{labelno}, so you would generate
3610 the name using @samp{LP%d} in a @code{fprintf}.
3613 @defmac PROFILE_HOOK
3614 A C statement or compound statement to output to @var{file} some assembly
3615 code to call the profiling subroutine @code{mcount} even the target does
3616 not support profiling.
3619 @defmac NO_PROFILE_COUNTERS
3620 Define this macro to be an expression with a nonzero value if the
3621 @code{mcount} subroutine on your system does not need a counter variable
3622 allocated for each function. This is true for almost all modern
3623 implementations. If you define this macro, you must not use the
3624 @var{labelno} argument to @code{FUNCTION_PROFILER}.
3627 @defmac PROFILE_BEFORE_PROLOGUE
3628 Define this macro if the code for function profiling should come before
3629 the function prologue. Normally, the profiling code comes after.
3632 @hook TARGET_KEEP_LEAF_WHEN_PROFILED
3635 @subsection Permitting tail calls
3638 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
3640 @hook TARGET_EXTRA_LIVE_ON_ENTRY
3642 @hook TARGET_SET_UP_BY_PROLOGUE
3644 @hook TARGET_WARN_FUNC_RETURN
3646 @node Shrink-wrapping separate components
3647 @subsection Shrink-wrapping separate components
3648 @cindex shrink-wrapping separate components
3650 The prologue may perform a variety of target dependent tasks such as
3651 saving callee-saved registers, saving the return address, aligning the
3652 stack, creating a stack frame, initializing the PIC register, setting
3653 up the static chain, etc.
3655 On some targets some of these tasks may be independent of others and
3656 thus may be shrink-wrapped separately. These independent tasks are
3657 referred to as components and are handled generically by the target
3658 independent parts of GCC.
3660 Using the following hooks those prologue or epilogue components can be
3661 shrink-wrapped separately, so that the initialization (and possibly
3662 teardown) those components do is not done as frequently on execution
3663 paths where this would unnecessary.
3665 What exactly those components are is up to the target code; the generic
3666 code treats them abstractly, as a bit in an @code{sbitmap}. These
3667 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
3668 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
3671 @hook TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
3673 @hook TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
3675 @hook TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
3677 @hook TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS
3679 @hook TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS
3681 @hook TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
3683 @node Stack Smashing Protection
3684 @subsection Stack smashing protection
3685 @cindex stack smashing protection
3687 @hook TARGET_STACK_PROTECT_GUARD
3689 @hook TARGET_STACK_PROTECT_FAIL
3691 @hook TARGET_STACK_PROTECT_RUNTIME_ENABLED_P
3693 @hook TARGET_SUPPORTS_SPLIT_STACK
3695 @hook TARGET_GET_VALID_OPTION_VALUES
3697 @node Miscellaneous Register Hooks
3698 @subsection Miscellaneous register hooks
3699 @cindex miscellaneous register hooks
3701 @hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
3704 @section Implementing the Varargs Macros
3705 @cindex varargs implementation
3707 GCC comes with an implementation of @code{<varargs.h>} and
3708 @code{<stdarg.h>} that work without change on machines that pass arguments
3709 on the stack. Other machines require their own implementations of
3710 varargs, and the two machine independent header files must have
3711 conditionals to include it.
3713 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
3714 the calling convention for @code{va_start}. The traditional
3715 implementation takes just one argument, which is the variable in which
3716 to store the argument pointer. The ISO implementation of
3717 @code{va_start} takes an additional second argument. The user is
3718 supposed to write the last named argument of the function here.
3720 However, @code{va_start} should not use this argument. The way to find
3721 the end of the named arguments is with the built-in functions described
3724 @defmac __builtin_saveregs ()
3725 Use this built-in function to save the argument registers in memory so
3726 that the varargs mechanism can access them. Both ISO and traditional
3727 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3728 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
3730 On some machines, @code{__builtin_saveregs} is open-coded under the
3731 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
3732 other machines, it calls a routine written in assembler language,
3733 found in @file{libgcc2.c}.
3735 Code generated for the call to @code{__builtin_saveregs} appears at the
3736 beginning of the function, as opposed to where the call to
3737 @code{__builtin_saveregs} is written, regardless of what the code is.
3738 This is because the registers must be saved before the function starts
3739 to use them for its own purposes.
3740 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3744 @defmac __builtin_next_arg (@var{lastarg})
3745 This builtin returns the address of the first anonymous stack
3746 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3747 returns the address of the location above the first anonymous stack
3748 argument. Use it in @code{va_start} to initialize the pointer for
3749 fetching arguments from the stack. Also use it in @code{va_start} to
3750 verify that the second parameter @var{lastarg} is the last named argument
3751 of the current function.
3754 @defmac __builtin_classify_type (@var{object})
3755 Since each machine has its own conventions for which data types are
3756 passed in which kind of register, your implementation of @code{va_arg}
3757 has to embody these conventions. The easiest way to categorize the
3758 specified data type is to use @code{__builtin_classify_type} together
3759 with @code{sizeof} and @code{__alignof__}.
3761 @code{__builtin_classify_type} ignores the value of @var{object},
3762 considering only its data type. It returns an integer describing what
3763 kind of type that is---integer, floating, pointer, structure, and so on.
3765 The file @file{typeclass.h} defines an enumeration that you can use to
3766 interpret the values of @code{__builtin_classify_type}.
3769 These machine description macros help implement varargs:
3771 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
3773 @hook TARGET_SETUP_INCOMING_VARARGS
3775 @hook TARGET_STRICT_ARGUMENT_NAMING
3777 @hook TARGET_CALL_ARGS
3779 @hook TARGET_END_CALL_ARGS
3781 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
3783 @hook TARGET_LOAD_BOUNDS_FOR_ARG
3785 @hook TARGET_STORE_BOUNDS_FOR_ARG
3787 @hook TARGET_LOAD_RETURNED_BOUNDS
3789 @hook TARGET_STORE_RETURNED_BOUNDS
3792 @section Support for Nested Functions
3793 @cindex support for nested functions
3794 @cindex trampolines for nested functions
3795 @cindex descriptors for nested functions
3796 @cindex nested functions, support for
3798 Taking the address of a nested function requires special compiler
3799 handling to ensure that the static chain register is loaded when
3800 the function is invoked via an indirect call.
3802 GCC has traditionally supported nested functions by creating an
3803 executable @dfn{trampoline} at run time when the address of a nested
3804 function is taken. This is a small piece of code which normally
3805 resides on the stack, in the stack frame of the containing function.
3806 The trampoline loads the static chain register and then jumps to the
3807 real address of the nested function.
3809 The use of trampolines requires an executable stack, which is a
3810 security risk. To avoid this problem, GCC also supports another
3811 strategy: using descriptors for nested functions. Under this model,
3812 taking the address of a nested function results in a pointer to a
3813 non-executable function descriptor object. Initializing the static chain
3814 from the descriptor is handled at indirect call sites.
3816 On some targets, including HPPA and IA-64, function descriptors may be
3817 mandated by the ABI or be otherwise handled in a target-specific way
3818 by the back end in its code generation strategy for indirect calls.
3819 GCC also provides its own generic descriptor implementation to support the
3820 @option{-fno-trampolines} option. In this case runtime detection of
3821 function descriptors at indirect call sites relies on descriptor
3822 pointers being tagged with a bit that is never set in bare function
3823 addresses. Since GCC's generic function descriptors are
3824 not ABI-compliant, this option is typically used only on a
3825 per-language basis (notably by Ada) or when it can otherwise be
3826 applied to the whole program.
3828 Define the following hook if your backend either implements ABI-specified
3829 descriptor support, or can use GCC's generic descriptor implementation
3830 for nested functions.
3832 @hook TARGET_CUSTOM_FUNCTION_DESCRIPTORS
3834 The following macros tell GCC how to generate code to allocate and
3835 initialize an executable trampoline. You can also use this interface
3836 if your back end needs to create ABI-specified non-executable descriptors; in
3837 this case the "trampoline" created is the descriptor containing data only.
3839 The instructions in an executable trampoline must do two things: load
3840 a constant address into the static chain register, and jump to the real
3841 address of the nested function. On CISC machines such as the m68k,
3842 this requires two instructions, a move immediate and a jump. Then the
3843 two addresses exist in the trampoline as word-long immediate operands.
3844 On RISC machines, it is often necessary to load each address into a
3845 register in two parts. Then pieces of each address form separate
3848 The code generated to initialize the trampoline must store the variable
3849 parts---the static chain value and the function address---into the
3850 immediate operands of the instructions. On a CISC machine, this is
3851 simply a matter of copying each address to a memory reference at the
3852 proper offset from the start of the trampoline. On a RISC machine, it
3853 may be necessary to take out pieces of the address and store them
3856 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
3858 @defmac TRAMPOLINE_SECTION
3859 Return the section into which the trampoline template is to be placed
3860 (@pxref{Sections}). The default value is @code{readonly_data_section}.
3863 @defmac TRAMPOLINE_SIZE
3864 A C expression for the size in bytes of the trampoline, as an integer.
3867 @defmac TRAMPOLINE_ALIGNMENT
3868 Alignment required for trampolines, in bits.
3870 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
3871 is used for aligning trampolines.
3874 @hook TARGET_TRAMPOLINE_INIT
3876 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
3878 Implementing trampolines is difficult on many machines because they have
3879 separate instruction and data caches. Writing into a stack location
3880 fails to clear the memory in the instruction cache, so when the program
3881 jumps to that location, it executes the old contents.
3883 Here are two possible solutions. One is to clear the relevant parts of
3884 the instruction cache whenever a trampoline is set up. The other is to
3885 make all trampolines identical, by having them jump to a standard
3886 subroutine. The former technique makes trampoline execution faster; the
3887 latter makes initialization faster.
3889 To clear the instruction cache when a trampoline is initialized, define
3890 the following macro.
3892 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
3893 If defined, expands to a C expression clearing the @emph{instruction
3894 cache} in the specified interval. The definition of this macro would
3895 typically be a series of @code{asm} statements. Both @var{beg} and
3896 @var{end} are both pointer expressions.
3899 To use a standard subroutine, define the following macro. In addition,
3900 you must make sure that the instructions in a trampoline fill an entire
3901 cache line with identical instructions, or else ensure that the
3902 beginning of the trampoline code is always aligned at the same point in
3903 its cache line. Look in @file{m68k.h} as a guide.
3905 @defmac TRANSFER_FROM_TRAMPOLINE
3906 Define this macro if trampolines need a special subroutine to do their
3907 work. The macro should expand to a series of @code{asm} statements
3908 which will be compiled with GCC@. They go in a library function named
3909 @code{__transfer_from_trampoline}.
3911 If you need to avoid executing the ordinary prologue code of a compiled
3912 C function when you jump to the subroutine, you can do so by placing a
3913 special label of your own in the assembler code. Use one @code{asm}
3914 statement to generate an assembler label, and another to make the label
3915 global. Then trampolines can use that label to jump directly to your
3916 special assembler code.
3920 @section Implicit Calls to Library Routines
3921 @cindex library subroutine names
3922 @cindex @file{libgcc.a}
3924 @c prevent bad page break with this line
3925 Here is an explanation of implicit calls to library routines.
3927 @defmac DECLARE_LIBRARY_RENAMES
3928 This macro, if defined, should expand to a piece of C code that will get
3929 expanded when compiling functions for libgcc.a. It can be used to
3930 provide alternate names for GCC's internal library functions if there
3931 are ABI-mandated names that the compiler should provide.
3934 @findex set_optab_libfunc
3935 @findex init_one_libfunc
3936 @hook TARGET_INIT_LIBFUNCS
3938 @hook TARGET_LIBFUNC_GNU_PREFIX
3940 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
3941 This macro should return @code{true} if the library routine that
3942 implements the floating point comparison operator @var{comparison} in
3943 mode @var{mode} will return a boolean, and @var{false} if it will
3946 GCC's own floating point libraries return tristates from the
3947 comparison operators, so the default returns false always. Most ports
3948 don't need to define this macro.
3951 @defmac TARGET_LIB_INT_CMP_BIASED
3952 This macro should evaluate to @code{true} if the integer comparison
3953 functions (like @code{__cmpdi2}) return 0 to indicate that the first
3954 operand is smaller than the second, 1 to indicate that they are equal,
3955 and 2 to indicate that the first operand is greater than the second.
3956 If this macro evaluates to @code{false} the comparison functions return
3957 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
3958 in @file{libgcc.a}, you do not need to define this macro.
3961 @defmac TARGET_HAS_NO_HW_DIVIDE
3962 This macro should be defined if the target has no hardware divide
3963 instructions. If this macro is defined, GCC will use an algorithm which
3964 make use of simple logical and arithmetic operations for 64-bit
3965 division. If the macro is not defined, GCC will use an algorithm which
3966 make use of a 64-bit by 32-bit divide primitive.
3969 @cindex @code{EDOM}, implicit usage
3972 The value of @code{EDOM} on the target machine, as a C integer constant
3973 expression. If you don't define this macro, GCC does not attempt to
3974 deposit the value of @code{EDOM} into @code{errno} directly. Look in
3975 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
3978 If you do not define @code{TARGET_EDOM}, then compiled code reports
3979 domain errors by calling the library function and letting it report the
3980 error. If mathematical functions on your system use @code{matherr} when
3981 there is an error, then you should leave @code{TARGET_EDOM} undefined so
3982 that @code{matherr} is used normally.
3985 @cindex @code{errno}, implicit usage
3986 @defmac GEN_ERRNO_RTX
3987 Define this macro as a C expression to create an rtl expression that
3988 refers to the global ``variable'' @code{errno}. (On certain systems,
3989 @code{errno} may not actually be a variable.) If you don't define this
3990 macro, a reasonable default is used.
3993 @hook TARGET_LIBC_HAS_FUNCTION
3995 @hook TARGET_LIBC_HAS_FAST_FUNCTION
3997 @defmac NEXT_OBJC_RUNTIME
3998 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
3999 by default. This calling convention involves passing the object, the selector
4000 and the method arguments all at once to the method-lookup library function.
4001 This is the usual setting when targeting Darwin/Mac OS X systems, which have
4002 the NeXT runtime installed.
4004 If the macro is set to 0, the "GNU" Objective-C message sending convention
4005 will be used by default. This convention passes just the object and the
4006 selector to the method-lookup function, which returns a pointer to the method.
4008 In either case, it remains possible to select code-generation for the alternate
4009 scheme, by means of compiler command line switches.
4012 @node Addressing Modes
4013 @section Addressing Modes
4014 @cindex addressing modes
4016 @c prevent bad page break with this line
4017 This is about addressing modes.
4019 @defmac HAVE_PRE_INCREMENT
4020 @defmacx HAVE_PRE_DECREMENT
4021 @defmacx HAVE_POST_INCREMENT
4022 @defmacx HAVE_POST_DECREMENT
4023 A C expression that is nonzero if the machine supports pre-increment,
4024 pre-decrement, post-increment, or post-decrement addressing respectively.
4027 @defmac HAVE_PRE_MODIFY_DISP
4028 @defmacx HAVE_POST_MODIFY_DISP
4029 A C expression that is nonzero if the machine supports pre- or
4030 post-address side-effect generation involving constants other than
4031 the size of the memory operand.
4034 @defmac HAVE_PRE_MODIFY_REG
4035 @defmacx HAVE_POST_MODIFY_REG
4036 A C expression that is nonzero if the machine supports pre- or
4037 post-address side-effect generation involving a register displacement.
4040 @defmac CONSTANT_ADDRESS_P (@var{x})
4041 A C expression that is 1 if the RTX @var{x} is a constant which
4042 is a valid address. On most machines the default definition of
4043 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4044 is acceptable, but a few machines are more restrictive as to which
4045 constant addresses are supported.
4048 @defmac CONSTANT_P (@var{x})
4049 @code{CONSTANT_P}, which is defined by target-independent code,
4050 accepts integer-values expressions whose values are not explicitly
4051 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4052 expressions and @code{const} arithmetic expressions, in addition to
4053 @code{const_int} and @code{const_double} expressions.
4056 @defmac MAX_REGS_PER_ADDRESS
4057 A number, the maximum number of registers that can appear in a valid
4058 memory address. Note that it is up to you to specify a value equal to
4059 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4063 @hook TARGET_LEGITIMATE_ADDRESS_P
4065 @defmac TARGET_MEM_CONSTRAINT
4066 A single character to be used instead of the default @code{'m'}
4067 character for general memory addresses. This defines the constraint
4068 letter which matches the memory addresses accepted by
4069 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
4070 support new address formats in your back end without changing the
4071 semantics of the @code{'m'} constraint. This is necessary in order to
4072 preserve functionality of inline assembly constructs using the
4073 @code{'m'} constraint.
4076 @defmac FIND_BASE_TERM (@var{x})
4077 A C expression to determine the base term of address @var{x},
4078 or to provide a simplified version of @var{x} from which @file{alias.c}
4079 can easily find the base term. This macro is used in only two places:
4080 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
4082 It is always safe for this macro to not be defined. It exists so
4083 that alias analysis can understand machine-dependent addresses.
4085 The typical use of this macro is to handle addresses containing
4086 a label_ref or symbol_ref within an UNSPEC@.
4089 @hook TARGET_LEGITIMIZE_ADDRESS
4091 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4092 A C compound statement that attempts to replace @var{x}, which is an address
4093 that needs reloading, with a valid memory address for an operand of mode
4094 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4095 It is not necessary to define this macro, but it might be useful for
4096 performance reasons.
4098 For example, on the i386, it is sometimes possible to use a single
4099 reload register instead of two by reloading a sum of two pseudo
4100 registers into a register. On the other hand, for number of RISC
4101 processors offsets are limited so that often an intermediate address
4102 needs to be generated in order to address a stack slot. By defining
4103 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4104 generated for adjacent some stack slots can be made identical, and thus
4107 @emph{Note}: This macro should be used with caution. It is necessary
4108 to know something of how reload works in order to effectively use this,
4109 and it is quite easy to produce macros that build in too much knowledge
4110 of reload internals.
4112 @emph{Note}: This macro must be able to reload an address created by a
4113 previous invocation of this macro. If it fails to handle such addresses
4114 then the compiler may generate incorrect code or abort.
4117 The macro definition should use @code{push_reload} to indicate parts that
4118 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4119 suitable to be passed unaltered to @code{push_reload}.
4121 The code generated by this macro must not alter the substructure of
4122 @var{x}. If it transforms @var{x} into a more legitimate form, it
4123 should assign @var{x} (which will always be a C variable) a new value.
4124 This also applies to parts that you change indirectly by calling
4127 @findex strict_memory_address_p
4128 The macro definition may use @code{strict_memory_address_p} to test if
4129 the address has become legitimate.
4132 If you want to change only a part of @var{x}, one standard way of doing
4133 this is to use @code{copy_rtx}. Note, however, that it unshares only a
4134 single level of rtl. Thus, if the part to be changed is not at the
4135 top level, you'll need to replace first the top level.
4136 It is not necessary for this macro to come up with a legitimate
4137 address; but often a machine-dependent strategy can generate better code.
4140 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
4142 @hook TARGET_LEGITIMATE_CONSTANT_P
4144 @hook TARGET_DELEGITIMIZE_ADDRESS
4146 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4148 @hook TARGET_CANNOT_FORCE_CONST_MEM
4150 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4152 @hook TARGET_USE_BLOCKS_FOR_DECL_P
4154 @hook TARGET_BUILTIN_RECIPROCAL
4156 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4158 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4160 @hook TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT
4162 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4164 @hook TARGET_VECTORIZE_VEC_PERM_CONST
4166 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4168 @hook TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION
4170 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4172 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4174 @hook TARGET_VECTORIZE_SPLIT_REDUCTION
4176 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_MODES
4178 @hook TARGET_VECTORIZE_RELATED_MODE
4180 @hook TARGET_VECTORIZE_GET_MASK_MODE
4182 @hook TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE
4184 @hook TARGET_VECTORIZE_INIT_COST
4186 @hook TARGET_VECTORIZE_ADD_STMT_COST
4188 @hook TARGET_VECTORIZE_FINISH_COST
4190 @hook TARGET_VECTORIZE_DESTROY_COST_DATA
4192 @hook TARGET_VECTORIZE_BUILTIN_GATHER
4194 @hook TARGET_VECTORIZE_BUILTIN_SCATTER
4196 @hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN
4198 @hook TARGET_SIMD_CLONE_ADJUST
4200 @hook TARGET_SIMD_CLONE_USABLE
4202 @hook TARGET_SIMT_VF
4204 @hook TARGET_OMP_DEVICE_KIND_ARCH_ISA
4206 @hook TARGET_GOACC_VALIDATE_DIMS
4208 @hook TARGET_GOACC_DIM_LIMIT
4210 @hook TARGET_GOACC_FORK_JOIN
4212 @hook TARGET_GOACC_REDUCTION
4214 @hook TARGET_PREFERRED_ELSE_VALUE
4216 @node Anchored Addresses
4217 @section Anchored Addresses
4218 @cindex anchored addresses
4219 @cindex @option{-fsection-anchors}
4221 GCC usually addresses every static object as a separate entity.
4222 For example, if we have:
4226 int foo (void) @{ return a + b + c; @}
4229 the code for @code{foo} will usually calculate three separate symbolic
4230 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
4231 it would be better to calculate just one symbolic address and access
4232 the three variables relative to it. The equivalent pseudocode would
4238 register int *xr = &x;
4239 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4243 (which isn't valid C). We refer to shared addresses like @code{x} as
4244 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
4246 The hooks below describe the target properties that GCC needs to know
4247 in order to make effective use of section anchors. It won't use
4248 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4249 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4251 @hook TARGET_MIN_ANCHOR_OFFSET
4253 @hook TARGET_MAX_ANCHOR_OFFSET
4255 @hook TARGET_ASM_OUTPUT_ANCHOR
4257 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4259 @node Condition Code
4260 @section Condition Code Status
4261 @cindex condition code status
4263 The macros in this section can be split in two families, according to the
4264 two ways of representing condition codes in GCC.
4266 The first representation is the so called @code{(cc0)} representation
4267 (@pxref{Jump Patterns}), where all instructions can have an implicit
4268 clobber of the condition codes. The second is the condition code
4269 register representation, which provides better schedulability for
4270 architectures that do have a condition code register, but on which
4271 most instructions do not affect it. The latter category includes
4274 The implicit clobbering poses a strong restriction on the placement of
4275 the definition and use of the condition code. In the past the definition
4276 and use were always adjacent. However, recent changes to support trapping
4277 arithmatic may result in the definition and user being in different blocks.
4278 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
4279 the definition may be the source of exception handling edges.
4281 These restrictions can prevent important
4282 optimizations on some machines. For example, on the IBM RS/6000, there
4283 is a delay for taken branches unless the condition code register is set
4284 three instructions earlier than the conditional branch. The instruction
4285 scheduler cannot perform this optimization if it is not permitted to
4286 separate the definition and use of the condition code register.
4288 For this reason, it is possible and suggested to use a register to
4289 represent the condition code for new ports. If there is a specific
4290 condition code register in the machine, use a hard register. If the
4291 condition code or comparison result can be placed in any general register,
4292 or if there are multiple condition registers, use a pseudo register.
4293 Registers used to store the condition code value will usually have a mode
4294 that is in class @code{MODE_CC}.
4296 Alternatively, you can use @code{BImode} if the comparison operator is
4297 specified already in the compare instruction. In this case, you are not
4298 interested in most macros in this section.
4301 * CC0 Condition Codes:: Old style representation of condition codes.
4302 * MODE_CC Condition Codes:: Modern representation of condition codes.
4305 @node CC0 Condition Codes
4306 @subsection Representation of condition codes using @code{(cc0)}
4310 The file @file{conditions.h} defines a variable @code{cc_status} to
4311 describe how the condition code was computed (in case the interpretation of
4312 the condition code depends on the instruction that it was set by). This
4313 variable contains the RTL expressions on which the condition code is
4314 currently based, and several standard flags.
4316 Sometimes additional machine-specific flags must be defined in the machine
4317 description header file. It can also add additional machine-specific
4318 information by defining @code{CC_STATUS_MDEP}.
4320 @defmac CC_STATUS_MDEP
4321 C code for a data type which is used for declaring the @code{mdep}
4322 component of @code{cc_status}. It defaults to @code{int}.
4324 This macro is not used on machines that do not use @code{cc0}.
4327 @defmac CC_STATUS_MDEP_INIT
4328 A C expression to initialize the @code{mdep} field to ``empty''.
4329 The default definition does nothing, since most machines don't use
4330 the field anyway. If you want to use the field, you should probably
4331 define this macro to initialize it.
4333 This macro is not used on machines that do not use @code{cc0}.
4336 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4337 A C compound statement to set the components of @code{cc_status}
4338 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4339 this macro's responsibility to recognize insns that set the condition
4340 code as a byproduct of other activity as well as those that explicitly
4343 This macro is not used on machines that do not use @code{cc0}.
4345 If there are insns that do not set the condition code but do alter
4346 other machine registers, this macro must check to see whether they
4347 invalidate the expressions that the condition code is recorded as
4348 reflecting. For example, on the 68000, insns that store in address
4349 registers do not set the condition code, which means that usually
4350 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4351 insns. But suppose that the previous insn set the condition code
4352 based on location @samp{a4@@(102)} and the current insn stores a new
4353 value in @samp{a4}. Although the condition code is not changed by
4354 this, it will no longer be true that it reflects the contents of
4355 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4356 @code{cc_status} in this case to say that nothing is known about the
4357 condition code value.
4359 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4360 with the results of peephole optimization: insns whose patterns are
4361 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4362 constants which are just the operands. The RTL structure of these
4363 insns is not sufficient to indicate what the insns actually do. What
4364 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4365 @code{CC_STATUS_INIT}.
4367 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4368 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4369 @samp{cc}. This avoids having detailed information about patterns in
4370 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4373 @node MODE_CC Condition Codes
4374 @subsection Representation of condition codes using registers
4378 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4379 On many machines, the condition code may be produced by other instructions
4380 than compares, for example the branch can use directly the condition
4381 code set by a subtract instruction. However, on some machines
4382 when the condition code is set this way some bits (such as the overflow
4383 bit) are not set in the same way as a test instruction, so that a different
4384 branch instruction must be used for some conditional branches. When
4385 this happens, use the machine mode of the condition code register to
4386 record different formats of the condition code register. Modes can
4387 also be used to record which compare instruction (e.g.@: a signed or an
4388 unsigned comparison) produced the condition codes.
4390 If other modes than @code{CCmode} are required, add them to
4391 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4392 a mode given an operand of a compare. This is needed because the modes
4393 have to be chosen not only during RTL generation but also, for example,
4394 by instruction combination. The result of @code{SELECT_CC_MODE} should
4395 be consistent with the mode used in the patterns; for example to support
4396 the case of the add on the SPARC discussed above, we have the pattern
4402 (plus:SI (match_operand:SI 0 "register_operand" "%r")
4403 (match_operand:SI 1 "arith_operand" "rI"))
4410 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
4411 for comparisons whose argument is a @code{plus}:
4414 #define SELECT_CC_MODE(OP,X,Y) \
4415 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4416 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
4417 ? CCFPEmode : CCFPmode) \
4418 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4419 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
4420 ? CCNZmode : CCmode))
4423 Another reason to use modes is to retain information on which operands
4424 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4427 You should define this macro if and only if you define extra CC modes
4428 in @file{@var{machine}-modes.def}.
4431 @hook TARGET_CANONICALIZE_COMPARISON
4433 @defmac REVERSIBLE_CC_MODE (@var{mode})
4434 A C expression whose value is one if it is always safe to reverse a
4435 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4436 can ever return @var{mode} for a floating-point inequality comparison,
4437 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4439 You need not define this macro if it would always returns zero or if the
4440 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4441 For example, here is the definition used on the SPARC, where floating-point
4442 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
4445 #define REVERSIBLE_CC_MODE(MODE) \
4446 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
4450 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
4451 A C expression whose value is reversed condition code of the @var{code} for
4452 comparison done in CC_MODE @var{mode}. The macro is used only in case
4453 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4454 machine has some non-standard way how to reverse certain conditionals. For
4455 instance in case all floating point conditions are non-trapping, compiler may
4456 freely convert unordered compares to ordered ones. Then definition may look
4460 #define REVERSE_CONDITION(CODE, MODE) \
4461 ((MODE) != CCFPmode ? reverse_condition (CODE) \
4462 : reverse_condition_maybe_unordered (CODE))
4466 @hook TARGET_FIXED_CONDITION_CODE_REGS
4468 @hook TARGET_CC_MODES_COMPATIBLE
4470 @hook TARGET_FLAGS_REGNUM
4473 @section Describing Relative Costs of Operations
4474 @cindex costs of instructions
4475 @cindex relative costs
4476 @cindex speed of instructions
4478 These macros let you describe the relative speed of various operations
4479 on the target machine.
4481 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4482 A C expression for the cost of moving data of mode @var{mode} from a
4483 register in class @var{from} to one in class @var{to}. The classes are
4484 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4485 value of 2 is the default; other values are interpreted relative to
4488 It is not required that the cost always equal 2 when @var{from} is the
4489 same as @var{to}; on some machines it is expensive to move between
4490 registers if they are not general registers.
4492 If reload sees an insn consisting of a single @code{set} between two
4493 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4494 classes returns a value of 2, reload does not check to ensure that the
4495 constraints of the insn are met. Setting a cost of other than 2 will
4496 allow reload to verify that the constraints are met. You should do this
4497 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4499 These macros are obsolete, new ports should use the target hook
4500 @code{TARGET_REGISTER_MOVE_COST} instead.
4503 @hook TARGET_REGISTER_MOVE_COST
4505 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4506 A C expression for the cost of moving data of mode @var{mode} between a
4507 register of class @var{class} and memory; @var{in} is zero if the value
4508 is to be written to memory, nonzero if it is to be read in. This cost
4509 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4510 registers and memory is more expensive than between two registers, you
4511 should define this macro to express the relative cost.
4513 If you do not define this macro, GCC uses a default cost of 4 plus
4514 the cost of copying via a secondary reload register, if one is
4515 needed. If your machine requires a secondary reload register to copy
4516 between memory and a register of @var{class} but the reload mechanism is
4517 more complex than copying via an intermediate, define this macro to
4518 reflect the actual cost of the move.
4520 GCC defines the function @code{memory_move_secondary_cost} if
4521 secondary reloads are needed. It computes the costs due to copying via
4522 a secondary register. If your machine copies from memory using a
4523 secondary register in the conventional way but the default base value of
4524 4 is not correct for your machine, define this macro to add some other
4525 value to the result of that function. The arguments to that function
4526 are the same as to this macro.
4528 These macros are obsolete, new ports should use the target hook
4529 @code{TARGET_MEMORY_MOVE_COST} instead.
4532 @hook TARGET_MEMORY_MOVE_COST
4534 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4535 A C expression for the cost of a branch instruction. A value of 1 is
4536 the default; other values are interpreted relative to that. Parameter
4537 @var{speed_p} is true when the branch in question should be optimized
4538 for speed. When it is false, @code{BRANCH_COST} should return a value
4539 optimal for code size rather than performance. @var{predictable_p} is
4540 true for well-predicted branches. On many architectures the
4541 @code{BRANCH_COST} can be reduced then.
4544 Here are additional macros which do not specify precise relative costs,
4545 but only that certain actions are more expensive than GCC would
4548 @defmac SLOW_BYTE_ACCESS
4549 Define this macro as a C expression which is nonzero if accessing less
4550 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4551 faster than accessing a word of memory, i.e., if such access
4552 require more than one instruction or if there is no difference in cost
4553 between byte and (aligned) word loads.
4555 When this macro is not defined, the compiler will access a field by
4556 finding the smallest containing object; when it is defined, a fullword
4557 load will be used if alignment permits. Unless bytes accesses are
4558 faster than word accesses, using word accesses is preferable since it
4559 may eliminate subsequent memory access if subsequent accesses occur to
4560 other fields in the same word of the structure, but to different bytes.
4563 @hook TARGET_SLOW_UNALIGNED_ACCESS
4565 @defmac MOVE_RATIO (@var{speed})
4566 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4567 which a sequence of insns should be generated instead of a
4568 string move insn or a library call. Increasing the value will always
4569 make code faster, but eventually incurs high cost in increased code size.
4571 Note that on machines where the corresponding move insn is a
4572 @code{define_expand} that emits a sequence of insns, this macro counts
4573 the number of such sequences.
4575 The parameter @var{speed} is true if the code is currently being
4576 optimized for speed rather than size.
4578 If you don't define this, a reasonable default is used.
4581 @hook TARGET_USE_BY_PIECES_INFRASTRUCTURE_P
4583 @hook TARGET_COMPARE_BY_PIECES_BRANCH_RATIO
4585 @defmac MOVE_MAX_PIECES
4586 A C expression used by @code{move_by_pieces} to determine the largest unit
4587 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4590 @defmac STORE_MAX_PIECES
4591 A C expression used by @code{store_by_pieces} to determine the largest unit
4592 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
4593 the size of @code{HOST_WIDE_INT}, whichever is smaller.
4596 @defmac COMPARE_MAX_PIECES
4597 A C expression used by @code{compare_by_pieces} to determine the largest unit
4598 a load or store used to compare memory is. Defaults to
4599 @code{MOVE_MAX_PIECES}.
4602 @defmac CLEAR_RATIO (@var{speed})
4603 The threshold of number of scalar move insns, @emph{below} which a sequence
4604 of insns should be generated to clear memory instead of a string clear insn
4605 or a library call. Increasing the value will always make code faster, but
4606 eventually incurs high cost in increased code size.
4608 The parameter @var{speed} is true if the code is currently being
4609 optimized for speed rather than size.
4611 If you don't define this, a reasonable default is used.
4614 @defmac SET_RATIO (@var{speed})
4615 The threshold of number of scalar move insns, @emph{below} which a sequence
4616 of insns should be generated to set memory to a constant value, instead of
4617 a block set insn or a library call.
4618 Increasing the value will always make code faster, but
4619 eventually incurs high cost in increased code size.
4621 The parameter @var{speed} is true if the code is currently being
4622 optimized for speed rather than size.
4624 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
4627 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
4628 A C expression used to determine whether a load postincrement is a good
4629 thing to use for a given mode. Defaults to the value of
4630 @code{HAVE_POST_INCREMENT}.
4633 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
4634 A C expression used to determine whether a load postdecrement is a good
4635 thing to use for a given mode. Defaults to the value of
4636 @code{HAVE_POST_DECREMENT}.
4639 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4640 A C expression used to determine whether a load preincrement is a good
4641 thing to use for a given mode. Defaults to the value of
4642 @code{HAVE_PRE_INCREMENT}.
4645 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4646 A C expression used to determine whether a load predecrement is a good
4647 thing to use for a given mode. Defaults to the value of
4648 @code{HAVE_PRE_DECREMENT}.
4651 @defmac USE_STORE_POST_INCREMENT (@var{mode})
4652 A C expression used to determine whether a store postincrement is a good
4653 thing to use for a given mode. Defaults to the value of
4654 @code{HAVE_POST_INCREMENT}.
4657 @defmac USE_STORE_POST_DECREMENT (@var{mode})
4658 A C expression used to determine whether a store postdecrement is a good
4659 thing to use for a given mode. Defaults to the value of
4660 @code{HAVE_POST_DECREMENT}.
4663 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
4664 This macro is used to determine whether a store preincrement is a good
4665 thing to use for a given mode. Defaults to the value of
4666 @code{HAVE_PRE_INCREMENT}.
4669 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
4670 This macro is used to determine whether a store predecrement is a good
4671 thing to use for a given mode. Defaults to the value of
4672 @code{HAVE_PRE_DECREMENT}.
4675 @defmac NO_FUNCTION_CSE
4676 Define this macro to be true if it is as good or better to call a constant
4677 function address than to call an address kept in a register.
4680 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4681 Define this macro if a non-short-circuit operation produced by
4682 @samp{fold_range_test ()} is optimal. This macro defaults to true if
4683 @code{BRANCH_COST} is greater than or equal to the value 2.
4686 @hook TARGET_OPTAB_SUPPORTED_P
4688 @hook TARGET_RTX_COSTS
4690 @hook TARGET_ADDRESS_COST
4692 @hook TARGET_INSN_COST
4694 @hook TARGET_MAX_NOCE_IFCVT_SEQ_COST
4696 @hook TARGET_NOCE_CONVERSION_PROFITABLE_P
4698 @hook TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P
4700 @hook TARGET_ESTIMATED_POLY_VALUE
4703 @section Adjusting the Instruction Scheduler
4705 The instruction scheduler may need a fair amount of machine-specific
4706 adjustment in order to produce good code. GCC provides several target
4707 hooks for this purpose. It is usually enough to define just a few of
4708 them: try the first ones in this list first.
4710 @hook TARGET_SCHED_ISSUE_RATE
4712 @hook TARGET_SCHED_VARIABLE_ISSUE
4714 @hook TARGET_SCHED_ADJUST_COST
4716 @hook TARGET_SCHED_ADJUST_PRIORITY
4718 @hook TARGET_SCHED_REORDER
4720 @hook TARGET_SCHED_REORDER2
4722 @hook TARGET_SCHED_MACRO_FUSION_P
4724 @hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4726 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4728 @hook TARGET_SCHED_INIT
4730 @hook TARGET_SCHED_FINISH
4732 @hook TARGET_SCHED_INIT_GLOBAL
4734 @hook TARGET_SCHED_FINISH_GLOBAL
4736 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4738 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4740 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4742 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4744 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4746 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4748 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4750 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4752 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4754 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4756 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4758 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4760 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4762 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4764 @hook TARGET_SCHED_DFA_NEW_CYCLE
4766 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4768 @hook TARGET_SCHED_H_I_D_EXTENDED
4770 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4772 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
4774 @hook TARGET_SCHED_SET_SCHED_CONTEXT
4776 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4778 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
4780 @hook TARGET_SCHED_SPECULATE_INSN
4782 @hook TARGET_SCHED_NEEDS_BLOCK_P
4784 @hook TARGET_SCHED_GEN_SPEC_CHECK
4786 @hook TARGET_SCHED_SET_SCHED_FLAGS
4788 @hook TARGET_SCHED_CAN_SPECULATE_INSN
4790 @hook TARGET_SCHED_SMS_RES_MII
4792 @hook TARGET_SCHED_DISPATCH
4794 @hook TARGET_SCHED_DISPATCH_DO
4796 @hook TARGET_SCHED_EXPOSED_PIPELINE
4798 @hook TARGET_SCHED_REASSOCIATION_WIDTH
4800 @hook TARGET_SCHED_FUSION_PRIORITY
4802 @hook TARGET_EXPAND_DIVMOD_LIBFUNC
4805 @section Dividing the Output into Sections (Texts, Data, @dots{})
4806 @c the above section title is WAY too long. maybe cut the part between
4807 @c the (...)? --mew 10feb93
4809 An object file is divided into sections containing different types of
4810 data. In the most common case, there are three sections: the @dfn{text
4811 section}, which holds instructions and read-only data; the @dfn{data
4812 section}, which holds initialized writable data; and the @dfn{bss
4813 section}, which holds uninitialized data. Some systems have other kinds
4816 @file{varasm.c} provides several well-known sections, such as
4817 @code{text_section}, @code{data_section} and @code{bss_section}.
4818 The normal way of controlling a @code{@var{foo}_section} variable
4819 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
4820 as described below. The macros are only read once, when @file{varasm.c}
4821 initializes itself, so their values must be run-time constants.
4822 They may however depend on command-line flags.
4824 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
4825 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
4826 to be string literals.
4828 Some assemblers require a different string to be written every time a
4829 section is selected. If your assembler falls into this category, you
4830 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
4831 @code{get_unnamed_section} to set up the sections.
4833 You must always create a @code{text_section}, either by defining
4834 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
4835 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
4836 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
4837 create a distinct @code{readonly_data_section}, the default is to
4838 reuse @code{text_section}.
4840 All the other @file{varasm.c} sections are optional, and are null
4841 if the target does not provide them.
4843 @defmac TEXT_SECTION_ASM_OP
4844 A C expression whose value is a string, including spacing, containing the
4845 assembler operation that should precede instructions and read-only data.
4846 Normally @code{"\t.text"} is right.
4849 @defmac HOT_TEXT_SECTION_NAME
4850 If defined, a C string constant for the name of the section containing most
4851 frequently executed functions of the program. If not defined, GCC will provide
4852 a default definition if the target supports named sections.
4855 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
4856 If defined, a C string constant for the name of the section containing unlikely
4857 executed functions in the program.
4860 @defmac DATA_SECTION_ASM_OP
4861 A C expression whose value is a string, including spacing, containing the
4862 assembler operation to identify the following data as writable initialized
4863 data. Normally @code{"\t.data"} is right.
4866 @defmac SDATA_SECTION_ASM_OP
4867 If defined, a C expression whose value is a string, including spacing,
4868 containing the assembler operation to identify the following data as
4869 initialized, writable small data.
4872 @defmac READONLY_DATA_SECTION_ASM_OP
4873 A C expression whose value is a string, including spacing, containing the
4874 assembler operation to identify the following data as read-only initialized
4878 @defmac BSS_SECTION_ASM_OP
4879 If defined, a C expression whose value is a string, including spacing,
4880 containing the assembler operation to identify the following data as
4881 uninitialized global data. If not defined, and
4882 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
4883 uninitialized global data will be output in the data section if
4884 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
4888 @defmac SBSS_SECTION_ASM_OP
4889 If defined, a C expression whose value is a string, including spacing,
4890 containing the assembler operation to identify the following data as
4891 uninitialized, writable small data.
4894 @defmac TLS_COMMON_ASM_OP
4895 If defined, a C expression whose value is a string containing the
4896 assembler operation to identify the following data as thread-local
4897 common data. The default is @code{".tls_common"}.
4900 @defmac TLS_SECTION_ASM_FLAG
4901 If defined, a C expression whose value is a character constant
4902 containing the flag used to mark a section as a TLS section. The
4903 default is @code{'T'}.
4906 @defmac INIT_SECTION_ASM_OP
4907 If defined, a C expression whose value is a string, including spacing,
4908 containing the assembler operation to identify the following data as
4909 initialization code. If not defined, GCC will assume such a section does
4910 not exist. This section has no corresponding @code{init_section}
4911 variable; it is used entirely in runtime code.
4914 @defmac FINI_SECTION_ASM_OP
4915 If defined, a C expression whose value is a string, including spacing,
4916 containing the assembler operation to identify the following data as
4917 finalization code. If not defined, GCC will assume such a section does
4918 not exist. This section has no corresponding @code{fini_section}
4919 variable; it is used entirely in runtime code.
4922 @defmac INIT_ARRAY_SECTION_ASM_OP
4923 If defined, a C expression whose value is a string, including spacing,
4924 containing the assembler operation to identify the following data as
4925 part of the @code{.init_array} (or equivalent) section. If not
4926 defined, GCC will assume such a section does not exist. Do not define
4927 both this macro and @code{INIT_SECTION_ASM_OP}.
4930 @defmac FINI_ARRAY_SECTION_ASM_OP
4931 If defined, a C expression whose value is a string, including spacing,
4932 containing the assembler operation to identify the following data as
4933 part of the @code{.fini_array} (or equivalent) section. If not
4934 defined, GCC will assume such a section does not exist. Do not define
4935 both this macro and @code{FINI_SECTION_ASM_OP}.
4938 @defmac MACH_DEP_SECTION_ASM_FLAG
4939 If defined, a C expression whose value is a character constant
4940 containing the flag used to mark a machine-dependent section. This
4941 corresponds to the @code{SECTION_MACH_DEP} section flag.
4944 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
4945 If defined, an ASM statement that switches to a different section
4946 via @var{section_op}, calls @var{function}, and switches back to
4947 the text section. This is used in @file{crtstuff.c} if
4948 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
4949 to initialization and finalization functions from the init and fini
4950 sections. By default, this macro uses a simple function call. Some
4951 ports need hand-crafted assembly code to avoid dependencies on
4952 registers initialized in the function prologue or to ensure that
4953 constant pools don't end up too far way in the text section.
4956 @defmac TARGET_LIBGCC_SDATA_SECTION
4957 If defined, a string which names the section into which small
4958 variables defined in crtstuff and libgcc should go. This is useful
4959 when the target has options for optimizing access to small data, and
4960 you want the crtstuff and libgcc routines to be conservative in what
4961 they expect of your application yet liberal in what your application
4962 expects. For example, for targets with a @code{.sdata} section (like
4963 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
4964 require small data support from your application, but use this macro
4965 to put small data into @code{.sdata} so that your application can
4966 access these variables whether it uses small data or not.
4969 @defmac FORCE_CODE_SECTION_ALIGN
4970 If defined, an ASM statement that aligns a code section to some
4971 arbitrary boundary. This is used to force all fragments of the
4972 @code{.init} and @code{.fini} sections to have to same alignment
4973 and thus prevent the linker from having to add any padding.
4976 @defmac JUMP_TABLES_IN_TEXT_SECTION
4977 Define this macro to be an expression with a nonzero value if jump
4978 tables (for @code{tablejump} insns) should be output in the text
4979 section, along with the assembler instructions. Otherwise, the
4980 readonly data section is used.
4982 This macro is irrelevant if there is no separate readonly data section.
4985 @hook TARGET_ASM_INIT_SECTIONS
4987 @hook TARGET_ASM_RELOC_RW_MASK
4989 @hook TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC
4991 @hook TARGET_ASM_SELECT_SECTION
4993 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
4994 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
4995 for @code{FUNCTION_DECL}s as well as for variables and constants.
4997 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
4998 function has been determined to be likely to be called, and nonzero if
4999 it is unlikely to be called.
5002 @hook TARGET_ASM_UNIQUE_SECTION
5004 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
5006 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
5008 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
5010 @hook TARGET_ASM_SELECT_RTX_SECTION
5012 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
5014 @hook TARGET_ENCODE_SECTION_INFO
5016 @hook TARGET_STRIP_NAME_ENCODING
5018 @hook TARGET_IN_SMALL_DATA_P
5020 @hook TARGET_HAVE_SRODATA_SECTION
5022 @hook TARGET_PROFILE_BEFORE_PROLOGUE
5024 @hook TARGET_BINDS_LOCAL_P
5026 @hook TARGET_HAVE_TLS
5030 @section Position Independent Code
5031 @cindex position independent code
5034 This section describes macros that help implement generation of position
5035 independent code. Simply defining these macros is not enough to
5036 generate valid PIC; you must also add support to the hook
5037 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
5038 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
5039 must modify the definition of @samp{movsi} to do something appropriate
5040 when the source operand contains a symbolic address. You may also
5041 need to alter the handling of switch statements so that they use
5043 @c i rearranged the order of the macros above to try to force one of
5044 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5046 @defmac PIC_OFFSET_TABLE_REGNUM
5047 The register number of the register used to address a table of static
5048 data addresses in memory. In some cases this register is defined by a
5049 processor's ``application binary interface'' (ABI)@. When this macro
5050 is defined, RTL is generated for this register once, as with the stack
5051 pointer and frame pointer registers. If this macro is not defined, it
5052 is up to the machine-dependent files to allocate such a register (if
5053 necessary). Note that this register must be fixed when in use (e.g.@:
5054 when @code{flag_pic} is true).
5057 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5058 A C expression that is nonzero if the register defined by
5059 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
5060 the default is zero. Do not define
5061 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5064 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
5065 A C expression that is nonzero if @var{x} is a legitimate immediate
5066 operand on the target machine when generating position independent code.
5067 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5068 check this. You can also assume @var{flag_pic} is true, so you need not
5069 check it either. You need not define this macro if all constants
5070 (including @code{SYMBOL_REF}) can be immediate operands when generating
5071 position independent code.
5074 @node Assembler Format
5075 @section Defining the Output Assembler Language
5077 This section describes macros whose principal purpose is to describe how
5078 to write instructions in assembler language---rather than what the
5082 * File Framework:: Structural information for the assembler file.
5083 * Data Output:: Output of constants (numbers, strings, addresses).
5084 * Uninitialized Data:: Output of uninitialized variables.
5085 * Label Output:: Output and generation of labels.
5086 * Initialization:: General principles of initialization
5087 and termination routines.
5088 * Macros for Initialization::
5089 Specific macros that control the handling of
5090 initialization and termination routines.
5091 * Instruction Output:: Output of actual instructions.
5092 * Dispatch Tables:: Output of jump tables.
5093 * Exception Region Output:: Output of exception region code.
5094 * Alignment Output:: Pseudo ops for alignment and skipping data.
5097 @node File Framework
5098 @subsection The Overall Framework of an Assembler File
5099 @cindex assembler format
5100 @cindex output of assembler code
5102 @c prevent bad page break with this line
5103 This describes the overall framework of an assembly file.
5105 @findex default_file_start
5106 @hook TARGET_ASM_FILE_START
5108 @hook TARGET_ASM_FILE_START_APP_OFF
5110 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5112 @hook TARGET_ASM_FILE_END
5114 @deftypefun void file_end_indicate_exec_stack ()
5115 Some systems use a common convention, the @samp{.note.GNU-stack}
5116 special section, to indicate whether or not an object file relies on
5117 the stack being executable. If your system uses this convention, you
5118 should define @code{TARGET_ASM_FILE_END} to this function. If you
5119 need to do other things in that hook, have your hook function call
5123 @hook TARGET_ASM_LTO_START
5125 @hook TARGET_ASM_LTO_END
5127 @hook TARGET_ASM_CODE_END
5129 @defmac ASM_COMMENT_START
5130 A C string constant describing how to begin a comment in the target
5131 assembler language. The compiler assumes that the comment will end at
5132 the end of the line.
5136 A C string constant for text to be output before each @code{asm}
5137 statement or group of consecutive ones. Normally this is
5138 @code{"#APP"}, which is a comment that has no effect on most
5139 assemblers but tells the GNU assembler that it must check the lines
5140 that follow for all valid assembler constructs.
5144 A C string constant for text to be output after each @code{asm}
5145 statement or group of consecutive ones. Normally this is
5146 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5147 time-saving assumptions that are valid for ordinary compiler output.
5150 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5151 A C statement to output COFF information or DWARF debugging information
5152 which indicates that filename @var{name} is the current source file to
5153 the stdio stream @var{stream}.
5155 This macro need not be defined if the standard form of output
5156 for the file format in use is appropriate.
5159 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5161 @hook TARGET_ASM_OUTPUT_IDENT
5163 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5164 A C statement to output the string @var{string} to the stdio stream
5165 @var{stream}. If you do not call the function @code{output_quoted_string}
5166 in your config files, GCC will only call it to output filenames to
5167 the assembler source. So you can use it to canonicalize the format
5168 of the filename using this macro.
5171 @hook TARGET_ASM_NAMED_SECTION
5173 @hook TARGET_ASM_ELF_FLAGS_NUMERIC
5175 @hook TARGET_ASM_FUNCTION_SECTION
5177 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5179 @hook TARGET_HAVE_NAMED_SECTIONS
5180 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5181 It must not be modified by command-line option processing.
5184 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5185 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5187 @hook TARGET_SECTION_TYPE_FLAGS
5189 @hook TARGET_ASM_RECORD_GCC_SWITCHES
5191 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5195 @subsection Output of Data
5198 @hook TARGET_ASM_BYTE_OP
5200 @hook TARGET_ASM_INTEGER
5202 @hook TARGET_ASM_DECL_END
5204 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5206 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5207 A C statement to output to the stdio stream @var{stream} an assembler
5208 instruction to assemble a string constant containing the @var{len}
5209 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5210 @code{char *} and @var{len} a C expression of type @code{int}.
5212 If the assembler has a @code{.ascii} pseudo-op as found in the
5213 Berkeley Unix assembler, do not define the macro
5214 @code{ASM_OUTPUT_ASCII}.
5217 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5218 A C statement to output word @var{n} of a function descriptor for
5219 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5220 is defined, and is otherwise unused.
5223 @defmac CONSTANT_POOL_BEFORE_FUNCTION
5224 You may define this macro as a C expression. You should define the
5225 expression to have a nonzero value if GCC should output the constant
5226 pool for a function before the code for the function, or a zero value if
5227 GCC should output the constant pool after the function. If you do
5228 not define this macro, the usual case, GCC will output the constant
5229 pool before the function.
5232 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5233 A C statement to output assembler commands to define the start of the
5234 constant pool for a function. @var{funname} is a string giving
5235 the name of the function. Should the return type of the function
5236 be required, it can be obtained via @var{fundecl}. @var{size}
5237 is the size, in bytes, of the constant pool that will be written
5238 immediately after this call.
5240 If no constant-pool prefix is required, the usual case, this macro need
5244 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5245 A C statement (with or without semicolon) to output a constant in the
5246 constant pool, if it needs special treatment. (This macro need not do
5247 anything for RTL expressions that can be output normally.)
5249 The argument @var{file} is the standard I/O stream to output the
5250 assembler code on. @var{x} is the RTL expression for the constant to
5251 output, and @var{mode} is the machine mode (in case @var{x} is a
5252 @samp{const_int}). @var{align} is the required alignment for the value
5253 @var{x}; you should output an assembler directive to force this much
5256 The argument @var{labelno} is a number to use in an internal label for
5257 the address of this pool entry. The definition of this macro is
5258 responsible for outputting the label definition at the proper place.
5259 Here is how to do this:
5262 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5265 When you output a pool entry specially, you should end with a
5266 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5267 entry from being output a second time in the usual manner.
5269 You need not define this macro if it would do nothing.
5272 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5273 A C statement to output assembler commands to at the end of the constant
5274 pool for a function. @var{funname} is a string giving the name of the
5275 function. Should the return type of the function be required, you can
5276 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5277 constant pool that GCC wrote immediately before this call.
5279 If no constant-pool epilogue is required, the usual case, you need not
5283 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5284 Define this macro as a C expression which is nonzero if @var{C} is
5285 used as a logical line separator by the assembler. @var{STR} points
5286 to the position in the string where @var{C} was found; this can be used if
5287 a line separator uses multiple characters.
5289 If you do not define this macro, the default is that only
5290 the character @samp{;} is treated as a logical line separator.
5293 @hook TARGET_ASM_OPEN_PAREN
5295 These macros are provided by @file{real.h} for writing the definitions
5296 of @code{ASM_OUTPUT_DOUBLE} and the like:
5298 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5299 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5300 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5301 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5302 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5303 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5304 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5305 target's floating point representation, and store its bit pattern in
5306 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5307 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5308 simple @code{long int}. For the others, it should be an array of
5309 @code{long int}. The number of elements in this array is determined
5310 by the size of the desired target floating point data type: 32 bits of
5311 it go in each @code{long int} array element. Each array element holds
5312 32 bits of the result, even if @code{long int} is wider than 32 bits
5313 on the host machine.
5315 The array element values are designed so that you can print them out
5316 using @code{fprintf} in the order they should appear in the target
5320 @node Uninitialized Data
5321 @subsection Output of Uninitialized Variables
5323 Each of the macros in this section is used to do the whole job of
5324 outputting a single uninitialized variable.
5326 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5327 A C statement (sans semicolon) to output to the stdio stream
5328 @var{stream} the assembler definition of a common-label named
5329 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5330 is the size rounded up to whatever alignment the caller wants. It is
5331 possible that @var{size} may be zero, for instance if a struct with no
5332 other member than a zero-length array is defined. In this case, the
5333 backend must output a symbol definition that allocates at least one
5334 byte, both so that the address of the resulting object does not compare
5335 equal to any other, and because some object formats cannot even express
5336 the concept of a zero-sized common symbol, as that is how they represent
5337 an ordinary undefined external.
5339 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5340 output the name itself; before and after that, output the additional
5341 assembler syntax for defining the name, and a newline.
5343 This macro controls how the assembler definitions of uninitialized
5344 common global variables are output.
5347 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5348 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5349 separate, explicit argument. If you define this macro, it is used in
5350 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5351 handling the required alignment of the variable. The alignment is specified
5352 as the number of bits.
5355 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5356 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5357 variable to be output, if there is one, or @code{NULL_TREE} if there
5358 is no corresponding variable. If you define this macro, GCC will use it
5359 in place of both @code{ASM_OUTPUT_COMMON} and
5360 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5361 the variable's decl in order to chose what to output.
5364 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5365 A C statement (sans semicolon) to output to the stdio stream
5366 @var{stream} the assembler definition of uninitialized global @var{decl} named
5367 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
5368 is the alignment specified as the number of bits.
5370 Try to use function @code{asm_output_aligned_bss} defined in file
5371 @file{varasm.c} when defining this macro. If unable, use the expression
5372 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5373 before and after that, output the additional assembler syntax for defining
5374 the name, and a newline.
5376 There are two ways of handling global BSS@. One is to define this macro.
5377 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5378 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5379 You do not need to do both.
5381 Some languages do not have @code{common} data, and require a
5382 non-common form of global BSS in order to handle uninitialized globals
5383 efficiently. C++ is one example of this. However, if the target does
5384 not support global BSS, the front end may choose to make globals
5385 common in order to save space in the object file.
5388 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5389 A C statement (sans semicolon) to output to the stdio stream
5390 @var{stream} the assembler definition of a local-common-label named
5391 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5392 is the size rounded up to whatever alignment the caller wants.
5394 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5395 output the name itself; before and after that, output the additional
5396 assembler syntax for defining the name, and a newline.
5398 This macro controls how the assembler definitions of uninitialized
5399 static variables are output.
5402 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5403 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5404 separate, explicit argument. If you define this macro, it is used in
5405 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5406 handling the required alignment of the variable. The alignment is specified
5407 as the number of bits.
5410 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5411 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5412 variable to be output, if there is one, or @code{NULL_TREE} if there
5413 is no corresponding variable. If you define this macro, GCC will use it
5414 in place of both @code{ASM_OUTPUT_DECL} and
5415 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5416 the variable's decl in order to chose what to output.
5420 @subsection Output and Generation of Labels
5422 @c prevent bad page break with this line
5423 This is about outputting labels.
5425 @findex assemble_name
5426 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5427 A C statement (sans semicolon) to output to the stdio stream
5428 @var{stream} the assembler definition of a label named @var{name}.
5429 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5430 output the name itself; before and after that, output the additional
5431 assembler syntax for defining the name, and a newline. A default
5432 definition of this macro is provided which is correct for most systems.
5435 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5436 A C statement (sans semicolon) to output to the stdio stream
5437 @var{stream} the assembler definition of a label named @var{name} of
5439 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5440 output the name itself; before and after that, output the additional
5441 assembler syntax for defining the name, and a newline. A default
5442 definition of this macro is provided which is correct for most systems.
5444 If this macro is not defined, then the function name is defined in the
5445 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5448 @findex assemble_name_raw
5449 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5450 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5451 to refer to a compiler-generated label. The default definition uses
5452 @code{assemble_name_raw}, which is like @code{assemble_name} except
5453 that it is more efficient.
5457 A C string containing the appropriate assembler directive to specify the
5458 size of a symbol, without any arguments. On systems that use ELF, the
5459 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5460 systems, the default is not to define this macro.
5462 Define this macro only if it is correct to use the default definitions
5463 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5464 for your system. If you need your own custom definitions of those
5465 macros, or if you do not need explicit symbol sizes at all, do not
5469 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5470 A C statement (sans semicolon) to output to the stdio stream
5471 @var{stream} a directive telling the assembler that the size of the
5472 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
5473 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5477 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5478 A C statement (sans semicolon) to output to the stdio stream
5479 @var{stream} a directive telling the assembler to calculate the size of
5480 the symbol @var{name} by subtracting its address from the current
5483 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5484 provided. The default assumes that the assembler recognizes a special
5485 @samp{.} symbol as referring to the current address, and can calculate
5486 the difference between this and another symbol. If your assembler does
5487 not recognize @samp{.} or cannot do calculations with it, you will need
5488 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5491 @defmac NO_DOLLAR_IN_LABEL
5492 Define this macro if the assembler does not accept the character
5493 @samp{$} in label names. By default constructors and destructors in
5494 G++ have @samp{$} in the identifiers. If this macro is defined,
5495 @samp{.} is used instead.
5498 @defmac NO_DOT_IN_LABEL
5499 Define this macro if the assembler does not accept the character
5500 @samp{.} in label names. By default constructors and destructors in G++
5501 have names that use @samp{.}. If this macro is defined, these names
5502 are rewritten to avoid @samp{.}.
5506 A C string containing the appropriate assembler directive to specify the
5507 type of a symbol, without any arguments. On systems that use ELF, the
5508 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5509 systems, the default is not to define this macro.
5511 Define this macro only if it is correct to use the default definition of
5512 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5513 custom definition of this macro, or if you do not need explicit symbol
5514 types at all, do not define this macro.
5517 @defmac TYPE_OPERAND_FMT
5518 A C string which specifies (using @code{printf} syntax) the format of
5519 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
5520 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5521 the default is not to define this macro.
5523 Define this macro only if it is correct to use the default definition of
5524 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5525 custom definition of this macro, or if you do not need explicit symbol
5526 types at all, do not define this macro.
5529 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5530 A C statement (sans semicolon) to output to the stdio stream
5531 @var{stream} a directive telling the assembler that the type of the
5532 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
5533 that string is always either @samp{"function"} or @samp{"object"}, but
5534 you should not count on this.
5536 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5537 definition of this macro is provided.
5540 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5541 A C statement (sans semicolon) to output to the stdio stream
5542 @var{stream} any text necessary for declaring the name @var{name} of a
5543 function which is being defined. This macro is responsible for
5544 outputting the label definition (perhaps using
5545 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5546 @code{FUNCTION_DECL} tree node representing the function.
5548 If this macro is not defined, then the function name is defined in the
5549 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5551 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5555 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5556 A C statement (sans semicolon) to output to the stdio stream
5557 @var{stream} any text necessary for declaring the size of a function
5558 which is being defined. The argument @var{name} is the name of the
5559 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5560 representing the function.
5562 If this macro is not defined, then the function size is not defined.
5564 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5568 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5569 A C statement (sans semicolon) to output to the stdio stream
5570 @var{stream} any text necessary for declaring the name @var{name} of a
5571 cold function partition which is being defined. This macro is responsible
5572 for outputting the label definition (perhaps using
5573 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5574 @code{FUNCTION_DECL} tree node representing the function.
5576 If this macro is not defined, then the cold partition name is defined in the
5577 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5579 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5583 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5584 A C statement (sans semicolon) to output to the stdio stream
5585 @var{stream} any text necessary for declaring the size of a cold function
5586 partition which is being defined. The argument @var{name} is the name of the
5587 cold partition of the function. The argument @var{decl} is the
5588 @code{FUNCTION_DECL} tree node representing the function.
5590 If this macro is not defined, then the partition size is not defined.
5592 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5596 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5597 A C statement (sans semicolon) to output to the stdio stream
5598 @var{stream} any text necessary for declaring the name @var{name} of an
5599 initialized variable which is being defined. This macro must output the
5600 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5601 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5603 If this macro is not defined, then the variable name is defined in the
5604 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5606 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5607 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5610 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
5612 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5613 A C statement (sans semicolon) to output to the stdio stream
5614 @var{stream} any text necessary for claiming a register @var{regno}
5615 for a global variable @var{decl} with name @var{name}.
5617 If you don't define this macro, that is equivalent to defining it to do
5621 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5622 A C statement (sans semicolon) to finish up declaring a variable name
5623 once the compiler has processed its initializer fully and thus has had a
5624 chance to determine the size of an array when controlled by an
5625 initializer. This is used on systems where it's necessary to declare
5626 something about the size of the object.
5628 If you don't define this macro, that is equivalent to defining it to do
5631 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5632 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5635 @hook TARGET_ASM_GLOBALIZE_LABEL
5637 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
5639 @hook TARGET_ASM_ASSEMBLE_UNDEFINED_DECL
5641 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5642 A C statement (sans semicolon) to output to the stdio stream
5643 @var{stream} some commands that will make the label @var{name} weak;
5644 that is, available for reference from other files but only used if
5645 no other definition is available. Use the expression
5646 @code{assemble_name (@var{stream}, @var{name})} to output the name
5647 itself; before and after that, output the additional assembler syntax
5648 for making that name weak, and a newline.
5650 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5651 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5655 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5656 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5657 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5658 or variable decl. If @var{value} is not @code{NULL}, this C statement
5659 should output to the stdio stream @var{stream} assembler code which
5660 defines (equates) the weak symbol @var{name} to have the value
5661 @var{value}. If @var{value} is @code{NULL}, it should output commands
5662 to make @var{name} weak.
5665 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5666 Outputs a directive that enables @var{name} to be used to refer to
5667 symbol @var{value} with weak-symbol semantics. @code{decl} is the
5668 declaration of @code{name}.
5671 @defmac SUPPORTS_WEAK
5672 A preprocessor constant expression which evaluates to true if the target
5673 supports weak symbols.
5675 If you don't define this macro, @file{defaults.h} provides a default
5676 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5677 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5680 @defmac TARGET_SUPPORTS_WEAK
5681 A C expression which evaluates to true if the target supports weak symbols.
5683 If you don't define this macro, @file{defaults.h} provides a default
5684 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
5685 this macro if you want to control weak symbol support with a compiler
5686 flag such as @option{-melf}.
5689 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
5690 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5691 public symbol such that extra copies in multiple translation units will
5692 be discarded by the linker. Define this macro if your object file
5693 format provides support for this concept, such as the @samp{COMDAT}
5694 section flags in the Microsoft Windows PE/COFF format, and this support
5695 requires changes to @var{decl}, such as putting it in a separate section.
5698 @defmac SUPPORTS_ONE_ONLY
5699 A C expression which evaluates to true if the target supports one-only
5702 If you don't define this macro, @file{varasm.c} provides a default
5703 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5704 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5705 you want to control one-only symbol support with a compiler flag, or if
5706 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5707 be emitted as one-only.
5710 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
5712 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5713 A C expression that evaluates to true if the target's linker expects
5714 that weak symbols do not appear in a static archive's table of contents.
5715 The default is @code{0}.
5717 Leaving weak symbols out of an archive's table of contents means that,
5718 if a symbol will only have a definition in one translation unit and
5719 will have undefined references from other translation units, that
5720 symbol should not be weak. Defining this macro to be nonzero will
5721 thus have the effect that certain symbols that would normally be weak
5722 (explicit template instantiations, and vtables for polymorphic classes
5723 with noninline key methods) will instead be nonweak.
5725 The C++ ABI requires this macro to be zero. Define this macro for
5726 targets where full C++ ABI compliance is impossible and where linker
5727 restrictions require weak symbols to be left out of a static archive's
5731 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5732 A C statement (sans semicolon) to output to the stdio stream
5733 @var{stream} any text necessary for declaring the name of an external
5734 symbol named @var{name} which is referenced in this compilation but
5735 not defined. The value of @var{decl} is the tree node for the
5738 This macro need not be defined if it does not need to output anything.
5739 The GNU assembler and most Unix assemblers don't require anything.
5742 @hook TARGET_ASM_EXTERNAL_LIBCALL
5744 @hook TARGET_ASM_MARK_DECL_PRESERVED
5746 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5747 A C statement (sans semicolon) to output to the stdio stream
5748 @var{stream} a reference in assembler syntax to a label named
5749 @var{name}. This should add @samp{_} to the front of the name, if that
5750 is customary on your operating system, as it is in most Berkeley Unix
5751 systems. This macro is used in @code{assemble_name}.
5754 @hook TARGET_MANGLE_ASSEMBLER_NAME
5756 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5757 A C statement (sans semicolon) to output a reference to
5758 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
5759 will be used to output the name of the symbol. This macro may be used
5760 to modify the way a symbol is referenced depending on information
5761 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5764 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5765 A C statement (sans semicolon) to output a reference to @var{buf}, the
5766 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
5767 @code{assemble_name} will be used to output the name of the symbol.
5768 This macro is not used by @code{output_asm_label}, or the @code{%l}
5769 specifier that calls it; the intention is that this macro should be set
5770 when it is necessary to output a label differently when its address is
5774 @hook TARGET_ASM_INTERNAL_LABEL
5776 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5777 A C statement to output to the stdio stream @var{stream} a debug info
5778 label whose name is made from the string @var{prefix} and the number
5779 @var{num}. This is useful for VLIW targets, where debug info labels
5780 may need to be treated differently than branch target labels. On some
5781 systems, branch target labels must be at the beginning of instruction
5782 bundles, but debug info labels can occur in the middle of instruction
5785 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5789 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5790 A C statement to store into the string @var{string} a label whose name
5791 is made from the string @var{prefix} and the number @var{num}.
5793 This string, when output subsequently by @code{assemble_name}, should
5794 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5795 with the same @var{prefix} and @var{num}.
5797 If the string begins with @samp{*}, then @code{assemble_name} will
5798 output the rest of the string unchanged. It is often convenient for
5799 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5800 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5801 to output the string, and may change it. (Of course,
5802 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5803 you should know what it does on your machine.)
5806 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5807 A C expression to assign to @var{outvar} (which is a variable of type
5808 @code{char *}) a newly allocated string made from the string
5809 @var{name} and the number @var{number}, with some suitable punctuation
5810 added. Use @code{alloca} to get space for the string.
5812 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5813 produce an assembler label for an internal static variable whose name is
5814 @var{name}. Therefore, the string must be such as to result in valid
5815 assembler code. The argument @var{number} is different each time this
5816 macro is executed; it prevents conflicts between similarly-named
5817 internal static variables in different scopes.
5819 Ideally this string should not be a valid C identifier, to prevent any
5820 conflict with the user's own symbols. Most assemblers allow periods
5821 or percent signs in assembler symbols; putting at least one of these
5822 between the name and the number will suffice.
5824 If this macro is not defined, a default definition will be provided
5825 which is correct for most systems.
5828 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5829 A C statement to output to the stdio stream @var{stream} assembler code
5830 which defines (equates) the symbol @var{name} to have the value @var{value}.
5833 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5834 correct for most systems.
5837 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5838 A C statement to output to the stdio stream @var{stream} assembler code
5839 which defines (equates) the symbol whose tree node is @var{decl_of_name}
5840 to have the value of the tree node @var{decl_of_value}. This macro will
5841 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5842 the tree nodes are available.
5845 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5846 correct for most systems.
5849 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
5850 A C statement that evaluates to true if the assembler code which defines
5851 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
5852 of the tree node @var{decl_of_value} should be emitted near the end of the
5853 current compilation unit. The default is to not defer output of defines.
5854 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
5855 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
5858 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5859 A C statement to output to the stdio stream @var{stream} assembler code
5860 which defines (equates) the weak symbol @var{name} to have the value
5861 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
5862 an undefined weak symbol.
5864 Define this macro if the target only supports weak aliases; define
5865 @code{ASM_OUTPUT_DEF} instead if possible.
5868 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5869 Define this macro to override the default assembler names used for
5870 Objective-C methods.
5872 The default name is a unique method number followed by the name of the
5873 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5874 the category is also included in the assembler name (e.g.@:
5877 These names are safe on most systems, but make debugging difficult since
5878 the method's selector is not present in the name. Therefore, particular
5879 systems define other ways of computing names.
5881 @var{buf} is an expression of type @code{char *} which gives you a
5882 buffer in which to store the name; its length is as long as
5883 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5884 50 characters extra.
5886 The argument @var{is_inst} specifies whether the method is an instance
5887 method or a class method; @var{class_name} is the name of the class;
5888 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
5889 in a category); and @var{sel_name} is the name of the selector.
5891 On systems where the assembler can handle quoted names, you can use this
5892 macro to provide more human-readable names.
5895 @node Initialization
5896 @subsection How Initialization Functions Are Handled
5897 @cindex initialization routines
5898 @cindex termination routines
5899 @cindex constructors, output of
5900 @cindex destructors, output of
5902 The compiled code for certain languages includes @dfn{constructors}
5903 (also called @dfn{initialization routines})---functions to initialize
5904 data in the program when the program is started. These functions need
5905 to be called before the program is ``started''---that is to say, before
5906 @code{main} is called.
5908 Compiling some languages generates @dfn{destructors} (also called
5909 @dfn{termination routines}) that should be called when the program
5912 To make the initialization and termination functions work, the compiler
5913 must output something in the assembler code to cause those functions to
5914 be called at the appropriate time. When you port the compiler to a new
5915 system, you need to specify how to do this.
5917 There are two major ways that GCC currently supports the execution of
5918 initialization and termination functions. Each way has two variants.
5919 Much of the structure is common to all four variations.
5921 @findex __CTOR_LIST__
5922 @findex __DTOR_LIST__
5923 The linker must build two lists of these functions---a list of
5924 initialization functions, called @code{__CTOR_LIST__}, and a list of
5925 termination functions, called @code{__DTOR_LIST__}.
5927 Each list always begins with an ignored function pointer (which may hold
5928 0, @minus{}1, or a count of the function pointers after it, depending on
5929 the environment). This is followed by a series of zero or more function
5930 pointers to constructors (or destructors), followed by a function
5931 pointer containing zero.
5933 Depending on the operating system and its executable file format, either
5934 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5935 time and exit time. Constructors are called in reverse order of the
5936 list; destructors in forward order.
5938 The best way to handle static constructors works only for object file
5939 formats which provide arbitrarily-named sections. A section is set
5940 aside for a list of constructors, and another for a list of destructors.
5941 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5942 object file that defines an initialization function also puts a word in
5943 the constructor section to point to that function. The linker
5944 accumulates all these words into one contiguous @samp{.ctors} section.
5945 Termination functions are handled similarly.
5947 This method will be chosen as the default by @file{target-def.h} if
5948 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
5949 support arbitrary sections, but does support special designated
5950 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
5951 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
5953 When arbitrary sections are available, there are two variants, depending
5954 upon how the code in @file{crtstuff.c} is called. On systems that
5955 support a @dfn{.init} section which is executed at program startup,
5956 parts of @file{crtstuff.c} are compiled into that section. The
5957 program is linked by the @command{gcc} driver like this:
5960 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
5963 The prologue of a function (@code{__init}) appears in the @code{.init}
5964 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
5965 for the function @code{__fini} in the @dfn{.fini} section. Normally these
5966 files are provided by the operating system or by the GNU C library, but
5967 are provided by GCC for a few targets.
5969 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
5970 compiled from @file{crtstuff.c}. They contain, among other things, code
5971 fragments within the @code{.init} and @code{.fini} sections that branch
5972 to routines in the @code{.text} section. The linker will pull all parts
5973 of a section together, which results in a complete @code{__init} function
5974 that invokes the routines we need at startup.
5976 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5979 If no init section is available, when GCC compiles any function called
5980 @code{main} (or more accurately, any function designated as a program
5981 entry point by the language front end calling @code{expand_main_function}),
5982 it inserts a procedure call to @code{__main} as the first executable code
5983 after the function prologue. The @code{__main} function is defined
5984 in @file{libgcc2.c} and runs the global constructors.
5986 In file formats that don't support arbitrary sections, there are again
5987 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5988 and an `a.out' format must be used. In this case,
5989 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
5990 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5991 and with the address of the void function containing the initialization
5992 code as its value. The GNU linker recognizes this as a request to add
5993 the value to a @dfn{set}; the values are accumulated, and are eventually
5994 placed in the executable as a vector in the format described above, with
5995 a leading (ignored) count and a trailing zero element.
5996 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
5997 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5998 the compilation of @code{main} to call @code{__main} as above, starting
5999 the initialization process.
6001 The last variant uses neither arbitrary sections nor the GNU linker.
6002 This is preferable when you want to do dynamic linking and when using
6003 file formats which the GNU linker does not support, such as `ECOFF'@. In
6004 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6005 termination functions are recognized simply by their names. This requires
6006 an extra program in the linkage step, called @command{collect2}. This program
6007 pretends to be the linker, for use with GCC; it does its job by running
6008 the ordinary linker, but also arranges to include the vectors of
6009 initialization and termination functions. These functions are called
6010 via @code{__main} as described above. In order to use this method,
6011 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6014 The following section describes the specific macros that control and
6015 customize the handling of initialization and termination functions.
6018 @node Macros for Initialization
6019 @subsection Macros Controlling Initialization Routines
6021 Here are the macros that control how the compiler handles initialization
6022 and termination functions:
6024 @defmac INIT_SECTION_ASM_OP
6025 If defined, a C string constant, including spacing, for the assembler
6026 operation to identify the following data as initialization code. If not
6027 defined, GCC will assume such a section does not exist. When you are
6028 using special sections for initialization and termination functions, this
6029 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6030 run the initialization functions.
6033 @defmac HAS_INIT_SECTION
6034 If defined, @code{main} will not call @code{__main} as described above.
6035 This macro should be defined for systems that control start-up code
6036 on a symbol-by-symbol basis, such as OSF/1, and should not
6037 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6040 @defmac LD_INIT_SWITCH
6041 If defined, a C string constant for a switch that tells the linker that
6042 the following symbol is an initialization routine.
6045 @defmac LD_FINI_SWITCH
6046 If defined, a C string constant for a switch that tells the linker that
6047 the following symbol is a finalization routine.
6050 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6051 If defined, a C statement that will write a function that can be
6052 automatically called when a shared library is loaded. The function
6053 should call @var{func}, which takes no arguments. If not defined, and
6054 the object format requires an explicit initialization function, then a
6055 function called @code{_GLOBAL__DI} will be generated.
6057 This function and the following one are used by collect2 when linking a
6058 shared library that needs constructors or destructors, or has DWARF2
6059 exception tables embedded in the code.
6062 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6063 If defined, a C statement that will write a function that can be
6064 automatically called when a shared library is unloaded. The function
6065 should call @var{func}, which takes no arguments. If not defined, and
6066 the object format requires an explicit finalization function, then a
6067 function called @code{_GLOBAL__DD} will be generated.
6070 @defmac INVOKE__main
6071 If defined, @code{main} will call @code{__main} despite the presence of
6072 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6073 where the init section is not actually run automatically, but is still
6074 useful for collecting the lists of constructors and destructors.
6077 @defmac SUPPORTS_INIT_PRIORITY
6078 If nonzero, the C++ @code{init_priority} attribute is supported and the
6079 compiler should emit instructions to control the order of initialization
6080 of objects. If zero, the compiler will issue an error message upon
6081 encountering an @code{init_priority} attribute.
6084 @hook TARGET_HAVE_CTORS_DTORS
6086 @hook TARGET_ASM_CONSTRUCTOR
6088 @hook TARGET_ASM_DESTRUCTOR
6090 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6091 generated for the generated object file will have static linkage.
6093 If your system uses @command{collect2} as the means of processing
6094 constructors, then that program normally uses @command{nm} to scan
6095 an object file for constructor functions to be called.
6097 On certain kinds of systems, you can define this macro to make
6098 @command{collect2} work faster (and, in some cases, make it work at all):
6100 @defmac OBJECT_FORMAT_COFF
6101 Define this macro if the system uses COFF (Common Object File Format)
6102 object files, so that @command{collect2} can assume this format and scan
6103 object files directly for dynamic constructor/destructor functions.
6105 This macro is effective only in a native compiler; @command{collect2} as
6106 part of a cross compiler always uses @command{nm} for the target machine.
6109 @defmac REAL_NM_FILE_NAME
6110 Define this macro as a C string constant containing the file name to use
6111 to execute @command{nm}. The default is to search the path normally for
6116 @command{collect2} calls @command{nm} to scan object files for static
6117 constructors and destructors and LTO info. By default, @option{-n} is
6118 passed. Define @code{NM_FLAGS} to a C string constant if other options
6119 are needed to get the same output format as GNU @command{nm -n}
6123 If your system supports shared libraries and has a program to list the
6124 dynamic dependencies of a given library or executable, you can define
6125 these macros to enable support for running initialization and
6126 termination functions in shared libraries:
6129 Define this macro to a C string constant containing the name of the program
6130 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6133 @defmac PARSE_LDD_OUTPUT (@var{ptr})
6134 Define this macro to be C code that extracts filenames from the output
6135 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6136 of type @code{char *} that points to the beginning of a line of output
6137 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6138 code must advance @var{ptr} to the beginning of the filename on that
6139 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6142 @defmac SHLIB_SUFFIX
6143 Define this macro to a C string constant containing the default shared
6144 library extension of the target (e.g., @samp{".so"}). @command{collect2}
6145 strips version information after this suffix when generating global
6146 constructor and destructor names. This define is only needed on targets
6147 that use @command{collect2} to process constructors and destructors.
6150 @node Instruction Output
6151 @subsection Output of Assembler Instructions
6153 @c prevent bad page break with this line
6154 This describes assembler instruction output.
6156 @defmac REGISTER_NAMES
6157 A C initializer containing the assembler's names for the machine
6158 registers, each one as a C string constant. This is what translates
6159 register numbers in the compiler into assembler language.
6162 @defmac ADDITIONAL_REGISTER_NAMES
6163 If defined, a C initializer for an array of structures containing a name
6164 and a register number. This macro defines additional names for hard
6165 registers, thus allowing the @code{asm} option in declarations to refer
6166 to registers using alternate names.
6169 @defmac OVERLAPPING_REGISTER_NAMES
6170 If defined, a C initializer for an array of structures containing a
6171 name, a register number and a count of the number of consecutive
6172 machine registers the name overlaps. This macro defines additional
6173 names for hard registers, thus allowing the @code{asm} option in
6174 declarations to refer to registers using alternate names. Unlike
6175 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6176 register name implies multiple underlying registers.
6178 This macro should be used when it is important that a clobber in an
6179 @code{asm} statement clobbers all the underlying values implied by the
6180 register name. For example, on ARM, clobbering the double-precision
6181 VFP register ``d0'' implies clobbering both single-precision registers
6185 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6186 Define this macro if you are using an unusual assembler that
6187 requires different names for the machine instructions.
6189 The definition is a C statement or statements which output an
6190 assembler instruction opcode to the stdio stream @var{stream}. The
6191 macro-operand @var{ptr} is a variable of type @code{char *} which
6192 points to the opcode name in its ``internal'' form---the form that is
6193 written in the machine description. The definition should output the
6194 opcode name to @var{stream}, performing any translation you desire, and
6195 increment the variable @var{ptr} to point at the end of the opcode
6196 so that it will not be output twice.
6198 In fact, your macro definition may process less than the entire opcode
6199 name, or more than the opcode name; but if you want to process text
6200 that includes @samp{%}-sequences to substitute operands, you must take
6201 care of the substitution yourself. Just be sure to increment
6202 @var{ptr} over whatever text should not be output normally.
6204 @findex recog_data.operand
6205 If you need to look at the operand values, they can be found as the
6206 elements of @code{recog_data.operand}.
6208 If the macro definition does nothing, the instruction is output
6212 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6213 If defined, a C statement to be executed just prior to the output of
6214 assembler code for @var{insn}, to modify the extracted operands so
6215 they will be output differently.
6217 Here the argument @var{opvec} is the vector containing the operands
6218 extracted from @var{insn}, and @var{noperands} is the number of
6219 elements of the vector which contain meaningful data for this insn.
6220 The contents of this vector are what will be used to convert the insn
6221 template into assembler code, so you can change the assembler output
6222 by changing the contents of the vector.
6224 This macro is useful when various assembler syntaxes share a single
6225 file of instruction patterns; by defining this macro differently, you
6226 can cause a large class of instructions to be output differently (such
6227 as with rearranged operands). Naturally, variations in assembler
6228 syntax affecting individual insn patterns ought to be handled by
6229 writing conditional output routines in those patterns.
6231 If this macro is not defined, it is equivalent to a null statement.
6234 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
6236 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6237 A C compound statement to output to stdio stream @var{stream} the
6238 assembler syntax for an instruction operand @var{x}. @var{x} is an
6241 @var{code} is a value that can be used to specify one of several ways
6242 of printing the operand. It is used when identical operands must be
6243 printed differently depending on the context. @var{code} comes from
6244 the @samp{%} specification that was used to request printing of the
6245 operand. If the specification was just @samp{%@var{digit}} then
6246 @var{code} is 0; if the specification was @samp{%@var{ltr}
6247 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6250 If @var{x} is a register, this macro should print the register's name.
6251 The names can be found in an array @code{reg_names} whose type is
6252 @code{char *[]}. @code{reg_names} is initialized from
6253 @code{REGISTER_NAMES}.
6255 When the machine description has a specification @samp{%@var{punct}}
6256 (a @samp{%} followed by a punctuation character), this macro is called
6257 with a null pointer for @var{x} and the punctuation character for
6261 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6262 A C expression which evaluates to true if @var{code} is a valid
6263 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6264 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6265 punctuation characters (except for the standard one, @samp{%}) are used
6269 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6270 A C compound statement to output to stdio stream @var{stream} the
6271 assembler syntax for an instruction operand that is a memory reference
6272 whose address is @var{x}. @var{x} is an RTL expression.
6274 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6275 On some machines, the syntax for a symbolic address depends on the
6276 section that the address refers to. On these machines, define the hook
6277 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6278 @code{symbol_ref}, and then check for it here. @xref{Assembler
6282 @findex dbr_sequence_length
6283 @defmac DBR_OUTPUT_SEQEND (@var{file})
6284 A C statement, to be executed after all slot-filler instructions have
6285 been output. If necessary, call @code{dbr_sequence_length} to
6286 determine the number of slots filled in a sequence (zero if not
6287 currently outputting a sequence), to decide how many no-ops to output,
6290 Don't define this macro if it has nothing to do, but it is helpful in
6291 reading assembly output if the extent of the delay sequence is made
6292 explicit (e.g.@: with white space).
6295 @findex final_sequence
6296 Note that output routines for instructions with delay slots must be
6297 prepared to deal with not being output as part of a sequence
6298 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6299 found.) The variable @code{final_sequence} is null when not
6300 processing a sequence, otherwise it contains the @code{sequence} rtx
6304 @defmac REGISTER_PREFIX
6305 @defmacx LOCAL_LABEL_PREFIX
6306 @defmacx USER_LABEL_PREFIX
6307 @defmacx IMMEDIATE_PREFIX
6308 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6309 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6310 @file{final.c}). These are useful when a single @file{md} file must
6311 support multiple assembler formats. In that case, the various @file{tm.h}
6312 files can define these macros differently.
6315 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6316 If defined this macro should expand to a series of @code{case}
6317 statements which will be parsed inside the @code{switch} statement of
6318 the @code{asm_fprintf} function. This allows targets to define extra
6319 printf formats which may useful when generating their assembler
6320 statements. Note that uppercase letters are reserved for future
6321 generic extensions to asm_fprintf, and so are not available to target
6322 specific code. The output file is given by the parameter @var{file}.
6323 The varargs input pointer is @var{argptr} and the rest of the format
6324 string, starting the character after the one that is being switched
6325 upon, is pointed to by @var{format}.
6328 @defmac ASSEMBLER_DIALECT
6329 If your target supports multiple dialects of assembler language (such as
6330 different opcodes), define this macro as a C expression that gives the
6331 numeric index of the assembler language dialect to use, with zero as the
6334 If this macro is defined, you may use constructs of the form
6336 @samp{@{option0|option1|option2@dots{}@}}
6339 in the output templates of patterns (@pxref{Output Template}) or in the
6340 first argument of @code{asm_fprintf}. This construct outputs
6341 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6342 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
6343 within these strings retain their usual meaning. If there are fewer
6344 alternatives within the braces than the value of
6345 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6346 to print curly braces or @samp{|} character in assembler output directly,
6347 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6349 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6350 @samp{@}} do not have any special meaning when used in templates or
6351 operands to @code{asm_fprintf}.
6353 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6354 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6355 the variations in assembler language syntax with that mechanism. Define
6356 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6357 if the syntax variant are larger and involve such things as different
6358 opcodes or operand order.
6361 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6362 A C expression to output to @var{stream} some assembler code
6363 which will push hard register number @var{regno} onto the stack.
6364 The code need not be optimal, since this macro is used only when
6368 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6369 A C expression to output to @var{stream} some assembler code
6370 which will pop hard register number @var{regno} off of the stack.
6371 The code need not be optimal, since this macro is used only when
6375 @node Dispatch Tables
6376 @subsection Output of Dispatch Tables
6378 @c prevent bad page break with this line
6379 This concerns dispatch tables.
6381 @cindex dispatch table
6382 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6383 A C statement to output to the stdio stream @var{stream} an assembler
6384 pseudo-instruction to generate a difference between two labels.
6385 @var{value} and @var{rel} are the numbers of two internal labels. The
6386 definitions of these labels are output using
6387 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6388 way here. For example,
6391 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6392 @var{value}, @var{rel})
6395 You must provide this macro on machines where the addresses in a
6396 dispatch table are relative to the table's own address. If defined, GCC
6397 will also use this macro on all machines when producing PIC@.
6398 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6399 mode and flags can be read.
6402 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6403 This macro should be provided on machines where the addresses
6404 in a dispatch table are absolute.
6406 The definition should be a C statement to output to the stdio stream
6407 @var{stream} an assembler pseudo-instruction to generate a reference to
6408 a label. @var{value} is the number of an internal label whose
6409 definition is output using @code{(*targetm.asm_out.internal_label)}.
6413 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6417 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6418 Define this if the label before a jump-table needs to be output
6419 specially. The first three arguments are the same as for
6420 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6421 jump-table which follows (a @code{jump_table_data} containing an
6422 @code{addr_vec} or @code{addr_diff_vec}).
6424 This feature is used on system V to output a @code{swbeg} statement
6427 If this macro is not defined, these labels are output with
6428 @code{(*targetm.asm_out.internal_label)}.
6431 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6432 Define this if something special must be output at the end of a
6433 jump-table. The definition should be a C statement to be executed
6434 after the assembler code for the table is written. It should write
6435 the appropriate code to stdio stream @var{stream}. The argument
6436 @var{table} is the jump-table insn, and @var{num} is the label-number
6437 of the preceding label.
6439 If this macro is not defined, nothing special is output at the end of
6443 @hook TARGET_ASM_POST_CFI_STARTPROC
6445 @hook TARGET_ASM_EMIT_UNWIND_LABEL
6447 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6449 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6451 @hook TARGET_ASM_UNWIND_EMIT
6453 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6455 @node Exception Region Output
6456 @subsection Assembler Commands for Exception Regions
6458 @c prevent bad page break with this line
6460 This describes commands marking the start and the end of an exception
6463 @defmac EH_FRAME_SECTION_NAME
6464 If defined, a C string constant for the name of the section containing
6465 exception handling frame unwind information. If not defined, GCC will
6466 provide a default definition if the target supports named sections.
6467 @file{crtstuff.c} uses this macro to switch to the appropriate section.
6469 You should define this symbol if your target supports DWARF 2 frame
6470 unwind information and the default definition does not work.
6473 @defmac EH_FRAME_THROUGH_COLLECT2
6474 If defined, DWARF 2 frame unwind information will identified by
6475 specially named labels. The collect2 process will locate these
6476 labels and generate code to register the frames.
6478 This might be necessary, for instance, if the system linker will not
6479 place the eh_frames in-between the sentinals from @file{crtstuff.c},
6480 or if the system linker does garbage collection and sections cannot
6481 be marked as not to be collected.
6484 @defmac EH_TABLES_CAN_BE_READ_ONLY
6485 Define this macro to 1 if your target is such that no frame unwind
6486 information encoding used with non-PIC code will ever require a
6487 runtime relocation, but the linker may not support merging read-only
6488 and read-write sections into a single read-write section.
6491 @defmac MASK_RETURN_ADDR
6492 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6493 that it does not contain any extraneous set bits in it.
6496 @defmac DWARF2_UNWIND_INFO
6497 Define this macro to 0 if your target supports DWARF 2 frame unwind
6498 information, but it does not yet work with exception handling.
6499 Otherwise, if your target supports this information (if it defines
6500 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6501 GCC will provide a default definition of 1.
6504 @hook TARGET_EXCEPT_UNWIND_INFO
6505 This hook defines the mechanism that will be used for exception handling
6506 by the target. If the target has ABI specified unwind tables, the hook
6507 should return @code{UI_TARGET}. If the target is to use the
6508 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6509 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
6510 information, the hook should return @code{UI_DWARF2}.
6512 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6513 This may end up simplifying other parts of target-specific code. The
6514 default implementation of this hook never returns @code{UI_NONE}.
6516 Note that the value returned by this hook should be constant. It should
6517 not depend on anything except the command-line switches described by
6518 @var{opts}. In particular, the
6519 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6520 macros and builtin functions related to exception handling are set up
6521 depending on this setting.
6523 The default implementation of the hook first honors the
6524 @option{--enable-sjlj-exceptions} configure option, then
6525 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
6526 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6527 must define this hook so that @var{opts} is used correctly.
6530 @hook TARGET_UNWIND_TABLES_DEFAULT
6531 This variable should be set to @code{true} if the target ABI requires unwinding
6532 tables even when exceptions are not used. It must not be modified by
6533 command-line option processing.
6536 @defmac DONT_USE_BUILTIN_SETJMP
6537 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6538 should use the @code{setjmp}/@code{longjmp} functions from the C library
6539 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6542 @defmac JMP_BUF_SIZE
6543 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6544 defined. Define this macro if the default size of @code{jmp_buf} buffer
6545 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6546 is not large enough, or if it is much too large.
6547 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6550 @defmac DWARF_CIE_DATA_ALIGNMENT
6551 This macro need only be defined if the target might save registers in the
6552 function prologue at an offset to the stack pointer that is not aligned to
6553 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6554 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
6555 minimum alignment otherwise. @xref{DWARF}. Only applicable if
6556 the target supports DWARF 2 frame unwind information.
6559 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6561 @hook TARGET_DWARF_REGISTER_SPAN
6563 @hook TARGET_DWARF_FRAME_REG_MODE
6565 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6567 @hook TARGET_ASM_TTYPE
6569 @hook TARGET_ARM_EABI_UNWINDER
6571 @node Alignment Output
6572 @subsection Assembler Commands for Alignment
6574 @c prevent bad page break with this line
6575 This describes commands for alignment.
6577 @defmac JUMP_ALIGN (@var{label})
6578 The alignment (log base 2) to put in front of @var{label}, which is
6579 a common destination of jumps and has no fallthru incoming edge.
6581 This macro need not be defined if you don't want any special alignment
6582 to be done at such a time. Most machine descriptions do not currently
6585 Unless it's necessary to inspect the @var{label} parameter, it is better
6586 to set the variable @var{align_jumps} in the target's
6587 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6588 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6591 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6592 The alignment (log base 2) to put in front of @var{label}, which follows
6595 This macro need not be defined if you don't want any special alignment
6596 to be done at such a time. Most machine descriptions do not currently
6600 @defmac LOOP_ALIGN (@var{label})
6601 The alignment (log base 2) to put in front of @var{label} that heads
6602 a frequently executed basic block (usually the header of a loop).
6604 This macro need not be defined if you don't want any special alignment
6605 to be done at such a time. Most machine descriptions do not currently
6608 Unless it's necessary to inspect the @var{label} parameter, it is better
6609 to set the variable @code{align_loops} in the target's
6610 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6611 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6614 @defmac LABEL_ALIGN (@var{label})
6615 The alignment (log base 2) to put in front of @var{label}.
6616 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6617 the maximum of the specified values is used.
6619 Unless it's necessary to inspect the @var{label} parameter, it is better
6620 to set the variable @code{align_labels} in the target's
6621 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6622 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6625 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6626 A C statement to output to the stdio stream @var{stream} an assembler
6627 instruction to advance the location counter by @var{nbytes} bytes.
6628 Those bytes should be zero when loaded. @var{nbytes} will be a C
6629 expression of type @code{unsigned HOST_WIDE_INT}.
6632 @defmac ASM_NO_SKIP_IN_TEXT
6633 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6634 text section because it fails to put zeros in the bytes that are skipped.
6635 This is true on many Unix systems, where the pseudo--op to skip bytes
6636 produces no-op instructions rather than zeros when used in the text
6640 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6641 A C statement to output to the stdio stream @var{stream} an assembler
6642 command to advance the location counter to a multiple of 2 to the
6643 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6646 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6647 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6648 for padding, if necessary.
6651 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6652 A C statement to output to the stdio stream @var{stream} an assembler
6653 command to advance the location counter to a multiple of 2 to the
6654 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6655 satisfy the alignment request. @var{power} and @var{max_skip} will be
6656 a C expression of type @code{int}.
6660 @node Debugging Info
6661 @section Controlling Debugging Information Format
6663 @c prevent bad page break with this line
6664 This describes how to specify debugging information.
6667 * All Debuggers:: Macros that affect all debugging formats uniformly.
6668 * DBX Options:: Macros enabling specific options in DBX format.
6669 * DBX Hooks:: Hook macros for varying DBX format.
6670 * File Names and DBX:: Macros controlling output of file names in DBX format.
6671 * DWARF:: Macros for DWARF format.
6672 * VMS Debug:: Macros for VMS debug format.
6676 @subsection Macros Affecting All Debugging Formats
6678 @c prevent bad page break with this line
6679 These macros affect all debugging formats.
6681 @defmac DBX_REGISTER_NUMBER (@var{regno})
6682 A C expression that returns the DBX register number for the compiler
6683 register number @var{regno}. In the default macro provided, the value
6684 of this expression will be @var{regno} itself. But sometimes there are
6685 some registers that the compiler knows about and DBX does not, or vice
6686 versa. In such cases, some register may need to have one number in the
6687 compiler and another for DBX@.
6689 If two registers have consecutive numbers inside GCC, and they can be
6690 used as a pair to hold a multiword value, then they @emph{must} have
6691 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6692 Otherwise, debuggers will be unable to access such a pair, because they
6693 expect register pairs to be consecutive in their own numbering scheme.
6695 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6696 does not preserve register pairs, then what you must do instead is
6697 redefine the actual register numbering scheme.
6700 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
6701 A C expression that returns the integer offset value for an automatic
6702 variable having address @var{x} (an RTL expression). The default
6703 computation assumes that @var{x} is based on the frame-pointer and
6704 gives the offset from the frame-pointer. This is required for targets
6705 that produce debugging output for DBX and allow the frame-pointer to be
6706 eliminated when the @option{-g} option is used.
6709 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6710 A C expression that returns the integer offset value for an argument
6711 having address @var{x} (an RTL expression). The nominal offset is
6715 @defmac PREFERRED_DEBUGGING_TYPE
6716 A C expression that returns the type of debugging output GCC should
6717 produce when the user specifies just @option{-g}. Define
6718 this if you have arranged for GCC to support more than one format of
6719 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6720 @code{DWARF2_DEBUG}, @code{XCOFF_DEBUG}, @code{VMS_DEBUG},
6721 and @code{VMS_AND_DWARF2_DEBUG}.
6723 When the user specifies @option{-ggdb}, GCC normally also uses the
6724 value of this macro to select the debugging output format, but with two
6725 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6726 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6727 defined, GCC uses @code{DBX_DEBUG}.
6729 The value of this macro only affects the default debugging output; the
6730 user can always get a specific type of output by using @option{-gstabs},
6731 @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
6735 @subsection Specific Options for DBX Output
6737 @c prevent bad page break with this line
6738 These are specific options for DBX output.
6740 @defmac DBX_DEBUGGING_INFO
6741 Define this macro if GCC should produce debugging output for DBX
6742 in response to the @option{-g} option.
6745 @defmac XCOFF_DEBUGGING_INFO
6746 Define this macro if GCC should produce XCOFF format debugging output
6747 in response to the @option{-g} option. This is a variant of DBX format.
6750 @defmac DEFAULT_GDB_EXTENSIONS
6751 Define this macro to control whether GCC should by default generate
6752 GDB's extended version of DBX debugging information (assuming DBX-format
6753 debugging information is enabled at all). If you don't define the
6754 macro, the default is 1: always generate the extended information
6755 if there is any occasion to.
6758 @defmac DEBUG_SYMS_TEXT
6759 Define this macro if all @code{.stabs} commands should be output while
6760 in the text section.
6763 @defmac ASM_STABS_OP
6764 A C string constant, including spacing, naming the assembler pseudo op to
6765 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6766 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
6767 applies only to DBX debugging information format.
6770 @defmac ASM_STABD_OP
6771 A C string constant, including spacing, naming the assembler pseudo op to
6772 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6773 value is the current location. If you don't define this macro,
6774 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
6778 @defmac ASM_STABN_OP
6779 A C string constant, including spacing, naming the assembler pseudo op to
6780 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6781 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
6782 macro applies only to DBX debugging information format.
6785 @defmac DBX_NO_XREFS
6786 Define this macro if DBX on your system does not support the construct
6787 @samp{xs@var{tagname}}. On some systems, this construct is used to
6788 describe a forward reference to a structure named @var{tagname}.
6789 On other systems, this construct is not supported at all.
6792 @defmac DBX_CONTIN_LENGTH
6793 A symbol name in DBX-format debugging information is normally
6794 continued (split into two separate @code{.stabs} directives) when it
6795 exceeds a certain length (by default, 80 characters). On some
6796 operating systems, DBX requires this splitting; on others, splitting
6797 must not be done. You can inhibit splitting by defining this macro
6798 with the value zero. You can override the default splitting-length by
6799 defining this macro as an expression for the length you desire.
6802 @defmac DBX_CONTIN_CHAR
6803 Normally continuation is indicated by adding a @samp{\} character to
6804 the end of a @code{.stabs} string when a continuation follows. To use
6805 a different character instead, define this macro as a character
6806 constant for the character you want to use. Do not define this macro
6807 if backslash is correct for your system.
6810 @defmac DBX_STATIC_STAB_DATA_SECTION
6811 Define this macro if it is necessary to go to the data section before
6812 outputting the @samp{.stabs} pseudo-op for a non-global static
6816 @defmac DBX_TYPE_DECL_STABS_CODE
6817 The value to use in the ``code'' field of the @code{.stabs} directive
6818 for a typedef. The default is @code{N_LSYM}.
6821 @defmac DBX_STATIC_CONST_VAR_CODE
6822 The value to use in the ``code'' field of the @code{.stabs} directive
6823 for a static variable located in the text section. DBX format does not
6824 provide any ``right'' way to do this. The default is @code{N_FUN}.
6827 @defmac DBX_REGPARM_STABS_CODE
6828 The value to use in the ``code'' field of the @code{.stabs} directive
6829 for a parameter passed in registers. DBX format does not provide any
6830 ``right'' way to do this. The default is @code{N_RSYM}.
6833 @defmac DBX_REGPARM_STABS_LETTER
6834 The letter to use in DBX symbol data to identify a symbol as a parameter
6835 passed in registers. DBX format does not customarily provide any way to
6836 do this. The default is @code{'P'}.
6839 @defmac DBX_FUNCTION_FIRST
6840 Define this macro if the DBX information for a function and its
6841 arguments should precede the assembler code for the function. Normally,
6842 in DBX format, the debugging information entirely follows the assembler
6846 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
6847 Define this macro, with value 1, if the value of a symbol describing
6848 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
6849 relative to the start of the enclosing function. Normally, GCC uses
6850 an absolute address.
6853 @defmac DBX_LINES_FUNCTION_RELATIVE
6854 Define this macro, with value 1, if the value of a symbol indicating
6855 the current line number (@code{N_SLINE}) should be relative to the
6856 start of the enclosing function. Normally, GCC uses an absolute address.
6859 @defmac DBX_USE_BINCL
6860 Define this macro if GCC should generate @code{N_BINCL} and
6861 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6862 macro also directs GCC to output a type number as a pair of a file
6863 number and a type number within the file. Normally, GCC does not
6864 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6865 number for a type number.
6869 @subsection Open-Ended Hooks for DBX Format
6871 @c prevent bad page break with this line
6872 These are hooks for DBX format.
6874 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6875 A C statement to output DBX debugging information before code for line
6876 number @var{line} of the current source file to the stdio stream
6877 @var{stream}. @var{counter} is the number of time the macro was
6878 invoked, including the current invocation; it is intended to generate
6879 unique labels in the assembly output.
6881 This macro should not be defined if the default output is correct, or
6882 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
6885 @defmac NO_DBX_FUNCTION_END
6886 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6887 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
6888 On those machines, define this macro to turn this feature off without
6889 disturbing the rest of the gdb extensions.
6892 @defmac NO_DBX_BNSYM_ENSYM
6893 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
6894 extension construct. On those machines, define this macro to turn this
6895 feature off without disturbing the rest of the gdb extensions.
6898 @node File Names and DBX
6899 @subsection File Names in DBX Format
6901 @c prevent bad page break with this line
6902 This describes file names in DBX format.
6904 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6905 A C statement to output DBX debugging information to the stdio stream
6906 @var{stream}, which indicates that file @var{name} is the main source
6907 file---the file specified as the input file for compilation.
6908 This macro is called only once, at the beginning of compilation.
6910 This macro need not be defined if the standard form of output
6911 for DBX debugging information is appropriate.
6913 It may be necessary to refer to a label equal to the beginning of the
6914 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
6915 to do so. If you do this, you must also set the variable
6916 @var{used_ltext_label_name} to @code{true}.
6919 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
6920 Define this macro, with value 1, if GCC should not emit an indication
6921 of the current directory for compilation and current source language at
6922 the beginning of the file.
6925 @defmac NO_DBX_GCC_MARKER
6926 Define this macro, with value 1, if GCC should not emit an indication
6927 that this object file was compiled by GCC@. The default is to emit
6928 an @code{N_OPT} stab at the beginning of every source file, with
6929 @samp{gcc2_compiled.} for the string and value 0.
6932 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6933 A C statement to output DBX debugging information at the end of
6934 compilation of the main source file @var{name}. Output should be
6935 written to the stdio stream @var{stream}.
6937 If you don't define this macro, nothing special is output at the end
6938 of compilation, which is correct for most machines.
6941 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
6942 Define this macro @emph{instead of} defining
6943 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
6944 the end of compilation is an @code{N_SO} stab with an empty string,
6945 whose value is the highest absolute text address in the file.
6950 @subsection Macros for DWARF Output
6952 @c prevent bad page break with this line
6953 Here are macros for DWARF output.
6955 @defmac DWARF2_DEBUGGING_INFO
6956 Define this macro if GCC should produce dwarf version 2 format
6957 debugging output in response to the @option{-g} option.
6959 @hook TARGET_DWARF_CALLING_CONVENTION
6961 To support optional call frame debugging information, you must also
6962 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6963 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6964 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6965 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
6968 @defmac DWARF2_FRAME_INFO
6969 Define this macro to a nonzero value if GCC should always output
6970 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
6971 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
6972 exceptions are enabled, GCC will output this information not matter
6973 how you define @code{DWARF2_FRAME_INFO}.
6976 @hook TARGET_DEBUG_UNWIND_INFO
6978 @defmac DWARF2_ASM_LINE_DEBUG_INFO
6979 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
6980 line debug info sections. This will result in much more compact line number
6981 tables, and hence is desirable if it works.
6984 @defmac DWARF2_ASM_VIEW_DEBUG_INFO
6985 Define this macro to be a nonzero value if the assembler supports view
6986 assignment and verification in @code{.loc}. If it does not, but the
6987 user enables location views, the compiler may have to fallback to
6988 internal line number tables.
6991 @hook TARGET_RESET_LOCATION_VIEW
6993 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
6995 @hook TARGET_DELAY_SCHED2
6997 @hook TARGET_DELAY_VARTRACK
6999 @hook TARGET_NO_REGISTER_ALLOCATION
7001 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7002 A C statement to issue assembly directives that create a difference
7003 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
7006 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
7007 A C statement to issue assembly directives that create a difference
7008 between the two given labels in system defined units, e.g.@: instruction
7009 slots on IA64 VMS, using an integer of the given size.
7012 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
7013 A C statement to issue assembly directives that create a
7014 section-relative reference to the given @var{label} plus @var{offset}, using
7015 an integer of the given @var{size}. The label is known to be defined in the
7016 given @var{section}.
7019 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
7020 A C statement to issue assembly directives that create a self-relative
7021 reference to the given @var{label}, using an integer of the given @var{size}.
7024 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
7025 A C statement to issue assembly directives that create a reference to the
7026 given @var{label} relative to the dbase, using an integer of the given @var{size}.
7029 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
7030 A C statement to issue assembly directives that create a reference to
7031 the DWARF table identifier @var{label} from the current section. This
7032 is used on some systems to avoid garbage collecting a DWARF table which
7033 is referenced by a function.
7036 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
7040 @subsection Macros for VMS Debug Format
7042 @c prevent bad page break with this line
7043 Here are macros for VMS debug format.
7045 @defmac VMS_DEBUGGING_INFO
7046 Define this macro if GCC should produce debugging output for VMS
7047 in response to the @option{-g} option. The default behavior for VMS
7048 is to generate minimal debug info for a traceback in the absence of
7049 @option{-g} unless explicitly overridden with @option{-g0}. This
7050 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
7051 @code{TARGET_OPTION_OVERRIDE}.
7054 @node Floating Point
7055 @section Cross Compilation and Floating Point
7056 @cindex cross compilation and floating point
7057 @cindex floating point and cross compilation
7059 While all modern machines use twos-complement representation for integers,
7060 there are a variety of representations for floating point numbers. This
7061 means that in a cross-compiler the representation of floating point numbers
7062 in the compiled program may be different from that used in the machine
7063 doing the compilation.
7065 Because different representation systems may offer different amounts of
7066 range and precision, all floating point constants must be represented in
7067 the target machine's format. Therefore, the cross compiler cannot
7068 safely use the host machine's floating point arithmetic; it must emulate
7069 the target's arithmetic. To ensure consistency, GCC always uses
7070 emulation to work with floating point values, even when the host and
7071 target floating point formats are identical.
7073 The following macros are provided by @file{real.h} for the compiler to
7074 use. All parts of the compiler which generate or optimize
7075 floating-point calculations must use these macros. They may evaluate
7076 their operands more than once, so operands must not have side effects.
7078 @defmac REAL_VALUE_TYPE
7079 The C data type to be used to hold a floating point value in the target
7080 machine's format. Typically this is a @code{struct} containing an
7081 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7085 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7086 Truncates @var{x} to a signed integer, rounding toward zero.
7089 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7090 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7091 @var{x} is negative, returns zero.
7094 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
7095 Converts @var{string} into a floating point number in the target machine's
7096 representation for mode @var{mode}. This routine can handle both
7097 decimal and hexadecimal floating point constants, using the syntax
7098 defined by the C language for both.
7101 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7102 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7105 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7106 Determines whether @var{x} represents infinity (positive or negative).
7109 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7110 Determines whether @var{x} represents a ``NaN'' (not-a-number).
7113 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
7114 Returns the negative of the floating point value @var{x}.
7117 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
7118 Returns the absolute value of @var{x}.
7121 @node Mode Switching
7122 @section Mode Switching Instructions
7123 @cindex mode switching
7124 The following macros control mode switching optimizations:
7126 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
7127 Define this macro if the port needs extra instructions inserted for mode
7128 switching in an optimizing compilation.
7130 For an example, the SH4 can perform both single and double precision
7131 floating point operations, but to perform a single precision operation,
7132 the FPSCR PR bit has to be cleared, while for a double precision
7133 operation, this bit has to be set. Changing the PR bit requires a general
7134 purpose register as a scratch register, hence these FPSCR sets have to
7135 be inserted before reload, i.e.@: you cannot put this into instruction emitting
7136 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
7138 You can have multiple entities that are mode-switched, and select at run time
7139 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7140 return nonzero for any @var{entity} that needs mode-switching.
7141 If you define this macro, you also have to define
7142 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
7143 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
7144 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
7148 @defmac NUM_MODES_FOR_MODE_SWITCHING
7149 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7150 initializer for an array of integers. Each initializer element
7151 N refers to an entity that needs mode switching, and specifies the number
7152 of different modes that might need to be set for this entity.
7153 The position of the initializer in the initializer---starting counting at
7154 zero---determines the integer that is used to refer to the mode-switched
7156 In macros that take mode arguments / yield a mode result, modes are
7157 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
7158 switch is needed / supplied.
7161 @hook TARGET_MODE_EMIT
7163 @hook TARGET_MODE_NEEDED
7165 @hook TARGET_MODE_AFTER
7167 @hook TARGET_MODE_ENTRY
7169 @hook TARGET_MODE_EXIT
7171 @hook TARGET_MODE_PRIORITY
7173 @node Target Attributes
7174 @section Defining target-specific uses of @code{__attribute__}
7175 @cindex target attributes
7176 @cindex machine attributes
7177 @cindex attributes, target-specific
7179 Target-specific attributes may be defined for functions, data and types.
7180 These are described using the following target hooks; they also need to
7181 be documented in @file{extend.texi}.
7183 @hook TARGET_ATTRIBUTE_TABLE
7185 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
7187 @hook TARGET_COMP_TYPE_ATTRIBUTES
7189 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
7191 @hook TARGET_MERGE_TYPE_ATTRIBUTES
7193 @hook TARGET_MERGE_DECL_ATTRIBUTES
7195 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7197 @defmac TARGET_DECLSPEC
7198 Define this macro to a nonzero value if you want to treat
7199 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
7200 default, this behavior is enabled only for targets that define
7201 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
7202 of @code{__declspec} is via a built-in macro, but you should not rely
7203 on this implementation detail.
7206 @hook TARGET_INSERT_ATTRIBUTES
7208 @hook TARGET_HANDLE_GENERIC_ATTRIBUTE
7210 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7212 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
7214 @hook TARGET_OPTION_SAVE
7216 @hook TARGET_OPTION_RESTORE
7218 @hook TARGET_OPTION_POST_STREAM_IN
7220 @hook TARGET_OPTION_PRINT
7222 @hook TARGET_OPTION_PRAGMA_PARSE
7224 @hook TARGET_OPTION_OVERRIDE
7226 @hook TARGET_OPTION_FUNCTION_VERSIONS
7228 @hook TARGET_CAN_INLINE_P
7230 @hook TARGET_RELAYOUT_FUNCTION
7233 @section Emulating TLS
7234 @cindex Emulated TLS
7236 For targets whose psABI does not provide Thread Local Storage via
7237 specific relocations and instruction sequences, an emulation layer is
7238 used. A set of target hooks allows this emulation layer to be
7239 configured for the requirements of a particular target. For instance
7240 the psABI may in fact specify TLS support in terms of an emulation
7243 The emulation layer works by creating a control object for every TLS
7244 object. To access the TLS object, a lookup function is provided
7245 which, when given the address of the control object, will return the
7246 address of the current thread's instance of the TLS object.
7248 @hook TARGET_EMUTLS_GET_ADDRESS
7250 @hook TARGET_EMUTLS_REGISTER_COMMON
7252 @hook TARGET_EMUTLS_VAR_SECTION
7254 @hook TARGET_EMUTLS_TMPL_SECTION
7256 @hook TARGET_EMUTLS_VAR_PREFIX
7258 @hook TARGET_EMUTLS_TMPL_PREFIX
7260 @hook TARGET_EMUTLS_VAR_FIELDS
7262 @hook TARGET_EMUTLS_VAR_INIT
7264 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7266 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7268 @node MIPS Coprocessors
7269 @section Defining coprocessor specifics for MIPS targets.
7270 @cindex MIPS coprocessor-definition macros
7272 The MIPS specification allows MIPS implementations to have as many as 4
7273 coprocessors, each with as many as 32 private registers. GCC supports
7274 accessing these registers and transferring values between the registers
7275 and memory using asm-ized variables. For example:
7278 register unsigned int cp0count asm ("c0r1");
7284 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7285 names may be added as described below, or the default names may be
7286 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7288 Coprocessor registers are assumed to be epilogue-used; sets to them will
7289 be preserved even if it does not appear that the register is used again
7290 later in the function.
7292 Another note: according to the MIPS spec, coprocessor 1 (if present) is
7293 the FPU@. One accesses COP1 registers through standard mips
7294 floating-point support; they are not included in this mechanism.
7297 @section Parameters for Precompiled Header Validity Checking
7298 @cindex parameters, precompiled headers
7300 @hook TARGET_GET_PCH_VALIDITY
7302 @hook TARGET_PCH_VALID_P
7304 @hook TARGET_CHECK_PCH_TARGET_FLAGS
7306 @hook TARGET_PREPARE_PCH_SAVE
7309 @section C++ ABI parameters
7310 @cindex parameters, c++ abi
7312 @hook TARGET_CXX_GUARD_TYPE
7314 @hook TARGET_CXX_GUARD_MASK_BIT
7316 @hook TARGET_CXX_GET_COOKIE_SIZE
7318 @hook TARGET_CXX_COOKIE_HAS_SIZE
7320 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
7322 @hook TARGET_CXX_CDTOR_RETURNS_THIS
7324 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7326 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7328 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7330 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7332 @hook TARGET_CXX_USE_AEABI_ATEXIT
7334 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7336 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7338 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
7340 @node D Language and ABI
7341 @section D ABI parameters
7342 @cindex parameters, d abi
7344 @hook TARGET_D_CPU_VERSIONS
7346 @hook TARGET_D_OS_VERSIONS
7348 @hook TARGET_D_CRITSEC_SIZE
7350 @node Named Address Spaces
7351 @section Adding support for named address spaces
7352 @cindex named address spaces
7354 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7355 standards committee, @cite{Programming Languages - C - Extensions to
7356 support embedded processors}, specifies a syntax for embedded
7357 processors to specify alternate address spaces. You can configure a
7358 GCC port to support section 5.1 of the draft report to add support for
7359 address spaces other than the default address space. These address
7360 spaces are new keywords that are similar to the @code{volatile} and
7361 @code{const} type attributes.
7363 Pointers to named address spaces can have a different size than
7364 pointers to the generic address space.
7366 For example, the SPU port uses the @code{__ea} address space to refer
7367 to memory in the host processor, rather than memory local to the SPU
7368 processor. Access to memory in the @code{__ea} address space involves
7369 issuing DMA operations to move data between the host processor and the
7370 local processor memory address space. Pointers in the @code{__ea}
7371 address space are either 32 bits or 64 bits based on the
7372 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
7375 Internally, address spaces are represented as a small integer in the
7376 range 0 to 15 with address space 0 being reserved for the generic
7379 To register a named address space qualifier keyword with the C front end,
7380 the target may call the @code{c_register_addr_space} routine. For example,
7381 the SPU port uses the following to declare @code{__ea} as the keyword for
7382 named address space #1:
7384 #define ADDR_SPACE_EA 1
7385 c_register_addr_space ("__ea", ADDR_SPACE_EA);
7388 @hook TARGET_ADDR_SPACE_POINTER_MODE
7390 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
7392 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7394 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7396 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7398 @hook TARGET_ADDR_SPACE_SUBSET_P
7400 @hook TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID
7402 @hook TARGET_ADDR_SPACE_CONVERT
7404 @hook TARGET_ADDR_SPACE_DEBUG
7406 @hook TARGET_ADDR_SPACE_DIAGNOSE_USAGE
7409 @section Miscellaneous Parameters
7410 @cindex parameters, miscellaneous
7412 @c prevent bad page break with this line
7413 Here are several miscellaneous parameters.
7415 @defmac HAS_LONG_COND_BRANCH
7416 Define this boolean macro to indicate whether or not your architecture
7417 has conditional branches that can span all of memory. It is used in
7418 conjunction with an optimization that partitions hot and cold basic
7419 blocks into separate sections of the executable. If this macro is
7420 set to false, gcc will convert any conditional branches that attempt
7421 to cross between sections into unconditional branches or indirect jumps.
7424 @defmac HAS_LONG_UNCOND_BRANCH
7425 Define this boolean macro to indicate whether or not your architecture
7426 has unconditional branches that can span all of memory. It is used in
7427 conjunction with an optimization that partitions hot and cold basic
7428 blocks into separate sections of the executable. If this macro is
7429 set to false, gcc will convert any unconditional branches that attempt
7430 to cross between sections into indirect jumps.
7433 @defmac CASE_VECTOR_MODE
7434 An alias for a machine mode name. This is the machine mode that
7435 elements of a jump-table should have.
7438 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7439 Optional: return the preferred mode for an @code{addr_diff_vec}
7440 when the minimum and maximum offset are known. If you define this,
7441 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7442 To make this work, you also have to define @code{INSN_ALIGN} and
7443 make the alignment for @code{addr_diff_vec} explicit.
7444 The @var{body} argument is provided so that the offset_unsigned and scale
7445 flags can be updated.
7448 @defmac CASE_VECTOR_PC_RELATIVE
7449 Define this macro to be a C expression to indicate when jump-tables
7450 should contain relative addresses. You need not define this macro if
7451 jump-tables never contain relative addresses, or jump-tables should
7452 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7456 @hook TARGET_CASE_VALUES_THRESHOLD
7458 @defmac WORD_REGISTER_OPERATIONS
7459 Define this macro to 1 if operations between registers with integral mode
7460 smaller than a word are always performed on the entire register. To be
7461 more explicit, if you start with a pair of @code{word_mode} registers with
7462 known values and you do a subword, for example @code{QImode}, addition on
7463 the low part of the registers, then the compiler may consider that the
7464 result has a known value in @code{word_mode} too if the macro is defined
7465 to 1. Most RISC machines have this property and most CISC machines do not.
7468 @hook TARGET_MIN_ARITHMETIC_PRECISION
7470 @defmac LOAD_EXTEND_OP (@var{mem_mode})
7471 Define this macro to be a C expression indicating when insns that read
7472 memory in @var{mem_mode}, an integral mode narrower than a word, set the
7473 bits outside of @var{mem_mode} to be either the sign-extension or the
7474 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7475 of @var{mem_mode} for which the
7476 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7477 @code{UNKNOWN} for other modes.
7479 This macro is not called with @var{mem_mode} non-integral or with a width
7480 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7481 value in this case. Do not define this macro if it would always return
7482 @code{UNKNOWN}. On machines where this macro is defined, you will normally
7483 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7485 You may return a non-@code{UNKNOWN} value even if for some hard registers
7486 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7487 of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
7488 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7489 integral mode larger than this but not larger than @code{word_mode}.
7491 You must return @code{UNKNOWN} if for some hard registers that allow this
7492 mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
7493 @code{word_mode}, but that they can change to another integral mode that
7494 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7497 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
7498 Define this macro to 1 if loading short immediate values into registers sign
7502 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7505 The maximum number of bytes that a single instruction can move quickly
7506 between memory and registers or between two memory locations.
7509 @defmac MAX_MOVE_MAX
7510 The maximum number of bytes that a single instruction can move quickly
7511 between memory and registers or between two memory locations. If this
7512 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7513 constant value that is the largest value that @code{MOVE_MAX} can have
7517 @defmac SHIFT_COUNT_TRUNCATED
7518 A C expression that is nonzero if on this machine the number of bits
7519 actually used for the count of a shift operation is equal to the number
7520 of bits needed to represent the size of the object being shifted. When
7521 this macro is nonzero, the compiler will assume that it is safe to omit
7522 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7523 truncates the count of a shift operation. On machines that have
7524 instructions that act on bit-fields at variable positions, which may
7525 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7526 also enables deletion of truncations of the values that serve as
7527 arguments to bit-field instructions.
7529 If both types of instructions truncate the count (for shifts) and
7530 position (for bit-field operations), or if no variable-position bit-field
7531 instructions exist, you should define this macro.
7533 However, on some machines, such as the 80386 and the 680x0, truncation
7534 only applies to shift operations and not the (real or pretended)
7535 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7536 such machines. Instead, add patterns to the @file{md} file that include
7537 the implied truncation of the shift instructions.
7539 You need not define this macro if it would always have the value of zero.
7542 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
7543 @hook TARGET_SHIFT_TRUNCATION_MASK
7545 @hook TARGET_TRULY_NOOP_TRUNCATION
7547 @hook TARGET_MODE_REP_EXTENDED
7549 @hook TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P
7551 @defmac STORE_FLAG_VALUE
7552 A C expression describing the value returned by a comparison operator
7553 with an integral mode and stored by a store-flag instruction
7554 (@samp{cstore@var{mode}4}) when the condition is true. This description must
7555 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7556 comparison operators whose results have a @code{MODE_INT} mode.
7558 A value of 1 or @minus{}1 means that the instruction implementing the
7559 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7560 and 0 when the comparison is false. Otherwise, the value indicates
7561 which bits of the result are guaranteed to be 1 when the comparison is
7562 true. This value is interpreted in the mode of the comparison
7563 operation, which is given by the mode of the first operand in the
7564 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
7565 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7568 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7569 generate code that depends only on the specified bits. It can also
7570 replace comparison operators with equivalent operations if they cause
7571 the required bits to be set, even if the remaining bits are undefined.
7572 For example, on a machine whose comparison operators return an
7573 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7574 @samp{0x80000000}, saying that just the sign bit is relevant, the
7578 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7585 (ashift:SI @var{x} (const_int @var{n}))
7589 where @var{n} is the appropriate shift count to move the bit being
7590 tested into the sign bit.
7592 There is no way to describe a machine that always sets the low-order bit
7593 for a true value, but does not guarantee the value of any other bits,
7594 but we do not know of any machine that has such an instruction. If you
7595 are trying to port GCC to such a machine, include an instruction to
7596 perform a logical-and of the result with 1 in the pattern for the
7597 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7599 Often, a machine will have multiple instructions that obtain a value
7600 from a comparison (or the condition codes). Here are rules to guide the
7601 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7606 Use the shortest sequence that yields a valid definition for
7607 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7608 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7609 comparison operators to do so because there may be opportunities to
7610 combine the normalization with other operations.
7613 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7614 slightly preferred on machines with expensive jumps and 1 preferred on
7618 As a second choice, choose a value of @samp{0x80000001} if instructions
7619 exist that set both the sign and low-order bits but do not define the
7623 Otherwise, use a value of @samp{0x80000000}.
7626 Many machines can produce both the value chosen for
7627 @code{STORE_FLAG_VALUE} and its negation in the same number of
7628 instructions. On those machines, you should also define a pattern for
7629 those cases, e.g., one matching
7632 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7635 Some machines can also perform @code{and} or @code{plus} operations on
7636 condition code values with less instructions than the corresponding
7637 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
7638 machines, define the appropriate patterns. Use the names @code{incscc}
7639 and @code{decscc}, respectively, for the patterns which perform
7640 @code{plus} or @code{minus} operations on condition code values. See
7641 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
7642 find such instruction sequences on other machines.
7644 If this macro is not defined, the default value, 1, is used. You need
7645 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7646 instructions, or if the value generated by these instructions is 1.
7649 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7650 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7651 returned when comparison operators with floating-point results are true.
7652 Define this macro on machines that have comparison operations that return
7653 floating-point values. If there are no such operations, do not define
7657 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7658 A C expression that gives a rtx representing the nonzero true element
7659 for vector comparisons. The returned rtx should be valid for the inner
7660 mode of @var{mode} which is guaranteed to be a vector mode. Define
7661 this macro on machines that have vector comparison operations that
7662 return a vector result. If there are no such operations, do not define
7663 this macro. Typically, this macro is defined as @code{const1_rtx} or
7664 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
7665 the compiler optimizing such vector comparison operations for the
7669 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7670 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7671 A C expression that indicates whether the architecture defines a value
7672 for @code{clz} or @code{ctz} with a zero operand.
7673 A result of @code{0} indicates the value is undefined.
7674 If the value is defined for only the RTL expression, the macro should
7675 evaluate to @code{1}; if the value applies also to the corresponding optab
7676 entry (which is normally the case if it expands directly into
7677 the corresponding RTL), then the macro should evaluate to @code{2}.
7678 In the cases where the value is defined, @var{value} should be set to
7681 If this macro is not defined, the value of @code{clz} or
7682 @code{ctz} at zero is assumed to be undefined.
7684 This macro must be defined if the target's expansion for @code{ffs}
7685 relies on a particular value to get correct results. Otherwise it
7686 is not necessary, though it may be used to optimize some corner cases, and
7687 to provide a default expansion for the @code{ffs} optab.
7689 Note that regardless of this macro the ``definedness'' of @code{clz}
7690 and @code{ctz} at zero do @emph{not} extend to the builtin functions
7691 visible to the user. Thus one may be free to adjust the value at will
7692 to match the target expansion of these operations without fear of
7697 An alias for the machine mode for pointers. On most machines, define
7698 this to be the integer mode corresponding to the width of a hardware
7699 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7700 On some machines you must define this to be one of the partial integer
7701 modes, such as @code{PSImode}.
7703 The width of @code{Pmode} must be at least as large as the value of
7704 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7705 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7709 @defmac FUNCTION_MODE
7710 An alias for the machine mode used for memory references to functions
7711 being called, in @code{call} RTL expressions. On most CISC machines,
7712 where an instruction can begin at any byte address, this should be
7713 @code{QImode}. On most RISC machines, where all instructions have fixed
7714 size and alignment, this should be a mode with the same size and alignment
7715 as the machine instruction words - typically @code{SImode} or @code{HImode}.
7718 @defmac STDC_0_IN_SYSTEM_HEADERS
7719 In normal operation, the preprocessor expands @code{__STDC__} to the
7720 constant 1, to signify that GCC conforms to ISO Standard C@. On some
7721 hosts, like Solaris, the system compiler uses a different convention,
7722 where @code{__STDC__} is normally 0, but is 1 if the user specifies
7723 strict conformance to the C Standard.
7725 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7726 convention when processing system header files, but when processing user
7727 files @code{__STDC__} will always expand to 1.
7730 @hook TARGET_C_PREINCLUDE
7732 @hook TARGET_CXX_IMPLICIT_EXTERN_C
7734 @defmac SYSTEM_IMPLICIT_EXTERN_C
7735 Define this macro if the system header files do not support C++@.
7736 This macro handles system header files by pretending that system
7737 header files are enclosed in @samp{extern "C" @{@dots{}@}}.
7742 @defmac REGISTER_TARGET_PRAGMAS ()
7743 Define this macro if you want to implement any target-specific pragmas.
7744 If defined, it is a C expression which makes a series of calls to
7745 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
7746 for each pragma. The macro may also do any
7747 setup required for the pragmas.
7749 The primary reason to define this macro is to provide compatibility with
7750 other compilers for the same target. In general, we discourage
7751 definition of target-specific pragmas for GCC@.
7753 If the pragma can be implemented by attributes then you should consider
7754 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
7756 Preprocessor macros that appear on pragma lines are not expanded. All
7757 @samp{#pragma} directives that do not match any registered pragma are
7758 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
7761 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7762 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7764 Each call to @code{c_register_pragma} or
7765 @code{c_register_pragma_with_expansion} establishes one pragma. The
7766 @var{callback} routine will be called when the preprocessor encounters a
7770 #pragma [@var{space}] @var{name} @dots{}
7773 @var{space} is the case-sensitive namespace of the pragma, or
7774 @code{NULL} to put the pragma in the global namespace. The callback
7775 routine receives @var{pfile} as its first argument, which can be passed
7776 on to cpplib's functions if necessary. You can lex tokens after the
7777 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
7778 callback will be silently ignored. The end of the line is indicated by
7779 a token of type @code{CPP_EOF}. Macro expansion occurs on the
7780 arguments of pragmas registered with
7781 @code{c_register_pragma_with_expansion} but not on the arguments of
7782 pragmas registered with @code{c_register_pragma}.
7784 Note that the use of @code{pragma_lex} is specific to the C and C++
7785 compilers. It will not work in the Java or Fortran compilers, or any
7786 other language compilers for that matter. Thus if @code{pragma_lex} is going
7787 to be called from target-specific code, it must only be done so when
7788 building the C and C++ compilers. This can be done by defining the
7789 variables @code{c_target_objs} and @code{cxx_target_objs} in the
7790 target entry in the @file{config.gcc} file. These variables should name
7791 the target-specific, language-specific object file which contains the
7792 code that uses @code{pragma_lex}. Note it will also be necessary to add a
7793 rule to the makefile fragment pointed to by @code{tmake_file} that shows
7794 how to build this object file.
7797 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
7798 Define this macro if macros should be expanded in the
7799 arguments of @samp{#pragma pack}.
7802 @defmac TARGET_DEFAULT_PACK_STRUCT
7803 If your target requires a structure packing default other than 0 (meaning
7804 the machine default), define this macro to the necessary value (in bytes).
7805 This must be a value that would also be valid to use with
7806 @samp{#pragma pack()} (that is, a small power of two).
7809 @defmac DOLLARS_IN_IDENTIFIERS
7810 Define this macro to control use of the character @samp{$} in
7811 identifier names for the C family of languages. 0 means @samp{$} is
7812 not allowed by default; 1 means it is allowed. 1 is the default;
7813 there is no need to define this macro in that case.
7816 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
7817 Define this macro as a C expression that is nonzero if it is safe for the
7818 delay slot scheduler to place instructions in the delay slot of @var{insn},
7819 even if they appear to use a resource set or clobbered in @var{insn}.
7820 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7821 every @code{call_insn} has this behavior. On machines where some @code{insn}
7822 or @code{jump_insn} is really a function call and hence has this behavior,
7823 you should define this macro.
7825 You need not define this macro if it would always return zero.
7828 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
7829 Define this macro as a C expression that is nonzero if it is safe for the
7830 delay slot scheduler to place instructions in the delay slot of @var{insn},
7831 even if they appear to set or clobber a resource referenced in @var{insn}.
7832 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7833 some @code{insn} or @code{jump_insn} is really a function call and its operands
7834 are registers whose use is actually in the subroutine it calls, you should
7835 define this macro. Doing so allows the delay slot scheduler to move
7836 instructions which copy arguments into the argument registers into the delay
7839 You need not define this macro if it would always return zero.
7842 @defmac MULTIPLE_SYMBOL_SPACES
7843 Define this macro as a C expression that is nonzero if, in some cases,
7844 global symbols from one translation unit may not be bound to undefined
7845 symbols in another translation unit without user intervention. For
7846 instance, under Microsoft Windows symbols must be explicitly imported
7847 from shared libraries (DLLs).
7849 You need not define this macro if it would always evaluate to zero.
7852 @hook TARGET_MD_ASM_ADJUST
7854 @defmac MATH_LIBRARY
7855 Define this macro as a C string constant for the linker argument to link
7856 in the system math library, minus the initial @samp{"-l"}, or
7857 @samp{""} if the target does not have a
7858 separate math library.
7860 You need only define this macro if the default of @samp{"m"} is wrong.
7863 @defmac LIBRARY_PATH_ENV
7864 Define this macro as a C string constant for the environment variable that
7865 specifies where the linker should look for libraries.
7867 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7871 @defmac TARGET_POSIX_IO
7872 Define this macro if the target supports the following POSIX@ file
7873 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
7874 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
7875 to use file locking when exiting a program, which avoids race conditions
7876 if the program has forked. It will also create directories at run-time
7877 for cross-profiling.
7880 @defmac MAX_CONDITIONAL_EXECUTE
7882 A C expression for the maximum number of instructions to execute via
7883 conditional execution instructions instead of a branch. A value of
7884 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
7885 1 if it does use cc0.
7888 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
7889 Used if the target needs to perform machine-dependent modifications on the
7890 conditionals used for turning basic blocks into conditionally executed code.
7891 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
7892 contains information about the currently processed blocks. @var{true_expr}
7893 and @var{false_expr} are the tests that are used for converting the
7894 then-block and the else-block, respectively. Set either @var{true_expr} or
7895 @var{false_expr} to a null pointer if the tests cannot be converted.
7898 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
7899 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
7900 if-statements into conditions combined by @code{and} and @code{or} operations.
7901 @var{bb} contains the basic block that contains the test that is currently
7902 being processed and about to be turned into a condition.
7905 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
7906 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
7907 be converted to conditional execution format. @var{ce_info} points to
7908 a data structure, @code{struct ce_if_block}, which contains information
7909 about the currently processed blocks.
7912 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
7913 A C expression to perform any final machine dependent modifications in
7914 converting code to conditional execution. The involved basic blocks
7915 can be found in the @code{struct ce_if_block} structure that is pointed
7916 to by @var{ce_info}.
7919 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
7920 A C expression to cancel any machine dependent modifications in
7921 converting code to conditional execution. The involved basic blocks
7922 can be found in the @code{struct ce_if_block} structure that is pointed
7923 to by @var{ce_info}.
7926 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
7927 A C expression to initialize any machine specific data for if-conversion
7928 of the if-block in the @code{struct ce_if_block} structure that is pointed
7929 to by @var{ce_info}.
7932 @hook TARGET_MACHINE_DEPENDENT_REORG
7934 @hook TARGET_INIT_BUILTINS
7936 @hook TARGET_BUILTIN_DECL
7938 @hook TARGET_EXPAND_BUILTIN
7940 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
7942 @hook TARGET_CHECK_BUILTIN_CALL
7944 @hook TARGET_FOLD_BUILTIN
7946 @hook TARGET_GIMPLE_FOLD_BUILTIN
7948 @hook TARGET_COMPARE_VERSION_PRIORITY
7950 @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
7952 @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
7954 @hook TARGET_PREDICT_DOLOOP_P
7956 @hook TARGET_HAVE_COUNT_REG_DECR_P
7958 @hook TARGET_DOLOOP_COST_FOR_GENERIC
7960 @hook TARGET_DOLOOP_COST_FOR_ADDRESS
7962 @hook TARGET_CAN_USE_DOLOOP_P
7964 @hook TARGET_INVALID_WITHIN_DOLOOP
7966 @hook TARGET_LEGITIMATE_COMBINED_INSN
7968 @hook TARGET_CAN_FOLLOW_JUMP
7970 @hook TARGET_COMMUTATIVE_P
7972 @hook TARGET_ALLOCATE_INITIAL_VALUE
7974 @hook TARGET_UNSPEC_MAY_TRAP_P
7976 @hook TARGET_SET_CURRENT_FUNCTION
7978 @defmac TARGET_OBJECT_SUFFIX
7979 Define this macro to be a C string representing the suffix for object
7980 files on your target machine. If you do not define this macro, GCC will
7981 use @samp{.o} as the suffix for object files.
7984 @defmac TARGET_EXECUTABLE_SUFFIX
7985 Define this macro to be a C string representing the suffix to be
7986 automatically added to executable files on your target machine. If you
7987 do not define this macro, GCC will use the null string as the suffix for
7991 @defmac COLLECT_EXPORT_LIST
7992 If defined, @code{collect2} will scan the individual object files
7993 specified on its command line and create an export list for the linker.
7994 Define this macro for systems like AIX, where the linker discards
7995 object files that are not referenced from @code{main} and uses export
7999 @hook TARGET_CANNOT_MODIFY_JUMPS_P
8001 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
8003 @hook TARGET_GEN_CCMP_FIRST
8005 @hook TARGET_GEN_CCMP_NEXT
8007 @hook TARGET_LOOP_UNROLL_ADJUST
8009 @defmac POWI_MAX_MULTS
8010 If defined, this macro is interpreted as a signed integer C expression
8011 that specifies the maximum number of floating point multiplications
8012 that should be emitted when expanding exponentiation by an integer
8013 constant inline. When this value is defined, exponentiation requiring
8014 more than this number of multiplications is implemented by calling the
8015 system library's @code{pow}, @code{powf} or @code{powl} routines.
8016 The default value places no upper bound on the multiplication count.
8019 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8020 This target hook should register any extra include files for the
8021 target. The parameter @var{stdinc} indicates if normal include files
8022 are present. The parameter @var{sysroot} is the system root directory.
8023 The parameter @var{iprefix} is the prefix for the gcc directory.
8026 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8027 This target hook should register any extra include files for the
8028 target before any standard headers. The parameter @var{stdinc}
8029 indicates if normal include files are present. The parameter
8030 @var{sysroot} is the system root directory. The parameter
8031 @var{iprefix} is the prefix for the gcc directory.
8034 @deftypefn Macro void TARGET_OPTF (char *@var{path})
8035 This target hook should register special include paths for the target.
8036 The parameter @var{path} is the include to register. On Darwin
8037 systems, this is used for Framework includes, which have semantics
8038 that are different from @option{-I}.
8041 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
8042 This target macro returns @code{true} if it is safe to use a local alias
8043 for a virtual function @var{fndecl} when constructing thunks,
8044 @code{false} otherwise. By default, the macro returns @code{true} for all
8045 functions, if a target supports aliases (i.e.@: defines
8046 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
8049 @defmac TARGET_FORMAT_TYPES
8050 If defined, this macro is the name of a global variable containing
8051 target-specific format checking information for the @option{-Wformat}
8052 option. The default is to have no target-specific format checks.
8055 @defmac TARGET_N_FORMAT_TYPES
8056 If defined, this macro is the number of entries in
8057 @code{TARGET_FORMAT_TYPES}.
8060 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
8061 If defined, this macro is the name of a global variable containing
8062 target-specific format overrides for the @option{-Wformat} option. The
8063 default is to have no target-specific format overrides. If defined,
8064 @code{TARGET_FORMAT_TYPES} must be defined, too.
8067 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
8068 If defined, this macro specifies the number of entries in
8069 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
8072 @defmac TARGET_OVERRIDES_FORMAT_INIT
8073 If defined, this macro specifies the optional initialization
8074 routine for target specific customizations of the system printf
8075 and scanf formatter settings.
8078 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
8080 @hook TARGET_INVALID_CONVERSION
8082 @hook TARGET_INVALID_UNARY_OP
8084 @hook TARGET_INVALID_BINARY_OP
8086 @hook TARGET_PROMOTED_TYPE
8088 @hook TARGET_CONVERT_TO_TYPE
8090 @hook TARGET_VERIFY_TYPE_CONTEXT
8093 This macro determines the size of the objective C jump buffer for the
8094 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
8097 @defmac LIBGCC2_UNWIND_ATTRIBUTE
8098 Define this macro if any target-specific attributes need to be attached
8099 to the functions in @file{libgcc} that provide low-level support for
8100 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
8101 and the associated definitions of those functions.
8104 @hook TARGET_UPDATE_STACK_BOUNDARY
8106 @hook TARGET_GET_DRAP_RTX
8108 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
8110 @hook TARGET_CONST_ANCHOR
8112 @hook TARGET_ASAN_SHADOW_OFFSET
8114 @hook TARGET_MEMMODEL_CHECK
8116 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
8118 @hook TARGET_HAS_IFUNC_P
8120 @hook TARGET_ATOMIC_ALIGN_FOR_MODE
8122 @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV
8124 @hook TARGET_RECORD_OFFLOAD_SYMBOL
8126 @hook TARGET_OFFLOAD_OPTIONS
8128 @defmac TARGET_SUPPORTS_WIDE_INT
8130 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
8131 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
8132 to indicate that large integers are stored in
8133 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
8134 very large integer constants to be represented. @code{CONST_DOUBLE}
8135 is limited to twice the size of the host's @code{HOST_WIDE_INT}
8138 Converting a port mostly requires looking for the places where
8139 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
8140 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
8141 const_double"} at the port level gets you to 95% of the changes that
8142 need to be made. There are a few places that require a deeper look.
8146 There is no equivalent to @code{hval} and @code{lval} for
8147 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
8148 language since there are a variable number of elements.
8150 Most ports only check that @code{hval} is either 0 or -1 to see if the
8151 value is small. As mentioned above, this will no longer be necessary
8152 since small constants are always @code{CONST_INT}. Of course there
8153 are still a few exceptions, the alpha's constraint used by the zap
8154 instruction certainly requires careful examination by C code.
8155 However, all the current code does is pass the hval and lval to C
8156 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
8157 not really a large change.
8160 Because there is no standard template that ports use to materialize
8161 constants, there is likely to be some futzing that is unique to each
8165 The rtx costs may have to be adjusted to properly account for larger
8166 constants that are represented as @code{CONST_WIDE_INT}.
8169 All and all it does not take long to convert ports that the
8170 maintainer is familiar with.
8174 @hook TARGET_HAVE_SPECULATION_SAFE_VALUE
8176 @hook TARGET_SPECULATION_SAFE_VALUE
8178 @hook TARGET_RUN_TARGET_SELFTESTS